summaryrefslogtreecommitdiffhomepage
diff options
context:
space:
mode:
-rw-r--r--.bazelrc4
-rw-r--r--.github/workflows/go.yml5
-rw-r--r--.travis.yml2
-rw-r--r--Makefile12
-rw-r--r--WORKSPACE7
-rw-r--r--benchmarks/workloads/ruby/Gemfile.lock12
-rw-r--r--go.mod2
-rw-r--r--go.sum3
-rw-r--r--images/basic/hostoverlaytest/Dockerfile3
-rw-r--r--images/basic/hostoverlaytest/copy_up_testfile.txt (renamed from images/basic/hostoverlaytest/testfile.txt)0
-rw-r--r--images/basic/hostoverlaytest/test_copy_up.c (renamed from images/basic/hostoverlaytest/test.c)2
-rw-r--r--images/basic/hostoverlaytest/test_rewinddir.c78
-rw-r--r--images/benchmarks/fio/Dockerfile7
-rw-r--r--images/benchmarks/hey/Dockerfile12
-rw-r--r--images/benchmarks/nginx/Dockerfile1
-rw-r--r--images/benchmarks/node/Dockerfile1
-rw-r--r--images/benchmarks/node/index.hbs8
-rw-r--r--images/benchmarks/node/index.js42
-rw-r--r--images/benchmarks/node/package-lock.json486
-rw-r--r--images/benchmarks/node/package.json19
-rw-r--r--images/benchmarks/runsc/Dockerfile24
-rw-r--r--pkg/abi/linux/fuse.go105
-rw-r--r--pkg/abi/linux/netfilter.go8
-rw-r--r--pkg/abi/linux/socket.go4
-rw-r--r--pkg/sentry/fdimport/fdimport.go5
-rw-r--r--pkg/sentry/fsimpl/fuse/BUILD4
-rw-r--r--pkg/sentry/fsimpl/fuse/connection.go234
-rw-r--r--pkg/sentry/fsimpl/fuse/dev.go13
-rw-r--r--pkg/sentry/fsimpl/fuse/dev_test.go17
-rw-r--r--pkg/sentry/fsimpl/fuse/fusefs.go22
-rw-r--r--pkg/sentry/fsimpl/fuse/init.go166
-rw-r--r--pkg/sentry/platform/kvm/kvm_const_arm64.go11
-rw-r--r--pkg/sentry/platform/kvm/machine_arm64.go72
-rw-r--r--pkg/sentry/platform/kvm/testutil/testutil_arm64.go4
-rw-r--r--pkg/sentry/platform/kvm/testutil/testutil_arm64.s14
-rw-r--r--pkg/sentry/socket/netstack/netstack.go17
-rw-r--r--pkg/sentry/strace/socket.go1
-rw-r--r--pkg/sentry/syscalls/linux/sys_file.go8
-rw-r--r--pkg/sentry/syscalls/linux/vfs2/fd.go4
-rw-r--r--pkg/sentry/syscalls/linux/vfs2/filesystem.go50
-rw-r--r--pkg/sentry/syscalls/linux/vfs2/setstat.go50
-rw-r--r--pkg/sentry/vfs/epoll.go4
-rw-r--r--pkg/sentry/vfs/file_description.go7
-rw-r--r--pkg/syserror/syserror.go1
-rw-r--r--pkg/tcpip/network/fragmentation/fragmentation.go94
-rw-r--r--pkg/tcpip/network/fragmentation/fragmentation_test.go139
-rw-r--r--pkg/tcpip/network/fragmentation/reassembler.go4
-rw-r--r--pkg/tcpip/network/fragmentation/reassembler_test.go2
-rw-r--r--pkg/tcpip/network/ipv4/ipv4.go19
-rw-r--r--pkg/tcpip/network/ipv4/ipv4_test.go193
-rw-r--r--pkg/tcpip/network/ipv6/BUILD1
-rw-r--r--pkg/tcpip/network/ipv6/ipv6.go17
-rw-r--r--pkg/tcpip/network/ipv6/ipv6_test.go357
-rw-r--r--pkg/tcpip/stack/BUILD25
-rw-r--r--pkg/tcpip/stack/conntrack.go26
-rw-r--r--pkg/tcpip/stack/fake_time_test.go209
-rw-r--r--pkg/tcpip/stack/iptables.go11
-rw-r--r--pkg/tcpip/stack/ndp.go16
-rw-r--r--pkg/tcpip/stack/neighbor_cache.go335
-rw-r--r--pkg/tcpip/stack/neighbor_cache_test.go1752
-rw-r--r--pkg/tcpip/stack/neighbor_entry.go482
-rw-r--r--pkg/tcpip/stack/neighbor_entry_test.go2770
-rw-r--r--pkg/tcpip/stack/neighborstate_string.go44
-rw-r--r--pkg/tcpip/stack/nic.go33
-rw-r--r--pkg/tcpip/stack/nud.go466
-rw-r--r--pkg/tcpip/stack/nud_test.go795
-rw-r--r--pkg/tcpip/stack/route.go10
-rw-r--r--pkg/tcpip/stack/stack.go63
-rw-r--r--pkg/tcpip/stack/stack_test.go223
-rw-r--r--pkg/tcpip/tcpip.go30
-rw-r--r--pkg/tcpip/transport/tcp/connect.go11
-rw-r--r--pkg/tcpip/transport/tcp/endpoint.go11
-rw-r--r--pkg/tcpip/transport/tcp/protocol.go2
-rw-r--r--pkg/tcpip/transport/tcp/tcp_test.go147
-rw-r--r--pkg/tcpip/transport/udp/endpoint.go8
-rw-r--r--pkg/tcpip/transport/udp/udp_test.go189
-rw-r--r--pkg/test/dockerutil/container.go16
-rw-r--r--runsc/boot/vfs.go13
-rw-r--r--runsc/fsgofer/fsgofer.go9
-rw-r--r--test/benchmarks/database/BUILD1
-rw-r--r--test/benchmarks/database/redis_test.go86
-rw-r--r--test/benchmarks/fs/BUILD12
-rw-r--r--test/benchmarks/fs/bazel_test.go30
-rw-r--r--test/benchmarks/fs/fio_test.go170
-rw-r--r--test/benchmarks/network/BUILD3
-rw-r--r--test/benchmarks/network/httpd_test.go166
-rw-r--r--test/benchmarks/network/iperf_test.go49
-rw-r--r--test/benchmarks/network/nginx_test.go104
-rw-r--r--test/benchmarks/network/node_test.go131
-rw-r--r--test/benchmarks/tools/BUILD29
-rw-r--r--test/benchmarks/tools/ab.go94
-rw-r--r--test/benchmarks/tools/ab_test.go90
-rw-r--r--test/benchmarks/tools/fio.go124
-rw-r--r--test/benchmarks/tools/fio_test.go122
-rw-r--r--test/benchmarks/tools/hey.go75
-rw-r--r--test/benchmarks/tools/hey_test.go81
-rw-r--r--test/benchmarks/tools/iperf.go56
-rw-r--r--test/benchmarks/tools/iperf_test.go34
-rw-r--r--test/benchmarks/tools/redis.go64
-rw-r--r--test/benchmarks/tools/redis_test.go87
-rw-r--r--test/benchmarks/tools/tools.go17
-rw-r--r--test/e2e/integration_test.go32
-rw-r--r--test/iptables/BUILD1
-rw-r--r--test/iptables/iptables_test.go8
-rw-r--r--test/iptables/iptables_unsafe.go63
-rw-r--r--test/iptables/iptables_util.go51
-rw-r--r--test/iptables/nat.go152
-rw-r--r--test/runtimes/BUILD1
-rw-r--r--test/runtimes/defs.bzl6
-rw-r--r--test/runtimes/exclude_go1.12.csv11
-rw-r--r--test/runtimes/exclude_java11.csv93
-rw-r--r--test/runtimes/proctor/go.go4
-rw-r--r--test/runtimes/proctor/java.go7
-rw-r--r--test/runtimes/runner/main.go2
-rw-r--r--test/syscalls/linux/epoll.cc22
-rw-r--r--test/syscalls/linux/fcntl.cc38
-rw-r--r--test/syscalls/linux/raw_socket_hdrincl.cc4
-rw-r--r--test/syscalls/linux/socket_inet_loopback_nogotsan.cc5
-rw-r--r--test/syscalls/linux/socket_netlink_route.cc5
-rw-r--r--tools/bazel.mk5
-rwxr-xr-xtools/make_release.sh3
121 files changed, 11467 insertions, 744 deletions
diff --git a/.bazelrc b/.bazelrc
index 17530a682..a2fe95822 100644
--- a/.bazelrc
+++ b/.bazelrc
@@ -50,14 +50,14 @@ build:remote --extra_toolchains=//tools/bazeldefs:cc-toolchain-clang-x86_64-defa
build:remote --extra_execution_platforms=//tools/bazeldefs:rbe_ubuntu1604
build:remote --platforms=//tools/bazeldefs:rbe_ubuntu1604
build:remote --crosstool_top=@rbe_default//cc:toolchain
-build:remote --jobs=50
+build:remote --jobs=100
build:remote --remote_timeout=3600
build:remote3 --host_platform=//tools/bazeldefs:rbe_ubuntu1604_bazel3
build:remote3 --extra_toolchains=//tools/bazeldefs:cc-toolchain-clang-x86_64-default_bazel3
build:remote3 --extra_execution_platforms=//tools/bazeldefs:rbe_ubuntu1604_bazel3
build:remote3 --platforms=//tools/bazeldefs:rbe_ubuntu1604_bazel3
build:remote3 --crosstool_top=@rbe_default//cc:toolchain
-build:remote3 --jobs=50
+build:remote3 --jobs=100
build:remote3 --remote_timeout=3600
# Set flags for uploading to BES in order to view results in the Bazel Build
diff --git a/.github/workflows/go.yml b/.github/workflows/go.yml
index a81d06cca..0869fb8db 100644
--- a/.github/workflows/go.yml
+++ b/.github/workflows/go.yml
@@ -14,10 +14,11 @@ jobs:
- id: setup
run: |
if ! [[ -z "${{ secrets.GO_TOKEN }}" ]]; then
- echo ::set-output has_token=true
+ echo ::set-output name=has_token::true
else
- echo ::set-output has_token=false
+ echo ::set-output name=has_token::false
fi
+ - run: |
jq -nc '{"state": "pending", "context": "go tests"}' | \
curl -sL -X POST -d @- \
-H "Content-Type: application/json" \
diff --git a/.travis.yml b/.travis.yml
index 1d955b05d..2d9fa80a1 100644
--- a/.travis.yml
+++ b/.travis.yml
@@ -40,7 +40,7 @@ script:
# On arm64, we need to create our own pipes for stderr and stdout,
# otherwise we will not be able to open /dev/stderr. This is probably
# due to AppArmor rules.
- - bash -xeo pipefail -c 'uname -a && make smoke-test 2>&1 | cat'
+ - bash -xeo pipefail -c 'uname -a && make smoke-tests 2>&1 | cat'
branches:
except:
# Skip copybara branches.
diff --git a/Makefile b/Makefile
index 8e97fc978..5a9367de8 100644
--- a/Makefile
+++ b/Makefile
@@ -117,7 +117,7 @@ debian: ## Builds the debian packages.
@$(call submake,build OPTIONS="-c opt" TARGETS="//runsc:runsc-debian")
.PHONY: debian
-smoke-test: ## Runs a simple smoke test after build runsc.
+smoke-tests: ## Runs a simple smoke test after build runsc.
@$(call submake,run DOCKER_PRIVILEGED="" ARGS="--alsologtostderr --network none --debug --TESTONLY-unsafe-nonroot=true --rootless do true")
.PHONY: smoke-tests
@@ -155,7 +155,7 @@ syscall-tests: syscall-ptrace-tests syscall-kvm-tests syscall-native-tests
%-runtime-tests: load-runtimes_%
@$(call submake,install-test-runtime)
- @$(call submake,test-runtime TARGETS="//test/runtimes:$*")
+ @$(call submake,test-runtime OPTIONS="--test_timeout=10800" TARGETS="//test/runtimes:$*")
do-tests: runsc
@$(call submake,run TARGETS="//runsc" ARGS="--rootless do true")
@@ -218,7 +218,7 @@ packetimpact-tests: load-packetimpact
root-tests: load-basic-images
@$(call submake,install-test-runtime)
@$(call submake,sudo TARGETS="//test/root:root_test" ARGS="-test.v")
-.PHONY: test-root
+.PHONY: root-tests
# Specific containerd version tests.
containerd-test-%: load-basic_alpine load-basic_python load-basic_busybox load-basic_resolv load-basic_httpd install-test-runtime
@@ -266,7 +266,7 @@ website-push: website-build ## Push a new image and update the service.
website-deploy: website-push ## Deploy a new version of the website.
@gcloud run deploy $(WEBSITE_SERVICE) --platform=managed --region=$(WEBSITE_REGION) --project=$(WEBSITE_PROJECT) --image=$(WEBSITE_IMAGE)
-.PHONY: website-push
+.PHONY: website-deploy
##
## Repository builders.
@@ -349,7 +349,7 @@ dev: ## Installs a set of local runtimes. Requires sudo.
refresh: ## Refreshes the runtime binary (for development only). Must have called 'dev' or 'install-test-runtime' first.
@mkdir -p "$(RUNTIME_DIR)"
@$(call submake,copy TARGETS=runsc DESTINATION="$(RUNTIME_BIN)")
-.PHONY: install
+.PHONY: refresh
install-test-runtime: ## Installs the runtime for testing. Requires sudo.
@$(call submake,refresh ARGS="--net-raw --TESTONLY-test-name-env=RUNSC_TEST_NAME --debug --strace --log-packets $(ARGS)")
@@ -370,5 +370,5 @@ configure: ## Configures a single runtime. Requires sudo. Typically called from
.PHONY: configure
test-runtime: ## A convenient wrapper around test that provides the runtime argument. Target must still be provided.
- @$(call submake,test OPTIONS="$(OPTIONS) --test_output=streamed --test_arg=--runtime=$(RUNTIME)")
+ @$(call submake,test OPTIONS="$(OPTIONS) --test_arg=--runtime=$(RUNTIME)")
.PHONY: test-runtime
diff --git a/WORKSPACE b/WORKSPACE
index 49f231755..058d1b306 100644
--- a/WORKSPACE
+++ b/WORKSPACE
@@ -1089,3 +1089,10 @@ go_repository(
sum = "h1:iVjPR7a6H0tWELX5NxNe7bYopibicUzc7uPribsnS6o=",
version = "v1.0.0",
)
+
+go_repository(
+ name = "com_github_dpjacques_clockwork",
+ importpath = "github.com/dpjacques/clockwork",
+ sum = "h1:7krODee+eIlZYoLiEDmP1kLFNCvd0bQ0eEXOympdN6U=",
+ version = "v0.1.1-0.20190114191937-d864eecc357b",
+)
diff --git a/benchmarks/workloads/ruby/Gemfile.lock b/benchmarks/workloads/ruby/Gemfile.lock
index ea9f0ea85..82bfc0c79 100644
--- a/benchmarks/workloads/ruby/Gemfile.lock
+++ b/benchmarks/workloads/ruby/Gemfile.lock
@@ -6,22 +6,23 @@ GEM
builder (>= 2.1.2, < 4.0.0)
i18n (>= 0.6.9)
nokogiri (~> 1.4)
- activesupport (5.2.3)
+ activesupport (6.0.3.2)
concurrent-ruby (~> 1.0, >= 1.0.2)
i18n (>= 0.7, < 2)
minitest (~> 5.1)
tzinfo (~> 1.1)
+ zeitwerk (~> 2.2, >= 2.2.2)
bcrypt (3.1.13)
builder (3.2.4)
cassandra-driver (3.2.3)
ione (~> 1.2)
- concurrent-ruby (1.1.5)
+ concurrent-ruby (1.1.6)
ffi (1.12.2)
- i18n (1.6.0)
+ i18n (1.8.5)
concurrent-ruby (~> 1.0)
ione (1.2.4)
mini_portile2 (2.4.0)
- minitest (5.11.3)
+ minitest (5.14.1)
mustermann (1.0.3)
nokogiri (1.10.8)
mini_portile2 (~> 2.4.0)
@@ -49,8 +50,9 @@ GEM
thread_safe (0.3.6)
tilt (2.0.9)
ttfunk (1.5.1)
- tzinfo (1.2.5)
+ tzinfo (1.2.7)
thread_safe (~> 0.1)
+ zeitwerk (2.4.0)
PLATFORMS
ruby
diff --git a/go.mod b/go.mod
index 0875b4ba0..2fcba5cc9 100644
--- a/go.mod
+++ b/go.mod
@@ -22,6 +22,7 @@ require (
github.com/docker/go-connections v0.3.0 // indirect
github.com/docker/go-events v0.0.0-20190806004212-e31b211e4f1c // indirect
github.com/docker/go-units v0.4.0 // indirect
+ github.com/dpjacques/clockwork v0.1.1-0.20190114191937-d864eecc357b // indirect
github.com/godbus/dbus v0.0.0-20190422162347-ade71ed3457e // indirect
github.com/gofrs/flock v0.6.1-0.20180915234121-886344bea079 // indirect
github.com/gogo/googleapis v1.4.0 // indirect
@@ -43,7 +44,6 @@ require (
github.com/vishvananda/netns v0.0.0-20200520041808-52d707b772fe // indirect
go.uber.org/atomic v1.6.0 // indirect
go.uber.org/multierr v1.2.0 // indirect
- golang.org/x/sys v0.0.0-20200302150141-5c8b2ff67527 // indirect
golang.org/x/time v0.0.0-20191024005414-555d28b269f0 // indirect
golang.org/x/tools v0.0.0-20200707200213-416e8f4faf8a // indirect
google.golang.org/grpc v1.29.0 // indirect
diff --git a/go.sum b/go.sum
index a90bca394..f98132971 100644
--- a/go.sum
+++ b/go.sum
@@ -74,6 +74,8 @@ github.com/docker/go-events v0.0.0-20190806004212-e31b211e4f1c h1:+pKlWGMw7gf6bQ
github.com/docker/go-events v0.0.0-20190806004212-e31b211e4f1c/go.mod h1:Uw6UezgYA44ePAFQYUehOuCzmy5zmg/+nl2ZfMWGkpA=
github.com/docker/go-units v0.4.0 h1:3uh0PgVws3nIA0Q+MwDC8yjEPf9zjRfZZWXZYDct3Tw=
github.com/docker/go-units v0.4.0/go.mod h1:fgPhTUdO+D/Jk86RDLlptpiXQzgHJF7gydDDbaIK4Dk=
+github.com/dpjacques/clockwork v0.1.1-0.20190114191937-d864eecc357b h1:7krODee+eIlZYoLiEDmP1kLFNCvd0bQ0eEXOympdN6U=
+github.com/dpjacques/clockwork v0.1.1-0.20190114191937-d864eecc357b/go.mod h1:D8mP2A8vVT2GkXqPorSBmhnshhkFBYgzhA90KmJt25Y=
github.com/dustin/go-humanize v0.0.0-20171111073723-bb3d318650d4/go.mod h1:HtrtbFcZ19U5GC7JDqmcUSB87Iq5E25KnS6fMYU6eOk=
github.com/envoyproxy/go-control-plane v0.9.0/go.mod h1:YTl/9mNaCwkRvm6d1a2C3ymFceY/DCBVvsKhRF0iEA4=
github.com/envoyproxy/go-control-plane v0.9.1-0.20191026205805-5f8ba28d4473/go.mod h1:YTl/9mNaCwkRvm6d1a2C3ymFceY/DCBVvsKhRF0iEA4=
@@ -288,6 +290,7 @@ golang.org/x/sys v0.0.0-20191210023423-ac6580df4449/go.mod h1:h1NjWce9XRLGQEsW7w
golang.org/x/sys v0.0.0-20200113162924-86b910548bc1/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
golang.org/x/sys v0.0.0-20200120151820-655fe14d7479/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
golang.org/x/sys v0.0.0-20200217220822-9197077df867/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
+golang.org/x/sys v0.0.0-20200302150141-5c8b2ff67527/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
golang.org/x/sys v0.0.0-20200323222414-85ca7c5b95cd/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
golang.org/x/sys v0.0.0-20200523222454-059865788121 h1:rITEj+UZHYC927n8GT97eC3zrpzXdb/voyeOuVKS46o=
golang.org/x/sys v0.0.0-20200523222454-059865788121/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
diff --git a/images/basic/hostoverlaytest/Dockerfile b/images/basic/hostoverlaytest/Dockerfile
index d83439e9c..6cef1a542 100644
--- a/images/basic/hostoverlaytest/Dockerfile
+++ b/images/basic/hostoverlaytest/Dockerfile
@@ -4,4 +4,5 @@ WORKDIR /root
COPY . .
RUN apt-get update && apt-get install -y gcc
-RUN gcc -O2 -o test test.c
+RUN gcc -O2 -o test_copy_up test_copy_up.c
+RUN gcc -O2 -o test_rewinddir test_rewinddir.c
diff --git a/images/basic/hostoverlaytest/testfile.txt b/images/basic/hostoverlaytest/copy_up_testfile.txt
index e4188c841..e4188c841 100644
--- a/images/basic/hostoverlaytest/testfile.txt
+++ b/images/basic/hostoverlaytest/copy_up_testfile.txt
diff --git a/images/basic/hostoverlaytest/test.c b/images/basic/hostoverlaytest/test_copy_up.c
index 088f90746..010b261dc 100644
--- a/images/basic/hostoverlaytest/test.c
+++ b/images/basic/hostoverlaytest/test_copy_up.c
@@ -6,7 +6,7 @@
#include <unistd.h>
int main(int argc, char** argv) {
- const char kTestFilePath[] = "testfile.txt";
+ const char kTestFilePath[] = "copy_up_testfile.txt";
const char kOldFileData[] = "old data\n";
const char kNewFileData[] = "new data\n";
const size_t kPageSize = sysconf(_SC_PAGE_SIZE);
diff --git a/images/basic/hostoverlaytest/test_rewinddir.c b/images/basic/hostoverlaytest/test_rewinddir.c
new file mode 100644
index 000000000..f1a4085e1
--- /dev/null
+++ b/images/basic/hostoverlaytest/test_rewinddir.c
@@ -0,0 +1,78 @@
+#include <dirent.h>
+#include <err.h>
+#include <errno.h>
+#include <stdlib.h>
+#include <string.h>
+#include <sys/stat.h>
+#include <sys/types.h>
+
+int main(int argc, char** argv) {
+ const char kDirPath[] = "rewinddir_test_dir";
+ const char kFileBasename[] = "rewinddir_test_file";
+
+ // Create the test directory.
+ if (mkdir(kDirPath, 0755) < 0) {
+ err(1, "mkdir(%s)", kDirPath);
+ }
+
+ // The test directory should initially be empty.
+ DIR* dir = opendir(kDirPath);
+ if (!dir) {
+ err(1, "opendir(%s)", kDirPath);
+ }
+ int failed = 0;
+ while (1) {
+ errno = 0;
+ struct dirent* d = readdir(dir);
+ if (!d) {
+ if (errno != 0) {
+ err(1, "readdir");
+ }
+ break;
+ }
+ if (strcmp(d->d_name, ".") != 0 && strcmp(d->d_name, "..") != 0) {
+ warnx("unexpected file %s in new directory", d->d_name);
+ failed = 1;
+ }
+ }
+
+ // Create a file in the test directory.
+ char* file_path = malloc(strlen(kDirPath) + 1 + strlen(kFileBasename));
+ if (!file_path) {
+ errx(1, "malloc");
+ }
+ strcpy(file_path, kDirPath);
+ file_path[strlen(kDirPath)] = '/';
+ strcpy(file_path + strlen(kDirPath) + 1, kFileBasename);
+ if (mknod(file_path, 0644, 0) < 0) {
+ err(1, "mknod(%s)", file_path);
+ }
+
+ // After rewinddir(), re-reading the directory stream should yield the new
+ // file.
+ rewinddir(dir);
+ size_t found_file = 0;
+ while (1) {
+ errno = 0;
+ struct dirent* d = readdir(dir);
+ if (!d) {
+ if (errno != 0) {
+ err(1, "readdir");
+ }
+ break;
+ }
+ if (strcmp(d->d_name, kFileBasename) == 0) {
+ found_file++;
+ } else if (strcmp(d->d_name, ".") != 0 && strcmp(d->d_name, "..") != 0) {
+ warnx("unexpected file %s in new directory", d->d_name);
+ failed = 1;
+ }
+ }
+ if (found_file != 1) {
+ warnx("readdir returned file %s %zu times, wanted 1", kFileBasename,
+ found_file);
+ failed = 1;
+ }
+
+ return failed;
+}
diff --git a/images/benchmarks/fio/Dockerfile b/images/benchmarks/fio/Dockerfile
new file mode 100644
index 000000000..9531df7fa
--- /dev/null
+++ b/images/benchmarks/fio/Dockerfile
@@ -0,0 +1,7 @@
+FROM ubuntu:18.04
+
+RUN set -x \
+ && apt-get update \
+ && apt-get install -y \
+ fio \
+ && rm -rf /var/lib/apt/lists/*
diff --git a/images/benchmarks/hey/Dockerfile b/images/benchmarks/hey/Dockerfile
new file mode 100644
index 000000000..f586978b6
--- /dev/null
+++ b/images/benchmarks/hey/Dockerfile
@@ -0,0 +1,12 @@
+FROM ubuntu:18.04
+
+RUN set -x \
+ && apt-get update \
+ && apt-get install -y \
+ wget \
+ && rm -rf /var/lib/apt/lists/*
+
+RUN wget https://storage.googleapis.com/hey-release/hey_linux_amd64 \
+ && chmod 777 hey_linux_amd64 \
+ && cp hey_linux_amd64 /bin/hey \
+ && rm hey_linux_amd64
diff --git a/images/benchmarks/nginx/Dockerfile b/images/benchmarks/nginx/Dockerfile
new file mode 100644
index 000000000..b64eb52ae
--- /dev/null
+++ b/images/benchmarks/nginx/Dockerfile
@@ -0,0 +1 @@
+FROM nginx:1.15.10
diff --git a/images/benchmarks/node/Dockerfile b/images/benchmarks/node/Dockerfile
new file mode 100644
index 000000000..bf45650a0
--- /dev/null
+++ b/images/benchmarks/node/Dockerfile
@@ -0,0 +1 @@
+FROM node:onbuild
diff --git a/images/benchmarks/node/index.hbs b/images/benchmarks/node/index.hbs
new file mode 100644
index 000000000..03feceb75
--- /dev/null
+++ b/images/benchmarks/node/index.hbs
@@ -0,0 +1,8 @@
+<!DOCTYPE html>
+<html>
+<body>
+ {{#each text}}
+ <p>{{this}}</p>
+ {{/each}}
+</body>
+</html>
diff --git a/images/benchmarks/node/index.js b/images/benchmarks/node/index.js
new file mode 100644
index 000000000..831015d18
--- /dev/null
+++ b/images/benchmarks/node/index.js
@@ -0,0 +1,42 @@
+const app = require('express')();
+const path = require('path');
+const redis = require('redis');
+const srs = require('secure-random-string');
+
+// The hostname is the first argument.
+const host_name = process.argv[2];
+
+var client = redis.createClient({host: host_name, detect_buffers: true});
+
+app.set('views', __dirname);
+app.set('view engine', 'hbs');
+
+app.get('/', (req, res) => {
+ var tmp = [];
+ /* Pull four random keys from the redis server. */
+ for (i = 0; i < 4; i++) {
+ client.get(Math.floor(Math.random() * (100)), function(err, reply) {
+ tmp.push(reply.toString());
+ });
+ }
+ res.render('index', {text: tmp});
+});
+
+/**
+ * Securely generate a random string.
+ * @param {number} len
+ * @return {string}
+ */
+function randomBody(len) {
+ return srs({alphanumeric: true, length: len});
+}
+
+/** Mutates one hundred keys randomly. */
+function generateText() {
+ for (i = 0; i < 100; i++) {
+ client.set(i, randomBody(1024));
+ }
+}
+
+generateText();
+app.listen(8080);
diff --git a/images/benchmarks/node/package-lock.json b/images/benchmarks/node/package-lock.json
new file mode 100644
index 000000000..580e68aa5
--- /dev/null
+++ b/images/benchmarks/node/package-lock.json
@@ -0,0 +1,486 @@
+{
+ "name": "nodedum",
+ "version": "1.0.0",
+ "lockfileVersion": 1,
+ "requires": true,
+ "dependencies": {
+ "accepts": {
+ "version": "1.3.5",
+ "resolved": "https://registry.npmjs.org/accepts/-/accepts-1.3.5.tgz",
+ "integrity": "sha1-63d99gEXI6OxTopywIBcjoZ0a9I=",
+ "requires": {
+ "mime-types": "~2.1.18",
+ "negotiator": "0.6.1"
+ }
+ },
+ "array-flatten": {
+ "version": "1.1.1",
+ "resolved": "https://registry.npmjs.org/array-flatten/-/array-flatten-1.1.1.tgz",
+ "integrity": "sha1-ml9pkFGx5wczKPKgCJaLZOopVdI="
+ },
+ "async": {
+ "version": "2.6.2",
+ "resolved": "https://registry.npmjs.org/async/-/async-2.6.2.tgz",
+ "integrity": "sha512-H1qVYh1MYhEEFLsP97cVKqCGo7KfCyTt6uEWqsTBr9SO84oK9Uwbyd/yCW+6rKJLHksBNUVWZDAjfS+Ccx0Bbg==",
+ "requires": {
+ "lodash": "^4.17.11"
+ }
+ },
+ "body-parser": {
+ "version": "1.18.3",
+ "resolved": "https://registry.npmjs.org/body-parser/-/body-parser-1.18.3.tgz",
+ "integrity": "sha1-WykhmP/dVTs6DyDe0FkrlWlVyLQ=",
+ "requires": {
+ "bytes": "3.0.0",
+ "content-type": "~1.0.4",
+ "debug": "2.6.9",
+ "depd": "~1.1.2",
+ "http-errors": "~1.6.3",
+ "iconv-lite": "0.4.23",
+ "on-finished": "~2.3.0",
+ "qs": "6.5.2",
+ "raw-body": "2.3.3",
+ "type-is": "~1.6.16"
+ }
+ },
+ "bytes": {
+ "version": "3.0.0",
+ "resolved": "https://registry.npmjs.org/bytes/-/bytes-3.0.0.tgz",
+ "integrity": "sha1-0ygVQE1olpn4Wk6k+odV3ROpYEg="
+ },
+ "commander": {
+ "version": "2.20.0",
+ "resolved": "https://registry.npmjs.org/commander/-/commander-2.20.0.tgz",
+ "integrity": "sha512-7j2y+40w61zy6YC2iRNpUe/NwhNyoXrYpHMrSunaMG64nRnaf96zO/KMQR4OyN/UnE5KLyEBnKHd4aG3rskjpQ==",
+ "optional": true
+ },
+ "content-disposition": {
+ "version": "0.5.2",
+ "resolved": "https://registry.npmjs.org/content-disposition/-/content-disposition-0.5.2.tgz",
+ "integrity": "sha1-DPaLud318r55YcOoUXjLhdunjLQ="
+ },
+ "content-type": {
+ "version": "1.0.4",
+ "resolved": "https://registry.npmjs.org/content-type/-/content-type-1.0.4.tgz",
+ "integrity": "sha512-hIP3EEPs8tB9AT1L+NUqtwOAps4mk2Zob89MWXMHjHWg9milF/j4osnnQLXBCBFBk/tvIG/tUc9mOUJiPBhPXA=="
+ },
+ "cookie": {
+ "version": "0.3.1",
+ "resolved": "https://registry.npmjs.org/cookie/-/cookie-0.3.1.tgz",
+ "integrity": "sha1-5+Ch+e9DtMi6klxcWpboBtFoc7s="
+ },
+ "cookie-signature": {
+ "version": "1.0.6",
+ "resolved": "https://registry.npmjs.org/cookie-signature/-/cookie-signature-1.0.6.tgz",
+ "integrity": "sha1-4wOogrNCzD7oylE6eZmXNNqzriw="
+ },
+ "debug": {
+ "version": "2.6.9",
+ "resolved": "https://registry.npmjs.org/debug/-/debug-2.6.9.tgz",
+ "integrity": "sha512-bC7ElrdJaJnPbAP+1EotYvqZsb3ecl5wi6Bfi6BJTUcNowp6cvspg0jXznRTKDjm/E7AdgFBVeAPVMNcKGsHMA==",
+ "requires": {
+ "ms": "2.0.0"
+ }
+ },
+ "depd": {
+ "version": "1.1.2",
+ "resolved": "https://registry.npmjs.org/depd/-/depd-1.1.2.tgz",
+ "integrity": "sha1-m81S4UwJd2PnSbJ0xDRu0uVgtak="
+ },
+ "destroy": {
+ "version": "1.0.4",
+ "resolved": "https://registry.npmjs.org/destroy/-/destroy-1.0.4.tgz",
+ "integrity": "sha1-l4hXRCxEdJ5CBmE+N5RiBYJqvYA="
+ },
+ "double-ended-queue": {
+ "version": "2.1.0-0",
+ "resolved": "https://registry.npmjs.org/double-ended-queue/-/double-ended-queue-2.1.0-0.tgz",
+ "integrity": "sha1-ED01J/0xUo9AGIEwyEHv3XgmTlw="
+ },
+ "ee-first": {
+ "version": "1.1.1",
+ "resolved": "https://registry.npmjs.org/ee-first/-/ee-first-1.1.1.tgz",
+ "integrity": "sha1-WQxhFWsK4vTwJVcyoViyZrxWsh0="
+ },
+ "encodeurl": {
+ "version": "1.0.2",
+ "resolved": "https://registry.npmjs.org/encodeurl/-/encodeurl-1.0.2.tgz",
+ "integrity": "sha1-rT/0yG7C0CkyL1oCw6mmBslbP1k="
+ },
+ "escape-html": {
+ "version": "1.0.3",
+ "resolved": "https://registry.npmjs.org/escape-html/-/escape-html-1.0.3.tgz",
+ "integrity": "sha1-Aljq5NPQwJdN4cFpGI7wBR0dGYg="
+ },
+ "etag": {
+ "version": "1.8.1",
+ "resolved": "https://registry.npmjs.org/etag/-/etag-1.8.1.tgz",
+ "integrity": "sha1-Qa4u62XvpiJorr/qg6x9eSmbCIc="
+ },
+ "express": {
+ "version": "4.16.4",
+ "resolved": "https://registry.npmjs.org/express/-/express-4.16.4.tgz",
+ "integrity": "sha512-j12Uuyb4FMrd/qQAm6uCHAkPtO8FDTRJZBDd5D2KOL2eLaz1yUNdUB/NOIyq0iU4q4cFarsUCrnFDPBcnksuOg==",
+ "requires": {
+ "accepts": "~1.3.5",
+ "array-flatten": "1.1.1",
+ "body-parser": "1.18.3",
+ "content-disposition": "0.5.2",
+ "content-type": "~1.0.4",
+ "cookie": "0.3.1",
+ "cookie-signature": "1.0.6",
+ "debug": "2.6.9",
+ "depd": "~1.1.2",
+ "encodeurl": "~1.0.2",
+ "escape-html": "~1.0.3",
+ "etag": "~1.8.1",
+ "finalhandler": "1.1.1",
+ "fresh": "0.5.2",
+ "merge-descriptors": "1.0.1",
+ "methods": "~1.1.2",
+ "on-finished": "~2.3.0",
+ "parseurl": "~1.3.2",
+ "path-to-regexp": "0.1.7",
+ "proxy-addr": "~2.0.4",
+ "qs": "6.5.2",
+ "range-parser": "~1.2.0",
+ "safe-buffer": "5.1.2",
+ "send": "0.16.2",
+ "serve-static": "1.13.2",
+ "setprototypeof": "1.1.0",
+ "statuses": "~1.4.0",
+ "type-is": "~1.6.16",
+ "utils-merge": "1.0.1",
+ "vary": "~1.1.2"
+ }
+ },
+ "finalhandler": {
+ "version": "1.1.1",
+ "resolved": "http://registry.npmjs.org/finalhandler/-/finalhandler-1.1.1.tgz",
+ "integrity": "sha512-Y1GUDo39ez4aHAw7MysnUD5JzYX+WaIj8I57kO3aEPT1fFRL4sr7mjei97FgnwhAyyzRYmQZaTHb2+9uZ1dPtg==",
+ "requires": {
+ "debug": "2.6.9",
+ "encodeurl": "~1.0.2",
+ "escape-html": "~1.0.3",
+ "on-finished": "~2.3.0",
+ "parseurl": "~1.3.2",
+ "statuses": "~1.4.0",
+ "unpipe": "~1.0.0"
+ }
+ },
+ "foreachasync": {
+ "version": "3.0.0",
+ "resolved": "https://registry.npmjs.org/foreachasync/-/foreachasync-3.0.0.tgz",
+ "integrity": "sha1-VQKYfchxS+M5IJfzLgBxyd7gfPY="
+ },
+ "forwarded": {
+ "version": "0.1.2",
+ "resolved": "https://registry.npmjs.org/forwarded/-/forwarded-0.1.2.tgz",
+ "integrity": "sha1-mMI9qxF1ZXuMBXPozszZGw/xjIQ="
+ },
+ "fresh": {
+ "version": "0.5.2",
+ "resolved": "https://registry.npmjs.org/fresh/-/fresh-0.5.2.tgz",
+ "integrity": "sha1-PYyt2Q2XZWn6g1qx+OSyOhBWBac="
+ },
+ "handlebars": {
+ "version": "4.0.14",
+ "resolved": "https://registry.npmjs.org/handlebars/-/handlebars-4.0.14.tgz",
+ "integrity": "sha512-E7tDoyAA8ilZIV3xDJgl18sX3M8xB9/fMw8+mfW4msLW8jlX97bAnWgT3pmaNXuvzIEgSBMnAHfuXsB2hdzfow==",
+ "requires": {
+ "async": "^2.5.0",
+ "optimist": "^0.6.1",
+ "source-map": "^0.6.1",
+ "uglify-js": "^3.1.4"
+ }
+ },
+ "hbs": {
+ "version": "4.0.4",
+ "resolved": "https://registry.npmjs.org/hbs/-/hbs-4.0.4.tgz",
+ "integrity": "sha512-esVlyV/V59mKkwFai5YmPRSNIWZzhqL5YMN0++ueMxyK1cCfPa5f6JiHtapPKAIVAhQR6rpGxow0troav9WMEg==",
+ "requires": {
+ "handlebars": "4.0.14",
+ "walk": "2.3.9"
+ }
+ },
+ "http-errors": {
+ "version": "1.6.3",
+ "resolved": "http://registry.npmjs.org/http-errors/-/http-errors-1.6.3.tgz",
+ "integrity": "sha1-i1VoC7S+KDoLW/TqLjhYC+HZMg0=",
+ "requires": {
+ "depd": "~1.1.2",
+ "inherits": "2.0.3",
+ "setprototypeof": "1.1.0",
+ "statuses": ">= 1.4.0 < 2"
+ }
+ },
+ "iconv-lite": {
+ "version": "0.4.23",
+ "resolved": "https://registry.npmjs.org/iconv-lite/-/iconv-lite-0.4.23.tgz",
+ "integrity": "sha512-neyTUVFtahjf0mB3dZT77u+8O0QB89jFdnBkd5P1JgYPbPaia3gXXOVL2fq8VyU2gMMD7SaN7QukTB/pmXYvDA==",
+ "requires": {
+ "safer-buffer": ">= 2.1.2 < 3"
+ }
+ },
+ "inherits": {
+ "version": "2.0.3",
+ "resolved": "https://registry.npmjs.org/inherits/-/inherits-2.0.3.tgz",
+ "integrity": "sha1-Yzwsg+PaQqUC9SRmAiSA9CCCYd4="
+ },
+ "ipaddr.js": {
+ "version": "1.8.0",
+ "resolved": "https://registry.npmjs.org/ipaddr.js/-/ipaddr.js-1.8.0.tgz",
+ "integrity": "sha1-6qM9bd16zo9/b+DJygRA5wZzix4="
+ },
+ "lodash": {
+ "version": "4.17.15",
+ "resolved": "https://registry.npmjs.org/lodash/-/lodash-4.17.15.tgz",
+ "integrity": "sha512-8xOcRHvCjnocdS5cpwXQXVzmmh5e5+saE2QGoeQmbKmRS6J3VQppPOIt0MnmE+4xlZoumy0GPG0D0MVIQbNA1A=="
+ },
+ "media-typer": {
+ "version": "0.3.0",
+ "resolved": "http://registry.npmjs.org/media-typer/-/media-typer-0.3.0.tgz",
+ "integrity": "sha1-hxDXrwqmJvj/+hzgAWhUUmMlV0g="
+ },
+ "merge-descriptors": {
+ "version": "1.0.1",
+ "resolved": "https://registry.npmjs.org/merge-descriptors/-/merge-descriptors-1.0.1.tgz",
+ "integrity": "sha1-sAqqVW3YtEVoFQ7J0blT8/kMu2E="
+ },
+ "methods": {
+ "version": "1.1.2",
+ "resolved": "https://registry.npmjs.org/methods/-/methods-1.1.2.tgz",
+ "integrity": "sha1-VSmk1nZUE07cxSZmVoNbD4Ua/O4="
+ },
+ "mime": {
+ "version": "1.4.1",
+ "resolved": "https://registry.npmjs.org/mime/-/mime-1.4.1.tgz",
+ "integrity": "sha512-KI1+qOZu5DcW6wayYHSzR/tXKCDC5Om4s1z2QJjDULzLcmf3DvzS7oluY4HCTrc+9FiKmWUgeNLg7W3uIQvxtQ=="
+ },
+ "mime-db": {
+ "version": "1.37.0",
+ "resolved": "https://registry.npmjs.org/mime-db/-/mime-db-1.37.0.tgz",
+ "integrity": "sha512-R3C4db6bgQhlIhPU48fUtdVmKnflq+hRdad7IyKhtFj06VPNVdk2RhiYL3UjQIlso8L+YxAtFkobT0VK+S/ybg=="
+ },
+ "mime-types": {
+ "version": "2.1.21",
+ "resolved": "https://registry.npmjs.org/mime-types/-/mime-types-2.1.21.tgz",
+ "integrity": "sha512-3iL6DbwpyLzjR3xHSFNFeb9Nz/M8WDkX33t1GFQnFOllWk8pOrh/LSrB5OXlnlW5P9LH73X6loW/eogc+F5lJg==",
+ "requires": {
+ "mime-db": "~1.37.0"
+ }
+ },
+ "minimist": {
+ "version": "0.0.10",
+ "resolved": "https://registry.npmjs.org/minimist/-/minimist-0.0.10.tgz",
+ "integrity": "sha1-3j+YVD2/lggr5IrRoMfNqDYwHc8="
+ },
+ "ms": {
+ "version": "2.0.0",
+ "resolved": "https://registry.npmjs.org/ms/-/ms-2.0.0.tgz",
+ "integrity": "sha1-VgiurfwAvmwpAd9fmGF4jeDVl8g="
+ },
+ "negotiator": {
+ "version": "0.6.1",
+ "resolved": "https://registry.npmjs.org/negotiator/-/negotiator-0.6.1.tgz",
+ "integrity": "sha1-KzJxhOiZIQEXeyhWP7XnECrNDKk="
+ },
+ "on-finished": {
+ "version": "2.3.0",
+ "resolved": "https://registry.npmjs.org/on-finished/-/on-finished-2.3.0.tgz",
+ "integrity": "sha1-IPEzZIGwg811M3mSoWlxqi2QaUc=",
+ "requires": {
+ "ee-first": "1.1.1"
+ }
+ },
+ "optimist": {
+ "version": "0.6.1",
+ "resolved": "https://registry.npmjs.org/optimist/-/optimist-0.6.1.tgz",
+ "integrity": "sha1-2j6nRob6IaGaERwybpDrFaAZZoY=",
+ "requires": {
+ "minimist": "~0.0.1",
+ "wordwrap": "~0.0.2"
+ }
+ },
+ "parseurl": {
+ "version": "1.3.2",
+ "resolved": "https://registry.npmjs.org/parseurl/-/parseurl-1.3.2.tgz",
+ "integrity": "sha1-/CidTtiZMRlGDBViUyYs3I3mW/M="
+ },
+ "path-to-regexp": {
+ "version": "0.1.7",
+ "resolved": "https://registry.npmjs.org/path-to-regexp/-/path-to-regexp-0.1.7.tgz",
+ "integrity": "sha1-32BBeABfUi8V60SQ5yR6G/qmf4w="
+ },
+ "proxy-addr": {
+ "version": "2.0.4",
+ "resolved": "https://registry.npmjs.org/proxy-addr/-/proxy-addr-2.0.4.tgz",
+ "integrity": "sha512-5erio2h9jp5CHGwcybmxmVqHmnCBZeewlfJ0pex+UW7Qny7OOZXTtH56TGNyBizkgiOwhJtMKrVzDTeKcySZwA==",
+ "requires": {
+ "forwarded": "~0.1.2",
+ "ipaddr.js": "1.8.0"
+ }
+ },
+ "qs": {
+ "version": "6.5.2",
+ "resolved": "https://registry.npmjs.org/qs/-/qs-6.5.2.tgz",
+ "integrity": "sha512-N5ZAX4/LxJmF+7wN74pUD6qAh9/wnvdQcjq9TZjevvXzSUo7bfmw91saqMjzGS2xq91/odN2dW/WOl7qQHNDGA=="
+ },
+ "range-parser": {
+ "version": "1.2.0",
+ "resolved": "https://registry.npmjs.org/range-parser/-/range-parser-1.2.0.tgz",
+ "integrity": "sha1-9JvmtIeJTdxA3MlKMi9hEJLgDV4="
+ },
+ "raw-body": {
+ "version": "2.3.3",
+ "resolved": "https://registry.npmjs.org/raw-body/-/raw-body-2.3.3.tgz",
+ "integrity": "sha512-9esiElv1BrZoI3rCDuOuKCBRbuApGGaDPQfjSflGxdy4oyzqghxu6klEkkVIvBje+FF0BX9coEv8KqW6X/7njw==",
+ "requires": {
+ "bytes": "3.0.0",
+ "http-errors": "1.6.3",
+ "iconv-lite": "0.4.23",
+ "unpipe": "1.0.0"
+ }
+ },
+ "redis": {
+ "version": "2.8.0",
+ "resolved": "https://registry.npmjs.org/redis/-/redis-2.8.0.tgz",
+ "integrity": "sha512-M1OkonEQwtRmZv4tEWF2VgpG0JWJ8Fv1PhlgT5+B+uNq2cA3Rt1Yt/ryoR+vQNOQcIEgdCdfH0jr3bDpihAw1A==",
+ "requires": {
+ "double-ended-queue": "^2.1.0-0",
+ "redis-commands": "^1.2.0",
+ "redis-parser": "^2.6.0"
+ },
+ "dependencies": {
+ "redis-commands": {
+ "version": "1.4.0",
+ "resolved": "https://registry.npmjs.org/redis-commands/-/redis-commands-1.4.0.tgz",
+ "integrity": "sha512-cu8EF+MtkwI4DLIT0x9P8qNTLFhQD4jLfxLR0cCNkeGzs87FN6879JOJwNQR/1zD7aSYNbU0hgsV9zGY71Itvw=="
+ },
+ "redis-parser": {
+ "version": "2.6.0",
+ "resolved": "https://registry.npmjs.org/redis-parser/-/redis-parser-2.6.0.tgz",
+ "integrity": "sha1-Uu0J2srBCPGmMcB+m2mUHnoZUEs="
+ }
+ }
+ },
+ "redis-commands": {
+ "version": "1.5.0",
+ "resolved": "https://registry.npmjs.org/redis-commands/-/redis-commands-1.5.0.tgz",
+ "integrity": "sha512-6KxamqpZ468MeQC3bkWmCB1fp56XL64D4Kf0zJSwDZbVLLm7KFkoIcHrgRvQ+sk8dnhySs7+yBg94yIkAK7aJg=="
+ },
+ "redis-parser": {
+ "version": "2.6.0",
+ "resolved": "https://registry.npmjs.org/redis-parser/-/redis-parser-2.6.0.tgz",
+ "integrity": "sha1-Uu0J2srBCPGmMcB+m2mUHnoZUEs="
+ },
+ "safe-buffer": {
+ "version": "5.1.2",
+ "resolved": "https://registry.npmjs.org/safe-buffer/-/safe-buffer-5.1.2.tgz",
+ "integrity": "sha512-Gd2UZBJDkXlY7GbJxfsE8/nvKkUEU1G38c1siN6QP6a9PT9MmHB8GnpscSmMJSoF8LOIrt8ud/wPtojys4G6+g=="
+ },
+ "safer-buffer": {
+ "version": "2.1.2",
+ "resolved": "https://registry.npmjs.org/safer-buffer/-/safer-buffer-2.1.2.tgz",
+ "integrity": "sha512-YZo3K82SD7Riyi0E1EQPojLz7kpepnSQI9IyPbHHg1XXXevb5dJI7tpyN2ADxGcQbHG7vcyRHk0cbwqcQriUtg=="
+ },
+ "secure-random-string": {
+ "version": "1.1.0",
+ "resolved": "https://registry.npmjs.org/secure-random-string/-/secure-random-string-1.1.0.tgz",
+ "integrity": "sha512-V/h8jqoz58zklNGybVhP++cWrxEPXlLM/6BeJ4e0a8zlb4BsbYRzFs16snrxByPa5LUxCVTD3M6EYIVIHR1fAg=="
+ },
+ "send": {
+ "version": "0.16.2",
+ "resolved": "https://registry.npmjs.org/send/-/send-0.16.2.tgz",
+ "integrity": "sha512-E64YFPUssFHEFBvpbbjr44NCLtI1AohxQ8ZSiJjQLskAdKuriYEP6VyGEsRDH8ScozGpkaX1BGvhanqCwkcEZw==",
+ "requires": {
+ "debug": "2.6.9",
+ "depd": "~1.1.2",
+ "destroy": "~1.0.4",
+ "encodeurl": "~1.0.2",
+ "escape-html": "~1.0.3",
+ "etag": "~1.8.1",
+ "fresh": "0.5.2",
+ "http-errors": "~1.6.2",
+ "mime": "1.4.1",
+ "ms": "2.0.0",
+ "on-finished": "~2.3.0",
+ "range-parser": "~1.2.0",
+ "statuses": "~1.4.0"
+ }
+ },
+ "serve-static": {
+ "version": "1.13.2",
+ "resolved": "https://registry.npmjs.org/serve-static/-/serve-static-1.13.2.tgz",
+ "integrity": "sha512-p/tdJrO4U387R9oMjb1oj7qSMaMfmOyd4j9hOFoxZe2baQszgHcSWjuya/CiT5kgZZKRudHNOA0pYXOl8rQ5nw==",
+ "requires": {
+ "encodeurl": "~1.0.2",
+ "escape-html": "~1.0.3",
+ "parseurl": "~1.3.2",
+ "send": "0.16.2"
+ }
+ },
+ "setprototypeof": {
+ "version": "1.1.0",
+ "resolved": "https://registry.npmjs.org/setprototypeof/-/setprototypeof-1.1.0.tgz",
+ "integrity": "sha512-BvE/TwpZX4FXExxOxZyRGQQv651MSwmWKZGqvmPcRIjDqWub67kTKuIMx43cZZrS/cBBzwBcNDWoFxt2XEFIpQ=="
+ },
+ "source-map": {
+ "version": "0.6.1",
+ "resolved": "https://registry.npmjs.org/source-map/-/source-map-0.6.1.tgz",
+ "integrity": "sha512-UjgapumWlbMhkBgzT7Ykc5YXUT46F0iKu8SGXq0bcwP5dz/h0Plj6enJqjz1Zbq2l5WaqYnrVbwWOWMyF3F47g=="
+ },
+ "statuses": {
+ "version": "1.4.0",
+ "resolved": "https://registry.npmjs.org/statuses/-/statuses-1.4.0.tgz",
+ "integrity": "sha512-zhSCtt8v2NDrRlPQpCNtw/heZLtfUDqxBM1udqikb/Hbk52LK4nQSwr10u77iopCW5LsyHpuXS0GnEc48mLeew=="
+ },
+ "type-is": {
+ "version": "1.6.16",
+ "resolved": "https://registry.npmjs.org/type-is/-/type-is-1.6.16.tgz",
+ "integrity": "sha512-HRkVv/5qY2G6I8iab9cI7v1bOIdhm94dVjQCPFElW9W+3GeDOSHmy2EBYe4VTApuzolPcmgFTN3ftVJRKR2J9Q==",
+ "requires": {
+ "media-typer": "0.3.0",
+ "mime-types": "~2.1.18"
+ }
+ },
+ "uglify-js": {
+ "version": "3.5.9",
+ "resolved": "https://registry.npmjs.org/uglify-js/-/uglify-js-3.5.9.tgz",
+ "integrity": "sha512-WpT0RqsDtAWPNJK955DEnb6xjymR8Fn0OlK4TT4pS0ASYsVPqr5ELhgwOwLCP5J5vHeJ4xmMmz3DEgdqC10JeQ==",
+ "optional": true,
+ "requires": {
+ "commander": "~2.20.0",
+ "source-map": "~0.6.1"
+ }
+ },
+ "unpipe": {
+ "version": "1.0.0",
+ "resolved": "https://registry.npmjs.org/unpipe/-/unpipe-1.0.0.tgz",
+ "integrity": "sha1-sr9O6FFKrmFltIF4KdIbLvSZBOw="
+ },
+ "utils-merge": {
+ "version": "1.0.1",
+ "resolved": "https://registry.npmjs.org/utils-merge/-/utils-merge-1.0.1.tgz",
+ "integrity": "sha1-n5VxD1CiZ5R7LMwSR0HBAoQn5xM="
+ },
+ "vary": {
+ "version": "1.1.2",
+ "resolved": "https://registry.npmjs.org/vary/-/vary-1.1.2.tgz",
+ "integrity": "sha1-IpnwLG3tMNSllhsLn3RSShj2NPw="
+ },
+ "walk": {
+ "version": "2.3.9",
+ "resolved": "https://registry.npmjs.org/walk/-/walk-2.3.9.tgz",
+ "integrity": "sha1-MbTbZnjyrgHDnqn7hyWpAx5Vins=",
+ "requires": {
+ "foreachasync": "^3.0.0"
+ }
+ },
+ "wordwrap": {
+ "version": "0.0.3",
+ "resolved": "https://registry.npmjs.org/wordwrap/-/wordwrap-0.0.3.tgz",
+ "integrity": "sha1-o9XabNXAvAAI03I0u68b7WMFkQc="
+ }
+ }
+}
diff --git a/images/benchmarks/node/package.json b/images/benchmarks/node/package.json
new file mode 100644
index 000000000..7dcadd523
--- /dev/null
+++ b/images/benchmarks/node/package.json
@@ -0,0 +1,19 @@
+{
+ "name": "nodedum",
+ "version": "1.0.0",
+ "description": "",
+ "main": "index.js",
+ "scripts": {
+ "test": "echo \"Error: no test specified\" && exit 1"
+ },
+ "author": "",
+ "license": "ISC",
+ "dependencies": {
+ "express": "^4.16.4",
+ "hbs": "^4.0.4",
+ "redis": "^2.8.0",
+ "redis-commands": "^1.2.0",
+ "redis-parser": "^2.6.0",
+ "secure-random-string": "^1.1.0"
+ }
+}
diff --git a/images/benchmarks/runsc/Dockerfile b/images/benchmarks/runsc/Dockerfile
new file mode 100644
index 000000000..6c3aafa57
--- /dev/null
+++ b/images/benchmarks/runsc/Dockerfile
@@ -0,0 +1,24 @@
+FROM ubuntu:18.04
+
+RUN set -x \
+ && apt-get update \
+ && apt-get install -y \
+ wget \
+ git \
+ pkg-config \
+ zip \
+ g++ \
+ zlib1g-dev \
+ unzip \
+ python-minimal \
+ python3 \
+ python3-pip \
+ && rm -rf /var/lib/apt/lists/*
+RUN wget https://github.com/bazelbuild/bazel/releases/download/3.4.1/bazel-3.4.1-installer-linux-x86_64.sh
+RUN chmod +x bazel-3.4.1-installer-linux-x86_64.sh
+RUN ./bazel-3.4.1-installer-linux-x86_64.sh
+
+# Download release-20200601.0
+RUN mkdir gvisor && cd gvisor \
+ && git init && git remote add origin https://github.com/google/gvisor.git \
+ && git fetch --depth 1 origin a9b47390c821942d60784e308f681f213645049c && git checkout FETCH_HEAD
diff --git a/pkg/abi/linux/fuse.go b/pkg/abi/linux/fuse.go
index d3ebbccc4..5c6ffe4a3 100644
--- a/pkg/abi/linux/fuse.go
+++ b/pkg/abi/linux/fuse.go
@@ -141,3 +141,108 @@ type FUSEWriteIn struct {
_ uint32
}
+
+// FUSE_INIT flags, consistent with the ones in include/uapi/linux/fuse.h.
+const (
+ FUSE_ASYNC_READ = 1 << 0
+ FUSE_POSIX_LOCKS = 1 << 1
+ FUSE_FILE_OPS = 1 << 2
+ FUSE_ATOMIC_O_TRUNC = 1 << 3
+ FUSE_EXPORT_SUPPORT = 1 << 4
+ FUSE_BIG_WRITES = 1 << 5
+ FUSE_DONT_MASK = 1 << 6
+ FUSE_SPLICE_WRITE = 1 << 7
+ FUSE_SPLICE_MOVE = 1 << 8
+ FUSE_SPLICE_READ = 1 << 9
+ FUSE_FLOCK_LOCKS = 1 << 10
+ FUSE_HAS_IOCTL_DIR = 1 << 11
+ FUSE_AUTO_INVAL_DATA = 1 << 12
+ FUSE_DO_READDIRPLUS = 1 << 13
+ FUSE_READDIRPLUS_AUTO = 1 << 14
+ FUSE_ASYNC_DIO = 1 << 15
+ FUSE_WRITEBACK_CACHE = 1 << 16
+ FUSE_NO_OPEN_SUPPORT = 1 << 17
+ FUSE_PARALLEL_DIROPS = 1 << 18
+ FUSE_HANDLE_KILLPRIV = 1 << 19
+ FUSE_POSIX_ACL = 1 << 20
+ FUSE_ABORT_ERROR = 1 << 21
+ FUSE_MAX_PAGES = 1 << 22
+ FUSE_CACHE_SYMLINKS = 1 << 23
+ FUSE_NO_OPENDIR_SUPPORT = 1 << 24
+ FUSE_EXPLICIT_INVAL_DATA = 1 << 25
+ FUSE_MAP_ALIGNMENT = 1 << 26
+)
+
+// currently supported FUSE protocol version numbers.
+const (
+ FUSE_KERNEL_VERSION = 7
+ FUSE_KERNEL_MINOR_VERSION = 31
+)
+
+// FUSEInitIn is the request sent by the kernel to the daemon,
+// to negotiate the version and flags.
+//
+// +marshal
+type FUSEInitIn struct {
+ // Major version supported by kernel.
+ Major uint32
+
+ // Minor version supported by the kernel.
+ Minor uint32
+
+ // MaxReadahead is the maximum number of bytes to read-ahead
+ // decided by the kernel.
+ MaxReadahead uint32
+
+ // Flags of this init request.
+ Flags uint32
+}
+
+// FUSEInitOut is the reply sent by the daemon to the kernel
+// for FUSEInitIn.
+//
+// +marshal
+type FUSEInitOut struct {
+ // Major version supported by daemon.
+ Major uint32
+
+ // Minor version supported by daemon.
+ Minor uint32
+
+ // MaxReadahead is the maximum number of bytes to read-ahead.
+ // Decided by the daemon, after receiving the value from kernel.
+ MaxReadahead uint32
+
+ // Flags of this init reply.
+ Flags uint32
+
+ // MaxBackground is the maximum number of pending background requests
+ // that the daemon wants.
+ MaxBackground uint16
+
+ // CongestionThreshold is the daemon-decided threshold for
+ // the number of the pending background requests.
+ CongestionThreshold uint16
+
+ // MaxWrite is the daemon's maximum size of a write buffer.
+ // Kernel adjusts it to the minimum (fuse/init.go:fuseMinMaxWrite).
+ // if the value from daemon is too small.
+ MaxWrite uint32
+
+ // TimeGran is the daemon's time granularity for mtime and ctime metadata.
+ // The unit is nanosecond.
+ // Value should be power of 10.
+ // 1 indicates full nanosecond granularity support.
+ TimeGran uint32
+
+ // MaxPages is the daemon's maximum number of pages for one write operation.
+ // Kernel adjusts it to the maximum (fuse/init.go:FUSE_MAX_MAX_PAGES).
+ // if the value from daemon is too large.
+ MaxPages uint16
+
+ // MapAlignment is an unknown field and not used by this package at this moment.
+ // Use as a placeholder to be consistent with the FUSE protocol.
+ MapAlignment uint16
+
+ _ [8]uint32
+}
diff --git a/pkg/abi/linux/netfilter.go b/pkg/abi/linux/netfilter.go
index a91f9f018..9c27f7bb2 100644
--- a/pkg/abi/linux/netfilter.go
+++ b/pkg/abi/linux/netfilter.go
@@ -59,7 +59,7 @@ var VerdictStrings = map[int32]string{
NF_RETURN: "RETURN",
}
-// Socket options. These correspond to values in
+// Socket options for SOL_SOCKET. These correspond to values in
// include/uapi/linux/netfilter_ipv4/ip_tables.h.
const (
IPT_BASE_CTL = 64
@@ -74,6 +74,12 @@ const (
IPT_SO_GET_MAX = IPT_SO_GET_REVISION_TARGET
)
+// Socket option for SOL_IP. This corresponds to the value in
+// include/uapi/linux/netfilter_ipv4.h.
+const (
+ SO_ORIGINAL_DST = 80
+)
+
// Name lengths. These correspond to values in
// include/uapi/linux/netfilter/x_tables.h.
const (
diff --git a/pkg/abi/linux/socket.go b/pkg/abi/linux/socket.go
index c24a8216e..d6946bb82 100644
--- a/pkg/abi/linux/socket.go
+++ b/pkg/abi/linux/socket.go
@@ -239,11 +239,13 @@ const SockAddrMax = 128
type InetAddr [4]byte
// SockAddrInet is struct sockaddr_in, from uapi/linux/in.h.
+//
+// +marshal
type SockAddrInet struct {
Family uint16
Port uint16
Addr InetAddr
- Zero [8]uint8 // pad to sizeof(struct sockaddr).
+ _ [8]uint8 // pad to sizeof(struct sockaddr).
}
// InetMulticastRequest is struct ip_mreq, from uapi/linux/in.h.
diff --git a/pkg/sentry/fdimport/fdimport.go b/pkg/sentry/fdimport/fdimport.go
index a4199f9e9..b8686adb4 100644
--- a/pkg/sentry/fdimport/fdimport.go
+++ b/pkg/sentry/fdimport/fdimport.go
@@ -15,6 +15,8 @@
package fdimport
import (
+ "fmt"
+
"gvisor.dev/gvisor/pkg/context"
"gvisor.dev/gvisor/pkg/sentry/fs"
"gvisor.dev/gvisor/pkg/sentry/fs/host"
@@ -84,6 +86,9 @@ func importFS(ctx context.Context, fdTable *kernel.FDTable, console bool, fds []
func importVFS2(ctx context.Context, fdTable *kernel.FDTable, console bool, stdioFDs []int) (*hostvfs2.TTYFileDescription, error) {
k := kernel.KernelFromContext(ctx)
+ if k == nil {
+ return nil, fmt.Errorf("cannot find kernel from context")
+ }
var ttyFile *vfs.FileDescription
for appFD, hostFD := range stdioFDs {
diff --git a/pkg/sentry/fsimpl/fuse/BUILD b/pkg/sentry/fsimpl/fuse/BUILD
index 67649e811..999111deb 100644
--- a/pkg/sentry/fsimpl/fuse/BUILD
+++ b/pkg/sentry/fsimpl/fuse/BUILD
@@ -21,6 +21,7 @@ go_library(
"connection.go",
"dev.go",
"fusefs.go",
+ "init.go",
"register.go",
"request_list.go",
],
@@ -44,14 +45,13 @@ go_library(
)
go_test(
- name = "dev_test",
+ name = "fuse_test",
size = "small",
srcs = ["dev_test.go"],
library = ":fuse",
deps = [
"//pkg/abi/linux",
"//pkg/sentry/fsimpl/testutil",
- "//pkg/sentry/fsimpl/tmpfs",
"//pkg/sentry/kernel",
"//pkg/sentry/kernel/auth",
"//pkg/sentry/vfs",
diff --git a/pkg/sentry/fsimpl/fuse/connection.go b/pkg/sentry/fsimpl/fuse/connection.go
index f330da0bd..6df2728ab 100644
--- a/pkg/sentry/fsimpl/fuse/connection.go
+++ b/pkg/sentry/fsimpl/fuse/connection.go
@@ -17,6 +17,8 @@ package fuse
import (
"errors"
"fmt"
+ "sync"
+ "sync/atomic"
"syscall"
"gvisor.dev/gvisor/pkg/abi/linux"
@@ -25,18 +27,29 @@ import (
"gvisor.dev/gvisor/pkg/sentry/kernel"
"gvisor.dev/gvisor/pkg/sentry/kernel/auth"
"gvisor.dev/gvisor/pkg/sentry/vfs"
+ "gvisor.dev/gvisor/pkg/syserror"
"gvisor.dev/gvisor/pkg/waiter"
"gvisor.dev/gvisor/tools/go_marshal/marshal"
)
-// MaxActiveRequestsDefault is the default setting controlling the upper bound
+// maxActiveRequestsDefault is the default setting controlling the upper bound
// on the number of active requests at any given time.
-const MaxActiveRequestsDefault = 10000
+const maxActiveRequestsDefault = 10000
-var (
- // Ordinary requests have even IDs, while interrupts IDs are odd.
- InitReqBit uint64 = 1
- ReqIDStep uint64 = 2
+// Ordinary requests have even IDs, while interrupts IDs are odd.
+// Used to increment the unique ID for each FUSE request.
+var reqIDStep uint64 = 2
+
+const (
+ // fuseDefaultMaxBackground is the default value for MaxBackground.
+ fuseDefaultMaxBackground = 12
+
+ // fuseDefaultCongestionThreshold is the default value for CongestionThreshold,
+ // and is 75% of the default maximum of MaxGround.
+ fuseDefaultCongestionThreshold = (fuseDefaultMaxBackground * 3 / 4)
+
+ // fuseDefaultMaxPagesPerReq is the default value for MaxPagesPerReq.
+ fuseDefaultMaxPagesPerReq = 32
)
// Request represents a FUSE operation request that hasn't been sent to the
@@ -61,17 +74,125 @@ type Response struct {
data []byte
}
-// Connection is the struct by which the sentry communicates with the FUSE server daemon.
-type Connection struct {
+// connection is the struct by which the sentry communicates with the FUSE server daemon.
+type connection struct {
fd *DeviceFD
- // MaxWrite is the daemon's maximum size of a write buffer.
- // This is negotiated during FUSE_INIT.
- MaxWrite uint32
+ // The following FUSE_INIT flags are currently unsupported by this implementation:
+ // - FUSE_ATOMIC_O_TRUNC: requires open(..., O_TRUNC)
+ // - FUSE_EXPORT_SUPPORT
+ // - FUSE_HANDLE_KILLPRIV
+ // - FUSE_POSIX_LOCKS: requires POSIX locks
+ // - FUSE_FLOCK_LOCKS: requires POSIX locks
+ // - FUSE_AUTO_INVAL_DATA: requires page caching eviction
+ // - FUSE_EXPLICIT_INVAL_DATA: requires page caching eviction
+ // - FUSE_DO_READDIRPLUS/FUSE_READDIRPLUS_AUTO: requires FUSE_READDIRPLUS implementation
+ // - FUSE_ASYNC_DIO
+ // - FUSE_POSIX_ACL: affects defaultPermissions, posixACL, xattr handler
+
+ // initialized after receiving FUSE_INIT reply.
+ // Until it's set, suspend sending FUSE requests.
+ // Use SetInitialized() and IsInitialized() for atomic access.
+ initialized int32
+
+ // initializedChan is used to block requests before initialization.
+ initializedChan chan struct{}
+
+ // blocked when there are too many outstading backgrounds requests (NumBackground == MaxBackground).
+ // TODO(gvisor.dev/issue/3185): update the numBackground accordingly; use a channel to block.
+ blocked bool
+
+ // connected (connection established) when a new FUSE file system is created.
+ // Set to false when:
+ // umount,
+ // connection abort,
+ // device release.
+ connected bool
+
+ // aborted via sysfs.
+ // TODO(gvisor.dev/issue/3185): abort all queued requests.
+ aborted bool
+
+ // connInitError if FUSE_INIT encountered error (major version mismatch).
+ // Only set in INIT.
+ connInitError bool
+
+ // connInitSuccess if FUSE_INIT is successful.
+ // Only set in INIT.
+ // Used for destory.
+ connInitSuccess bool
+
+ // TODO(gvisor.dev/issue/3185): All the queue logic are working in progress.
+
+ // NumberBackground is the number of requests in the background.
+ numBackground uint16
+
+ // congestionThreshold for NumBackground.
+ // Negotiated in FUSE_INIT.
+ congestionThreshold uint16
+
+ // maxBackground is the maximum number of NumBackground.
+ // Block connection when it is reached.
+ // Negotiated in FUSE_INIT.
+ maxBackground uint16
+
+ // numActiveBackground is the number of requests in background and has being marked as active.
+ numActiveBackground uint16
+
+ // numWating is the number of requests waiting for completion.
+ numWaiting uint32
+
+ // TODO(gvisor.dev/issue/3185): BgQueue
+ // some queue for background queued requests.
+
+ // bgLock protects:
+ // MaxBackground, CongestionThreshold, NumBackground,
+ // NumActiveBackground, BgQueue, Blocked.
+ bgLock sync.Mutex
+
+ // maxRead is the maximum size of a read buffer in in bytes.
+ maxRead uint32
+
+ // maxWrite is the maximum size of a write buffer in bytes.
+ // Negotiated in FUSE_INIT.
+ maxWrite uint32
+
+ // maxPages is the maximum number of pages for a single request to use.
+ // Negotiated in FUSE_INIT.
+ maxPages uint16
+
+ // minor version of the FUSE protocol.
+ // Negotiated and only set in INIT.
+ minor uint32
+
+ // asyncRead if read pages asynchronously.
+ // Negotiated and only set in INIT.
+ asyncRead bool
+
+ // abortErr is true if kernel need to return an unique read error after abort.
+ // Negotiated and only set in INIT.
+ abortErr bool
+
+ // writebackCache is true for write-back cache policy,
+ // false for write-through policy.
+ // Negotiated and only set in INIT.
+ writebackCache bool
+
+ // cacheSymlinks if filesystem needs to cache READLINK responses in page cache.
+ // Negotiated and only set in INIT.
+ cacheSymlinks bool
+
+ // bigWrites if doing multi-page cached writes.
+ // Negotiated and only set in INIT.
+ bigWrites bool
+
+ // dontMask if filestestem does not apply umask to creation modes.
+ // Negotiated in INIT.
+ dontMask bool
}
-// NewFUSEConnection creates a FUSE connection to fd
-func NewFUSEConnection(_ context.Context, fd *vfs.FileDescription, maxInFlightRequests uint64) (*Connection, error) {
+// newFUSEConnection creates a FUSE connection to fd.
+func newFUSEConnection(_ context.Context, fd *vfs.FileDescription, maxInFlightRequests uint64) (*connection, error) {
// Mark the device as ready so it can be used. /dev/fuse can only be used if the FD was used to
// mount a FUSE filesystem.
fuseFD := fd.Impl().(*DeviceFD)
@@ -84,16 +205,41 @@ func NewFUSEConnection(_ context.Context, fd *vfs.FileDescription, maxInFlightRe
fuseFD.fullQueueCh = make(chan struct{}, maxInFlightRequests)
fuseFD.writeCursor = 0
- return &Connection{
- fd: fuseFD,
+ return &connection{
+ fd: fuseFD,
+ maxBackground: fuseDefaultMaxBackground,
+ congestionThreshold: fuseDefaultCongestionThreshold,
+ maxPages: fuseDefaultMaxPagesPerReq,
+ initializedChan: make(chan struct{}),
+ connected: true,
}, nil
}
+// SetInitialized atomically sets the connection as initialized.
+func (conn *connection) SetInitialized() {
+ // Unblock the requests sent before INIT.
+ close(conn.initializedChan)
+
+ // Close the channel first to avoid the non-atomic situation
+ // where conn.initialized is true but there are
+ // tasks being blocked on the channel.
+ // And it prevents the newer tasks from gaining
+ // unnecessary higher chance to be issued before the blocked one.
+
+ atomic.StoreInt32(&(conn.initialized), int32(1))
+}
+
+// IsInitialized atomically check if the connection is initialized.
+// pairs with SetInitialized().
+func (conn *connection) Initialized() bool {
+ return atomic.LoadInt32(&(conn.initialized)) != 0
+}
+
// NewRequest creates a new request that can be sent to the FUSE server.
-func (conn *Connection) NewRequest(creds *auth.Credentials, pid uint32, ino uint64, opcode linux.FUSEOpcode, payload marshal.Marshallable) (*Request, error) {
+func (conn *connection) NewRequest(creds *auth.Credentials, pid uint32, ino uint64, opcode linux.FUSEOpcode, payload marshal.Marshallable) (*Request, error) {
conn.fd.mu.Lock()
defer conn.fd.mu.Unlock()
- conn.fd.nextOpID += linux.FUSEOpID(ReqIDStep)
+ conn.fd.nextOpID += linux.FUSEOpID(reqIDStep)
hdrLen := (*linux.FUSEHeaderIn)(nil).SizeBytes()
hdr := linux.FUSEHeaderIn{
@@ -118,13 +264,49 @@ func (conn *Connection) NewRequest(creds *auth.Credentials, pid uint32, ino uint
}
// Call makes a request to the server and blocks the invoking task until a
-// server responds with a response.
-// NOTE: If no task is provided then the Call will simply enqueue the request
-// and return a nil response. No blocking will happen in this case. Instead,
-// this is used to signify that the processing of this request will happen by
-// the kernel.Task that writes the response. See FUSE_INIT for such an
-// invocation.
-func (conn *Connection) Call(t *kernel.Task, r *Request) (*Response, error) {
+// server responds with a response. Task should never be nil.
+// Requests will not be sent before the connection is initialized.
+// For async tasks, use CallAsync().
+func (conn *connection) Call(t *kernel.Task, r *Request) (*Response, error) {
+ // Block requests sent before connection is initalized.
+ if !conn.Initialized() {
+ if err := t.Block(conn.initializedChan); err != nil {
+ return nil, err
+ }
+ }
+
+ return conn.call(t, r)
+}
+
+// CallAsync makes an async (aka background) request.
+// Those requests either do not expect a response (e.g. release) or
+// the response should be handled by others (e.g. init).
+// Return immediately unless the connection is blocked (before initialization).
+// Async call example: init, release, forget, aio, interrupt.
+// When the Request is FUSE_INIT, it will not be blocked before initialization.
+func (conn *connection) CallAsync(t *kernel.Task, r *Request) error {
+ // Block requests sent before connection is initalized.
+ if !conn.Initialized() && r.hdr.Opcode != linux.FUSE_INIT {
+ if err := t.Block(conn.initializedChan); err != nil {
+ return err
+ }
+ }
+
+ // This should be the only place that invokes call() with a nil task.
+ _, err := conn.call(nil, r)
+ return err
+}
+
+// call makes a call without blocking checks.
+func (conn *connection) call(t *kernel.Task, r *Request) (*Response, error) {
+ if !conn.connected {
+ return nil, syserror.ENOTCONN
+ }
+
+ if conn.connInitError {
+ return nil, syserror.ECONNREFUSED
+ }
+
fut, err := conn.callFuture(t, r)
if err != nil {
return nil, err
@@ -160,7 +342,7 @@ func (r *Response) UnmarshalPayload(m marshal.Marshallable) error {
// callFuture makes a request to the server and returns a future response.
// Call resolve() when the response needs to be fulfilled.
-func (conn *Connection) callFuture(t *kernel.Task, r *Request) (*futureResponse, error) {
+func (conn *connection) callFuture(t *kernel.Task, r *Request) (*futureResponse, error) {
conn.fd.mu.Lock()
defer conn.fd.mu.Unlock()
@@ -195,7 +377,7 @@ func (conn *Connection) callFuture(t *kernel.Task, r *Request) (*futureResponse,
}
// callFutureLocked makes a request to the server and returns a future response.
-func (conn *Connection) callFutureLocked(t *kernel.Task, r *Request) (*futureResponse, error) {
+func (conn *connection) callFutureLocked(t *kernel.Task, r *Request) (*futureResponse, error) {
conn.fd.queue.PushBack(r)
conn.fd.numActiveRequests += 1
fut := newFutureResponse(r.hdr.Opcode)
diff --git a/pkg/sentry/fsimpl/fuse/dev.go b/pkg/sentry/fsimpl/fuse/dev.go
index f3443ac71..2225076bc 100644
--- a/pkg/sentry/fsimpl/fuse/dev.go
+++ b/pkg/sentry/fsimpl/fuse/dev.go
@@ -21,6 +21,7 @@ import (
"gvisor.dev/gvisor/pkg/context"
"gvisor.dev/gvisor/pkg/log"
"gvisor.dev/gvisor/pkg/sentry/kernel"
+ "gvisor.dev/gvisor/pkg/sentry/kernel/auth"
"gvisor.dev/gvisor/pkg/sentry/vfs"
"gvisor.dev/gvisor/pkg/sync"
"gvisor.dev/gvisor/pkg/syserror"
@@ -98,7 +99,9 @@ type DeviceFD struct {
}
// Release implements vfs.FileDescriptionImpl.Release.
-func (fd *DeviceFD) Release() {}
+func (fd *DeviceFD) Release() {
+ fd.fs.conn.connected = false
+}
// PRead implements vfs.FileDescriptionImpl.PRead.
func (fd *DeviceFD) PRead(ctx context.Context, dst usermem.IOSequence, offset int64, opts vfs.ReadOptions) (int64, error) {
@@ -124,7 +127,7 @@ func (fd *DeviceFD) Read(ctx context.Context, dst usermem.IOSequence, opts vfs.R
minBuffSize := linux.FUSE_MIN_READ_BUFFER
inHdrLen := uint32((*linux.FUSEHeaderIn)(nil).SizeBytes())
writeHdrLen := uint32((*linux.FUSEWriteIn)(nil).SizeBytes())
- negotiatedMinBuffSize := inHdrLen + writeHdrLen + fd.fs.conn.MaxWrite
+ negotiatedMinBuffSize := inHdrLen + writeHdrLen + fd.fs.conn.maxWrite
if minBuffSize < negotiatedMinBuffSize {
minBuffSize = negotiatedMinBuffSize
}
@@ -385,9 +388,9 @@ func (fd *DeviceFD) sendError(ctx context.Context, errno int32, req *Request) er
// FUSE_INIT.
func (fd *DeviceFD) noReceiverAction(ctx context.Context, r *Response) error {
if r.opcode == linux.FUSE_INIT {
- // TODO: process init response here.
- // Maybe get the creds from the context?
- // creds := auth.CredentialsFromContext(ctx)
+ creds := auth.CredentialsFromContext(ctx)
+ rootUserNs := kernel.KernelFromContext(ctx).RootUserNamespace()
+ return fd.fs.conn.InitRecv(r, creds.HasCapabilityIn(linux.CAP_SYS_ADMIN, rootUserNs))
}
return nil
diff --git a/pkg/sentry/fsimpl/fuse/dev_test.go b/pkg/sentry/fsimpl/fuse/dev_test.go
index fcd77832a..84c222ad6 100644
--- a/pkg/sentry/fsimpl/fuse/dev_test.go
+++ b/pkg/sentry/fsimpl/fuse/dev_test.go
@@ -22,7 +22,6 @@ import (
"gvisor.dev/gvisor/pkg/abi/linux"
"gvisor.dev/gvisor/pkg/sentry/fsimpl/testutil"
- "gvisor.dev/gvisor/pkg/sentry/fsimpl/tmpfs"
"gvisor.dev/gvisor/pkg/sentry/kernel"
"gvisor.dev/gvisor/pkg/sentry/kernel/auth"
"gvisor.dev/gvisor/pkg/sentry/vfs"
@@ -60,25 +59,25 @@ func TestFUSECommunication(t *testing.T) {
Name: "SingleClientSingleServer",
NumClients: 1,
NumServers: 1,
- MaxActiveRequests: MaxActiveRequestsDefault,
+ MaxActiveRequests: maxActiveRequestsDefault,
},
{
Name: "SingleClientMultipleServers",
NumClients: 1,
NumServers: 10,
- MaxActiveRequests: MaxActiveRequestsDefault,
+ MaxActiveRequests: maxActiveRequestsDefault,
},
{
Name: "MultipleClientsSingleServer",
NumClients: 10,
NumServers: 1,
- MaxActiveRequests: MaxActiveRequestsDefault,
+ MaxActiveRequests: maxActiveRequestsDefault,
},
{
Name: "MultipleClientsMultipleServers",
NumClients: 10,
NumServers: 10,
- MaxActiveRequests: MaxActiveRequestsDefault,
+ MaxActiveRequests: maxActiveRequestsDefault,
},
{
Name: "RequestCapacityFull",
@@ -145,7 +144,7 @@ func TestFUSECommunication(t *testing.T) {
// CallTest makes a request to the server and blocks the invoking
// goroutine until a server responds with a response. Doesn't block
// a kernel.Task. Analogous to Connection.Call but used for testing.
-func CallTest(conn *Connection, t *kernel.Task, r *Request, i uint32) (*Response, error) {
+func CallTest(conn *connection, t *kernel.Task, r *Request, i uint32) (*Response, error) {
conn.fd.mu.Lock()
// Wait until we're certain that a new request can be processed.
@@ -214,7 +213,7 @@ func ReadTest(serverTask *kernel.Task, fd *vfs.FileDescription, inIOseq usermem.
// fuseClientRun emulates all the actions of a normal FUSE request. It creates
// a header, a payload, calls the server, waits for the response, and processes
// the response.
-func fuseClientRun(t *testing.T, s *testutil.System, k *kernel.Kernel, conn *Connection, creds *auth.Credentials, pid uint32, inode uint64, clientDone chan struct{}) {
+func fuseClientRun(t *testing.T, s *testutil.System, k *kernel.Kernel, conn *connection, creds *auth.Credentials, pid uint32, inode uint64, clientDone chan struct{}) {
defer func() { clientDone <- struct{}{} }()
tc := k.NewThreadGroup(nil, k.RootPIDNamespace(), kernel.NewSignalHandlers(), linux.SIGCHLD, k.GlobalInit().Limits())
@@ -343,7 +342,7 @@ func setup(t *testing.T) *testutil.System {
AllowUserMount: true,
})
- mntns, err := k.VFS().NewMountNamespace(ctx, creds, "", tmpfs.Name, &vfs.GetFilesystemOptions{})
+ mntns, err := k.VFS().NewMountNamespace(ctx, creds, "", "tmpfs", &vfs.GetFilesystemOptions{})
if err != nil {
t.Fatalf("NewMountNamespace(): %v", err)
}
@@ -353,7 +352,7 @@ func setup(t *testing.T) *testutil.System {
// newTestConnection creates a fuse connection that the sentry can communicate with
// and the FD for the server to communicate with.
-func newTestConnection(system *testutil.System, k *kernel.Kernel, maxActiveRequests uint64) (*Connection, *vfs.FileDescription, error) {
+func newTestConnection(system *testutil.System, k *kernel.Kernel, maxActiveRequests uint64) (*connection, *vfs.FileDescription, error) {
vfsObj := &vfs.VirtualFilesystem{}
fuseDev := &DeviceFD{}
diff --git a/pkg/sentry/fsimpl/fuse/fusefs.go b/pkg/sentry/fsimpl/fuse/fusefs.go
index 911b6f7cb..200a93bbf 100644
--- a/pkg/sentry/fsimpl/fuse/fusefs.go
+++ b/pkg/sentry/fsimpl/fuse/fusefs.go
@@ -65,7 +65,7 @@ type filesystem struct {
// conn is used for communication between the FUSE server
// daemon and the sentry fusefs.
- conn *Connection
+ conn *connection
// opts is the options the fusefs is initialized with.
opts *filesystemOptions
@@ -140,7 +140,7 @@ func (fsType FilesystemType) GetFilesystem(ctx context.Context, vfsObj *vfs.Virt
fsopts.rootMode = rootMode
// Set the maxInFlightRequests option.
- fsopts.maxActiveRequests = MaxActiveRequestsDefault
+ fsopts.maxActiveRequests = maxActiveRequestsDefault
// Check for unparsed options.
if len(mopts) != 0 {
@@ -157,8 +157,12 @@ func (fsType FilesystemType) GetFilesystem(ctx context.Context, vfsObj *vfs.Virt
fs.VFSFilesystem().Init(vfsObj, &fsType, fs)
- // TODO: dispatch a FUSE_INIT request to the FUSE daemon server before
- // returning. Mount will not block on this dispatched request.
+ // Send a FUSE_INIT request to the FUSE daemon server before returning.
+ // This call is not blocking.
+ if err := fs.conn.InitSend(creds, uint32(kernelTask.ThreadID())); err != nil {
+ log.Warningf("%s.InitSend: failed with error: %v", fsType.Name(), err)
+ return nil, nil, err
+ }
// root is the fusefs root directory.
root := fs.newInode(creds, fsopts.rootMode)
@@ -173,7 +177,7 @@ func NewFUSEFilesystem(ctx context.Context, devMinor uint32, opts *filesystemOpt
opts: opts,
}
- conn, err := NewFUSEConnection(ctx, device, opts.maxActiveRequests)
+ conn, err := newFUSEConnection(ctx, device, opts.maxActiveRequests)
if err != nil {
log.Warningf("fuse.NewFUSEFilesystem: NewFUSEConnection failed with error: %v", err)
return nil, syserror.EINVAL
@@ -192,8 +196,8 @@ func (fs *filesystem) Release() {
fs.Filesystem.Release()
}
-// Inode implements kernfs.Inode.
-type Inode struct {
+// inode implements kernfs.Inode.
+type inode struct {
kernfs.InodeAttrs
kernfs.InodeNoDynamicLookup
kernfs.InodeNotSymlink
@@ -206,7 +210,7 @@ type Inode struct {
}
func (fs *filesystem) newInode(creds *auth.Credentials, mode linux.FileMode) *kernfs.Dentry {
- i := &Inode{}
+ i := &inode{}
i.InodeAttrs.Init(creds, linux.UNNAMED_MAJOR, fs.devMinor, fs.NextIno(), linux.ModeDirectory|0755)
i.OrderedChildren.Init(kernfs.OrderedChildrenOptions{})
i.dentry.Init(i)
@@ -215,7 +219,7 @@ func (fs *filesystem) newInode(creds *auth.Credentials, mode linux.FileMode) *ke
}
// Open implements kernfs.Inode.Open.
-func (i *Inode) Open(ctx context.Context, rp *vfs.ResolvingPath, vfsd *vfs.Dentry, opts vfs.OpenOptions) (*vfs.FileDescription, error) {
+func (i *inode) Open(ctx context.Context, rp *vfs.ResolvingPath, vfsd *vfs.Dentry, opts vfs.OpenOptions) (*vfs.FileDescription, error) {
fd, err := kernfs.NewGenericDirectoryFD(rp.Mount(), vfsd, &i.OrderedChildren, &i.locks, &opts)
if err != nil {
return nil, err
diff --git a/pkg/sentry/fsimpl/fuse/init.go b/pkg/sentry/fsimpl/fuse/init.go
new file mode 100644
index 000000000..779c2bd3f
--- /dev/null
+++ b/pkg/sentry/fsimpl/fuse/init.go
@@ -0,0 +1,166 @@
+// Copyright 2020 The gVisor Authors.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+// http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package fuse
+
+import (
+ "gvisor.dev/gvisor/pkg/abi/linux"
+ "gvisor.dev/gvisor/pkg/sentry/kernel/auth"
+)
+
+// consts used by FUSE_INIT negotiation.
+const (
+ // fuseMaxMaxPages is the maximum value for MaxPages received in InitOut.
+ // Follow the same behavior as unix fuse implementation.
+ fuseMaxMaxPages = 256
+
+ // Maximum value for the time granularity for file time stamps, 1s.
+ // Follow the same behavior as unix fuse implementation.
+ fuseMaxTimeGranNs = 1000000000
+
+ // Minimum value for MaxWrite.
+ // Follow the same behavior as unix fuse implementation.
+ fuseMinMaxWrite = 4096
+
+ // Temporary default value for max readahead, 128kb.
+ fuseDefaultMaxReadahead = 131072
+
+ // The FUSE_INIT_IN flags sent to the daemon.
+ // TODO(gvisor.dev/issue/3199): complete the flags.
+ fuseDefaultInitFlags = linux.FUSE_MAX_PAGES
+)
+
+// Adjustable maximums for Connection's cogestion control parameters.
+// Used as the upperbound of the config values.
+// Currently we do not support adjustment to them.
+var (
+ MaxUserBackgroundRequest uint16 = fuseDefaultMaxBackground
+ MaxUserCongestionThreshold uint16 = fuseDefaultCongestionThreshold
+)
+
+// InitSend sends a FUSE_INIT request.
+func (conn *connection) InitSend(creds *auth.Credentials, pid uint32) error {
+ in := linux.FUSEInitIn{
+ Major: linux.FUSE_KERNEL_VERSION,
+ Minor: linux.FUSE_KERNEL_MINOR_VERSION,
+ // TODO(gvisor.dev/issue/3196): find appropriate way to calculate this
+ MaxReadahead: fuseDefaultMaxReadahead,
+ Flags: fuseDefaultInitFlags,
+ }
+
+ req, err := conn.NewRequest(creds, pid, 0, linux.FUSE_INIT, &in)
+ if err != nil {
+ return err
+ }
+
+ // Since there is no task to block on and FUSE_INIT is the request
+ // to unblock other requests, use nil.
+ return conn.CallAsync(nil, req)
+}
+
+// InitRecv receives a FUSE_INIT reply and process it.
+func (conn *connection) InitRecv(res *Response, hasSysAdminCap bool) error {
+ if err := res.Error(); err != nil {
+ return err
+ }
+
+ var out linux.FUSEInitOut
+ if err := res.UnmarshalPayload(&out); err != nil {
+ return err
+ }
+
+ return conn.initProcessReply(&out, hasSysAdminCap)
+}
+
+// Process the FUSE_INIT reply from the FUSE server.
+func (conn *connection) initProcessReply(out *linux.FUSEInitOut, hasSysAdminCap bool) error {
+ // No support for old major fuse versions.
+ if out.Major != linux.FUSE_KERNEL_VERSION {
+ conn.connInitError = true
+
+ // Set the connection as initialized and unblock the blocked requests
+ // (i.e. return error for them).
+ conn.SetInitialized()
+
+ return nil
+ }
+
+ // Start processing the reply.
+ conn.connInitSuccess = true
+ conn.minor = out.Minor
+
+ // No support for limits before minor version 13.
+ if out.Minor >= 13 {
+ conn.bgLock.Lock()
+
+ if out.MaxBackground > 0 {
+ conn.maxBackground = out.MaxBackground
+
+ if !hasSysAdminCap &&
+ conn.maxBackground > MaxUserBackgroundRequest {
+ conn.maxBackground = MaxUserBackgroundRequest
+ }
+ }
+
+ if out.CongestionThreshold > 0 {
+ conn.congestionThreshold = out.CongestionThreshold
+
+ if !hasSysAdminCap &&
+ conn.congestionThreshold > MaxUserCongestionThreshold {
+ conn.congestionThreshold = MaxUserCongestionThreshold
+ }
+ }
+
+ conn.bgLock.Unlock()
+ }
+
+ // No support for the following flags before minor version 6.
+ if out.Minor >= 6 {
+ conn.asyncRead = out.Flags&linux.FUSE_ASYNC_READ != 0
+ conn.bigWrites = out.Flags&linux.FUSE_BIG_WRITES != 0
+ conn.dontMask = out.Flags&linux.FUSE_DONT_MASK != 0
+ conn.writebackCache = out.Flags&linux.FUSE_WRITEBACK_CACHE != 0
+ conn.cacheSymlinks = out.Flags&linux.FUSE_CACHE_SYMLINKS != 0
+ conn.abortErr = out.Flags&linux.FUSE_ABORT_ERROR != 0
+
+ // TODO(gvisor.dev/issue/3195): figure out how to use TimeGran (0 < TimeGran <= fuseMaxTimeGranNs).
+
+ if out.Flags&linux.FUSE_MAX_PAGES != 0 {
+ maxPages := out.MaxPages
+ if maxPages < 1 {
+ maxPages = 1
+ }
+ if maxPages > fuseMaxMaxPages {
+ maxPages = fuseMaxMaxPages
+ }
+ conn.maxPages = maxPages
+ }
+ }
+
+ // No support for negotiating MaxWrite before minor version 5.
+ if out.Minor >= 5 {
+ conn.maxWrite = out.MaxWrite
+ } else {
+ conn.maxWrite = fuseMinMaxWrite
+ }
+ if conn.maxWrite < fuseMinMaxWrite {
+ conn.maxWrite = fuseMinMaxWrite
+ }
+
+ // Set connection as initialized and unblock the requests
+ // issued before init.
+ conn.SetInitialized()
+
+ return nil
+}
diff --git a/pkg/sentry/platform/kvm/kvm_const_arm64.go b/pkg/sentry/platform/kvm/kvm_const_arm64.go
index 6f0539c29..fdc599477 100644
--- a/pkg/sentry/platform/kvm/kvm_const_arm64.go
+++ b/pkg/sentry/platform/kvm/kvm_const_arm64.go
@@ -116,6 +116,17 @@ const (
// Arm64: Exception Syndrome Register EL1.
const (
+ _ESR_ELx_EC_SHIFT = 26
+ _ESR_ELx_EC_MASK = 0x3F << _ESR_ELx_EC_SHIFT
+
+ _ESR_ELx_EC_IMP_DEF = 0x1f
+ _ESR_ELx_EC_IABT_LOW = 0x20
+ _ESR_ELx_EC_IABT_CUR = 0x21
+ _ESR_ELx_EC_PC_ALIGN = 0x22
+
+ _ESR_ELx_CM = 1 << 8
+ _ESR_ELx_WNR = 1 << 6
+
_ESR_ELx_FSC = 0x3F
_ESR_SEGV_MAPERR_L0 = 0x4
diff --git a/pkg/sentry/platform/kvm/machine_arm64.go b/pkg/sentry/platform/kvm/machine_arm64.go
index f3bf973de..9db171af9 100644
--- a/pkg/sentry/platform/kvm/machine_arm64.go
+++ b/pkg/sentry/platform/kvm/machine_arm64.go
@@ -125,71 +125,59 @@ func nonCanonical(addr uint64, signal int32, info *arch.SignalInfo) (usermem.Acc
return usermem.NoAccess, platform.ErrContextSignal
}
+// isInstructionAbort returns true if it is an instruction abort.
+//
+//go:nosplit
+func isInstructionAbort(code uint64) bool {
+ value := (code & _ESR_ELx_EC_MASK) >> _ESR_ELx_EC_SHIFT
+ return value == _ESR_ELx_EC_IABT_LOW
+}
+
+// isWriteFault returns whether it is a write fault.
+//
+//go:nosplit
+func isWriteFault(code uint64) bool {
+ if isInstructionAbort(code) {
+ return false
+ }
+
+ return (code & _ESR_ELx_WNR) != 0
+}
+
// fault generates an appropriate fault return.
//
//go:nosplit
func (c *vCPU) fault(signal int32, info *arch.SignalInfo) (usermem.AccessType, error) {
+ bluepill(c) // Probably no-op, but may not be.
faultAddr := c.GetFaultAddr()
code, user := c.ErrorCode()
+ if !user {
+ // The last fault serviced by this CPU was not a user
+ // fault, so we can't reliably trust the faultAddr or
+ // the code provided here. We need to re-execute.
+ return usermem.NoAccess, platform.ErrContextInterrupt
+ }
+
// Reset the pointed SignalInfo.
*info = arch.SignalInfo{Signo: signal}
info.SetAddr(uint64(faultAddr))
- read := true
- write := false
- execute := true
-
ret := code & _ESR_ELx_FSC
switch ret {
case _ESR_SEGV_MAPERR_L0, _ESR_SEGV_MAPERR_L1, _ESR_SEGV_MAPERR_L2, _ESR_SEGV_MAPERR_L3:
info.Code = 1 //SEGV_MAPERR
- read = false
- write = true
- execute = false
case _ESR_SEGV_ACCERR_L1, _ESR_SEGV_ACCERR_L2, _ESR_SEGV_ACCERR_L3, _ESR_SEGV_PEMERR_L1, _ESR_SEGV_PEMERR_L2, _ESR_SEGV_PEMERR_L3:
info.Code = 2 // SEGV_ACCERR.
- read = true
- write = false
- execute = false
default:
info.Code = 2
}
- if !user {
- read = true
- write = false
- execute = true
-
- }
accessType := usermem.AccessType{
- Read: read,
- Write: write,
- Execute: execute,
+ Read: !isWriteFault(uint64(code)),
+ Write: isWriteFault(uint64(code)),
+ Execute: isInstructionAbort(uint64(code)),
}
return accessType, platform.ErrContextSignal
}
-
-// retryInGuest runs the given function in guest mode.
-//
-// If the function does not complete in guest mode (due to execution of a
-// system call due to a GC stall, for example), then it will be retried. The
-// given function must be idempotent as a result of the retry mechanism.
-func (m *machine) retryInGuest(fn func()) {
- c := m.Get()
- defer m.Put(c)
- for {
- c.ClearErrorCode() // See below.
- bluepill(c) // Force guest mode.
- fn() // Execute the given function.
- _, user := c.ErrorCode()
- if user {
- // If user is set, then we haven't bailed back to host
- // mode via a kernel exception or system call. We
- // consider the full function to have executed in guest
- // mode and we can return.
- break
- }
- }
-}
diff --git a/pkg/sentry/platform/kvm/testutil/testutil_arm64.go b/pkg/sentry/platform/kvm/testutil/testutil_arm64.go
index ca902c8c1..4dad877ba 100644
--- a/pkg/sentry/platform/kvm/testutil/testutil_arm64.go
+++ b/pkg/sentry/platform/kvm/testutil/testutil_arm64.go
@@ -56,5 +56,9 @@ func CheckTestRegs(regs *arch.Registers, full bool) (err error) {
err = addRegisterMismatch(err, fmt.Sprintf("R%d", i), regs.Regs[i], need)
}
}
+ // Check tls.
+ if need := ^uint64(11); regs.TPIDR_EL0 != need {
+ err = addRegisterMismatch(err, "tpdir_el0", regs.TPIDR_EL0, need)
+ }
return
}
diff --git a/pkg/sentry/platform/kvm/testutil/testutil_arm64.s b/pkg/sentry/platform/kvm/testutil/testutil_arm64.s
index 07658144e..6caf7282d 100644
--- a/pkg/sentry/platform/kvm/testutil/testutil_arm64.s
+++ b/pkg/sentry/platform/kvm/testutil/testutil_arm64.s
@@ -52,6 +52,8 @@ start:
TEXT ·FloatingPointWorks(SB),NOSPLIT,$0-8
NO_LOCAL_POINTERS
+ // gc will touch fpsimd, so we should test it.
+ // such as in <runtime.deductSweepCredit>.
FMOVD $(9.9), F0
MOVD $SYS_GETPID, R8 // getpid
SVC
@@ -102,11 +104,15 @@ isNaN:
TEXT ·TwiddleRegsSyscall(SB),NOSPLIT,$0
TWIDDLE_REGS()
+ MSR R10, TPIDR_EL0
+ // Trapped in el0_svc.
SVC
RET // never reached
TEXT ·TwiddleRegsFault(SB),NOSPLIT,$0
- TWIDDLE_REGS()
- // Branch to Register branches unconditionally to an address in <Rn>.
- JMP (R4) // <=> br x4, must fault
- RET // never reached
+ TWIDDLE_REGS()
+ MSR R10, TPIDR_EL0
+ // Trapped in el0_ia.
+ // Branch to Register branches unconditionally to an address in <Rn>.
+ JMP (R6) // <=> br x6, must fault
+ RET // never reached
diff --git a/pkg/sentry/socket/netstack/netstack.go b/pkg/sentry/socket/netstack/netstack.go
index f86e6cd7a..31a168f7e 100644
--- a/pkg/sentry/socket/netstack/netstack.go
+++ b/pkg/sentry/socket/netstack/netstack.go
@@ -1490,6 +1490,10 @@ func getSockOptIPv6(t *kernel.Task, ep commonEndpoint, name, outLen int) (marsha
vP := primitive.Int32(boolToInt32(v))
return &vP, nil
+ case linux.SO_ORIGINAL_DST:
+ // TODO(gvisor.dev/issue/170): ip6tables.
+ return nil, syserr.ErrInvalidArgument
+
default:
emitUnimplementedEventIPv6(t, name)
}
@@ -1600,6 +1604,19 @@ func getSockOptIP(t *kernel.Task, ep commonEndpoint, name, outLen int, family in
vP := primitive.Int32(boolToInt32(v))
return &vP, nil
+ case linux.SO_ORIGINAL_DST:
+ if outLen < int(binary.Size(linux.SockAddrInet{})) {
+ return nil, syserr.ErrInvalidArgument
+ }
+
+ var v tcpip.OriginalDestinationOption
+ if err := ep.GetSockOpt(&v); err != nil {
+ return nil, syserr.TranslateNetstackError(err)
+ }
+
+ a, _ := ConvertAddress(linux.AF_INET, tcpip.FullAddress(v))
+ return a.(*linux.SockAddrInet), nil
+
default:
emitUnimplementedEventIP(t, name)
}
diff --git a/pkg/sentry/strace/socket.go b/pkg/sentry/strace/socket.go
index c0512de89..b51c4c941 100644
--- a/pkg/sentry/strace/socket.go
+++ b/pkg/sentry/strace/socket.go
@@ -521,6 +521,7 @@ var sockOptNames = map[uint64]abi.ValueSet{
linux.IP_ROUTER_ALERT: "IP_ROUTER_ALERT",
linux.IP_PKTOPTIONS: "IP_PKTOPTIONS",
linux.IP_MTU: "IP_MTU",
+ linux.SO_ORIGINAL_DST: "SO_ORIGINAL_DST",
},
linux.SOL_SOCKET: {
linux.SO_ERROR: "SO_ERROR",
diff --git a/pkg/sentry/syscalls/linux/sys_file.go b/pkg/sentry/syscalls/linux/sys_file.go
index 2797c6a72..8cf6401e7 100644
--- a/pkg/sentry/syscalls/linux/sys_file.go
+++ b/pkg/sentry/syscalls/linux/sys_file.go
@@ -1057,7 +1057,7 @@ func Fcntl(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Syscall
case linux.F_SETOWN_EX:
addr := args[2].Pointer()
var owner linux.FOwnerEx
- n, err := t.CopyIn(addr, &owner)
+ _, err := t.CopyIn(addr, &owner)
if err != nil {
return 0, nil, err
}
@@ -1069,21 +1069,21 @@ func Fcntl(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Syscall
return 0, nil, syserror.ESRCH
}
a.SetOwnerTask(t, task)
- return uintptr(n), nil, nil
+ return 0, nil, nil
case linux.F_OWNER_PID:
tg := t.PIDNamespace().ThreadGroupWithID(kernel.ThreadID(owner.PID))
if tg == nil {
return 0, nil, syserror.ESRCH
}
a.SetOwnerThreadGroup(t, tg)
- return uintptr(n), nil, nil
+ return 0, nil, nil
case linux.F_OWNER_PGRP:
pg := t.PIDNamespace().ProcessGroupWithID(kernel.ProcessGroupID(owner.PID))
if pg == nil {
return 0, nil, syserror.ESRCH
}
a.SetOwnerProcessGroup(t, pg)
- return uintptr(n), nil, nil
+ return 0, nil, nil
default:
return 0, nil, syserror.EINVAL
}
diff --git a/pkg/sentry/syscalls/linux/vfs2/fd.go b/pkg/sentry/syscalls/linux/vfs2/fd.go
index 517394ba9..67f191551 100644
--- a/pkg/sentry/syscalls/linux/vfs2/fd.go
+++ b/pkg/sentry/syscalls/linux/vfs2/fd.go
@@ -185,11 +185,11 @@ func Fcntl(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Syscall
return 0, nil, err
case linux.F_SETOWN_EX:
var owner linux.FOwnerEx
- n, err := t.CopyIn(args[2].Pointer(), &owner)
+ _, err := t.CopyIn(args[2].Pointer(), &owner)
if err != nil {
return 0, nil, err
}
- return uintptr(n), nil, setAsyncOwner(t, file, owner.Type, owner.PID)
+ return 0, nil, setAsyncOwner(t, file, owner.Type, owner.PID)
case linux.F_GETPIPE_SZ:
pipefile, ok := file.Impl().(*pipe.VFSPipeFD)
if !ok {
diff --git a/pkg/sentry/syscalls/linux/vfs2/filesystem.go b/pkg/sentry/syscalls/linux/vfs2/filesystem.go
index 6b14c2bef..b6d2ddd65 100644
--- a/pkg/sentry/syscalls/linux/vfs2/filesystem.go
+++ b/pkg/sentry/syscalls/linux/vfs2/filesystem.go
@@ -18,7 +18,6 @@ import (
"gvisor.dev/gvisor/pkg/abi/linux"
"gvisor.dev/gvisor/pkg/sentry/arch"
"gvisor.dev/gvisor/pkg/sentry/kernel"
- "gvisor.dev/gvisor/pkg/sentry/limits"
"gvisor.dev/gvisor/pkg/sentry/vfs"
"gvisor.dev/gvisor/pkg/syserror"
"gvisor.dev/gvisor/pkg/usermem"
@@ -245,55 +244,6 @@ func renameat(t *kernel.Task, olddirfd int32, oldpathAddr usermem.Addr, newdirfd
})
}
-// Fallocate implements linux system call fallocate(2).
-func Fallocate(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.SyscallControl, error) {
- fd := args[0].Int()
- mode := args[1].Uint64()
- offset := args[2].Int64()
- length := args[3].Int64()
-
- file := t.GetFileVFS2(fd)
-
- if file == nil {
- return 0, nil, syserror.EBADF
- }
- defer file.DecRef()
-
- if !file.IsWritable() {
- return 0, nil, syserror.EBADF
- }
-
- if mode != 0 {
- return 0, nil, syserror.ENOTSUP
- }
-
- if offset < 0 || length <= 0 {
- return 0, nil, syserror.EINVAL
- }
-
- size := offset + length
-
- if size < 0 {
- return 0, nil, syserror.EFBIG
- }
-
- limit := limits.FromContext(t).Get(limits.FileSize).Cur
-
- if uint64(size) >= limit {
- t.SendSignal(&arch.SignalInfo{
- Signo: int32(linux.SIGXFSZ),
- Code: arch.SignalInfoUser,
- })
- return 0, nil, syserror.EFBIG
- }
-
- return 0, nil, file.Impl().Allocate(t, mode, uint64(offset), uint64(length))
-
- // File length modified, generate notification.
- // TODO(gvisor.dev/issue/1479): Reenable when Inotify is ported.
- // file.Dirent.InotifyEvent(linux.IN_MODIFY, 0)
-}
-
// Rmdir implements Linux syscall rmdir(2).
func Rmdir(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.SyscallControl, error) {
pathAddr := args[0].Pointer()
diff --git a/pkg/sentry/syscalls/linux/vfs2/setstat.go b/pkg/sentry/syscalls/linux/vfs2/setstat.go
index 6daedd173..37fa56c19 100644
--- a/pkg/sentry/syscalls/linux/vfs2/setstat.go
+++ b/pkg/sentry/syscalls/linux/vfs2/setstat.go
@@ -20,6 +20,7 @@ import (
"gvisor.dev/gvisor/pkg/sentry/arch"
"gvisor.dev/gvisor/pkg/sentry/kernel"
"gvisor.dev/gvisor/pkg/sentry/kernel/auth"
+ "gvisor.dev/gvisor/pkg/sentry/limits"
"gvisor.dev/gvisor/pkg/sentry/vfs"
"gvisor.dev/gvisor/pkg/syserror"
"gvisor.dev/gvisor/pkg/usermem"
@@ -211,6 +212,55 @@ func Ftruncate(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Sys
return 0, nil, handleSetSizeError(t, err)
}
+// Fallocate implements linux system call fallocate(2).
+func Fallocate(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.SyscallControl, error) {
+ fd := args[0].Int()
+ mode := args[1].Uint64()
+ offset := args[2].Int64()
+ length := args[3].Int64()
+
+ file := t.GetFileVFS2(fd)
+
+ if file == nil {
+ return 0, nil, syserror.EBADF
+ }
+ defer file.DecRef()
+
+ if !file.IsWritable() {
+ return 0, nil, syserror.EBADF
+ }
+
+ if mode != 0 {
+ return 0, nil, syserror.ENOTSUP
+ }
+
+ if offset < 0 || length <= 0 {
+ return 0, nil, syserror.EINVAL
+ }
+
+ size := offset + length
+
+ if size < 0 {
+ return 0, nil, syserror.EFBIG
+ }
+
+ limit := limits.FromContext(t).Get(limits.FileSize).Cur
+
+ if uint64(size) >= limit {
+ t.SendSignal(&arch.SignalInfo{
+ Signo: int32(linux.SIGXFSZ),
+ Code: arch.SignalInfoUser,
+ })
+ return 0, nil, syserror.EFBIG
+ }
+
+ return 0, nil, file.Allocate(t, mode, uint64(offset), uint64(length))
+
+ // File length modified, generate notification.
+ // TODO(gvisor.dev/issue/1479): Reenable when Inotify is ported.
+ // file.Dirent.InotifyEvent(linux.IN_MODIFY, 0)
+}
+
// Utime implements Linux syscall utime(2).
func Utime(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.SyscallControl, error) {
pathAddr := args[0].Pointer()
diff --git a/pkg/sentry/vfs/epoll.go b/pkg/sentry/vfs/epoll.go
index 599c3131c..5b009b928 100644
--- a/pkg/sentry/vfs/epoll.go
+++ b/pkg/sentry/vfs/epoll.go
@@ -186,7 +186,7 @@ func (ep *EpollInstance) AddInterest(file *FileDescription, num int32, event lin
}
// Register interest in file.
- mask := event.Events | linux.EPOLLERR | linux.EPOLLRDHUP
+ mask := event.Events | linux.EPOLLERR | linux.EPOLLHUP
epi := &epollInterest{
epoll: ep,
key: key,
@@ -257,7 +257,7 @@ func (ep *EpollInstance) ModifyInterest(file *FileDescription, num int32, event
}
// Update epi for the next call to ep.ReadEvents().
- mask := event.Events | linux.EPOLLERR | linux.EPOLLRDHUP
+ mask := event.Events | linux.EPOLLERR | linux.EPOLLHUP
ep.mu.Lock()
epi.mask = mask
epi.userData = event.Data
diff --git a/pkg/sentry/vfs/file_description.go b/pkg/sentry/vfs/file_description.go
index 0c42574db..93861fb4a 100644
--- a/pkg/sentry/vfs/file_description.go
+++ b/pkg/sentry/vfs/file_description.go
@@ -354,7 +354,7 @@ type FileDescriptionImpl interface {
// represented by the FileDescription.
StatFS(ctx context.Context) (linux.Statfs, error)
- // Allocate grows file represented by FileDescription to offset + length bytes.
+ // Allocate grows the file to offset + length bytes.
// Only mode == 0 is supported currently.
Allocate(ctx context.Context, mode, offset, length uint64) error
@@ -563,6 +563,11 @@ func (fd *FileDescription) StatFS(ctx context.Context) (linux.Statfs, error) {
return fd.impl.StatFS(ctx)
}
+// Allocate grows file represented by FileDescription to offset + length bytes.
+func (fd *FileDescription) Allocate(ctx context.Context, mode, offset, length uint64) error {
+ return fd.impl.Allocate(ctx, mode, offset, length)
+}
+
// Readiness implements waiter.Waitable.Readiness.
//
// It returns fd's I/O readiness.
diff --git a/pkg/syserror/syserror.go b/pkg/syserror/syserror.go
index c73072c42..798e07b01 100644
--- a/pkg/syserror/syserror.go
+++ b/pkg/syserror/syserror.go
@@ -61,6 +61,7 @@ var (
ENOMEM = error(syscall.ENOMEM)
ENOSPC = error(syscall.ENOSPC)
ENOSYS = error(syscall.ENOSYS)
+ ENOTCONN = error(syscall.ENOTCONN)
ENOTDIR = error(syscall.ENOTDIR)
ENOTEMPTY = error(syscall.ENOTEMPTY)
ENOTSOCK = error(syscall.ENOTSOCK)
diff --git a/pkg/tcpip/network/fragmentation/fragmentation.go b/pkg/tcpip/network/fragmentation/fragmentation.go
index 2982450f8..1827666c5 100644
--- a/pkg/tcpip/network/fragmentation/fragmentation.go
+++ b/pkg/tcpip/network/fragmentation/fragmentation.go
@@ -17,28 +17,58 @@
package fragmentation
import (
+ "errors"
"fmt"
"log"
"time"
"gvisor.dev/gvisor/pkg/sync"
+ "gvisor.dev/gvisor/pkg/tcpip"
"gvisor.dev/gvisor/pkg/tcpip/buffer"
)
-// DefaultReassembleTimeout is based on the linux stack: net.ipv4.ipfrag_time.
-const DefaultReassembleTimeout = 30 * time.Second
+const (
+ // DefaultReassembleTimeout is based on the linux stack: net.ipv4.ipfrag_time.
+ DefaultReassembleTimeout = 30 * time.Second
-// HighFragThreshold is the threshold at which we start trimming old
-// fragmented packets. Linux uses a default value of 4 MB. See
-// net.ipv4.ipfrag_high_thresh for more information.
-const HighFragThreshold = 4 << 20 // 4MB
+ // HighFragThreshold is the threshold at which we start trimming old
+ // fragmented packets. Linux uses a default value of 4 MB. See
+ // net.ipv4.ipfrag_high_thresh for more information.
+ HighFragThreshold = 4 << 20 // 4MB
-// LowFragThreshold is the threshold we reach to when we start dropping
-// older fragmented packets. It's important that we keep enough room for newer
-// packets to be re-assembled. Hence, this needs to be lower than
-// HighFragThreshold enough. Linux uses a default value of 3 MB. See
-// net.ipv4.ipfrag_low_thresh for more information.
-const LowFragThreshold = 3 << 20 // 3MB
+ // LowFragThreshold is the threshold we reach to when we start dropping
+ // older fragmented packets. It's important that we keep enough room for newer
+ // packets to be re-assembled. Hence, this needs to be lower than
+ // HighFragThreshold enough. Linux uses a default value of 3 MB. See
+ // net.ipv4.ipfrag_low_thresh for more information.
+ LowFragThreshold = 3 << 20 // 3MB
+
+ // minBlockSize is the minimum block size for fragments.
+ minBlockSize = 1
+)
+
+var (
+ // ErrInvalidArgs indicates to the caller that that an invalid argument was
+ // provided.
+ ErrInvalidArgs = errors.New("invalid args")
+)
+
+// FragmentID is the identifier for a fragment.
+type FragmentID struct {
+ // Source is the source address of the fragment.
+ Source tcpip.Address
+
+ // Destination is the destination address of the fragment.
+ Destination tcpip.Address
+
+ // ID is the identification value of the fragment.
+ //
+ // This is a uint32 because IPv6 uses a 32-bit identification value.
+ ID uint32
+
+ // The protocol for the packet.
+ Protocol uint8
+}
// Fragmentation is the main structure that other modules
// of the stack should use to implement IP Fragmentation.
@@ -46,14 +76,17 @@ type Fragmentation struct {
mu sync.Mutex
highLimit int
lowLimit int
- reassemblers map[uint32]*reassembler
+ reassemblers map[FragmentID]*reassembler
rList reassemblerList
size int
timeout time.Duration
+ blockSize uint16
}
// NewFragmentation creates a new Fragmentation.
//
+// blockSize specifies the fragment block size, in bytes.
+//
// highMemoryLimit specifies the limit on the memory consumed
// by the fragments stored by Fragmentation (overhead of internal data-structures
// is not accounted). Fragments are dropped when the limit is reached.
@@ -64,7 +97,7 @@ type Fragmentation struct {
// reassemblingTimeout specifies the maximum time allowed to reassemble a packet.
// Fragments are lazily evicted only when a new a packet with an
// already existing fragmentation-id arrives after the timeout.
-func NewFragmentation(highMemoryLimit, lowMemoryLimit int, reassemblingTimeout time.Duration) *Fragmentation {
+func NewFragmentation(blockSize uint16, highMemoryLimit, lowMemoryLimit int, reassemblingTimeout time.Duration) *Fragmentation {
if lowMemoryLimit >= highMemoryLimit {
lowMemoryLimit = highMemoryLimit
}
@@ -73,17 +106,46 @@ func NewFragmentation(highMemoryLimit, lowMemoryLimit int, reassemblingTimeout t
lowMemoryLimit = 0
}
+ if blockSize < minBlockSize {
+ blockSize = minBlockSize
+ }
+
return &Fragmentation{
- reassemblers: make(map[uint32]*reassembler),
+ reassemblers: make(map[FragmentID]*reassembler),
highLimit: highMemoryLimit,
lowLimit: lowMemoryLimit,
timeout: reassemblingTimeout,
+ blockSize: blockSize,
}
}
// Process processes an incoming fragment belonging to an ID and returns a
// complete packet when all the packets belonging to that ID have been received.
-func (f *Fragmentation) Process(id uint32, first, last uint16, more bool, vv buffer.VectorisedView) (buffer.VectorisedView, bool, error) {
+//
+// [first, last] is the range of the fragment bytes.
+//
+// first must be a multiple of the block size f is configured with. The size
+// of the fragment data must be a multiple of the block size, unless there are
+// no fragments following this fragment (more set to false).
+func (f *Fragmentation) Process(id FragmentID, first, last uint16, more bool, vv buffer.VectorisedView) (buffer.VectorisedView, bool, error) {
+ if first > last {
+ return buffer.VectorisedView{}, false, fmt.Errorf("first=%d is greater than last=%d: %w", first, last, ErrInvalidArgs)
+ }
+
+ if first%f.blockSize != 0 {
+ return buffer.VectorisedView{}, false, fmt.Errorf("first=%d is not a multiple of block size=%d: %w", first, f.blockSize, ErrInvalidArgs)
+ }
+
+ fragmentSize := last - first + 1
+ if more && fragmentSize%f.blockSize != 0 {
+ return buffer.VectorisedView{}, false, fmt.Errorf("fragment size=%d bytes is not a multiple of block size=%d on non-final fragment: %w", fragmentSize, f.blockSize, ErrInvalidArgs)
+ }
+
+ if l := vv.Size(); l < int(fragmentSize) {
+ return buffer.VectorisedView{}, false, fmt.Errorf("got fragment size=%d bytes less than the expected fragment size=%d bytes (first=%d last=%d): %w", l, fragmentSize, first, last, ErrInvalidArgs)
+ }
+ vv.CapLength(int(fragmentSize))
+
f.mu.Lock()
r, ok := f.reassemblers[id]
if ok && r.tooOld(f.timeout) {
diff --git a/pkg/tcpip/network/fragmentation/fragmentation_test.go b/pkg/tcpip/network/fragmentation/fragmentation_test.go
index 72c0f53be..9eedd33c4 100644
--- a/pkg/tcpip/network/fragmentation/fragmentation_test.go
+++ b/pkg/tcpip/network/fragmentation/fragmentation_test.go
@@ -15,6 +15,7 @@
package fragmentation
import (
+ "errors"
"reflect"
"testing"
"time"
@@ -33,7 +34,7 @@ func vv(size int, pieces ...string) buffer.VectorisedView {
}
type processInput struct {
- id uint32
+ id FragmentID
first uint16
last uint16
more bool
@@ -53,8 +54,8 @@ var processTestCases = []struct {
{
comment: "One ID",
in: []processInput{
- {id: 0, first: 0, last: 1, more: true, vv: vv(2, "01")},
- {id: 0, first: 2, last: 3, more: false, vv: vv(2, "23")},
+ {id: FragmentID{ID: 0}, first: 0, last: 1, more: true, vv: vv(2, "01")},
+ {id: FragmentID{ID: 0}, first: 2, last: 3, more: false, vv: vv(2, "23")},
},
out: []processOutput{
{vv: buffer.VectorisedView{}, done: false},
@@ -64,10 +65,10 @@ var processTestCases = []struct {
{
comment: "Two IDs",
in: []processInput{
- {id: 0, first: 0, last: 1, more: true, vv: vv(2, "01")},
- {id: 1, first: 0, last: 1, more: true, vv: vv(2, "ab")},
- {id: 1, first: 2, last: 3, more: false, vv: vv(2, "cd")},
- {id: 0, first: 2, last: 3, more: false, vv: vv(2, "23")},
+ {id: FragmentID{ID: 0}, first: 0, last: 1, more: true, vv: vv(2, "01")},
+ {id: FragmentID{ID: 1}, first: 0, last: 1, more: true, vv: vv(2, "ab")},
+ {id: FragmentID{ID: 1}, first: 2, last: 3, more: false, vv: vv(2, "cd")},
+ {id: FragmentID{ID: 0}, first: 2, last: 3, more: false, vv: vv(2, "23")},
},
out: []processOutput{
{vv: buffer.VectorisedView{}, done: false},
@@ -81,7 +82,7 @@ var processTestCases = []struct {
func TestFragmentationProcess(t *testing.T) {
for _, c := range processTestCases {
t.Run(c.comment, func(t *testing.T) {
- f := NewFragmentation(1024, 512, DefaultReassembleTimeout)
+ f := NewFragmentation(minBlockSize, 1024, 512, DefaultReassembleTimeout)
for i, in := range c.in {
vv, done, err := f.Process(in.id, in.first, in.last, in.more, in.vv)
if err != nil {
@@ -110,14 +111,14 @@ func TestFragmentationProcess(t *testing.T) {
func TestReassemblingTimeout(t *testing.T) {
timeout := time.Millisecond
- f := NewFragmentation(1024, 512, timeout)
+ f := NewFragmentation(minBlockSize, 1024, 512, timeout)
// Send first fragment with id = 0, first = 0, last = 0, and more = true.
- f.Process(0, 0, 0, true, vv(1, "0"))
+ f.Process(FragmentID{}, 0, 0, true, vv(1, "0"))
// Sleep more than the timeout.
time.Sleep(2 * timeout)
// Send another fragment that completes a packet.
// However, no packet should be reassembled because the fragment arrived after the timeout.
- _, done, err := f.Process(0, 1, 1, false, vv(1, "1"))
+ _, done, err := f.Process(FragmentID{}, 1, 1, false, vv(1, "1"))
if err != nil {
t.Fatalf("f.Process(0, 1, 1, false, vv(1, \"1\")) failed: %v", err)
}
@@ -127,35 +128,35 @@ func TestReassemblingTimeout(t *testing.T) {
}
func TestMemoryLimits(t *testing.T) {
- f := NewFragmentation(3, 1, DefaultReassembleTimeout)
+ f := NewFragmentation(minBlockSize, 3, 1, DefaultReassembleTimeout)
// Send first fragment with id = 0.
- f.Process(0, 0, 0, true, vv(1, "0"))
+ f.Process(FragmentID{ID: 0}, 0, 0, true, vv(1, "0"))
// Send first fragment with id = 1.
- f.Process(1, 0, 0, true, vv(1, "1"))
+ f.Process(FragmentID{ID: 1}, 0, 0, true, vv(1, "1"))
// Send first fragment with id = 2.
- f.Process(2, 0, 0, true, vv(1, "2"))
+ f.Process(FragmentID{ID: 2}, 0, 0, true, vv(1, "2"))
// Send first fragment with id = 3. This should caused id = 0 and id = 1 to be
// evicted.
- f.Process(3, 0, 0, true, vv(1, "3"))
+ f.Process(FragmentID{ID: 3}, 0, 0, true, vv(1, "3"))
- if _, ok := f.reassemblers[0]; ok {
+ if _, ok := f.reassemblers[FragmentID{ID: 0}]; ok {
t.Errorf("Memory limits are not respected: id=0 has not been evicted.")
}
- if _, ok := f.reassemblers[1]; ok {
+ if _, ok := f.reassemblers[FragmentID{ID: 1}]; ok {
t.Errorf("Memory limits are not respected: id=1 has not been evicted.")
}
- if _, ok := f.reassemblers[3]; !ok {
+ if _, ok := f.reassemblers[FragmentID{ID: 3}]; !ok {
t.Errorf("Implementation of memory limits is wrong: id=3 is not present.")
}
}
func TestMemoryLimitsIgnoresDuplicates(t *testing.T) {
- f := NewFragmentation(1, 0, DefaultReassembleTimeout)
+ f := NewFragmentation(minBlockSize, 1, 0, DefaultReassembleTimeout)
// Send first fragment with id = 0.
- f.Process(0, 0, 0, true, vv(1, "0"))
+ f.Process(FragmentID{}, 0, 0, true, vv(1, "0"))
// Send the same packet again.
- f.Process(0, 0, 0, true, vv(1, "0"))
+ f.Process(FragmentID{}, 0, 0, true, vv(1, "0"))
got := f.size
want := 1
@@ -163,3 +164,97 @@ func TestMemoryLimitsIgnoresDuplicates(t *testing.T) {
t.Errorf("Wrong size, duplicates are not handled correctly: got=%d, want=%d.", got, want)
}
}
+
+func TestErrors(t *testing.T) {
+ tests := []struct {
+ name string
+ blockSize uint16
+ first uint16
+ last uint16
+ more bool
+ data string
+ err error
+ }{
+ {
+ name: "exact block size without more",
+ blockSize: 2,
+ first: 2,
+ last: 3,
+ more: false,
+ data: "01",
+ },
+ {
+ name: "exact block size with more",
+ blockSize: 2,
+ first: 2,
+ last: 3,
+ more: true,
+ data: "01",
+ },
+ {
+ name: "exact block size with more and extra data",
+ blockSize: 2,
+ first: 2,
+ last: 3,
+ more: true,
+ data: "012",
+ },
+ {
+ name: "exact block size with more and too little data",
+ blockSize: 2,
+ first: 2,
+ last: 3,
+ more: true,
+ data: "0",
+ err: ErrInvalidArgs,
+ },
+ {
+ name: "not exact block size with more",
+ blockSize: 2,
+ first: 2,
+ last: 2,
+ more: true,
+ data: "0",
+ err: ErrInvalidArgs,
+ },
+ {
+ name: "not exact block size without more",
+ blockSize: 2,
+ first: 2,
+ last: 2,
+ more: false,
+ data: "0",
+ },
+ {
+ name: "first not a multiple of block size",
+ blockSize: 2,
+ first: 3,
+ last: 4,
+ more: true,
+ data: "01",
+ err: ErrInvalidArgs,
+ },
+ {
+ name: "first more than last",
+ blockSize: 2,
+ first: 4,
+ last: 3,
+ more: true,
+ data: "01",
+ err: ErrInvalidArgs,
+ },
+ }
+
+ for _, test := range tests {
+ t.Run(test.name, func(t *testing.T) {
+ f := NewFragmentation(test.blockSize, HighFragThreshold, LowFragThreshold, DefaultReassembleTimeout)
+ _, done, err := f.Process(FragmentID{}, test.first, test.last, test.more, vv(len(test.data), test.data))
+ if !errors.Is(err, test.err) {
+ t.Errorf("got Proceess(_, %d, %d, %t, %q) = (_, _, %v), want = (_, _, %v)", test.first, test.last, test.more, test.data, err, test.err)
+ }
+ if done {
+ t.Errorf("got Proceess(_, %d, %d, %t, %q) = (_, true, _), want = (_, false, _)", test.first, test.last, test.more, test.data)
+ }
+ })
+ }
+}
diff --git a/pkg/tcpip/network/fragmentation/reassembler.go b/pkg/tcpip/network/fragmentation/reassembler.go
index 0a83d81f2..50d30bbf0 100644
--- a/pkg/tcpip/network/fragmentation/reassembler.go
+++ b/pkg/tcpip/network/fragmentation/reassembler.go
@@ -32,7 +32,7 @@ type hole struct {
type reassembler struct {
reassemblerEntry
- id uint32
+ id FragmentID
size int
mu sync.Mutex
holes []hole
@@ -42,7 +42,7 @@ type reassembler struct {
creationTime time.Time
}
-func newReassembler(id uint32) *reassembler {
+func newReassembler(id FragmentID) *reassembler {
r := &reassembler{
id: id,
holes: make([]hole, 0, 16),
diff --git a/pkg/tcpip/network/fragmentation/reassembler_test.go b/pkg/tcpip/network/fragmentation/reassembler_test.go
index 7eee0710d..dff7c9dcb 100644
--- a/pkg/tcpip/network/fragmentation/reassembler_test.go
+++ b/pkg/tcpip/network/fragmentation/reassembler_test.go
@@ -94,7 +94,7 @@ var holesTestCases = []struct {
func TestUpdateHoles(t *testing.T) {
for _, c := range holesTestCases {
- r := newReassembler(0)
+ r := newReassembler(FragmentID{})
for _, i := range c.in {
r.updateHoles(i.first, i.last, i.more)
}
diff --git a/pkg/tcpip/network/ipv4/ipv4.go b/pkg/tcpip/network/ipv4/ipv4.go
index b1776e5ee..d5f5d38f7 100644
--- a/pkg/tcpip/network/ipv4/ipv4.go
+++ b/pkg/tcpip/network/ipv4/ipv4.go
@@ -45,6 +45,10 @@ const (
// buckets is the number of identifier buckets.
buckets = 2048
+
+ // The size of a fragment block, in bytes, as per RFC 791 section 3.1,
+ // page 14.
+ fragmentblockSize = 8
)
type endpoint struct {
@@ -66,7 +70,7 @@ func (p *protocol) NewEndpoint(nicID tcpip.NICID, addrWithPrefix tcpip.AddressWi
prefixLen: addrWithPrefix.PrefixLen,
linkEP: linkEP,
dispatcher: dispatcher,
- fragmentation: fragmentation.NewFragmentation(fragmentation.HighFragThreshold, fragmentation.LowFragThreshold, fragmentation.DefaultReassembleTimeout),
+ fragmentation: fragmentation.NewFragmentation(fragmentblockSize, fragmentation.HighFragThreshold, fragmentation.LowFragThreshold, fragmentation.DefaultReassembleTimeout),
protocol: p,
stack: st,
}
@@ -438,7 +442,18 @@ func (e *endpoint) HandlePacket(r *stack.Route, pkt *stack.PacketBuffer) {
}
var ready bool
var err error
- pkt.Data, ready, err = e.fragmentation.Process(hash.IPv4FragmentHash(h), h.FragmentOffset(), last, h.More(), pkt.Data)
+ pkt.Data, ready, err = e.fragmentation.Process(
+ fragmentation.FragmentID{
+ Source: h.SourceAddress(),
+ Destination: h.DestinationAddress(),
+ ID: uint32(h.ID()),
+ Protocol: h.Protocol(),
+ },
+ h.FragmentOffset(),
+ last,
+ h.More(),
+ pkt.Data,
+ )
if err != nil {
r.Stats().IP.MalformedPacketsReceived.Increment()
r.Stats().IP.MalformedFragmentsReceived.Increment()
diff --git a/pkg/tcpip/network/ipv4/ipv4_test.go b/pkg/tcpip/network/ipv4/ipv4_test.go
index 11e579c4b..ded97ac64 100644
--- a/pkg/tcpip/network/ipv4/ipv4_test.go
+++ b/pkg/tcpip/network/ipv4/ipv4_test.go
@@ -486,12 +486,16 @@ func TestInvalidFragments(t *testing.T) {
// TestReceiveFragments feeds fragments in through the incoming packet path to
// test reassembly
func TestReceiveFragments(t *testing.T) {
- const addr1 = "\x0c\xa8\x00\x01" // 192.168.0.1
- const addr2 = "\x0c\xa8\x00\x02" // 192.168.0.2
- const nicID = 1
+ const (
+ nicID = 1
+
+ addr1 = "\x0c\xa8\x00\x01" // 192.168.0.1
+ addr2 = "\x0c\xa8\x00\x02" // 192.168.0.2
+ addr3 = "\x0c\xa8\x00\x03" // 192.168.0.3
+ )
// Build and return a UDP header containing payload.
- udpGen := func(payloadLen int, multiplier uint8) buffer.View {
+ udpGen := func(payloadLen int, multiplier uint8, src, dst tcpip.Address) buffer.View {
payload := buffer.NewView(payloadLen)
for i := 0; i < len(payload); i++ {
payload[i] = uint8(i) * multiplier
@@ -507,20 +511,29 @@ func TestReceiveFragments(t *testing.T) {
Length: uint16(udpLength),
})
copy(u.Payload(), payload)
- sum := header.PseudoHeaderChecksum(udp.ProtocolNumber, addr1, addr2, uint16(udpLength))
+ sum := header.PseudoHeaderChecksum(udp.ProtocolNumber, src, dst, uint16(udpLength))
sum = header.Checksum(payload, sum)
u.SetChecksum(^u.CalculateChecksum(sum))
return hdr.View()
}
// UDP header plus a payload of 0..256
- ipv4Payload1 := udpGen(256, 1)
- udpPayload1 := ipv4Payload1[header.UDPMinimumSize:]
+ ipv4Payload1Addr1ToAddr2 := udpGen(256, 1, addr1, addr2)
+ udpPayload1Addr1ToAddr2 := ipv4Payload1Addr1ToAddr2[header.UDPMinimumSize:]
+ ipv4Payload1Addr3ToAddr2 := udpGen(256, 1, addr3, addr2)
+ udpPayload1Addr3ToAddr2 := ipv4Payload1Addr3ToAddr2[header.UDPMinimumSize:]
// UDP header plus a payload of 0..256 in increments of 2.
- ipv4Payload2 := udpGen(128, 2)
- udpPayload2 := ipv4Payload2[header.UDPMinimumSize:]
+ ipv4Payload2Addr1ToAddr2 := udpGen(128, 2, addr1, addr2)
+ udpPayload2Addr1ToAddr2 := ipv4Payload2Addr1ToAddr2[header.UDPMinimumSize:]
+ // UDP header plus a payload of 0..256 in increments of 3.
+ // Used to test cases where the fragment blocks are not a multiple of
+ // the fragment block size of 8 (RFC 791 section 3.1 page 14).
+ ipv4Payload3Addr1ToAddr2 := udpGen(127, 3, addr1, addr2)
+ udpPayload3Addr1ToAddr2 := ipv4Payload3Addr1ToAddr2[header.UDPMinimumSize:]
type fragmentData struct {
+ srcAddr tcpip.Address
+ dstAddr tcpip.Address
id uint16
flags uint8
fragmentOffset uint16
@@ -536,22 +549,40 @@ func TestReceiveFragments(t *testing.T) {
name: "No fragmentation",
fragments: []fragmentData{
{
+ srcAddr: addr1,
+ dstAddr: addr2,
+ id: 1,
+ flags: 0,
+ fragmentOffset: 0,
+ payload: ipv4Payload1Addr1ToAddr2,
+ },
+ },
+ expectedPayloads: [][]byte{udpPayload1Addr1ToAddr2},
+ },
+ {
+ name: "No fragmentation with size not a multiple of fragment block size",
+ fragments: []fragmentData{
+ {
+ srcAddr: addr1,
+ dstAddr: addr2,
id: 1,
flags: 0,
fragmentOffset: 0,
- payload: ipv4Payload1,
+ payload: ipv4Payload3Addr1ToAddr2,
},
},
- expectedPayloads: [][]byte{udpPayload1},
+ expectedPayloads: [][]byte{udpPayload3Addr1ToAddr2},
},
{
name: "More fragments without payload",
fragments: []fragmentData{
{
+ srcAddr: addr1,
+ dstAddr: addr2,
id: 1,
flags: header.IPv4FlagMoreFragments,
fragmentOffset: 0,
- payload: ipv4Payload1,
+ payload: ipv4Payload1Addr1ToAddr2,
},
},
expectedPayloads: nil,
@@ -560,10 +591,12 @@ func TestReceiveFragments(t *testing.T) {
name: "Non-zero fragment offset without payload",
fragments: []fragmentData{
{
+ srcAddr: addr1,
+ dstAddr: addr2,
id: 1,
flags: 0,
fragmentOffset: 8,
- payload: ipv4Payload1,
+ payload: ipv4Payload1Addr1ToAddr2,
},
},
expectedPayloads: nil,
@@ -572,34 +605,86 @@ func TestReceiveFragments(t *testing.T) {
name: "Two fragments",
fragments: []fragmentData{
{
+ srcAddr: addr1,
+ dstAddr: addr2,
+ id: 1,
+ flags: header.IPv4FlagMoreFragments,
+ fragmentOffset: 0,
+ payload: ipv4Payload1Addr1ToAddr2[:64],
+ },
+ {
+ srcAddr: addr1,
+ dstAddr: addr2,
+ id: 1,
+ flags: 0,
+ fragmentOffset: 64,
+ payload: ipv4Payload1Addr1ToAddr2[64:],
+ },
+ },
+ expectedPayloads: [][]byte{udpPayload1Addr1ToAddr2},
+ },
+ {
+ name: "Two fragments with last fragment size not a multiple of fragment block size",
+ fragments: []fragmentData{
+ {
+ srcAddr: addr1,
+ dstAddr: addr2,
id: 1,
flags: header.IPv4FlagMoreFragments,
fragmentOffset: 0,
- payload: ipv4Payload1[:64],
+ payload: ipv4Payload3Addr1ToAddr2[:64],
},
{
+ srcAddr: addr1,
+ dstAddr: addr2,
id: 1,
flags: 0,
fragmentOffset: 64,
- payload: ipv4Payload1[64:],
+ payload: ipv4Payload3Addr1ToAddr2[64:],
},
},
- expectedPayloads: [][]byte{udpPayload1},
+ expectedPayloads: [][]byte{udpPayload3Addr1ToAddr2},
+ },
+ {
+ name: "Two fragments with first fragment size not a multiple of fragment block size",
+ fragments: []fragmentData{
+ {
+ srcAddr: addr1,
+ dstAddr: addr2,
+ id: 1,
+ flags: header.IPv4FlagMoreFragments,
+ fragmentOffset: 0,
+ payload: ipv4Payload3Addr1ToAddr2[:63],
+ },
+ {
+ srcAddr: addr1,
+ dstAddr: addr2,
+ id: 1,
+ flags: 0,
+ fragmentOffset: 63,
+ payload: ipv4Payload3Addr1ToAddr2[63:],
+ },
+ },
+ expectedPayloads: nil,
},
{
name: "Second fragment has MoreFlags set",
fragments: []fragmentData{
{
+ srcAddr: addr1,
+ dstAddr: addr2,
id: 1,
flags: header.IPv4FlagMoreFragments,
fragmentOffset: 0,
- payload: ipv4Payload1[:64],
+ payload: ipv4Payload1Addr1ToAddr2[:64],
},
{
+ srcAddr: addr1,
+ dstAddr: addr2,
id: 1,
flags: header.IPv4FlagMoreFragments,
fragmentOffset: 64,
- payload: ipv4Payload1[64:],
+ payload: ipv4Payload1Addr1ToAddr2[64:],
},
},
expectedPayloads: nil,
@@ -608,16 +693,20 @@ func TestReceiveFragments(t *testing.T) {
name: "Two fragments with different IDs",
fragments: []fragmentData{
{
+ srcAddr: addr1,
+ dstAddr: addr2,
id: 1,
flags: header.IPv4FlagMoreFragments,
fragmentOffset: 0,
- payload: ipv4Payload1[:64],
+ payload: ipv4Payload1Addr1ToAddr2[:64],
},
{
+ srcAddr: addr1,
+ dstAddr: addr2,
id: 2,
flags: 0,
fragmentOffset: 64,
- payload: ipv4Payload1[64:],
+ payload: ipv4Payload1Addr1ToAddr2[64:],
},
},
expectedPayloads: nil,
@@ -626,40 +715,88 @@ func TestReceiveFragments(t *testing.T) {
name: "Two interleaved fragmented packets",
fragments: []fragmentData{
{
+ srcAddr: addr1,
+ dstAddr: addr2,
id: 1,
flags: header.IPv4FlagMoreFragments,
fragmentOffset: 0,
- payload: ipv4Payload1[:64],
+ payload: ipv4Payload1Addr1ToAddr2[:64],
},
{
+ srcAddr: addr1,
+ dstAddr: addr2,
id: 2,
flags: header.IPv4FlagMoreFragments,
fragmentOffset: 0,
- payload: ipv4Payload2[:64],
+ payload: ipv4Payload2Addr1ToAddr2[:64],
},
{
+ srcAddr: addr1,
+ dstAddr: addr2,
id: 1,
flags: 0,
fragmentOffset: 64,
- payload: ipv4Payload1[64:],
+ payload: ipv4Payload1Addr1ToAddr2[64:],
},
{
+ srcAddr: addr1,
+ dstAddr: addr2,
id: 2,
flags: 0,
fragmentOffset: 64,
- payload: ipv4Payload2[64:],
+ payload: ipv4Payload2Addr1ToAddr2[64:],
+ },
+ },
+ expectedPayloads: [][]byte{udpPayload1Addr1ToAddr2, udpPayload2Addr1ToAddr2},
+ },
+ {
+ name: "Two interleaved fragmented packets from different sources but with same ID",
+ fragments: []fragmentData{
+ {
+ srcAddr: addr1,
+ dstAddr: addr2,
+ id: 1,
+ flags: header.IPv4FlagMoreFragments,
+ fragmentOffset: 0,
+ payload: ipv4Payload1Addr1ToAddr2[:64],
+ },
+ {
+ srcAddr: addr3,
+ dstAddr: addr2,
+ id: 1,
+ flags: header.IPv4FlagMoreFragments,
+ fragmentOffset: 0,
+ payload: ipv4Payload1Addr3ToAddr2[:32],
+ },
+ {
+ srcAddr: addr1,
+ dstAddr: addr2,
+ id: 1,
+ flags: 0,
+ fragmentOffset: 64,
+ payload: ipv4Payload1Addr1ToAddr2[64:],
+ },
+ {
+ srcAddr: addr3,
+ dstAddr: addr2,
+ id: 1,
+ flags: 0,
+ fragmentOffset: 32,
+ payload: ipv4Payload1Addr3ToAddr2[32:],
},
},
- expectedPayloads: [][]byte{udpPayload1, udpPayload2},
+ expectedPayloads: [][]byte{udpPayload1Addr1ToAddr2, udpPayload1Addr3ToAddr2},
},
{
name: "Fragment without followup",
fragments: []fragmentData{
{
+ srcAddr: addr1,
+ dstAddr: addr2,
id: 1,
flags: header.IPv4FlagMoreFragments,
fragmentOffset: 0,
- payload: ipv4Payload1[:64],
+ payload: ipv4Payload1Addr1ToAddr2[:64],
},
},
expectedPayloads: nil,
@@ -711,8 +848,8 @@ func TestReceiveFragments(t *testing.T) {
FragmentOffset: frag.fragmentOffset,
TTL: 64,
Protocol: uint8(header.UDPProtocolNumber),
- SrcAddr: addr1,
- DstAddr: addr2,
+ SrcAddr: frag.srcAddr,
+ DstAddr: frag.dstAddr,
})
vv := hdr.View().ToVectorisedView()
diff --git a/pkg/tcpip/network/ipv6/BUILD b/pkg/tcpip/network/ipv6/BUILD
index feada63dc..bcc64994e 100644
--- a/pkg/tcpip/network/ipv6/BUILD
+++ b/pkg/tcpip/network/ipv6/BUILD
@@ -14,7 +14,6 @@ go_library(
"//pkg/tcpip/buffer",
"//pkg/tcpip/header",
"//pkg/tcpip/network/fragmentation",
- "//pkg/tcpip/network/hash",
"//pkg/tcpip/stack",
],
)
diff --git a/pkg/tcpip/network/ipv6/ipv6.go b/pkg/tcpip/network/ipv6/ipv6.go
index 95fbcf2d1..a0a5c9c01 100644
--- a/pkg/tcpip/network/ipv6/ipv6.go
+++ b/pkg/tcpip/network/ipv6/ipv6.go
@@ -28,7 +28,6 @@ import (
"gvisor.dev/gvisor/pkg/tcpip/buffer"
"gvisor.dev/gvisor/pkg/tcpip/header"
"gvisor.dev/gvisor/pkg/tcpip/network/fragmentation"
- "gvisor.dev/gvisor/pkg/tcpip/network/hash"
"gvisor.dev/gvisor/pkg/tcpip/stack"
)
@@ -343,7 +342,19 @@ func (e *endpoint) HandlePacket(r *stack.Route, pkt *stack.PacketBuffer) {
var ready bool
// Note that pkt doesn't have its transport header set after reassembly,
// and won't until DeliverNetworkPacket sets it.
- pkt.Data, ready, err = e.fragmentation.Process(hash.IPv6FragmentHash(h, extHdr.ID()), start, last, extHdr.More(), rawPayload.Buf)
+ pkt.Data, ready, err = e.fragmentation.Process(
+ // IPv6 ignores the Protocol field since the ID only needs to be unique
+ // across source-destination pairs, as per RFC 8200 section 4.5.
+ fragmentation.FragmentID{
+ Source: h.SourceAddress(),
+ Destination: h.DestinationAddress(),
+ ID: extHdr.ID(),
+ },
+ start,
+ last,
+ extHdr.More(),
+ rawPayload.Buf,
+ )
if err != nil {
r.Stats().IP.MalformedPacketsReceived.Increment()
r.Stats().IP.MalformedFragmentsReceived.Increment()
@@ -467,7 +478,7 @@ func (p *protocol) NewEndpoint(nicID tcpip.NICID, addrWithPrefix tcpip.AddressWi
linkEP: linkEP,
linkAddrCache: linkAddrCache,
dispatcher: dispatcher,
- fragmentation: fragmentation.NewFragmentation(fragmentation.HighFragThreshold, fragmentation.LowFragThreshold, fragmentation.DefaultReassembleTimeout),
+ fragmentation: fragmentation.NewFragmentation(header.IPv6FragmentExtHdrFragmentOffsetBytesPerUnit, fragmentation.HighFragThreshold, fragmentation.LowFragThreshold, fragmentation.DefaultReassembleTimeout),
protocol: p,
}, nil
}
diff --git a/pkg/tcpip/network/ipv6/ipv6_test.go b/pkg/tcpip/network/ipv6/ipv6_test.go
index 213ff64f2..3d65814de 100644
--- a/pkg/tcpip/network/ipv6/ipv6_test.go
+++ b/pkg/tcpip/network/ipv6/ipv6_test.go
@@ -673,20 +673,27 @@ func TestReceiveIPv6ExtHdrs(t *testing.T) {
// fragmentData holds the IPv6 payload for a fragmented IPv6 packet.
type fragmentData struct {
+ srcAddr tcpip.Address
+ dstAddr tcpip.Address
nextHdr uint8
data buffer.VectorisedView
}
func TestReceiveIPv6Fragments(t *testing.T) {
- const nicID = 1
- const udpPayload1Length = 256
- const udpPayload2Length = 128
- const fragmentExtHdrLen = 8
- // Note, not all routing extension headers will be 8 bytes but this test
- // uses 8 byte routing extension headers for most sub tests.
- const routingExtHdrLen = 8
-
- udpGen := func(payload []byte, multiplier uint8) buffer.View {
+ const (
+ nicID = 1
+ udpPayload1Length = 256
+ udpPayload2Length = 128
+ // Used to test cases where the fragment blocks are not a multiple of
+ // the fragment block size of 8 (RFC 8200 section 4.5).
+ udpPayload3Length = 127
+ fragmentExtHdrLen = 8
+ // Note, not all routing extension headers will be 8 bytes but this test
+ // uses 8 byte routing extension headers for most sub tests.
+ routingExtHdrLen = 8
+ )
+
+ udpGen := func(payload []byte, multiplier uint8, src, dst tcpip.Address) buffer.View {
payloadLen := len(payload)
for i := 0; i < payloadLen; i++ {
payload[i] = uint8(i) * multiplier
@@ -702,19 +709,27 @@ func TestReceiveIPv6Fragments(t *testing.T) {
Length: uint16(udpLength),
})
copy(u.Payload(), payload)
- sum := header.PseudoHeaderChecksum(udp.ProtocolNumber, addr1, addr2, uint16(udpLength))
+ sum := header.PseudoHeaderChecksum(udp.ProtocolNumber, src, dst, uint16(udpLength))
sum = header.Checksum(payload, sum)
u.SetChecksum(^u.CalculateChecksum(sum))
return hdr.View()
}
- var udpPayload1Buf [udpPayload1Length]byte
- udpPayload1 := udpPayload1Buf[:]
- ipv6Payload1 := udpGen(udpPayload1, 1)
+ var udpPayload1Addr1ToAddr2Buf [udpPayload1Length]byte
+ udpPayload1Addr1ToAddr2 := udpPayload1Addr1ToAddr2Buf[:]
+ ipv6Payload1Addr1ToAddr2 := udpGen(udpPayload1Addr1ToAddr2, 1, addr1, addr2)
+
+ var udpPayload1Addr3ToAddr2Buf [udpPayload1Length]byte
+ udpPayload1Addr3ToAddr2 := udpPayload1Addr3ToAddr2Buf[:]
+ ipv6Payload1Addr3ToAddr2 := udpGen(udpPayload1Addr3ToAddr2, 4, addr3, addr2)
- var udpPayload2Buf [udpPayload2Length]byte
- udpPayload2 := udpPayload2Buf[:]
- ipv6Payload2 := udpGen(udpPayload2, 2)
+ var udpPayload2Addr1ToAddr2Buf [udpPayload2Length]byte
+ udpPayload2Addr1ToAddr2 := udpPayload2Addr1ToAddr2Buf[:]
+ ipv6Payload2Addr1ToAddr2 := udpGen(udpPayload2Addr1ToAddr2, 2, addr1, addr2)
+
+ var udpPayload3Addr1ToAddr2Buf [udpPayload3Length]byte
+ udpPayload3Addr1ToAddr2 := udpPayload3Addr1ToAddr2Buf[:]
+ ipv6Payload3Addr1ToAddr2 := udpGen(udpPayload3Addr1ToAddr2, 3, addr1, addr2)
tests := []struct {
name string
@@ -726,34 +741,98 @@ func TestReceiveIPv6Fragments(t *testing.T) {
name: "No fragmentation",
fragments: []fragmentData{
{
+ srcAddr: addr1,
+ dstAddr: addr2,
nextHdr: uint8(header.UDPProtocolNumber),
- data: ipv6Payload1.ToVectorisedView(),
+ data: ipv6Payload1Addr1ToAddr2.ToVectorisedView(),
},
},
- expectedPayloads: [][]byte{udpPayload1},
+ expectedPayloads: [][]byte{udpPayload1Addr1ToAddr2},
},
{
name: "Atomic fragment",
fragments: []fragmentData{
{
+ srcAddr: addr1,
+ dstAddr: addr2,
+ nextHdr: fragmentExtHdrID,
+ data: buffer.NewVectorisedView(
+ fragmentExtHdrLen+len(ipv6Payload1Addr1ToAddr2),
+ []buffer.View{
+ // Fragment extension header.
+ buffer.View([]byte{uint8(header.UDPProtocolNumber), 0, 0, 0, 0, 0, 0, 0}),
+
+ ipv6Payload1Addr1ToAddr2,
+ },
+ ),
+ },
+ },
+ expectedPayloads: [][]byte{udpPayload1Addr1ToAddr2},
+ },
+ {
+ name: "Atomic fragment with size not a multiple of fragment block size",
+ fragments: []fragmentData{
+ {
+ srcAddr: addr1,
+ dstAddr: addr2,
nextHdr: fragmentExtHdrID,
data: buffer.NewVectorisedView(
- fragmentExtHdrLen+len(ipv6Payload1),
+ fragmentExtHdrLen+len(ipv6Payload3Addr1ToAddr2),
[]buffer.View{
// Fragment extension header.
buffer.View([]byte{uint8(header.UDPProtocolNumber), 0, 0, 0, 0, 0, 0, 0}),
- ipv6Payload1,
+ ipv6Payload3Addr1ToAddr2,
},
),
},
},
- expectedPayloads: [][]byte{udpPayload1},
+ expectedPayloads: [][]byte{udpPayload3Addr1ToAddr2},
},
{
name: "Two fragments",
fragments: []fragmentData{
{
+ srcAddr: addr1,
+ dstAddr: addr2,
+ nextHdr: fragmentExtHdrID,
+ data: buffer.NewVectorisedView(
+ fragmentExtHdrLen+64,
+ []buffer.View{
+ // Fragment extension header.
+ //
+ // Fragment offset = 0, More = true, ID = 1
+ buffer.View([]byte{uint8(header.UDPProtocolNumber), 0, 0, 1, 0, 0, 0, 1}),
+
+ ipv6Payload1Addr1ToAddr2[:64],
+ },
+ ),
+ },
+ {
+ srcAddr: addr1,
+ dstAddr: addr2,
+ nextHdr: fragmentExtHdrID,
+ data: buffer.NewVectorisedView(
+ fragmentExtHdrLen+len(ipv6Payload1Addr1ToAddr2)-64,
+ []buffer.View{
+ // Fragment extension header.
+ //
+ // Fragment offset = 8, More = false, ID = 1
+ buffer.View([]byte{uint8(header.UDPProtocolNumber), 0, 0, 64, 0, 0, 0, 1}),
+
+ ipv6Payload1Addr1ToAddr2[64:],
+ },
+ ),
+ },
+ },
+ expectedPayloads: [][]byte{udpPayload1Addr1ToAddr2},
+ },
+ {
+ name: "Two fragments with last fragment size not a multiple of fragment block size",
+ fragments: []fragmentData{
+ {
+ srcAddr: addr1,
+ dstAddr: addr2,
nextHdr: fragmentExtHdrID,
data: buffer.NewVectorisedView(
fragmentExtHdrLen+64,
@@ -763,31 +842,73 @@ func TestReceiveIPv6Fragments(t *testing.T) {
// Fragment offset = 0, More = true, ID = 1
buffer.View([]byte{uint8(header.UDPProtocolNumber), 0, 0, 1, 0, 0, 0, 1}),
- ipv6Payload1[:64],
+ ipv6Payload3Addr1ToAddr2[:64],
},
),
},
{
+ srcAddr: addr1,
+ dstAddr: addr2,
nextHdr: fragmentExtHdrID,
data: buffer.NewVectorisedView(
- fragmentExtHdrLen+len(ipv6Payload1)-64,
+ fragmentExtHdrLen+len(ipv6Payload3Addr1ToAddr2)-64,
[]buffer.View{
// Fragment extension header.
//
// Fragment offset = 8, More = false, ID = 1
buffer.View([]byte{uint8(header.UDPProtocolNumber), 0, 0, 64, 0, 0, 0, 1}),
- ipv6Payload1[64:],
+ ipv6Payload3Addr1ToAddr2[64:],
},
),
},
},
- expectedPayloads: [][]byte{udpPayload1},
+ expectedPayloads: [][]byte{udpPayload3Addr1ToAddr2},
+ },
+ {
+ name: "Two fragments with first fragment size not a multiple of fragment block size",
+ fragments: []fragmentData{
+ {
+ srcAddr: addr1,
+ dstAddr: addr2,
+ nextHdr: fragmentExtHdrID,
+ data: buffer.NewVectorisedView(
+ fragmentExtHdrLen+63,
+ []buffer.View{
+ // Fragment extension header.
+ //
+ // Fragment offset = 0, More = true, ID = 1
+ buffer.View([]byte{uint8(header.UDPProtocolNumber), 0, 0, 1, 0, 0, 0, 1}),
+
+ ipv6Payload3Addr1ToAddr2[:63],
+ },
+ ),
+ },
+ {
+ srcAddr: addr1,
+ dstAddr: addr2,
+ nextHdr: fragmentExtHdrID,
+ data: buffer.NewVectorisedView(
+ fragmentExtHdrLen+len(ipv6Payload3Addr1ToAddr2)-63,
+ []buffer.View{
+ // Fragment extension header.
+ //
+ // Fragment offset = 8, More = false, ID = 1
+ buffer.View([]byte{uint8(header.UDPProtocolNumber), 0, 0, 64, 0, 0, 0, 1}),
+
+ ipv6Payload3Addr1ToAddr2[63:],
+ },
+ ),
+ },
+ },
+ expectedPayloads: nil,
},
{
name: "Two fragments with different IDs",
fragments: []fragmentData{
{
+ srcAddr: addr1,
+ dstAddr: addr2,
nextHdr: fragmentExtHdrID,
data: buffer.NewVectorisedView(
fragmentExtHdrLen+64,
@@ -797,21 +918,23 @@ func TestReceiveIPv6Fragments(t *testing.T) {
// Fragment offset = 0, More = true, ID = 1
buffer.View([]byte{uint8(header.UDPProtocolNumber), 0, 0, 1, 0, 0, 0, 1}),
- ipv6Payload1[:64],
+ ipv6Payload1Addr1ToAddr2[:64],
},
),
},
{
+ srcAddr: addr1,
+ dstAddr: addr2,
nextHdr: fragmentExtHdrID,
data: buffer.NewVectorisedView(
- fragmentExtHdrLen+len(ipv6Payload1)-64,
+ fragmentExtHdrLen+len(ipv6Payload1Addr1ToAddr2)-64,
[]buffer.View{
// Fragment extension header.
//
// Fragment offset = 8, More = false, ID = 2
buffer.View([]byte{uint8(header.UDPProtocolNumber), 0, 0, 64, 0, 0, 0, 2}),
- ipv6Payload1[64:],
+ ipv6Payload1Addr1ToAddr2[64:],
},
),
},
@@ -822,6 +945,8 @@ func TestReceiveIPv6Fragments(t *testing.T) {
name: "Two fragments with per-fragment routing header with zero segments left",
fragments: []fragmentData{
{
+ srcAddr: addr1,
+ dstAddr: addr2,
nextHdr: routingExtHdrID,
data: buffer.NewVectorisedView(
routingExtHdrLen+fragmentExtHdrLen+64,
@@ -836,14 +961,16 @@ func TestReceiveIPv6Fragments(t *testing.T) {
// Fragment offset = 0, More = true, ID = 1
buffer.View([]byte{uint8(header.UDPProtocolNumber), 0, 0, 1, 0, 0, 0, 1}),
- ipv6Payload1[:64],
+ ipv6Payload1Addr1ToAddr2[:64],
},
),
},
{
+ srcAddr: addr1,
+ dstAddr: addr2,
nextHdr: routingExtHdrID,
data: buffer.NewVectorisedView(
- routingExtHdrLen+fragmentExtHdrLen+len(ipv6Payload1)-64,
+ routingExtHdrLen+fragmentExtHdrLen+len(ipv6Payload1Addr1ToAddr2)-64,
[]buffer.View{
// Routing extension header.
//
@@ -855,17 +982,19 @@ func TestReceiveIPv6Fragments(t *testing.T) {
// Fragment offset = 8, More = false, ID = 1
buffer.View([]byte{uint8(header.UDPProtocolNumber), 0, 0, 64, 0, 0, 0, 1}),
- ipv6Payload1[64:],
+ ipv6Payload1Addr1ToAddr2[64:],
},
),
},
},
- expectedPayloads: [][]byte{udpPayload1},
+ expectedPayloads: [][]byte{udpPayload1Addr1ToAddr2},
},
{
name: "Two fragments with per-fragment routing header with non-zero segments left",
fragments: []fragmentData{
{
+ srcAddr: addr1,
+ dstAddr: addr2,
nextHdr: routingExtHdrID,
data: buffer.NewVectorisedView(
routingExtHdrLen+fragmentExtHdrLen+64,
@@ -880,14 +1009,16 @@ func TestReceiveIPv6Fragments(t *testing.T) {
// Fragment offset = 0, More = true, ID = 1
buffer.View([]byte{uint8(header.UDPProtocolNumber), 0, 0, 1, 0, 0, 0, 1}),
- ipv6Payload1[:64],
+ ipv6Payload1Addr1ToAddr2[:64],
},
),
},
{
+ srcAddr: addr1,
+ dstAddr: addr2,
nextHdr: routingExtHdrID,
data: buffer.NewVectorisedView(
- routingExtHdrLen+fragmentExtHdrLen+len(ipv6Payload1)-64,
+ routingExtHdrLen+fragmentExtHdrLen+len(ipv6Payload1Addr1ToAddr2)-64,
[]buffer.View{
// Routing extension header.
//
@@ -899,7 +1030,7 @@ func TestReceiveIPv6Fragments(t *testing.T) {
// Fragment offset = 9, More = false, ID = 1
buffer.View([]byte{uint8(header.UDPProtocolNumber), 0, 0, 72, 0, 0, 0, 1}),
- ipv6Payload1[64:],
+ ipv6Payload1Addr1ToAddr2[64:],
},
),
},
@@ -910,6 +1041,8 @@ func TestReceiveIPv6Fragments(t *testing.T) {
name: "Two fragments with routing header with zero segments left",
fragments: []fragmentData{
{
+ srcAddr: addr1,
+ dstAddr: addr2,
nextHdr: fragmentExtHdrID,
data: buffer.NewVectorisedView(
routingExtHdrLen+fragmentExtHdrLen+64,
@@ -924,31 +1057,35 @@ func TestReceiveIPv6Fragments(t *testing.T) {
// Segments left = 0.
buffer.View([]byte{uint8(header.UDPProtocolNumber), 0, 1, 0, 2, 3, 4, 5}),
- ipv6Payload1[:64],
+ ipv6Payload1Addr1ToAddr2[:64],
},
),
},
{
+ srcAddr: addr1,
+ dstAddr: addr2,
nextHdr: fragmentExtHdrID,
data: buffer.NewVectorisedView(
- fragmentExtHdrLen+len(ipv6Payload1)-64,
+ fragmentExtHdrLen+len(ipv6Payload1Addr1ToAddr2)-64,
[]buffer.View{
// Fragment extension header.
//
// Fragment offset = 9, More = false, ID = 1
buffer.View([]byte{routingExtHdrID, 0, 0, 72, 0, 0, 0, 1}),
- ipv6Payload1[64:],
+ ipv6Payload1Addr1ToAddr2[64:],
},
),
},
},
- expectedPayloads: [][]byte{udpPayload1},
+ expectedPayloads: [][]byte{udpPayload1Addr1ToAddr2},
},
{
name: "Two fragments with routing header with non-zero segments left",
fragments: []fragmentData{
{
+ srcAddr: addr1,
+ dstAddr: addr2,
nextHdr: fragmentExtHdrID,
data: buffer.NewVectorisedView(
routingExtHdrLen+fragmentExtHdrLen+64,
@@ -963,21 +1100,23 @@ func TestReceiveIPv6Fragments(t *testing.T) {
// Segments left = 1.
buffer.View([]byte{uint8(header.UDPProtocolNumber), 0, 1, 1, 2, 3, 4, 5}),
- ipv6Payload1[:64],
+ ipv6Payload1Addr1ToAddr2[:64],
},
),
},
{
+ srcAddr: addr1,
+ dstAddr: addr2,
nextHdr: fragmentExtHdrID,
data: buffer.NewVectorisedView(
- fragmentExtHdrLen+len(ipv6Payload1)-64,
+ fragmentExtHdrLen+len(ipv6Payload1Addr1ToAddr2)-64,
[]buffer.View{
// Fragment extension header.
//
// Fragment offset = 9, More = false, ID = 1
buffer.View([]byte{routingExtHdrID, 0, 0, 72, 0, 0, 0, 1}),
- ipv6Payload1[64:],
+ ipv6Payload1Addr1ToAddr2[64:],
},
),
},
@@ -988,6 +1127,8 @@ func TestReceiveIPv6Fragments(t *testing.T) {
name: "Two fragments with routing header with zero segments left across fragments",
fragments: []fragmentData{
{
+ srcAddr: addr1,
+ dstAddr: addr2,
nextHdr: fragmentExtHdrID,
data: buffer.NewVectorisedView(
// The length of this payload is fragmentExtHdrLen+8 because the
@@ -1008,12 +1149,14 @@ func TestReceiveIPv6Fragments(t *testing.T) {
),
},
{
+ srcAddr: addr1,
+ dstAddr: addr2,
nextHdr: fragmentExtHdrID,
data: buffer.NewVectorisedView(
// The length of this payload is
- // fragmentExtHdrLen+8+len(ipv6Payload1) because the last 8 bytes of
+ // fragmentExtHdrLen+8+len(ipv6Payload1Addr1ToAddr2) because the last 8 bytes of
// the 16 byte routing extension header is in this fagment.
- fragmentExtHdrLen+8+len(ipv6Payload1),
+ fragmentExtHdrLen+8+len(ipv6Payload1Addr1ToAddr2),
[]buffer.View{
// Fragment extension header.
//
@@ -1023,7 +1166,7 @@ func TestReceiveIPv6Fragments(t *testing.T) {
// Routing extension header (part 2)
buffer.View([]byte{6, 7, 8, 9, 10, 11, 12, 13}),
- ipv6Payload1,
+ ipv6Payload1Addr1ToAddr2,
},
),
},
@@ -1034,6 +1177,8 @@ func TestReceiveIPv6Fragments(t *testing.T) {
name: "Two fragments with routing header with non-zero segments left across fragments",
fragments: []fragmentData{
{
+ srcAddr: addr1,
+ dstAddr: addr2,
nextHdr: fragmentExtHdrID,
data: buffer.NewVectorisedView(
// The length of this payload is fragmentExtHdrLen+8 because the
@@ -1054,12 +1199,14 @@ func TestReceiveIPv6Fragments(t *testing.T) {
),
},
{
+ srcAddr: addr1,
+ dstAddr: addr2,
nextHdr: fragmentExtHdrID,
data: buffer.NewVectorisedView(
// The length of this payload is
- // fragmentExtHdrLen+8+len(ipv6Payload1) because the last 8 bytes of
+ // fragmentExtHdrLen+8+len(ipv6Payload1Addr1ToAddr2) because the last 8 bytes of
// the 16 byte routing extension header is in this fagment.
- fragmentExtHdrLen+8+len(ipv6Payload1),
+ fragmentExtHdrLen+8+len(ipv6Payload1Addr1ToAddr2),
[]buffer.View{
// Fragment extension header.
//
@@ -1069,7 +1216,7 @@ func TestReceiveIPv6Fragments(t *testing.T) {
// Routing extension header (part 2)
buffer.View([]byte{6, 7, 8, 9, 10, 11, 12, 13}),
- ipv6Payload1,
+ ipv6Payload1Addr1ToAddr2,
},
),
},
@@ -1082,6 +1229,8 @@ func TestReceiveIPv6Fragments(t *testing.T) {
name: "Two fragments with atomic",
fragments: []fragmentData{
{
+ srcAddr: addr1,
+ dstAddr: addr2,
nextHdr: fragmentExtHdrID,
data: buffer.NewVectorisedView(
fragmentExtHdrLen+64,
@@ -1091,47 +1240,53 @@ func TestReceiveIPv6Fragments(t *testing.T) {
// Fragment offset = 0, More = true, ID = 1
buffer.View([]byte{uint8(header.UDPProtocolNumber), 0, 0, 1, 0, 0, 0, 1}),
- ipv6Payload1[:64],
+ ipv6Payload1Addr1ToAddr2[:64],
},
),
},
// This fragment has the same ID as the other fragments but is an atomic
// fragment. It should not interfere with the other fragments.
{
+ srcAddr: addr1,
+ dstAddr: addr2,
nextHdr: fragmentExtHdrID,
data: buffer.NewVectorisedView(
- fragmentExtHdrLen+len(ipv6Payload2),
+ fragmentExtHdrLen+len(ipv6Payload2Addr1ToAddr2),
[]buffer.View{
// Fragment extension header.
//
// Fragment offset = 0, More = false, ID = 1
buffer.View([]byte{uint8(header.UDPProtocolNumber), 0, 0, 0, 0, 0, 0, 1}),
- ipv6Payload2,
+ ipv6Payload2Addr1ToAddr2,
},
),
},
{
+ srcAddr: addr1,
+ dstAddr: addr2,
nextHdr: fragmentExtHdrID,
data: buffer.NewVectorisedView(
- fragmentExtHdrLen+len(ipv6Payload1)-64,
+ fragmentExtHdrLen+len(ipv6Payload1Addr1ToAddr2)-64,
[]buffer.View{
// Fragment extension header.
//
// Fragment offset = 8, More = false, ID = 1
buffer.View([]byte{uint8(header.UDPProtocolNumber), 0, 0, 64, 0, 0, 0, 1}),
- ipv6Payload1[64:],
+ ipv6Payload1Addr1ToAddr2[64:],
},
),
},
},
- expectedPayloads: [][]byte{udpPayload2, udpPayload1},
+ expectedPayloads: [][]byte{udpPayload2Addr1ToAddr2, udpPayload1Addr1ToAddr2},
},
{
name: "Two interleaved fragmented packets",
fragments: []fragmentData{
{
+ srcAddr: addr1,
+ dstAddr: addr2,
nextHdr: fragmentExtHdrID,
data: buffer.NewVectorisedView(
fragmentExtHdrLen+64,
@@ -1141,11 +1296,13 @@ func TestReceiveIPv6Fragments(t *testing.T) {
// Fragment offset = 0, More = true, ID = 1
buffer.View([]byte{uint8(header.UDPProtocolNumber), 0, 0, 1, 0, 0, 0, 1}),
- ipv6Payload1[:64],
+ ipv6Payload1Addr1ToAddr2[:64],
},
),
},
{
+ srcAddr: addr1,
+ dstAddr: addr2,
nextHdr: fragmentExtHdrID,
data: buffer.NewVectorisedView(
fragmentExtHdrLen+32,
@@ -1155,40 +1312,114 @@ func TestReceiveIPv6Fragments(t *testing.T) {
// Fragment offset = 0, More = true, ID = 2
buffer.View([]byte{uint8(header.UDPProtocolNumber), 0, 0, 1, 0, 0, 0, 2}),
- ipv6Payload2[:32],
+ ipv6Payload2Addr1ToAddr2[:32],
},
),
},
{
+ srcAddr: addr1,
+ dstAddr: addr2,
nextHdr: fragmentExtHdrID,
data: buffer.NewVectorisedView(
- fragmentExtHdrLen+len(ipv6Payload1)-64,
+ fragmentExtHdrLen+len(ipv6Payload1Addr1ToAddr2)-64,
[]buffer.View{
// Fragment extension header.
//
// Fragment offset = 8, More = false, ID = 1
buffer.View([]byte{uint8(header.UDPProtocolNumber), 0, 0, 64, 0, 0, 0, 1}),
- ipv6Payload1[64:],
+ ipv6Payload1Addr1ToAddr2[64:],
},
),
},
{
+ srcAddr: addr1,
+ dstAddr: addr2,
nextHdr: fragmentExtHdrID,
data: buffer.NewVectorisedView(
- fragmentExtHdrLen+len(ipv6Payload2)-32,
+ fragmentExtHdrLen+len(ipv6Payload2Addr1ToAddr2)-32,
[]buffer.View{
// Fragment extension header.
//
// Fragment offset = 4, More = false, ID = 2
buffer.View([]byte{uint8(header.UDPProtocolNumber), 0, 0, 32, 0, 0, 0, 2}),
- ipv6Payload2[32:],
+ ipv6Payload2Addr1ToAddr2[32:],
+ },
+ ),
+ },
+ },
+ expectedPayloads: [][]byte{udpPayload1Addr1ToAddr2, udpPayload2Addr1ToAddr2},
+ },
+ {
+ name: "Two interleaved fragmented packets from different sources but with same ID",
+ fragments: []fragmentData{
+ {
+ srcAddr: addr1,
+ dstAddr: addr2,
+ nextHdr: fragmentExtHdrID,
+ data: buffer.NewVectorisedView(
+ fragmentExtHdrLen+64,
+ []buffer.View{
+ // Fragment extension header.
+ //
+ // Fragment offset = 0, More = true, ID = 1
+ buffer.View([]byte{uint8(header.UDPProtocolNumber), 0, 0, 1, 0, 0, 0, 1}),
+
+ ipv6Payload1Addr1ToAddr2[:64],
+ },
+ ),
+ },
+ {
+ srcAddr: addr3,
+ dstAddr: addr2,
+ nextHdr: fragmentExtHdrID,
+ data: buffer.NewVectorisedView(
+ fragmentExtHdrLen+32,
+ []buffer.View{
+ // Fragment extension header.
+ //
+ // Fragment offset = 0, More = true, ID = 1
+ buffer.View([]byte{uint8(header.UDPProtocolNumber), 0, 0, 1, 0, 0, 0, 1}),
+
+ ipv6Payload1Addr3ToAddr2[:32],
+ },
+ ),
+ },
+ {
+ srcAddr: addr1,
+ dstAddr: addr2,
+ nextHdr: fragmentExtHdrID,
+ data: buffer.NewVectorisedView(
+ fragmentExtHdrLen+len(ipv6Payload1Addr1ToAddr2)-64,
+ []buffer.View{
+ // Fragment extension header.
+ //
+ // Fragment offset = 8, More = false, ID = 1
+ buffer.View([]byte{uint8(header.UDPProtocolNumber), 0, 0, 64, 0, 0, 0, 1}),
+
+ ipv6Payload1Addr1ToAddr2[64:],
+ },
+ ),
+ },
+ {
+ srcAddr: addr3,
+ dstAddr: addr2,
+ nextHdr: fragmentExtHdrID,
+ data: buffer.NewVectorisedView(
+ fragmentExtHdrLen+len(ipv6Payload1Addr1ToAddr2)-32,
+ []buffer.View{
+ // Fragment extension header.
+ //
+ // Fragment offset = 4, More = false, ID = 1
+ buffer.View([]byte{uint8(header.UDPProtocolNumber), 0, 0, 32, 0, 0, 0, 1}),
+
+ ipv6Payload1Addr3ToAddr2[32:],
},
),
},
},
- expectedPayloads: [][]byte{udpPayload1, udpPayload2},
+ expectedPayloads: [][]byte{udpPayload1Addr1ToAddr2, udpPayload1Addr3ToAddr2},
},
}
@@ -1231,8 +1462,8 @@ func TestReceiveIPv6Fragments(t *testing.T) {
PayloadLength: uint16(f.data.Size()),
NextHeader: f.nextHdr,
HopLimit: 255,
- SrcAddr: addr1,
- DstAddr: addr2,
+ SrcAddr: f.srcAddr,
+ DstAddr: f.dstAddr,
})
vv := hdr.View().ToVectorisedView()
diff --git a/pkg/tcpip/stack/BUILD b/pkg/tcpip/stack/BUILD
index 6b9a6b316..1c58bed2d 100644
--- a/pkg/tcpip/stack/BUILD
+++ b/pkg/tcpip/stack/BUILD
@@ -16,6 +16,18 @@ go_template_instance(
)
go_template_instance(
+ name = "neighbor_entry_list",
+ out = "neighbor_entry_list.go",
+ package = "stack",
+ prefix = "neighborEntry",
+ template = "//pkg/ilist:generic_list",
+ types = {
+ "Element": "*neighborEntry",
+ "Linker": "*neighborEntry",
+ },
+)
+
+go_template_instance(
name = "packet_buffer_list",
out = "packet_buffer_list.go",
package = "stack",
@@ -53,7 +65,12 @@ go_library(
"linkaddrcache.go",
"linkaddrentry_list.go",
"ndp.go",
+ "neighbor_cache.go",
+ "neighbor_entry.go",
+ "neighbor_entry_list.go",
+ "neighborstate_string.go",
"nic.go",
+ "nud.go",
"packet_buffer.go",
"packet_buffer_list.go",
"rand.go",
@@ -89,6 +106,7 @@ go_test(
size = "medium",
srcs = [
"ndp_test.go",
+ "nud_test.go",
"stack_test.go",
"transport_demuxer_test.go",
"transport_test.go",
@@ -110,6 +128,7 @@ go_test(
"//pkg/tcpip/transport/udp",
"//pkg/waiter",
"@com_github_google_go_cmp//cmp:go_default_library",
+ "@com_github_google_go_cmp//cmp/cmpopts:go_default_library",
],
)
@@ -117,8 +136,11 @@ go_test(
name = "stack_test",
size = "small",
srcs = [
+ "fake_time_test.go",
"forwarder_test.go",
"linkaddrcache_test.go",
+ "neighbor_cache_test.go",
+ "neighbor_entry_test.go",
"nic_test.go",
],
library = ":stack",
@@ -128,5 +150,8 @@ go_test(
"//pkg/tcpip",
"//pkg/tcpip/buffer",
"//pkg/tcpip/header",
+ "@com_github_dpjacques_clockwork//:go_default_library",
+ "@com_github_google_go_cmp//cmp:go_default_library",
+ "@com_github_google_go_cmp//cmp/cmpopts:go_default_library",
],
)
diff --git a/pkg/tcpip/stack/conntrack.go b/pkg/tcpip/stack/conntrack.go
index 559a1c4dd..470c265aa 100644
--- a/pkg/tcpip/stack/conntrack.go
+++ b/pkg/tcpip/stack/conntrack.go
@@ -240,7 +240,10 @@ func (ct *ConnTrack) connFor(pkt *PacketBuffer) (*conn, direction) {
if err != nil {
return nil, dirOriginal
}
+ return ct.connForTID(tid)
+}
+func (ct *ConnTrack) connForTID(tid tupleID) (*conn, direction) {
bucket := ct.bucket(tid)
now := time.Now()
@@ -604,3 +607,26 @@ func (ct *ConnTrack) reapTupleLocked(tuple *tuple, bucket int, now time.Time) bo
return true
}
+
+func (ct *ConnTrack) originalDst(epID TransportEndpointID) (tcpip.Address, uint16, *tcpip.Error) {
+ // Lookup the connection. The reply's original destination
+ // describes the original address.
+ tid := tupleID{
+ srcAddr: epID.LocalAddress,
+ srcPort: epID.LocalPort,
+ dstAddr: epID.RemoteAddress,
+ dstPort: epID.RemotePort,
+ transProto: header.TCPProtocolNumber,
+ netProto: header.IPv4ProtocolNumber,
+ }
+ conn, _ := ct.connForTID(tid)
+ if conn == nil {
+ // Not a tracked connection.
+ return "", 0, tcpip.ErrNotConnected
+ } else if conn.manip == manipNone {
+ // Unmanipulated connection.
+ return "", 0, tcpip.ErrInvalidOptionValue
+ }
+
+ return conn.original.dstAddr, conn.original.dstPort, nil
+}
diff --git a/pkg/tcpip/stack/fake_time_test.go b/pkg/tcpip/stack/fake_time_test.go
new file mode 100644
index 000000000..92c8cb534
--- /dev/null
+++ b/pkg/tcpip/stack/fake_time_test.go
@@ -0,0 +1,209 @@
+// Copyright 2020 The gVisor Authors.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+// http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package stack
+
+import (
+ "container/heap"
+ "sync"
+ "time"
+
+ "github.com/dpjacques/clockwork"
+ "gvisor.dev/gvisor/pkg/tcpip"
+)
+
+type fakeClock struct {
+ clock clockwork.FakeClock
+
+ // mu protects the fields below.
+ mu sync.RWMutex
+
+ // times is min-heap of times. A heap is used for quick retrieval of the next
+ // upcoming time of scheduled work.
+ times *timeHeap
+
+ // waitGroups stores one WaitGroup for all work scheduled to execute at the
+ // same time via AfterFunc. This allows parallel execution of all functions
+ // passed to AfterFunc scheduled for the same time.
+ waitGroups map[time.Time]*sync.WaitGroup
+}
+
+func newFakeClock() *fakeClock {
+ return &fakeClock{
+ clock: clockwork.NewFakeClock(),
+ times: &timeHeap{},
+ waitGroups: make(map[time.Time]*sync.WaitGroup),
+ }
+}
+
+var _ tcpip.Clock = (*fakeClock)(nil)
+
+// NowNanoseconds implements tcpip.Clock.NowNanoseconds.
+func (fc *fakeClock) NowNanoseconds() int64 {
+ return fc.clock.Now().UnixNano()
+}
+
+// NowMonotonic implements tcpip.Clock.NowMonotonic.
+func (fc *fakeClock) NowMonotonic() int64 {
+ return fc.NowNanoseconds()
+}
+
+// AfterFunc implements tcpip.Clock.AfterFunc.
+func (fc *fakeClock) AfterFunc(d time.Duration, f func()) tcpip.Timer {
+ until := fc.clock.Now().Add(d)
+ wg := fc.addWait(until)
+ return &fakeTimer{
+ clock: fc,
+ until: until,
+ timer: fc.clock.AfterFunc(d, func() {
+ defer wg.Done()
+ f()
+ }),
+ }
+}
+
+// addWait adds an additional wait to the WaitGroup for parallel execution of
+// all work scheduled for t. Returns a reference to the WaitGroup modified.
+func (fc *fakeClock) addWait(t time.Time) *sync.WaitGroup {
+ fc.mu.RLock()
+ wg, ok := fc.waitGroups[t]
+ fc.mu.RUnlock()
+
+ if ok {
+ wg.Add(1)
+ return wg
+ }
+
+ fc.mu.Lock()
+ heap.Push(fc.times, t)
+ fc.mu.Unlock()
+
+ wg = &sync.WaitGroup{}
+ wg.Add(1)
+
+ fc.mu.Lock()
+ fc.waitGroups[t] = wg
+ fc.mu.Unlock()
+
+ return wg
+}
+
+// removeWait removes a wait from the WaitGroup for parallel execution of all
+// work scheduled for t.
+func (fc *fakeClock) removeWait(t time.Time) {
+ fc.mu.RLock()
+ defer fc.mu.RUnlock()
+
+ wg := fc.waitGroups[t]
+ wg.Done()
+}
+
+// advance executes all work that have been scheduled to execute within d from
+// the current fake time. Blocks until all work has completed execution.
+func (fc *fakeClock) advance(d time.Duration) {
+ // Block until all the work is done
+ until := fc.clock.Now().Add(d)
+ for {
+ fc.mu.Lock()
+ if fc.times.Len() == 0 {
+ fc.mu.Unlock()
+ return
+ }
+
+ t := heap.Pop(fc.times).(time.Time)
+ if t.After(until) {
+ // No work to do
+ heap.Push(fc.times, t)
+ fc.mu.Unlock()
+ return
+ }
+ fc.mu.Unlock()
+
+ diff := t.Sub(fc.clock.Now())
+ fc.clock.Advance(diff)
+
+ fc.mu.RLock()
+ wg := fc.waitGroups[t]
+ fc.mu.RUnlock()
+
+ wg.Wait()
+
+ fc.mu.Lock()
+ delete(fc.waitGroups, t)
+ fc.mu.Unlock()
+ }
+}
+
+type fakeTimer struct {
+ clock *fakeClock
+ timer clockwork.Timer
+
+ mu sync.RWMutex
+ until time.Time
+}
+
+var _ tcpip.Timer = (*fakeTimer)(nil)
+
+// Reset implements tcpip.Timer.Reset.
+func (ft *fakeTimer) Reset(d time.Duration) {
+ if !ft.timer.Reset(d) {
+ return
+ }
+
+ ft.mu.Lock()
+ defer ft.mu.Unlock()
+
+ ft.clock.removeWait(ft.until)
+ ft.until = ft.clock.clock.Now().Add(d)
+ ft.clock.addWait(ft.until)
+}
+
+// Stop implements tcpip.Timer.Stop.
+func (ft *fakeTimer) Stop() bool {
+ if !ft.timer.Stop() {
+ return false
+ }
+
+ ft.mu.RLock()
+ defer ft.mu.RUnlock()
+
+ ft.clock.removeWait(ft.until)
+ return true
+}
+
+type timeHeap []time.Time
+
+var _ heap.Interface = (*timeHeap)(nil)
+
+func (h timeHeap) Len() int {
+ return len(h)
+}
+
+func (h timeHeap) Less(i, j int) bool {
+ return h[i].Before(h[j])
+}
+
+func (h timeHeap) Swap(i, j int) {
+ h[i], h[j] = h[j], h[i]
+}
+
+func (h *timeHeap) Push(x interface{}) {
+ *h = append(*h, x.(time.Time))
+}
+
+func (h *timeHeap) Pop() interface{} {
+ last := (*h)[len(*h)-1]
+ *h = (*h)[:len(*h)-1]
+ return last
+}
diff --git a/pkg/tcpip/stack/iptables.go b/pkg/tcpip/stack/iptables.go
index cbbae4224..110ba073d 100644
--- a/pkg/tcpip/stack/iptables.go
+++ b/pkg/tcpip/stack/iptables.go
@@ -218,19 +218,16 @@ func (it *IPTables) Check(hook Hook, pkt *PacketBuffer, gso *GSO, r *Route, addr
// Many users never configure iptables. Spare them the cost of rule
// traversal if rules have never been set.
it.mu.RLock()
+ defer it.mu.RUnlock()
if !it.modified {
- it.mu.RUnlock()
return true
}
- it.mu.RUnlock()
// Packets are manipulated only if connection and matching
// NAT rule exists.
shouldTrack := it.connections.handlePacket(pkt, hook, gso, r)
// Go through each table containing the hook.
- it.mu.RLock()
- defer it.mu.RUnlock()
priorities := it.priorities[hook]
for _, tableID := range priorities {
// If handlePacket already NATed the packet, we don't need to
@@ -418,3 +415,9 @@ func (it *IPTables) checkRule(hook Hook, pkt *PacketBuffer, table Table, ruleIdx
// All the matchers matched, so run the target.
return rule.Target.Action(pkt, &it.connections, hook, gso, r, address)
}
+
+// OriginalDst returns the original destination of redirected connections. It
+// returns an error if the connection doesn't exist or isn't redirected.
+func (it *IPTables) OriginalDst(epID TransportEndpointID) (tcpip.Address, uint16, *tcpip.Error) {
+ return it.connections.originalDst(epID)
+}
diff --git a/pkg/tcpip/stack/ndp.go b/pkg/tcpip/stack/ndp.go
index 9dce11a97..5174e639c 100644
--- a/pkg/tcpip/stack/ndp.go
+++ b/pkg/tcpip/stack/ndp.go
@@ -33,12 +33,6 @@ const (
// Default = 1 (from RFC 4862 section 5.1)
defaultDupAddrDetectTransmits = 1
- // defaultRetransmitTimer is the default amount of time to wait between
- // sending NDP Neighbor solicitation messages.
- //
- // Default = 1s (from RFC 4861 section 10).
- defaultRetransmitTimer = time.Second
-
// defaultMaxRtrSolicitations is the default number of Router
// Solicitation messages to send when a NIC becomes enabled.
//
@@ -79,16 +73,6 @@ const (
// Default = true.
defaultAutoGenGlobalAddresses = true
- // minimumRetransmitTimer is the minimum amount of time to wait between
- // sending NDP Neighbor solicitation messages. Note, RFC 4861 does
- // not impose a minimum Retransmit Timer, but we do here to make sure
- // the messages are not sent all at once. We also come to this value
- // because in the RetransmitTimer field of a Router Advertisement, a
- // value of 0 means unspecified, so the smallest valid value is 1.
- // Note, the unit of the RetransmitTimer field in the Router
- // Advertisement is milliseconds.
- minimumRetransmitTimer = time.Millisecond
-
// minimumRtrSolicitationInterval is the minimum amount of time to wait
// between sending Router Solicitation messages. This limit is imposed
// to make sure that Router Solicitation messages are not sent all at
diff --git a/pkg/tcpip/stack/neighbor_cache.go b/pkg/tcpip/stack/neighbor_cache.go
new file mode 100644
index 000000000..1d37716c2
--- /dev/null
+++ b/pkg/tcpip/stack/neighbor_cache.go
@@ -0,0 +1,335 @@
+// Copyright 2020 The gVisor Authors.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+// http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package stack
+
+import (
+ "fmt"
+ "time"
+
+ "gvisor.dev/gvisor/pkg/sleep"
+ "gvisor.dev/gvisor/pkg/sync"
+ "gvisor.dev/gvisor/pkg/tcpip"
+)
+
+const neighborCacheSize = 512 // max entries per interface
+
+// neighborCache maps IP addresses to link addresses. It uses the Least
+// Recently Used (LRU) eviction strategy to implement a bounded cache for
+// dynmically acquired entries. It contains the state machine and configuration
+// for running Neighbor Unreachability Detection (NUD).
+//
+// There are two types of entries in the neighbor cache:
+// 1. Dynamic entries are discovered automatically by neighbor discovery
+// protocols (e.g. ARP, NDP). These protocols will attempt to reconfirm
+// reachability with the device once the entry's state becomes Stale.
+// 2. Static entries are explicitly added by a user and have no expiration.
+// Their state is always Static. The amount of static entries stored in the
+// cache is unbounded.
+//
+// neighborCache implements NUDHandler.
+type neighborCache struct {
+ nic *NIC
+ state *NUDState
+
+ // mu protects the fields below.
+ mu sync.RWMutex
+
+ cache map[tcpip.Address]*neighborEntry
+ dynamic struct {
+ lru neighborEntryList
+
+ // count tracks the amount of dynamic entries in the cache. This is
+ // needed since static entries do not count towards the LRU cache
+ // eviction strategy.
+ count uint16
+ }
+}
+
+var _ NUDHandler = (*neighborCache)(nil)
+
+// getOrCreateEntry retrieves a cache entry associated with addr. The
+// returned entry is always refreshed in the cache (it is reachable via the
+// map, and its place is bumped in LRU).
+//
+// If a matching entry exists in the cache, it is returned. If no matching
+// entry exists and the cache is full, an existing entry is evicted via LRU,
+// reset to state incomplete, and returned. If no matching entry exists and the
+// cache is not full, a new entry with state incomplete is allocated and
+// returned.
+func (n *neighborCache) getOrCreateEntry(remoteAddr, localAddr tcpip.Address, linkRes LinkAddressResolver) *neighborEntry {
+ n.mu.Lock()
+ defer n.mu.Unlock()
+
+ if entry, ok := n.cache[remoteAddr]; ok {
+ entry.mu.RLock()
+ if entry.neigh.State != Static {
+ n.dynamic.lru.Remove(entry)
+ n.dynamic.lru.PushFront(entry)
+ }
+ entry.mu.RUnlock()
+ return entry
+ }
+
+ // The entry that needs to be created must be dynamic since all static
+ // entries are directly added to the cache via addStaticEntry.
+ entry := newNeighborEntry(n.nic, remoteAddr, localAddr, n.state, linkRes)
+ if n.dynamic.count == neighborCacheSize {
+ e := n.dynamic.lru.Back()
+ e.mu.Lock()
+
+ delete(n.cache, e.neigh.Addr)
+ n.dynamic.lru.Remove(e)
+ n.dynamic.count--
+
+ e.dispatchRemoveEventLocked()
+ e.setStateLocked(Unknown)
+ e.notifyWakersLocked()
+ e.mu.Unlock()
+ }
+ n.cache[remoteAddr] = entry
+ n.dynamic.lru.PushFront(entry)
+ n.dynamic.count++
+ return entry
+}
+
+// entry looks up the neighbor cache for translating address to link address
+// (e.g. IP -> MAC). If the LinkEndpoint requests address resolution and there
+// is a LinkAddressResolver registered with the network protocol, the cache
+// attempts to resolve the address and returns ErrWouldBlock. If a Waker is
+// provided, it will be notified when address resolution is complete (success
+// or not).
+//
+// If address resolution is required, ErrNoLinkAddress and a notification
+// channel is returned for the top level caller to block. Channel is closed
+// once address resolution is complete (success or not).
+func (n *neighborCache) entry(remoteAddr, localAddr tcpip.Address, linkRes LinkAddressResolver, w *sleep.Waker) (NeighborEntry, <-chan struct{}, *tcpip.Error) {
+ if linkRes != nil {
+ if linkAddr, ok := linkRes.ResolveStaticAddress(remoteAddr); ok {
+ e := NeighborEntry{
+ Addr: remoteAddr,
+ LocalAddr: localAddr,
+ LinkAddr: linkAddr,
+ State: Static,
+ UpdatedAt: time.Now(),
+ }
+ return e, nil, nil
+ }
+ }
+
+ entry := n.getOrCreateEntry(remoteAddr, localAddr, linkRes)
+ entry.mu.Lock()
+ defer entry.mu.Unlock()
+
+ switch s := entry.neigh.State; s {
+ case Reachable, Static:
+ return entry.neigh, nil, nil
+
+ case Unknown, Incomplete, Stale, Delay, Probe:
+ entry.addWakerLocked(w)
+
+ if entry.done == nil {
+ // Address resolution needs to be initiated.
+ if linkRes == nil {
+ return entry.neigh, nil, tcpip.ErrNoLinkAddress
+ }
+ entry.done = make(chan struct{})
+ }
+
+ entry.handlePacketQueuedLocked()
+ return entry.neigh, entry.done, tcpip.ErrWouldBlock
+
+ case Failed:
+ return entry.neigh, nil, tcpip.ErrNoLinkAddress
+
+ default:
+ panic(fmt.Sprintf("Invalid cache entry state: %s", s))
+ }
+}
+
+// removeWaker removes a waker that has been added when link resolution for
+// addr was requested.
+func (n *neighborCache) removeWaker(addr tcpip.Address, waker *sleep.Waker) {
+ n.mu.Lock()
+ if entry, ok := n.cache[addr]; ok {
+ delete(entry.wakers, waker)
+ }
+ n.mu.Unlock()
+}
+
+// entries returns all entries in the neighbor cache.
+func (n *neighborCache) entries() []NeighborEntry {
+ entries := make([]NeighborEntry, 0, len(n.cache))
+ n.mu.RLock()
+ for _, entry := range n.cache {
+ entry.mu.RLock()
+ entries = append(entries, entry.neigh)
+ entry.mu.RUnlock()
+ }
+ n.mu.RUnlock()
+ return entries
+}
+
+// addStaticEntry adds a static entry to the neighbor cache, mapping an IP
+// address to a link address. If a dynamic entry exists in the neighbor cache
+// with the same address, it will be replaced with this static entry. If a
+// static entry exists with the same address but different link address, it
+// will be updated with the new link address. If a static entry exists with the
+// same address and link address, nothing will happen.
+func (n *neighborCache) addStaticEntry(addr tcpip.Address, linkAddr tcpip.LinkAddress) {
+ n.mu.Lock()
+ defer n.mu.Unlock()
+
+ if entry, ok := n.cache[addr]; ok {
+ entry.mu.Lock()
+ if entry.neigh.State != Static {
+ // Dynamic entry found with the same address.
+ n.dynamic.lru.Remove(entry)
+ n.dynamic.count--
+ } else if entry.neigh.LinkAddr == linkAddr {
+ // Static entry found with the same address and link address.
+ entry.mu.Unlock()
+ return
+ } else {
+ // Static entry found with the same address but different link address.
+ entry.neigh.LinkAddr = linkAddr
+ entry.dispatchChangeEventLocked(entry.neigh.State)
+ entry.mu.Unlock()
+ return
+ }
+
+ // Notify that resolution has been interrupted, just in case the entry was
+ // in the Incomplete or Probe state.
+ entry.dispatchRemoveEventLocked()
+ entry.setStateLocked(Unknown)
+ entry.notifyWakersLocked()
+ entry.mu.Unlock()
+ }
+
+ entry := newStaticNeighborEntry(n.nic, addr, linkAddr, n.state)
+ n.cache[addr] = entry
+}
+
+// removeEntryLocked removes the specified entry from the neighbor cache.
+func (n *neighborCache) removeEntryLocked(entry *neighborEntry) {
+ if entry.neigh.State != Static {
+ n.dynamic.lru.Remove(entry)
+ n.dynamic.count--
+ }
+ if entry.neigh.State != Failed {
+ entry.dispatchRemoveEventLocked()
+ }
+ entry.setStateLocked(Unknown)
+ entry.notifyWakersLocked()
+
+ delete(n.cache, entry.neigh.Addr)
+}
+
+// removeEntry removes a dynamic or static entry by address from the neighbor
+// cache. Returns true if the entry was found and deleted.
+func (n *neighborCache) removeEntry(addr tcpip.Address) bool {
+ n.mu.Lock()
+ defer n.mu.Unlock()
+
+ entry, ok := n.cache[addr]
+ if !ok {
+ return false
+ }
+
+ entry.mu.Lock()
+ defer entry.mu.Unlock()
+
+ n.removeEntryLocked(entry)
+ return true
+}
+
+// clear removes all dynamic and static entries from the neighbor cache.
+func (n *neighborCache) clear() {
+ n.mu.Lock()
+ defer n.mu.Unlock()
+
+ for _, entry := range n.cache {
+ entry.mu.Lock()
+ entry.dispatchRemoveEventLocked()
+ entry.setStateLocked(Unknown)
+ entry.notifyWakersLocked()
+ entry.mu.Unlock()
+ }
+
+ n.dynamic.lru = neighborEntryList{}
+ n.cache = make(map[tcpip.Address]*neighborEntry)
+ n.dynamic.count = 0
+}
+
+// config returns the NUD configuration.
+func (n *neighborCache) config() NUDConfigurations {
+ return n.state.Config()
+}
+
+// setConfig changes the NUD configuration.
+//
+// If config contains invalid NUD configuration values, it will be fixed to
+// use default values for the erroneous values.
+func (n *neighborCache) setConfig(config NUDConfigurations) {
+ config.resetInvalidFields()
+ n.state.SetConfig(config)
+}
+
+// HandleProbe implements NUDHandler.HandleProbe by following the logic defined
+// in RFC 4861 section 7.2.3. Validation of the probe is expected to be handled
+// by the caller.
+func (n *neighborCache) HandleProbe(remoteAddr, localAddr tcpip.Address, protocol tcpip.NetworkProtocolNumber, remoteLinkAddr tcpip.LinkAddress) {
+ entry := n.getOrCreateEntry(remoteAddr, localAddr, nil)
+ entry.mu.Lock()
+ entry.handleProbeLocked(remoteLinkAddr)
+ entry.mu.Unlock()
+}
+
+// HandleConfirmation implements NUDHandler.HandleConfirmation by following the
+// logic defined in RFC 4861 section 7.2.5.
+//
+// TODO(gvisor.dev/issue/2277): To protect against ARP poisoning and other
+// attacks against NDP functions, Secure Neighbor Discovery (SEND) Protocol
+// should be deployed where preventing access to the broadcast segment might
+// not be possible. SEND uses RSA key pairs to produce cryptographically
+// generated addresses, as defined in RFC 3972, Cryptographically Generated
+// Addresses (CGA). This ensures that the claimed source of an NDP message is
+// the owner of the claimed address.
+func (n *neighborCache) HandleConfirmation(addr tcpip.Address, linkAddr tcpip.LinkAddress, flags ReachabilityConfirmationFlags) {
+ n.mu.RLock()
+ entry, ok := n.cache[addr]
+ n.mu.RUnlock()
+ if ok {
+ entry.mu.Lock()
+ entry.handleConfirmationLocked(linkAddr, flags)
+ entry.mu.Unlock()
+ }
+ // The confirmation SHOULD be silently discarded if the recipient did not
+ // initiate any communication with the target. This is indicated if there is
+ // no matching entry for the remote address.
+}
+
+// HandleUpperLevelConfirmation implements
+// NUDHandler.HandleUpperLevelConfirmation by following the logic defined in
+// RFC 4861 section 7.3.1.
+func (n *neighborCache) HandleUpperLevelConfirmation(addr tcpip.Address) {
+ n.mu.RLock()
+ entry, ok := n.cache[addr]
+ n.mu.RUnlock()
+ if ok {
+ entry.mu.Lock()
+ entry.handleUpperLevelConfirmationLocked()
+ entry.mu.Unlock()
+ }
+}
diff --git a/pkg/tcpip/stack/neighbor_cache_test.go b/pkg/tcpip/stack/neighbor_cache_test.go
new file mode 100644
index 000000000..4cb2c9c6b
--- /dev/null
+++ b/pkg/tcpip/stack/neighbor_cache_test.go
@@ -0,0 +1,1752 @@
+// Copyright 2019 The gVisor Authors.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+// http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package stack
+
+import (
+ "bytes"
+ "encoding/binary"
+ "fmt"
+ "math"
+ "math/rand"
+ "strings"
+ "sync"
+ "sync/atomic"
+ "testing"
+ "time"
+
+ "github.com/google/go-cmp/cmp"
+ "github.com/google/go-cmp/cmp/cmpopts"
+ "gvisor.dev/gvisor/pkg/sleep"
+ "gvisor.dev/gvisor/pkg/tcpip"
+)
+
+const (
+ // entryStoreSize is the default number of entries that will be generated and
+ // added to the entry store. This number needs to be larger than the size of
+ // the neighbor cache to give ample opportunity for verifying behavior during
+ // cache overflows. Four times the size of the neighbor cache allows for
+ // three complete cache overflows.
+ entryStoreSize = 4 * neighborCacheSize
+
+ // typicalLatency is the typical latency for an ARP or NDP packet to travel
+ // to a router and back.
+ typicalLatency = time.Millisecond
+
+ // testEntryBroadcastAddr is a special address that indicates a packet should
+ // be sent to all nodes.
+ testEntryBroadcastAddr = tcpip.Address("broadcast")
+
+ // testEntryLocalAddr is the source address of neighbor probes.
+ testEntryLocalAddr = tcpip.Address("local_addr")
+
+ // testEntryBroadcastLinkAddr is a special link address sent back to
+ // multicast neighbor probes.
+ testEntryBroadcastLinkAddr = tcpip.LinkAddress("mac_broadcast")
+
+ // infiniteDuration indicates that a task will not occur in our lifetime.
+ infiniteDuration = time.Duration(math.MaxInt64)
+)
+
+// entryDiffOpts returns the options passed to cmp.Diff to compare neighbor
+// entries. The UpdatedAt field is ignored due to a lack of a deterministic
+// method to predict the time that an event will be dispatched.
+func entryDiffOpts() []cmp.Option {
+ return []cmp.Option{
+ cmpopts.IgnoreFields(NeighborEntry{}, "UpdatedAt"),
+ }
+}
+
+// entryDiffOptsWithSort is like entryDiffOpts but also includes an option to
+// sort slices of entries for cases where ordering must be ignored.
+func entryDiffOptsWithSort() []cmp.Option {
+ return []cmp.Option{
+ cmpopts.IgnoreFields(NeighborEntry{}, "UpdatedAt"),
+ cmpopts.SortSlices(func(a, b NeighborEntry) bool {
+ return strings.Compare(string(a.Addr), string(b.Addr)) < 0
+ }),
+ }
+}
+
+func newTestNeighborCache(nudDisp NUDDispatcher, config NUDConfigurations, clock tcpip.Clock) *neighborCache {
+ config.resetInvalidFields()
+ rng := rand.New(rand.NewSource(time.Now().UnixNano()))
+ return &neighborCache{
+ nic: &NIC{
+ stack: &Stack{
+ clock: clock,
+ nudDisp: nudDisp,
+ },
+ id: 1,
+ },
+ state: NewNUDState(config, rng),
+ cache: make(map[tcpip.Address]*neighborEntry, neighborCacheSize),
+ }
+}
+
+// testEntryStore contains a set of IP to NeighborEntry mappings.
+type testEntryStore struct {
+ mu sync.RWMutex
+ entriesMap map[tcpip.Address]NeighborEntry
+}
+
+func toAddress(i int) tcpip.Address {
+ buf := new(bytes.Buffer)
+ binary.Write(buf, binary.BigEndian, uint8(1))
+ binary.Write(buf, binary.BigEndian, uint8(0))
+ binary.Write(buf, binary.BigEndian, uint16(i))
+ return tcpip.Address(buf.String())
+}
+
+func toLinkAddress(i int) tcpip.LinkAddress {
+ buf := new(bytes.Buffer)
+ binary.Write(buf, binary.BigEndian, uint8(1))
+ binary.Write(buf, binary.BigEndian, uint8(0))
+ binary.Write(buf, binary.BigEndian, uint32(i))
+ return tcpip.LinkAddress(buf.String())
+}
+
+// newTestEntryStore returns a testEntryStore pre-populated with entries.
+func newTestEntryStore() *testEntryStore {
+ store := &testEntryStore{
+ entriesMap: make(map[tcpip.Address]NeighborEntry),
+ }
+ for i := 0; i < entryStoreSize; i++ {
+ addr := toAddress(i)
+ linkAddr := toLinkAddress(i)
+
+ store.entriesMap[addr] = NeighborEntry{
+ Addr: addr,
+ LocalAddr: testEntryLocalAddr,
+ LinkAddr: linkAddr,
+ }
+ }
+ return store
+}
+
+// size returns the number of entries in the store.
+func (s *testEntryStore) size() int {
+ s.mu.RLock()
+ defer s.mu.RUnlock()
+ return len(s.entriesMap)
+}
+
+// entry returns the entry at index i. Returns an empty entry and false if i is
+// out of bounds.
+func (s *testEntryStore) entry(i int) (NeighborEntry, bool) {
+ return s.entryByAddr(toAddress(i))
+}
+
+// entryByAddr returns the entry matching addr for situations when the index is
+// not available. Returns an empty entry and false if no entries match addr.
+func (s *testEntryStore) entryByAddr(addr tcpip.Address) (NeighborEntry, bool) {
+ s.mu.RLock()
+ defer s.mu.RUnlock()
+ entry, ok := s.entriesMap[addr]
+ return entry, ok
+}
+
+// entries returns all entries in the store.
+func (s *testEntryStore) entries() []NeighborEntry {
+ entries := make([]NeighborEntry, 0, len(s.entriesMap))
+ s.mu.RLock()
+ defer s.mu.RUnlock()
+ for i := 0; i < entryStoreSize; i++ {
+ addr := toAddress(i)
+ if entry, ok := s.entriesMap[addr]; ok {
+ entries = append(entries, entry)
+ }
+ }
+ return entries
+}
+
+// set modifies the link addresses of an entry.
+func (s *testEntryStore) set(i int, linkAddr tcpip.LinkAddress) {
+ addr := toAddress(i)
+ s.mu.Lock()
+ defer s.mu.Unlock()
+ if entry, ok := s.entriesMap[addr]; ok {
+ entry.LinkAddr = linkAddr
+ s.entriesMap[addr] = entry
+ }
+}
+
+// testNeighborResolver implements LinkAddressResolver to emulate sending a
+// neighbor probe.
+type testNeighborResolver struct {
+ clock tcpip.Clock
+ neigh *neighborCache
+ entries *testEntryStore
+ delay time.Duration
+ onLinkAddressRequest func()
+}
+
+var _ LinkAddressResolver = (*testNeighborResolver)(nil)
+
+func (r *testNeighborResolver) LinkAddressRequest(addr, localAddr tcpip.Address, linkAddr tcpip.LinkAddress, linkEP LinkEndpoint) *tcpip.Error {
+ // Delay handling the request to emulate network latency.
+ r.clock.AfterFunc(r.delay, func() {
+ r.fakeRequest(addr)
+ })
+
+ // Execute post address resolution action, if available.
+ if f := r.onLinkAddressRequest; f != nil {
+ f()
+ }
+ return nil
+}
+
+// fakeRequest emulates handling a response for a link address request.
+func (r *testNeighborResolver) fakeRequest(addr tcpip.Address) {
+ if entry, ok := r.entries.entryByAddr(addr); ok {
+ r.neigh.HandleConfirmation(addr, entry.LinkAddr, ReachabilityConfirmationFlags{
+ Solicited: true,
+ Override: false,
+ IsRouter: false,
+ })
+ }
+}
+
+func (*testNeighborResolver) ResolveStaticAddress(addr tcpip.Address) (tcpip.LinkAddress, bool) {
+ if addr == testEntryBroadcastAddr {
+ return testEntryBroadcastLinkAddr, true
+ }
+ return "", false
+}
+
+func (*testNeighborResolver) LinkAddressProtocol() tcpip.NetworkProtocolNumber {
+ return 0
+}
+
+type entryEvent struct {
+ nicID tcpip.NICID
+ address tcpip.Address
+ linkAddr tcpip.LinkAddress
+ state NeighborState
+}
+
+func TestNeighborCacheGetConfig(t *testing.T) {
+ nudDisp := testNUDDispatcher{}
+ c := DefaultNUDConfigurations()
+ clock := newFakeClock()
+ neigh := newTestNeighborCache(&nudDisp, c, clock)
+
+ if got, want := neigh.config(), c; got != want {
+ t.Errorf("got neigh.config() = %+v, want = %+v", got, want)
+ }
+
+ // No events should have been dispatched.
+ nudDisp.mu.Lock()
+ defer nudDisp.mu.Unlock()
+ if diff := cmp.Diff(nudDisp.events, []testEntryEventInfo(nil)); diff != "" {
+ t.Errorf("nud dispatcher events mismatch (-got, +want):\n%s", diff)
+ }
+}
+
+func TestNeighborCacheSetConfig(t *testing.T) {
+ nudDisp := testNUDDispatcher{}
+ c := DefaultNUDConfigurations()
+ clock := newFakeClock()
+ neigh := newTestNeighborCache(&nudDisp, c, clock)
+
+ c.MinRandomFactor = 1
+ c.MaxRandomFactor = 1
+ neigh.setConfig(c)
+
+ if got, want := neigh.config(), c; got != want {
+ t.Errorf("got neigh.config() = %+v, want = %+v", got, want)
+ }
+
+ // No events should have been dispatched.
+ nudDisp.mu.Lock()
+ defer nudDisp.mu.Unlock()
+ if diff := cmp.Diff(nudDisp.events, []testEntryEventInfo(nil)); diff != "" {
+ t.Errorf("nud dispatcher events mismatch (-got, +want):\n%s", diff)
+ }
+}
+
+func TestNeighborCacheEntry(t *testing.T) {
+ c := DefaultNUDConfigurations()
+ nudDisp := testNUDDispatcher{}
+ clock := newFakeClock()
+ neigh := newTestNeighborCache(&nudDisp, c, clock)
+ store := newTestEntryStore()
+ linkRes := &testNeighborResolver{
+ clock: clock,
+ neigh: neigh,
+ entries: store,
+ delay: typicalLatency,
+ }
+
+ entry, ok := store.entry(0)
+ if !ok {
+ t.Fatalf("store.entry(0) not found")
+ }
+ _, _, err := neigh.entry(entry.Addr, entry.LocalAddr, linkRes, nil)
+ if err != tcpip.ErrWouldBlock {
+ t.Errorf("got neigh.entry(%s, %s, _, nil) = %v, want = %s", entry.Addr, entry.LocalAddr, err, tcpip.ErrWouldBlock)
+ }
+
+ clock.advance(typicalLatency)
+
+ wantEvents := []testEntryEventInfo{
+ {
+ EventType: entryTestAdded,
+ NICID: 1,
+ Addr: entry.Addr,
+ State: Incomplete,
+ },
+ {
+ EventType: entryTestChanged,
+ NICID: 1,
+ Addr: entry.Addr,
+ LinkAddr: entry.LinkAddr,
+ State: Reachable,
+ },
+ }
+ nudDisp.mu.Lock()
+ diff := cmp.Diff(nudDisp.events, wantEvents, eventDiffOpts()...)
+ nudDisp.events = nil
+ nudDisp.mu.Unlock()
+ if diff != "" {
+ t.Fatalf("nud dispatcher events mismatch (-got, +want):\n%s", diff)
+ }
+
+ if _, _, err := neigh.entry(entry.Addr, entry.LocalAddr, linkRes, nil); err != nil {
+ t.Fatalf("unexpected error from neigh.entry(%s, %s, _, nil): %s", entry.Addr, entry.LocalAddr, err)
+ }
+
+ // No more events should have been dispatched.
+ nudDisp.mu.Lock()
+ defer nudDisp.mu.Unlock()
+ if diff := cmp.Diff(nudDisp.events, []testEntryEventInfo(nil)); diff != "" {
+ t.Errorf("nud dispatcher events mismatch (-got, +want):\n%s", diff)
+ }
+}
+
+// TestNeighborCacheEntryNoLinkAddress verifies calling entry() without a
+// LinkAddressResolver returns ErrNoLinkAddress.
+func TestNeighborCacheEntryNoLinkAddress(t *testing.T) {
+ nudDisp := testNUDDispatcher{}
+ c := DefaultNUDConfigurations()
+ clock := newFakeClock()
+ neigh := newTestNeighborCache(&nudDisp, c, clock)
+ store := newTestEntryStore()
+
+ entry, ok := store.entry(0)
+ if !ok {
+ t.Fatalf("store.entry(0) not found")
+ }
+ _, _, err := neigh.entry(entry.Addr, entry.LocalAddr, nil, nil)
+ if err != tcpip.ErrNoLinkAddress {
+ t.Errorf("got neigh.entry(%s, %s, nil, nil) = %v, want = %s", entry.Addr, entry.LocalAddr, err, tcpip.ErrNoLinkAddress)
+ }
+
+ // No events should have been dispatched.
+ nudDisp.mu.Lock()
+ defer nudDisp.mu.Unlock()
+ if diff := cmp.Diff(nudDisp.events, []testEntryEventInfo(nil)); diff != "" {
+ t.Errorf("nud dispatcher events mismatch (-got, +want):\n%s", diff)
+ }
+}
+
+func TestNeighborCacheRemoveEntry(t *testing.T) {
+ config := DefaultNUDConfigurations()
+
+ nudDisp := testNUDDispatcher{}
+ clock := newFakeClock()
+ neigh := newTestNeighborCache(&nudDisp, config, clock)
+ store := newTestEntryStore()
+ linkRes := &testNeighborResolver{
+ clock: clock,
+ neigh: neigh,
+ entries: store,
+ delay: typicalLatency,
+ }
+
+ entry, ok := store.entry(0)
+ if !ok {
+ t.Fatalf("store.entry(0) not found")
+ }
+ _, _, err := neigh.entry(entry.Addr, entry.LocalAddr, linkRes, nil)
+ if err != tcpip.ErrWouldBlock {
+ t.Errorf("got neigh.entry(%s, %s, _, nil) = %v, want = %s", entry.Addr, entry.LocalAddr, err, tcpip.ErrWouldBlock)
+ }
+
+ clock.advance(typicalLatency)
+
+ wantEvents := []testEntryEventInfo{
+ {
+ EventType: entryTestAdded,
+ NICID: 1,
+ Addr: entry.Addr,
+ State: Incomplete,
+ },
+ {
+ EventType: entryTestChanged,
+ NICID: 1,
+ Addr: entry.Addr,
+ LinkAddr: entry.LinkAddr,
+ State: Reachable,
+ },
+ }
+ nudDisp.mu.Lock()
+ diff := cmp.Diff(nudDisp.events, wantEvents, eventDiffOpts()...)
+ nudDisp.events = nil
+ nudDisp.mu.Unlock()
+ if diff != "" {
+ t.Fatalf("nud dispatcher events mismatch (-got, +want):\n%s", diff)
+ }
+
+ neigh.removeEntry(entry.Addr)
+
+ {
+ wantEvents := []testEntryEventInfo{
+ {
+ EventType: entryTestRemoved,
+ NICID: 1,
+ Addr: entry.Addr,
+ LinkAddr: entry.LinkAddr,
+ State: Reachable,
+ },
+ }
+ nudDisp.mu.Lock()
+ diff := cmp.Diff(nudDisp.events, wantEvents, eventDiffOpts()...)
+ nudDisp.mu.Unlock()
+ if diff != "" {
+ t.Fatalf("nud dispatcher events mismatch (-got, +want):\n%s", diff)
+ }
+ }
+
+ if _, _, err := neigh.entry(entry.Addr, entry.LocalAddr, linkRes, nil); err != tcpip.ErrWouldBlock {
+ t.Errorf("got neigh.entry(%s, %s, _, nil) = %v, want = %s", entry.Addr, entry.LocalAddr, err, tcpip.ErrWouldBlock)
+ }
+}
+
+type testContext struct {
+ clock *fakeClock
+ neigh *neighborCache
+ store *testEntryStore
+ linkRes *testNeighborResolver
+ nudDisp *testNUDDispatcher
+}
+
+func newTestContext(c NUDConfigurations) testContext {
+ nudDisp := &testNUDDispatcher{}
+ clock := newFakeClock()
+ neigh := newTestNeighborCache(nudDisp, c, clock)
+ store := newTestEntryStore()
+ linkRes := &testNeighborResolver{
+ clock: clock,
+ neigh: neigh,
+ entries: store,
+ delay: typicalLatency,
+ }
+
+ return testContext{
+ clock: clock,
+ neigh: neigh,
+ store: store,
+ linkRes: linkRes,
+ nudDisp: nudDisp,
+ }
+}
+
+type overflowOptions struct {
+ startAtEntryIndex int
+ wantStaticEntries []NeighborEntry
+}
+
+func (c *testContext) overflowCache(opts overflowOptions) error {
+ // Fill the neighbor cache to capacity to verify the LRU eviction strategy is
+ // working properly after the entry removal.
+ for i := opts.startAtEntryIndex; i < c.store.size(); i++ {
+ // Add a new entry
+ entry, ok := c.store.entry(i)
+ if !ok {
+ return fmt.Errorf("c.store.entry(%d) not found", i)
+ }
+ if _, _, err := c.neigh.entry(entry.Addr, entry.LocalAddr, c.linkRes, nil); err != tcpip.ErrWouldBlock {
+ return fmt.Errorf("got c.neigh.entry(%s, %s, _, nil) = %v, want = %s", entry.Addr, entry.LocalAddr, err, tcpip.ErrWouldBlock)
+ }
+ c.clock.advance(c.neigh.config().RetransmitTimer)
+
+ var wantEvents []testEntryEventInfo
+
+ // When beyond the full capacity, the cache will evict an entry as per the
+ // LRU eviction strategy. Note that the number of static entries should not
+ // affect the total number of dynamic entries that can be added.
+ if i >= neighborCacheSize+opts.startAtEntryIndex {
+ removedEntry, ok := c.store.entry(i - neighborCacheSize)
+ if !ok {
+ return fmt.Errorf("store.entry(%d) not found", i-neighborCacheSize)
+ }
+ wantEvents = append(wantEvents, testEntryEventInfo{
+ EventType: entryTestRemoved,
+ NICID: 1,
+ Addr: removedEntry.Addr,
+ LinkAddr: removedEntry.LinkAddr,
+ State: Reachable,
+ })
+ }
+
+ wantEvents = append(wantEvents, testEntryEventInfo{
+ EventType: entryTestAdded,
+ NICID: 1,
+ Addr: entry.Addr,
+ State: Incomplete,
+ }, testEntryEventInfo{
+ EventType: entryTestChanged,
+ NICID: 1,
+ Addr: entry.Addr,
+ LinkAddr: entry.LinkAddr,
+ State: Reachable,
+ })
+
+ c.nudDisp.mu.Lock()
+ diff := cmp.Diff(c.nudDisp.events, wantEvents, eventDiffOpts()...)
+ c.nudDisp.events = nil
+ c.nudDisp.mu.Unlock()
+ if diff != "" {
+ return fmt.Errorf("nud dispatcher events mismatch (-got, +want):\n%s", diff)
+ }
+ }
+
+ // Expect to find only the most recent entries. The order of entries reported
+ // by entries() is undeterministic, so entries have to be sorted before
+ // comparison.
+ wantUnsortedEntries := opts.wantStaticEntries
+ for i := c.store.size() - neighborCacheSize; i < c.store.size(); i++ {
+ entry, ok := c.store.entry(i)
+ if !ok {
+ return fmt.Errorf("c.store.entry(%d) not found", i)
+ }
+ wantEntry := NeighborEntry{
+ Addr: entry.Addr,
+ LocalAddr: entry.LocalAddr,
+ LinkAddr: entry.LinkAddr,
+ State: Reachable,
+ }
+ wantUnsortedEntries = append(wantUnsortedEntries, wantEntry)
+ }
+
+ if diff := cmp.Diff(c.neigh.entries(), wantUnsortedEntries, entryDiffOptsWithSort()...); diff != "" {
+ return fmt.Errorf("neighbor entries mismatch (-got, +want):\n%s", diff)
+ }
+
+ // No more events should have been dispatched.
+ c.nudDisp.mu.Lock()
+ defer c.nudDisp.mu.Unlock()
+ if diff := cmp.Diff(c.nudDisp.events, []testEntryEventInfo(nil)); diff != "" {
+ return fmt.Errorf("nud dispatcher events mismatch (-got, +want):\n%s", diff)
+ }
+
+ return nil
+}
+
+// TestNeighborCacheOverflow verifies that the LRU cache eviction strategy
+// respects the dynamic entry count.
+func TestNeighborCacheOverflow(t *testing.T) {
+ config := DefaultNUDConfigurations()
+ // Stay in Reachable so the cache can overflow
+ config.BaseReachableTime = infiniteDuration
+ config.MinRandomFactor = 1
+ config.MaxRandomFactor = 1
+
+ c := newTestContext(config)
+ opts := overflowOptions{
+ startAtEntryIndex: 0,
+ }
+ if err := c.overflowCache(opts); err != nil {
+ t.Errorf("c.overflowCache(%+v): %s", opts, err)
+ }
+}
+
+// TestNeighborCacheRemoveEntryThenOverflow verifies that the LRU cache
+// eviction strategy respects the dynamic entry count when an entry is removed.
+func TestNeighborCacheRemoveEntryThenOverflow(t *testing.T) {
+ config := DefaultNUDConfigurations()
+ // Stay in Reachable so the cache can overflow
+ config.BaseReachableTime = infiniteDuration
+ config.MinRandomFactor = 1
+ config.MaxRandomFactor = 1
+
+ c := newTestContext(config)
+
+ // Add a dynamic entry
+ entry, ok := c.store.entry(0)
+ if !ok {
+ t.Fatalf("c.store.entry(0) not found")
+ }
+ _, _, err := c.neigh.entry(entry.Addr, entry.LocalAddr, c.linkRes, nil)
+ if err != tcpip.ErrWouldBlock {
+ t.Errorf("got c.neigh.entry(%s, %s, _, nil) = %v, want = %s", entry.Addr, entry.LocalAddr, err, tcpip.ErrWouldBlock)
+ }
+ c.clock.advance(c.neigh.config().RetransmitTimer)
+ wantEvents := []testEntryEventInfo{
+ {
+ EventType: entryTestAdded,
+ NICID: 1,
+ Addr: entry.Addr,
+ State: Incomplete,
+ },
+ {
+ EventType: entryTestChanged,
+ NICID: 1,
+ Addr: entry.Addr,
+ LinkAddr: entry.LinkAddr,
+ State: Reachable,
+ },
+ }
+ c.nudDisp.mu.Lock()
+ diff := cmp.Diff(c.nudDisp.events, wantEvents, eventDiffOpts()...)
+ c.nudDisp.events = nil
+ c.nudDisp.mu.Unlock()
+ if diff != "" {
+ t.Fatalf("nud dispatcher events mismatch (-got, +want):\n%s", diff)
+ }
+
+ // Remove the entry
+ c.neigh.removeEntry(entry.Addr)
+
+ {
+ wantEvents := []testEntryEventInfo{
+ {
+ EventType: entryTestRemoved,
+ NICID: 1,
+ Addr: entry.Addr,
+ LinkAddr: entry.LinkAddr,
+ State: Reachable,
+ },
+ }
+ c.nudDisp.mu.Lock()
+ diff := cmp.Diff(c.nudDisp.events, wantEvents, eventDiffOpts()...)
+ c.nudDisp.events = nil
+ c.nudDisp.mu.Unlock()
+ if diff != "" {
+ t.Fatalf("nud dispatcher events mismatch (-got, +want):\n%s", diff)
+ }
+ }
+
+ opts := overflowOptions{
+ startAtEntryIndex: 0,
+ }
+ if err := c.overflowCache(opts); err != nil {
+ t.Errorf("c.overflowCache(%+v): %s", opts, err)
+ }
+}
+
+// TestNeighborCacheDuplicateStaticEntryWithSameLinkAddress verifies that
+// adding a duplicate static entry with the same link address does not dispatch
+// any events.
+func TestNeighborCacheDuplicateStaticEntryWithSameLinkAddress(t *testing.T) {
+ config := DefaultNUDConfigurations()
+ c := newTestContext(config)
+
+ // Add a static entry
+ entry, ok := c.store.entry(0)
+ if !ok {
+ t.Fatalf("c.store.entry(0) not found")
+ }
+ staticLinkAddr := entry.LinkAddr + "static"
+ c.neigh.addStaticEntry(entry.Addr, staticLinkAddr)
+ wantEvents := []testEntryEventInfo{
+ {
+ EventType: entryTestAdded,
+ NICID: 1,
+ Addr: entry.Addr,
+ LinkAddr: staticLinkAddr,
+ State: Static,
+ },
+ }
+ c.nudDisp.mu.Lock()
+ diff := cmp.Diff(c.nudDisp.events, wantEvents, eventDiffOpts()...)
+ c.nudDisp.events = nil
+ c.nudDisp.mu.Unlock()
+ if diff != "" {
+ t.Fatalf("nud dispatcher events mismatch (-got, +want):\n%s", diff)
+ }
+
+ // Remove the static entry that was just added
+ c.neigh.addStaticEntry(entry.Addr, staticLinkAddr)
+
+ // No more events should have been dispatched.
+ c.nudDisp.mu.Lock()
+ defer c.nudDisp.mu.Unlock()
+ if diff := cmp.Diff(c.nudDisp.events, []testEntryEventInfo(nil)); diff != "" {
+ t.Errorf("nud dispatcher events mismatch (-got, +want):\n%s", diff)
+ }
+}
+
+// TestNeighborCacheDuplicateStaticEntryWithDifferentLinkAddress verifies that
+// adding a duplicate static entry with a different link address dispatches a
+// change event.
+func TestNeighborCacheDuplicateStaticEntryWithDifferentLinkAddress(t *testing.T) {
+ config := DefaultNUDConfigurations()
+ c := newTestContext(config)
+
+ // Add a static entry
+ entry, ok := c.store.entry(0)
+ if !ok {
+ t.Fatalf("c.store.entry(0) not found")
+ }
+ staticLinkAddr := entry.LinkAddr + "static"
+ c.neigh.addStaticEntry(entry.Addr, staticLinkAddr)
+ wantEvents := []testEntryEventInfo{
+ {
+ EventType: entryTestAdded,
+ NICID: 1,
+ Addr: entry.Addr,
+ LinkAddr: staticLinkAddr,
+ State: Static,
+ },
+ }
+ c.nudDisp.mu.Lock()
+ diff := cmp.Diff(c.nudDisp.events, wantEvents, eventDiffOpts()...)
+ c.nudDisp.events = nil
+ c.nudDisp.mu.Unlock()
+ if diff != "" {
+ t.Fatalf("nud dispatcher events mismatch (-got, +want):\n%s", diff)
+ }
+
+ // Add a duplicate entry with a different link address
+ staticLinkAddr += "duplicate"
+ c.neigh.addStaticEntry(entry.Addr, staticLinkAddr)
+ {
+ wantEvents := []testEntryEventInfo{
+ {
+ EventType: entryTestChanged,
+ NICID: 1,
+ Addr: entry.Addr,
+ LinkAddr: staticLinkAddr,
+ State: Static,
+ },
+ }
+ c.nudDisp.mu.Lock()
+ defer c.nudDisp.mu.Unlock()
+ if diff := cmp.Diff(c.nudDisp.events, wantEvents, eventDiffOpts()...); diff != "" {
+ t.Errorf("nud dispatcher events mismatch (-got, +want):\n%s", diff)
+ }
+ }
+}
+
+// TestNeighborCacheRemoveStaticEntryThenOverflow verifies that the LRU cache
+// eviction strategy respects the dynamic entry count when a static entry is
+// added then removed. In this case, the dynamic entry count shouldn't have
+// been touched.
+func TestNeighborCacheRemoveStaticEntryThenOverflow(t *testing.T) {
+ config := DefaultNUDConfigurations()
+ // Stay in Reachable so the cache can overflow
+ config.BaseReachableTime = infiniteDuration
+ config.MinRandomFactor = 1
+ config.MaxRandomFactor = 1
+
+ c := newTestContext(config)
+
+ // Add a static entry
+ entry, ok := c.store.entry(0)
+ if !ok {
+ t.Fatalf("c.store.entry(0) not found")
+ }
+ staticLinkAddr := entry.LinkAddr + "static"
+ c.neigh.addStaticEntry(entry.Addr, staticLinkAddr)
+ wantEvents := []testEntryEventInfo{
+ {
+ EventType: entryTestAdded,
+ NICID: 1,
+ Addr: entry.Addr,
+ LinkAddr: staticLinkAddr,
+ State: Static,
+ },
+ }
+ c.nudDisp.mu.Lock()
+ diff := cmp.Diff(c.nudDisp.events, wantEvents, eventDiffOpts()...)
+ c.nudDisp.events = nil
+ c.nudDisp.mu.Unlock()
+ if diff != "" {
+ t.Fatalf("nud dispatcher events mismatch (-got, +want):\n%s", diff)
+ }
+
+ // Remove the static entry that was just added
+ c.neigh.removeEntry(entry.Addr)
+ {
+ wantEvents := []testEntryEventInfo{
+ {
+ EventType: entryTestRemoved,
+ NICID: 1,
+ Addr: entry.Addr,
+ LinkAddr: staticLinkAddr,
+ State: Static,
+ },
+ }
+ c.nudDisp.mu.Lock()
+ diff := cmp.Diff(c.nudDisp.events, wantEvents, eventDiffOpts()...)
+ c.nudDisp.events = nil
+ c.nudDisp.mu.Unlock()
+ if diff != "" {
+ t.Fatalf("nud dispatcher events mismatch (-got, +want):\n%s", diff)
+ }
+ }
+
+ opts := overflowOptions{
+ startAtEntryIndex: 0,
+ }
+ if err := c.overflowCache(opts); err != nil {
+ t.Errorf("c.overflowCache(%+v): %s", opts, err)
+ }
+}
+
+// TestNeighborCacheOverwriteWithStaticEntryThenOverflow verifies that the LRU
+// cache eviction strategy keeps count of the dynamic entry count when an entry
+// is overwritten by a static entry. Static entries should not count towards
+// the size of the LRU cache.
+func TestNeighborCacheOverwriteWithStaticEntryThenOverflow(t *testing.T) {
+ config := DefaultNUDConfigurations()
+ // Stay in Reachable so the cache can overflow
+ config.BaseReachableTime = infiniteDuration
+ config.MinRandomFactor = 1
+ config.MaxRandomFactor = 1
+
+ c := newTestContext(config)
+
+ // Add a dynamic entry
+ entry, ok := c.store.entry(0)
+ if !ok {
+ t.Fatalf("c.store.entry(0) not found")
+ }
+ _, _, err := c.neigh.entry(entry.Addr, entry.LocalAddr, c.linkRes, nil)
+ if err != tcpip.ErrWouldBlock {
+ t.Errorf("got c.neigh.entry(%s, %s, _, nil) = %v, want = %s", entry.Addr, entry.LocalAddr, err, tcpip.ErrWouldBlock)
+ }
+ c.clock.advance(typicalLatency)
+ wantEvents := []testEntryEventInfo{
+ {
+ EventType: entryTestAdded,
+ NICID: 1,
+ Addr: entry.Addr,
+ State: Incomplete,
+ },
+ {
+ EventType: entryTestChanged,
+ NICID: 1,
+ Addr: entry.Addr,
+ LinkAddr: entry.LinkAddr,
+ State: Reachable,
+ },
+ }
+ c.nudDisp.mu.Lock()
+ diff := cmp.Diff(c.nudDisp.events, wantEvents, eventDiffOpts()...)
+ c.nudDisp.events = nil
+ c.nudDisp.mu.Unlock()
+ if diff != "" {
+ t.Fatalf("nud dispatcher events mismatch (-got, +want):\n%s", diff)
+ }
+
+ // Override the entry with a static one using the same address
+ staticLinkAddr := entry.LinkAddr + "static"
+ c.neigh.addStaticEntry(entry.Addr, staticLinkAddr)
+ {
+ wantEvents := []testEntryEventInfo{
+ {
+ EventType: entryTestRemoved,
+ NICID: 1,
+ Addr: entry.Addr,
+ LinkAddr: entry.LinkAddr,
+ State: Reachable,
+ },
+ {
+ EventType: entryTestAdded,
+ NICID: 1,
+ Addr: entry.Addr,
+ LinkAddr: staticLinkAddr,
+ State: Static,
+ },
+ }
+ c.nudDisp.mu.Lock()
+ diff := cmp.Diff(c.nudDisp.events, wantEvents, eventDiffOpts()...)
+ c.nudDisp.events = nil
+ c.nudDisp.mu.Unlock()
+ if diff != "" {
+ t.Fatalf("nud dispatcher events mismatch (-got, +want):\n%s", diff)
+ }
+ }
+
+ opts := overflowOptions{
+ startAtEntryIndex: 1,
+ wantStaticEntries: []NeighborEntry{
+ {
+ Addr: entry.Addr,
+ LocalAddr: "", // static entries don't need a local address
+ LinkAddr: staticLinkAddr,
+ State: Static,
+ },
+ },
+ }
+ if err := c.overflowCache(opts); err != nil {
+ t.Errorf("c.overflowCache(%+v): %s", opts, err)
+ }
+}
+
+func TestNeighborCacheNotifiesWaker(t *testing.T) {
+ config := DefaultNUDConfigurations()
+
+ nudDisp := testNUDDispatcher{}
+ clock := newFakeClock()
+ neigh := newTestNeighborCache(&nudDisp, config, clock)
+ store := newTestEntryStore()
+ linkRes := &testNeighborResolver{
+ clock: clock,
+ neigh: neigh,
+ entries: store,
+ delay: typicalLatency,
+ }
+
+ w := sleep.Waker{}
+ s := sleep.Sleeper{}
+ const wakerID = 1
+ s.AddWaker(&w, wakerID)
+
+ entry, ok := store.entry(0)
+ if !ok {
+ t.Fatalf("store.entry(0) not found")
+ }
+ _, doneCh, err := neigh.entry(entry.Addr, entry.LocalAddr, linkRes, &w)
+ if err != tcpip.ErrWouldBlock {
+ t.Fatalf("got neigh.entry(%s, %s, _, _ = %v, want = %s", entry.Addr, entry.LocalAddr, err, tcpip.ErrWouldBlock)
+ }
+ if doneCh == nil {
+ t.Fatalf("expected done channel from neigh.entry(%s, %s, _, _)", entry.Addr, entry.LocalAddr)
+ }
+ clock.advance(typicalLatency)
+
+ select {
+ case <-doneCh:
+ default:
+ t.Fatal("expected notification from done channel")
+ }
+
+ id, ok := s.Fetch(false /* block */)
+ if !ok {
+ t.Errorf("expected waker to be notified after neigh.entry(%s, %s, _, _)", entry.Addr, entry.LocalAddr)
+ }
+ if id != wakerID {
+ t.Errorf("got s.Fetch(false) = %d, want = %d", id, wakerID)
+ }
+
+ wantEvents := []testEntryEventInfo{
+ {
+ EventType: entryTestAdded,
+ NICID: 1,
+ Addr: entry.Addr,
+ State: Incomplete,
+ },
+ {
+ EventType: entryTestChanged,
+ NICID: 1,
+ Addr: entry.Addr,
+ LinkAddr: entry.LinkAddr,
+ State: Reachable,
+ },
+ }
+ nudDisp.mu.Lock()
+ defer nudDisp.mu.Unlock()
+ if diff := cmp.Diff(nudDisp.events, wantEvents, eventDiffOpts()...); diff != "" {
+ t.Errorf("nud dispatcher events mismatch (-got, +want):\n%s", diff)
+ }
+}
+
+func TestNeighborCacheRemoveWaker(t *testing.T) {
+ config := DefaultNUDConfigurations()
+
+ nudDisp := testNUDDispatcher{}
+ clock := newFakeClock()
+ neigh := newTestNeighborCache(&nudDisp, config, clock)
+ store := newTestEntryStore()
+ linkRes := &testNeighborResolver{
+ clock: clock,
+ neigh: neigh,
+ entries: store,
+ delay: typicalLatency,
+ }
+
+ w := sleep.Waker{}
+ s := sleep.Sleeper{}
+ const wakerID = 1
+ s.AddWaker(&w, wakerID)
+
+ entry, ok := store.entry(0)
+ if !ok {
+ t.Fatalf("store.entry(0) not found")
+ }
+ _, doneCh, err := neigh.entry(entry.Addr, entry.LocalAddr, linkRes, &w)
+ if err != tcpip.ErrWouldBlock {
+ t.Fatalf("got neigh.entry(%s, %s, _, _) = %v, want = %s", entry.Addr, entry.LocalAddr, err, tcpip.ErrWouldBlock)
+ }
+ if doneCh == nil {
+ t.Fatalf("expected done channel from neigh.entry(%s, %s, _, _)", entry.Addr, entry.LocalAddr)
+ }
+
+ // Remove the waker before the neighbor cache has the opportunity to send a
+ // notification.
+ neigh.removeWaker(entry.Addr, &w)
+ clock.advance(typicalLatency)
+
+ select {
+ case <-doneCh:
+ default:
+ t.Fatal("expected notification from done channel")
+ }
+
+ if id, ok := s.Fetch(false /* block */); ok {
+ t.Errorf("unexpected notification from waker with id %d", id)
+ }
+
+ wantEvents := []testEntryEventInfo{
+ {
+ EventType: entryTestAdded,
+ NICID: 1,
+ Addr: entry.Addr,
+ State: Incomplete,
+ },
+ {
+ EventType: entryTestChanged,
+ NICID: 1,
+ Addr: entry.Addr,
+ LinkAddr: entry.LinkAddr,
+ State: Reachable,
+ },
+ }
+ nudDisp.mu.Lock()
+ defer nudDisp.mu.Unlock()
+ if diff := cmp.Diff(nudDisp.events, wantEvents, eventDiffOpts()...); diff != "" {
+ t.Errorf("nud dispatcher events mismatch (-got, +want):\n%s", diff)
+ }
+}
+
+func TestNeighborCacheAddStaticEntryThenOverflow(t *testing.T) {
+ config := DefaultNUDConfigurations()
+ // Stay in Reachable so the cache can overflow
+ config.BaseReachableTime = infiniteDuration
+ config.MinRandomFactor = 1
+ config.MaxRandomFactor = 1
+
+ c := newTestContext(config)
+
+ entry, ok := c.store.entry(0)
+ if !ok {
+ t.Fatalf("c.store.entry(0) not found")
+ }
+ c.neigh.addStaticEntry(entry.Addr, entry.LinkAddr)
+ e, _, err := c.neigh.entry(entry.Addr, "", nil, nil)
+ if err != nil {
+ t.Errorf("unexpected error from c.neigh.entry(%s, \"\", nil nil): %s", entry.Addr, err)
+ }
+ want := NeighborEntry{
+ Addr: entry.Addr,
+ LocalAddr: "", // static entries don't need a local address
+ LinkAddr: entry.LinkAddr,
+ State: Static,
+ }
+ if diff := cmp.Diff(e, want, entryDiffOpts()...); diff != "" {
+ t.Errorf("c.neigh.entry(%s, \"\", nil, nil) mismatch (-got, +want):\n%s", entry.Addr, diff)
+ }
+
+ wantEvents := []testEntryEventInfo{
+ {
+ EventType: entryTestAdded,
+ NICID: 1,
+ Addr: entry.Addr,
+ LinkAddr: entry.LinkAddr,
+ State: Static,
+ },
+ }
+ c.nudDisp.mu.Lock()
+ diff := cmp.Diff(c.nudDisp.events, wantEvents, eventDiffOpts()...)
+ c.nudDisp.events = nil
+ c.nudDisp.mu.Unlock()
+ if diff != "" {
+ t.Fatalf("nud dispatcher events mismatch (-got, +want):\n%s", diff)
+ }
+
+ opts := overflowOptions{
+ startAtEntryIndex: 1,
+ wantStaticEntries: []NeighborEntry{
+ {
+ Addr: entry.Addr,
+ LocalAddr: "", // static entries don't need a local address
+ LinkAddr: entry.LinkAddr,
+ State: Static,
+ },
+ },
+ }
+ if err := c.overflowCache(opts); err != nil {
+ t.Errorf("c.overflowCache(%+v): %s", opts, err)
+ }
+}
+
+func TestNeighborCacheClear(t *testing.T) {
+ config := DefaultNUDConfigurations()
+
+ nudDisp := testNUDDispatcher{}
+ clock := newFakeClock()
+ neigh := newTestNeighborCache(&nudDisp, config, clock)
+ store := newTestEntryStore()
+ linkRes := &testNeighborResolver{
+ clock: clock,
+ neigh: neigh,
+ entries: store,
+ delay: typicalLatency,
+ }
+
+ // Add a dynamic entry.
+ entry, ok := store.entry(0)
+ if !ok {
+ t.Fatalf("store.entry(0) not found")
+ }
+ _, _, err := neigh.entry(entry.Addr, entry.LocalAddr, linkRes, nil)
+ if err != tcpip.ErrWouldBlock {
+ t.Errorf("got neigh.entry(%s, %s, _, nil) = %v, want = %s", entry.Addr, entry.LocalAddr, err, tcpip.ErrWouldBlock)
+ }
+ clock.advance(typicalLatency)
+
+ wantEvents := []testEntryEventInfo{
+ {
+ EventType: entryTestAdded,
+ NICID: 1,
+ Addr: entry.Addr,
+ State: Incomplete,
+ },
+ {
+ EventType: entryTestChanged,
+ NICID: 1,
+ Addr: entry.Addr,
+ LinkAddr: entry.LinkAddr,
+ State: Reachable,
+ },
+ }
+ nudDisp.mu.Lock()
+ diff := cmp.Diff(nudDisp.events, wantEvents, eventDiffOpts()...)
+ nudDisp.events = nil
+ nudDisp.mu.Unlock()
+ if diff != "" {
+ t.Fatalf("nud dispatcher events mismatch (-got, +want):\n%s", diff)
+ }
+
+ // Add a static entry.
+ neigh.addStaticEntry(entryTestAddr1, entryTestLinkAddr1)
+
+ {
+ wantEvents := []testEntryEventInfo{
+ {
+ EventType: entryTestAdded,
+ NICID: 1,
+ Addr: entryTestAddr1,
+ LinkAddr: entryTestLinkAddr1,
+ State: Static,
+ },
+ }
+ nudDisp.mu.Lock()
+ diff := cmp.Diff(nudDisp.events, wantEvents, eventDiffOpts()...)
+ nudDisp.events = nil
+ nudDisp.mu.Unlock()
+ if diff != "" {
+ t.Fatalf("nud dispatcher events mismatch (-got, +want):\n%s", diff)
+ }
+ }
+
+ // Clear shoud remove both dynamic and static entries.
+ neigh.clear()
+
+ // Remove events dispatched from clear() have no deterministic order so they
+ // need to be sorted beforehand.
+ wantUnsortedEvents := []testEntryEventInfo{
+ {
+ EventType: entryTestRemoved,
+ NICID: 1,
+ Addr: entry.Addr,
+ LinkAddr: entry.LinkAddr,
+ State: Reachable,
+ },
+ {
+ EventType: entryTestRemoved,
+ NICID: 1,
+ Addr: entryTestAddr1,
+ LinkAddr: entryTestLinkAddr1,
+ State: Static,
+ },
+ }
+ nudDisp.mu.Lock()
+ defer nudDisp.mu.Unlock()
+ if diff := cmp.Diff(nudDisp.events, wantUnsortedEvents, eventDiffOptsWithSort()...); diff != "" {
+ t.Errorf("nud dispatcher events mismatch (-got, +want):\n%s", diff)
+ }
+}
+
+// TestNeighborCacheClearThenOverflow verifies that the LRU cache eviction
+// strategy keeps count of the dynamic entry count when all entries are
+// cleared.
+func TestNeighborCacheClearThenOverflow(t *testing.T) {
+ config := DefaultNUDConfigurations()
+ // Stay in Reachable so the cache can overflow
+ config.BaseReachableTime = infiniteDuration
+ config.MinRandomFactor = 1
+ config.MaxRandomFactor = 1
+
+ c := newTestContext(config)
+
+ // Add a dynamic entry
+ entry, ok := c.store.entry(0)
+ if !ok {
+ t.Fatalf("c.store.entry(0) not found")
+ }
+ _, _, err := c.neigh.entry(entry.Addr, entry.LocalAddr, c.linkRes, nil)
+ if err != tcpip.ErrWouldBlock {
+ t.Errorf("got c.neigh.entry(%s, %s, _, nil) = %v, want = %s", entry.Addr, entry.LocalAddr, err, tcpip.ErrWouldBlock)
+ }
+ c.clock.advance(typicalLatency)
+ wantEvents := []testEntryEventInfo{
+ {
+ EventType: entryTestAdded,
+ NICID: 1,
+ Addr: entry.Addr,
+ State: Incomplete,
+ },
+ {
+ EventType: entryTestChanged,
+ NICID: 1,
+ Addr: entry.Addr,
+ LinkAddr: entry.LinkAddr,
+ State: Reachable,
+ },
+ }
+ c.nudDisp.mu.Lock()
+ diff := cmp.Diff(c.nudDisp.events, wantEvents, eventDiffOpts()...)
+ c.nudDisp.events = nil
+ c.nudDisp.mu.Unlock()
+ if diff != "" {
+ t.Fatalf("nud dispatcher events mismatch (-got, +want):\n%s", diff)
+ }
+
+ // Clear the cache.
+ c.neigh.clear()
+ {
+ wantEvents := []testEntryEventInfo{
+ {
+ EventType: entryTestRemoved,
+ NICID: 1,
+ Addr: entry.Addr,
+ LinkAddr: entry.LinkAddr,
+ State: Reachable,
+ },
+ }
+ c.nudDisp.mu.Lock()
+ diff := cmp.Diff(c.nudDisp.events, wantEvents, eventDiffOpts()...)
+ c.nudDisp.events = nil
+ c.nudDisp.mu.Unlock()
+ if diff != "" {
+ t.Fatalf("nud dispatcher events mismatch (-got, +want):\n%s", diff)
+ }
+ }
+
+ opts := overflowOptions{
+ startAtEntryIndex: 0,
+ }
+ if err := c.overflowCache(opts); err != nil {
+ t.Errorf("c.overflowCache(%+v): %s", opts, err)
+ }
+}
+
+func TestNeighborCacheKeepFrequentlyUsed(t *testing.T) {
+ config := DefaultNUDConfigurations()
+ // Stay in Reachable so the cache can overflow
+ config.BaseReachableTime = infiniteDuration
+ config.MinRandomFactor = 1
+ config.MaxRandomFactor = 1
+
+ nudDisp := testNUDDispatcher{}
+ clock := newFakeClock()
+ neigh := newTestNeighborCache(&nudDisp, config, clock)
+ store := newTestEntryStore()
+ linkRes := &testNeighborResolver{
+ clock: clock,
+ neigh: neigh,
+ entries: store,
+ delay: typicalLatency,
+ }
+
+ frequentlyUsedEntry, ok := store.entry(0)
+ if !ok {
+ t.Fatalf("store.entry(0) not found")
+ }
+
+ // The following logic is very similar to overflowCache, but
+ // periodically refreshes the frequently used entry.
+
+ // Fill the neighbor cache to capacity
+ for i := 0; i < neighborCacheSize; i++ {
+ entry, ok := store.entry(i)
+ if !ok {
+ t.Fatalf("store.entry(%d) not found", i)
+ }
+ _, doneCh, err := neigh.entry(entry.Addr, entry.LocalAddr, linkRes, nil)
+ if err != tcpip.ErrWouldBlock {
+ t.Errorf("got neigh.entry(%s, %s, _, nil) = %v, want = %s", entry.Addr, entry.LocalAddr, err, tcpip.ErrWouldBlock)
+ }
+ clock.advance(typicalLatency)
+ select {
+ case <-doneCh:
+ default:
+ t.Fatalf("expected notification from done channel returned by neigh.entry(%s, %s, _, nil)", entry.Addr, entry.LocalAddr)
+ }
+ wantEvents := []testEntryEventInfo{
+ {
+ EventType: entryTestAdded,
+ NICID: 1,
+ Addr: entry.Addr,
+ State: Incomplete,
+ },
+ {
+ EventType: entryTestChanged,
+ NICID: 1,
+ Addr: entry.Addr,
+ LinkAddr: entry.LinkAddr,
+ State: Reachable,
+ },
+ }
+ nudDisp.mu.Lock()
+ diff := cmp.Diff(nudDisp.events, wantEvents, eventDiffOpts()...)
+ nudDisp.events = nil
+ nudDisp.mu.Unlock()
+ if diff != "" {
+ t.Fatalf("nud dispatcher events mismatch (-got, +want):\n%s", diff)
+ }
+ }
+
+ // Keep adding more entries
+ for i := neighborCacheSize; i < store.size(); i++ {
+ // Periodically refresh the frequently used entry
+ if i%(neighborCacheSize/2) == 0 {
+ _, _, err := neigh.entry(frequentlyUsedEntry.Addr, frequentlyUsedEntry.LocalAddr, linkRes, nil)
+ if err != nil {
+ t.Errorf("unexpected error from neigh.entry(%s, %s, _, nil): %s", frequentlyUsedEntry.Addr, frequentlyUsedEntry.LocalAddr, err)
+ }
+ }
+
+ entry, ok := store.entry(i)
+ if !ok {
+ t.Fatalf("store.entry(%d) not found", i)
+ }
+ _, doneCh, err := neigh.entry(entry.Addr, entry.LocalAddr, linkRes, nil)
+ if err != tcpip.ErrWouldBlock {
+ t.Errorf("got neigh.entry(%s, %s, _, nil) = %v, want = %s", entry.Addr, entry.LocalAddr, err, tcpip.ErrWouldBlock)
+ }
+ clock.advance(typicalLatency)
+ select {
+ case <-doneCh:
+ default:
+ t.Fatalf("expected notification from done channel returned by neigh.entry(%s, %s, _, nil)", entry.Addr, entry.LocalAddr)
+ }
+
+ // An entry should have been removed, as per the LRU eviction strategy
+ removedEntry, ok := store.entry(i - neighborCacheSize + 1)
+ if !ok {
+ t.Fatalf("store.entry(%d) not found", i-neighborCacheSize+1)
+ }
+ wantEvents := []testEntryEventInfo{
+ {
+ EventType: entryTestRemoved,
+ NICID: 1,
+ Addr: removedEntry.Addr,
+ LinkAddr: removedEntry.LinkAddr,
+ State: Reachable,
+ },
+ {
+ EventType: entryTestAdded,
+ NICID: 1,
+ Addr: entry.Addr,
+ State: Incomplete,
+ },
+ {
+ EventType: entryTestChanged,
+ NICID: 1,
+ Addr: entry.Addr,
+ LinkAddr: entry.LinkAddr,
+ State: Reachable,
+ },
+ }
+ nudDisp.mu.Lock()
+ diff := cmp.Diff(nudDisp.events, wantEvents, eventDiffOpts()...)
+ nudDisp.events = nil
+ nudDisp.mu.Unlock()
+ if diff != "" {
+ t.Fatalf("nud dispatcher events mismatch (-got, +want):\n%s", diff)
+ }
+ }
+
+ // Expect to find only the frequently used entry and the most recent entries.
+ // The order of entries reported by entries() is undeterministic, so entries
+ // have to be sorted before comparison.
+ wantUnsortedEntries := []NeighborEntry{
+ {
+ Addr: frequentlyUsedEntry.Addr,
+ LocalAddr: frequentlyUsedEntry.LocalAddr,
+ LinkAddr: frequentlyUsedEntry.LinkAddr,
+ State: Reachable,
+ },
+ }
+
+ for i := store.size() - neighborCacheSize + 1; i < store.size(); i++ {
+ entry, ok := store.entry(i)
+ if !ok {
+ t.Fatalf("store.entry(%d) not found", i)
+ }
+ wantEntry := NeighborEntry{
+ Addr: entry.Addr,
+ LocalAddr: entry.LocalAddr,
+ LinkAddr: entry.LinkAddr,
+ State: Reachable,
+ }
+ wantUnsortedEntries = append(wantUnsortedEntries, wantEntry)
+ }
+
+ if diff := cmp.Diff(neigh.entries(), wantUnsortedEntries, entryDiffOptsWithSort()...); diff != "" {
+ t.Errorf("neighbor entries mismatch (-got, +want):\n%s", diff)
+ }
+
+ // No more events should have been dispatched.
+ nudDisp.mu.Lock()
+ defer nudDisp.mu.Unlock()
+ if diff := cmp.Diff(nudDisp.events, []testEntryEventInfo(nil)); diff != "" {
+ t.Errorf("nud dispatcher events mismatch (-got, +want):\n%s", diff)
+ }
+}
+
+func TestNeighborCacheConcurrent(t *testing.T) {
+ const concurrentProcesses = 16
+
+ config := DefaultNUDConfigurations()
+
+ nudDisp := testNUDDispatcher{}
+ clock := newFakeClock()
+ neigh := newTestNeighborCache(&nudDisp, config, clock)
+ store := newTestEntryStore()
+ linkRes := &testNeighborResolver{
+ clock: clock,
+ neigh: neigh,
+ entries: store,
+ delay: typicalLatency,
+ }
+
+ storeEntries := store.entries()
+ for _, entry := range storeEntries {
+ var wg sync.WaitGroup
+ for r := 0; r < concurrentProcesses; r++ {
+ wg.Add(1)
+ go func(entry NeighborEntry) {
+ defer wg.Done()
+ e, _, err := neigh.entry(entry.Addr, entry.LocalAddr, linkRes, nil)
+ if err != nil && err != tcpip.ErrWouldBlock {
+ t.Errorf("got neigh.entry(%s, %s, _, nil) = (%+v, _, %s), want (_, _, nil) or (_, _, %s)", entry.Addr, entry.LocalAddr, e, err, tcpip.ErrWouldBlock)
+ }
+ }(entry)
+ }
+
+ // Wait for all gorountines to send a request
+ wg.Wait()
+
+ // Process all the requests for a single entry concurrently
+ clock.advance(typicalLatency)
+ }
+
+ // All goroutines add in the same order and add more values than can fit in
+ // the cache. Our eviction strategy requires that the last entries are
+ // present, up to the size of the neighbor cache, and the rest are missing.
+ // The order of entries reported by entries() is undeterministic, so entries
+ // have to be sorted before comparison.
+ var wantUnsortedEntries []NeighborEntry
+ for i := store.size() - neighborCacheSize; i < store.size(); i++ {
+ entry, ok := store.entry(i)
+ if !ok {
+ t.Errorf("store.entry(%d) not found", i)
+ }
+ wantEntry := NeighborEntry{
+ Addr: entry.Addr,
+ LocalAddr: entry.LocalAddr,
+ LinkAddr: entry.LinkAddr,
+ State: Reachable,
+ }
+ wantUnsortedEntries = append(wantUnsortedEntries, wantEntry)
+ }
+
+ if diff := cmp.Diff(neigh.entries(), wantUnsortedEntries, entryDiffOptsWithSort()...); diff != "" {
+ t.Errorf("neighbor entries mismatch (-got, +want):\n%s", diff)
+ }
+}
+
+func TestNeighborCacheReplace(t *testing.T) {
+ config := DefaultNUDConfigurations()
+
+ nudDisp := testNUDDispatcher{}
+ clock := newFakeClock()
+ neigh := newTestNeighborCache(&nudDisp, config, clock)
+ store := newTestEntryStore()
+ linkRes := &testNeighborResolver{
+ clock: clock,
+ neigh: neigh,
+ entries: store,
+ delay: typicalLatency,
+ }
+
+ // Add an entry
+ entry, ok := store.entry(0)
+ if !ok {
+ t.Fatalf("store.entry(0) not found")
+ }
+ _, doneCh, err := neigh.entry(entry.Addr, entry.LocalAddr, linkRes, nil)
+ if err != tcpip.ErrWouldBlock {
+ t.Fatalf("got neigh.entry(%s, %s, _, nil) = %v, want = %s", entry.Addr, entry.LocalAddr, err, tcpip.ErrWouldBlock)
+ }
+ clock.advance(typicalLatency)
+ select {
+ case <-doneCh:
+ default:
+ t.Fatalf("expected notification from done channel returned by neigh.entry(%s, %s, _, nil)", entry.Addr, entry.LocalAddr)
+ }
+
+ // Verify the entry exists
+ e, doneCh, err := neigh.entry(entry.Addr, entry.LocalAddr, linkRes, nil)
+ if err != nil {
+ t.Errorf("unexpected error from neigh.entry(%s, %s, _, nil): %s", entry.Addr, entry.LocalAddr, err)
+ }
+ if doneCh != nil {
+ t.Errorf("unexpected done channel from neigh.entry(%s, %s, _, nil): %v", entry.Addr, entry.LocalAddr, doneCh)
+ }
+ if t.Failed() {
+ t.FailNow()
+ }
+ want := NeighborEntry{
+ Addr: entry.Addr,
+ LocalAddr: entry.LocalAddr,
+ LinkAddr: entry.LinkAddr,
+ State: Reachable,
+ }
+ if diff := cmp.Diff(e, want, entryDiffOpts()...); diff != "" {
+ t.Errorf("neigh.entry(%s, %s, _, nil) mismatch (-got, +want):\n%s", entry.Addr, entry.LinkAddr, diff)
+ }
+
+ // Notify of a link address change
+ var updatedLinkAddr tcpip.LinkAddress
+ {
+ entry, ok := store.entry(1)
+ if !ok {
+ t.Fatalf("store.entry(1) not found")
+ }
+ updatedLinkAddr = entry.LinkAddr
+ }
+ store.set(0, updatedLinkAddr)
+ neigh.HandleConfirmation(entry.Addr, updatedLinkAddr, ReachabilityConfirmationFlags{
+ Solicited: false,
+ Override: true,
+ IsRouter: false,
+ })
+
+ // Requesting the entry again should start address resolution
+ {
+ _, doneCh, err := neigh.entry(entry.Addr, entry.LocalAddr, linkRes, nil)
+ if err != tcpip.ErrWouldBlock {
+ t.Fatalf("got neigh.entry(%s, %s, _, nil) = %v, want = %s", entry.Addr, entry.LocalAddr, err, tcpip.ErrWouldBlock)
+ }
+ clock.advance(config.DelayFirstProbeTime + typicalLatency)
+ select {
+ case <-doneCh:
+ default:
+ t.Fatalf("expected notification from done channel returned by neigh.entry(%s, %s, _, nil)", entry.Addr, entry.LocalAddr)
+ }
+ }
+
+ // Verify the entry's new link address
+ {
+ e, _, err := neigh.entry(entry.Addr, entry.LocalAddr, linkRes, nil)
+ clock.advance(typicalLatency)
+ if err != nil {
+ t.Errorf("unexpected error from neigh.entry(%s, %s, _, nil): %s", entry.Addr, entry.LocalAddr, err)
+ }
+ want = NeighborEntry{
+ Addr: entry.Addr,
+ LocalAddr: entry.LocalAddr,
+ LinkAddr: updatedLinkAddr,
+ State: Reachable,
+ }
+ if diff := cmp.Diff(e, want, entryDiffOpts()...); diff != "" {
+ t.Errorf("neigh.entry(%s, %s, _, nil) mismatch (-got, +want):\n%s", entry.Addr, entry.LocalAddr, diff)
+ }
+ }
+}
+
+func TestNeighborCacheResolutionFailed(t *testing.T) {
+ config := DefaultNUDConfigurations()
+
+ nudDisp := testNUDDispatcher{}
+ clock := newFakeClock()
+ neigh := newTestNeighborCache(&nudDisp, config, clock)
+ store := newTestEntryStore()
+
+ var requestCount uint32
+ linkRes := &testNeighborResolver{
+ clock: clock,
+ neigh: neigh,
+ entries: store,
+ delay: typicalLatency,
+ onLinkAddressRequest: func() {
+ atomic.AddUint32(&requestCount, 1)
+ },
+ }
+
+ // First, sanity check that resolution is working
+ entry, ok := store.entry(0)
+ if !ok {
+ t.Fatalf("store.entry(0) not found")
+ }
+ if _, _, err := neigh.entry(entry.Addr, entry.LocalAddr, linkRes, nil); err != tcpip.ErrWouldBlock {
+ t.Fatalf("got neigh.entry(%s, %s, _, nil) = %v, want = %s", entry.Addr, entry.LocalAddr, err, tcpip.ErrWouldBlock)
+ }
+ clock.advance(typicalLatency)
+ got, _, err := neigh.entry(entry.Addr, entry.LocalAddr, linkRes, nil)
+ if err != nil {
+ t.Fatalf("unexpected error from neigh.entry(%s, %s, _, nil): %s", entry.Addr, entry.LocalAddr, err)
+ }
+ want := NeighborEntry{
+ Addr: entry.Addr,
+ LocalAddr: entry.LocalAddr,
+ LinkAddr: entry.LinkAddr,
+ State: Reachable,
+ }
+ if diff := cmp.Diff(got, want, entryDiffOpts()...); diff != "" {
+ t.Errorf("neigh.entry(%s, %s, _, nil) mismatch (-got, +want):\n%s", entry.Addr, entry.LocalAddr, diff)
+ }
+
+ // Verify that address resolution for an unknown address returns ErrNoLinkAddress
+ before := atomic.LoadUint32(&requestCount)
+
+ entry.Addr += "2"
+ if _, _, err := neigh.entry(entry.Addr, entry.LocalAddr, linkRes, nil); err != tcpip.ErrWouldBlock {
+ t.Fatalf("got neigh.entry(%s, %s, _, nil) = %v, want = %s", entry.Addr, entry.LocalAddr, err, tcpip.ErrWouldBlock)
+ }
+ waitFor := config.DelayFirstProbeTime + typicalLatency*time.Duration(config.MaxMulticastProbes)
+ clock.advance(waitFor)
+ if _, _, err := neigh.entry(entry.Addr, entry.LocalAddr, linkRes, nil); err != tcpip.ErrNoLinkAddress {
+ t.Fatalf("got neigh.entry(%s, %s, _, nil) = %v, want = %s", entry.Addr, entry.LocalAddr, err, tcpip.ErrNoLinkAddress)
+ }
+
+ maxAttempts := neigh.config().MaxUnicastProbes
+ if got, want := atomic.LoadUint32(&requestCount)-before, maxAttempts; got != want {
+ t.Errorf("got link address request count = %d, want = %d", got, want)
+ }
+}
+
+// TestNeighborCacheResolutionTimeout simulates sending MaxMulticastProbes
+// probes and not retrieving a confirmation before the duration defined by
+// MaxMulticastProbes * RetransmitTimer.
+func TestNeighborCacheResolutionTimeout(t *testing.T) {
+ config := DefaultNUDConfigurations()
+ config.RetransmitTimer = time.Millisecond // small enough to cause timeout
+
+ clock := newFakeClock()
+ neigh := newTestNeighborCache(nil, config, clock)
+ store := newTestEntryStore()
+ linkRes := &testNeighborResolver{
+ clock: clock,
+ neigh: neigh,
+ entries: store,
+ delay: time.Minute, // large enough to cause timeout
+ }
+
+ entry, ok := store.entry(0)
+ if !ok {
+ t.Fatalf("store.entry(0) not found")
+ }
+ if _, _, err := neigh.entry(entry.Addr, entry.LocalAddr, linkRes, nil); err != tcpip.ErrWouldBlock {
+ t.Fatalf("got neigh.entry(%s, %s, _, nil) = %v, want = %s", entry.Addr, entry.LocalAddr, err, tcpip.ErrWouldBlock)
+ }
+ waitFor := config.RetransmitTimer * time.Duration(config.MaxMulticastProbes)
+ clock.advance(waitFor)
+ if _, _, err := neigh.entry(entry.Addr, entry.LocalAddr, linkRes, nil); err != tcpip.ErrNoLinkAddress {
+ t.Fatalf("got neigh.entry(%s, %s, _, nil) = %v, want = %s", entry.Addr, entry.LocalAddr, err, tcpip.ErrNoLinkAddress)
+ }
+}
+
+// TestNeighborCacheStaticResolution checks that static link addresses are
+// resolved immediately and don't send resolution requests.
+func TestNeighborCacheStaticResolution(t *testing.T) {
+ config := DefaultNUDConfigurations()
+ clock := newFakeClock()
+ neigh := newTestNeighborCache(nil, config, clock)
+ store := newTestEntryStore()
+ linkRes := &testNeighborResolver{
+ clock: clock,
+ neigh: neigh,
+ entries: store,
+ delay: typicalLatency,
+ }
+
+ got, _, err := neigh.entry(testEntryBroadcastAddr, testEntryLocalAddr, linkRes, nil)
+ if err != nil {
+ t.Fatalf("unexpected error from neigh.entry(%s, %s, _, nil): %s", testEntryBroadcastAddr, testEntryLocalAddr, err)
+ }
+ want := NeighborEntry{
+ Addr: testEntryBroadcastAddr,
+ LocalAddr: testEntryLocalAddr,
+ LinkAddr: testEntryBroadcastLinkAddr,
+ State: Static,
+ }
+ if diff := cmp.Diff(got, want, entryDiffOpts()...); diff != "" {
+ t.Errorf("neigh.entry(%s, %s, _, nil) mismatch (-got, +want):\n%s", testEntryBroadcastAddr, testEntryLocalAddr, diff)
+ }
+}
+
+func BenchmarkCacheClear(b *testing.B) {
+ b.StopTimer()
+ config := DefaultNUDConfigurations()
+ clock := &tcpip.StdClock{}
+ neigh := newTestNeighborCache(nil, config, clock)
+ store := newTestEntryStore()
+ linkRes := &testNeighborResolver{
+ clock: clock,
+ neigh: neigh,
+ entries: store,
+ delay: 0,
+ }
+
+ // Clear for every possible size of the cache
+ for cacheSize := 0; cacheSize < neighborCacheSize; cacheSize++ {
+ // Fill the neighbor cache to capacity.
+ for i := 0; i < cacheSize; i++ {
+ entry, ok := store.entry(i)
+ if !ok {
+ b.Fatalf("store.entry(%d) not found", i)
+ }
+ _, doneCh, err := neigh.entry(entry.Addr, entry.LocalAddr, linkRes, nil)
+ if err != tcpip.ErrWouldBlock {
+ b.Fatalf("got neigh.entry(%s, %s, _, nil) = %v, want = %s", entry.Addr, entry.LocalAddr, err, tcpip.ErrWouldBlock)
+ }
+ if doneCh != nil {
+ <-doneCh
+ }
+ }
+
+ b.StartTimer()
+ neigh.clear()
+ b.StopTimer()
+ }
+}
diff --git a/pkg/tcpip/stack/neighbor_entry.go b/pkg/tcpip/stack/neighbor_entry.go
new file mode 100644
index 000000000..0068cacb8
--- /dev/null
+++ b/pkg/tcpip/stack/neighbor_entry.go
@@ -0,0 +1,482 @@
+// Copyright 2020 The gVisor Authors.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+// http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package stack
+
+import (
+ "fmt"
+ "sync"
+ "time"
+
+ "gvisor.dev/gvisor/pkg/sleep"
+ "gvisor.dev/gvisor/pkg/tcpip"
+)
+
+// NeighborEntry describes a neighboring device in the local network.
+type NeighborEntry struct {
+ Addr tcpip.Address
+ LocalAddr tcpip.Address
+ LinkAddr tcpip.LinkAddress
+ State NeighborState
+ UpdatedAt time.Time
+}
+
+// NeighborState defines the state of a NeighborEntry within the Neighbor
+// Unreachability Detection state machine, as per RFC 4861 section 7.3.2.
+type NeighborState uint8
+
+const (
+ // Unknown means reachability has not been verified yet. This is the initial
+ // state of entries that have been created automatically by the Neighbor
+ // Unreachability Detection state machine.
+ Unknown NeighborState = iota
+ // Incomplete means that there is an outstanding request to resolve the
+ // address.
+ Incomplete
+ // Reachable means the path to the neighbor is functioning properly for both
+ // receive and transmit paths.
+ Reachable
+ // Stale means reachability to the neighbor is unknown, but packets are still
+ // able to be transmitted to the possibly stale link address.
+ Stale
+ // Delay means reachability to the neighbor is unknown and pending
+ // confirmation from an upper-level protocol like TCP, but packets are still
+ // able to be transmitted to the possibly stale link address.
+ Delay
+ // Probe means a reachability confirmation is actively being sought by
+ // periodically retransmitting reachability probes until a reachability
+ // confirmation is received, or until the max amount of probes has been sent.
+ Probe
+ // Static describes entries that have been explicitly added by the user. They
+ // do not expire and are not deleted until explicitly removed.
+ Static
+ // Failed means traffic should not be sent to this neighbor since attempts of
+ // reachability have returned inconclusive.
+ Failed
+)
+
+// neighborEntry implements a neighbor entry's individual node behavior, as per
+// RFC 4861 section 7.3.3. Neighbor Unreachability Detection operates in
+// parallel with the sending of packets to a neighbor, necessitating the
+// entry's lock to be acquired for all operations.
+type neighborEntry struct {
+ neighborEntryEntry
+
+ nic *NIC
+ protocol tcpip.NetworkProtocolNumber
+
+ // linkRes provides the functionality to send reachability probes, used in
+ // Neighbor Unreachability Detection.
+ linkRes LinkAddressResolver
+
+ // nudState points to the Neighbor Unreachability Detection configuration.
+ nudState *NUDState
+
+ // mu protects the fields below.
+ mu sync.RWMutex
+
+ neigh NeighborEntry
+
+ // wakers is a set of waiters for address resolution result. Anytime state
+ // transitions out of incomplete these waiters are notified. It is nil iff
+ // address resolution is ongoing and no clients are waiting for the result.
+ wakers map[*sleep.Waker]struct{}
+
+ // done is used to allow callers to wait on address resolution. It is nil
+ // iff nudState is not Reachable and address resolution is not yet in
+ // progress.
+ done chan struct{}
+
+ isRouter bool
+ job *tcpip.Job
+}
+
+// newNeighborEntry creates a neighbor cache entry starting at the default
+// state, Unknown. Transition out of Unknown by calling either
+// `handlePacketQueuedLocked` or `handleProbeLocked` on the newly created
+// neighborEntry.
+func newNeighborEntry(nic *NIC, remoteAddr tcpip.Address, localAddr tcpip.Address, nudState *NUDState, linkRes LinkAddressResolver) *neighborEntry {
+ return &neighborEntry{
+ nic: nic,
+ linkRes: linkRes,
+ nudState: nudState,
+ neigh: NeighborEntry{
+ Addr: remoteAddr,
+ LocalAddr: localAddr,
+ State: Unknown,
+ },
+ }
+}
+
+// newStaticNeighborEntry creates a neighbor cache entry starting at the Static
+// state. The entry can only transition out of Static by directly calling
+// `setStateLocked`.
+func newStaticNeighborEntry(nic *NIC, addr tcpip.Address, linkAddr tcpip.LinkAddress, state *NUDState) *neighborEntry {
+ if nic.stack.nudDisp != nil {
+ nic.stack.nudDisp.OnNeighborAdded(nic.id, addr, linkAddr, Static, time.Now())
+ }
+ return &neighborEntry{
+ nic: nic,
+ nudState: state,
+ neigh: NeighborEntry{
+ Addr: addr,
+ LinkAddr: linkAddr,
+ State: Static,
+ UpdatedAt: time.Now(),
+ },
+ }
+}
+
+// addWaker adds w to the list of wakers waiting for address resolution.
+// Assumes the entry has already been appropriately locked.
+func (e *neighborEntry) addWakerLocked(w *sleep.Waker) {
+ if w == nil {
+ return
+ }
+ if e.wakers == nil {
+ e.wakers = make(map[*sleep.Waker]struct{})
+ }
+ e.wakers[w] = struct{}{}
+}
+
+// notifyWakersLocked notifies those waiting for address resolution, whether it
+// succeeded or failed. Assumes the entry has already been appropriately locked.
+func (e *neighborEntry) notifyWakersLocked() {
+ for w := range e.wakers {
+ w.Assert()
+ }
+ e.wakers = nil
+ if ch := e.done; ch != nil {
+ close(ch)
+ e.done = nil
+ }
+}
+
+// dispatchAddEventLocked signals to stack's NUD Dispatcher that the entry has
+// been added.
+func (e *neighborEntry) dispatchAddEventLocked(nextState NeighborState) {
+ if nudDisp := e.nic.stack.nudDisp; nudDisp != nil {
+ nudDisp.OnNeighborAdded(e.nic.id, e.neigh.Addr, e.neigh.LinkAddr, nextState, time.Now())
+ }
+}
+
+// dispatchChangeEventLocked signals to stack's NUD Dispatcher that the entry
+// has changed state or link-layer address.
+func (e *neighborEntry) dispatchChangeEventLocked(nextState NeighborState) {
+ if nudDisp := e.nic.stack.nudDisp; nudDisp != nil {
+ nudDisp.OnNeighborChanged(e.nic.id, e.neigh.Addr, e.neigh.LinkAddr, nextState, time.Now())
+ }
+}
+
+// dispatchRemoveEventLocked signals to stack's NUD Dispatcher that the entry
+// has been removed.
+func (e *neighborEntry) dispatchRemoveEventLocked() {
+ if nudDisp := e.nic.stack.nudDisp; nudDisp != nil {
+ nudDisp.OnNeighborRemoved(e.nic.id, e.neigh.Addr, e.neigh.LinkAddr, e.neigh.State, time.Now())
+ }
+}
+
+// setStateLocked transitions the entry to the specified state immediately.
+//
+// Follows the logic defined in RFC 4861 section 7.3.3.
+//
+// e.mu MUST be locked.
+func (e *neighborEntry) setStateLocked(next NeighborState) {
+ // Cancel the previously scheduled action, if there is one. Entries in
+ // Unknown, Stale, or Static state do not have scheduled actions.
+ if timer := e.job; timer != nil {
+ timer.Cancel()
+ }
+
+ prev := e.neigh.State
+ e.neigh.State = next
+ e.neigh.UpdatedAt = time.Now()
+ config := e.nudState.Config()
+
+ switch next {
+ case Incomplete:
+ var retryCounter uint32
+ var sendMulticastProbe func()
+
+ sendMulticastProbe = func() {
+ if retryCounter == config.MaxMulticastProbes {
+ // "If no Neighbor Advertisement is received after
+ // MAX_MULTICAST_SOLICIT solicitations, address resolution has failed.
+ // The sender MUST return ICMP destination unreachable indications with
+ // code 3 (Address Unreachable) for each packet queued awaiting address
+ // resolution." - RFC 4861 section 7.2.2
+ //
+ // There is no need to send an ICMP destination unreachable indication
+ // since the failure to resolve the address is expected to only occur
+ // on this node. Thus, redirecting traffic is currently not supported.
+ //
+ // "If the error occurs on a node other than the node originating the
+ // packet, an ICMP error message is generated. If the error occurs on
+ // the originating node, an implementation is not required to actually
+ // create and send an ICMP error packet to the source, as long as the
+ // upper-layer sender is notified through an appropriate mechanism
+ // (e.g. return value from a procedure call). Note, however, that an
+ // implementation may find it convenient in some cases to return errors
+ // to the sender by taking the offending packet, generating an ICMP
+ // error message, and then delivering it (locally) through the generic
+ // error-handling routines.' - RFC 4861 section 2.1
+ e.dispatchRemoveEventLocked()
+ e.setStateLocked(Failed)
+ return
+ }
+
+ if err := e.linkRes.LinkAddressRequest(e.neigh.Addr, e.neigh.LocalAddr, "", e.nic.linkEP); err != nil {
+ // There is no need to log the error here; the NUD implementation may
+ // assume a working link. A valid link should be the responsibility of
+ // the NIC/stack.LinkEndpoint.
+ e.dispatchRemoveEventLocked()
+ e.setStateLocked(Failed)
+ return
+ }
+
+ retryCounter++
+ e.job = e.nic.stack.newJob(&e.mu, sendMulticastProbe)
+ e.job.Schedule(config.RetransmitTimer)
+ }
+
+ sendMulticastProbe()
+
+ case Reachable:
+ e.job = e.nic.stack.newJob(&e.mu, func() {
+ e.dispatchChangeEventLocked(Stale)
+ e.setStateLocked(Stale)
+ })
+ e.job.Schedule(e.nudState.ReachableTime())
+
+ case Delay:
+ e.job = e.nic.stack.newJob(&e.mu, func() {
+ e.dispatchChangeEventLocked(Probe)
+ e.setStateLocked(Probe)
+ })
+ e.job.Schedule(config.DelayFirstProbeTime)
+
+ case Probe:
+ var retryCounter uint32
+ var sendUnicastProbe func()
+
+ sendUnicastProbe = func() {
+ if retryCounter == config.MaxUnicastProbes {
+ e.dispatchRemoveEventLocked()
+ e.setStateLocked(Failed)
+ return
+ }
+
+ if err := e.linkRes.LinkAddressRequest(e.neigh.Addr, e.neigh.LocalAddr, e.neigh.LinkAddr, e.nic.linkEP); err != nil {
+ e.dispatchRemoveEventLocked()
+ e.setStateLocked(Failed)
+ return
+ }
+
+ retryCounter++
+ if retryCounter == config.MaxUnicastProbes {
+ e.dispatchRemoveEventLocked()
+ e.setStateLocked(Failed)
+ return
+ }
+
+ e.job = e.nic.stack.newJob(&e.mu, sendUnicastProbe)
+ e.job.Schedule(config.RetransmitTimer)
+ }
+
+ sendUnicastProbe()
+
+ case Failed:
+ e.notifyWakersLocked()
+ e.job = e.nic.stack.newJob(&e.mu, func() {
+ e.nic.neigh.removeEntryLocked(e)
+ })
+ e.job.Schedule(config.UnreachableTime)
+
+ case Unknown, Stale, Static:
+ // Do nothing
+
+ default:
+ panic(fmt.Sprintf("Invalid state transition from %q to %q", prev, next))
+ }
+}
+
+// handlePacketQueuedLocked advances the state machine according to a packet
+// being queued for outgoing transmission.
+//
+// Follows the logic defined in RFC 4861 section 7.3.3.
+func (e *neighborEntry) handlePacketQueuedLocked() {
+ switch e.neigh.State {
+ case Unknown:
+ e.dispatchAddEventLocked(Incomplete)
+ e.setStateLocked(Incomplete)
+
+ case Stale:
+ e.dispatchChangeEventLocked(Delay)
+ e.setStateLocked(Delay)
+
+ case Incomplete, Reachable, Delay, Probe, Static, Failed:
+ // Do nothing
+
+ default:
+ panic(fmt.Sprintf("Invalid cache entry state: %s", e.neigh.State))
+ }
+}
+
+// handleProbeLocked processes an incoming neighbor probe (e.g. ARP request or
+// Neighbor Solicitation for ARP or NDP, respectively).
+//
+// Follows the logic defined in RFC 4861 section 7.2.3.
+func (e *neighborEntry) handleProbeLocked(remoteLinkAddr tcpip.LinkAddress) {
+ // Probes MUST be silently discarded if the target address is tentative, does
+ // not exist, or not bound to the NIC as per RFC 4861 section 7.2.3. These
+ // checks MUST be done by the NetworkEndpoint.
+
+ switch e.neigh.State {
+ case Unknown, Incomplete, Failed:
+ e.neigh.LinkAddr = remoteLinkAddr
+ e.dispatchAddEventLocked(Stale)
+ e.setStateLocked(Stale)
+ e.notifyWakersLocked()
+
+ case Reachable, Delay, Probe:
+ if e.neigh.LinkAddr != remoteLinkAddr {
+ e.neigh.LinkAddr = remoteLinkAddr
+ e.dispatchChangeEventLocked(Stale)
+ e.setStateLocked(Stale)
+ }
+
+ case Stale:
+ if e.neigh.LinkAddr != remoteLinkAddr {
+ e.neigh.LinkAddr = remoteLinkAddr
+ e.dispatchChangeEventLocked(Stale)
+ }
+
+ case Static:
+ // Do nothing
+
+ default:
+ panic(fmt.Sprintf("Invalid cache entry state: %s", e.neigh.State))
+ }
+}
+
+// handleConfirmationLocked processes an incoming neighbor confirmation
+// (e.g. ARP reply or Neighbor Advertisement for ARP or NDP, respectively).
+//
+// Follows the state machine defined by RFC 4861 section 7.2.5.
+//
+// TODO(gvisor.dev/issue/2277): To protect against ARP poisoning and other
+// attacks against NDP functions, Secure Neighbor Discovery (SEND) Protocol
+// should be deployed where preventing access to the broadcast segment might
+// not be possible. SEND uses RSA key pairs to produce Cryptographically
+// Generated Addresses (CGA), as defined in RFC 3972. This ensures that the
+// claimed source of an NDP message is the owner of the claimed address.
+func (e *neighborEntry) handleConfirmationLocked(linkAddr tcpip.LinkAddress, flags ReachabilityConfirmationFlags) {
+ switch e.neigh.State {
+ case Incomplete:
+ if len(linkAddr) == 0 {
+ // "If the link layer has addresses and no Target Link-Layer Address
+ // option is included, the receiving node SHOULD silently discard the
+ // received advertisement." - RFC 4861 section 7.2.5
+ break
+ }
+
+ e.neigh.LinkAddr = linkAddr
+ if flags.Solicited {
+ e.dispatchChangeEventLocked(Reachable)
+ e.setStateLocked(Reachable)
+ } else {
+ e.dispatchChangeEventLocked(Stale)
+ e.setStateLocked(Stale)
+ }
+ e.isRouter = flags.IsRouter
+ e.notifyWakersLocked()
+
+ // "Note that the Override flag is ignored if the entry is in the
+ // INCOMPLETE state." - RFC 4861 section 7.2.5
+
+ case Reachable, Stale, Delay, Probe:
+ sameLinkAddr := e.neigh.LinkAddr == linkAddr
+
+ if !sameLinkAddr {
+ if !flags.Override {
+ if e.neigh.State == Reachable {
+ e.dispatchChangeEventLocked(Stale)
+ e.setStateLocked(Stale)
+ }
+ break
+ }
+
+ e.neigh.LinkAddr = linkAddr
+
+ if !flags.Solicited {
+ if e.neigh.State != Stale {
+ e.dispatchChangeEventLocked(Stale)
+ e.setStateLocked(Stale)
+ } else {
+ // Notify the LinkAddr change, even though NUD state hasn't changed.
+ e.dispatchChangeEventLocked(e.neigh.State)
+ }
+ break
+ }
+ }
+
+ if flags.Solicited && (flags.Override || sameLinkAddr) {
+ if e.neigh.State != Reachable {
+ e.dispatchChangeEventLocked(Reachable)
+ }
+ // Set state to Reachable again to refresh timers.
+ e.setStateLocked(Reachable)
+ e.notifyWakersLocked()
+ }
+
+ if e.isRouter && !flags.IsRouter {
+ // "In those cases where the IsRouter flag changes from TRUE to FALSE as
+ // a result of this update, the node MUST remove that router from the
+ // Default Router List and update the Destination Cache entries for all
+ // destinations using that neighbor as a router as specified in Section
+ // 7.3.3. This is needed to detect when a node that is used as a router
+ // stops forwarding packets due to being configured as a host."
+ // - RFC 4861 section 7.2.5
+ e.nic.mu.Lock()
+ e.nic.mu.ndp.invalidateDefaultRouter(e.neigh.Addr)
+ e.nic.mu.Unlock()
+ }
+ e.isRouter = flags.IsRouter
+
+ case Unknown, Failed, Static:
+ // Do nothing
+
+ default:
+ panic(fmt.Sprintf("Invalid cache entry state: %s", e.neigh.State))
+ }
+}
+
+// handleUpperLevelConfirmationLocked processes an incoming upper-level protocol
+// (e.g. TCP acknowledgements) reachability confirmation.
+func (e *neighborEntry) handleUpperLevelConfirmationLocked() {
+ switch e.neigh.State {
+ case Reachable, Stale, Delay, Probe:
+ if e.neigh.State != Reachable {
+ e.dispatchChangeEventLocked(Reachable)
+ // Set state to Reachable again to refresh timers.
+ }
+ e.setStateLocked(Reachable)
+
+ case Unknown, Incomplete, Failed, Static:
+ // Do nothing
+
+ default:
+ panic(fmt.Sprintf("Invalid cache entry state: %s", e.neigh.State))
+ }
+}
diff --git a/pkg/tcpip/stack/neighbor_entry_test.go b/pkg/tcpip/stack/neighbor_entry_test.go
new file mode 100644
index 000000000..08c9ccd25
--- /dev/null
+++ b/pkg/tcpip/stack/neighbor_entry_test.go
@@ -0,0 +1,2770 @@
+// Copyright 2020 The gVisor Authors.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+// http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package stack
+
+import (
+ "fmt"
+ "math"
+ "math/rand"
+ "strings"
+ "sync"
+ "testing"
+ "time"
+
+ "github.com/google/go-cmp/cmp"
+ "github.com/google/go-cmp/cmp/cmpopts"
+ "gvisor.dev/gvisor/pkg/sleep"
+ "gvisor.dev/gvisor/pkg/tcpip"
+)
+
+const (
+ entryTestNetNumber tcpip.NetworkProtocolNumber = math.MaxUint32
+
+ entryTestNICID tcpip.NICID = 1
+ entryTestAddr1 = tcpip.Address("\x00\x0a\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x01")
+ entryTestAddr2 = tcpip.Address("\x00\x0a\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x02")
+
+ entryTestLinkAddr1 = tcpip.LinkAddress("\x0a\x00\x00\x00\x00\x01")
+ entryTestLinkAddr2 = tcpip.LinkAddress("\x0a\x00\x00\x00\x00\x02")
+
+ // entryTestNetDefaultMTU is the MTU, in bytes, used throughout the tests,
+ // except where another value is explicitly used. It is chosen to match the
+ // MTU of loopback interfaces on Linux systems.
+ entryTestNetDefaultMTU = 65536
+)
+
+// eventDiffOpts are the options passed to cmp.Diff to compare entry events.
+// The UpdatedAt field is ignored due to a lack of a deterministic method to
+// predict the time that an event will be dispatched.
+func eventDiffOpts() []cmp.Option {
+ return []cmp.Option{
+ cmpopts.IgnoreFields(testEntryEventInfo{}, "UpdatedAt"),
+ }
+}
+
+// eventDiffOptsWithSort is like eventDiffOpts but also includes an option to
+// sort slices of events for cases where ordering must be ignored.
+func eventDiffOptsWithSort() []cmp.Option {
+ return []cmp.Option{
+ cmpopts.IgnoreFields(testEntryEventInfo{}, "UpdatedAt"),
+ cmpopts.SortSlices(func(a, b testEntryEventInfo) bool {
+ return strings.Compare(string(a.Addr), string(b.Addr)) < 0
+ }),
+ }
+}
+
+// The following unit tests exercise every state transition and verify its
+// behavior with RFC 4681.
+//
+// | From | To | Cause | Action | Event |
+// | ========== | ========== | ========================================== | =============== | ======= |
+// | Unknown | Unknown | Confirmation w/ unknown address | | Added |
+// | Unknown | Incomplete | Packet queued to unknown address | Send probe | Added |
+// | Unknown | Stale | Probe w/ unknown address | | Added |
+// | Incomplete | Incomplete | Retransmit timer expired | Send probe | Changed |
+// | Incomplete | Reachable | Solicited confirmation | Notify wakers | Changed |
+// | Incomplete | Stale | Unsolicited confirmation | Notify wakers | Changed |
+// | Incomplete | Failed | Max probes sent without reply | Notify wakers | Removed |
+// | Reachable | Reachable | Confirmation w/ different isRouter flag | Update IsRouter | |
+// | Reachable | Stale | Reachable timer expired | | Changed |
+// | Reachable | Stale | Probe or confirmation w/ different address | | Changed |
+// | Stale | Reachable | Solicited override confirmation | Update LinkAddr | Changed |
+// | Stale | Stale | Override confirmation | Update LinkAddr | Changed |
+// | Stale | Stale | Probe w/ different address | Update LinkAddr | Changed |
+// | Stale | Delay | Packet sent | | Changed |
+// | Delay | Reachable | Upper-layer confirmation | | Changed |
+// | Delay | Reachable | Solicited override confirmation | Update LinkAddr | Changed |
+// | Delay | Stale | Probe or confirmation w/ different address | | Changed |
+// | Delay | Probe | Delay timer expired | Send probe | Changed |
+// | Probe | Reachable | Solicited override confirmation | Update LinkAddr | Changed |
+// | Probe | Reachable | Solicited confirmation w/ same address | Notify wakers | Changed |
+// | Probe | Stale | Probe or confirmation w/ different address | | Changed |
+// | Probe | Probe | Retransmit timer expired | Send probe | Changed |
+// | Probe | Failed | Max probes sent without reply | Notify wakers | Removed |
+// | Failed | | Unreachability timer expired | Delete entry | |
+
+type testEntryEventType uint8
+
+const (
+ entryTestAdded testEntryEventType = iota
+ entryTestChanged
+ entryTestRemoved
+)
+
+func (t testEntryEventType) String() string {
+ switch t {
+ case entryTestAdded:
+ return "add"
+ case entryTestChanged:
+ return "change"
+ case entryTestRemoved:
+ return "remove"
+ default:
+ return fmt.Sprintf("unknown (%d)", t)
+ }
+}
+
+// Fields are exported for use with cmp.Diff.
+type testEntryEventInfo struct {
+ EventType testEntryEventType
+ NICID tcpip.NICID
+ Addr tcpip.Address
+ LinkAddr tcpip.LinkAddress
+ State NeighborState
+ UpdatedAt time.Time
+}
+
+func (e testEntryEventInfo) String() string {
+ return fmt.Sprintf("%s event for NIC #%d, addr=%q, linkAddr=%q, state=%q", e.EventType, e.NICID, e.Addr, e.LinkAddr, e.State)
+}
+
+// testNUDDispatcher implements NUDDispatcher to validate the dispatching of
+// events upon certain NUD state machine events.
+type testNUDDispatcher struct {
+ mu sync.Mutex
+ events []testEntryEventInfo
+}
+
+var _ NUDDispatcher = (*testNUDDispatcher)(nil)
+
+func (d *testNUDDispatcher) queueEvent(e testEntryEventInfo) {
+ d.mu.Lock()
+ defer d.mu.Unlock()
+ d.events = append(d.events, e)
+}
+
+func (d *testNUDDispatcher) OnNeighborAdded(nicID tcpip.NICID, addr tcpip.Address, linkAddr tcpip.LinkAddress, state NeighborState, updatedAt time.Time) {
+ d.queueEvent(testEntryEventInfo{
+ EventType: entryTestAdded,
+ NICID: nicID,
+ Addr: addr,
+ LinkAddr: linkAddr,
+ State: state,
+ UpdatedAt: updatedAt,
+ })
+}
+
+func (d *testNUDDispatcher) OnNeighborChanged(nicID tcpip.NICID, addr tcpip.Address, linkAddr tcpip.LinkAddress, state NeighborState, updatedAt time.Time) {
+ d.queueEvent(testEntryEventInfo{
+ EventType: entryTestChanged,
+ NICID: nicID,
+ Addr: addr,
+ LinkAddr: linkAddr,
+ State: state,
+ UpdatedAt: updatedAt,
+ })
+}
+
+func (d *testNUDDispatcher) OnNeighborRemoved(nicID tcpip.NICID, addr tcpip.Address, linkAddr tcpip.LinkAddress, state NeighborState, updatedAt time.Time) {
+ d.queueEvent(testEntryEventInfo{
+ EventType: entryTestRemoved,
+ NICID: nicID,
+ Addr: addr,
+ LinkAddr: linkAddr,
+ State: state,
+ UpdatedAt: updatedAt,
+ })
+}
+
+type entryTestLinkResolver struct {
+ mu sync.Mutex
+ probes []entryTestProbeInfo
+}
+
+var _ LinkAddressResolver = (*entryTestLinkResolver)(nil)
+
+type entryTestProbeInfo struct {
+ RemoteAddress tcpip.Address
+ RemoteLinkAddress tcpip.LinkAddress
+ LocalAddress tcpip.Address
+}
+
+func (p entryTestProbeInfo) String() string {
+ return fmt.Sprintf("probe with RemoteAddress=%q, RemoteLinkAddress=%q, LocalAddress=%q", p.RemoteAddress, p.RemoteLinkAddress, p.LocalAddress)
+}
+
+// LinkAddressRequest sends a request for the LinkAddress of addr. Broadcasts
+// to the local network if linkAddr is the zero value.
+func (r *entryTestLinkResolver) LinkAddressRequest(addr, localAddr tcpip.Address, linkAddr tcpip.LinkAddress, linkEP LinkEndpoint) *tcpip.Error {
+ p := entryTestProbeInfo{
+ RemoteAddress: addr,
+ RemoteLinkAddress: linkAddr,
+ LocalAddress: localAddr,
+ }
+ r.mu.Lock()
+ defer r.mu.Unlock()
+ r.probes = append(r.probes, p)
+ return nil
+}
+
+// ResolveStaticAddress attempts to resolve address without sending requests.
+// It either resolves the name immediately or returns the empty LinkAddress.
+func (r *entryTestLinkResolver) ResolveStaticAddress(addr tcpip.Address) (tcpip.LinkAddress, bool) {
+ return "", false
+}
+
+// LinkAddressProtocol returns the network protocol of the addresses this
+// resolver can resolve.
+func (r *entryTestLinkResolver) LinkAddressProtocol() tcpip.NetworkProtocolNumber {
+ return entryTestNetNumber
+}
+
+func entryTestSetup(c NUDConfigurations) (*neighborEntry, *testNUDDispatcher, *entryTestLinkResolver, *fakeClock) {
+ clock := newFakeClock()
+ disp := testNUDDispatcher{}
+ nic := NIC{
+ id: entryTestNICID,
+ linkEP: nil, // entryTestLinkResolver doesn't use a LinkEndpoint
+ stack: &Stack{
+ clock: clock,
+ nudDisp: &disp,
+ },
+ }
+
+ rng := rand.New(rand.NewSource(time.Now().UnixNano()))
+ nudState := NewNUDState(c, rng)
+ linkRes := entryTestLinkResolver{}
+ entry := newNeighborEntry(&nic, entryTestAddr1, entryTestAddr2, nudState, &linkRes)
+
+ // Stub out ndpState to verify modification of default routers.
+ nic.mu.ndp = ndpState{
+ nic: &nic,
+ defaultRouters: make(map[tcpip.Address]defaultRouterState),
+ }
+
+ // Stub out the neighbor cache to verify deletion from the cache.
+ nic.neigh = &neighborCache{
+ nic: &nic,
+ state: nudState,
+ cache: make(map[tcpip.Address]*neighborEntry, neighborCacheSize),
+ }
+ nic.neigh.cache[entryTestAddr1] = entry
+
+ return entry, &disp, &linkRes, clock
+}
+
+// TestEntryInitiallyUnknown verifies that the state of a newly created
+// neighborEntry is Unknown.
+func TestEntryInitiallyUnknown(t *testing.T) {
+ c := DefaultNUDConfigurations()
+ e, nudDisp, linkRes, clock := entryTestSetup(c)
+
+ e.mu.Lock()
+ if got, want := e.neigh.State, Unknown; got != want {
+ t.Errorf("got e.neigh.State = %q, want = %q", got, want)
+ }
+ e.mu.Unlock()
+
+ clock.advance(c.RetransmitTimer)
+
+ // No probes should have been sent.
+ linkRes.mu.Lock()
+ diff := cmp.Diff(linkRes.probes, []entryTestProbeInfo(nil))
+ linkRes.mu.Unlock()
+ if diff != "" {
+ t.Fatalf("link address resolver probes mismatch (-got, +want):\n%s", diff)
+ }
+
+ // No events should have been dispatched.
+ nudDisp.mu.Lock()
+ if diff := cmp.Diff(nudDisp.events, []testEntryEventInfo(nil)); diff != "" {
+ t.Errorf("nud dispatcher events mismatch (-got, +want):\n%s", diff)
+ }
+ nudDisp.mu.Unlock()
+}
+
+func TestEntryUnknownToUnknownWhenConfirmationWithUnknownAddress(t *testing.T) {
+ c := DefaultNUDConfigurations()
+ e, nudDisp, linkRes, clock := entryTestSetup(c)
+
+ e.mu.Lock()
+ e.handleConfirmationLocked(entryTestLinkAddr1, ReachabilityConfirmationFlags{
+ Solicited: false,
+ Override: false,
+ IsRouter: false,
+ })
+ if got, want := e.neigh.State, Unknown; got != want {
+ t.Errorf("got e.neigh.State = %q, want = %q", got, want)
+ }
+ e.mu.Unlock()
+
+ clock.advance(time.Hour)
+
+ // No probes should have been sent.
+ linkRes.mu.Lock()
+ diff := cmp.Diff(linkRes.probes, []entryTestProbeInfo(nil))
+ linkRes.mu.Unlock()
+ if diff != "" {
+ t.Fatalf("link address resolver probes mismatch (-got, +want):\n%s", diff)
+ }
+
+ // No events should have been dispatched.
+ nudDisp.mu.Lock()
+ if diff := cmp.Diff(nudDisp.events, []testEntryEventInfo(nil)); diff != "" {
+ t.Errorf("nud dispatcher events mismatch (-got, +want):\n%s", diff)
+ }
+ nudDisp.mu.Unlock()
+}
+
+func TestEntryUnknownToIncomplete(t *testing.T) {
+ c := DefaultNUDConfigurations()
+ e, nudDisp, linkRes, _ := entryTestSetup(c)
+
+ e.mu.Lock()
+ e.handlePacketQueuedLocked()
+ if got, want := e.neigh.State, Incomplete; got != want {
+ t.Errorf("got e.neigh.State = %q, want = %q", got, want)
+ }
+ e.mu.Unlock()
+
+ wantProbes := []entryTestProbeInfo{
+ {
+ RemoteAddress: entryTestAddr1,
+ RemoteLinkAddress: tcpip.LinkAddress(""),
+ LocalAddress: entryTestAddr2,
+ },
+ }
+ linkRes.mu.Lock()
+ diff := cmp.Diff(linkRes.probes, wantProbes)
+ linkRes.mu.Unlock()
+ if diff != "" {
+ t.Fatalf("link address resolver probes mismatch (-got, +want):\n%s", diff)
+ }
+
+ wantEvents := []testEntryEventInfo{
+ {
+ EventType: entryTestAdded,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: tcpip.LinkAddress(""),
+ State: Incomplete,
+ },
+ }
+ {
+ nudDisp.mu.Lock()
+ diff := cmp.Diff(nudDisp.events, wantEvents, eventDiffOpts()...)
+ nudDisp.mu.Unlock()
+ if diff != "" {
+ t.Fatalf("nud dispatcher events mismatch (-got, +want):\n%s", diff)
+ }
+ }
+}
+
+func TestEntryUnknownToStale(t *testing.T) {
+ c := DefaultNUDConfigurations()
+ e, nudDisp, linkRes, _ := entryTestSetup(c)
+
+ e.mu.Lock()
+ e.handleProbeLocked(entryTestLinkAddr1)
+ if got, want := e.neigh.State, Stale; got != want {
+ t.Errorf("got e.neigh.State = %q, want = %q", got, want)
+ }
+ e.mu.Unlock()
+
+ // No probes should have been sent.
+ linkRes.mu.Lock()
+ diff := cmp.Diff(linkRes.probes, []entryTestProbeInfo(nil))
+ linkRes.mu.Unlock()
+ if diff != "" {
+ t.Fatalf("link address resolver probes mismatch (-got, +want):\n%s", diff)
+ }
+
+ wantEvents := []testEntryEventInfo{
+ {
+ EventType: entryTestAdded,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: entryTestLinkAddr1,
+ State: Stale,
+ },
+ }
+ nudDisp.mu.Lock()
+ if diff := cmp.Diff(nudDisp.events, wantEvents, eventDiffOpts()...); diff != "" {
+ t.Errorf("nud dispatcher events mismatch (-got, +want):\n%s", diff)
+ }
+ nudDisp.mu.Unlock()
+}
+
+func TestEntryIncompleteToIncompleteDoesNotChangeUpdatedAt(t *testing.T) {
+ c := DefaultNUDConfigurations()
+ c.MaxMulticastProbes = 3
+ e, nudDisp, linkRes, clock := entryTestSetup(c)
+
+ e.mu.Lock()
+ e.handlePacketQueuedLocked()
+ if got, want := e.neigh.State, Incomplete; got != want {
+ t.Errorf("got e.neigh.State = %q, want = %q", got, want)
+ }
+ updatedAt := e.neigh.UpdatedAt
+ e.mu.Unlock()
+
+ clock.advance(c.RetransmitTimer)
+
+ // UpdatedAt should remain the same during address resolution.
+ wantProbes := []entryTestProbeInfo{
+ {
+ RemoteAddress: entryTestAddr1,
+ RemoteLinkAddress: tcpip.LinkAddress(""),
+ LocalAddress: entryTestAddr2,
+ },
+ {
+ RemoteAddress: entryTestAddr1,
+ RemoteLinkAddress: tcpip.LinkAddress(""),
+ LocalAddress: entryTestAddr2,
+ },
+ }
+ linkRes.mu.Lock()
+ diff := cmp.Diff(linkRes.probes, wantProbes)
+ linkRes.probes = nil
+ linkRes.mu.Unlock()
+ if diff != "" {
+ t.Fatalf("link address resolver probes mismatch (-got, +want):\n%s", diff)
+ }
+
+ e.mu.Lock()
+ if got, want := e.neigh.UpdatedAt, updatedAt; got != want {
+ t.Errorf("got e.neigh.UpdatedAt = %q, want = %q", got, want)
+ }
+ e.mu.Unlock()
+
+ clock.advance(c.RetransmitTimer)
+
+ // UpdatedAt should change after failing address resolution. Timing out after
+ // sending the last probe transitions the entry to Failed.
+ {
+ wantProbes := []entryTestProbeInfo{
+ {
+ RemoteAddress: entryTestAddr1,
+ RemoteLinkAddress: tcpip.LinkAddress(""),
+ LocalAddress: entryTestAddr2,
+ },
+ }
+ linkRes.mu.Lock()
+ diff := cmp.Diff(linkRes.probes, wantProbes)
+ linkRes.mu.Unlock()
+ if diff != "" {
+ t.Fatalf("link address resolver probes mismatch (-got, +want):\n%s", diff)
+ }
+ }
+
+ clock.advance(c.RetransmitTimer)
+
+ wantEvents := []testEntryEventInfo{
+ {
+ EventType: entryTestAdded,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: tcpip.LinkAddress(""),
+ State: Incomplete,
+ },
+ {
+ EventType: entryTestRemoved,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: tcpip.LinkAddress(""),
+ State: Incomplete,
+ },
+ }
+ nudDisp.mu.Lock()
+ if diff := cmp.Diff(nudDisp.events, wantEvents, eventDiffOpts()...); diff != "" {
+ t.Errorf("nud dispatcher events mismatch (-got, +want):\n%s", diff)
+ }
+ nudDisp.mu.Unlock()
+
+ e.mu.Lock()
+ if got, notWant := e.neigh.UpdatedAt, updatedAt; got == notWant {
+ t.Errorf("expected e.neigh.UpdatedAt to change, got = %q", got)
+ }
+ e.mu.Unlock()
+}
+
+func TestEntryIncompleteToReachable(t *testing.T) {
+ c := DefaultNUDConfigurations()
+ e, nudDisp, linkRes, _ := entryTestSetup(c)
+
+ e.mu.Lock()
+ e.handlePacketQueuedLocked()
+ if got, want := e.neigh.State, Incomplete; got != want {
+ t.Errorf("got e.neigh.State = %q, want = %q", got, want)
+ }
+ e.handleConfirmationLocked(entryTestLinkAddr1, ReachabilityConfirmationFlags{
+ Solicited: true,
+ Override: false,
+ IsRouter: false,
+ })
+ if got, want := e.neigh.State, Reachable; got != want {
+ t.Errorf("got e.neigh.State = %q, want = %q", got, want)
+ }
+ e.mu.Unlock()
+
+ wantProbes := []entryTestProbeInfo{
+ {
+ RemoteAddress: entryTestAddr1,
+ RemoteLinkAddress: tcpip.LinkAddress(""),
+ LocalAddress: entryTestAddr2,
+ },
+ }
+ linkRes.mu.Lock()
+ diff := cmp.Diff(linkRes.probes, wantProbes)
+ linkRes.mu.Unlock()
+ if diff != "" {
+ t.Fatalf("link address resolver probes mismatch (-got, +want):\n%s", diff)
+ }
+
+ wantEvents := []testEntryEventInfo{
+ {
+ EventType: entryTestAdded,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: tcpip.LinkAddress(""),
+ State: Incomplete,
+ },
+ {
+ EventType: entryTestChanged,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: entryTestLinkAddr1,
+ State: Reachable,
+ },
+ }
+ nudDisp.mu.Lock()
+ if diff := cmp.Diff(nudDisp.events, wantEvents, eventDiffOpts()...); diff != "" {
+ t.Errorf("nud dispatcher events mismatch (-got, +want):\n%s", diff)
+ }
+ nudDisp.mu.Unlock()
+}
+
+// TestEntryAddsAndClearsWakers verifies that wakers are added when
+// addWakerLocked is called and cleared when address resolution finishes. In
+// this case, address resolution will finish when transitioning from Incomplete
+// to Reachable.
+func TestEntryAddsAndClearsWakers(t *testing.T) {
+ c := DefaultNUDConfigurations()
+ e, nudDisp, linkRes, _ := entryTestSetup(c)
+
+ w := sleep.Waker{}
+ s := sleep.Sleeper{}
+ s.AddWaker(&w, 123)
+ defer s.Done()
+
+ e.mu.Lock()
+ e.handlePacketQueuedLocked()
+ if got := e.wakers; got != nil {
+ t.Errorf("got e.wakers = %v, want = nil", got)
+ }
+ e.addWakerLocked(&w)
+ if got, want := w.IsAsserted(), false; got != want {
+ t.Errorf("waker.IsAsserted() = %t, want = %t", got, want)
+ }
+ if e.wakers == nil {
+ t.Error("expected e.wakers to be non-nil")
+ }
+ e.handleConfirmationLocked(entryTestLinkAddr1, ReachabilityConfirmationFlags{
+ Solicited: true,
+ Override: false,
+ IsRouter: false,
+ })
+ if e.wakers != nil {
+ t.Errorf("got e.wakers = %v, want = nil", e.wakers)
+ }
+ if got, want := w.IsAsserted(), true; got != want {
+ t.Errorf("waker.IsAsserted() = %t, want = %t", got, want)
+ }
+ e.mu.Unlock()
+
+ wantProbes := []entryTestProbeInfo{
+ {
+ RemoteAddress: entryTestAddr1,
+ RemoteLinkAddress: tcpip.LinkAddress(""),
+ LocalAddress: entryTestAddr2,
+ },
+ }
+ linkRes.mu.Lock()
+ diff := cmp.Diff(linkRes.probes, wantProbes)
+ linkRes.mu.Unlock()
+ if diff != "" {
+ t.Fatalf("link address resolver probes mismatch (-got, +want):\n%s", diff)
+ }
+
+ wantEvents := []testEntryEventInfo{
+ {
+ EventType: entryTestAdded,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: tcpip.LinkAddress(""),
+ State: Incomplete,
+ },
+ {
+ EventType: entryTestChanged,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: entryTestLinkAddr1,
+ State: Reachable,
+ },
+ }
+ nudDisp.mu.Lock()
+ if diff := cmp.Diff(nudDisp.events, wantEvents, eventDiffOpts()...); diff != "" {
+ t.Errorf("nud dispatcher events mismatch (-got, +want):\n%s", diff)
+ }
+ nudDisp.mu.Unlock()
+}
+
+func TestEntryIncompleteToReachableWithRouterFlag(t *testing.T) {
+ c := DefaultNUDConfigurations()
+ e, nudDisp, linkRes, _ := entryTestSetup(c)
+
+ e.mu.Lock()
+ e.handlePacketQueuedLocked()
+ if got, want := e.neigh.State, Incomplete; got != want {
+ t.Errorf("got e.neigh.State = %q, want = %q", got, want)
+ }
+ e.handleConfirmationLocked(entryTestLinkAddr1, ReachabilityConfirmationFlags{
+ Solicited: true,
+ Override: false,
+ IsRouter: true,
+ })
+ if got, want := e.neigh.State, Reachable; got != want {
+ t.Errorf("got e.neigh.State = %q, want = %q", got, want)
+ }
+ if got, want := e.isRouter, true; got != want {
+ t.Errorf("got e.isRouter = %t, want = %t", got, want)
+ }
+ e.mu.Unlock()
+
+ wantProbes := []entryTestProbeInfo{
+ {
+ RemoteAddress: entryTestAddr1,
+ RemoteLinkAddress: tcpip.LinkAddress(""),
+ LocalAddress: entryTestAddr2,
+ },
+ }
+ linkRes.mu.Lock()
+ if diff := cmp.Diff(linkRes.probes, wantProbes); diff != "" {
+ t.Fatalf("link address resolver probes mismatch (-got, +want):\n%s", diff)
+ }
+ linkRes.mu.Unlock()
+
+ wantEvents := []testEntryEventInfo{
+ {
+ EventType: entryTestAdded,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: tcpip.LinkAddress(""),
+ State: Incomplete,
+ },
+ {
+ EventType: entryTestChanged,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: entryTestLinkAddr1,
+ State: Reachable,
+ },
+ }
+ nudDisp.mu.Lock()
+ if diff := cmp.Diff(nudDisp.events, wantEvents, eventDiffOpts()...); diff != "" {
+ t.Errorf("nud dispatcher events mismatch (-got, +want):\n%s", diff)
+ }
+ nudDisp.mu.Unlock()
+}
+
+func TestEntryIncompleteToStale(t *testing.T) {
+ c := DefaultNUDConfigurations()
+ e, nudDisp, linkRes, _ := entryTestSetup(c)
+
+ e.mu.Lock()
+ e.handlePacketQueuedLocked()
+ if got, want := e.neigh.State, Incomplete; got != want {
+ t.Errorf("got e.neigh.State = %q, want = %q", got, want)
+ }
+ e.handleConfirmationLocked(entryTestLinkAddr1, ReachabilityConfirmationFlags{
+ Solicited: false,
+ Override: false,
+ IsRouter: false,
+ })
+ if got, want := e.neigh.State, Stale; got != want {
+ t.Errorf("got e.neigh.State = %q, want = %q", got, want)
+ }
+ e.mu.Unlock()
+
+ wantProbes := []entryTestProbeInfo{
+ {
+ RemoteAddress: entryTestAddr1,
+ RemoteLinkAddress: tcpip.LinkAddress(""),
+ LocalAddress: entryTestAddr2,
+ },
+ }
+ linkRes.mu.Lock()
+ diff := cmp.Diff(linkRes.probes, wantProbes)
+ linkRes.mu.Unlock()
+ if diff != "" {
+ t.Fatalf("link address resolver probes mismatch (-got, +want):\n%s", diff)
+ }
+
+ wantEvents := []testEntryEventInfo{
+ {
+ EventType: entryTestAdded,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: tcpip.LinkAddress(""),
+ State: Incomplete,
+ },
+ {
+ EventType: entryTestChanged,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: entryTestLinkAddr1,
+ State: Stale,
+ },
+ }
+ nudDisp.mu.Lock()
+ if diff := cmp.Diff(nudDisp.events, wantEvents, eventDiffOpts()...); diff != "" {
+ t.Errorf("nud dispatcher events mismatch (-got, +want):\n%s", diff)
+ }
+ nudDisp.mu.Unlock()
+}
+
+func TestEntryIncompleteToFailed(t *testing.T) {
+ c := DefaultNUDConfigurations()
+ c.MaxMulticastProbes = 3
+ e, nudDisp, linkRes, clock := entryTestSetup(c)
+
+ e.mu.Lock()
+ e.handlePacketQueuedLocked()
+ if got, want := e.neigh.State, Incomplete; got != want {
+ t.Errorf("got e.neigh.State = %q, want = %q", got, want)
+ }
+ e.mu.Unlock()
+
+ waitFor := c.RetransmitTimer * time.Duration(c.MaxMulticastProbes)
+ clock.advance(waitFor)
+
+ wantProbes := []entryTestProbeInfo{
+ // The Incomplete-to-Incomplete state transition is tested here by
+ // verifying that 3 reachability probes were sent.
+ {
+ RemoteAddress: entryTestAddr1,
+ RemoteLinkAddress: tcpip.LinkAddress(""),
+ LocalAddress: entryTestAddr2,
+ },
+ {
+ RemoteAddress: entryTestAddr1,
+ RemoteLinkAddress: tcpip.LinkAddress(""),
+ LocalAddress: entryTestAddr2,
+ },
+ {
+ RemoteAddress: entryTestAddr1,
+ RemoteLinkAddress: tcpip.LinkAddress(""),
+ LocalAddress: entryTestAddr2,
+ },
+ }
+ linkRes.mu.Lock()
+ diff := cmp.Diff(linkRes.probes, wantProbes)
+ linkRes.mu.Unlock()
+ if diff != "" {
+ t.Fatalf("link address resolver probes mismatch (-got, +want):\n%s", diff)
+ }
+
+ wantEvents := []testEntryEventInfo{
+ {
+ EventType: entryTestAdded,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: tcpip.LinkAddress(""),
+ State: Incomplete,
+ },
+ {
+ EventType: entryTestRemoved,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: tcpip.LinkAddress(""),
+ State: Incomplete,
+ },
+ }
+ nudDisp.mu.Lock()
+ if diff := cmp.Diff(nudDisp.events, wantEvents, eventDiffOpts()...); diff != "" {
+ t.Errorf("nud dispatcher events mismatch (-got, +want):\n%s", diff)
+ }
+ nudDisp.mu.Unlock()
+
+ e.mu.Lock()
+ if got, want := e.neigh.State, Failed; got != want {
+ t.Errorf("got e.neigh.State = %q, want = %q", got, want)
+ }
+ e.mu.Unlock()
+}
+
+type testLocker struct{}
+
+var _ sync.Locker = (*testLocker)(nil)
+
+func (*testLocker) Lock() {}
+func (*testLocker) Unlock() {}
+
+func TestEntryStaysReachableWhenConfirmationWithRouterFlag(t *testing.T) {
+ c := DefaultNUDConfigurations()
+ e, nudDisp, linkRes, _ := entryTestSetup(c)
+
+ e.mu.Lock()
+ e.handlePacketQueuedLocked()
+ e.handleConfirmationLocked(entryTestLinkAddr1, ReachabilityConfirmationFlags{
+ Solicited: true,
+ Override: false,
+ IsRouter: true,
+ })
+ if got, want := e.neigh.State, Reachable; got != want {
+ t.Errorf("got e.neigh.State = %q, want = %q", got, want)
+ }
+ if got, want := e.isRouter, true; got != want {
+ t.Errorf("got e.isRouter = %t, want = %t", got, want)
+ }
+ e.nic.mu.ndp.defaultRouters[entryTestAddr1] = defaultRouterState{
+ invalidationJob: e.nic.stack.newJob(&testLocker{}, func() {}),
+ }
+ e.handleConfirmationLocked(entryTestLinkAddr1, ReachabilityConfirmationFlags{
+ Solicited: false,
+ Override: false,
+ IsRouter: false,
+ })
+ if got, want := e.isRouter, false; got != want {
+ t.Errorf("got e.isRouter = %t, want = %t", got, want)
+ }
+ if _, ok := e.nic.mu.ndp.defaultRouters[entryTestAddr1]; ok {
+ t.Errorf("unexpected defaultRouter for %s", entryTestAddr1)
+ }
+ e.mu.Unlock()
+
+ wantProbes := []entryTestProbeInfo{
+ {
+ RemoteAddress: entryTestAddr1,
+ RemoteLinkAddress: tcpip.LinkAddress(""),
+ LocalAddress: entryTestAddr2,
+ },
+ }
+ linkRes.mu.Lock()
+ diff := cmp.Diff(linkRes.probes, wantProbes)
+ linkRes.mu.Unlock()
+ if diff != "" {
+ t.Fatalf("link address resolver probes mismatch (-got, +want):\n%s", diff)
+ }
+
+ wantEvents := []testEntryEventInfo{
+ {
+ EventType: entryTestAdded,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: tcpip.LinkAddress(""),
+ State: Incomplete,
+ },
+ {
+ EventType: entryTestChanged,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: entryTestLinkAddr1,
+ State: Reachable,
+ },
+ }
+ nudDisp.mu.Lock()
+ if diff := cmp.Diff(nudDisp.events, wantEvents, eventDiffOpts()...); diff != "" {
+ t.Errorf("nud dispatcher events mismatch (-got, +want):\n%s", diff)
+ }
+ nudDisp.mu.Unlock()
+
+ e.mu.Lock()
+ if got, want := e.neigh.State, Reachable; got != want {
+ t.Errorf("got e.neigh.State = %q, want = %q", got, want)
+ }
+ e.mu.Unlock()
+}
+
+func TestEntryStaysReachableWhenProbeWithSameAddress(t *testing.T) {
+ c := DefaultNUDConfigurations()
+ e, nudDisp, linkRes, _ := entryTestSetup(c)
+
+ e.mu.Lock()
+ e.handlePacketQueuedLocked()
+ e.handleConfirmationLocked(entryTestLinkAddr1, ReachabilityConfirmationFlags{
+ Solicited: true,
+ Override: false,
+ IsRouter: false,
+ })
+ if got, want := e.neigh.State, Reachable; got != want {
+ t.Errorf("got e.neigh.State = %q, want = %q", got, want)
+ }
+ e.handleProbeLocked(entryTestLinkAddr1)
+ if got, want := e.neigh.State, Reachable; got != want {
+ t.Errorf("got e.neigh.State = %q, want = %q", got, want)
+ }
+ if got, want := e.neigh.LinkAddr, entryTestLinkAddr1; got != want {
+ t.Errorf("got e.neigh.LinkAddr = %q, want = %q", got, want)
+ }
+ e.mu.Unlock()
+
+ wantProbes := []entryTestProbeInfo{
+ {
+ RemoteAddress: entryTestAddr1,
+ RemoteLinkAddress: tcpip.LinkAddress(""),
+ LocalAddress: entryTestAddr2,
+ },
+ }
+ linkRes.mu.Lock()
+ diff := cmp.Diff(linkRes.probes, wantProbes)
+ linkRes.mu.Unlock()
+ if diff != "" {
+ t.Fatalf("link address resolver probes mismatch (-got, +want):\n%s", diff)
+ }
+
+ wantEvents := []testEntryEventInfo{
+ {
+ EventType: entryTestAdded,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: tcpip.LinkAddress(""),
+ State: Incomplete,
+ },
+ {
+ EventType: entryTestChanged,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: entryTestLinkAddr1,
+ State: Reachable,
+ },
+ }
+ nudDisp.mu.Lock()
+ if diff := cmp.Diff(nudDisp.events, wantEvents, eventDiffOpts()...); diff != "" {
+ t.Errorf("nud dispatcher events mismatch (-got, +want):\n%s", diff)
+ }
+ nudDisp.mu.Unlock()
+}
+
+func TestEntryReachableToStaleWhenTimeout(t *testing.T) {
+ c := DefaultNUDConfigurations()
+ // Eliminate random factors from ReachableTime computation so the transition
+ // from Stale to Reachable will only take BaseReachableTime duration.
+ c.MinRandomFactor = 1
+ c.MaxRandomFactor = 1
+
+ e, nudDisp, linkRes, clock := entryTestSetup(c)
+
+ e.mu.Lock()
+ e.handlePacketQueuedLocked()
+ e.handleConfirmationLocked(entryTestLinkAddr1, ReachabilityConfirmationFlags{
+ Solicited: true,
+ Override: false,
+ IsRouter: false,
+ })
+ if got, want := e.neigh.State, Reachable; got != want {
+ t.Errorf("got e.neigh.State = %q, want = %q", got, want)
+ }
+ e.mu.Unlock()
+
+ wantProbes := []entryTestProbeInfo{
+ {
+ RemoteAddress: entryTestAddr1,
+ RemoteLinkAddress: tcpip.LinkAddress(""),
+ LocalAddress: entryTestAddr2,
+ },
+ }
+ linkRes.mu.Lock()
+ diff := cmp.Diff(linkRes.probes, wantProbes)
+ linkRes.mu.Unlock()
+ if diff != "" {
+ t.Fatalf("link address resolver probes mismatch (-got, +want):\n%s", diff)
+ }
+
+ clock.advance(c.BaseReachableTime)
+
+ wantEvents := []testEntryEventInfo{
+ {
+ EventType: entryTestAdded,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: tcpip.LinkAddress(""),
+ State: Incomplete,
+ },
+ {
+ EventType: entryTestChanged,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: entryTestLinkAddr1,
+ State: Reachable,
+ },
+ {
+ EventType: entryTestChanged,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: entryTestLinkAddr1,
+ State: Stale,
+ },
+ }
+ nudDisp.mu.Lock()
+ if diff := cmp.Diff(nudDisp.events, wantEvents, eventDiffOpts()...); diff != "" {
+ t.Errorf("nud dispatcher events mismatch (-got, +want):\n%s", diff)
+ }
+ nudDisp.mu.Unlock()
+
+ e.mu.Lock()
+ if got, want := e.neigh.State, Stale; got != want {
+ t.Errorf("got e.neigh.State = %q, want = %q", got, want)
+ }
+ e.mu.Unlock()
+}
+
+func TestEntryReachableToStaleWhenProbeWithDifferentAddress(t *testing.T) {
+ c := DefaultNUDConfigurations()
+ e, nudDisp, linkRes, _ := entryTestSetup(c)
+
+ e.mu.Lock()
+ e.handlePacketQueuedLocked()
+ e.handleConfirmationLocked(entryTestLinkAddr1, ReachabilityConfirmationFlags{
+ Solicited: true,
+ Override: false,
+ IsRouter: false,
+ })
+ if got, want := e.neigh.State, Reachable; got != want {
+ t.Errorf("got e.neigh.State = %q, want = %q", got, want)
+ }
+ e.handleProbeLocked(entryTestLinkAddr2)
+ if got, want := e.neigh.State, Stale; got != want {
+ t.Errorf("got e.neigh.State = %q, want = %q", got, want)
+ }
+ e.mu.Unlock()
+
+ wantProbes := []entryTestProbeInfo{
+ {
+ RemoteAddress: entryTestAddr1,
+ RemoteLinkAddress: tcpip.LinkAddress(""),
+ LocalAddress: entryTestAddr2,
+ },
+ }
+ linkRes.mu.Lock()
+ diff := cmp.Diff(linkRes.probes, wantProbes)
+ linkRes.mu.Unlock()
+ if diff != "" {
+ t.Fatalf("link address resolver probes mismatch (-got, +want):\n%s", diff)
+ }
+
+ wantEvents := []testEntryEventInfo{
+ {
+ EventType: entryTestAdded,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: tcpip.LinkAddress(""),
+ State: Incomplete,
+ },
+ {
+ EventType: entryTestChanged,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: entryTestLinkAddr1,
+ State: Reachable,
+ },
+ {
+ EventType: entryTestChanged,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: entryTestLinkAddr2,
+ State: Stale,
+ },
+ }
+ nudDisp.mu.Lock()
+ if diff := cmp.Diff(nudDisp.events, wantEvents, eventDiffOpts()...); diff != "" {
+ t.Errorf("nud dispatcher events mismatch (-got, +want):\n%s", diff)
+ }
+ nudDisp.mu.Unlock()
+
+ e.mu.Lock()
+ if got, want := e.neigh.State, Stale; got != want {
+ t.Errorf("got e.neigh.State = %q, want = %q", got, want)
+ }
+ e.mu.Unlock()
+}
+
+func TestEntryReachableToStaleWhenConfirmationWithDifferentAddress(t *testing.T) {
+ c := DefaultNUDConfigurations()
+ e, nudDisp, linkRes, _ := entryTestSetup(c)
+
+ e.mu.Lock()
+ e.handlePacketQueuedLocked()
+ e.handleConfirmationLocked(entryTestLinkAddr1, ReachabilityConfirmationFlags{
+ Solicited: true,
+ Override: false,
+ IsRouter: false,
+ })
+ if got, want := e.neigh.State, Reachable; got != want {
+ t.Errorf("got e.neigh.State = %q, want = %q", got, want)
+ }
+ e.handleConfirmationLocked(entryTestLinkAddr2, ReachabilityConfirmationFlags{
+ Solicited: false,
+ Override: false,
+ IsRouter: false,
+ })
+ if got, want := e.neigh.State, Stale; got != want {
+ t.Errorf("got e.neigh.State = %q, want = %q", got, want)
+ }
+ e.mu.Unlock()
+
+ wantProbes := []entryTestProbeInfo{
+ {
+ RemoteAddress: entryTestAddr1,
+ RemoteLinkAddress: tcpip.LinkAddress(""),
+ LocalAddress: entryTestAddr2,
+ },
+ }
+ linkRes.mu.Lock()
+ diff := cmp.Diff(linkRes.probes, wantProbes)
+ linkRes.mu.Unlock()
+ if diff != "" {
+ t.Fatalf("link address resolver probes mismatch (-got, +want):\n%s", diff)
+ }
+
+ wantEvents := []testEntryEventInfo{
+ {
+ EventType: entryTestAdded,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: tcpip.LinkAddress(""),
+ State: Incomplete,
+ },
+ {
+ EventType: entryTestChanged,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: entryTestLinkAddr1,
+ State: Reachable,
+ },
+ {
+ EventType: entryTestChanged,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: entryTestLinkAddr1,
+ State: Stale,
+ },
+ }
+ nudDisp.mu.Lock()
+ if diff := cmp.Diff(nudDisp.events, wantEvents, eventDiffOpts()...); diff != "" {
+ t.Errorf("nud dispatcher events mismatch (-got, +want):\n%s", diff)
+ }
+ nudDisp.mu.Unlock()
+
+ e.mu.Lock()
+ if got, want := e.neigh.State, Stale; got != want {
+ t.Errorf("got e.neigh.State = %q, want = %q", got, want)
+ }
+ e.mu.Unlock()
+}
+
+func TestEntryReachableToStaleWhenConfirmationWithDifferentAddressAndOverride(t *testing.T) {
+ c := DefaultNUDConfigurations()
+ e, nudDisp, linkRes, _ := entryTestSetup(c)
+
+ e.mu.Lock()
+ e.handlePacketQueuedLocked()
+ e.handleConfirmationLocked(entryTestLinkAddr1, ReachabilityConfirmationFlags{
+ Solicited: true,
+ Override: false,
+ IsRouter: false,
+ })
+ if got, want := e.neigh.State, Reachable; got != want {
+ t.Errorf("got e.neigh.State = %q, want = %q", got, want)
+ }
+ e.handleConfirmationLocked(entryTestLinkAddr2, ReachabilityConfirmationFlags{
+ Solicited: false,
+ Override: true,
+ IsRouter: false,
+ })
+ if got, want := e.neigh.State, Stale; got != want {
+ t.Errorf("got e.neigh.State = %q, want = %q", got, want)
+ }
+ e.mu.Unlock()
+
+ wantProbes := []entryTestProbeInfo{
+ {
+ RemoteAddress: entryTestAddr1,
+ RemoteLinkAddress: tcpip.LinkAddress(""),
+ LocalAddress: entryTestAddr2,
+ },
+ }
+ linkRes.mu.Lock()
+ diff := cmp.Diff(linkRes.probes, wantProbes)
+ linkRes.mu.Unlock()
+ if diff != "" {
+ t.Fatalf("link address resolver probes mismatch (-got, +want):\n%s", diff)
+ }
+
+ wantEvents := []testEntryEventInfo{
+ {
+ EventType: entryTestAdded,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: tcpip.LinkAddress(""),
+ State: Incomplete,
+ },
+ {
+ EventType: entryTestChanged,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: entryTestLinkAddr1,
+ State: Reachable,
+ },
+ {
+ EventType: entryTestChanged,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: entryTestLinkAddr2,
+ State: Stale,
+ },
+ }
+ nudDisp.mu.Lock()
+ if diff := cmp.Diff(nudDisp.events, wantEvents, eventDiffOpts()...); diff != "" {
+ t.Errorf("nud dispatcher events mismatch (-got, +want):\n%s", diff)
+ }
+ nudDisp.mu.Unlock()
+
+ e.mu.Lock()
+ if got, want := e.neigh.State, Stale; got != want {
+ t.Errorf("got e.neigh.State = %q, want = %q", got, want)
+ }
+ e.mu.Unlock()
+}
+
+func TestEntryStaysStaleWhenProbeWithSameAddress(t *testing.T) {
+ c := DefaultNUDConfigurations()
+ e, nudDisp, linkRes, _ := entryTestSetup(c)
+
+ e.mu.Lock()
+ e.handlePacketQueuedLocked()
+ e.handleConfirmationLocked(entryTestLinkAddr1, ReachabilityConfirmationFlags{
+ Solicited: false,
+ Override: false,
+ IsRouter: false,
+ })
+ if got, want := e.neigh.State, Stale; got != want {
+ t.Errorf("got e.neigh.State = %q, want = %q", got, want)
+ }
+ e.handleProbeLocked(entryTestLinkAddr1)
+ if got, want := e.neigh.State, Stale; got != want {
+ t.Errorf("got e.neigh.State = %q, want = %q", got, want)
+ }
+ if got, want := e.neigh.LinkAddr, entryTestLinkAddr1; got != want {
+ t.Errorf("got e.neigh.LinkAddr = %q, want = %q", got, want)
+ }
+ e.mu.Unlock()
+
+ wantProbes := []entryTestProbeInfo{
+ {
+ RemoteAddress: entryTestAddr1,
+ RemoteLinkAddress: tcpip.LinkAddress(""),
+ LocalAddress: entryTestAddr2,
+ },
+ }
+ linkRes.mu.Lock()
+ diff := cmp.Diff(linkRes.probes, wantProbes)
+ linkRes.mu.Unlock()
+ if diff != "" {
+ t.Fatalf("link address resolver probes mismatch (-got, +want):\n%s", diff)
+ }
+
+ wantEvents := []testEntryEventInfo{
+ {
+ EventType: entryTestAdded,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: tcpip.LinkAddress(""),
+ State: Incomplete,
+ },
+ {
+ EventType: entryTestChanged,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: entryTestLinkAddr1,
+ State: Stale,
+ },
+ }
+ nudDisp.mu.Lock()
+ if diff := cmp.Diff(nudDisp.events, wantEvents, eventDiffOpts()...); diff != "" {
+ t.Errorf("nud dispatcher events mismatch (-got, +want):\n%s", diff)
+ }
+ nudDisp.mu.Unlock()
+}
+
+func TestEntryStaleToReachableWhenSolicitedOverrideConfirmation(t *testing.T) {
+ c := DefaultNUDConfigurations()
+ e, nudDisp, linkRes, _ := entryTestSetup(c)
+
+ e.mu.Lock()
+ e.handlePacketQueuedLocked()
+ e.handleConfirmationLocked(entryTestLinkAddr1, ReachabilityConfirmationFlags{
+ Solicited: false,
+ Override: false,
+ IsRouter: false,
+ })
+ if got, want := e.neigh.State, Stale; got != want {
+ t.Errorf("got e.neigh.State = %q, want = %q", got, want)
+ }
+ e.handleConfirmationLocked(entryTestLinkAddr2, ReachabilityConfirmationFlags{
+ Solicited: true,
+ Override: true,
+ IsRouter: false,
+ })
+ if got, want := e.neigh.State, Reachable; got != want {
+ t.Errorf("got e.neigh.State = %q, want = %q", got, want)
+ }
+ if got, want := e.neigh.LinkAddr, entryTestLinkAddr2; got != want {
+ t.Errorf("got e.neigh.LinkAddr = %q, want = %q", got, want)
+ }
+ e.mu.Unlock()
+
+ wantProbes := []entryTestProbeInfo{
+ {
+ RemoteAddress: entryTestAddr1,
+ RemoteLinkAddress: tcpip.LinkAddress(""),
+ LocalAddress: entryTestAddr2,
+ },
+ }
+ linkRes.mu.Lock()
+ diff := cmp.Diff(linkRes.probes, wantProbes)
+ linkRes.mu.Unlock()
+ if diff != "" {
+ t.Fatalf("link address resolver probes mismatch (-got, +want):\n%s", diff)
+ }
+
+ wantEvents := []testEntryEventInfo{
+ {
+ EventType: entryTestAdded,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: tcpip.LinkAddress(""),
+ State: Incomplete,
+ },
+ {
+ EventType: entryTestChanged,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: entryTestLinkAddr1,
+ State: Stale,
+ },
+ {
+ EventType: entryTestChanged,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: entryTestLinkAddr2,
+ State: Reachable,
+ },
+ }
+ nudDisp.mu.Lock()
+ if diff := cmp.Diff(nudDisp.events, wantEvents, eventDiffOpts()...); diff != "" {
+ t.Errorf("nud dispatcher events mismatch (-got, +want):\n%s", diff)
+ }
+ nudDisp.mu.Unlock()
+}
+
+func TestEntryStaleToStaleWhenOverrideConfirmation(t *testing.T) {
+ c := DefaultNUDConfigurations()
+ e, nudDisp, linkRes, _ := entryTestSetup(c)
+
+ e.mu.Lock()
+ e.handlePacketQueuedLocked()
+ e.handleConfirmationLocked(entryTestLinkAddr1, ReachabilityConfirmationFlags{
+ Solicited: false,
+ Override: false,
+ IsRouter: false,
+ })
+ if got, want := e.neigh.State, Stale; got != want {
+ t.Errorf("got e.neigh.State = %q, want = %q", got, want)
+ }
+ e.handleConfirmationLocked(entryTestLinkAddr2, ReachabilityConfirmationFlags{
+ Solicited: false,
+ Override: true,
+ IsRouter: false,
+ })
+ if got, want := e.neigh.State, Stale; got != want {
+ t.Errorf("got e.neigh.State = %q, want = %q", got, want)
+ }
+ if got, want := e.neigh.LinkAddr, entryTestLinkAddr2; got != want {
+ t.Errorf("got e.neigh.LinkAddr = %q, want = %q", got, want)
+ }
+ e.mu.Unlock()
+
+ wantProbes := []entryTestProbeInfo{
+ {
+ RemoteAddress: entryTestAddr1,
+ RemoteLinkAddress: tcpip.LinkAddress(""),
+ LocalAddress: entryTestAddr2,
+ },
+ }
+ linkRes.mu.Lock()
+ diff := cmp.Diff(linkRes.probes, wantProbes)
+ linkRes.mu.Unlock()
+ if diff != "" {
+ t.Fatalf("link address resolver probes mismatch (-got, +want):\n%s", diff)
+ }
+
+ wantEvents := []testEntryEventInfo{
+ {
+ EventType: entryTestAdded,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: tcpip.LinkAddress(""),
+ State: Incomplete,
+ },
+ {
+ EventType: entryTestChanged,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: entryTestLinkAddr1,
+ State: Stale,
+ },
+ {
+ EventType: entryTestChanged,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: entryTestLinkAddr2,
+ State: Stale,
+ },
+ }
+ nudDisp.mu.Lock()
+ if diff := cmp.Diff(nudDisp.events, wantEvents, eventDiffOpts()...); diff != "" {
+ t.Errorf("nud dispatcher events mismatch (-got, +want):\n%s", diff)
+ }
+ nudDisp.mu.Unlock()
+}
+
+func TestEntryStaleToStaleWhenProbeUpdateAddress(t *testing.T) {
+ c := DefaultNUDConfigurations()
+ e, nudDisp, linkRes, _ := entryTestSetup(c)
+
+ e.mu.Lock()
+ e.handlePacketQueuedLocked()
+ e.handleConfirmationLocked(entryTestLinkAddr1, ReachabilityConfirmationFlags{
+ Solicited: false,
+ Override: false,
+ IsRouter: false,
+ })
+ if got, want := e.neigh.State, Stale; got != want {
+ t.Errorf("got e.neigh.State = %q, want = %q", got, want)
+ }
+ e.handleProbeLocked(entryTestLinkAddr2)
+ if got, want := e.neigh.State, Stale; got != want {
+ t.Errorf("got e.neigh.State = %q, want = %q", got, want)
+ }
+ if got, want := e.neigh.LinkAddr, entryTestLinkAddr2; got != want {
+ t.Errorf("got e.neigh.LinkAddr = %q, want = %q", got, want)
+ }
+ e.mu.Unlock()
+
+ wantProbes := []entryTestProbeInfo{
+ {
+ RemoteAddress: entryTestAddr1,
+ RemoteLinkAddress: tcpip.LinkAddress(""),
+ LocalAddress: entryTestAddr2,
+ },
+ }
+ linkRes.mu.Lock()
+ diff := cmp.Diff(linkRes.probes, wantProbes)
+ linkRes.mu.Unlock()
+ if diff != "" {
+ t.Fatalf("link address resolver probes mismatch (-got, +want):\n%s", diff)
+ }
+
+ wantEvents := []testEntryEventInfo{
+ {
+ EventType: entryTestAdded,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: tcpip.LinkAddress(""),
+ State: Incomplete,
+ },
+ {
+ EventType: entryTestChanged,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: entryTestLinkAddr1,
+ State: Stale,
+ },
+ {
+ EventType: entryTestChanged,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: entryTestLinkAddr2,
+ State: Stale,
+ },
+ }
+ nudDisp.mu.Lock()
+ if diff := cmp.Diff(nudDisp.events, wantEvents, eventDiffOpts()...); diff != "" {
+ t.Errorf("nud dispatcher events mismatch (-got, +want):\n%s", diff)
+ }
+ nudDisp.mu.Unlock()
+}
+
+func TestEntryStaleToDelay(t *testing.T) {
+ c := DefaultNUDConfigurations()
+ e, nudDisp, linkRes, _ := entryTestSetup(c)
+
+ e.mu.Lock()
+ e.handlePacketQueuedLocked()
+ e.handleConfirmationLocked(entryTestLinkAddr1, ReachabilityConfirmationFlags{
+ Solicited: false,
+ Override: false,
+ IsRouter: false,
+ })
+ if got, want := e.neigh.State, Stale; got != want {
+ t.Errorf("got e.neigh.State = %q, want = %q", got, want)
+ }
+ e.handlePacketQueuedLocked()
+ if got, want := e.neigh.State, Delay; got != want {
+ t.Errorf("got e.neigh.State = %q, want = %q", got, want)
+ }
+ e.mu.Unlock()
+
+ wantProbes := []entryTestProbeInfo{
+ {
+ RemoteAddress: entryTestAddr1,
+ RemoteLinkAddress: tcpip.LinkAddress(""),
+ LocalAddress: entryTestAddr2,
+ },
+ }
+ linkRes.mu.Lock()
+ diff := cmp.Diff(linkRes.probes, wantProbes)
+ linkRes.mu.Unlock()
+ if diff != "" {
+ t.Fatalf("link address resolver probes mismatch (-got, +want):\n%s", diff)
+ }
+
+ wantEvents := []testEntryEventInfo{
+ {
+ EventType: entryTestAdded,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: tcpip.LinkAddress(""),
+ State: Incomplete,
+ },
+ {
+ EventType: entryTestChanged,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: entryTestLinkAddr1,
+ State: Stale,
+ },
+ {
+ EventType: entryTestChanged,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: entryTestLinkAddr1,
+ State: Delay,
+ },
+ }
+ nudDisp.mu.Lock()
+ if diff := cmp.Diff(nudDisp.events, wantEvents, eventDiffOpts()...); diff != "" {
+ t.Errorf("nud dispatcher events mismatch (-got, +want):\n%s", diff)
+ }
+ nudDisp.mu.Unlock()
+}
+
+func TestEntryDelayToReachableWhenUpperLevelConfirmation(t *testing.T) {
+ c := DefaultNUDConfigurations()
+ // Eliminate random factors from ReachableTime computation so the transition
+ // from Stale to Reachable will only take BaseReachableTime duration.
+ c.MinRandomFactor = 1
+ c.MaxRandomFactor = 1
+
+ e, nudDisp, linkRes, clock := entryTestSetup(c)
+
+ e.mu.Lock()
+ e.handlePacketQueuedLocked()
+ e.handleConfirmationLocked(entryTestLinkAddr1, ReachabilityConfirmationFlags{
+ Solicited: false,
+ Override: false,
+ IsRouter: false,
+ })
+ e.handlePacketQueuedLocked()
+ if got, want := e.neigh.State, Delay; got != want {
+ t.Errorf("got e.neigh.State = %q, want = %q", got, want)
+ }
+ e.handleUpperLevelConfirmationLocked()
+ if got, want := e.neigh.State, Reachable; got != want {
+ t.Errorf("got e.neigh.State = %q, want = %q", got, want)
+ }
+ e.mu.Unlock()
+
+ wantProbes := []entryTestProbeInfo{
+ {
+ RemoteAddress: entryTestAddr1,
+ RemoteLinkAddress: tcpip.LinkAddress(""),
+ LocalAddress: entryTestAddr2,
+ },
+ }
+ linkRes.mu.Lock()
+ diff := cmp.Diff(linkRes.probes, wantProbes)
+ linkRes.mu.Unlock()
+ if diff != "" {
+ t.Fatalf("link address resolver probes mismatch (-got, +want):\n%s", diff)
+ }
+
+ clock.advance(c.BaseReachableTime)
+
+ wantEvents := []testEntryEventInfo{
+ {
+ EventType: entryTestAdded,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: tcpip.LinkAddress(""),
+ State: Incomplete,
+ },
+ {
+ EventType: entryTestChanged,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: entryTestLinkAddr1,
+ State: Stale,
+ },
+ {
+ EventType: entryTestChanged,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: entryTestLinkAddr1,
+ State: Delay,
+ },
+ {
+ EventType: entryTestChanged,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: entryTestLinkAddr1,
+ State: Reachable,
+ },
+ {
+ EventType: entryTestChanged,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: entryTestLinkAddr1,
+ State: Stale,
+ },
+ }
+ nudDisp.mu.Lock()
+ if diff := cmp.Diff(nudDisp.events, wantEvents, eventDiffOpts()...); diff != "" {
+ t.Errorf("nud dispatcher events mismatch (-got, +want):\n%s", diff)
+ }
+ nudDisp.mu.Unlock()
+}
+
+func TestEntryDelayToReachableWhenSolicitedOverrideConfirmation(t *testing.T) {
+ c := DefaultNUDConfigurations()
+ c.MaxMulticastProbes = 1
+ // Eliminate random factors from ReachableTime computation so the transition
+ // from Stale to Reachable will only take BaseReachableTime duration.
+ c.MinRandomFactor = 1
+ c.MaxRandomFactor = 1
+
+ e, nudDisp, linkRes, clock := entryTestSetup(c)
+
+ e.mu.Lock()
+ e.handlePacketQueuedLocked()
+ e.handleConfirmationLocked(entryTestLinkAddr1, ReachabilityConfirmationFlags{
+ Solicited: false,
+ Override: false,
+ IsRouter: false,
+ })
+ e.handlePacketQueuedLocked()
+ if got, want := e.neigh.State, Delay; got != want {
+ t.Errorf("got e.neigh.State = %q, want = %q", got, want)
+ }
+ e.handleConfirmationLocked(entryTestLinkAddr2, ReachabilityConfirmationFlags{
+ Solicited: true,
+ Override: true,
+ IsRouter: false,
+ })
+ if got, want := e.neigh.State, Reachable; got != want {
+ t.Errorf("got e.neigh.State = %q, want = %q", got, want)
+ }
+ if got, want := e.neigh.LinkAddr, entryTestLinkAddr2; got != want {
+ t.Errorf("got e.neigh.LinkAddr = %q, want = %q", got, want)
+ }
+ e.mu.Unlock()
+
+ wantProbes := []entryTestProbeInfo{
+ {
+ RemoteAddress: entryTestAddr1,
+ RemoteLinkAddress: tcpip.LinkAddress(""),
+ LocalAddress: entryTestAddr2,
+ },
+ }
+ linkRes.mu.Lock()
+ diff := cmp.Diff(linkRes.probes, wantProbes)
+ linkRes.mu.Unlock()
+ if diff != "" {
+ t.Fatalf("link address resolver probes mismatch (-got, +want):\n%s", diff)
+ }
+
+ clock.advance(c.BaseReachableTime)
+
+ wantEvents := []testEntryEventInfo{
+ {
+ EventType: entryTestAdded,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: tcpip.LinkAddress(""),
+ State: Incomplete,
+ },
+ {
+ EventType: entryTestChanged,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: entryTestLinkAddr1,
+ State: Stale,
+ },
+ {
+ EventType: entryTestChanged,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: entryTestLinkAddr1,
+ State: Delay,
+ },
+ {
+ EventType: entryTestChanged,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: entryTestLinkAddr2,
+ State: Reachable,
+ },
+ {
+ EventType: entryTestChanged,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: entryTestLinkAddr2,
+ State: Stale,
+ },
+ }
+ nudDisp.mu.Lock()
+ if diff := cmp.Diff(nudDisp.events, wantEvents, eventDiffOpts()...); diff != "" {
+ t.Errorf("nud dispatcher events mismatch (-got, +want):\n%s", diff)
+ }
+ nudDisp.mu.Unlock()
+}
+
+func TestEntryStaysDelayWhenOverrideConfirmationWithSameAddress(t *testing.T) {
+ c := DefaultNUDConfigurations()
+ e, nudDisp, linkRes, _ := entryTestSetup(c)
+
+ e.mu.Lock()
+ e.handlePacketQueuedLocked()
+ e.handleConfirmationLocked(entryTestLinkAddr1, ReachabilityConfirmationFlags{
+ Solicited: false,
+ Override: false,
+ IsRouter: false,
+ })
+ e.handlePacketQueuedLocked()
+ if got, want := e.neigh.State, Delay; got != want {
+ t.Errorf("got e.neigh.State = %q, want = %q", got, want)
+ }
+ e.handleConfirmationLocked(entryTestLinkAddr1, ReachabilityConfirmationFlags{
+ Solicited: false,
+ Override: true,
+ IsRouter: false,
+ })
+ if got, want := e.neigh.State, Delay; got != want {
+ t.Errorf("got e.neigh.State = %q, want = %q", got, want)
+ }
+ if got, want := e.neigh.LinkAddr, entryTestLinkAddr1; got != want {
+ t.Errorf("got e.neigh.LinkAddr = %q, want = %q", got, want)
+ }
+ e.mu.Unlock()
+
+ wantProbes := []entryTestProbeInfo{
+ {
+ RemoteAddress: entryTestAddr1,
+ RemoteLinkAddress: tcpip.LinkAddress(""),
+ LocalAddress: entryTestAddr2,
+ },
+ }
+ linkRes.mu.Lock()
+ diff := cmp.Diff(linkRes.probes, wantProbes)
+ linkRes.mu.Unlock()
+ if diff != "" {
+ t.Fatalf("link address resolver probes mismatch (-got, +want):\n%s", diff)
+ }
+
+ wantEvents := []testEntryEventInfo{
+ {
+ EventType: entryTestAdded,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: tcpip.LinkAddress(""),
+ State: Incomplete,
+ },
+ {
+ EventType: entryTestChanged,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: entryTestLinkAddr1,
+ State: Stale,
+ },
+ {
+ EventType: entryTestChanged,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: entryTestLinkAddr1,
+ State: Delay,
+ },
+ }
+ nudDisp.mu.Lock()
+ if diff := cmp.Diff(nudDisp.events, wantEvents, eventDiffOpts()...); diff != "" {
+ t.Errorf("nud dispatcher events mismatch (-got, +want):\n%s", diff)
+ }
+ nudDisp.mu.Unlock()
+}
+
+func TestEntryDelayToStaleWhenProbeWithDifferentAddress(t *testing.T) {
+ c := DefaultNUDConfigurations()
+ e, nudDisp, linkRes, _ := entryTestSetup(c)
+
+ e.mu.Lock()
+ e.handlePacketQueuedLocked()
+ e.handleConfirmationLocked(entryTestLinkAddr1, ReachabilityConfirmationFlags{
+ Solicited: false,
+ Override: false,
+ IsRouter: false,
+ })
+ e.handlePacketQueuedLocked()
+ if got, want := e.neigh.State, Delay; got != want {
+ t.Errorf("got e.neigh.State = %q, want = %q", got, want)
+ }
+ e.handleProbeLocked(entryTestLinkAddr2)
+ if got, want := e.neigh.State, Stale; got != want {
+ t.Errorf("got e.neigh.State = %q, want = %q", got, want)
+ }
+ e.mu.Unlock()
+
+ wantProbes := []entryTestProbeInfo{
+ {
+ RemoteAddress: entryTestAddr1,
+ RemoteLinkAddress: tcpip.LinkAddress(""),
+ LocalAddress: entryTestAddr2,
+ },
+ }
+ linkRes.mu.Lock()
+ diff := cmp.Diff(linkRes.probes, wantProbes)
+ linkRes.mu.Unlock()
+ if diff != "" {
+ t.Fatalf("link address resolver probes mismatch (-got, +want):\n%s", diff)
+ }
+
+ wantEvents := []testEntryEventInfo{
+ {
+ EventType: entryTestAdded,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: tcpip.LinkAddress(""),
+ State: Incomplete,
+ },
+ {
+ EventType: entryTestChanged,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: entryTestLinkAddr1,
+ State: Stale,
+ },
+ {
+ EventType: entryTestChanged,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: entryTestLinkAddr1,
+ State: Delay,
+ },
+ {
+ EventType: entryTestChanged,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: entryTestLinkAddr2,
+ State: Stale,
+ },
+ }
+ nudDisp.mu.Lock()
+ if diff := cmp.Diff(nudDisp.events, wantEvents, eventDiffOpts()...); diff != "" {
+ t.Errorf("nud dispatcher events mismatch (-got, +want):\n%s", diff)
+ }
+ nudDisp.mu.Unlock()
+}
+
+func TestEntryDelayToStaleWhenConfirmationWithDifferentAddress(t *testing.T) {
+ c := DefaultNUDConfigurations()
+ e, nudDisp, linkRes, _ := entryTestSetup(c)
+
+ e.mu.Lock()
+ e.handlePacketQueuedLocked()
+ e.handleConfirmationLocked(entryTestLinkAddr1, ReachabilityConfirmationFlags{
+ Solicited: false,
+ Override: false,
+ IsRouter: false,
+ })
+ e.handlePacketQueuedLocked()
+ if got, want := e.neigh.State, Delay; got != want {
+ t.Errorf("got e.neigh.State = %q, want = %q", got, want)
+ }
+ e.handleConfirmationLocked(entryTestLinkAddr2, ReachabilityConfirmationFlags{
+ Solicited: false,
+ Override: true,
+ IsRouter: false,
+ })
+ if got, want := e.neigh.State, Stale; got != want {
+ t.Errorf("got e.neigh.State = %q, want = %q", got, want)
+ }
+ e.mu.Unlock()
+
+ wantProbes := []entryTestProbeInfo{
+ {
+ RemoteAddress: entryTestAddr1,
+ RemoteLinkAddress: tcpip.LinkAddress(""),
+ LocalAddress: entryTestAddr2,
+ },
+ }
+ linkRes.mu.Lock()
+ diff := cmp.Diff(linkRes.probes, wantProbes)
+ linkRes.mu.Unlock()
+ if diff != "" {
+ t.Fatalf("link address resolver probes mismatch (-got, +want):\n%s", diff)
+ }
+
+ wantEvents := []testEntryEventInfo{
+ {
+ EventType: entryTestAdded,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: tcpip.LinkAddress(""),
+ State: Incomplete,
+ },
+ {
+ EventType: entryTestChanged,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: entryTestLinkAddr1,
+ State: Stale,
+ },
+ {
+ EventType: entryTestChanged,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: entryTestLinkAddr1,
+ State: Delay,
+ },
+ {
+ EventType: entryTestChanged,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: entryTestLinkAddr2,
+ State: Stale,
+ },
+ }
+ nudDisp.mu.Lock()
+ if diff := cmp.Diff(nudDisp.events, wantEvents, eventDiffOpts()...); diff != "" {
+ t.Errorf("nud dispatcher events mismatch (-got, +want):\n%s", diff)
+ }
+ nudDisp.mu.Unlock()
+}
+
+func TestEntryDelayToProbe(t *testing.T) {
+ c := DefaultNUDConfigurations()
+ e, nudDisp, linkRes, clock := entryTestSetup(c)
+
+ e.mu.Lock()
+ e.handlePacketQueuedLocked()
+ e.handleConfirmationLocked(entryTestLinkAddr1, ReachabilityConfirmationFlags{
+ Solicited: false,
+ Override: false,
+ IsRouter: false,
+ })
+ e.handlePacketQueuedLocked()
+ if got, want := e.neigh.State, Delay; got != want {
+ t.Errorf("got e.neigh.State = %q, want = %q", got, want)
+ }
+ e.mu.Unlock()
+
+ clock.advance(c.DelayFirstProbeTime)
+
+ wantProbes := []entryTestProbeInfo{
+ // The first probe is caused by the Unknown-to-Incomplete transition.
+ {
+ RemoteAddress: entryTestAddr1,
+ RemoteLinkAddress: tcpip.LinkAddress(""),
+ LocalAddress: entryTestAddr2,
+ },
+ // The second probe is caused by the Delay-to-Probe transition.
+ {
+ RemoteAddress: entryTestAddr1,
+ RemoteLinkAddress: entryTestLinkAddr1,
+ LocalAddress: entryTestAddr2,
+ },
+ }
+ linkRes.mu.Lock()
+ diff := cmp.Diff(linkRes.probes, wantProbes)
+ linkRes.mu.Unlock()
+ if diff != "" {
+ t.Fatalf("link address resolver probes mismatch (-got, +want):\n%s", diff)
+ }
+
+ wantEvents := []testEntryEventInfo{
+ {
+ EventType: entryTestAdded,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: tcpip.LinkAddress(""),
+ State: Incomplete,
+ },
+ {
+ EventType: entryTestChanged,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: entryTestLinkAddr1,
+ State: Stale,
+ },
+ {
+ EventType: entryTestChanged,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: entryTestLinkAddr1,
+ State: Delay,
+ },
+ {
+ EventType: entryTestChanged,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: entryTestLinkAddr1,
+ State: Probe,
+ },
+ }
+ nudDisp.mu.Lock()
+ if diff := cmp.Diff(nudDisp.events, wantEvents, eventDiffOpts()...); diff != "" {
+ t.Errorf("nud dispatcher events mismatch (-got, +want):\n%s", diff)
+ }
+ nudDisp.mu.Unlock()
+
+ e.mu.Lock()
+ if got, want := e.neigh.State, Probe; got != want {
+ t.Errorf("got e.neigh.State = %q, want = %q", got, want)
+ }
+ e.mu.Unlock()
+}
+
+func TestEntryProbeToStaleWhenProbeWithDifferentAddress(t *testing.T) {
+ c := DefaultNUDConfigurations()
+ e, nudDisp, linkRes, clock := entryTestSetup(c)
+
+ e.mu.Lock()
+ e.handlePacketQueuedLocked()
+ e.handleConfirmationLocked(entryTestLinkAddr1, ReachabilityConfirmationFlags{
+ Solicited: false,
+ Override: false,
+ IsRouter: false,
+ })
+ e.handlePacketQueuedLocked()
+ e.mu.Unlock()
+
+ clock.advance(c.DelayFirstProbeTime)
+
+ wantProbes := []entryTestProbeInfo{
+ // The first probe is caused by the Unknown-to-Incomplete transition.
+ {
+ RemoteAddress: entryTestAddr1,
+ RemoteLinkAddress: tcpip.LinkAddress(""),
+ LocalAddress: entryTestAddr2,
+ },
+ // The second probe is caused by the Delay-to-Probe transition.
+ {
+ RemoteAddress: entryTestAddr1,
+ RemoteLinkAddress: entryTestLinkAddr1,
+ LocalAddress: entryTestAddr2,
+ },
+ }
+ linkRes.mu.Lock()
+ diff := cmp.Diff(linkRes.probes, wantProbes)
+ linkRes.mu.Unlock()
+ if diff != "" {
+ t.Fatalf("link address resolver probes mismatch (-got, +want):\n%s", diff)
+ }
+
+ e.mu.Lock()
+ if got, want := e.neigh.State, Probe; got != want {
+ t.Errorf("got e.neigh.State = %q, want = %q", got, want)
+ }
+ e.handleProbeLocked(entryTestLinkAddr2)
+ if got, want := e.neigh.State, Stale; got != want {
+ t.Errorf("got e.neigh.State = %q, want = %q", got, want)
+ }
+ e.mu.Unlock()
+
+ wantEvents := []testEntryEventInfo{
+ {
+ EventType: entryTestAdded,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: tcpip.LinkAddress(""),
+ State: Incomplete,
+ },
+ {
+ EventType: entryTestChanged,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: entryTestLinkAddr1,
+ State: Stale,
+ },
+ {
+ EventType: entryTestChanged,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: entryTestLinkAddr1,
+ State: Delay,
+ },
+ {
+ EventType: entryTestChanged,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: entryTestLinkAddr1,
+ State: Probe,
+ },
+ {
+ EventType: entryTestChanged,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: entryTestLinkAddr2,
+ State: Stale,
+ },
+ }
+ nudDisp.mu.Lock()
+ if diff := cmp.Diff(nudDisp.events, wantEvents, eventDiffOpts()...); diff != "" {
+ t.Errorf("nud dispatcher events mismatch (-got, +want):\n%s", diff)
+ }
+ nudDisp.mu.Unlock()
+
+ e.mu.Lock()
+ if got, want := e.neigh.State, Stale; got != want {
+ t.Errorf("got e.neigh.State = %q, want = %q", got, want)
+ }
+ e.mu.Unlock()
+}
+
+func TestEntryProbeToStaleWhenConfirmationWithDifferentAddress(t *testing.T) {
+ c := DefaultNUDConfigurations()
+ e, nudDisp, linkRes, clock := entryTestSetup(c)
+
+ e.mu.Lock()
+ e.handlePacketQueuedLocked()
+ e.handleConfirmationLocked(entryTestLinkAddr1, ReachabilityConfirmationFlags{
+ Solicited: false,
+ Override: false,
+ IsRouter: false,
+ })
+ e.handlePacketQueuedLocked()
+ e.mu.Unlock()
+
+ clock.advance(c.DelayFirstProbeTime)
+
+ wantProbes := []entryTestProbeInfo{
+ // The first probe is caused by the Unknown-to-Incomplete transition.
+ {
+ RemoteAddress: entryTestAddr1,
+ RemoteLinkAddress: tcpip.LinkAddress(""),
+ LocalAddress: entryTestAddr2,
+ },
+ // The second probe is caused by the Delay-to-Probe transition.
+ {
+ RemoteAddress: entryTestAddr1,
+ RemoteLinkAddress: entryTestLinkAddr1,
+ LocalAddress: entryTestAddr2,
+ },
+ }
+ linkRes.mu.Lock()
+ diff := cmp.Diff(linkRes.probes, wantProbes)
+ linkRes.mu.Unlock()
+ if diff != "" {
+ t.Fatalf("link address resolver probes mismatch (-got, +want):\n%s", diff)
+ }
+
+ e.mu.Lock()
+ if got, want := e.neigh.State, Probe; got != want {
+ t.Errorf("got e.neigh.State = %q, want = %q", got, want)
+ }
+ e.handleConfirmationLocked(entryTestLinkAddr2, ReachabilityConfirmationFlags{
+ Solicited: false,
+ Override: true,
+ IsRouter: false,
+ })
+ if got, want := e.neigh.State, Stale; got != want {
+ t.Errorf("got e.neigh.State = %q, want = %q", got, want)
+ }
+ e.mu.Unlock()
+
+ wantEvents := []testEntryEventInfo{
+ {
+ EventType: entryTestAdded,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: tcpip.LinkAddress(""),
+ State: Incomplete,
+ },
+ {
+ EventType: entryTestChanged,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: entryTestLinkAddr1,
+ State: Stale,
+ },
+ {
+ EventType: entryTestChanged,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: entryTestLinkAddr1,
+ State: Delay,
+ },
+ {
+ EventType: entryTestChanged,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: entryTestLinkAddr1,
+ State: Probe,
+ },
+ {
+ EventType: entryTestChanged,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: entryTestLinkAddr2,
+ State: Stale,
+ },
+ }
+ nudDisp.mu.Lock()
+ if diff := cmp.Diff(nudDisp.events, wantEvents, eventDiffOpts()...); diff != "" {
+ t.Errorf("nud dispatcher events mismatch (-got, +want):\n%s", diff)
+ }
+ nudDisp.mu.Unlock()
+
+ e.mu.Lock()
+ if got, want := e.neigh.State, Stale; got != want {
+ t.Errorf("got e.neigh.State = %q, want = %q", got, want)
+ }
+ e.mu.Unlock()
+}
+
+func TestEntryStaysProbeWhenOverrideConfirmationWithSameAddress(t *testing.T) {
+ c := DefaultNUDConfigurations()
+ e, nudDisp, linkRes, clock := entryTestSetup(c)
+
+ e.mu.Lock()
+ e.handlePacketQueuedLocked()
+ e.handleConfirmationLocked(entryTestLinkAddr1, ReachabilityConfirmationFlags{
+ Solicited: false,
+ Override: false,
+ IsRouter: false,
+ })
+ e.handlePacketQueuedLocked()
+ e.mu.Unlock()
+
+ clock.advance(c.DelayFirstProbeTime)
+
+ wantProbes := []entryTestProbeInfo{
+ // The first probe is caused by the Unknown-to-Incomplete transition.
+ {
+ RemoteAddress: entryTestAddr1,
+ RemoteLinkAddress: tcpip.LinkAddress(""),
+ LocalAddress: entryTestAddr2,
+ },
+ // The second probe is caused by the Delay-to-Probe transition.
+ {
+ RemoteAddress: entryTestAddr1,
+ RemoteLinkAddress: entryTestLinkAddr1,
+ LocalAddress: entryTestAddr2,
+ },
+ }
+ linkRes.mu.Lock()
+ diff := cmp.Diff(linkRes.probes, wantProbes)
+ linkRes.mu.Unlock()
+ if diff != "" {
+ t.Fatalf("link address resolver probes mismatch (-got, +want):\n%s", diff)
+ }
+
+ e.mu.Lock()
+ if got, want := e.neigh.State, Probe; got != want {
+ t.Errorf("got e.neigh.State = %q, want = %q", got, want)
+ }
+ e.handleConfirmationLocked(entryTestLinkAddr1, ReachabilityConfirmationFlags{
+ Solicited: false,
+ Override: true,
+ IsRouter: false,
+ })
+ if got, want := e.neigh.State, Probe; got != want {
+ t.Errorf("got e.neigh.State = %q, want = %q", got, want)
+ }
+ if got, want := e.neigh.LinkAddr, entryTestLinkAddr1; got != want {
+ t.Errorf("got e.neigh.LinkAddr = %q, want = %q", got, want)
+ }
+ e.mu.Unlock()
+
+ wantEvents := []testEntryEventInfo{
+ {
+ EventType: entryTestAdded,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: tcpip.LinkAddress(""),
+ State: Incomplete,
+ },
+ {
+ EventType: entryTestChanged,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: entryTestLinkAddr1,
+ State: Stale,
+ },
+ {
+ EventType: entryTestChanged,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: entryTestLinkAddr1,
+ State: Delay,
+ },
+ {
+ EventType: entryTestChanged,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: entryTestLinkAddr1,
+ State: Probe,
+ },
+ }
+ nudDisp.mu.Lock()
+ if diff := cmp.Diff(nudDisp.events, wantEvents, eventDiffOpts()...); diff != "" {
+ t.Errorf("nud dispatcher events mismatch (-got, +want):\n%s", diff)
+ }
+ nudDisp.mu.Unlock()
+}
+
+func TestEntryProbeToReachableWhenSolicitedOverrideConfirmation(t *testing.T) {
+ c := DefaultNUDConfigurations()
+ // Eliminate random factors from ReachableTime computation so the transition
+ // from Stale to Reachable will only take BaseReachableTime duration.
+ c.MinRandomFactor = 1
+ c.MaxRandomFactor = 1
+
+ e, nudDisp, linkRes, clock := entryTestSetup(c)
+
+ e.mu.Lock()
+ e.handlePacketQueuedLocked()
+ e.handleConfirmationLocked(entryTestLinkAddr1, ReachabilityConfirmationFlags{
+ Solicited: false,
+ Override: false,
+ IsRouter: false,
+ })
+ e.handlePacketQueuedLocked()
+ e.mu.Unlock()
+
+ clock.advance(c.DelayFirstProbeTime)
+
+ wantProbes := []entryTestProbeInfo{
+ // The first probe is caused by the Unknown-to-Incomplete transition.
+ {
+ RemoteAddress: entryTestAddr1,
+ RemoteLinkAddress: tcpip.LinkAddress(""),
+ LocalAddress: entryTestAddr2,
+ },
+ // The second probe is caused by the Delay-to-Probe transition.
+ {
+ RemoteAddress: entryTestAddr1,
+ RemoteLinkAddress: entryTestLinkAddr1,
+ LocalAddress: entryTestAddr2,
+ },
+ }
+ linkRes.mu.Lock()
+ diff := cmp.Diff(linkRes.probes, wantProbes)
+ linkRes.mu.Unlock()
+ if diff != "" {
+ t.Fatalf("link address resolver probes mismatch (-got, +want):\n%s", diff)
+ }
+
+ e.mu.Lock()
+ if got, want := e.neigh.State, Probe; got != want {
+ t.Errorf("got e.neigh.State = %q, want = %q", got, want)
+ }
+ e.handleConfirmationLocked(entryTestLinkAddr2, ReachabilityConfirmationFlags{
+ Solicited: true,
+ Override: true,
+ IsRouter: false,
+ })
+ if got, want := e.neigh.State, Reachable; got != want {
+ t.Errorf("got e.neigh.State = %q, want = %q", got, want)
+ }
+ if got, want := e.neigh.LinkAddr, entryTestLinkAddr2; got != want {
+ t.Errorf("got e.neigh.LinkAddr = %q, want = %q", got, want)
+ }
+ e.mu.Unlock()
+
+ clock.advance(c.BaseReachableTime)
+
+ wantEvents := []testEntryEventInfo{
+ {
+ EventType: entryTestAdded,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: tcpip.LinkAddress(""),
+ State: Incomplete,
+ },
+ {
+ EventType: entryTestChanged,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: entryTestLinkAddr1,
+ State: Stale,
+ },
+ {
+ EventType: entryTestChanged,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: entryTestLinkAddr1,
+ State: Delay,
+ },
+ {
+ EventType: entryTestChanged,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: entryTestLinkAddr1,
+ State: Probe,
+ },
+ {
+ EventType: entryTestChanged,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: entryTestLinkAddr2,
+ State: Reachable,
+ },
+ {
+ EventType: entryTestChanged,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: entryTestLinkAddr2,
+ State: Stale,
+ },
+ }
+ nudDisp.mu.Lock()
+ if diff := cmp.Diff(nudDisp.events, wantEvents, eventDiffOpts()...); diff != "" {
+ t.Errorf("nud dispatcher events mismatch (-got, +want):\n%s", diff)
+ }
+ nudDisp.mu.Unlock()
+}
+
+func TestEntryProbeToReachableWhenSolicitedConfirmationWithSameAddress(t *testing.T) {
+ c := DefaultNUDConfigurations()
+ // Eliminate random factors from ReachableTime computation so the transition
+ // from Stale to Reachable will only take BaseReachableTime duration.
+ c.MinRandomFactor = 1
+ c.MaxRandomFactor = 1
+
+ e, nudDisp, linkRes, clock := entryTestSetup(c)
+
+ e.mu.Lock()
+ e.handlePacketQueuedLocked()
+ e.handleConfirmationLocked(entryTestLinkAddr1, ReachabilityConfirmationFlags{
+ Solicited: false,
+ Override: false,
+ IsRouter: false,
+ })
+ e.handlePacketQueuedLocked()
+ e.mu.Unlock()
+
+ clock.advance(c.DelayFirstProbeTime)
+
+ wantProbes := []entryTestProbeInfo{
+ // The first probe is caused by the Unknown-to-Incomplete transition.
+ {
+ RemoteAddress: entryTestAddr1,
+ RemoteLinkAddress: tcpip.LinkAddress(""),
+ LocalAddress: entryTestAddr2,
+ },
+ // The second probe is caused by the Delay-to-Probe transition.
+ {
+ RemoteAddress: entryTestAddr1,
+ RemoteLinkAddress: entryTestLinkAddr1,
+ LocalAddress: entryTestAddr2,
+ },
+ }
+ linkRes.mu.Lock()
+ diff := cmp.Diff(linkRes.probes, wantProbes)
+ linkRes.mu.Unlock()
+ if diff != "" {
+ t.Fatalf("link address resolver probes mismatch (-got, +want):\n%s", diff)
+ }
+
+ e.mu.Lock()
+ if got, want := e.neigh.State, Probe; got != want {
+ t.Errorf("got e.neigh.State = %q, want = %q", got, want)
+ }
+ e.handleConfirmationLocked(entryTestLinkAddr1, ReachabilityConfirmationFlags{
+ Solicited: true,
+ Override: false,
+ IsRouter: false,
+ })
+ if got, want := e.neigh.State, Reachable; got != want {
+ t.Errorf("got e.neigh.State = %q, want = %q", got, want)
+ }
+ e.mu.Unlock()
+
+ clock.advance(c.BaseReachableTime)
+
+ wantEvents := []testEntryEventInfo{
+ {
+ EventType: entryTestAdded,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: tcpip.LinkAddress(""),
+ State: Incomplete,
+ },
+ {
+ EventType: entryTestChanged,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: entryTestLinkAddr1,
+ State: Stale,
+ },
+ {
+ EventType: entryTestChanged,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: entryTestLinkAddr1,
+ State: Delay,
+ },
+ {
+ EventType: entryTestChanged,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: entryTestLinkAddr1,
+ State: Probe,
+ },
+ {
+ EventType: entryTestChanged,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: entryTestLinkAddr1,
+ State: Reachable,
+ },
+ {
+ EventType: entryTestChanged,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: entryTestLinkAddr1,
+ State: Stale,
+ },
+ }
+ nudDisp.mu.Lock()
+ if diff := cmp.Diff(nudDisp.events, wantEvents, eventDiffOpts()...); diff != "" {
+ t.Errorf("nud dispatcher events mismatch (-got, +want):\n%s", diff)
+ }
+ nudDisp.mu.Unlock()
+}
+
+func TestEntryProbeToFailed(t *testing.T) {
+ c := DefaultNUDConfigurations()
+ c.MaxMulticastProbes = 3
+ c.MaxUnicastProbes = 3
+ e, nudDisp, linkRes, clock := entryTestSetup(c)
+
+ e.mu.Lock()
+ e.handlePacketQueuedLocked()
+ e.handleConfirmationLocked(entryTestLinkAddr1, ReachabilityConfirmationFlags{
+ Solicited: false,
+ Override: false,
+ IsRouter: false,
+ })
+ e.handlePacketQueuedLocked()
+ e.mu.Unlock()
+
+ waitFor := c.DelayFirstProbeTime + c.RetransmitTimer*time.Duration(c.MaxUnicastProbes)
+ clock.advance(waitFor)
+
+ wantProbes := []entryTestProbeInfo{
+ // The first probe is caused by the Unknown-to-Incomplete transition.
+ {
+ RemoteAddress: entryTestAddr1,
+ RemoteLinkAddress: tcpip.LinkAddress(""),
+ LocalAddress: entryTestAddr2,
+ },
+ // The next three probe are caused by the Delay-to-Probe transition.
+ {
+ RemoteAddress: entryTestAddr1,
+ RemoteLinkAddress: entryTestLinkAddr1,
+ LocalAddress: entryTestAddr2,
+ },
+ {
+ RemoteAddress: entryTestAddr1,
+ RemoteLinkAddress: entryTestLinkAddr1,
+ LocalAddress: entryTestAddr2,
+ },
+ {
+ RemoteAddress: entryTestAddr1,
+ RemoteLinkAddress: entryTestLinkAddr1,
+ LocalAddress: entryTestAddr2,
+ },
+ }
+ linkRes.mu.Lock()
+ diff := cmp.Diff(linkRes.probes, wantProbes)
+ linkRes.mu.Unlock()
+ if diff != "" {
+ t.Fatalf("link address resolver probes mismatch (-got, +want):\n%s", diff)
+ }
+
+ wantEvents := []testEntryEventInfo{
+ {
+ EventType: entryTestAdded,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: tcpip.LinkAddress(""),
+ State: Incomplete,
+ },
+ {
+ EventType: entryTestChanged,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: entryTestLinkAddr1,
+ State: Stale,
+ },
+ {
+ EventType: entryTestChanged,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: entryTestLinkAddr1,
+ State: Delay,
+ },
+ {
+ EventType: entryTestChanged,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: entryTestLinkAddr1,
+ State: Probe,
+ },
+ {
+ EventType: entryTestRemoved,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: entryTestLinkAddr1,
+ State: Probe,
+ },
+ }
+ nudDisp.mu.Lock()
+ if diff := cmp.Diff(nudDisp.events, wantEvents, eventDiffOpts()...); diff != "" {
+ t.Errorf("nud dispatcher events mismatch (-got, +want):\n%s", diff)
+ }
+ nudDisp.mu.Unlock()
+
+ e.mu.Lock()
+ if got, want := e.neigh.State, Failed; got != want {
+ t.Errorf("got e.neigh.State = %q, want = %q", got, want)
+ }
+ e.mu.Unlock()
+}
+
+func TestEntryFailedGetsDeleted(t *testing.T) {
+ c := DefaultNUDConfigurations()
+ c.MaxMulticastProbes = 3
+ c.MaxUnicastProbes = 3
+ e, nudDisp, linkRes, clock := entryTestSetup(c)
+
+ // Verify the cache contains the entry.
+ if _, ok := e.nic.neigh.cache[entryTestAddr1]; !ok {
+ t.Errorf("expected entry %q to exist in the neighbor cache", entryTestAddr1)
+ }
+
+ e.mu.Lock()
+ e.handlePacketQueuedLocked()
+ e.handleConfirmationLocked(entryTestLinkAddr1, ReachabilityConfirmationFlags{
+ Solicited: false,
+ Override: false,
+ IsRouter: false,
+ })
+ e.handlePacketQueuedLocked()
+ e.mu.Unlock()
+
+ waitFor := c.DelayFirstProbeTime + c.RetransmitTimer*time.Duration(c.MaxUnicastProbes) + c.UnreachableTime
+ clock.advance(waitFor)
+
+ wantProbes := []entryTestProbeInfo{
+ // The first probe is caused by the Unknown-to-Incomplete transition.
+ {
+ RemoteAddress: entryTestAddr1,
+ RemoteLinkAddress: tcpip.LinkAddress(""),
+ LocalAddress: entryTestAddr2,
+ },
+ // The next three probe are caused by the Delay-to-Probe transition.
+ {
+ RemoteAddress: entryTestAddr1,
+ RemoteLinkAddress: entryTestLinkAddr1,
+ LocalAddress: entryTestAddr2,
+ },
+ {
+ RemoteAddress: entryTestAddr1,
+ RemoteLinkAddress: entryTestLinkAddr1,
+ LocalAddress: entryTestAddr2,
+ },
+ {
+ RemoteAddress: entryTestAddr1,
+ RemoteLinkAddress: entryTestLinkAddr1,
+ LocalAddress: entryTestAddr2,
+ },
+ }
+ linkRes.mu.Lock()
+ diff := cmp.Diff(linkRes.probes, wantProbes)
+ linkRes.mu.Unlock()
+ if diff != "" {
+ t.Fatalf("link address resolver probes mismatch (-got, +want):\n%s", diff)
+ }
+
+ wantEvents := []testEntryEventInfo{
+ {
+ EventType: entryTestAdded,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: tcpip.LinkAddress(""),
+ State: Incomplete,
+ },
+ {
+ EventType: entryTestChanged,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: entryTestLinkAddr1,
+ State: Stale,
+ },
+ {
+ EventType: entryTestChanged,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: entryTestLinkAddr1,
+ State: Delay,
+ },
+ {
+ EventType: entryTestChanged,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: entryTestLinkAddr1,
+ State: Probe,
+ },
+ {
+ EventType: entryTestRemoved,
+ NICID: entryTestNICID,
+ Addr: entryTestAddr1,
+ LinkAddr: entryTestLinkAddr1,
+ State: Probe,
+ },
+ }
+ nudDisp.mu.Lock()
+ if diff := cmp.Diff(nudDisp.events, wantEvents, eventDiffOpts()...); diff != "" {
+ t.Errorf("nud dispatcher events mismatch (-got, +want):\n%s", diff)
+ }
+ nudDisp.mu.Unlock()
+
+ // Verify the cache no longer contains the entry.
+ if _, ok := e.nic.neigh.cache[entryTestAddr1]; ok {
+ t.Errorf("entry %q should have been deleted from the neighbor cache", entryTestAddr1)
+ }
+}
diff --git a/pkg/tcpip/stack/neighborstate_string.go b/pkg/tcpip/stack/neighborstate_string.go
new file mode 100644
index 000000000..aa7311ec6
--- /dev/null
+++ b/pkg/tcpip/stack/neighborstate_string.go
@@ -0,0 +1,44 @@
+// Copyright 2020 The gVisor Authors.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+// http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+// Code generated by "stringer -type NeighborState"; DO NOT EDIT.
+
+package stack
+
+import "strconv"
+
+func _() {
+ // An "invalid array index" compiler error signifies that the constant values have changed.
+ // Re-run the stringer command to generate them again.
+ var x [1]struct{}
+ _ = x[Unknown-0]
+ _ = x[Incomplete-1]
+ _ = x[Reachable-2]
+ _ = x[Stale-3]
+ _ = x[Delay-4]
+ _ = x[Probe-5]
+ _ = x[Static-6]
+ _ = x[Failed-7]
+}
+
+const _NeighborState_name = "UnknownIncompleteReachableStaleDelayProbeStaticFailed"
+
+var _NeighborState_index = [...]uint8{0, 7, 17, 26, 31, 36, 41, 47, 53}
+
+func (i NeighborState) String() string {
+ if i >= NeighborState(len(_NeighborState_index)-1) {
+ return "NeighborState(" + strconv.FormatInt(int64(i), 10) + ")"
+ }
+ return _NeighborState_name[_NeighborState_index[i]:_NeighborState_index[i+1]]
+}
diff --git a/pkg/tcpip/stack/nic.go b/pkg/tcpip/stack/nic.go
index 9256d4d43..f21066fce 100644
--- a/pkg/tcpip/stack/nic.go
+++ b/pkg/tcpip/stack/nic.go
@@ -16,6 +16,7 @@ package stack
import (
"fmt"
+ "math/rand"
"reflect"
"sort"
"strings"
@@ -45,6 +46,7 @@ type NIC struct {
context NICContext
stats NICStats
+ neigh *neighborCache
mu struct {
sync.RWMutex
@@ -141,6 +143,16 @@ func newNIC(stack *Stack, id tcpip.NICID, name string, ep LinkEndpoint, ctx NICC
nic.mu.packetEPs[netProto.Number()] = []PacketEndpoint{}
}
+ // Check for Neighbor Unreachability Detection support.
+ if ep.Capabilities()&CapabilityResolutionRequired != 0 && len(stack.linkAddrResolvers) != 0 {
+ rng := rand.New(rand.NewSource(stack.clock.NowNanoseconds()))
+ nic.neigh = &neighborCache{
+ nic: nic,
+ state: NewNUDState(stack.nudConfigs, rng),
+ cache: make(map[tcpip.Address]*neighborEntry, neighborCacheSize),
+ }
+ }
+
nic.linkEP.Attach(nic)
return nic
@@ -1540,6 +1552,27 @@ func (n *NIC) setNDPConfigs(c NDPConfigurations) {
n.mu.Unlock()
}
+// NUDConfigs gets the NUD configurations for n.
+func (n *NIC) NUDConfigs() (NUDConfigurations, *tcpip.Error) {
+ if n.neigh == nil {
+ return NUDConfigurations{}, tcpip.ErrNotSupported
+ }
+ return n.neigh.config(), nil
+}
+
+// setNUDConfigs sets the NUD configurations for n.
+//
+// Note, if c contains invalid NUD configuration values, it will be fixed to
+// use default values for the erroneous values.
+func (n *NIC) setNUDConfigs(c NUDConfigurations) *tcpip.Error {
+ if n.neigh == nil {
+ return tcpip.ErrNotSupported
+ }
+ c.resetInvalidFields()
+ n.neigh.setConfig(c)
+ return nil
+}
+
// handleNDPRA handles an NDP Router Advertisement message that arrived on n.
func (n *NIC) handleNDPRA(ip tcpip.Address, ra header.NDPRouterAdvert) {
n.mu.Lock()
diff --git a/pkg/tcpip/stack/nud.go b/pkg/tcpip/stack/nud.go
new file mode 100644
index 000000000..f848d50ad
--- /dev/null
+++ b/pkg/tcpip/stack/nud.go
@@ -0,0 +1,466 @@
+// Copyright 2020 The gVisor Authors.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+// http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package stack
+
+import (
+ "math"
+ "sync"
+ "time"
+
+ "gvisor.dev/gvisor/pkg/tcpip"
+)
+
+const (
+ // defaultBaseReachableTime is the default base duration for computing the
+ // random reachable time.
+ //
+ // Reachable time is the duration for which a neighbor is considered
+ // reachable after a positive reachability confirmation is received. It is a
+ // function of a uniformly distributed random value between the minimum and
+ // maximum random factors, multiplied by the base reachable time. Using a
+ // random component eliminates the possibility that Neighbor Unreachability
+ // Detection messages will synchronize with each other.
+ //
+ // Default taken from REACHABLE_TIME of RFC 4861 section 10.
+ defaultBaseReachableTime = 30 * time.Second
+
+ // minimumBaseReachableTime is the minimum base duration for computing the
+ // random reachable time.
+ //
+ // Minimum = 1ms
+ minimumBaseReachableTime = time.Millisecond
+
+ // defaultMinRandomFactor is the default minimum value of the random factor
+ // used for computing reachable time.
+ //
+ // Default taken from MIN_RANDOM_FACTOR of RFC 4861 section 10.
+ defaultMinRandomFactor = 0.5
+
+ // defaultMaxRandomFactor is the default maximum value of the random factor
+ // used for computing reachable time.
+ //
+ // The default value depends on the value of MinRandomFactor.
+ // If MinRandomFactor is less than MAX_RANDOM_FACTOR of RFC 4861 section 10,
+ // the value from the RFC will be used; otherwise, the default is
+ // MinRandomFactor multiplied by three.
+ defaultMaxRandomFactor = 1.5
+
+ // defaultRetransmitTimer is the default amount of time to wait between
+ // sending reachability probes.
+ //
+ // Default taken from RETRANS_TIMER of RFC 4861 section 10.
+ defaultRetransmitTimer = time.Second
+
+ // minimumRetransmitTimer is the minimum amount of time to wait between
+ // sending reachability probes.
+ //
+ // Note, RFC 4861 does not impose a minimum Retransmit Timer, but we do here
+ // to make sure the messages are not sent all at once. We also come to this
+ // value because in the RetransmitTimer field of a Router Advertisement, a
+ // value of 0 means unspecified, so the smallest valid value is 1. Note, the
+ // unit of the RetransmitTimer field in the Router Advertisement is
+ // milliseconds.
+ minimumRetransmitTimer = time.Millisecond
+
+ // defaultDelayFirstProbeTime is the default duration to wait for a
+ // non-Neighbor-Discovery related protocol to reconfirm reachability after
+ // entering the DELAY state. After this time, a reachability probe will be
+ // sent and the entry will transition to the PROBE state.
+ //
+ // Default taken from DELAY_FIRST_PROBE_TIME of RFC 4861 section 10.
+ defaultDelayFirstProbeTime = 5 * time.Second
+
+ // defaultMaxMulticastProbes is the default number of reachabililty probes
+ // to send before concluding negative reachability and deleting the neighbor
+ // entry from the INCOMPLETE state.
+ //
+ // Default taken from MAX_MULTICAST_SOLICIT of RFC 4861 section 10.
+ defaultMaxMulticastProbes = 3
+
+ // defaultMaxUnicastProbes is the default number of reachability probes to
+ // send before concluding retransmission from within the PROBE state should
+ // cease and the entry SHOULD be deleted.
+ //
+ // Default taken from MAX_UNICASE_SOLICIT of RFC 4861 section 10.
+ defaultMaxUnicastProbes = 3
+
+ // defaultMaxAnycastDelayTime is the default time in which the stack SHOULD
+ // delay sending a response for a random time between 0 and this time, if the
+ // target address is an anycast address.
+ //
+ // Default taken from MAX_ANYCAST_DELAY_TIME of RFC 4861 section 10.
+ defaultMaxAnycastDelayTime = time.Second
+
+ // defaultMaxReachbilityConfirmations is the default amount of unsolicited
+ // reachability confirmation messages a node MAY send to all-node multicast
+ // address when it determines its link-layer address has changed.
+ //
+ // Default taken from MAX_NEIGHBOR_ADVERTISEMENT of RFC 4861 section 10.
+ defaultMaxReachbilityConfirmations = 3
+
+ // defaultUnreachableTime is the default duration for how long an entry will
+ // remain in the FAILED state before being removed from the neighbor cache.
+ //
+ // Note, there is no equivalent protocol constant defined in RFC 4861. It
+ // leaves the specifics of any garbage collection mechanism up to the
+ // implementation.
+ defaultUnreachableTime = 5 * time.Second
+)
+
+// NUDDispatcher is the interface integrators of netstack must implement to
+// receive and handle NUD related events.
+type NUDDispatcher interface {
+ // OnNeighborAdded will be called when a new entry is added to a NIC's (with
+ // ID nicID) neighbor table.
+ //
+ // This function is permitted to block indefinitely without interfering with
+ // the stack's operation.
+ //
+ // May be called concurrently.
+ OnNeighborAdded(nicID tcpip.NICID, ipAddr tcpip.Address, linkAddr tcpip.LinkAddress, state NeighborState, updatedAt time.Time)
+
+ // OnNeighborChanged will be called when an entry in a NIC's (with ID nicID)
+ // neighbor table changes state and/or link address.
+ //
+ // This function is permitted to block indefinitely without interfering with
+ // the stack's operation.
+ //
+ // May be called concurrently.
+ OnNeighborChanged(nicID tcpip.NICID, ipAddr tcpip.Address, linkAddr tcpip.LinkAddress, state NeighborState, updatedAt time.Time)
+
+ // OnNeighborRemoved will be called when an entry is removed from a NIC's
+ // (with ID nicID) neighbor table.
+ //
+ // This function is permitted to block indefinitely without interfering with
+ // the stack's operation.
+ //
+ // May be called concurrently.
+ OnNeighborRemoved(nicID tcpip.NICID, ipAddr tcpip.Address, linkAddr tcpip.LinkAddress, state NeighborState, updatedAt time.Time)
+}
+
+// ReachabilityConfirmationFlags describes the flags used within a reachability
+// confirmation (e.g. ARP reply or Neighbor Advertisement for ARP or NDP,
+// respectively).
+type ReachabilityConfirmationFlags struct {
+ // Solicited indicates that the advertisement was sent in response to a
+ // reachability probe.
+ Solicited bool
+
+ // Override indicates that the reachability confirmation should override an
+ // existing neighbor cache entry and update the cached link-layer address.
+ // When Override is not set the confirmation will not update a cached
+ // link-layer address, but will update an existing neighbor cache entry for
+ // which no link-layer address is known.
+ Override bool
+
+ // IsRouter indicates that the sender is a router.
+ IsRouter bool
+}
+
+// NUDHandler communicates external events to the Neighbor Unreachability
+// Detection state machine, which is implemented per-interface. This is used by
+// network endpoints to inform the Neighbor Cache of probes and confirmations.
+type NUDHandler interface {
+ // HandleProbe processes an incoming neighbor probe (e.g. ARP request or
+ // Neighbor Solicitation for ARP or NDP, respectively). Validation of the
+ // probe needs to be performed before calling this function since the
+ // Neighbor Cache doesn't have access to view the NIC's assigned addresses.
+ HandleProbe(remoteAddr, localAddr tcpip.Address, protocol tcpip.NetworkProtocolNumber, remoteLinkAddr tcpip.LinkAddress)
+
+ // HandleConfirmation processes an incoming neighbor confirmation (e.g. ARP
+ // reply or Neighbor Advertisement for ARP or NDP, respectively).
+ HandleConfirmation(addr tcpip.Address, linkAddr tcpip.LinkAddress, flags ReachabilityConfirmationFlags)
+
+ // HandleUpperLevelConfirmation processes an incoming upper-level protocol
+ // (e.g. TCP acknowledgements) reachability confirmation.
+ HandleUpperLevelConfirmation(addr tcpip.Address)
+}
+
+// NUDConfigurations is the NUD configurations for the netstack. This is used
+// by the neighbor cache to operate the NUD state machine on each device in the
+// local network.
+type NUDConfigurations struct {
+ // BaseReachableTime is the base duration for computing the random reachable
+ // time.
+ //
+ // Reachable time is the duration for which a neighbor is considered
+ // reachable after a positive reachability confirmation is received. It is a
+ // function of uniformly distributed random value between minRandomFactor and
+ // maxRandomFactor multiplied by baseReachableTime. Using a random component
+ // eliminates the possibility that Neighbor Unreachability Detection messages
+ // will synchronize with each other.
+ //
+ // After this time, a neighbor entry will transition from REACHABLE to STALE
+ // state.
+ //
+ // Must be greater than 0.
+ BaseReachableTime time.Duration
+
+ // LearnBaseReachableTime enables learning BaseReachableTime during runtime
+ // from the neighbor discovery protocol, if supported.
+ //
+ // TODO(gvisor.dev/issue/2240): Implement this NUD configuration option.
+ LearnBaseReachableTime bool
+
+ // MinRandomFactor is the minimum value of the random factor used for
+ // computing reachable time.
+ //
+ // See BaseReachbleTime for more information on computing the reachable time.
+ //
+ // Must be greater than 0.
+ MinRandomFactor float32
+
+ // MaxRandomFactor is the maximum value of the random factor used for
+ // computing reachabile time.
+ //
+ // See BaseReachbleTime for more information on computing the reachable time.
+ //
+ // Must be great than or equal to MinRandomFactor.
+ MaxRandomFactor float32
+
+ // RetransmitTimer is the duration between retransmission of reachability
+ // probes in the PROBE state.
+ RetransmitTimer time.Duration
+
+ // LearnRetransmitTimer enables learning RetransmitTimer during runtime from
+ // the neighbor discovery protocol, if supported.
+ //
+ // TODO(gvisor.dev/issue/2241): Implement this NUD configuration option.
+ LearnRetransmitTimer bool
+
+ // DelayFirstProbeTime is the duration to wait for a non-Neighbor-Discovery
+ // related protocol to reconfirm reachability after entering the DELAY state.
+ // After this time, a reachability probe will be sent and the entry will
+ // transition to the PROBE state.
+ //
+ // Must be greater than 0.
+ DelayFirstProbeTime time.Duration
+
+ // MaxMulticastProbes is the number of reachability probes to send before
+ // concluding negative reachability and deleting the neighbor entry from the
+ // INCOMPLETE state.
+ //
+ // Must be greater than 0.
+ MaxMulticastProbes uint32
+
+ // MaxUnicastProbes is the number of reachability probes to send before
+ // concluding retransmission from within the PROBE state should cease and
+ // entry SHOULD be deleted.
+ //
+ // Must be greater than 0.
+ MaxUnicastProbes uint32
+
+ // MaxAnycastDelayTime is the time in which the stack SHOULD delay sending a
+ // response for a random time between 0 and this time, if the target address
+ // is an anycast address.
+ //
+ // TODO(gvisor.dev/issue/2242): Use this option when sending solicited
+ // neighbor confirmations to anycast addresses and proxying neighbor
+ // confirmations.
+ MaxAnycastDelayTime time.Duration
+
+ // MaxReachabilityConfirmations is the number of unsolicited reachability
+ // confirmation messages a node MAY send to all-node multicast address when
+ // it determines its link-layer address has changed.
+ //
+ // TODO(gvisor.dev/issue/2246): Discuss if implementation of this NUD
+ // configuration option is necessary.
+ MaxReachabilityConfirmations uint32
+
+ // UnreachableTime describes how long an entry will remain in the FAILED
+ // state before being removed from the neighbor cache.
+ UnreachableTime time.Duration
+}
+
+// DefaultNUDConfigurations returns a NUDConfigurations populated with default
+// values defined by RFC 4861 section 10.
+func DefaultNUDConfigurations() NUDConfigurations {
+ return NUDConfigurations{
+ BaseReachableTime: defaultBaseReachableTime,
+ LearnBaseReachableTime: true,
+ MinRandomFactor: defaultMinRandomFactor,
+ MaxRandomFactor: defaultMaxRandomFactor,
+ RetransmitTimer: defaultRetransmitTimer,
+ LearnRetransmitTimer: true,
+ DelayFirstProbeTime: defaultDelayFirstProbeTime,
+ MaxMulticastProbes: defaultMaxMulticastProbes,
+ MaxUnicastProbes: defaultMaxUnicastProbes,
+ MaxAnycastDelayTime: defaultMaxAnycastDelayTime,
+ MaxReachabilityConfirmations: defaultMaxReachbilityConfirmations,
+ UnreachableTime: defaultUnreachableTime,
+ }
+}
+
+// resetInvalidFields modifies an invalid NDPConfigurations with valid values.
+// If invalid values are present in c, the corresponding default values will be
+// used instead. This is needed to check, and conditionally fix, user-specified
+// NUDConfigurations.
+func (c *NUDConfigurations) resetInvalidFields() {
+ if c.BaseReachableTime < minimumBaseReachableTime {
+ c.BaseReachableTime = defaultBaseReachableTime
+ }
+ if c.MinRandomFactor <= 0 {
+ c.MinRandomFactor = defaultMinRandomFactor
+ }
+ if c.MaxRandomFactor < c.MinRandomFactor {
+ c.MaxRandomFactor = calcMaxRandomFactor(c.MinRandomFactor)
+ }
+ if c.RetransmitTimer < minimumRetransmitTimer {
+ c.RetransmitTimer = defaultRetransmitTimer
+ }
+ if c.DelayFirstProbeTime == 0 {
+ c.DelayFirstProbeTime = defaultDelayFirstProbeTime
+ }
+ if c.MaxMulticastProbes == 0 {
+ c.MaxMulticastProbes = defaultMaxMulticastProbes
+ }
+ if c.MaxUnicastProbes == 0 {
+ c.MaxUnicastProbes = defaultMaxUnicastProbes
+ }
+ if c.UnreachableTime == 0 {
+ c.UnreachableTime = defaultUnreachableTime
+ }
+}
+
+// calcMaxRandomFactor calculates the maximum value of the random factor used
+// for computing reachable time. This function is necessary for when the
+// default specified in RFC 4861 section 10 is less than the current
+// MinRandomFactor.
+//
+// Assumes minRandomFactor is positive since validation of the minimum value
+// should come before the validation of the maximum.
+func calcMaxRandomFactor(minRandomFactor float32) float32 {
+ if minRandomFactor > defaultMaxRandomFactor {
+ return minRandomFactor * 3
+ }
+ return defaultMaxRandomFactor
+}
+
+// A Rand is a source of random numbers.
+type Rand interface {
+ // Float32 returns, as a float32, a pseudo-random number in [0.0,1.0).
+ Float32() float32
+}
+
+// NUDState stores states needed for calculating reachable time.
+type NUDState struct {
+ rng Rand
+
+ // mu protects the fields below.
+ //
+ // It is necessary for NUDState to handle its own locking since neighbor
+ // entries may access the NUD state from within the goroutine spawned by
+ // time.AfterFunc(). This goroutine may run concurrently with the main
+ // process for controlling the neighbor cache and would otherwise introduce
+ // race conditions if NUDState was not locked properly.
+ mu sync.RWMutex
+
+ config NUDConfigurations
+
+ // reachableTime is the duration to wait for a REACHABLE entry to
+ // transition into STALE after inactivity. This value is calculated with
+ // the algorithm defined in RFC 4861 section 6.3.2.
+ reachableTime time.Duration
+
+ expiration time.Time
+ prevBaseReachableTime time.Duration
+ prevMinRandomFactor float32
+ prevMaxRandomFactor float32
+}
+
+// NewNUDState returns new NUDState using c as configuration and the specified
+// random number generator for use in recomputing ReachableTime.
+func NewNUDState(c NUDConfigurations, rng Rand) *NUDState {
+ s := &NUDState{
+ rng: rng,
+ }
+ s.config = c
+ return s
+}
+
+// Config returns the NUD configuration.
+func (s *NUDState) Config() NUDConfigurations {
+ s.mu.RLock()
+ defer s.mu.RUnlock()
+ return s.config
+}
+
+// SetConfig replaces the existing NUD configurations with c.
+func (s *NUDState) SetConfig(c NUDConfigurations) {
+ s.mu.Lock()
+ defer s.mu.Unlock()
+ s.config = c
+}
+
+// ReachableTime returns the duration to wait for a REACHABLE entry to
+// transition into STALE after inactivity. This value is recalculated for new
+// values of BaseReachableTime, MinRandomFactor, and MaxRandomFactor using the
+// algorithm defined in RFC 4861 section 6.3.2.
+func (s *NUDState) ReachableTime() time.Duration {
+ s.mu.Lock()
+ defer s.mu.Unlock()
+
+ if time.Now().After(s.expiration) ||
+ s.config.BaseReachableTime != s.prevBaseReachableTime ||
+ s.config.MinRandomFactor != s.prevMinRandomFactor ||
+ s.config.MaxRandomFactor != s.prevMaxRandomFactor {
+ return s.recomputeReachableTimeLocked()
+ }
+ return s.reachableTime
+}
+
+// recomputeReachableTimeLocked forces a recalculation of ReachableTime using
+// the algorithm defined in RFC 4861 section 6.3.2.
+//
+// This SHOULD automatically be invoked during certain situations, as per
+// RFC 4861 section 6.3.4:
+//
+// If the received Reachable Time value is non-zero, the host SHOULD set its
+// BaseReachableTime variable to the received value. If the new value
+// differs from the previous value, the host SHOULD re-compute a new random
+// ReachableTime value. ReachableTime is computed as a uniformly
+// distributed random value between MIN_RANDOM_FACTOR and MAX_RANDOM_FACTOR
+// times the BaseReachableTime. Using a random component eliminates the
+// possibility that Neighbor Unreachability Detection messages will
+// synchronize with each other.
+//
+// In most cases, the advertised Reachable Time value will be the same in
+// consecutive Router Advertisements, and a host's BaseReachableTime rarely
+// changes. In such cases, an implementation SHOULD ensure that a new
+// random value gets re-computed at least once every few hours.
+//
+// s.mu MUST be locked for writing.
+func (s *NUDState) recomputeReachableTimeLocked() time.Duration {
+ s.prevBaseReachableTime = s.config.BaseReachableTime
+ s.prevMinRandomFactor = s.config.MinRandomFactor
+ s.prevMaxRandomFactor = s.config.MaxRandomFactor
+
+ randomFactor := s.config.MinRandomFactor + s.rng.Float32()*(s.config.MaxRandomFactor-s.config.MinRandomFactor)
+
+ // Check for overflow, given that minRandomFactor and maxRandomFactor are
+ // guaranteed to be positive numbers.
+ if float32(math.MaxInt64)/randomFactor < float32(s.config.BaseReachableTime) {
+ s.reachableTime = time.Duration(math.MaxInt64)
+ } else if randomFactor == 1 {
+ // Avoid loss of precision when a large base reachable time is used.
+ s.reachableTime = s.config.BaseReachableTime
+ } else {
+ reachableTime := int64(float32(s.config.BaseReachableTime) * randomFactor)
+ s.reachableTime = time.Duration(reachableTime)
+ }
+
+ s.expiration = time.Now().Add(2 * time.Hour)
+ return s.reachableTime
+}
diff --git a/pkg/tcpip/stack/nud_test.go b/pkg/tcpip/stack/nud_test.go
new file mode 100644
index 000000000..2494ee610
--- /dev/null
+++ b/pkg/tcpip/stack/nud_test.go
@@ -0,0 +1,795 @@
+// Copyright 2020 The gVisor Authors.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+// http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package stack_test
+
+import (
+ "math"
+ "testing"
+ "time"
+
+ "gvisor.dev/gvisor/pkg/tcpip"
+ "gvisor.dev/gvisor/pkg/tcpip/link/channel"
+ "gvisor.dev/gvisor/pkg/tcpip/network/ipv6"
+ "gvisor.dev/gvisor/pkg/tcpip/stack"
+)
+
+const (
+ defaultBaseReachableTime = 30 * time.Second
+ minimumBaseReachableTime = time.Millisecond
+ defaultMinRandomFactor = 0.5
+ defaultMaxRandomFactor = 1.5
+ defaultRetransmitTimer = time.Second
+ minimumRetransmitTimer = time.Millisecond
+ defaultDelayFirstProbeTime = 5 * time.Second
+ defaultMaxMulticastProbes = 3
+ defaultMaxUnicastProbes = 3
+ defaultMaxAnycastDelayTime = time.Second
+ defaultMaxReachbilityConfirmations = 3
+ defaultUnreachableTime = 5 * time.Second
+
+ defaultFakeRandomNum = 0.5
+)
+
+// fakeRand is a deterministic random number generator.
+type fakeRand struct {
+ num float32
+}
+
+var _ stack.Rand = (*fakeRand)(nil)
+
+func (f *fakeRand) Float32() float32 {
+ return f.num
+}
+
+// TestSetNUDConfigurationFailsForBadNICID tests to make sure we get an error if
+// we attempt to update NUD configurations using an invalid NICID.
+func TestSetNUDConfigurationFailsForBadNICID(t *testing.T) {
+ s := stack.New(stack.Options{
+ // A neighbor cache is required to store NUDConfigurations. The networking
+ // stack will only allocate neighbor caches if a protocol providing link
+ // address resolution is specified (e.g. ARP or IPv6).
+ NetworkProtocols: []stack.NetworkProtocol{ipv6.NewProtocol()},
+ })
+
+ // No NIC with ID 1 yet.
+ config := stack.NUDConfigurations{}
+ if err := s.SetNUDConfigurations(1, config); err != tcpip.ErrUnknownNICID {
+ t.Fatalf("got s.SetNDPConfigurations(1, %+v) = %v, want = %s", config, err, tcpip.ErrUnknownNICID)
+ }
+}
+
+// TestNUDConfigurationFailsForNotSupported tests to make sure we get a
+// NotSupported error if we attempt to retrieve NUD configurations when the
+// stack doesn't support NUD.
+//
+// The stack will report to not support NUD if a neighbor cache for a given NIC
+// is not allocated. The networking stack will only allocate neighbor caches if
+// a protocol providing link address resolution is specified (e.g. ARP, IPv6).
+func TestNUDConfigurationFailsForNotSupported(t *testing.T) {
+ const nicID = 1
+
+ e := channel.New(0, 1280, linkAddr1)
+ e.LinkEPCapabilities |= stack.CapabilityResolutionRequired
+
+ s := stack.New(stack.Options{
+ NUDConfigs: stack.DefaultNUDConfigurations(),
+ })
+ if err := s.CreateNIC(nicID, e); err != nil {
+ t.Fatalf("CreateNIC(%d, _) = %s", nicID, err)
+ }
+ if _, err := s.NUDConfigurations(nicID); err != tcpip.ErrNotSupported {
+ t.Fatalf("got s.NDPConfigurations(%d) = %v, want = %s", nicID, err, tcpip.ErrNotSupported)
+ }
+}
+
+// TestNUDConfigurationFailsForNotSupported tests to make sure we get a
+// NotSupported error if we attempt to set NUD configurations when the stack
+// doesn't support NUD.
+//
+// The stack will report to not support NUD if a neighbor cache for a given NIC
+// is not allocated. The networking stack will only allocate neighbor caches if
+// a protocol providing link address resolution is specified (e.g. ARP, IPv6).
+func TestSetNUDConfigurationFailsForNotSupported(t *testing.T) {
+ const nicID = 1
+
+ e := channel.New(0, 1280, linkAddr1)
+ e.LinkEPCapabilities |= stack.CapabilityResolutionRequired
+
+ s := stack.New(stack.Options{
+ NUDConfigs: stack.DefaultNUDConfigurations(),
+ })
+ if err := s.CreateNIC(nicID, e); err != nil {
+ t.Fatalf("CreateNIC(%d, _) = %s", nicID, err)
+ }
+
+ config := stack.NUDConfigurations{}
+ if err := s.SetNUDConfigurations(nicID, config); err != tcpip.ErrNotSupported {
+ t.Fatalf("got s.SetNDPConfigurations(%d, %+v) = %v, want = %s", nicID, config, err, tcpip.ErrNotSupported)
+ }
+}
+
+// TestDefaultNUDConfigurationIsValid verifies that calling
+// resetInvalidFields() on the result of DefaultNUDConfigurations() does not
+// change anything. DefaultNUDConfigurations() should return a valid
+// NUDConfigurations.
+func TestDefaultNUDConfigurations(t *testing.T) {
+ const nicID = 1
+
+ e := channel.New(0, 1280, linkAddr1)
+ e.LinkEPCapabilities |= stack.CapabilityResolutionRequired
+
+ s := stack.New(stack.Options{
+ // A neighbor cache is required to store NUDConfigurations. The networking
+ // stack will only allocate neighbor caches if a protocol providing link
+ // address resolution is specified (e.g. ARP or IPv6).
+ NetworkProtocols: []stack.NetworkProtocol{ipv6.NewProtocol()},
+ NUDConfigs: stack.DefaultNUDConfigurations(),
+ })
+ if err := s.CreateNIC(nicID, e); err != nil {
+ t.Fatalf("CreateNIC(%d, _) = %s", nicID, err)
+ }
+ c, err := s.NUDConfigurations(nicID)
+ if err != nil {
+ t.Fatalf("got stack.NUDConfigurations(%d) = %s", nicID, err)
+ }
+ if got, want := c, stack.DefaultNUDConfigurations(); got != want {
+ t.Errorf("got stack.NUDConfigurations(%d) = %+v, want = %+v", nicID, got, want)
+ }
+}
+
+func TestNUDConfigurationsBaseReachableTime(t *testing.T) {
+ tests := []struct {
+ name string
+ baseReachableTime time.Duration
+ want time.Duration
+ }{
+ // Invalid cases
+ {
+ name: "EqualToZero",
+ baseReachableTime: 0,
+ want: defaultBaseReachableTime,
+ },
+ // Valid cases
+ {
+ name: "MoreThanZero",
+ baseReachableTime: time.Millisecond,
+ want: time.Millisecond,
+ },
+ {
+ name: "MoreThanDefaultBaseReachableTime",
+ baseReachableTime: 2 * defaultBaseReachableTime,
+ want: 2 * defaultBaseReachableTime,
+ },
+ }
+
+ for _, test := range tests {
+ t.Run(test.name, func(t *testing.T) {
+ const nicID = 1
+
+ c := stack.DefaultNUDConfigurations()
+ c.BaseReachableTime = test.baseReachableTime
+
+ e := channel.New(0, 1280, linkAddr1)
+ e.LinkEPCapabilities |= stack.CapabilityResolutionRequired
+
+ s := stack.New(stack.Options{
+ // A neighbor cache is required to store NUDConfigurations. The
+ // networking stack will only allocate neighbor caches if a protocol
+ // providing link address resolution is specified (e.g. ARP or IPv6).
+ NetworkProtocols: []stack.NetworkProtocol{ipv6.NewProtocol()},
+ NUDConfigs: c,
+ })
+ if err := s.CreateNIC(nicID, e); err != nil {
+ t.Fatalf("CreateNIC(%d, _) = %s", nicID, err)
+ }
+ sc, err := s.NUDConfigurations(nicID)
+ if err != nil {
+ t.Fatalf("got stack.NUDConfigurations(%d) = %s", nicID, err)
+ }
+ if got := sc.BaseReachableTime; got != test.want {
+ t.Errorf("got BaseReachableTime = %q, want = %q", got, test.want)
+ }
+ })
+ }
+}
+
+func TestNUDConfigurationsMinRandomFactor(t *testing.T) {
+ tests := []struct {
+ name string
+ minRandomFactor float32
+ want float32
+ }{
+ // Invalid cases
+ {
+ name: "LessThanZero",
+ minRandomFactor: -1,
+ want: defaultMinRandomFactor,
+ },
+ {
+ name: "EqualToZero",
+ minRandomFactor: 0,
+ want: defaultMinRandomFactor,
+ },
+ // Valid cases
+ {
+ name: "MoreThanZero",
+ minRandomFactor: 1,
+ want: 1,
+ },
+ }
+
+ for _, test := range tests {
+ t.Run(test.name, func(t *testing.T) {
+ const nicID = 1
+
+ c := stack.DefaultNUDConfigurations()
+ c.MinRandomFactor = test.minRandomFactor
+
+ e := channel.New(0, 1280, linkAddr1)
+ e.LinkEPCapabilities |= stack.CapabilityResolutionRequired
+
+ s := stack.New(stack.Options{
+ // A neighbor cache is required to store NUDConfigurations. The
+ // networking stack will only allocate neighbor caches if a protocol
+ // providing link address resolution is specified (e.g. ARP or IPv6).
+ NetworkProtocols: []stack.NetworkProtocol{ipv6.NewProtocol()},
+ NUDConfigs: c,
+ })
+ if err := s.CreateNIC(nicID, e); err != nil {
+ t.Fatalf("CreateNIC(%d, _) = %s", nicID, err)
+ }
+ sc, err := s.NUDConfigurations(nicID)
+ if err != nil {
+ t.Fatalf("got stack.NUDConfigurations(%d) = %s", nicID, err)
+ }
+ if got := sc.MinRandomFactor; got != test.want {
+ t.Errorf("got MinRandomFactor = %f, want = %f", got, test.want)
+ }
+ })
+ }
+}
+
+func TestNUDConfigurationsMaxRandomFactor(t *testing.T) {
+ tests := []struct {
+ name string
+ minRandomFactor float32
+ maxRandomFactor float32
+ want float32
+ }{
+ // Invalid cases
+ {
+ name: "LessThanZero",
+ minRandomFactor: defaultMinRandomFactor,
+ maxRandomFactor: -1,
+ want: defaultMaxRandomFactor,
+ },
+ {
+ name: "EqualToZero",
+ minRandomFactor: defaultMinRandomFactor,
+ maxRandomFactor: 0,
+ want: defaultMaxRandomFactor,
+ },
+ {
+ name: "LessThanMinRandomFactor",
+ minRandomFactor: defaultMinRandomFactor,
+ maxRandomFactor: defaultMinRandomFactor * 0.99,
+ want: defaultMaxRandomFactor,
+ },
+ {
+ name: "MoreThanMinRandomFactorWhenMinRandomFactorIsLargerThanMaxRandomFactorDefault",
+ minRandomFactor: defaultMaxRandomFactor * 2,
+ maxRandomFactor: defaultMaxRandomFactor,
+ want: defaultMaxRandomFactor * 6,
+ },
+ // Valid cases
+ {
+ name: "EqualToMinRandomFactor",
+ minRandomFactor: defaultMinRandomFactor,
+ maxRandomFactor: defaultMinRandomFactor,
+ want: defaultMinRandomFactor,
+ },
+ {
+ name: "MoreThanMinRandomFactor",
+ minRandomFactor: defaultMinRandomFactor,
+ maxRandomFactor: defaultMinRandomFactor * 1.1,
+ want: defaultMinRandomFactor * 1.1,
+ },
+ }
+
+ for _, test := range tests {
+ t.Run(test.name, func(t *testing.T) {
+ const nicID = 1
+
+ c := stack.DefaultNUDConfigurations()
+ c.MinRandomFactor = test.minRandomFactor
+ c.MaxRandomFactor = test.maxRandomFactor
+
+ e := channel.New(0, 1280, linkAddr1)
+ e.LinkEPCapabilities |= stack.CapabilityResolutionRequired
+
+ s := stack.New(stack.Options{
+ // A neighbor cache is required to store NUDConfigurations. The
+ // networking stack will only allocate neighbor caches if a protocol
+ // providing link address resolution is specified (e.g. ARP or IPv6).
+ NetworkProtocols: []stack.NetworkProtocol{ipv6.NewProtocol()},
+ NUDConfigs: c,
+ })
+ if err := s.CreateNIC(nicID, e); err != nil {
+ t.Fatalf("CreateNIC(%d, _) = %s", nicID, err)
+ }
+ sc, err := s.NUDConfigurations(nicID)
+ if err != nil {
+ t.Fatalf("got stack.NUDConfigurations(%d) = %s", nicID, err)
+ }
+ if got := sc.MaxRandomFactor; got != test.want {
+ t.Errorf("got MaxRandomFactor = %f, want = %f", got, test.want)
+ }
+ })
+ }
+}
+
+func TestNUDConfigurationsRetransmitTimer(t *testing.T) {
+ tests := []struct {
+ name string
+ retransmitTimer time.Duration
+ want time.Duration
+ }{
+ // Invalid cases
+ {
+ name: "EqualToZero",
+ retransmitTimer: 0,
+ want: defaultRetransmitTimer,
+ },
+ {
+ name: "LessThanMinimumRetransmitTimer",
+ retransmitTimer: minimumRetransmitTimer - time.Nanosecond,
+ want: defaultRetransmitTimer,
+ },
+ // Valid cases
+ {
+ name: "EqualToMinimumRetransmitTimer",
+ retransmitTimer: minimumRetransmitTimer,
+ want: minimumBaseReachableTime,
+ },
+ {
+ name: "LargetThanMinimumRetransmitTimer",
+ retransmitTimer: 2 * minimumBaseReachableTime,
+ want: 2 * minimumBaseReachableTime,
+ },
+ }
+
+ for _, test := range tests {
+ t.Run(test.name, func(t *testing.T) {
+ const nicID = 1
+
+ c := stack.DefaultNUDConfigurations()
+ c.RetransmitTimer = test.retransmitTimer
+
+ e := channel.New(0, 1280, linkAddr1)
+ e.LinkEPCapabilities |= stack.CapabilityResolutionRequired
+
+ s := stack.New(stack.Options{
+ // A neighbor cache is required to store NUDConfigurations. The
+ // networking stack will only allocate neighbor caches if a protocol
+ // providing link address resolution is specified (e.g. ARP or IPv6).
+ NetworkProtocols: []stack.NetworkProtocol{ipv6.NewProtocol()},
+ NUDConfigs: c,
+ })
+ if err := s.CreateNIC(nicID, e); err != nil {
+ t.Fatalf("CreateNIC(%d, _) = %s", nicID, err)
+ }
+ sc, err := s.NUDConfigurations(nicID)
+ if err != nil {
+ t.Fatalf("got stack.NUDConfigurations(%d) = %s", nicID, err)
+ }
+ if got := sc.RetransmitTimer; got != test.want {
+ t.Errorf("got RetransmitTimer = %q, want = %q", got, test.want)
+ }
+ })
+ }
+}
+
+func TestNUDConfigurationsDelayFirstProbeTime(t *testing.T) {
+ tests := []struct {
+ name string
+ delayFirstProbeTime time.Duration
+ want time.Duration
+ }{
+ // Invalid cases
+ {
+ name: "EqualToZero",
+ delayFirstProbeTime: 0,
+ want: defaultDelayFirstProbeTime,
+ },
+ // Valid cases
+ {
+ name: "MoreThanZero",
+ delayFirstProbeTime: time.Millisecond,
+ want: time.Millisecond,
+ },
+ }
+
+ for _, test := range tests {
+ t.Run(test.name, func(t *testing.T) {
+ const nicID = 1
+
+ c := stack.DefaultNUDConfigurations()
+ c.DelayFirstProbeTime = test.delayFirstProbeTime
+
+ e := channel.New(0, 1280, linkAddr1)
+ e.LinkEPCapabilities |= stack.CapabilityResolutionRequired
+
+ s := stack.New(stack.Options{
+ // A neighbor cache is required to store NUDConfigurations. The
+ // networking stack will only allocate neighbor caches if a protocol
+ // providing link address resolution is specified (e.g. ARP or IPv6).
+ NetworkProtocols: []stack.NetworkProtocol{ipv6.NewProtocol()},
+ NUDConfigs: c,
+ })
+ if err := s.CreateNIC(nicID, e); err != nil {
+ t.Fatalf("CreateNIC(%d, _) = %s", nicID, err)
+ }
+ sc, err := s.NUDConfigurations(nicID)
+ if err != nil {
+ t.Fatalf("got stack.NUDConfigurations(%d) = %s", nicID, err)
+ }
+ if got := sc.DelayFirstProbeTime; got != test.want {
+ t.Errorf("got DelayFirstProbeTime = %q, want = %q", got, test.want)
+ }
+ })
+ }
+}
+
+func TestNUDConfigurationsMaxMulticastProbes(t *testing.T) {
+ tests := []struct {
+ name string
+ maxMulticastProbes uint32
+ want uint32
+ }{
+ // Invalid cases
+ {
+ name: "EqualToZero",
+ maxMulticastProbes: 0,
+ want: defaultMaxMulticastProbes,
+ },
+ // Valid cases
+ {
+ name: "MoreThanZero",
+ maxMulticastProbes: 1,
+ want: 1,
+ },
+ }
+
+ for _, test := range tests {
+ t.Run(test.name, func(t *testing.T) {
+ const nicID = 1
+
+ c := stack.DefaultNUDConfigurations()
+ c.MaxMulticastProbes = test.maxMulticastProbes
+
+ e := channel.New(0, 1280, linkAddr1)
+ e.LinkEPCapabilities |= stack.CapabilityResolutionRequired
+
+ s := stack.New(stack.Options{
+ // A neighbor cache is required to store NUDConfigurations. The
+ // networking stack will only allocate neighbor caches if a protocol
+ // providing link address resolution is specified (e.g. ARP or IPv6).
+ NetworkProtocols: []stack.NetworkProtocol{ipv6.NewProtocol()},
+ NUDConfigs: c,
+ })
+ if err := s.CreateNIC(nicID, e); err != nil {
+ t.Fatalf("CreateNIC(%d, _) = %s", nicID, err)
+ }
+ sc, err := s.NUDConfigurations(nicID)
+ if err != nil {
+ t.Fatalf("got stack.NUDConfigurations(%d) = %s", nicID, err)
+ }
+ if got := sc.MaxMulticastProbes; got != test.want {
+ t.Errorf("got MaxMulticastProbes = %q, want = %q", got, test.want)
+ }
+ })
+ }
+}
+
+func TestNUDConfigurationsMaxUnicastProbes(t *testing.T) {
+ tests := []struct {
+ name string
+ maxUnicastProbes uint32
+ want uint32
+ }{
+ // Invalid cases
+ {
+ name: "EqualToZero",
+ maxUnicastProbes: 0,
+ want: defaultMaxUnicastProbes,
+ },
+ // Valid cases
+ {
+ name: "MoreThanZero",
+ maxUnicastProbes: 1,
+ want: 1,
+ },
+ }
+
+ for _, test := range tests {
+ t.Run(test.name, func(t *testing.T) {
+ const nicID = 1
+
+ c := stack.DefaultNUDConfigurations()
+ c.MaxUnicastProbes = test.maxUnicastProbes
+
+ e := channel.New(0, 1280, linkAddr1)
+ e.LinkEPCapabilities |= stack.CapabilityResolutionRequired
+
+ s := stack.New(stack.Options{
+ // A neighbor cache is required to store NUDConfigurations. The
+ // networking stack will only allocate neighbor caches if a protocol
+ // providing link address resolution is specified (e.g. ARP or IPv6).
+ NetworkProtocols: []stack.NetworkProtocol{ipv6.NewProtocol()},
+ NUDConfigs: c,
+ })
+ if err := s.CreateNIC(nicID, e); err != nil {
+ t.Fatalf("CreateNIC(%d, _) = %s", nicID, err)
+ }
+ sc, err := s.NUDConfigurations(nicID)
+ if err != nil {
+ t.Fatalf("got stack.NUDConfigurations(%d) = %s", nicID, err)
+ }
+ if got := sc.MaxUnicastProbes; got != test.want {
+ t.Errorf("got MaxUnicastProbes = %q, want = %q", got, test.want)
+ }
+ })
+ }
+}
+
+func TestNUDConfigurationsUnreachableTime(t *testing.T) {
+ tests := []struct {
+ name string
+ unreachableTime time.Duration
+ want time.Duration
+ }{
+ // Invalid cases
+ {
+ name: "EqualToZero",
+ unreachableTime: 0,
+ want: defaultUnreachableTime,
+ },
+ // Valid cases
+ {
+ name: "MoreThanZero",
+ unreachableTime: time.Millisecond,
+ want: time.Millisecond,
+ },
+ }
+
+ for _, test := range tests {
+ t.Run(test.name, func(t *testing.T) {
+ const nicID = 1
+
+ c := stack.DefaultNUDConfigurations()
+ c.UnreachableTime = test.unreachableTime
+
+ e := channel.New(0, 1280, linkAddr1)
+ e.LinkEPCapabilities |= stack.CapabilityResolutionRequired
+
+ s := stack.New(stack.Options{
+ // A neighbor cache is required to store NUDConfigurations. The
+ // networking stack will only allocate neighbor caches if a protocol
+ // providing link address resolution is specified (e.g. ARP or IPv6).
+ NetworkProtocols: []stack.NetworkProtocol{ipv6.NewProtocol()},
+ NUDConfigs: c,
+ })
+ if err := s.CreateNIC(nicID, e); err != nil {
+ t.Fatalf("CreateNIC(%d, _) = %s", nicID, err)
+ }
+ sc, err := s.NUDConfigurations(nicID)
+ if err != nil {
+ t.Fatalf("got stack.NUDConfigurations(%d) = %s", nicID, err)
+ }
+ if got := sc.UnreachableTime; got != test.want {
+ t.Errorf("got UnreachableTime = %q, want = %q", got, test.want)
+ }
+ })
+ }
+}
+
+// TestNUDStateReachableTime verifies the correctness of the ReachableTime
+// computation.
+func TestNUDStateReachableTime(t *testing.T) {
+ tests := []struct {
+ name string
+ baseReachableTime time.Duration
+ minRandomFactor float32
+ maxRandomFactor float32
+ want time.Duration
+ }{
+ {
+ name: "AllZeros",
+ baseReachableTime: 0,
+ minRandomFactor: 0,
+ maxRandomFactor: 0,
+ want: 0,
+ },
+ {
+ name: "ZeroMaxRandomFactor",
+ baseReachableTime: time.Second,
+ minRandomFactor: 0,
+ maxRandomFactor: 0,
+ want: 0,
+ },
+ {
+ name: "ZeroMinRandomFactor",
+ baseReachableTime: time.Second,
+ minRandomFactor: 0,
+ maxRandomFactor: 1,
+ want: time.Duration(defaultFakeRandomNum * float32(time.Second)),
+ },
+ {
+ name: "FractionalRandomFactor",
+ baseReachableTime: time.Duration(math.MaxInt64),
+ minRandomFactor: 0.001,
+ maxRandomFactor: 0.002,
+ want: time.Duration((0.001 + (0.001 * defaultFakeRandomNum)) * float32(math.MaxInt64)),
+ },
+ {
+ name: "MinAndMaxRandomFactorsEqual",
+ baseReachableTime: time.Second,
+ minRandomFactor: 1,
+ maxRandomFactor: 1,
+ want: time.Second,
+ },
+ {
+ name: "MinAndMaxRandomFactorsDifferent",
+ baseReachableTime: time.Second,
+ minRandomFactor: 1,
+ maxRandomFactor: 2,
+ want: time.Duration((1.0 + defaultFakeRandomNum) * float32(time.Second)),
+ },
+ {
+ name: "MaxInt64",
+ baseReachableTime: time.Duration(math.MaxInt64),
+ minRandomFactor: 1,
+ maxRandomFactor: 1,
+ want: time.Duration(math.MaxInt64),
+ },
+ {
+ name: "Overflow",
+ baseReachableTime: time.Duration(math.MaxInt64),
+ minRandomFactor: 1.5,
+ maxRandomFactor: 1.5,
+ want: time.Duration(math.MaxInt64),
+ },
+ {
+ name: "DoubleOverflow",
+ baseReachableTime: time.Duration(math.MaxInt64),
+ minRandomFactor: 2.5,
+ maxRandomFactor: 2.5,
+ want: time.Duration(math.MaxInt64),
+ },
+ }
+
+ for _, test := range tests {
+ t.Run(test.name, func(t *testing.T) {
+ c := stack.NUDConfigurations{
+ BaseReachableTime: test.baseReachableTime,
+ MinRandomFactor: test.minRandomFactor,
+ MaxRandomFactor: test.maxRandomFactor,
+ }
+ // A fake random number generator is used to ensure deterministic
+ // results.
+ rng := fakeRand{
+ num: defaultFakeRandomNum,
+ }
+ s := stack.NewNUDState(c, &rng)
+ if got, want := s.ReachableTime(), test.want; got != want {
+ t.Errorf("got ReachableTime = %q, want = %q", got, want)
+ }
+ })
+ }
+}
+
+// TestNUDStateRecomputeReachableTime exercises the ReachableTime function
+// twice to verify recomputation of reachable time when the min random factor,
+// max random factor, or base reachable time changes.
+func TestNUDStateRecomputeReachableTime(t *testing.T) {
+ const defaultBase = time.Second
+ const defaultMin = 2.0 * defaultMaxRandomFactor
+ const defaultMax = 3.0 * defaultMaxRandomFactor
+
+ tests := []struct {
+ name string
+ baseReachableTime time.Duration
+ minRandomFactor float32
+ maxRandomFactor float32
+ want time.Duration
+ }{
+ {
+ name: "BaseReachableTime",
+ baseReachableTime: 2 * defaultBase,
+ minRandomFactor: defaultMin,
+ maxRandomFactor: defaultMax,
+ want: time.Duration((defaultMin + (defaultMax-defaultMin)*defaultFakeRandomNum) * float32(2*defaultBase)),
+ },
+ {
+ name: "MinRandomFactor",
+ baseReachableTime: defaultBase,
+ minRandomFactor: defaultMax,
+ maxRandomFactor: defaultMax,
+ want: time.Duration(defaultMax * float32(defaultBase)),
+ },
+ {
+ name: "MaxRandomFactor",
+ baseReachableTime: defaultBase,
+ minRandomFactor: defaultMin,
+ maxRandomFactor: defaultMin,
+ want: time.Duration(defaultMin * float32(defaultBase)),
+ },
+ {
+ name: "BothRandomFactor",
+ baseReachableTime: defaultBase,
+ minRandomFactor: 2 * defaultMin,
+ maxRandomFactor: 2 * defaultMax,
+ want: time.Duration((2*defaultMin + (2*defaultMax-2*defaultMin)*defaultFakeRandomNum) * float32(defaultBase)),
+ },
+ {
+ name: "BaseReachableTimeAndBothRandomFactors",
+ baseReachableTime: 2 * defaultBase,
+ minRandomFactor: 2 * defaultMin,
+ maxRandomFactor: 2 * defaultMax,
+ want: time.Duration((2*defaultMin + (2*defaultMax-2*defaultMin)*defaultFakeRandomNum) * float32(2*defaultBase)),
+ },
+ }
+
+ for _, test := range tests {
+ t.Run(test.name, func(t *testing.T) {
+ c := stack.DefaultNUDConfigurations()
+ c.BaseReachableTime = defaultBase
+ c.MinRandomFactor = defaultMin
+ c.MaxRandomFactor = defaultMax
+
+ // A fake random number generator is used to ensure deterministic
+ // results.
+ rng := fakeRand{
+ num: defaultFakeRandomNum,
+ }
+ s := stack.NewNUDState(c, &rng)
+ old := s.ReachableTime()
+
+ if got, want := s.ReachableTime(), old; got != want {
+ t.Errorf("got ReachableTime = %q, want = %q", got, want)
+ }
+
+ // Check for recomputation when changing the min random factor, the max
+ // random factor, the base reachability time, or any permutation of those
+ // three options.
+ c.BaseReachableTime = test.baseReachableTime
+ c.MinRandomFactor = test.minRandomFactor
+ c.MaxRandomFactor = test.maxRandomFactor
+ s.SetConfig(c)
+
+ if got, want := s.ReachableTime(), test.want; got != want {
+ t.Errorf("got ReachableTime = %q, want = %q", got, want)
+ }
+
+ // Verify that ReachableTime isn't recomputed when none of the
+ // configuration options change. The random factor is changed so that if
+ // a recompution were to occur, ReachableTime would change.
+ rng.num = defaultFakeRandomNum / 2.0
+ if got, want := s.ReachableTime(), test.want; got != want {
+ t.Errorf("got ReachableTime = %q, want = %q", got, want)
+ }
+ })
+ }
+}
diff --git a/pkg/tcpip/stack/route.go b/pkg/tcpip/stack/route.go
index d65f8049e..91e0110f1 100644
--- a/pkg/tcpip/stack/route.go
+++ b/pkg/tcpip/stack/route.go
@@ -48,6 +48,10 @@ type Route struct {
// Loop controls where WritePacket should send packets.
Loop PacketLooping
+
+ // directedBroadcast indicates whether this route is sending a directed
+ // broadcast packet.
+ directedBroadcast bool
}
// makeRoute initializes a new route. It takes ownership of the provided
@@ -275,6 +279,12 @@ func (r *Route) Stack() *Stack {
return r.ref.stack()
}
+// IsBroadcast returns true if the route is to send a broadcast packet.
+func (r *Route) IsBroadcast() bool {
+ // Only IPv4 has a notion of broadcast.
+ return r.directedBroadcast || r.RemoteAddress == header.IPv4Broadcast
+}
+
// ReverseRoute returns new route with given source and destination address.
func (r *Route) ReverseRoute(src tcpip.Address, dst tcpip.Address) Route {
return Route{
diff --git a/pkg/tcpip/stack/stack.go b/pkg/tcpip/stack/stack.go
index a6faa22c2..3f07e4159 100644
--- a/pkg/tcpip/stack/stack.go
+++ b/pkg/tcpip/stack/stack.go
@@ -445,6 +445,9 @@ type Stack struct {
// ndpConfigs is the default NDP configurations used by interfaces.
ndpConfigs NDPConfigurations
+ // nudConfigs is the default NUD configurations used by interfaces.
+ nudConfigs NUDConfigurations
+
// autoGenIPv6LinkLocal determines whether or not the stack will attempt
// to auto-generate an IPv6 link-local address for newly enabled non-loopback
// NICs. See the AutoGenIPv6LinkLocal field of Options for more details.
@@ -454,6 +457,10 @@ type Stack struct {
// integrator NDP related events.
ndpDisp NDPDispatcher
+ // nudDisp is the NUD event dispatcher that is used to send the netstack
+ // integrator NUD related events.
+ nudDisp NUDDispatcher
+
// uniqueIDGenerator is a generator of unique identifiers.
uniqueIDGenerator UniqueID
@@ -518,6 +525,9 @@ type Options struct {
// before assigning an address to a NIC.
NDPConfigs NDPConfigurations
+ // NUDConfigs is the default NUD configurations used by interfaces.
+ NUDConfigs NUDConfigurations
+
// AutoGenIPv6LinkLocal determines whether or not the stack will attempt to
// auto-generate an IPv6 link-local address for newly enabled non-loopback
// NICs.
@@ -536,6 +546,10 @@ type Options struct {
// receive NDP related events.
NDPDisp NDPDispatcher
+ // NUDDisp is the NUD event dispatcher that an integrator can provide to
+ // receive NUD related events.
+ NUDDisp NUDDispatcher
+
// RawFactory produces raw endpoints. Raw endpoints are enabled only if
// this is non-nil.
RawFactory RawFactory
@@ -670,6 +684,8 @@ func New(opts Options) *Stack {
// Make sure opts.NDPConfigs contains valid values only.
opts.NDPConfigs.validate()
+ opts.NUDConfigs.resetInvalidFields()
+
s := &Stack{
transportProtocols: make(map[tcpip.TransportProtocolNumber]*transportProtocolState),
networkProtocols: make(map[tcpip.NetworkProtocolNumber]NetworkProtocol),
@@ -685,9 +701,11 @@ func New(opts Options) *Stack {
icmpRateLimiter: NewICMPRateLimiter(),
seed: generateRandUint32(),
ndpConfigs: opts.NDPConfigs,
+ nudConfigs: opts.NUDConfigs,
autoGenIPv6LinkLocal: opts.AutoGenIPv6LinkLocal,
uniqueIDGenerator: opts.UniqueID,
ndpDisp: opts.NDPDisp,
+ nudDisp: opts.NUDDisp,
opaqueIIDOpts: opts.OpaqueIIDOpts,
tempIIDSeed: opts.TempIIDSeed,
forwarder: newForwardQueue(),
@@ -1284,9 +1302,9 @@ func (s *Stack) FindRoute(id tcpip.NICID, localAddr, remoteAddr tcpip.Address, n
s.mu.RLock()
defer s.mu.RUnlock()
- isBroadcast := remoteAddr == header.IPv4Broadcast
+ isLocalBroadcast := remoteAddr == header.IPv4Broadcast
isMulticast := header.IsV4MulticastAddress(remoteAddr) || header.IsV6MulticastAddress(remoteAddr)
- needRoute := !(isBroadcast || isMulticast || header.IsV6LinkLocalAddress(remoteAddr))
+ needRoute := !(isLocalBroadcast || isMulticast || header.IsV6LinkLocalAddress(remoteAddr))
if id != 0 && !needRoute {
if nic, ok := s.nics[id]; ok && nic.enabled() {
if ref := s.getRefEP(nic, localAddr, remoteAddr, netProto); ref != nil {
@@ -1307,9 +1325,16 @@ func (s *Stack) FindRoute(id tcpip.NICID, localAddr, remoteAddr tcpip.Address, n
}
r := makeRoute(netProto, ref.ep.ID().LocalAddress, remoteAddr, nic.linkEP.LinkAddress(), ref, s.handleLocal && !nic.isLoopback(), multicastLoop && !nic.isLoopback())
- if needRoute {
- r.NextHop = route.Gateway
+ r.directedBroadcast = route.Destination.IsBroadcast(remoteAddr)
+
+ if len(route.Gateway) > 0 {
+ if needRoute {
+ r.NextHop = route.Gateway
+ }
+ } else if r.directedBroadcast {
+ r.RemoteLinkAddress = header.EthernetBroadcastAddress
}
+
return r, nil
}
}
@@ -1862,10 +1887,38 @@ func (s *Stack) SetNDPConfigurations(id tcpip.NICID, c NDPConfigurations) *tcpip
}
nic.setNDPConfigs(c)
-
return nil
}
+// NUDConfigurations gets the per-interface NUD configurations.
+func (s *Stack) NUDConfigurations(id tcpip.NICID) (NUDConfigurations, *tcpip.Error) {
+ s.mu.RLock()
+ nic, ok := s.nics[id]
+ s.mu.RUnlock()
+
+ if !ok {
+ return NUDConfigurations{}, tcpip.ErrUnknownNICID
+ }
+
+ return nic.NUDConfigs()
+}
+
+// SetNUDConfigurations sets the per-interface NUD configurations.
+//
+// Note, if c contains invalid NUD configuration values, it will be fixed to
+// use default values for the erroneous values.
+func (s *Stack) SetNUDConfigurations(id tcpip.NICID, c NUDConfigurations) *tcpip.Error {
+ s.mu.RLock()
+ nic, ok := s.nics[id]
+ s.mu.RUnlock()
+
+ if !ok {
+ return tcpip.ErrUnknownNICID
+ }
+
+ return nic.setNUDConfigs(c)
+}
+
// HandleNDPRA provides a NIC with ID id a validated NDP Router Advertisement
// message that it needs to handle.
func (s *Stack) HandleNDPRA(id tcpip.NICID, ip tcpip.Address, ra header.NDPRouterAdvert) *tcpip.Error {
diff --git a/pkg/tcpip/stack/stack_test.go b/pkg/tcpip/stack/stack_test.go
index 101ca2206..f22062889 100644
--- a/pkg/tcpip/stack/stack_test.go
+++ b/pkg/tcpip/stack/stack_test.go
@@ -27,6 +27,7 @@ import (
"time"
"github.com/google/go-cmp/cmp"
+ "github.com/google/go-cmp/cmp/cmpopts"
"gvisor.dev/gvisor/pkg/rand"
"gvisor.dev/gvisor/pkg/tcpip"
"gvisor.dev/gvisor/pkg/tcpip/buffer"
@@ -3418,3 +3419,225 @@ func TestStackSendBufferSizeOption(t *testing.T) {
})
}
}
+
+func TestOutgoingSubnetBroadcast(t *testing.T) {
+ const (
+ unspecifiedNICID = 0
+ nicID1 = 1
+ )
+
+ defaultAddr := tcpip.AddressWithPrefix{
+ Address: header.IPv4Any,
+ PrefixLen: 0,
+ }
+ defaultSubnet := defaultAddr.Subnet()
+ ipv4Addr := tcpip.AddressWithPrefix{
+ Address: "\xc0\xa8\x01\x3a",
+ PrefixLen: 24,
+ }
+ ipv4Subnet := ipv4Addr.Subnet()
+ ipv4SubnetBcast := ipv4Subnet.Broadcast()
+ ipv4Gateway := tcpip.Address("\xc0\xa8\x01\x01")
+ ipv4AddrPrefix31 := tcpip.AddressWithPrefix{
+ Address: "\xc0\xa8\x01\x3a",
+ PrefixLen: 31,
+ }
+ ipv4Subnet31 := ipv4AddrPrefix31.Subnet()
+ ipv4Subnet31Bcast := ipv4Subnet31.Broadcast()
+ ipv4AddrPrefix32 := tcpip.AddressWithPrefix{
+ Address: "\xc0\xa8\x01\x3a",
+ PrefixLen: 32,
+ }
+ ipv4Subnet32 := ipv4AddrPrefix32.Subnet()
+ ipv4Subnet32Bcast := ipv4Subnet32.Broadcast()
+ ipv6Addr := tcpip.AddressWithPrefix{
+ Address: "\x20\x0a\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x01",
+ PrefixLen: 64,
+ }
+ ipv6Subnet := ipv6Addr.Subnet()
+ ipv6SubnetBcast := ipv6Subnet.Broadcast()
+ remNetAddr := tcpip.AddressWithPrefix{
+ Address: "\x64\x0a\x7b\x18",
+ PrefixLen: 24,
+ }
+ remNetSubnet := remNetAddr.Subnet()
+ remNetSubnetBcast := remNetSubnet.Broadcast()
+
+ tests := []struct {
+ name string
+ nicAddr tcpip.ProtocolAddress
+ routes []tcpip.Route
+ remoteAddr tcpip.Address
+ expectedRoute stack.Route
+ }{
+ // Broadcast to a locally attached subnet populates the broadcast MAC.
+ {
+ name: "IPv4 Broadcast to local subnet",
+ nicAddr: tcpip.ProtocolAddress{
+ Protocol: header.IPv4ProtocolNumber,
+ AddressWithPrefix: ipv4Addr,
+ },
+ routes: []tcpip.Route{
+ {
+ Destination: ipv4Subnet,
+ NIC: nicID1,
+ },
+ },
+ remoteAddr: ipv4SubnetBcast,
+ expectedRoute: stack.Route{
+ LocalAddress: ipv4Addr.Address,
+ RemoteAddress: ipv4SubnetBcast,
+ RemoteLinkAddress: header.EthernetBroadcastAddress,
+ NetProto: header.IPv4ProtocolNumber,
+ Loop: stack.PacketOut,
+ },
+ },
+ // Broadcast to a locally attached /31 subnet does not populate the
+ // broadcast MAC.
+ {
+ name: "IPv4 Broadcast to local /31 subnet",
+ nicAddr: tcpip.ProtocolAddress{
+ Protocol: header.IPv4ProtocolNumber,
+ AddressWithPrefix: ipv4AddrPrefix31,
+ },
+ routes: []tcpip.Route{
+ {
+ Destination: ipv4Subnet31,
+ NIC: nicID1,
+ },
+ },
+ remoteAddr: ipv4Subnet31Bcast,
+ expectedRoute: stack.Route{
+ LocalAddress: ipv4AddrPrefix31.Address,
+ RemoteAddress: ipv4Subnet31Bcast,
+ NetProto: header.IPv4ProtocolNumber,
+ Loop: stack.PacketOut,
+ },
+ },
+ // Broadcast to a locally attached /32 subnet does not populate the
+ // broadcast MAC.
+ {
+ name: "IPv4 Broadcast to local /32 subnet",
+ nicAddr: tcpip.ProtocolAddress{
+ Protocol: header.IPv4ProtocolNumber,
+ AddressWithPrefix: ipv4AddrPrefix32,
+ },
+ routes: []tcpip.Route{
+ {
+ Destination: ipv4Subnet32,
+ NIC: nicID1,
+ },
+ },
+ remoteAddr: ipv4Subnet32Bcast,
+ expectedRoute: stack.Route{
+ LocalAddress: ipv4AddrPrefix32.Address,
+ RemoteAddress: ipv4Subnet32Bcast,
+ NetProto: header.IPv4ProtocolNumber,
+ Loop: stack.PacketOut,
+ },
+ },
+ // IPv6 has no notion of a broadcast.
+ {
+ name: "IPv6 'Broadcast' to local subnet",
+ nicAddr: tcpip.ProtocolAddress{
+ Protocol: header.IPv6ProtocolNumber,
+ AddressWithPrefix: ipv6Addr,
+ },
+ routes: []tcpip.Route{
+ {
+ Destination: ipv6Subnet,
+ NIC: nicID1,
+ },
+ },
+ remoteAddr: ipv6SubnetBcast,
+ expectedRoute: stack.Route{
+ LocalAddress: ipv6Addr.Address,
+ RemoteAddress: ipv6SubnetBcast,
+ NetProto: header.IPv6ProtocolNumber,
+ Loop: stack.PacketOut,
+ },
+ },
+ // Broadcast to a remote subnet in the route table is send to the next-hop
+ // gateway.
+ {
+ name: "IPv4 Broadcast to remote subnet",
+ nicAddr: tcpip.ProtocolAddress{
+ Protocol: header.IPv4ProtocolNumber,
+ AddressWithPrefix: ipv4Addr,
+ },
+ routes: []tcpip.Route{
+ {
+ Destination: remNetSubnet,
+ Gateway: ipv4Gateway,
+ NIC: nicID1,
+ },
+ },
+ remoteAddr: remNetSubnetBcast,
+ expectedRoute: stack.Route{
+ LocalAddress: ipv4Addr.Address,
+ RemoteAddress: remNetSubnetBcast,
+ NextHop: ipv4Gateway,
+ NetProto: header.IPv4ProtocolNumber,
+ Loop: stack.PacketOut,
+ },
+ },
+ // Broadcast to an unknown subnet follows the default route. Note that this
+ // is essentially just routing an unknown destination IP, because w/o any
+ // subnet prefix information a subnet broadcast address is just a normal IP.
+ {
+ name: "IPv4 Broadcast to unknown subnet",
+ nicAddr: tcpip.ProtocolAddress{
+ Protocol: header.IPv4ProtocolNumber,
+ AddressWithPrefix: ipv4Addr,
+ },
+ routes: []tcpip.Route{
+ {
+ Destination: defaultSubnet,
+ Gateway: ipv4Gateway,
+ NIC: nicID1,
+ },
+ },
+ remoteAddr: remNetSubnetBcast,
+ expectedRoute: stack.Route{
+ LocalAddress: ipv4Addr.Address,
+ RemoteAddress: remNetSubnetBcast,
+ NextHop: ipv4Gateway,
+ NetProto: header.IPv4ProtocolNumber,
+ Loop: stack.PacketOut,
+ },
+ },
+ }
+
+ for _, test := range tests {
+ t.Run(test.name, func(t *testing.T) {
+ s := stack.New(stack.Options{
+ NetworkProtocols: []stack.NetworkProtocol{ipv4.NewProtocol(), ipv6.NewProtocol()},
+ })
+ ep := channel.New(0, defaultMTU, "")
+ if err := s.CreateNIC(nicID1, ep); err != nil {
+ t.Fatalf("CreateNIC(%d, _): %s", nicID1, err)
+ }
+ if err := s.AddProtocolAddress(nicID1, test.nicAddr); err != nil {
+ t.Fatalf("AddProtocolAddress(%d, %+v): %s", nicID1, test.nicAddr, err)
+ }
+
+ s.SetRouteTable(test.routes)
+
+ var netProto tcpip.NetworkProtocolNumber
+ switch l := len(test.remoteAddr); l {
+ case header.IPv4AddressSize:
+ netProto = header.IPv4ProtocolNumber
+ case header.IPv6AddressSize:
+ netProto = header.IPv6ProtocolNumber
+ default:
+ t.Fatalf("got unexpected address length = %d bytes", l)
+ }
+
+ if r, err := s.FindRoute(unspecifiedNICID, "" /* localAddr */, test.remoteAddr, netProto, false /* multicastLoop */); err != nil {
+ t.Fatalf("FindRoute(%d, '', %s, %d): %s", unspecifiedNICID, test.remoteAddr, netProto, err)
+ } else if diff := cmp.Diff(r, test.expectedRoute, cmpopts.IgnoreUnexported(r)); diff != "" {
+ t.Errorf("route mismatch (-want +got):\n%s", diff)
+ }
+ })
+ }
+}
diff --git a/pkg/tcpip/tcpip.go b/pkg/tcpip/tcpip.go
index 21aafb0a2..45f59b60f 100644
--- a/pkg/tcpip/tcpip.go
+++ b/pkg/tcpip/tcpip.go
@@ -43,6 +43,9 @@ import (
"gvisor.dev/gvisor/pkg/waiter"
)
+// Using header.IPv4AddressSize would cause an import cycle.
+const ipv4AddressSize = 4
+
// Error represents an error in the netstack error space. Using a special type
// ensures that errors outside of this space are not accidentally introduced.
//
@@ -320,6 +323,29 @@ func (s *Subnet) Broadcast() Address {
return Address(addr)
}
+// IsBroadcast returns true if the address is considered a broadcast address.
+func (s *Subnet) IsBroadcast(address Address) bool {
+ // Only IPv4 supports the notion of a broadcast address.
+ if len(address) != ipv4AddressSize {
+ return false
+ }
+
+ // Normally, we would just compare address with the subnet's broadcast
+ // address but there is an exception where a simple comparison is not
+ // correct. This exception is for /31 and /32 IPv4 subnets where all
+ // addresses are considered valid host addresses.
+ //
+ // For /31 subnets, the case is easy. RFC 3021 Section 2.1 states that
+ // both addresses in a /31 subnet "MUST be interpreted as host addresses."
+ //
+ // For /32, the case is a bit more vague. RFC 3021 makes no mention of /32
+ // subnets. However, the same reasoning applies - if an exception is not
+ // made, then there do not exist any host addresses in a /32 subnet. RFC
+ // 4632 Section 3.1 also vaguely implies this interpretation by referring
+ // to addresses in /32 subnets as "host routes."
+ return s.Prefix() <= 30 && s.Broadcast() == address
+}
+
// Equal returns true if s equals o.
//
// Needed to use cmp.Equal on Subnet as its fields are unexported.
@@ -928,6 +954,10 @@ type DefaultTTLOption uint8
// classic BPF filter on a given endpoint.
type SocketDetachFilterOption int
+// OriginalDestinationOption is used to get the original destination address
+// and port of a redirected packet.
+type OriginalDestinationOption FullAddress
+
// IPPacketInfo is the message structure for IP_PKTINFO.
//
// +stateify savable
diff --git a/pkg/tcpip/transport/tcp/connect.go b/pkg/tcpip/transport/tcp/connect.go
index 1798510bc..6e5e55b6f 100644
--- a/pkg/tcpip/transport/tcp/connect.go
+++ b/pkg/tcpip/transport/tcp/connect.go
@@ -1024,14 +1024,19 @@ func (e *endpoint) transitionToStateEstablishedLocked(h *handshake) {
// delivered to this endpoint from the demuxer when the endpoint
// is transitioned to StateClose.
func (e *endpoint) transitionToStateCloseLocked() {
- if e.EndpointState() == StateClose {
+ s := e.EndpointState()
+ if s == StateClose {
return
}
+
+ if s.connected() {
+ e.stack.Stats().TCP.CurrentConnected.Decrement()
+ e.stack.Stats().TCP.EstablishedClosed.Increment()
+ }
+
// Mark the endpoint as fully closed for reads/writes.
e.cleanupLocked()
e.setEndpointState(StateClose)
- e.stack.Stats().TCP.CurrentConnected.Decrement()
- e.stack.Stats().TCP.EstablishedClosed.Increment()
}
// tryDeliverSegmentFromClosedEndpoint attempts to deliver the parsed
diff --git a/pkg/tcpip/transport/tcp/endpoint.go b/pkg/tcpip/transport/tcp/endpoint.go
index 0f7487963..682687ebe 100644
--- a/pkg/tcpip/transport/tcp/endpoint.go
+++ b/pkg/tcpip/transport/tcp/endpoint.go
@@ -2017,6 +2017,17 @@ func (e *endpoint) GetSockOpt(opt interface{}) *tcpip.Error {
*o = tcpip.TCPDeferAcceptOption(e.deferAccept)
e.UnlockUser()
+ case *tcpip.OriginalDestinationOption:
+ ipt := e.stack.IPTables()
+ addr, port, err := ipt.OriginalDst(e.ID)
+ if err != nil {
+ return err
+ }
+ *o = tcpip.OriginalDestinationOption{
+ Addr: addr,
+ Port: port,
+ }
+
default:
return tcpip.ErrUnknownProtocolOption
}
diff --git a/pkg/tcpip/transport/tcp/protocol.go b/pkg/tcpip/transport/tcp/protocol.go
index 5d6174a59..b34e47bbd 100644
--- a/pkg/tcpip/transport/tcp/protocol.go
+++ b/pkg/tcpip/transport/tcp/protocol.go
@@ -49,7 +49,7 @@ const (
// DefaultReceiveBufferSize is the default size of the receive buffer
// for an endpoint.
- DefaultReceiveBufferSize = 32 << 10 // 32KB
+ DefaultReceiveBufferSize = 1 << 20 // 1MB
// MaxBufferSize is the largest size a receive/send buffer can grow to.
MaxBufferSize = 4 << 20 // 4MB
diff --git a/pkg/tcpip/transport/tcp/tcp_test.go b/pkg/tcpip/transport/tcp/tcp_test.go
index e67ec42b1..fb25b86b9 100644
--- a/pkg/tcpip/transport/tcp/tcp_test.go
+++ b/pkg/tcpip/transport/tcp/tcp_test.go
@@ -146,6 +146,24 @@ func TestActiveFailedConnectionAttemptIncrement(t *testing.T) {
}
}
+func TestCloseWithoutConnect(t *testing.T) {
+ c := context.New(t, defaultMTU)
+ defer c.Cleanup()
+
+ // Create TCP endpoint.
+ var err *tcpip.Error
+ c.EP, err = c.Stack().NewEndpoint(tcp.ProtocolNumber, ipv4.ProtocolNumber, &c.WQ)
+ if err != nil {
+ t.Fatalf("NewEndpoint failed: %s", err)
+ }
+
+ c.EP.Close()
+
+ if got := c.Stack().Stats().TCP.CurrentConnected.Value(); got != 0 {
+ t.Errorf("got stats.TCP.CurrentConnected.Value() = %d, want = 0", got)
+ }
+}
+
func TestTCPSegmentsSentIncrement(t *testing.T) {
c := context.New(t, defaultMTU)
defer c.Cleanup()
@@ -1276,68 +1294,91 @@ func TestConnectBindToDevice(t *testing.T) {
}
}
-func TestRstOnSynSent(t *testing.T) {
- c := context.New(t, defaultMTU)
- defer c.Cleanup()
+func TestSynSent(t *testing.T) {
+ for _, test := range []struct {
+ name string
+ reset bool
+ }{
+ {"RstOnSynSent", true},
+ {"CloseOnSynSent", false},
+ } {
+ t.Run(test.name, func(t *testing.T) {
+ c := context.New(t, defaultMTU)
+ defer c.Cleanup()
- // Create an endpoint, don't handshake because we want to interfere with the
- // handshake process.
- c.Create(-1)
+ // Create an endpoint, don't handshake because we want to interfere with the
+ // handshake process.
+ c.Create(-1)
- // Start connection attempt.
- waitEntry, ch := waiter.NewChannelEntry(nil)
- c.WQ.EventRegister(&waitEntry, waiter.EventOut)
- defer c.WQ.EventUnregister(&waitEntry)
+ // Start connection attempt.
+ waitEntry, ch := waiter.NewChannelEntry(nil)
+ c.WQ.EventRegister(&waitEntry, waiter.EventOut)
+ defer c.WQ.EventUnregister(&waitEntry)
- addr := tcpip.FullAddress{Addr: context.TestAddr, Port: context.TestPort}
- if err := c.EP.Connect(addr); err != tcpip.ErrConnectStarted {
- t.Fatalf("got Connect(%+v) = %s, want %s", addr, err, tcpip.ErrConnectStarted)
- }
+ addr := tcpip.FullAddress{Addr: context.TestAddr, Port: context.TestPort}
+ if err := c.EP.Connect(addr); err != tcpip.ErrConnectStarted {
+ t.Fatalf("got Connect(%+v) = %s, want %s", addr, err, tcpip.ErrConnectStarted)
+ }
- // Receive SYN packet.
- b := c.GetPacket()
- checker.IPv4(t, b,
- checker.TCP(
- checker.DstPort(context.TestPort),
- checker.TCPFlags(header.TCPFlagSyn),
- ),
- )
+ // Receive SYN packet.
+ b := c.GetPacket()
+ checker.IPv4(t, b,
+ checker.TCP(
+ checker.DstPort(context.TestPort),
+ checker.TCPFlags(header.TCPFlagSyn),
+ ),
+ )
- // Ensure that we've reached SynSent state
- if got, want := tcp.EndpointState(c.EP.State()), tcp.StateSynSent; got != want {
- t.Fatalf("got State() = %s, want %s", got, want)
- }
- tcpHdr := header.TCP(header.IPv4(b).Payload())
- c.IRS = seqnum.Value(tcpHdr.SequenceNumber())
+ if got, want := tcp.EndpointState(c.EP.State()), tcp.StateSynSent; got != want {
+ t.Fatalf("got State() = %s, want %s", got, want)
+ }
+ tcpHdr := header.TCP(header.IPv4(b).Payload())
+ c.IRS = seqnum.Value(tcpHdr.SequenceNumber())
- // Send a packet with a proper ACK and a RST flag to cause the socket
- // to Error and close out
- iss := seqnum.Value(789)
- rcvWnd := seqnum.Size(30000)
- c.SendPacket(nil, &context.Headers{
- SrcPort: tcpHdr.DestinationPort(),
- DstPort: tcpHdr.SourcePort(),
- Flags: header.TCPFlagRst | header.TCPFlagAck,
- SeqNum: iss,
- AckNum: c.IRS.Add(1),
- RcvWnd: rcvWnd,
- TCPOpts: nil,
- })
+ if test.reset {
+ // Send a packet with a proper ACK and a RST flag to cause the socket
+ // to error and close out.
+ iss := seqnum.Value(789)
+ rcvWnd := seqnum.Size(30000)
+ c.SendPacket(nil, &context.Headers{
+ SrcPort: tcpHdr.DestinationPort(),
+ DstPort: tcpHdr.SourcePort(),
+ Flags: header.TCPFlagRst | header.TCPFlagAck,
+ SeqNum: iss,
+ AckNum: c.IRS.Add(1),
+ RcvWnd: rcvWnd,
+ TCPOpts: nil,
+ })
+ } else {
+ c.EP.Close()
+ }
- // Wait for receive to be notified.
- select {
- case <-ch:
- case <-time.After(3 * time.Second):
- t.Fatal("timed out waiting for packet to arrive")
- }
+ // Wait for receive to be notified.
+ select {
+ case <-ch:
+ case <-time.After(3 * time.Second):
+ t.Fatal("timed out waiting for packet to arrive")
+ }
- if _, _, err := c.EP.Read(nil); err != tcpip.ErrConnectionRefused {
- t.Fatalf("got c.EP.Read(nil) = %s, want = %s", err, tcpip.ErrConnectionRefused)
- }
+ if test.reset {
+ if _, _, err := c.EP.Read(nil); err != tcpip.ErrConnectionRefused {
+ t.Fatalf("got c.EP.Read(nil) = %s, want = %s", err, tcpip.ErrConnectionRefused)
+ }
+ } else {
+ if _, _, err := c.EP.Read(nil); err != tcpip.ErrAborted {
+ t.Fatalf("got c.EP.Read(nil) = %s, want = %s", err, tcpip.ErrAborted)
+ }
+ }
- // Due to the RST the endpoint should be in an error state.
- if got, want := tcp.EndpointState(c.EP.State()), tcp.StateError; got != want {
- t.Fatalf("got State() = %s, want %s", got, want)
+ if got := c.Stack().Stats().TCP.CurrentConnected.Value(); got != 0 {
+ t.Errorf("got stats.TCP.CurrentConnected.Value() = %d, want = 0", got)
+ }
+
+ // Due to the RST the endpoint should be in an error state.
+ if got, want := tcp.EndpointState(c.EP.State()), tcp.StateError; got != want {
+ t.Fatalf("got State() = %s, want %s", got, want)
+ }
+ })
}
}
diff --git a/pkg/tcpip/transport/udp/endpoint.go b/pkg/tcpip/transport/udp/endpoint.go
index 6e692da07..b7d735889 100644
--- a/pkg/tcpip/transport/udp/endpoint.go
+++ b/pkg/tcpip/transport/udp/endpoint.go
@@ -483,10 +483,6 @@ func (e *endpoint) write(p tcpip.Payloader, opts tcpip.WriteOptions) (int64, <-c
nicID = e.BindNICID
}
- if to.Addr == header.IPv4Broadcast && !e.broadcast {
- return 0, nil, tcpip.ErrBroadcastDisabled
- }
-
dst, netProto, err := e.checkV4MappedLocked(*to)
if err != nil {
return 0, nil, err
@@ -503,6 +499,10 @@ func (e *endpoint) write(p tcpip.Payloader, opts tcpip.WriteOptions) (int64, <-c
resolve = route.Resolve
}
+ if !e.broadcast && route.IsBroadcast() {
+ return 0, nil, tcpip.ErrBroadcastDisabled
+ }
+
if route.IsResolutionRequired() {
if ch, err := resolve(nil); err != nil {
if err == tcpip.ErrWouldBlock {
diff --git a/pkg/tcpip/transport/udp/udp_test.go b/pkg/tcpip/transport/udp/udp_test.go
index 90781cf49..66e8911c8 100644
--- a/pkg/tcpip/transport/udp/udp_test.go
+++ b/pkg/tcpip/transport/udp/udp_test.go
@@ -2142,3 +2142,192 @@ func (c *testContext) checkEndpointReadStats(incr uint64, want tcpip.TransportEn
c.t.Errorf("Endpoint stats not matching for error %s got %+v want %+v", err, got, want)
}
}
+
+func TestOutgoingSubnetBroadcast(t *testing.T) {
+ const nicID1 = 1
+
+ ipv4Addr := tcpip.AddressWithPrefix{
+ Address: "\xc0\xa8\x01\x3a",
+ PrefixLen: 24,
+ }
+ ipv4Subnet := ipv4Addr.Subnet()
+ ipv4SubnetBcast := ipv4Subnet.Broadcast()
+ ipv4Gateway := tcpip.Address("\xc0\xa8\x01\x01")
+ ipv4AddrPrefix31 := tcpip.AddressWithPrefix{
+ Address: "\xc0\xa8\x01\x3a",
+ PrefixLen: 31,
+ }
+ ipv4Subnet31 := ipv4AddrPrefix31.Subnet()
+ ipv4Subnet31Bcast := ipv4Subnet31.Broadcast()
+ ipv4AddrPrefix32 := tcpip.AddressWithPrefix{
+ Address: "\xc0\xa8\x01\x3a",
+ PrefixLen: 32,
+ }
+ ipv4Subnet32 := ipv4AddrPrefix32.Subnet()
+ ipv4Subnet32Bcast := ipv4Subnet32.Broadcast()
+ ipv6Addr := tcpip.AddressWithPrefix{
+ Address: "\x20\x0a\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x01",
+ PrefixLen: 64,
+ }
+ ipv6Subnet := ipv6Addr.Subnet()
+ ipv6SubnetBcast := ipv6Subnet.Broadcast()
+ remNetAddr := tcpip.AddressWithPrefix{
+ Address: "\x64\x0a\x7b\x18",
+ PrefixLen: 24,
+ }
+ remNetSubnet := remNetAddr.Subnet()
+ remNetSubnetBcast := remNetSubnet.Broadcast()
+
+ tests := []struct {
+ name string
+ nicAddr tcpip.ProtocolAddress
+ routes []tcpip.Route
+ remoteAddr tcpip.Address
+ requiresBroadcastOpt bool
+ }{
+ {
+ name: "IPv4 Broadcast to local subnet",
+ nicAddr: tcpip.ProtocolAddress{
+ Protocol: header.IPv4ProtocolNumber,
+ AddressWithPrefix: ipv4Addr,
+ },
+ routes: []tcpip.Route{
+ {
+ Destination: ipv4Subnet,
+ NIC: nicID1,
+ },
+ },
+ remoteAddr: ipv4SubnetBcast,
+ requiresBroadcastOpt: true,
+ },
+ {
+ name: "IPv4 Broadcast to local /31 subnet",
+ nicAddr: tcpip.ProtocolAddress{
+ Protocol: header.IPv4ProtocolNumber,
+ AddressWithPrefix: ipv4AddrPrefix31,
+ },
+ routes: []tcpip.Route{
+ {
+ Destination: ipv4Subnet31,
+ NIC: nicID1,
+ },
+ },
+ remoteAddr: ipv4Subnet31Bcast,
+ requiresBroadcastOpt: false,
+ },
+ {
+ name: "IPv4 Broadcast to local /32 subnet",
+ nicAddr: tcpip.ProtocolAddress{
+ Protocol: header.IPv4ProtocolNumber,
+ AddressWithPrefix: ipv4AddrPrefix32,
+ },
+ routes: []tcpip.Route{
+ {
+ Destination: ipv4Subnet32,
+ NIC: nicID1,
+ },
+ },
+ remoteAddr: ipv4Subnet32Bcast,
+ requiresBroadcastOpt: false,
+ },
+ // IPv6 has no notion of a broadcast.
+ {
+ name: "IPv6 'Broadcast' to local subnet",
+ nicAddr: tcpip.ProtocolAddress{
+ Protocol: header.IPv6ProtocolNumber,
+ AddressWithPrefix: ipv6Addr,
+ },
+ routes: []tcpip.Route{
+ {
+ Destination: ipv6Subnet,
+ NIC: nicID1,
+ },
+ },
+ remoteAddr: ipv6SubnetBcast,
+ requiresBroadcastOpt: false,
+ },
+ {
+ name: "IPv4 Broadcast to remote subnet",
+ nicAddr: tcpip.ProtocolAddress{
+ Protocol: header.IPv4ProtocolNumber,
+ AddressWithPrefix: ipv4Addr,
+ },
+ routes: []tcpip.Route{
+ {
+ Destination: remNetSubnet,
+ Gateway: ipv4Gateway,
+ NIC: nicID1,
+ },
+ },
+ remoteAddr: remNetSubnetBcast,
+ requiresBroadcastOpt: true,
+ },
+ }
+
+ for _, test := range tests {
+ t.Run(test.name, func(t *testing.T) {
+ s := stack.New(stack.Options{
+ NetworkProtocols: []stack.NetworkProtocol{ipv4.NewProtocol(), ipv6.NewProtocol()},
+
+ TransportProtocols: []stack.TransportProtocol{udp.NewProtocol()},
+ })
+ e := channel.New(0, defaultMTU, "")
+ if err := s.CreateNIC(nicID1, e); err != nil {
+ t.Fatalf("CreateNIC(%d, _): %s", nicID1, err)
+ }
+ if err := s.AddProtocolAddress(nicID1, test.nicAddr); err != nil {
+ t.Fatalf("AddProtocolAddress(%d, %+v): %s", nicID1, test.nicAddr, err)
+ }
+
+ s.SetRouteTable(test.routes)
+
+ var netProto tcpip.NetworkProtocolNumber
+ switch l := len(test.remoteAddr); l {
+ case header.IPv4AddressSize:
+ netProto = header.IPv4ProtocolNumber
+ case header.IPv6AddressSize:
+ netProto = header.IPv6ProtocolNumber
+ default:
+ t.Fatalf("got unexpected address length = %d bytes", l)
+ }
+
+ wq := waiter.Queue{}
+ ep, err := s.NewEndpoint(udp.ProtocolNumber, netProto, &wq)
+ if err != nil {
+ t.Fatalf("NewEndpoint(%d, %d, _): %s", udp.ProtocolNumber, netProto, err)
+ }
+ defer ep.Close()
+
+ data := tcpip.SlicePayload([]byte{1, 2, 3, 4})
+ to := tcpip.FullAddress{
+ Addr: test.remoteAddr,
+ Port: 80,
+ }
+ opts := tcpip.WriteOptions{To: &to}
+ expectedErrWithoutBcastOpt := tcpip.ErrBroadcastDisabled
+ if !test.requiresBroadcastOpt {
+ expectedErrWithoutBcastOpt = nil
+ }
+
+ if n, _, err := ep.Write(data, opts); err != expectedErrWithoutBcastOpt {
+ t.Fatalf("got ep.Write(_, _) = (%d, _, %v), want = (_, _, %v)", n, err, expectedErrWithoutBcastOpt)
+ }
+
+ if err := ep.SetSockOptBool(tcpip.BroadcastOption, true); err != nil {
+ t.Fatalf("got SetSockOptBool(BroadcastOption, true): %s", err)
+ }
+
+ if n, _, err := ep.Write(data, opts); err != nil {
+ t.Fatalf("got ep.Write(_, _) = (%d, _, %s), want = (_, _, nil)", n, err)
+ }
+
+ if err := ep.SetSockOptBool(tcpip.BroadcastOption, false); err != nil {
+ t.Fatalf("got SetSockOptBool(BroadcastOption, false): %s", err)
+ }
+
+ if n, _, err := ep.Write(data, opts); err != expectedErrWithoutBcastOpt {
+ t.Fatalf("got ep.Write(_, _) = (%d, _, %v), want = (_, _, %v)", n, err, expectedErrWithoutBcastOpt)
+ }
+ })
+ }
+}
diff --git a/pkg/test/dockerutil/container.go b/pkg/test/dockerutil/container.go
index 441173ec2..5a2157951 100644
--- a/pkg/test/dockerutil/container.go
+++ b/pkg/test/dockerutil/container.go
@@ -454,15 +454,19 @@ func (c *Container) Wait(ctx context.Context) error {
// WaitTimeout waits for the container to exit with a timeout.
func (c *Container) WaitTimeout(ctx context.Context, timeout time.Duration) error {
- timeoutChan := time.After(timeout)
+ ctx, cancel := context.WithTimeout(ctx, timeout)
+ defer cancel()
statusChan, errChan := c.client.ContainerWait(ctx, c.id, container.WaitConditionNotRunning)
select {
+ case <-ctx.Done():
+ if ctx.Err() == context.DeadlineExceeded {
+ return fmt.Errorf("container %s timed out after %v seconds", c.Name, timeout.Seconds())
+ }
+ return nil
case err := <-errChan:
return err
case <-statusChan:
return nil
- case <-timeoutChan:
- return fmt.Errorf("container %s timed out after %v seconds", c.Name, timeout.Seconds())
}
}
@@ -487,6 +491,12 @@ func (c *Container) WaitForOutputSubmatch(ctx context.Context, pattern string, t
}
for exp := time.Now().Add(timeout); time.Now().Before(exp); {
+ select {
+ case <-ctx.Done():
+ return nil, ctx.Err()
+ default:
+ }
+
c.streams.Conn.SetDeadline(time.Now().Add(50 * time.Millisecond))
_, err := stdcopy.StdCopy(&c.streamBuf, &c.streamBuf, c.streams.Reader)
diff --git a/runsc/boot/vfs.go b/runsc/boot/vfs.go
index 9a1ed8e9e..cfe2d36aa 100644
--- a/runsc/boot/vfs.go
+++ b/runsc/boot/vfs.go
@@ -171,10 +171,19 @@ func (c *containerMounter) setupVFS2(ctx context.Context, conf *Config, procArgs
func (c *containerMounter) createMountNamespaceVFS2(ctx context.Context, conf *Config, creds *auth.Credentials) (*vfs.MountNamespace, error) {
fd := c.fds.remove()
- opts := strings.Join(p9MountData(fd, conf.FileAccess, true /* vfs2 */), ",")
+ opts := p9MountData(fd, conf.FileAccess, true /* vfs2 */)
+
+ if conf.OverlayfsStaleRead {
+ // We can't check for overlayfs here because sandbox is chroot'ed and gofer
+ // can only send mount options for specs.Mounts (specs.Root is missing
+ // Options field). So assume root is always on top of overlayfs.
+ opts = append(opts, "overlayfs_stale_read")
+ }
log.Infof("Mounting root over 9P, ioFD: %d", fd)
- mns, err := c.k.VFS().NewMountNamespace(ctx, creds, "", gofer.Name, &vfs.GetFilesystemOptions{Data: opts})
+ mns, err := c.k.VFS().NewMountNamespace(ctx, creds, "", gofer.Name, &vfs.GetFilesystemOptions{
+ Data: strings.Join(opts, ","),
+ })
if err != nil {
return nil, fmt.Errorf("setting up mount namespace: %w", err)
}
diff --git a/runsc/fsgofer/fsgofer.go b/runsc/fsgofer/fsgofer.go
index ebefeacf2..c6694c278 100644
--- a/runsc/fsgofer/fsgofer.go
+++ b/runsc/fsgofer/fsgofer.go
@@ -979,9 +979,12 @@ func (l *localFile) Readdir(offset uint64, count uint32) ([]p9.Dirent, error) {
skip := uint64(0)
- // Check if the file is at the correct position already. If not, seek to the
- // beginning and read the entire directory again.
- if l.lastDirentOffset != offset {
+ // Check if the file is at the correct position already. If not, seek to
+ // the beginning and read the entire directory again. We always seek if
+ // offset is 0, since this is side-effectual (equivalent to rewinddir(3),
+ // which causes the directory stream to resynchronize with the directory's
+ // current contents).
+ if l.lastDirentOffset != offset || offset == 0 {
if _, err := syscall.Seek(l.file.FD(), 0, 0); err != nil {
return nil, extractErrno(err)
}
diff --git a/test/benchmarks/database/BUILD b/test/benchmarks/database/BUILD
index 5e33465cd..572db665f 100644
--- a/test/benchmarks/database/BUILD
+++ b/test/benchmarks/database/BUILD
@@ -24,5 +24,6 @@ go_test(
deps = [
"//pkg/test/dockerutil",
"//test/benchmarks/harness",
+ "//test/benchmarks/tools",
],
)
diff --git a/test/benchmarks/database/redis_test.go b/test/benchmarks/database/redis_test.go
index 6d39f4d66..394fce820 100644
--- a/test/benchmarks/database/redis_test.go
+++ b/test/benchmarks/database/redis_test.go
@@ -16,15 +16,12 @@ package database
import (
"context"
- "fmt"
- "regexp"
- "strconv"
- "strings"
"testing"
"time"
"gvisor.dev/gvisor/pkg/test/dockerutil"
"gvisor.dev/gvisor/test/benchmarks/harness"
+ "gvisor.dev/gvisor/test/benchmarks/tools"
)
// All possible operations from redis. Note: "ping" will
@@ -99,16 +96,10 @@ func BenchmarkRedis(b *testing.B) {
b.Fatalf("failed to start redis with: %v", err)
}
- // runs redis benchmark -t operation for 100K requests against server.
- cmd := strings.Split(
- fmt.Sprintf("redis-benchmark --csv -t %s -h %s -p %d", operation, ip, serverPort), " ")
-
- // There is no -t PING_BULK for redis-benchmark, so adjust the command in that case.
- // Note that "ping" will run both PING_INLINE and PING_BULK.
- if operation == "PING_BULK" {
- cmd = strings.Split(
- fmt.Sprintf("redis-benchmark --csv -t ping -h %s -p %d", ip, serverPort), " ")
+ redis := tools.Redis{
+ Operation: operation,
}
+
// Reset profiles and timer to begin the measurement.
server.RestartProfiles()
b.ResetTimer()
@@ -117,81 +108,16 @@ func BenchmarkRedis(b *testing.B) {
defer client.CleanUp(ctx)
out, err := client.Run(ctx, dockerutil.RunOpts{
Image: "benchmarks/redis",
- }, cmd...)
+ }, redis.MakeCmd(ip, serverPort)...)
if err != nil {
b.Fatalf("redis-benchmark failed with: %v", err)
}
// Stop time while we parse results.
b.StopTimer()
- result, err := parseOperation(operation, out)
- if err != nil {
- b.Fatalf("parsing result %s failed with err: %v", out, err)
- }
- b.ReportMetric(result, operation) // operations per second
+ redis.Report(b, out)
b.StartTimer()
}
})
}
}
-
-// parseOperation grabs the metric operations per second from redis-benchmark output.
-func parseOperation(operation, data string) (float64, error) {
- re := regexp.MustCompile(fmt.Sprintf(`"%s( .*)?","(\d*\.\d*)"`, operation))
- match := re.FindStringSubmatch(data)
- // If no match, simply don't add it to the result map.
- if len(match) < 3 {
- return 0.0, fmt.Errorf("could not find %s in %s", operation, data)
- }
- return strconv.ParseFloat(match[2], 64)
-}
-
-// TestParser tests the parser on sample data.
-func TestParser(t *testing.T) {
- sampleData := `
- "PING_INLINE","48661.80"
- "PING_BULK","50301.81"
- "SET","48923.68"
- "GET","49382.71"
- "INCR","49975.02"
- "LPUSH","49875.31"
- "RPUSH","50276.52"
- "LPOP","50327.12"
- "RPOP","50556.12"
- "SADD","49504.95"
- "HSET","49504.95"
- "SPOP","50025.02"
- "LPUSH (needed to benchmark LRANGE)","48875.86"
- "LRANGE_100 (first 100 elements)","33955.86"
- "LRANGE_300 (first 300 elements)","16550.81"
- "LRANGE_500 (first 450 elements)","13653.74"
- "LRANGE_600 (first 600 elements)","11219.57"
- "MSET (10 keys)","44682.75"
- `
- wants := map[string]float64{
- "PING_INLINE": 48661.80,
- "PING_BULK": 50301.81,
- "SET": 48923.68,
- "GET": 49382.71,
- "INCR": 49975.02,
- "LPUSH": 49875.31,
- "RPUSH": 50276.52,
- "LPOP": 50327.12,
- "RPOP": 50556.12,
- "SADD": 49504.95,
- "HSET": 49504.95,
- "SPOP": 50025.02,
- "LRANGE_100": 33955.86,
- "LRANGE_300": 16550.81,
- "LRANGE_500": 13653.74,
- "LRANGE_600": 11219.57,
- "MSET": 44682.75,
- }
- for op, want := range wants {
- if got, err := parseOperation(op, sampleData); err != nil {
- t.Fatalf("failed to parse %s: %v", op, err)
- } else if want != got {
- t.Fatalf("wanted %f for op %s, got %f", want, op, got)
- }
- }
-}
diff --git a/test/benchmarks/fs/BUILD b/test/benchmarks/fs/BUILD
index 2874cdbb3..20654d88f 100644
--- a/test/benchmarks/fs/BUILD
+++ b/test/benchmarks/fs/BUILD
@@ -12,12 +12,20 @@ go_library(
go_test(
name = "fs_test",
size = "large",
- srcs = ["bazel_test.go"],
+ srcs = [
+ "bazel_test.go",
+ "fio_test.go",
+ ],
library = ":fs",
tags = [
# Requires docker and runsc to be configured before test runs.
"local",
"manual",
],
- deps = ["//pkg/test/dockerutil"],
+ deps = [
+ "//pkg/test/dockerutil",
+ "//test/benchmarks/harness",
+ "//test/benchmarks/tools",
+ "@com_github_docker_docker//api/types/mount:go_default_library",
+ ],
)
diff --git a/test/benchmarks/fs/bazel_test.go b/test/benchmarks/fs/bazel_test.go
index 9b652fd43..f4236ba37 100644
--- a/test/benchmarks/fs/bazel_test.go
+++ b/test/benchmarks/fs/bazel_test.go
@@ -20,10 +20,22 @@ import (
"testing"
"gvisor.dev/gvisor/pkg/test/dockerutil"
+ "gvisor.dev/gvisor/test/benchmarks/harness"
)
// Note: CleanCache versions of this test require running with root permissions.
-func BenchmarkABSL(b *testing.B) {
+func BenchmarkBuildABSL(b *testing.B) {
+ runBuildBenchmark(b, "benchmarks/absl", "/abseil-cpp", "absl/base/...")
+}
+
+// Note: CleanCache versions of this test require running with root permissions.
+// Note: This test takes on the order of 10m per permutation for runsc on kvm.
+func BenchmarkBuildRunsc(b *testing.B) {
+ runBuildBenchmark(b, "benchmarks/runsc", "/gvisor", "runsc:runsc")
+}
+
+func runBuildBenchmark(b *testing.B, image, workdir, target string) {
+ b.Helper()
// Get a machine from the Harness on which to run.
machine, err := h.GetMachine()
if err != nil {
@@ -50,20 +62,18 @@ func BenchmarkABSL(b *testing.B) {
container := machine.GetContainer(ctx, b)
defer container.CleanUp(ctx)
- workdir := "/abseil-cpp"
-
// Start a container and sleep by an order of b.N.
if err := container.Spawn(ctx, dockerutil.RunOpts{
- Image: "benchmarks/absl",
+ Image: image,
}, "sleep", fmt.Sprintf("%d", 1000000)); err != nil {
b.Fatalf("run failed with: %v", err)
}
// If we are running on a tmpfs, copy to /tmp which is a tmpfs.
if bm.tmpfs {
- if _, err := container.Exec(ctx, dockerutil.ExecOpts{},
- "cp", "-r", "/abseil-cpp", "/tmp/."); err != nil {
- b.Fatal("failed to copy directory: %v", err)
+ if out, err := container.Exec(ctx, dockerutil.ExecOpts{},
+ "cp", "-r", workdir, "/tmp/."); err != nil {
+ b.Fatal("failed to copy directory: %v %s", err, out)
}
workdir = "/tmp" + workdir
}
@@ -77,15 +87,15 @@ func BenchmarkABSL(b *testing.B) {
b.StopTimer()
// Drop Caches for clear cache runs.
if bm.clearCache {
- if out, err := machine.RunCommand("/bin/sh", "-c", "sync && sysctl vm.drop_caches=3"); err != nil {
- b.Skipf("failed to drop caches: %v %s. You probably need root.", err, out)
+ if err := harness.DropCaches(machine); err != nil {
+ b.Skipf("failed to drop caches: %v. You probably need root.", err)
}
}
b.StartTimer()
got, err := container.Exec(ctx, dockerutil.ExecOpts{
WorkDir: workdir,
- }, "bazel", "build", "-c", "opt", "absl/base/...")
+ }, "bazel", "build", "-c", "opt", target)
if err != nil {
b.Fatalf("build failed with: %v", err)
}
diff --git a/test/benchmarks/fs/fio_test.go b/test/benchmarks/fs/fio_test.go
new file mode 100644
index 000000000..65874ed8b
--- /dev/null
+++ b/test/benchmarks/fs/fio_test.go
@@ -0,0 +1,170 @@
+// Copyright 2020 The gVisor Authors.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+// http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+package fs
+
+import (
+ "context"
+ "fmt"
+ "path/filepath"
+ "strings"
+ "testing"
+
+ "github.com/docker/docker/api/types/mount"
+ "gvisor.dev/gvisor/pkg/test/dockerutil"
+ "gvisor.dev/gvisor/test/benchmarks/harness"
+ "gvisor.dev/gvisor/test/benchmarks/tools"
+)
+
+// BenchmarkFio runs fio on the runtime under test. There are 4 basic test
+// cases each run on a tmpfs mount and a bind mount. Fio requires root so that
+// caches can be dropped.
+func BenchmarkFio(b *testing.B) {
+ testCases := []tools.Fio{
+ tools.Fio{
+ Test: "write",
+ Size: "5G",
+ Blocksize: "1M",
+ Iodepth: 4,
+ },
+ tools.Fio{
+ Test: "read",
+ Size: "5G",
+ Blocksize: "1M",
+ Iodepth: 4,
+ },
+ tools.Fio{
+ Test: "randwrite",
+ Size: "5G",
+ Blocksize: "4K",
+ Iodepth: 4,
+ Time: 30,
+ },
+ tools.Fio{
+ Test: "randread",
+ Size: "5G",
+ Blocksize: "4K",
+ Iodepth: 4,
+ Time: 30,
+ },
+ }
+
+ machine, err := h.GetMachine()
+ if err != nil {
+ b.Fatalf("failed to get machine with: %v", err)
+ }
+ defer machine.CleanUp()
+
+ for _, fsType := range []mount.Type{mount.TypeBind, mount.TypeTmpfs} {
+ for _, tc := range testCases {
+ testName := strings.Title(tc.Test) + strings.Title(string(fsType))
+ b.Run(testName, func(b *testing.B) {
+ ctx := context.Background()
+ container := machine.GetContainer(ctx, b)
+ defer container.CleanUp(ctx)
+
+ // Directory and filename inside container where fio will read/write.
+ outdir := "/data"
+ outfile := filepath.Join(outdir, "test.txt")
+
+ // Make the required mount and grab a cleanup for bind mounts
+ // as they are backed by a temp directory (mktemp).
+ mnt, mountCleanup, err := makeMount(machine, fsType, outdir)
+ if err != nil {
+ b.Fatalf("failed to make mount: %v", err)
+ }
+ defer mountCleanup()
+
+ // Start the container with the mount.
+ if err := container.Spawn(
+ ctx,
+ dockerutil.RunOpts{
+ Image: "benchmarks/fio",
+ Mounts: []mount.Mount{
+ mnt,
+ },
+ },
+ // Sleep on the order of b.N.
+ "sleep", fmt.Sprintf("%d", 1000*b.N),
+ ); err != nil {
+ b.Fatalf("failed to start fio container with: %v", err)
+ }
+
+ // For reads, we need a file to read so make one inside the container.
+ if strings.Contains(tc.Test, "read") {
+ fallocateCmd := fmt.Sprintf("fallocate -l %s %s", tc.Size, outfile)
+ if out, err := container.Exec(ctx, dockerutil.ExecOpts{},
+ strings.Split(fallocateCmd, " ")...); err != nil {
+ b.Fatalf("failed to create readable file on mount: %v, %s", err, out)
+ }
+ }
+
+ // Drop caches just before running.
+ if err := harness.DropCaches(machine); err != nil {
+ b.Skipf("failed to drop caches with %v. You probably need root.", err)
+ }
+ cmd := tc.MakeCmd(outfile)
+ container.RestartProfiles()
+ b.ResetTimer()
+ for i := 0; i < b.N; i++ {
+ // Run fio.
+ data, err := container.Exec(ctx, dockerutil.ExecOpts{}, cmd...)
+ if err != nil {
+ b.Fatalf("failed to run cmd %v: %v", cmd, err)
+ }
+ b.StopTimer()
+ tc.Report(b, data)
+ // If b.N is used (i.e. we run for an hour), we should drop caches
+ // after each run.
+ if err := harness.DropCaches(machine); err != nil {
+ b.Fatalf("failed to drop caches: %v", err)
+ }
+ b.StartTimer()
+ }
+ })
+ }
+ }
+}
+
+// makeMount makes a mount and cleanup based on the requested type. Bind
+// and volume mounts are backed by a temp directory made with mktemp.
+// tmpfs mounts require no such backing and are just made.
+// It is up to the caller to call the returned cleanup.
+func makeMount(machine harness.Machine, mountType mount.Type, target string) (mount.Mount, func(), error) {
+ switch mountType {
+ case mount.TypeVolume, mount.TypeBind:
+ dir, err := machine.RunCommand("mktemp", "-d")
+ if err != nil {
+ return mount.Mount{}, func() {}, fmt.Errorf("failed to create tempdir: %v", err)
+ }
+ dir = strings.TrimSuffix(dir, "\n")
+
+ out, err := machine.RunCommand("chmod", "777", dir)
+ if err != nil {
+ machine.RunCommand("rm", "-rf", dir)
+ return mount.Mount{}, func() {}, fmt.Errorf("failed modify directory: %v %s", err, out)
+ }
+ return mount.Mount{
+ Target: target,
+ Source: dir,
+ Type: mount.TypeBind,
+ }, func() { machine.RunCommand("rm", "-rf", dir) }, nil
+ case mount.TypeTmpfs:
+ return mount.Mount{
+ Target: target,
+ Type: mount.TypeTmpfs,
+ }, func() {}, nil
+ default:
+ return mount.Mount{}, func() {}, fmt.Errorf("illegal mount time not supported: %v", mountType)
+ }
+}
diff --git a/test/benchmarks/network/BUILD b/test/benchmarks/network/BUILD
index 363041fb7..d15cd55ee 100644
--- a/test/benchmarks/network/BUILD
+++ b/test/benchmarks/network/BUILD
@@ -15,6 +15,8 @@ go_test(
srcs = [
"httpd_test.go",
"iperf_test.go",
+ "nginx_test.go",
+ "node_test.go",
],
library = ":network",
tags = [
@@ -26,5 +28,6 @@ go_test(
"//pkg/test/dockerutil",
"//pkg/test/testutil",
"//test/benchmarks/harness",
+ "//test/benchmarks/tools",
],
)
diff --git a/test/benchmarks/network/httpd_test.go b/test/benchmarks/network/httpd_test.go
index fe23ca949..07833f9cd 100644
--- a/test/benchmarks/network/httpd_test.go
+++ b/test/benchmarks/network/httpd_test.go
@@ -16,12 +16,11 @@ package network
import (
"context"
"fmt"
- "regexp"
- "strconv"
"testing"
"gvisor.dev/gvisor/pkg/test/dockerutil"
"gvisor.dev/gvisor/test/benchmarks/harness"
+ "gvisor.dev/gvisor/test/benchmarks/tools"
)
// see Dockerfile '//images/benchmarks/httpd'.
@@ -52,13 +51,16 @@ func BenchmarkHttpdConcurrency(b *testing.B) {
defer serverMachine.CleanUp()
// The test iterates over client concurrency, so set other parameters.
- requests := 10000
concurrency := []int{1, 5, 10, 25}
- doc := docs["10Kb"]
for _, c := range concurrency {
b.Run(fmt.Sprintf("%d", c), func(b *testing.B) {
- runHttpd(b, clientMachine, serverMachine, doc, requests, c)
+ hey := &tools.Hey{
+ Requests: 10000,
+ Concurrency: c,
+ Doc: docs["10Kb"],
+ }
+ runHttpd(b, clientMachine, serverMachine, hey)
})
}
}
@@ -78,18 +80,20 @@ func BenchmarkHttpdDocSize(b *testing.B) {
}
defer serverMachine.CleanUp()
- requests := 10000
- concurrency := 1
-
for name, filename := range docs {
b.Run(name, func(b *testing.B) {
- runHttpd(b, clientMachine, serverMachine, filename, requests, concurrency)
+ hey := &tools.Hey{
+ Requests: 10000,
+ Concurrency: 1,
+ Doc: filename,
+ }
+ runHttpd(b, clientMachine, serverMachine, hey)
})
}
}
// runHttpd runs a single test run.
-func runHttpd(b *testing.B, clientMachine, serverMachine harness.Machine, doc string, requests, concurrency int) {
+func runHttpd(b *testing.B, clientMachine, serverMachine harness.Machine, hey *tools.Hey) {
b.Helper()
// Grab a container from the server.
@@ -98,11 +102,11 @@ func runHttpd(b *testing.B, clientMachine, serverMachine harness.Machine, doc st
defer server.CleanUp(ctx)
// Copy the docs to /tmp and serve from there.
- cmd := "mkdir -p /tmp/html; cp -r /local /tmp/html/.; apache2 -X"
+ cmd := "mkdir -p /tmp/html; cp -r /local/* /tmp/html/.; apache2 -X"
port := 80
// Start the server.
- server.Spawn(ctx, dockerutil.RunOpts{
+ if err := server.Spawn(ctx, dockerutil.RunOpts{
Image: "benchmarks/httpd",
Ports: []int{port},
Env: []string{
@@ -113,7 +117,9 @@ func runHttpd(b *testing.B, clientMachine, serverMachine harness.Machine, doc st
"APACHE_LOG_DIR=/tmp",
"APACHE_PID_FILE=/tmp/apache.pid",
},
- }, "sh", "-c", cmd)
+ }, "sh", "-c", cmd); err != nil {
+ b.Fatalf("failed to start server: %v")
+ }
ip, err := serverMachine.IPAddress()
if err != nil {
@@ -132,146 +138,18 @@ func runHttpd(b *testing.B, clientMachine, serverMachine harness.Machine, doc st
client := clientMachine.GetNativeContainer(ctx, b)
defer client.CleanUp(ctx)
- path := fmt.Sprintf("http://%s:%d/%s", ip, servingPort, doc)
- // See apachebench (ab) for flags.
- cmd = fmt.Sprintf("ab -n %d -c %d %s", requests, concurrency, path)
-
b.ResetTimer()
server.RestartProfiles()
for i := 0; i < b.N; i++ {
out, err := client.Run(ctx, dockerutil.RunOpts{
- Image: "benchmarks/ab",
- }, "sh", "-c", cmd)
+ Image: "benchmarks/hey",
+ }, hey.MakeCmd(ip, servingPort)...)
if err != nil {
b.Fatalf("run failed with: %v", err)
}
b.StopTimer()
-
- // Parse and report custom metrics.
- transferRate, err := parseTransferRate(out)
- if err != nil {
- b.Logf("failed to parse transferrate: %v", err)
- }
- b.ReportMetric(transferRate*1024, "transfer_rate") // Convert from Kb/s to b/s.
-
- latency, err := parseLatency(out)
- if err != nil {
- b.Logf("failed to parse latency: %v", err)
- }
- b.ReportMetric(latency/1000, "mean_latency") // Convert from ms to s.
-
- reqPerSecond, err := parseRequestsPerSecond(out)
- if err != nil {
- b.Logf("failed to parse requests per second: %v", err)
- }
- b.ReportMetric(reqPerSecond, "requests_per_second")
-
+ hey.Report(b, out)
b.StartTimer()
}
}
-
-var transferRateRE = regexp.MustCompile(`Transfer rate:\s+(\d+\.?\d+?)\s+\[Kbytes/sec\]\s+received`)
-
-// parseTransferRate parses transfer rate from apachebench output.
-func parseTransferRate(data string) (float64, error) {
- match := transferRateRE.FindStringSubmatch(data)
- if len(match) < 2 {
- return 0, fmt.Errorf("failed get bandwidth: %s", data)
- }
- return strconv.ParseFloat(match[1], 64)
-}
-
-var latencyRE = regexp.MustCompile(`Total:\s+\d+\s+(\d+)\s+(\d+\.?\d+?)\s+\d+\s+\d+\s`)
-
-// parseLatency parses latency from apachebench output.
-func parseLatency(data string) (float64, error) {
- match := latencyRE.FindStringSubmatch(data)
- if len(match) < 2 {
- return 0, fmt.Errorf("failed get bandwidth: %s", data)
- }
- return strconv.ParseFloat(match[1], 64)
-}
-
-var requestsPerSecondRE = regexp.MustCompile(`Requests per second:\s+(\d+\.?\d+?)\s+`)
-
-// parseRequestsPerSecond parses requests per second from apachebench output.
-func parseRequestsPerSecond(data string) (float64, error) {
- match := requestsPerSecondRE.FindStringSubmatch(data)
- if len(match) < 2 {
- return 0, fmt.Errorf("failed get bandwidth: %s", data)
- }
- return strconv.ParseFloat(match[1], 64)
-}
-
-// Sample output from apachebench.
-const sampleData = `This is ApacheBench, Version 2.3 <$Revision: 1826891 $>
-Copyright 1996 Adam Twiss, Zeus Technology Ltd, http://www.zeustech.net/
-Licensed to The Apache Software Foundation, http://www.apache.org/
-
-Benchmarking 10.10.10.10 (be patient).....done
-
-
-Server Software: Apache/2.4.38
-Server Hostname: 10.10.10.10
-Server Port: 80
-
-Document Path: /latin10k.txt
-Document Length: 210 bytes
-
-Concurrency Level: 1
-Time taken for tests: 0.180 seconds
-Complete requests: 100
-Failed requests: 0
-Non-2xx responses: 100
-Total transferred: 38800 bytes
-HTML transferred: 21000 bytes
-Requests per second: 556.44 [#/sec] (mean)
-Time per request: 1.797 [ms] (mean)
-Time per request: 1.797 [ms] (mean, across all concurrent requests)
-Transfer rate: 210.84 [Kbytes/sec] received
-
-Connection Times (ms)
- min mean[+/-sd] median max
-Connect: 0 0 0.2 0 2
-Processing: 1 2 1.0 1 8
-Waiting: 1 1 1.0 1 7
-Total: 1 2 1.2 1 10
-
-Percentage of the requests served within a certain time (ms)
- 50% 1
- 66% 2
- 75% 2
- 80% 2
- 90% 2
- 95% 3
- 98% 7
- 99% 10
- 100% 10 (longest request)`
-
-// TestParsers checks the parsers work.
-func TestParsers(t *testing.T) {
- want := 210.84
- got, err := parseTransferRate(sampleData)
- if err != nil {
- t.Fatalf("failed to parse transfer rate with error: %v", err)
- } else if got != want {
- t.Fatalf("parseTransferRate got: %f, want: %f", got, want)
- }
-
- want = 2.0
- got, err = parseLatency(sampleData)
- if err != nil {
- t.Fatalf("failed to parse transfer rate with error: %v", err)
- } else if got != want {
- t.Fatalf("parseLatency got: %f, want: %f", got, want)
- }
-
- want = 556.44
- got, err = parseRequestsPerSecond(sampleData)
- if err != nil {
- t.Fatalf("failed to parse transfer rate with error: %v", err)
- } else if got != want {
- t.Fatalf("parseRequestsPerSecond got: %f, want: %f", got, want)
- }
-}
diff --git a/test/benchmarks/network/iperf_test.go b/test/benchmarks/network/iperf_test.go
index a5e198e14..b8ab7dfb8 100644
--- a/test/benchmarks/network/iperf_test.go
+++ b/test/benchmarks/network/iperf_test.go
@@ -15,19 +15,18 @@ package network
import (
"context"
- "fmt"
- "regexp"
- "strconv"
- "strings"
"testing"
"gvisor.dev/gvisor/pkg/test/dockerutil"
"gvisor.dev/gvisor/pkg/test/testutil"
"gvisor.dev/gvisor/test/benchmarks/harness"
+ "gvisor.dev/gvisor/test/benchmarks/tools"
)
func BenchmarkIperf(b *testing.B) {
- const time = 10 // time in seconds to run the client.
+ iperf := tools.Iperf{
+ Time: 10, // time in seconds to run client.
+ }
clientMachine, err := h.GetMachine()
if err != nil {
@@ -92,10 +91,6 @@ func BenchmarkIperf(b *testing.B) {
if err := harness.WaitUntilServing(ctx, clientMachine, ip, servingPort); err != nil {
b.Fatalf("failed to wait for server: %v", err)
}
-
- // iperf report in Kb realtime
- cmd := fmt.Sprintf("iperf -f K --realtime --time %d -c %s -p %d", time, ip.String(), servingPort)
-
// Run the client.
b.ResetTimer()
@@ -105,46 +100,14 @@ func BenchmarkIperf(b *testing.B) {
for i := 0; i < b.N; i++ {
out, err := client.Run(ctx, dockerutil.RunOpts{
Image: "benchmarks/iperf",
- }, strings.Split(cmd, " ")...)
+ }, iperf.MakeCmd(ip, servingPort)...)
if err != nil {
b.Fatalf("failed to run client: %v", err)
}
b.StopTimer()
-
- // Parse bandwidth and report it.
- bW, err := bandwidth(out)
- if err != nil {
- b.Fatalf("failed to parse bandwitdth from %s: %v", out, err)
- }
- b.ReportMetric(bW*1024, "bandwidth") // Convert from Kb/s to b/s.
+ iperf.Report(b, out)
b.StartTimer()
}
})
}
}
-
-// bandwidth parses the Bandwidth number from an iperf report. A sample is below.
-func bandwidth(data string) (float64, error) {
- re := regexp.MustCompile(`\[\s*\d+\][^\n]+\s+(\d+\.?\d*)\s+KBytes/sec`)
- match := re.FindStringSubmatch(data)
- if len(match) < 1 {
- return 0, fmt.Errorf("failed get bandwidth: %s", data)
- }
- return strconv.ParseFloat(match[1], 64)
-}
-
-func TestParser(t *testing.T) {
- sampleData := `
-------------------------------------------------------------
-Client connecting to 10.138.15.215, TCP port 32779
-TCP window size: 45.0 KByte (default)
-------------------------------------------------------------
-[ 3] local 10.138.15.216 port 32866 connected with 10.138.15.215 port 32779
-[ ID] Interval Transfer Bandwidth
-[ 3] 0.0-10.0 sec 459520 KBytes 45900 KBytes/sec
-`
- bandwidth, err := bandwidth(sampleData)
- if err != nil || bandwidth != 45900 {
- t.Fatalf("failed with: %v and %f", err, bandwidth)
- }
-}
diff --git a/test/benchmarks/network/nginx_test.go b/test/benchmarks/network/nginx_test.go
new file mode 100644
index 000000000..5965652a5
--- /dev/null
+++ b/test/benchmarks/network/nginx_test.go
@@ -0,0 +1,104 @@
+// Copyright 2020 The gVisor Authors.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+// http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+package network
+
+import (
+ "context"
+ "fmt"
+ "testing"
+
+ "gvisor.dev/gvisor/pkg/test/dockerutil"
+ "gvisor.dev/gvisor/test/benchmarks/harness"
+ "gvisor.dev/gvisor/test/benchmarks/tools"
+)
+
+// BenchmarkNginxConcurrency iterates the concurrency argument and tests
+// how well the runtime under test handles requests in parallel.
+// TODO(zkoopmans): Update with different doc sizes like Httpd.
+func BenchmarkNginxConcurrency(b *testing.B) {
+ // Grab a machine for the client and server.
+ clientMachine, err := h.GetMachine()
+ if err != nil {
+ b.Fatalf("failed to get client: %v", err)
+ }
+ defer clientMachine.CleanUp()
+
+ serverMachine, err := h.GetMachine()
+ if err != nil {
+ b.Fatalf("failed to get server: %v", err)
+ }
+ defer serverMachine.CleanUp()
+
+ concurrency := []int{1, 5, 10, 25}
+ for _, c := range concurrency {
+ b.Run(fmt.Sprintf("%d", c), func(b *testing.B) {
+ hey := &tools.Hey{
+ Requests: 10000,
+ Concurrency: c,
+ }
+ runNginx(b, clientMachine, serverMachine, hey)
+ })
+ }
+}
+
+// runHttpd runs a single test run.
+func runNginx(b *testing.B, clientMachine, serverMachine harness.Machine, hey *tools.Hey) {
+ b.Helper()
+
+ // Grab a container from the server.
+ ctx := context.Background()
+ server := serverMachine.GetContainer(ctx, b)
+ defer server.CleanUp(ctx)
+
+ port := 80
+ // Start the server.
+ if err := server.Spawn(ctx,
+ dockerutil.RunOpts{
+ Image: "benchmarks/nginx",
+ Ports: []int{port},
+ }); err != nil {
+ b.Fatalf("server failed to start: %v", err)
+ }
+
+ ip, err := serverMachine.IPAddress()
+ if err != nil {
+ b.Fatalf("failed to find server ip: %v", err)
+ }
+
+ servingPort, err := server.FindPort(ctx, port)
+ if err != nil {
+ b.Fatalf("failed to find server port %d: %v", port, err)
+ }
+
+ // Check the server is serving.
+ harness.WaitUntilServing(ctx, clientMachine, ip, servingPort)
+
+ // Grab a client.
+ client := clientMachine.GetNativeContainer(ctx, b)
+ defer client.CleanUp(ctx)
+
+ b.ResetTimer()
+ server.RestartProfiles()
+ for i := 0; i < b.N; i++ {
+ out, err := client.Run(ctx, dockerutil.RunOpts{
+ Image: "benchmarks/hey",
+ }, hey.MakeCmd(ip, servingPort)...)
+ if err != nil {
+ b.Fatalf("run failed with: %v", err)
+ }
+ b.StopTimer()
+ hey.Report(b, out)
+ b.StartTimer()
+ }
+}
diff --git a/test/benchmarks/network/node_test.go b/test/benchmarks/network/node_test.go
new file mode 100644
index 000000000..5b568cfe5
--- /dev/null
+++ b/test/benchmarks/network/node_test.go
@@ -0,0 +1,131 @@
+// Copyright 2020 The gVisor Authors.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+// http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+package network
+
+import (
+ "context"
+ "fmt"
+ "testing"
+ "time"
+
+ "gvisor.dev/gvisor/pkg/test/dockerutil"
+ "gvisor.dev/gvisor/test/benchmarks/harness"
+ "gvisor.dev/gvisor/test/benchmarks/tools"
+)
+
+// BenchmarkNode runs 10K requests using 'hey' against a Node server run on
+// 'runtime'. The server responds to requests by grabbing some data in a
+// redis instance and returns the data in its reponse. The test loops through
+// increasing amounts of concurency for requests.
+func BenchmarkNode(b *testing.B) {
+ requests := 10000
+ concurrency := []int{1, 5, 10, 25}
+
+ for _, c := range concurrency {
+ b.Run(fmt.Sprintf("Concurrency%d", c), func(b *testing.B) {
+ hey := &tools.Hey{
+ Requests: requests,
+ Concurrency: c,
+ }
+ runNode(b, hey)
+ })
+ }
+}
+
+// runNode runs the test for a given # of requests and concurrency.
+func runNode(b *testing.B, hey *tools.Hey) {
+ b.Helper()
+
+ // The machine to hold Redis and the Node Server.
+ serverMachine, err := h.GetMachine()
+ if err != nil {
+ b.Fatal("failed to get machine with: %v", err)
+ }
+ defer serverMachine.CleanUp()
+
+ // The machine to run 'hey'.
+ clientMachine, err := h.GetMachine()
+ if err != nil {
+ b.Fatal("failed to get machine with: %v", err)
+ }
+ defer clientMachine.CleanUp()
+
+ ctx := context.Background()
+
+ // Spawn a redis instance for the app to use.
+ redis := serverMachine.GetNativeContainer(ctx, b)
+ if err := redis.Spawn(ctx, dockerutil.RunOpts{
+ Image: "benchmarks/redis",
+ }); err != nil {
+ b.Fatalf("failed to spwan redis instance: %v", err)
+ }
+ defer redis.CleanUp(ctx)
+
+ if out, err := redis.WaitForOutput(ctx, "Ready to accept connections", 3*time.Second); err != nil {
+ b.Fatalf("failed to start redis server: %v %s", err, out)
+ }
+ redisIP, err := redis.FindIP(ctx, false)
+ if err != nil {
+ b.Fatalf("failed to get IP from redis instance: %v", err)
+ }
+
+ // Node runs on port 8080.
+ port := 8080
+
+ // Start-up the Node server.
+ nodeApp := serverMachine.GetContainer(ctx, b)
+ if err := nodeApp.Spawn(ctx, dockerutil.RunOpts{
+ Image: "benchmarks/node",
+ WorkDir: "/usr/src/app",
+ Links: []string{redis.MakeLink("redis")},
+ Ports: []int{port},
+ }, "node", "index.js", redisIP.String()); err != nil {
+ b.Fatalf("failed to spawn node instance: %v", err)
+ }
+ defer nodeApp.CleanUp(ctx)
+
+ servingIP, err := serverMachine.IPAddress()
+ if err != nil {
+ b.Fatalf("failed to get ip from server: %v", err)
+ }
+
+ servingPort, err := nodeApp.FindPort(ctx, port)
+ if err != nil {
+ b.Fatalf("failed to port from node instance: %v", err)
+ }
+
+ // Wait until the Client sees the server as up.
+ harness.WaitUntilServing(ctx, clientMachine, servingIP, servingPort)
+
+ heyCmd := hey.MakeCmd(servingIP, servingPort)
+
+ nodeApp.RestartProfiles()
+ b.ResetTimer()
+
+ for i := 0; i < b.N; i++ {
+ // the client should run on Native.
+ client := clientMachine.GetNativeContainer(ctx, b)
+ out, err := client.Run(ctx, dockerutil.RunOpts{
+ Image: "benchmarks/hey",
+ }, heyCmd...)
+ if err != nil {
+ b.Fatalf("hey container failed: %v logs: %s", err, out)
+ }
+
+ // Stop the timer to parse the data and report stats.
+ b.StopTimer()
+ hey.Report(b, out)
+ b.StartTimer()
+ }
+}
diff --git a/test/benchmarks/tools/BUILD b/test/benchmarks/tools/BUILD
new file mode 100644
index 000000000..4358551bc
--- /dev/null
+++ b/test/benchmarks/tools/BUILD
@@ -0,0 +1,29 @@
+load("//tools:defs.bzl", "go_library", "go_test")
+
+package(licenses = ["notice"])
+
+go_library(
+ name = "tools",
+ srcs = [
+ "ab.go",
+ "fio.go",
+ "hey.go",
+ "iperf.go",
+ "redis.go",
+ "tools.go",
+ ],
+ visibility = ["//:sandbox"],
+)
+
+go_test(
+ name = "tools_test",
+ size = "small",
+ srcs = [
+ "ab_test.go",
+ "fio_test.go",
+ "hey_test.go",
+ "iperf_test.go",
+ "redis_test.go",
+ ],
+ library = ":tools",
+)
diff --git a/test/benchmarks/tools/ab.go b/test/benchmarks/tools/ab.go
new file mode 100644
index 000000000..4cc9c3bce
--- /dev/null
+++ b/test/benchmarks/tools/ab.go
@@ -0,0 +1,94 @@
+// Copyright 2020 The gVisor Authors.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+// http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package tools
+
+import (
+ "fmt"
+ "net"
+ "regexp"
+ "strconv"
+ "testing"
+)
+
+// ApacheBench is for the client application ApacheBench.
+type ApacheBench struct {
+ Requests int
+ Concurrency int
+ Doc string
+ // TODO(zkoopmans): support KeepAlive and pass option to enable.
+}
+
+// MakeCmd makes an ApacheBench command.
+func (a *ApacheBench) MakeCmd(ip net.IP, port int) []string {
+ path := fmt.Sprintf("http://%s:%d/%s", ip, port, a.Doc)
+ // See apachebench (ab) for flags.
+ cmd := fmt.Sprintf("ab -n %d -c %d %s", a.Requests, a.Concurrency, path)
+ return []string{"sh", "-c", cmd}
+}
+
+// Report parses and reports metrics from ApacheBench output.
+func (a *ApacheBench) Report(b *testing.B, output string) {
+ // Parse and report custom metrics.
+ transferRate, err := a.parseTransferRate(output)
+ if err != nil {
+ b.Logf("failed to parse transferrate: %v", err)
+ }
+ b.ReportMetric(transferRate*1024, "transfer_rate_b/s") // Convert from Kb/s to b/s.
+
+ latency, err := a.parseLatency(output)
+ if err != nil {
+ b.Logf("failed to parse latency: %v", err)
+ }
+ b.ReportMetric(latency/1000, "mean_latency_secs") // Convert from ms to s.
+
+ reqPerSecond, err := a.parseRequestsPerSecond(output)
+ if err != nil {
+ b.Logf("failed to parse requests per second: %v", err)
+ }
+ b.ReportMetric(reqPerSecond, "requests_per_second")
+}
+
+var transferRateRE = regexp.MustCompile(`Transfer rate:\s+(\d+\.?\d+?)\s+\[Kbytes/sec\]\s+received`)
+
+// parseTransferRate parses transfer rate from ApacheBench output.
+func (a *ApacheBench) parseTransferRate(data string) (float64, error) {
+ match := transferRateRE.FindStringSubmatch(data)
+ if len(match) < 2 {
+ return 0, fmt.Errorf("failed get bandwidth: %s", data)
+ }
+ return strconv.ParseFloat(match[1], 64)
+}
+
+var latencyRE = regexp.MustCompile(`Total:\s+\d+\s+(\d+)\s+(\d+\.?\d+?)\s+\d+\s+\d+\s`)
+
+// parseLatency parses latency from ApacheBench output.
+func (a *ApacheBench) parseLatency(data string) (float64, error) {
+ match := latencyRE.FindStringSubmatch(data)
+ if len(match) < 2 {
+ return 0, fmt.Errorf("failed get bandwidth: %s", data)
+ }
+ return strconv.ParseFloat(match[1], 64)
+}
+
+var requestsPerSecondRE = regexp.MustCompile(`Requests per second:\s+(\d+\.?\d+?)\s+`)
+
+// parseRequestsPerSecond parses requests per second from ApacheBench output.
+func (a *ApacheBench) parseRequestsPerSecond(data string) (float64, error) {
+ match := requestsPerSecondRE.FindStringSubmatch(data)
+ if len(match) < 2 {
+ return 0, fmt.Errorf("failed get bandwidth: %s", data)
+ }
+ return strconv.ParseFloat(match[1], 64)
+}
diff --git a/test/benchmarks/tools/ab_test.go b/test/benchmarks/tools/ab_test.go
new file mode 100644
index 000000000..28ee66ec1
--- /dev/null
+++ b/test/benchmarks/tools/ab_test.go
@@ -0,0 +1,90 @@
+// Copyright 2020 The gVisor Authors.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+// http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package tools
+
+import "testing"
+
+// TestApacheBench checks the ApacheBench parsers on sample output.
+func TestApacheBench(t *testing.T) {
+ // Sample output from apachebench.
+ sampleData := `This is ApacheBench, Version 2.3 <$Revision: 1826891 $>
+Copyright 1996 Adam Twiss, Zeus Technology Ltd, http://www.zeustech.net/
+Licensed to The Apache Software Foundation, http://www.apache.org/
+
+Benchmarking 10.10.10.10 (be patient).....done
+
+
+Server Software: Apache/2.4.38
+Server Hostname: 10.10.10.10
+Server Port: 80
+
+Document Path: /latin10k.txt
+Document Length: 210 bytes
+
+Concurrency Level: 1
+Time taken for tests: 0.180 seconds
+Complete requests: 100
+Failed requests: 0
+Non-2xx responses: 100
+Total transferred: 38800 bytes
+HTML transferred: 21000 bytes
+Requests per second: 556.44 [#/sec] (mean)
+Time per request: 1.797 [ms] (mean)
+Time per request: 1.797 [ms] (mean, across all concurrent requests)
+Transfer rate: 210.84 [Kbytes/sec] received
+
+Connection Times (ms)
+ min mean[+/-sd] median max
+Connect: 0 0 0.2 0 2
+Processing: 1 2 1.0 1 8
+Waiting: 1 1 1.0 1 7
+Total: 1 2 1.2 1 10
+
+Percentage of the requests served within a certain time (ms)
+ 50% 1
+ 66% 2
+ 75% 2
+ 80% 2
+ 90% 2
+ 95% 3
+ 98% 7
+ 99% 10
+ 100% 10 (longest request)`
+
+ ab := ApacheBench{}
+ want := 210.84
+ got, err := ab.parseTransferRate(sampleData)
+ if err != nil {
+ t.Fatalf("failed to parse transfer rate with error: %v", err)
+ } else if got != want {
+ t.Fatalf("parseTransferRate got: %f, want: %f", got, want)
+ }
+
+ want = 2.0
+ got, err = ab.parseLatency(sampleData)
+ if err != nil {
+ t.Fatalf("failed to parse transfer rate with error: %v", err)
+ } else if got != want {
+ t.Fatalf("parseLatency got: %f, want: %f", got, want)
+ }
+
+ want = 556.44
+ got, err = ab.parseRequestsPerSecond(sampleData)
+ if err != nil {
+ t.Fatalf("failed to parse transfer rate with error: %v", err)
+ } else if got != want {
+ t.Fatalf("parseRequestsPerSecond got: %f, want: %f", got, want)
+ }
+}
diff --git a/test/benchmarks/tools/fio.go b/test/benchmarks/tools/fio.go
new file mode 100644
index 000000000..20000db16
--- /dev/null
+++ b/test/benchmarks/tools/fio.go
@@ -0,0 +1,124 @@
+// Copyright 2020 The gVisor Authors.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+// http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package tools
+
+import (
+ "encoding/json"
+ "fmt"
+ "strconv"
+ "strings"
+ "testing"
+)
+
+// Fio makes 'fio' commands and parses their output.
+type Fio struct {
+ Test string // test to run: read, write, randread, randwrite.
+ Size string // total size to be read/written of format N[GMK] (e.g. 5G).
+ Blocksize string // blocksize to be read/write of format N[GMK] (e.g. 4K).
+ Iodepth int // iodepth for reads/writes.
+ Time int // time to run the test in seconds, usually for rand(read/write).
+}
+
+// MakeCmd makes a 'fio' command.
+func (f *Fio) MakeCmd(filename string) []string {
+ cmd := []string{"fio", "--output-format=json", "--ioengine=sync"}
+ cmd = append(cmd, fmt.Sprintf("--name=%s", f.Test))
+ cmd = append(cmd, fmt.Sprintf("--size=%s", f.Size))
+ cmd = append(cmd, fmt.Sprintf("--blocksize=%s", f.Blocksize))
+ cmd = append(cmd, fmt.Sprintf("--filename=%s", filename))
+ cmd = append(cmd, fmt.Sprintf("--iodepth=%d", f.Iodepth))
+ cmd = append(cmd, fmt.Sprintf("--rw=%s", f.Test))
+ if f.Time != 0 {
+ cmd = append(cmd, "--time_based")
+ cmd = append(cmd, fmt.Sprintf("--runtime=%d", f.Time))
+ }
+ return cmd
+}
+
+// Report reports metrics based on output from an 'fio' command.
+func (f *Fio) Report(b *testing.B, output string) {
+ b.Helper()
+ // Parse the output and report the metrics.
+ isRead := strings.Contains(f.Test, "read")
+ bw, err := f.parseBandwidth(output, isRead)
+ if err != nil {
+ b.Fatalf("failed to parse bandwidth from %s with: %v", output, err)
+ }
+ b.ReportMetric(bw, "bandwidth_b/s") // in b/s.
+
+ iops, err := f.parseIOps(output, isRead)
+ if err != nil {
+ b.Fatalf("failed to parse iops from %s with: %v", output, err)
+ }
+ b.ReportMetric(iops, "iops")
+}
+
+// parseBandwidth reports the bandwidth in b/s.
+func (f *Fio) parseBandwidth(data string, isRead bool) (float64, error) {
+ if isRead {
+ result, err := f.parseFioJSON(data, "read", "bw")
+ if err != nil {
+ return 0, err
+ }
+ return 1024 * result, nil
+ }
+ result, err := f.parseFioJSON(data, "write", "bw")
+ if err != nil {
+ return 0, err
+ }
+ return 1024 * result, nil
+}
+
+// parseIOps reports the write IO per second metric.
+func (f *Fio) parseIOps(data string, isRead bool) (float64, error) {
+ if isRead {
+ return f.parseFioJSON(data, "read", "iops")
+ }
+ return f.parseFioJSON(data, "write", "iops")
+}
+
+// fioResult is for parsing FioJSON.
+type fioResult struct {
+ Jobs []fioJob
+}
+
+// fioJob is for parsing FioJSON.
+type fioJob map[string]json.RawMessage
+
+// fioMetrics is for parsing FioJSON.
+type fioMetrics map[string]json.RawMessage
+
+// parseFioJSON parses data and grabs "op" (read or write) and "metric"
+// (bw or iops) from the JSON.
+func (f *Fio) parseFioJSON(data, op, metric string) (float64, error) {
+ var result fioResult
+ if err := json.Unmarshal([]byte(data), &result); err != nil {
+ return 0, fmt.Errorf("could not unmarshal data: %v", err)
+ }
+
+ if len(result.Jobs) < 1 {
+ return 0, fmt.Errorf("no jobs present to parse")
+ }
+
+ var metrics fioMetrics
+ if err := json.Unmarshal(result.Jobs[0][op], &metrics); err != nil {
+ return 0, fmt.Errorf("could not unmarshal jobs: %v", err)
+ }
+
+ if _, ok := metrics[metric]; !ok {
+ return 0, fmt.Errorf("no metric found for op: %s", op)
+ }
+ return strconv.ParseFloat(string(metrics[metric]), 64)
+}
diff --git a/test/benchmarks/tools/fio_test.go b/test/benchmarks/tools/fio_test.go
new file mode 100644
index 000000000..a98277150
--- /dev/null
+++ b/test/benchmarks/tools/fio_test.go
@@ -0,0 +1,122 @@
+// Copyright 2020 The gVisor Authors.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+// http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package tools
+
+import "testing"
+
+// TestFio checks the Fio parsers on sample output.
+func TestFio(t *testing.T) {
+ sampleData := `
+{
+ "fio version" : "fio-3.1",
+ "timestamp" : 1554837456,
+ "timestamp_ms" : 1554837456621,
+ "time" : "Tue Apr 9 19:17:36 2019",
+ "jobs" : [
+ {
+ "jobname" : "test",
+ "groupid" : 0,
+ "error" : 0,
+ "eta" : 2147483647,
+ "elapsed" : 1,
+ "job options" : {
+ "name" : "test",
+ "ioengine" : "sync",
+ "size" : "1073741824",
+ "filename" : "/disk/file.dat",
+ "iodepth" : "4",
+ "bs" : "4096",
+ "rw" : "write"
+ },
+ "read" : {
+ "io_bytes" : 0,
+ "io_kbytes" : 0,
+ "bw" : 123456,
+ "iops" : 1234.5678,
+ "runtime" : 0,
+ "total_ios" : 0,
+ "short_ios" : 0,
+ "bw_min" : 0,
+ "bw_max" : 0,
+ "bw_agg" : 0.000000,
+ "bw_mean" : 0.000000,
+ "bw_dev" : 0.000000,
+ "bw_samples" : 0,
+ "iops_min" : 0,
+ "iops_max" : 0,
+ "iops_mean" : 0.000000,
+ "iops_stddev" : 0.000000,
+ "iops_samples" : 0
+ },
+ "write" : {
+ "io_bytes" : 1073741824,
+ "io_kbytes" : 1048576,
+ "bw" : 1753471,
+ "iops" : 438367.892977,
+ "runtime" : 598,
+ "total_ios" : 262144,
+ "bw_min" : 1731120,
+ "bw_max" : 1731120,
+ "bw_agg" : 98.725328,
+ "bw_mean" : 1731120.000000,
+ "bw_dev" : 0.000000,
+ "bw_samples" : 1,
+ "iops_min" : 432780,
+ "iops_max" : 432780,
+ "iops_mean" : 432780.000000,
+ "iops_stddev" : 0.000000,
+ "iops_samples" : 1
+ }
+ }
+ ]
+}
+`
+ fio := Fio{}
+ // WriteBandwidth.
+ got, err := fio.parseBandwidth(sampleData, false)
+ var want float64 = 1753471.0 * 1024
+ if err != nil {
+ t.Fatalf("parse failed with err: %v", err)
+ } else if got != want {
+ t.Fatalf("got: %f, want: %f", got, want)
+ }
+
+ // ReadBandwidth.
+ got, err = fio.parseBandwidth(sampleData, true)
+ want = 123456 * 1024
+ if err != nil {
+ t.Fatalf("parse failed with err: %v", err)
+ } else if got != want {
+ t.Fatalf("got: %f, want: %f", got, want)
+ }
+
+ // WriteIOps.
+ got, err = fio.parseIOps(sampleData, false)
+ want = 438367.892977
+ if err != nil {
+ t.Fatalf("parse failed with err: %v", err)
+ } else if got != want {
+ t.Fatalf("got: %f, want: %f", got, want)
+ }
+
+ // ReadIOps.
+ got, err = fio.parseIOps(sampleData, true)
+ want = 1234.5678
+ if err != nil {
+ t.Fatalf("parse failed with err: %v", err)
+ } else if got != want {
+ t.Fatalf("got: %f, want: %f", got, want)
+ }
+}
diff --git a/test/benchmarks/tools/hey.go b/test/benchmarks/tools/hey.go
new file mode 100644
index 000000000..699497c64
--- /dev/null
+++ b/test/benchmarks/tools/hey.go
@@ -0,0 +1,75 @@
+// Copyright 2020 The gVisor Authors.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+// http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package tools
+
+import (
+ "fmt"
+ "net"
+ "regexp"
+ "strconv"
+ "strings"
+ "testing"
+)
+
+// Hey is for the client application 'hey'.
+type Hey struct {
+ Requests int
+ Concurrency int
+ Doc string
+}
+
+// MakeCmd returns a 'hey' command.
+func (h *Hey) MakeCmd(ip net.IP, port int) []string {
+ return strings.Split(fmt.Sprintf("hey -n %d -c %d http://%s:%d/%s",
+ h.Requests, h.Concurrency, ip, port, h.Doc), " ")
+}
+
+// Report parses output from 'hey' and reports metrics.
+func (h *Hey) Report(b *testing.B, output string) {
+ b.Helper()
+ requests, err := h.parseRequestsPerSecond(output)
+ if err != nil {
+ b.Fatalf("failed to parse requests per second: %v", err)
+ }
+ b.ReportMetric(requests, "requests_per_second")
+
+ ave, err := h.parseAverageLatency(output)
+ if err != nil {
+ b.Fatalf("failed to parse average latency: %v", err)
+ }
+ b.ReportMetric(ave, "average_latency_secs")
+}
+
+var heyReqPerSecondRE = regexp.MustCompile(`Requests/sec:\s*(\d+\.?\d+?)\s+`)
+
+// parseRequestsPerSecond finds requests per second from 'hey' output.
+func (h *Hey) parseRequestsPerSecond(data string) (float64, error) {
+ match := heyReqPerSecondRE.FindStringSubmatch(data)
+ if len(match) < 2 {
+ return 0, fmt.Errorf("failed get bandwidth: %s", data)
+ }
+ return strconv.ParseFloat(match[1], 64)
+}
+
+var heyAverageLatencyRE = regexp.MustCompile(`Average:\s*(\d+\.?\d+?)\s+secs`)
+
+// parseHeyAverageLatency finds Average Latency in seconds form 'hey' output.
+func (h *Hey) parseAverageLatency(data string) (float64, error) {
+ match := heyAverageLatencyRE.FindStringSubmatch(data)
+ if len(match) < 2 {
+ return 0, fmt.Errorf("failed get average latency match%d : %s", len(match), data)
+ }
+ return strconv.ParseFloat(match[1], 64)
+}
diff --git a/test/benchmarks/tools/hey_test.go b/test/benchmarks/tools/hey_test.go
new file mode 100644
index 000000000..e0cab1f52
--- /dev/null
+++ b/test/benchmarks/tools/hey_test.go
@@ -0,0 +1,81 @@
+// Copyright 2020 The gVisor Authors.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+// http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package tools
+
+import "testing"
+
+// TestHey checks the Hey parsers on sample output.
+func TestHey(t *testing.T) {
+ sampleData := `
+ Summary:
+ Total: 2.2391 secs
+ Slowest: 1.6292 secs
+ Fastest: 0.0066 secs
+ Average: 0.5351 secs
+ Requests/sec: 89.3202
+
+ Total data: 841200 bytes
+ Size/request: 4206 bytes
+
+ Response time histogram:
+ 0.007 [1] |
+ 0.169 [0] |
+ 0.331 [149] |â– â– â– â– â– â– â– â– â– â– â– â– â– â– â– â– â– â– â– â– â– â– â– â– â– â– â– â– â– â– â– â– â– â– â– â– â– â– â– â– 
+ 0.493 [0] |
+ 0.656 [0] |
+ 0.818 [0] |
+ 0.980 [0] |
+ 1.142 [0] |
+ 1.305 [0] |
+ 1.467 [49] |â– â– â– â– â– â– â– â– â– â– â– â– â– 
+ 1.629 [1] |
+
+
+ Latency distribution:
+ 10% in 0.2149 secs
+ 25% in 0.2449 secs
+ 50% in 0.2703 secs
+ 75% in 1.3315 secs
+ 90% in 1.4045 secs
+ 95% in 1.4232 secs
+ 99% in 1.4362 secs
+
+ Details (average, fastest, slowest):
+ DNS+dialup: 0.0002 secs, 0.0066 secs, 1.6292 secs
+ DNS-lookup: 0.0000 secs, 0.0000 secs, 0.0000 secs
+ req write: 0.0000 secs, 0.0000 secs, 0.0012 secs
+ resp wait: 0.5225 secs, 0.0064 secs, 1.4346 secs
+ resp read: 0.0122 secs, 0.0001 secs, 0.2006 secs
+
+ Status code distribution:
+ [200] 200 responses
+ `
+ hey := Hey{}
+ want := 89.3202
+ got, err := hey.parseRequestsPerSecond(sampleData)
+ if err != nil {
+ t.Fatalf("failed to parse request per second with: %v", err)
+ } else if got != want {
+ t.Fatalf("got: %f, want: %f", got, want)
+ }
+
+ want = 0.5351
+ got, err = hey.parseAverageLatency(sampleData)
+ if err != nil {
+ t.Fatalf("failed to parse average latency with: %v", err)
+ } else if got != want {
+ t.Fatalf("got: %f, want: %f", got, want)
+ }
+}
diff --git a/test/benchmarks/tools/iperf.go b/test/benchmarks/tools/iperf.go
new file mode 100644
index 000000000..df3d9349b
--- /dev/null
+++ b/test/benchmarks/tools/iperf.go
@@ -0,0 +1,56 @@
+// Copyright 2020 The gVisor Authors.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+// http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package tools
+
+import (
+ "fmt"
+ "net"
+ "regexp"
+ "strconv"
+ "strings"
+ "testing"
+)
+
+// Iperf is for the client side of `iperf`.
+type Iperf struct {
+ Time int
+}
+
+// MakeCmd returns a iperf client command.
+func (i *Iperf) MakeCmd(ip net.IP, port int) []string {
+ // iperf report in Kb realtime
+ return strings.Split(fmt.Sprintf("iperf -f K --realtime --time %d -c %s -p %d", i.Time, ip, port), " ")
+}
+
+// Report parses output from iperf client and reports metrics.
+func (i *Iperf) Report(b *testing.B, output string) {
+ b.Helper()
+ // Parse bandwidth and report it.
+ bW, err := i.bandwidth(output)
+ if err != nil {
+ b.Fatalf("failed to parse bandwitdth from %s: %v", output, err)
+ }
+ b.ReportMetric(bW*1024, "bandwidth_b/s") // Convert from Kb/s to b/s.
+}
+
+// bandwidth parses the Bandwidth number from an iperf report. A sample is below.
+func (i *Iperf) bandwidth(data string) (float64, error) {
+ re := regexp.MustCompile(`\[\s*\d+\][^\n]+\s+(\d+\.?\d*)\s+KBytes/sec`)
+ match := re.FindStringSubmatch(data)
+ if len(match) < 1 {
+ return 0, fmt.Errorf("failed get bandwidth: %s", data)
+ }
+ return strconv.ParseFloat(match[1], 64)
+}
diff --git a/test/benchmarks/tools/iperf_test.go b/test/benchmarks/tools/iperf_test.go
new file mode 100644
index 000000000..03bb30d05
--- /dev/null
+++ b/test/benchmarks/tools/iperf_test.go
@@ -0,0 +1,34 @@
+// Copyright 2020 The gVisor Authors.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+// http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+package tools
+
+import "testing"
+
+// TestIperf checks the Iperf parsers on sample output.
+func TestIperf(t *testing.T) {
+ sampleData := `
+------------------------------------------------------------
+Client connecting to 10.138.15.215, TCP port 32779
+TCP window size: 45.0 KByte (default)
+------------------------------------------------------------
+[ 3] local 10.138.15.216 port 32866 connected with 10.138.15.215 port 32779
+[ ID] Interval Transfer Bandwidth
+[ 3] 0.0-10.0 sec 459520 KBytes 45900 KBytes/sec
+`
+ i := Iperf{}
+ bandwidth, err := i.bandwidth(sampleData)
+ if err != nil || bandwidth != 45900 {
+ t.Fatalf("failed with: %v and %f", err, bandwidth)
+ }
+}
diff --git a/test/benchmarks/tools/redis.go b/test/benchmarks/tools/redis.go
new file mode 100644
index 000000000..db32460ec
--- /dev/null
+++ b/test/benchmarks/tools/redis.go
@@ -0,0 +1,64 @@
+// Copyright 2020 The gVisor Authors.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+// http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package tools
+
+import (
+ "fmt"
+ "net"
+ "regexp"
+ "strconv"
+ "strings"
+ "testing"
+)
+
+// Redis is for the client 'redis-benchmark'.
+type Redis struct {
+ Operation string
+}
+
+// MakeCmd returns a redis-benchmark client command.
+func (r *Redis) MakeCmd(ip net.IP, port int) []string {
+ // There is no -t PING_BULK for redis-benchmark, so adjust the command in that case.
+ // Note that "ping" will run both PING_INLINE and PING_BULK.
+ if r.Operation == "PING_BULK" {
+ return strings.Split(
+ fmt.Sprintf("redis-benchmark --csv -t ping -h %s -p %d", ip, port), " ")
+ }
+
+ // runs redis-benchmark -t operation for 100K requests against server.
+ return strings.Split(
+ fmt.Sprintf("redis-benchmark --csv -t %s -h %s -p %d", r.Operation, ip, port), " ")
+}
+
+// Report parses output from redis-benchmark client and reports metrics.
+func (r *Redis) Report(b *testing.B, output string) {
+ b.Helper()
+ result, err := r.parseOperation(output)
+ if err != nil {
+ b.Fatalf("parsing result %s failed with err: %v", output, err)
+ }
+ b.ReportMetric(result, r.Operation) // operations per second
+}
+
+// parseOperation grabs the metric operations per second from redis-benchmark output.
+func (r *Redis) parseOperation(data string) (float64, error) {
+ re := regexp.MustCompile(fmt.Sprintf(`"%s( .*)?","(\d*\.\d*)"`, r.Operation))
+ match := re.FindStringSubmatch(data)
+ // If no match, simply don't add it to the result map.
+ if len(match) < 3 {
+ return 0.0, fmt.Errorf("could not find %s in %s", r.Operation, data)
+ }
+ return strconv.ParseFloat(match[2], 64)
+}
diff --git a/test/benchmarks/tools/redis_test.go b/test/benchmarks/tools/redis_test.go
new file mode 100644
index 000000000..4bafda66f
--- /dev/null
+++ b/test/benchmarks/tools/redis_test.go
@@ -0,0 +1,87 @@
+// Copyright 2020 The gVisor Authors.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+// http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package tools
+
+import (
+ "testing"
+)
+
+// TestRedis checks the Redis parsers on sample output.
+func TestRedis(t *testing.T) {
+ sampleData := `
+ "PING_INLINE","48661.80"
+ "PING_BULK","50301.81"
+ "SET","48923.68"
+ "GET","49382.71"
+ "INCR","49975.02"
+ "LPUSH","49875.31"
+ "RPUSH","50276.52"
+ "LPOP","50327.12"
+ "RPOP","50556.12"
+ "SADD","49504.95"
+ "HSET","49504.95"
+ "SPOP","50025.02"
+ "LPUSH (needed to benchmark LRANGE)","48875.86"
+ "LRANGE_100 (first 100 elements)","33955.86"
+ "LRANGE_300 (first 300 elements)","16550.81"// Copyright 2020 The gVisor Authors.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+// http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package tools
+
+ "LRANGE_500 (first 450 elements)","13653.74"
+ "LRANGE_600 (first 600 elements)","11219.57"
+ "MSET (10 keys)","44682.75"
+ `
+ wants := map[string]float64{
+ "PING_INLINE": 48661.80,
+ "PING_BULK": 50301.81,
+ "SET": 48923.68,
+ "GET": 49382.71,
+ "INCR": 49975.02,
+ "LPUSH": 49875.31,
+ "RPUSH": 50276.52,
+ "LPOP": 50327.12,
+ "RPOP": 50556.12,
+ "SADD": 49504.95,
+ "HSET": 49504.95,
+ "SPOP": 50025.02,
+ "LRANGE_100": 33955.86,
+ "LRANGE_300": 16550.81,
+ "LRANGE_500": 13653.74,
+ "LRANGE_600": 11219.57,
+ "MSET": 44682.75,
+ }
+ for op, want := range wants {
+ redis := Redis{
+ Operation: op,
+ }
+ if got, err := redis.parseOperation(sampleData); err != nil {
+ t.Fatalf("failed to parse %s: %v", op, err)
+ } else if want != got {
+ t.Fatalf("wanted %f for op %s, got %f", want, op, got)
+ }
+ }
+}
diff --git a/test/benchmarks/tools/tools.go b/test/benchmarks/tools/tools.go
new file mode 100644
index 000000000..eb61c0136
--- /dev/null
+++ b/test/benchmarks/tools/tools.go
@@ -0,0 +1,17 @@
+// Copyright 2020 The gVisor Authors.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+// http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+// Package tools holds tooling to couple command formatting and output parsers
+// together.
+package tools
diff --git a/test/e2e/integration_test.go b/test/e2e/integration_test.go
index ef42b689a..6fe6d304f 100644
--- a/test/e2e/integration_test.go
+++ b/test/e2e/integration_test.go
@@ -431,11 +431,39 @@ func TestHostOverlayfsCopyUp(t *testing.T) {
d := dockerutil.MakeContainer(ctx, t)
defer d.CleanUp(ctx)
- if _, err := d.Run(ctx, dockerutil.RunOpts{
+ if got, err := d.Run(ctx, dockerutil.RunOpts{
Image: "basic/hostoverlaytest",
WorkDir: "/root",
- }, "./test"); err != nil {
+ }, "./test_copy_up"); err != nil {
t.Fatalf("docker run failed: %v", err)
+ } else if got != "" {
+ t.Errorf("test failed:\n%s", got)
+ }
+}
+
+// TestHostOverlayfsRewindDir tests that rewinddir() "causes the directory
+// stream to refer to the current state of the corresponding directory, as a
+// call to opendir() would have done" as required by POSIX, when the directory
+// in question is host overlayfs.
+//
+// This test specifically targets host overlayfs because, per POSIX, "if a file
+// is removed from or added to the directory after the most recent call to
+// opendir() or rewinddir(), whether a subsequent call to readdir() returns an
+// entry for that file is unspecified"; the host filesystems used by other
+// automated tests yield newly-added files from readdir() even if the fsgofer
+// does not explicitly rewinddir(), but overlayfs does not.
+func TestHostOverlayfsRewindDir(t *testing.T) {
+ ctx := context.Background()
+ d := dockerutil.MakeContainer(ctx, t)
+ defer d.CleanUp(ctx)
+
+ if got, err := d.Run(ctx, dockerutil.RunOpts{
+ Image: "basic/hostoverlaytest",
+ WorkDir: "/root",
+ }, "./test_rewinddir"); err != nil {
+ t.Fatalf("docker run failed: %v", err)
+ } else if got != "" {
+ t.Errorf("test failed:\n%s", got)
}
}
diff --git a/test/iptables/BUILD b/test/iptables/BUILD
index 40b63ebbe..66453772a 100644
--- a/test/iptables/BUILD
+++ b/test/iptables/BUILD
@@ -9,6 +9,7 @@ go_library(
"filter_input.go",
"filter_output.go",
"iptables.go",
+ "iptables_unsafe.go",
"iptables_util.go",
"nat.go",
],
diff --git a/test/iptables/iptables_test.go b/test/iptables/iptables_test.go
index 550b6198a..fda5f694f 100644
--- a/test/iptables/iptables_test.go
+++ b/test/iptables/iptables_test.go
@@ -371,3 +371,11 @@ func TestFilterAddrs(t *testing.T) {
}
}
}
+
+func TestNATPreOriginalDst(t *testing.T) {
+ singleTest(t, NATPreOriginalDst{})
+}
+
+func TestNATOutOriginalDst(t *testing.T) {
+ singleTest(t, NATOutOriginalDst{})
+}
diff --git a/test/iptables/iptables_unsafe.go b/test/iptables/iptables_unsafe.go
new file mode 100644
index 000000000..bd85a8fea
--- /dev/null
+++ b/test/iptables/iptables_unsafe.go
@@ -0,0 +1,63 @@
+// Copyright 2020 The gVisor Authors.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+// http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package iptables
+
+import (
+ "fmt"
+ "syscall"
+ "unsafe"
+)
+
+type originalDstError struct {
+ errno syscall.Errno
+}
+
+func (e originalDstError) Error() string {
+ return fmt.Sprintf("errno (%d) when calling getsockopt(SO_ORIGINAL_DST): %v", int(e.errno), e.errno.Error())
+}
+
+// SO_ORIGINAL_DST gets the original destination of a redirected packet via
+// getsockopt.
+const SO_ORIGINAL_DST = 80
+
+func originalDestination4(connfd int) (syscall.RawSockaddrInet4, error) {
+ var addr syscall.RawSockaddrInet4
+ var addrLen uint32 = syscall.SizeofSockaddrInet4
+ if errno := originalDestination(connfd, syscall.SOL_IP, unsafe.Pointer(&addr), &addrLen); errno != 0 {
+ return syscall.RawSockaddrInet4{}, originalDstError{errno}
+ }
+ return addr, nil
+}
+
+func originalDestination6(connfd int) (syscall.RawSockaddrInet6, error) {
+ var addr syscall.RawSockaddrInet6
+ var addrLen uint32 = syscall.SizeofSockaddrInet6
+ if errno := originalDestination(connfd, syscall.SOL_IPV6, unsafe.Pointer(&addr), &addrLen); errno != 0 {
+ return syscall.RawSockaddrInet6{}, originalDstError{errno}
+ }
+ return addr, nil
+}
+
+func originalDestination(connfd int, level uintptr, optval unsafe.Pointer, optlen *uint32) syscall.Errno {
+ _, _, errno := syscall.Syscall6(
+ syscall.SYS_GETSOCKOPT,
+ uintptr(connfd),
+ level,
+ SO_ORIGINAL_DST,
+ uintptr(optval),
+ uintptr(unsafe.Pointer(optlen)),
+ 0)
+ return errno
+}
diff --git a/test/iptables/iptables_util.go b/test/iptables/iptables_util.go
index ca80a4b5f..5125fe47b 100644
--- a/test/iptables/iptables_util.go
+++ b/test/iptables/iptables_util.go
@@ -15,6 +15,8 @@
package iptables
import (
+ "encoding/binary"
+ "errors"
"fmt"
"net"
"os/exec"
@@ -218,17 +220,58 @@ func filterAddrs(addrs []string, ipv6 bool) []string {
// getInterfaceName returns the name of the interface other than loopback.
func getInterfaceName() (string, bool) {
- var ifname string
+ iface, ok := getNonLoopbackInterface()
+ if !ok {
+ return "", false
+ }
+ return iface.Name, true
+}
+
+func getInterfaceAddrs(ipv6 bool) ([]net.IP, error) {
+ iface, ok := getNonLoopbackInterface()
+ if !ok {
+ return nil, errors.New("no non-loopback interface found")
+ }
+ addrs, err := iface.Addrs()
+ if err != nil {
+ return nil, err
+ }
+
+ // Get only IPv4 or IPv6 addresses.
+ ips := make([]net.IP, 0, len(addrs))
+ for _, addr := range addrs {
+ parts := strings.Split(addr.String(), "/")
+ var ip net.IP
+ // To16() returns IPv4 addresses as IPv4-mapped IPv6 addresses.
+ // So we check whether To4() returns nil to test whether the
+ // address is v4 or v6.
+ if v4 := net.ParseIP(parts[0]).To4(); ipv6 && v4 == nil {
+ ip = net.ParseIP(parts[0]).To16()
+ } else {
+ ip = v4
+ }
+ if ip != nil {
+ ips = append(ips, ip)
+ }
+ }
+ return ips, nil
+}
+
+func getNonLoopbackInterface() (net.Interface, bool) {
if interfaces, err := net.Interfaces(); err == nil {
for _, intf := range interfaces {
if intf.Name != "lo" {
- ifname = intf.Name
- break
+ return intf, true
}
}
}
+ return net.Interface{}, false
+}
- return ifname, ifname != ""
+func htons(x uint16) uint16 {
+ buf := make([]byte, 2)
+ binary.BigEndian.PutUint16(buf, x)
+ return binary.LittleEndian.Uint16(buf)
}
func localIP(ipv6 bool) string {
diff --git a/test/iptables/nat.go b/test/iptables/nat.go
index ac0d91bb2..b7fea2527 100644
--- a/test/iptables/nat.go
+++ b/test/iptables/nat.go
@@ -18,12 +18,11 @@ import (
"errors"
"fmt"
"net"
+ "syscall"
"time"
)
-const (
- redirectPort = 42
-)
+const redirectPort = 42
func init() {
RegisterTestCase(NATPreRedirectUDPPort{})
@@ -42,6 +41,8 @@ func init() {
RegisterTestCase(NATOutRedirectInvert{})
RegisterTestCase(NATRedirectRequiresProtocol{})
RegisterTestCase(NATLoopbackSkipsPrerouting{})
+ RegisterTestCase(NATPreOriginalDst{})
+ RegisterTestCase(NATOutOriginalDst{})
}
// NATPreRedirectUDPPort tests that packets are redirected to different port.
@@ -471,6 +472,151 @@ func (NATLoopbackSkipsPrerouting) LocalAction(ip net.IP, ipv6 bool) error {
return nil
}
+// NATPreOriginalDst tests that SO_ORIGINAL_DST returns the pre-NAT destination
+// of PREROUTING NATted packets.
+type NATPreOriginalDst struct{}
+
+// Name implements TestCase.Name.
+func (NATPreOriginalDst) Name() string {
+ return "NATPreOriginalDst"
+}
+
+// ContainerAction implements TestCase.ContainerAction.
+func (NATPreOriginalDst) ContainerAction(ip net.IP, ipv6 bool) error {
+ // Redirect incoming TCP connections to acceptPort.
+ if err := natTable(ipv6, "-A", "PREROUTING",
+ "-p", "tcp",
+ "--destination-port", fmt.Sprintf("%d", dropPort),
+ "-j", "REDIRECT", "--to-port", fmt.Sprintf("%d", acceptPort)); err != nil {
+ return err
+ }
+
+ addrs, err := getInterfaceAddrs(ipv6)
+ if err != nil {
+ return err
+ }
+ return listenForRedirectedConn(ipv6, addrs)
+}
+
+// LocalAction implements TestCase.LocalAction.
+func (NATPreOriginalDst) LocalAction(ip net.IP, ipv6 bool) error {
+ return connectTCP(ip, dropPort, sendloopDuration)
+}
+
+// NATOutOriginalDst tests that SO_ORIGINAL_DST returns the pre-NAT destination
+// of OUTBOUND NATted packets.
+type NATOutOriginalDst struct{}
+
+// Name implements TestCase.Name.
+func (NATOutOriginalDst) Name() string {
+ return "NATOutOriginalDst"
+}
+
+// ContainerAction implements TestCase.ContainerAction.
+func (NATOutOriginalDst) ContainerAction(ip net.IP, ipv6 bool) error {
+ // Redirect incoming TCP connections to acceptPort.
+ if err := natTable(ipv6, "-A", "OUTPUT", "-p", "tcp", "-j", "REDIRECT", "--to-port", fmt.Sprintf("%d", acceptPort)); err != nil {
+ return err
+ }
+
+ connCh := make(chan error)
+ go func() {
+ connCh <- connectTCP(ip, dropPort, sendloopDuration)
+ }()
+
+ if err := listenForRedirectedConn(ipv6, []net.IP{ip}); err != nil {
+ return err
+ }
+ return <-connCh
+}
+
+// LocalAction implements TestCase.LocalAction.
+func (NATOutOriginalDst) LocalAction(ip net.IP, ipv6 bool) error {
+ // No-op.
+ return nil
+}
+
+func listenForRedirectedConn(ipv6 bool, originalDsts []net.IP) error {
+ // The net package doesn't give guarantee access to the connection's
+ // underlying FD, and thus we cannot call getsockopt. We have to use
+ // traditional syscalls for SO_ORIGINAL_DST.
+
+ // Create the listening socket, bind, listen, and accept.
+ family := syscall.AF_INET
+ if ipv6 {
+ family = syscall.AF_INET6
+ }
+ sockfd, err := syscall.Socket(family, syscall.SOCK_STREAM, 0)
+ if err != nil {
+ return err
+ }
+ defer syscall.Close(sockfd)
+
+ var bindAddr syscall.Sockaddr
+ if ipv6 {
+ bindAddr = &syscall.SockaddrInet6{
+ Port: acceptPort,
+ Addr: [16]byte{0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}, // in6addr_any
+ }
+ } else {
+ bindAddr = &syscall.SockaddrInet4{
+ Port: acceptPort,
+ Addr: [4]byte{0, 0, 0, 0}, // INADDR_ANY
+ }
+ }
+ if err := syscall.Bind(sockfd, bindAddr); err != nil {
+ return err
+ }
+
+ if err := syscall.Listen(sockfd, 1); err != nil {
+ return err
+ }
+
+ connfd, _, err := syscall.Accept(sockfd)
+ if err != nil {
+ return err
+ }
+ defer syscall.Close(connfd)
+
+ // Verify that, despite listening on acceptPort, SO_ORIGINAL_DST
+ // indicates the packet was sent to originalDst:dropPort.
+ if ipv6 {
+ got, err := originalDestination6(connfd)
+ if err != nil {
+ return err
+ }
+ // The original destination could be any of our IPs.
+ for _, dst := range originalDsts {
+ want := syscall.RawSockaddrInet6{
+ Family: syscall.AF_INET6,
+ Port: htons(dropPort),
+ }
+ copy(want.Addr[:], dst.To16())
+ if got == want {
+ return nil
+ }
+ }
+ return fmt.Errorf("SO_ORIGINAL_DST returned %+v, but wanted one of %+v (note: port numbers are in network byte order)", got, originalDsts)
+ } else {
+ got, err := originalDestination4(connfd)
+ if err != nil {
+ return err
+ }
+ // The original destination could be any of our IPs.
+ for _, dst := range originalDsts {
+ want := syscall.RawSockaddrInet4{
+ Family: syscall.AF_INET,
+ Port: htons(dropPort),
+ }
+ copy(want.Addr[:], dst.To4())
+ if got == want {
+ return nil
+ }
+ }
+ return fmt.Errorf("SO_ORIGINAL_DST returned %+v, but wanted one of %+v (note: port numbers are in network byte order)", got, originalDsts)
+ }
+}
+
// loopbackTests runs an iptables rule and ensures that packets sent to
// dest:dropPort are received by localhost:acceptPort.
func loopbackTest(ipv6 bool, dest net.IP, args ...string) error {
diff --git a/test/runtimes/BUILD b/test/runtimes/BUILD
index 1728161ce..3be123d94 100644
--- a/test/runtimes/BUILD
+++ b/test/runtimes/BUILD
@@ -11,6 +11,7 @@ runtime_test(
runtime_test(
name = "java11",
+ batch = 100,
exclude_file = "exclude_java11.csv",
lang = "java",
shard_count = 20,
diff --git a/test/runtimes/defs.bzl b/test/runtimes/defs.bzl
index 5779b9591..db22029a8 100644
--- a/test/runtimes/defs.bzl
+++ b/test/runtimes/defs.bzl
@@ -9,6 +9,8 @@ def _runtime_test_impl(ctx):
ctx.attr.lang,
"--image",
ctx.attr.image,
+ "--batch",
+ str(ctx.attr.batch),
]
if ctx.attr.exclude_file:
args += [
@@ -47,6 +49,10 @@ _runtime_test = rule(
mandatory = False,
allow_single_file = True,
),
+ "batch": attr.int(
+ default = 50,
+ mandatory = False,
+ ),
"_runner": attr.label(
default = "//test/runtimes/runner:runner",
),
diff --git a/test/runtimes/exclude_go1.12.csv b/test/runtimes/exclude_go1.12.csv
index 8c8ae0c5d..81e02cf64 100644
--- a/test/runtimes/exclude_go1.12.csv
+++ b/test/runtimes/exclude_go1.12.csv
@@ -2,15 +2,12 @@ test name,bug id,comment
cgo_errors,,FLAKY
cgo_test,,FLAKY
go_test:cmd/go,,FLAKY
-go_test:cmd/vendor/golang.org/x/sys/unix,b/118783622,/dev devices missing
-go_test:net,b/118784196,socket: invalid argument. Works as intended: see bug.
+go_test:net,b/162473575,setsockopt: protocol not available.
go_test:os,b/118780122,we have a pollable filesystem but that's a surprise
-go_test:os/signal,b/118780860,/dev/pts not properly supported
-go_test:runtime,b/118782341,sigtrap not reported or caught or something
-go_test:syscall,b/118781998,bad bytes -- bad mem addr
-race,b/118782931,thread sanitizer. Works as intended: b/62219744.
+go_test:os/signal,b/118780860,/dev/pts not properly supported. Also being tracked in b/29356795.
+go_test:runtime,b/118782341,sigtrap not reported or caught or something. Also being tracked in b/33003106.
+go_test:syscall,b/118781998,bad bytes -- bad mem addr; FcntlFlock(F_GETLK) not supported.
runtime:cpu124,b/118778254,segmentation fault
test:0_1,,FLAKY
-testasan,,
testcarchive,b/118782924,no sigpipe
testshared,,FLAKY
diff --git a/test/runtimes/exclude_java11.csv b/test/runtimes/exclude_java11.csv
index c012e5a56..4d62f7d3a 100644
--- a/test/runtimes/exclude_java11.csv
+++ b/test/runtimes/exclude_java11.csv
@@ -15,10 +15,9 @@ java/lang/Character/CheckScript.java,,Fails in Docker
java/lang/Character/CheckUnicode.java,,Fails in Docker
java/lang/Class/GetPackageBootLoaderChildLayer.java,,
java/lang/ClassLoader/nativeLibrary/NativeLibraryTest.java,,Fails in Docker
+java/lang/module/ModuleDescriptorTest.java,,
java/lang/String/nativeEncoding/StringPlatformChars.java,,
-java/net/DatagramSocket/ReuseAddressTest.java,,
-java/net/DatagramSocket/SendDatagramToBadAddress.java,b/78473345,
-java/net/Inet4Address/PingThis.java,,
+java/net/CookieHandler/B6791927.java,,java.lang.RuntimeException: Expiration date shouldn't be 0
java/net/InterfaceAddress/NetworkPrefixLength.java,b/78507103,
java/net/MulticastSocket/MulticastTTL.java,,
java/net/MulticastSocket/Promiscuous.java,,
@@ -29,23 +28,16 @@ java/net/MulticastSocket/TestDefaults.java,,
java/net/MulticastSocket/TimeToLive.java,,
java/net/NetworkInterface/NetworkInterfaceStreamTest.java,,
java/net/Socket/SetSoLinger.java,b/78527327,SO_LINGER is not yet supported
-java/net/Socket/TrafficClass.java,b/78527818,Not supported on gVisor
java/net/Socket/UrgentDataTest.java,b/111515323,
-java/net/Socket/setReuseAddress/Basic.java,b/78519214,SO_REUSEADDR enabled by default
java/net/SocketOption/OptionsTest.java,,Fails in Docker
-java/net/SocketOption/TcpKeepAliveTest.java,,
java/net/SocketPermission/SocketPermissionTest.java,,
java/net/URLConnection/6212146/TestDriver.java,,Fails in Docker
java/net/httpclient/RequestBuilderTest.java,,Fails in Docker
-java/net/httpclient/ShortResponseBody.java,,
-java/net/httpclient/ShortResponseBodyWithRetry.java,,
-java/nio/channels/AsyncCloseAndInterrupt.java,,
-java/nio/channels/AsynchronousServerSocketChannel/Basic.java,,
-java/nio/channels/AsynchronousSocketChannel/Basic.java,b/77921528,SO_KEEPALIVE is not settable
java/nio/channels/DatagramChannel/BasicMulticastTests.java,,
-java/nio/channels/DatagramChannel/SocketOptionTests.java,,Fails in Docker
+java/nio/channels/DatagramChannel/SocketOptionTests.java,,java.net.SocketException: Invalid argument
java/nio/channels/DatagramChannel/UseDGWithIPv6.java,,
java/nio/channels/FileChannel/directio/DirectIOTest.java,,Fails in Docker
+java/nio/channels/FileChannel/directio/PwriteDirect.java,,java.io.IOException: Invalid argument
java/nio/channels/Selector/OutOfBand.java,,
java/nio/channels/Selector/SelectWithConsumer.java,,Flaky
java/nio/channels/ServerSocketChannel/SocketOptionTests.java,,
@@ -59,15 +51,89 @@ java/text/Format/NumberFormat/CurrencyFormat.java,,Fails in Docker
java/util/Calendar/JapaneseEraNameTest.java,,
java/util/Currency/CurrencyTest.java,,Fails in Docker
java/util/Currency/ValidateISO4217.java,,Fails in Docker
+java/util/EnumSet/BogusEnumSet.java,,"java.io.InvalidClassException: java.util.EnumSet; local class incompatible: stream classdesc serialVersionUID = -2409567991088730183, local class serialVersionUID = 1009687484059888093"
+java/util/Locale/Bug8040211.java,,java.lang.RuntimeException: Failed.
java/util/Locale/LSRDataTest.java,,
+java/util/Properties/CompatibilityTest.java,,"java.lang.RuntimeException: jdk.internal.org.xml.sax.SAXParseException; Internal DTD subset is not allowed. The Properties XML document must have the following DOCTYPE declaration: <!DOCTYPE properties SYSTEM ""http://java.sun.com/dtd/properties.dtd"">"
+java/util/ResourceBundle/Control/XMLResourceBundleTest.java,,java.util.MissingResourceException: Can't find bundle for base name XmlRB locale
+java/util/ResourceBundle/modules/xmlformat/xmlformat.sh,,Timeout reached: 60000. Process is not alive!
+java/util/TimeZone/TimeZoneTest.java,,Uncaught exception thrown in test method TestShortZoneIDs
java/util/concurrent/locks/Lock/TimedAcquireLeak.java,,
java/util/jar/JarFile/mrjar/MultiReleaseJarAPI.java,,Fails in Docker
java/util/logging/LogManager/Configuration/updateConfiguration/SimpleUpdateConfigWithInputStreamTest.java,,
java/util/logging/TestLoggerWeakRefLeak.java,,
+java/util/spi/ResourceBundleControlProvider/UserDefaultControlTest.java,,java.util.MissingResourceException: Can't find bundle for base name com.foo.XmlRB locale
javax/imageio/AppletResourceTest.java,,
+javax/imageio/plugins/jpeg/JPEGsNotAcceleratedTest.java,,java.awt.HeadlessException: No X11 DISPLAY variable was set but this program performed an operation which requires it.
javax/management/security/HashedPasswordFileTest.java,,
+javax/net/ssl/DTLS/DTLSBufferOverflowUnderflowTest.java,,Compilation failed
+javax/net/ssl/DTLS/DTLSDataExchangeTest.java,,Compilation failed
+javax/net/ssl/DTLS/DTLSEnginesClosureTest.java,,Compilation failed
+javax/net/ssl/DTLS/DTLSHandshakeTest.java,,Compilation failed
+javax/net/ssl/DTLS/DTLSHandshakeWithReplicatedPacketsTest.java,,Compilation failed
+javax/net/ssl/DTLS/DTLSIncorrectAppDataTest.java,,Compilation failed
+javax/net/ssl/DTLS/DTLSMFLNTest.java,,Compilation failed
+javax/net/ssl/DTLS/DTLSNotEnabledRC4Test.java,,Compilation failed
+javax/net/ssl/DTLS/DTLSRehandshakeTest.java,,Compilation failed
+javax/net/ssl/DTLS/DTLSRehandshakeWithCipherChangeTest.java,,Compilation failed
+javax/net/ssl/DTLS/DTLSRehandshakeWithDataExTest.java,,Compilation failed
+javax/net/ssl/DTLS/DTLSSequenceNumberTest.java,,Compilation failed
+javax/net/ssl/DTLS/DTLSUnsupportedCiphersTest.java,,Compilation failed
+javax/net/ssl/DTLSv10/DTLSv10BufferOverflowUnderflowTest.java,,Compilation failed
+javax/net/ssl/DTLSv10/DTLSv10DataExchangeTest.java,,Compilation failed
+javax/net/ssl/DTLSv10/DTLSv10EnginesClosureTest.java,,Compilation failed
+javax/net/ssl/DTLSv10/DTLSv10HandshakeTest.java,,Compilation failed
+javax/net/ssl/DTLSv10/DTLSv10HandshakeWithReplicatedPacketsTest.java,,Compilation failed
+javax/net/ssl/DTLSv10/DTLSv10IncorrectAppDataTest.java,,Compilation failed
+javax/net/ssl/DTLSv10/DTLSv10MFLNTest.java,,Compilation failed
+javax/net/ssl/DTLSv10/DTLSv10NotEnabledRC4Test.java,,Compilation failed
+javax/net/ssl/DTLSv10/DTLSv10RehandshakeTest.java,,Compilation failed
+javax/net/ssl/DTLSv10/DTLSv10RehandshakeWithCipherChangeTest.java,,Compilation failed
+javax/net/ssl/DTLSv10/DTLSv10RehandshakeWithDataExTest.java,,Compilation failed
+javax/net/ssl/DTLSv10/DTLSv10SequenceNumberTest.java,,Compilation failed
+javax/net/ssl/DTLSv10/DTLSv10UnsupportedCiphersTest.java,,Compilation failed
javax/net/ssl/SSLSession/JSSERenegotiate.java,,Fails in Docker
+javax/net/ssl/TLS/TLSDataExchangeTest.java,,Compilation failed
+javax/net/ssl/TLS/TLSEnginesClosureTest.java,,Compilation failed
+javax/net/ssl/TLS/TLSHandshakeTest.java,,Compilation failed
+javax/net/ssl/TLS/TLSMFLNTest.java,,Compilation failed
+javax/net/ssl/TLS/TLSNotEnabledRC4Test.java,,Compilation failed
+javax/net/ssl/TLS/TLSRehandshakeTest.java,,Compilation failed
+javax/net/ssl/TLS/TLSRehandshakeWithCipherChangeTest.java,,Compilation failed
+javax/net/ssl/TLS/TLSRehandshakeWithDataExTest.java,,Compilation failed
+javax/net/ssl/TLS/TLSUnsupportedCiphersTest.java,,Compilation failed
+javax/net/ssl/TLSv1/TLSDataExchangeTest.java,,Compilation failed
+javax/net/ssl/TLSv1/TLSEnginesClosureTest.java,,Compilation failed
+javax/net/ssl/TLSv1/TLSHandshakeTest.java,,Compilation failed
+javax/net/ssl/TLSv1/TLSMFLNTest.java,,Compilation failed
+javax/net/ssl/TLSv1/TLSNotEnabledRC4Test.java,,Compilation failed
+javax/net/ssl/TLSv1/TLSRehandshakeTest.java,,Compilation failed
+javax/net/ssl/TLSv1/TLSRehandshakeWithCipherChangeTest.java,,Compilation failed
+javax/net/ssl/TLSv1/TLSRehandshakeWithDataExTest.java,,Compilation failed
+javax/net/ssl/TLSv1/TLSUnsupportedCiphersTest.java,,Compilation failed
+javax/net/ssl/TLSv11/TLSDataExchangeTest.java,,Compilation failed
+javax/net/ssl/TLSv11/TLSEnginesClosureTest.java,,Compilation failed
+javax/net/ssl/TLSv11/TLSHandshakeTest.java,,Compilation failed
+javax/net/ssl/TLSv11/TLSMFLNTest.java,,Compilation failed
+javax/net/ssl/TLSv11/TLSNotEnabledRC4Test.java,,Compilation failed
+javax/net/ssl/TLSv11/TLSRehandshakeTest.java,,Compilation failed
+javax/net/ssl/TLSv11/TLSRehandshakeWithCipherChangeTest.java,,Compilation failed
+javax/net/ssl/TLSv11/TLSRehandshakeWithDataExTest.java,,Compilation failed
+javax/net/ssl/TLSv11/TLSUnsupportedCiphersTest.java,,Compilation failed
+javax/net/ssl/TLSv12/TLSEnginesClosureTest.java,,Compilation failed
javax/sound/sampled/AudioInputStream/FrameLengthAfterConversion.java,,
+jdk/jfr/cmd/TestHelp.java,,java.lang.RuntimeException: 'Available commands are:' missing from stdout/stderr
+jdk/jfr/cmd/TestPrint.java,,Missing file' missing from stdout/stderr
+jdk/jfr/cmd/TestPrintDefault.java,,java.lang.RuntimeException: 'JVMInformation' missing from stdout/stderr
+jdk/jfr/cmd/TestPrintJSON.java,,javax.script.ScriptException: <eval>:1:17 Expected an operand but found eof var jsonObject = ^ in <eval> at line number 1 at column number 17
+jdk/jfr/cmd/TestPrintXML.java,,org.xml.sax.SAXParseException; lineNumber: 1; columnNumber: 1; Premature end of file.
+jdk/jfr/cmd/TestReconstruct.java,,java.lang.RuntimeException: 'Too few arguments' missing from stdout/stderr
+jdk/jfr/cmd/TestSplit.java,,java.lang.RuntimeException: 'Missing file' missing from stdout/stderr
+jdk/jfr/cmd/TestSummary.java,,java.lang.RuntimeException: 'Missing file' missing from stdout/stderr
+jdk/jfr/event/compiler/TestCompilerStats.java,,java.lang.RuntimeException: Field nmetodsSize not in event
+jdk/jfr/event/metadata/TestDefaultConfigurations.java,,Setting 'threshold' in event 'jdk.SecurityPropertyModification' was not configured in the configuration 'default'
+jdk/jfr/event/runtime/TestActiveSettingEvent.java,,java.lang.Exception: Could not find setting with name jdk.X509Validation#threshold
+jdk/jfr/event/runtime/TestModuleEvents.java,,java.lang.RuntimeException: assertEquals: expected jdk.proxy1 to equal java.base
jdk/jfr/event/runtime/TestNetworkUtilizationEvent.java,,
jdk/jfr/event/runtime/TestThreadParkEvent.java,,
jdk/jfr/event/sampling/TestNative.java,,
@@ -86,6 +152,7 @@ jdk/jfr/jcmd/TestJcmdStartStopDefault.java,,
jdk/jfr/jcmd/TestJcmdStartWithOptions.java,,
jdk/jfr/jcmd/TestJcmdStartWithSettings.java,,
jdk/jfr/jcmd/TestJcmdStopInvalidFile.java,,
+jdk/jfr/jvm/TestGetAllEventClasses.java,,Compilation failed
jdk/jfr/jvm/TestJfrJavaBase.java,,
jdk/jfr/startupargs/TestStartRecording.java,,
jdk/modules/incubator/ImageModules.java,,
@@ -123,4 +190,6 @@ tools/jlink/JLinkTest.java,,
tools/jlink/plugins/IncludeLocalesPluginTest.java,,
tools/jmod/hashes/HashesTest.java,,
tools/launcher/BigJar.java,b/111611473,
+tools/launcher/HelpFlagsTest.java,,java.lang.AssertionError: HelpFlagsTest failed: Tool jfr not covered by this test. Add specification to jdkTools array!
+tools/launcher/VersionCheck.java,,java.lang.AssertionError: VersionCheck failed: testToolVersion: [jfr];
tools/launcher/modules/patch/systemmodules/PatchSystemModules.java,,
diff --git a/test/runtimes/proctor/go.go b/test/runtimes/proctor/go.go
index 073c2959d..d0ae844e6 100644
--- a/test/runtimes/proctor/go.go
+++ b/test/runtimes/proctor/go.go
@@ -81,13 +81,13 @@ func (goRunner) TestCmds(tests []string) []*exec.Cmd {
if strings.HasSuffix(test, ".go") {
onDiskTests = append(onDiskTests, test)
} else {
- toolTests = append(toolTests, test)
+ toolTests = append(toolTests, "^"+test+"$")
}
}
var cmds []*exec.Cmd
if len(toolTests) > 0 {
- cmds = append(cmds, exec.Command("go", "tool", "dist", "test", "-run", strings.Join(toolTests, "\\|")))
+ cmds = append(cmds, exec.Command("go", "tool", "dist", "test", "-v", "-no-rebuild", "-run", strings.Join(toolTests, "\\|")))
}
if len(onDiskTests) > 0 {
cmd := exec.Command("go", append([]string{"run", "run.go", "-v", "--"}, onDiskTests...)...)
diff --git a/test/runtimes/proctor/java.go b/test/runtimes/proctor/java.go
index 737fbe23e..d456fa681 100644
--- a/test/runtimes/proctor/java.go
+++ b/test/runtimes/proctor/java.go
@@ -64,8 +64,11 @@ func (javaRunner) ListTests() ([]string, error) {
func (javaRunner) TestCmds(tests []string) []*exec.Cmd {
args := append(
[]string{
- "-noreport",
- "-dir:" + javaTestDir,
+ "-agentvm", // Execute each action using a pool of reusable JVMs.
+ "-dir:" + javaTestDir, // Base directory for test files and directories.
+ "-noreport", // Do not generate a final report.
+ "-timeoutFactor:20", // Extend the default timeout (2 min) of all tests by this factor.
+ "-verbose:nopass", // Verbose output but supress it for tests that passed.
},
tests...,
)
diff --git a/test/runtimes/runner/main.go b/test/runtimes/runner/main.go
index e230912c9..948e7cf9c 100644
--- a/test/runtimes/runner/main.go
+++ b/test/runtimes/runner/main.go
@@ -40,7 +40,7 @@ var (
)
// Wait time for each test to run.
-const timeout = 45 * time.Minute
+const timeout = 90 * time.Minute
func main() {
flag.Parse()
diff --git a/test/syscalls/linux/epoll.cc b/test/syscalls/linux/epoll.cc
index f57d38dc7..2101e5c9f 100644
--- a/test/syscalls/linux/epoll.cc
+++ b/test/syscalls/linux/epoll.cc
@@ -422,6 +422,28 @@ TEST(EpollTest, CloseFile) {
SyscallSucceedsWithValue(0));
}
+TEST(EpollTest, PipeReaderHupAfterWriterClosed) {
+ auto epollfd = ASSERT_NO_ERRNO_AND_VALUE(NewEpollFD());
+ int pipefds[2];
+ ASSERT_THAT(pipe(pipefds), SyscallSucceeds());
+ FileDescriptor rfd(pipefds[0]);
+ FileDescriptor wfd(pipefds[1]);
+
+ ASSERT_NO_ERRNO(RegisterEpollFD(epollfd.get(), rfd.get(), 0, kMagicConstant));
+ struct epoll_event result[kFDsPerEpoll];
+ // Initially, rfd should not generate any events of interest.
+ ASSERT_THAT(epoll_wait(epollfd.get(), result, kFDsPerEpoll, 0),
+ SyscallSucceedsWithValue(0));
+ // Close the write end of the pipe.
+ wfd.reset();
+ // rfd should now generate EPOLLHUP, which EPOLL_CTL_ADD unconditionally adds
+ // to the set of events of interest.
+ ASSERT_THAT(epoll_wait(epollfd.get(), result, kFDsPerEpoll, 0),
+ SyscallSucceedsWithValue(1));
+ EXPECT_EQ(result[0].events, EPOLLHUP);
+ EXPECT_EQ(result[0].data.u64, kMagicConstant);
+}
+
} // namespace
} // namespace testing
diff --git a/test/syscalls/linux/fcntl.cc b/test/syscalls/linux/fcntl.cc
index 5467fa2c8..34016d4bd 100644
--- a/test/syscalls/linux/fcntl.cc
+++ b/test/syscalls/linux/fcntl.cc
@@ -1004,7 +1004,8 @@ TEST(FcntlTest, SetOwnPid) {
pid_t pid;
EXPECT_THAT(pid = getpid(), SyscallSucceeds());
- ASSERT_THAT(syscall(__NR_fcntl, s.get(), F_SETOWN, pid), SyscallSucceeds());
+ ASSERT_THAT(syscall(__NR_fcntl, s.get(), F_SETOWN, pid),
+ SyscallSucceedsWithValue(0));
EXPECT_THAT(syscall(__NR_fcntl, s.get(), F_GETOWN),
SyscallSucceedsWithValue(pid));
@@ -1018,7 +1019,8 @@ TEST(FcntlTest, SetOwnPgrp) {
pid_t pgid;
EXPECT_THAT(pgid = getpgrp(), SyscallSucceeds());
- ASSERT_THAT(syscall(__NR_fcntl, s.get(), F_SETOWN, -pgid), SyscallSucceeds());
+ ASSERT_THAT(syscall(__NR_fcntl, s.get(), F_SETOWN, -pgid),
+ SyscallSucceedsWithValue(0));
// Verify with F_GETOWN_EX; using F_GETOWN on Linux may incorrectly treat the
// negative return value as an error, converting the return value to -1 and
@@ -1038,8 +1040,10 @@ TEST(FcntlTest, SetOwnUnset) {
// Set and unset pid.
pid_t pid;
EXPECT_THAT(pid = getpid(), SyscallSucceeds());
- ASSERT_THAT(syscall(__NR_fcntl, s.get(), F_SETOWN, pid), SyscallSucceeds());
- ASSERT_THAT(syscall(__NR_fcntl, s.get(), F_SETOWN, 0), SyscallSucceeds());
+ ASSERT_THAT(syscall(__NR_fcntl, s.get(), F_SETOWN, pid),
+ SyscallSucceedsWithValue(0));
+ ASSERT_THAT(syscall(__NR_fcntl, s.get(), F_SETOWN, 0),
+ SyscallSucceedsWithValue(0));
EXPECT_THAT(syscall(__NR_fcntl, s.get(), F_GETOWN),
SyscallSucceedsWithValue(0));
@@ -1047,8 +1051,10 @@ TEST(FcntlTest, SetOwnUnset) {
// Set and unset pgid.
pid_t pgid;
EXPECT_THAT(pgid = getpgrp(), SyscallSucceeds());
- ASSERT_THAT(syscall(__NR_fcntl, s.get(), F_SETOWN, -pgid), SyscallSucceeds());
- ASSERT_THAT(syscall(__NR_fcntl, s.get(), F_SETOWN, 0), SyscallSucceeds());
+ ASSERT_THAT(syscall(__NR_fcntl, s.get(), F_SETOWN, -pgid),
+ SyscallSucceedsWithValue(0));
+ ASSERT_THAT(syscall(__NR_fcntl, s.get(), F_SETOWN, 0),
+ SyscallSucceedsWithValue(0));
EXPECT_THAT(syscall(__NR_fcntl, s.get(), F_GETOWN),
SyscallSucceedsWithValue(0));
@@ -1120,7 +1126,7 @@ TEST(FcntlTest, SetOwnExTid) {
EXPECT_THAT(owner.pid = syscall(__NR_gettid), SyscallSucceeds());
ASSERT_THAT(syscall(__NR_fcntl, s.get(), F_SETOWN_EX, &owner),
- SyscallSucceeds());
+ SyscallSucceedsWithValue(0));
EXPECT_THAT(syscall(__NR_fcntl, s.get(), F_GETOWN),
SyscallSucceedsWithValue(owner.pid));
@@ -1136,7 +1142,7 @@ TEST(FcntlTest, SetOwnExPid) {
EXPECT_THAT(owner.pid = getpid(), SyscallSucceeds());
ASSERT_THAT(syscall(__NR_fcntl, s.get(), F_SETOWN_EX, &owner),
- SyscallSucceeds());
+ SyscallSucceedsWithValue(0));
EXPECT_THAT(syscall(__NR_fcntl, s.get(), F_GETOWN),
SyscallSucceedsWithValue(owner.pid));
@@ -1152,7 +1158,7 @@ TEST(FcntlTest, SetOwnExPgrp) {
EXPECT_THAT(set_owner.pid = getpgrp(), SyscallSucceeds());
ASSERT_THAT(syscall(__NR_fcntl, s.get(), F_SETOWN_EX, &set_owner),
- SyscallSucceeds());
+ SyscallSucceedsWithValue(0));
// Verify with F_GETOWN_EX; using F_GETOWN on Linux may incorrectly treat the
// negative return value as an error, converting the return value to -1 and
@@ -1176,10 +1182,10 @@ TEST(FcntlTest, SetOwnExUnset) {
owner.type = F_OWNER_PID;
EXPECT_THAT(owner.pid = getpid(), SyscallSucceeds());
ASSERT_THAT(syscall(__NR_fcntl, s.get(), F_SETOWN_EX, &owner),
- SyscallSucceeds());
+ SyscallSucceedsWithValue(0));
owner.pid = 0;
ASSERT_THAT(syscall(__NR_fcntl, s.get(), F_SETOWN_EX, &owner),
- SyscallSucceeds());
+ SyscallSucceedsWithValue(0));
EXPECT_THAT(syscall(__NR_fcntl, s.get(), F_GETOWN),
SyscallSucceedsWithValue(0));
@@ -1188,10 +1194,10 @@ TEST(FcntlTest, SetOwnExUnset) {
owner.type = F_OWNER_PGRP;
EXPECT_THAT(owner.pid = getpgrp(), SyscallSucceeds());
ASSERT_THAT(syscall(__NR_fcntl, s.get(), F_SETOWN_EX, &owner),
- SyscallSucceeds());
+ SyscallSucceedsWithValue(0));
owner.pid = 0;
ASSERT_THAT(syscall(__NR_fcntl, s.get(), F_SETOWN_EX, &owner),
- SyscallSucceeds());
+ SyscallSucceedsWithValue(0));
EXPECT_THAT(syscall(__NR_fcntl, s.get(), F_GETOWN),
SyscallSucceedsWithValue(0));
@@ -1207,7 +1213,7 @@ TEST(FcntlTest, GetOwnExTid) {
EXPECT_THAT(set_owner.pid = syscall(__NR_gettid), SyscallSucceeds());
ASSERT_THAT(syscall(__NR_fcntl, s.get(), F_SETOWN_EX, &set_owner),
- SyscallSucceeds());
+ SyscallSucceedsWithValue(0));
f_owner_ex got_owner = {};
ASSERT_THAT(syscall(__NR_fcntl, s.get(), F_GETOWN_EX, &got_owner),
@@ -1225,7 +1231,7 @@ TEST(FcntlTest, GetOwnExPid) {
EXPECT_THAT(set_owner.pid = getpid(), SyscallSucceeds());
ASSERT_THAT(syscall(__NR_fcntl, s.get(), F_SETOWN_EX, &set_owner),
- SyscallSucceeds());
+ SyscallSucceedsWithValue(0));
f_owner_ex got_owner = {};
ASSERT_THAT(syscall(__NR_fcntl, s.get(), F_GETOWN_EX, &got_owner),
@@ -1243,7 +1249,7 @@ TEST(FcntlTest, GetOwnExPgrp) {
EXPECT_THAT(set_owner.pid = getpgrp(), SyscallSucceeds());
ASSERT_THAT(syscall(__NR_fcntl, s.get(), F_SETOWN_EX, &set_owner),
- SyscallSucceeds());
+ SyscallSucceedsWithValue(0));
f_owner_ex got_owner = {};
ASSERT_THAT(syscall(__NR_fcntl, s.get(), F_GETOWN_EX, &got_owner),
diff --git a/test/syscalls/linux/raw_socket_hdrincl.cc b/test/syscalls/linux/raw_socket_hdrincl.cc
index 97f0467aa..2f25aceb2 100644
--- a/test/syscalls/linux/raw_socket_hdrincl.cc
+++ b/test/syscalls/linux/raw_socket_hdrincl.cc
@@ -178,7 +178,7 @@ TEST_F(RawHDRINCL, ConnectToLoopback) {
}
TEST_F(RawHDRINCL, SendWithoutConnectSucceeds) {
- // FIXME(github.dev/issue/3159): Test currently flaky.
+ // FIXME(gvisor.dev/issue/3159): Test currently flaky.
SKIP_IF(true);
struct iphdr hdr = LoopbackHeader();
@@ -284,7 +284,7 @@ TEST_F(RawHDRINCL, SendAndReceive) {
// Send and receive a packet where the sendto address is not the same as the
// provided destination.
TEST_F(RawHDRINCL, SendAndReceiveDifferentAddress) {
- // FIXME(github.dev/issue/3160): Test currently flaky.
+ // FIXME(gvisor.dev/issue/3160): Test currently flaky.
SKIP_IF(true);
int port = 40000;
diff --git a/test/syscalls/linux/socket_inet_loopback_nogotsan.cc b/test/syscalls/linux/socket_inet_loopback_nogotsan.cc
index 2324c7f6a..791e2bd51 100644
--- a/test/syscalls/linux/socket_inet_loopback_nogotsan.cc
+++ b/test/syscalls/linux/socket_inet_loopback_nogotsan.cc
@@ -82,8 +82,11 @@ using SocketInetLoopbackTest = ::testing::TestWithParam<TestParam>;
// This test verifies that connect returns EADDRNOTAVAIL if all local ephemeral
// ports are already in use for a given destination ip/port.
+//
// We disable S/R because this test creates a large number of sockets.
-TEST_P(SocketInetLoopbackTest, TestTCPPortExhaustion_NoRandomSave) {
+//
+// FIXME(b/162475855): This test is failing reliably.
+TEST_P(SocketInetLoopbackTest, DISABLED_TestTCPPortExhaustion_NoRandomSave) {
auto const& param = GetParam();
TestAddress const& listener = param.listener;
TestAddress const& connector = param.connector;
diff --git a/test/syscalls/linux/socket_netlink_route.cc b/test/syscalls/linux/socket_netlink_route.cc
index e6647a1c3..b3fcf8e7c 100644
--- a/test/syscalls/linux/socket_netlink_route.cc
+++ b/test/syscalls/linux/socket_netlink_route.cc
@@ -577,7 +577,10 @@ TEST(NetlinkRouteTest, GetRouteDump) {
std::cout << std::endl;
- if (msg->rtm_table == RT_TABLE_MAIN) {
+ // If the test is running in a new network namespace, it will have only
+ // the local route table.
+ if (msg->rtm_table == RT_TABLE_MAIN ||
+ (!IsRunningOnGvisor() && msg->rtm_table == RT_TABLE_LOCAL)) {
routeFound = true;
dstFound = rtDstFound && dstFound;
}
diff --git a/tools/bazel.mk b/tools/bazel.mk
index 88faa0190..82fe11a03 100644
--- a/tools/bazel.mk
+++ b/tools/bazel.mk
@@ -48,7 +48,10 @@ FULL_DOCKER_RUN_OPTIONS += -v "$(BAZEL_CACHE):$(BAZEL_CACHE)"
FULL_DOCKER_RUN_OPTIONS += -v "$(GCLOUD_CONFIG):$(GCLOUD_CONFIG)"
FULL_DOCKER_RUN_OPTIONS += -v "/tmp:/tmp"
FULL_DOCKER_EXEC_OPTIONS := --user $(UID):$(GID)
-FULL_DOCKER_EXEC_OPTIONS += -i
+FULL_DOCKER_EXEC_OPTIONS += --interactive
+ifeq (true,$(shell [[ -t 0 ]] && echo true))
+FULL_DOCKER_EXEC_OPTIONS += --tty
+endif
# Add docker passthrough options.
ifneq ($(DOCKER_PRIVILEGED),)
diff --git a/tools/make_release.sh b/tools/make_release.sh
index b1cdd47b0..9137dd9bb 100755
--- a/tools/make_release.sh
+++ b/tools/make_release.sh
@@ -43,8 +43,7 @@ install_raw() {
# Copy the raw file & generate a sha512sum.
name=$(basename "${binary}")
cp -f "${binary}" "${root}/$1"
- sha512sum "${root}/$1/${name}" | \
- awk "{print $$1 \" ${name}\"}" > "${root}/$1/${name}.sha512"
+ (cd "${root}/$1" && sha512sum "${name}" > "${name}.sha512")
done
}