summaryrefslogtreecommitdiffhomepage
path: root/pkg
diff options
context:
space:
mode:
Diffstat (limited to 'pkg')
-rw-r--r--pkg/context/context.go4
-rw-r--r--pkg/eventchannel/event_test.go4
-rw-r--r--pkg/log/glog.go6
-rw-r--r--pkg/log/json.go2
-rw-r--r--pkg/log/json_k8s.go4
-rw-r--r--pkg/log/log.go2
-rw-r--r--pkg/log/log_test.go6
-rw-r--r--pkg/p9/client.go45
-rw-r--r--pkg/p9/client_test.go7
-rw-r--r--pkg/p9/handlers.go2
-rw-r--r--pkg/p9/transport_flipcall.go2
-rw-r--r--pkg/safecopy/memcpy_amd64.s111
-rw-r--r--pkg/segment/test/segment_test.go2
-rw-r--r--pkg/sentry/arch/arch.go3
-rw-r--r--pkg/sentry/arch/stack.go3
-rw-r--r--pkg/sentry/arch/syscalls_amd64.go7
-rw-r--r--pkg/sentry/arch/syscalls_arm64.go13
-rw-r--r--pkg/sentry/contexttest/contexttest.go4
-rw-r--r--pkg/sentry/fs/dirent.go6
-rw-r--r--pkg/sentry/fs/fdpipe/pipe_test.go4
-rw-r--r--pkg/sentry/fs/gofer/file_state.go1
-rw-r--r--pkg/sentry/fs/gofer/handles.go1
-rw-r--r--pkg/sentry/fs/gofer/inode.go5
-rw-r--r--pkg/sentry/fs/gofer/inode_state.go1
-rw-r--r--pkg/sentry/fs/gofer/session_state.go1
-rw-r--r--pkg/sentry/fs/gofer/util.go16
-rw-r--r--pkg/sentry/fs/host/inode.go3
-rw-r--r--pkg/sentry/fs/host/socket_test.go6
-rw-r--r--pkg/sentry/fs/inode.go3
-rw-r--r--pkg/sentry/fs/proc/sys_net.go4
-rw-r--r--pkg/sentry/fs/proc/task.go28
-rw-r--r--pkg/sentry/fs/tmpfs/fs.go3
-rw-r--r--pkg/sentry/fsimpl/ext/filesystem.go4
-rw-r--r--pkg/sentry/fsimpl/gofer/filesystem.go12
-rw-r--r--pkg/sentry/fsimpl/gofer/gofer.go60
-rw-r--r--pkg/sentry/fsimpl/gofer/p9file.go14
-rw-r--r--pkg/sentry/fsimpl/kernfs/filesystem.go6
-rw-r--r--pkg/sentry/fsimpl/proc/task.go16
-rw-r--r--pkg/sentry/fsimpl/proc/task_fds.go125
-rw-r--r--pkg/sentry/fsimpl/proc/task_files.go56
-rw-r--r--pkg/sentry/fsimpl/proc/task_net.go4
-rw-r--r--pkg/sentry/fsimpl/tmpfs/BUILD1
-rw-r--r--pkg/sentry/fsimpl/tmpfs/filesystem.go24
-rw-r--r--pkg/sentry/fsimpl/tmpfs/stat_test.go8
-rw-r--r--pkg/sentry/fsimpl/tmpfs/tmpfs.go77
-rw-r--r--pkg/sentry/kernel/fd_table.go55
-rw-r--r--pkg/sentry/kernel/ptrace.go1
-rw-r--r--pkg/sentry/kernel/shm/shm.go2
-rw-r--r--pkg/sentry/kernel/syscalls.go36
-rw-r--r--pkg/sentry/kernel/syscalls_state.go36
-rw-r--r--pkg/sentry/kernel/task.go9
-rw-r--r--pkg/sentry/kernel/task_context.go3
-rw-r--r--pkg/sentry/kernel/task_identity.go2
-rw-r--r--pkg/sentry/kernel/task_signals.go2
-rw-r--r--pkg/sentry/kernel/task_syscall.go14
-rw-r--r--pkg/sentry/kernel/time/time.go10
-rw-r--r--pkg/sentry/mm/address_space.go6
-rw-r--r--pkg/sentry/mm/aio_context.go101
-rw-r--r--pkg/sentry/mm/aio_context_state.go2
-rw-r--r--pkg/sentry/platform/kvm/kvm_arm64.go4
-rw-r--r--pkg/sentry/platform/kvm/machine_arm64_unsafe.go63
-rw-r--r--pkg/sentry/platform/ring0/BUILD2
-rw-r--r--pkg/sentry/platform/ring0/lib_arm64.go7
-rw-r--r--pkg/sentry/platform/ring0/lib_arm64_unsafe.go108
-rw-r--r--pkg/sentry/socket/netstack/netstack.go159
-rw-r--r--pkg/sentry/socket/unix/transport/BUILD1
-rw-r--r--pkg/sentry/socket/unix/transport/unix.go50
-rw-r--r--pkg/sentry/strace/strace.go3
-rw-r--r--pkg/sentry/syscalls/linux/sys_aio.go34
-rw-r--r--pkg/sentry/syscalls/linux/sys_prctl.go4
-rw-r--r--pkg/sentry/syscalls/linux/sys_read.go8
-rw-r--r--pkg/sentry/syscalls/linux/sys_socket.go9
-rw-r--r--pkg/sentry/syscalls/linux/sys_splice.go8
-rw-r--r--pkg/sentry/syscalls/linux/sys_write.go4
-rw-r--r--pkg/sentry/syscalls/linux/vfs2/BUILD6
-rw-r--r--pkg/sentry/syscalls/linux/vfs2/getdents.go20
-rw-r--r--pkg/sentry/syscalls/linux/vfs2/linux64_override_amd64.go40
-rw-r--r--pkg/sentry/syscalls/linux/vfs2/read_write.go8
-rw-r--r--pkg/sentry/syscalls/linux/vfs2/socket.go1138
-rw-r--r--pkg/sentry/syscalls/linux/vfs2/xattr.go13
-rw-r--r--pkg/sentry/vfs/anonfs.go4
-rw-r--r--pkg/sentry/vfs/file_description.go32
-rw-r--r--pkg/sentry/vfs/file_description_impl_util.go4
-rw-r--r--pkg/sentry/vfs/filesystem.go26
-rw-r--r--pkg/sentry/vfs/memxattr/BUILD15
-rw-r--r--pkg/sentry/vfs/memxattr/xattr.go102
-rw-r--r--pkg/sentry/vfs/mount.go12
-rw-r--r--pkg/sentry/vfs/mount_test.go2
-rw-r--r--pkg/sentry/vfs/options.go14
-rw-r--r--pkg/sentry/vfs/vfs.go8
-rw-r--r--pkg/state/state.go5
-rw-r--r--pkg/tcpip/checker/checker.go81
-rw-r--r--pkg/tcpip/header/eth_test.go2
-rw-r--r--pkg/tcpip/header/ndp_test.go2
-rw-r--r--pkg/tcpip/link/fdbased/endpoint.go2
-rw-r--r--pkg/tcpip/network/arp/arp_test.go3
-rw-r--r--pkg/tcpip/network/ipv6/BUILD1
-rw-r--r--pkg/tcpip/network/ipv6/icmp.go125
-rw-r--r--pkg/tcpip/network/ipv6/icmp_test.go3
-rw-r--r--pkg/tcpip/network/ipv6/ndp_test.go252
-rw-r--r--pkg/tcpip/stack/ndp_test.go18
-rw-r--r--pkg/tcpip/stack/stack_test.go32
-rw-r--r--pkg/tcpip/stack/transport_demuxer_test.go35
-rw-r--r--pkg/tcpip/tcpip.go145
-rw-r--r--pkg/tcpip/tcpip_test.go2
-rw-r--r--pkg/tcpip/transport/icmp/endpoint.go47
-rw-r--r--pkg/tcpip/transport/raw/endpoint.go18
-rw-r--r--pkg/tcpip/transport/tcp/endpoint.go387
-rw-r--r--pkg/tcpip/transport/tcp/tcp_test.go114
-rw-r--r--pkg/tcpip/transport/tcp/testing/context/context.go9
-rw-r--r--pkg/tcpip/transport/udp/endpoint.go231
-rw-r--r--pkg/tcpip/transport/udp/udp_test.go77
-rw-r--r--pkg/usermem/usermem.go3
113 files changed, 3261 insertions, 1236 deletions
diff --git a/pkg/context/context.go b/pkg/context/context.go
index 23e009ef3..5319b6d8d 100644
--- a/pkg/context/context.go
+++ b/pkg/context/context.go
@@ -127,10 +127,6 @@ func (logContext) Value(key interface{}) interface{} {
var bgContext = &logContext{Logger: log.Log()}
// Background returns an empty context using the default logger.
-//
-// Users should be wary of using a Background context. Please tag any use with
-// FIXME(b/38173783) and a note to remove this use.
-//
// Generally, one should use the Task as their context when available, or avoid
// having to use a context in places where a Task is unavailable.
//
diff --git a/pkg/eventchannel/event_test.go b/pkg/eventchannel/event_test.go
index 7f41b4a27..43750360b 100644
--- a/pkg/eventchannel/event_test.go
+++ b/pkg/eventchannel/event_test.go
@@ -78,7 +78,7 @@ func TestMultiEmitter(t *testing.T) {
for _, name := range names {
m := testMessage{name: name}
if _, err := me.Emit(m); err != nil {
- t.Fatal("me.Emit(%v) failed: %v", m, err)
+ t.Fatalf("me.Emit(%v) failed: %v", m, err)
}
}
@@ -96,7 +96,7 @@ func TestMultiEmitter(t *testing.T) {
// Close multiEmitter.
if err := me.Close(); err != nil {
- t.Fatal("me.Close() failed: %v", err)
+ t.Fatalf("me.Close() failed: %v", err)
}
// All testEmitters should be closed.
diff --git a/pkg/log/glog.go b/pkg/log/glog.go
index b4f7bb5a4..f57c4427b 100644
--- a/pkg/log/glog.go
+++ b/pkg/log/glog.go
@@ -25,7 +25,7 @@ import (
// GoogleEmitter is a wrapper that emits logs in a format compatible with
// package github.com/golang/glog.
type GoogleEmitter struct {
- Writer
+ *Writer
}
// pid is used for the threadid component of the header.
@@ -46,7 +46,7 @@ var pid = os.Getpid()
// line The line number
// msg The user-supplied message
//
-func (g *GoogleEmitter) Emit(depth int, level Level, timestamp time.Time, format string, args ...interface{}) {
+func (g GoogleEmitter) Emit(depth int, level Level, timestamp time.Time, format string, args ...interface{}) {
// Log level.
prefix := byte('?')
switch level {
@@ -81,5 +81,5 @@ func (g *GoogleEmitter) Emit(depth int, level Level, timestamp time.Time, format
message := fmt.Sprintf(format, args...)
// Emit the formatted result.
- fmt.Fprintf(&g.Writer, "%c%02d%02d %02d:%02d:%02d.%06d % 7d %s:%d] %s\n", prefix, int(month), day, hour, minute, second, microsecond, pid, file, line, message)
+ fmt.Fprintf(g.Writer, "%c%02d%02d %02d:%02d:%02d.%06d % 7d %s:%d] %s\n", prefix, int(month), day, hour, minute, second, microsecond, pid, file, line, message)
}
diff --git a/pkg/log/json.go b/pkg/log/json.go
index 0943db1cc..bdf9d691e 100644
--- a/pkg/log/json.go
+++ b/pkg/log/json.go
@@ -58,7 +58,7 @@ func (lv *Level) UnmarshalJSON(b []byte) error {
// JSONEmitter logs messages in json format.
type JSONEmitter struct {
- Writer
+ *Writer
}
// Emit implements Emitter.Emit.
diff --git a/pkg/log/json_k8s.go b/pkg/log/json_k8s.go
index 6c6fc8b6f..5883e95e1 100644
--- a/pkg/log/json_k8s.go
+++ b/pkg/log/json_k8s.go
@@ -29,11 +29,11 @@ type k8sJSONLog struct {
// K8sJSONEmitter logs messages in json format that is compatible with
// Kubernetes fluent configuration.
type K8sJSONEmitter struct {
- Writer
+ *Writer
}
// Emit implements Emitter.Emit.
-func (e *K8sJSONEmitter) Emit(_ int, level Level, timestamp time.Time, format string, v ...interface{}) {
+func (e K8sJSONEmitter) Emit(_ int, level Level, timestamp time.Time, format string, v ...interface{}) {
j := k8sJSONLog{
Log: fmt.Sprintf(format, v...),
Level: level,
diff --git a/pkg/log/log.go b/pkg/log/log.go
index a794da1aa..37e0605ad 100644
--- a/pkg/log/log.go
+++ b/pkg/log/log.go
@@ -374,5 +374,5 @@ func CopyStandardLogTo(l Level) error {
func init() {
// Store the initial value for the log.
- log.Store(&BasicLogger{Level: Info, Emitter: &GoogleEmitter{Writer{Next: os.Stderr}}})
+ log.Store(&BasicLogger{Level: Info, Emitter: GoogleEmitter{&Writer{Next: os.Stderr}}})
}
diff --git a/pkg/log/log_test.go b/pkg/log/log_test.go
index 402cc29ae..9ff18559b 100644
--- a/pkg/log/log_test.go
+++ b/pkg/log/log_test.go
@@ -52,7 +52,7 @@ func TestDropMessages(t *testing.T) {
t.Fatalf("Write should have failed")
}
- fmt.Printf("writer: %+v\n", w)
+ fmt.Printf("writer: %#v\n", &w)
tw.fail = false
if _, err := w.Write([]byte("line 2\n")); err != nil {
@@ -76,7 +76,7 @@ func TestDropMessages(t *testing.T) {
func TestCaller(t *testing.T) {
tw := &testWriter{}
- e := &GoogleEmitter{Writer: Writer{Next: tw}}
+ e := GoogleEmitter{Writer: &Writer{Next: tw}}
bl := &BasicLogger{
Emitter: e,
Level: Debug,
@@ -94,7 +94,7 @@ func BenchmarkGoogleLogging(b *testing.B) {
tw := &testWriter{
limit: 1, // Only record one message.
}
- e := &GoogleEmitter{Writer: Writer{Next: tw}}
+ e := GoogleEmitter{Writer: &Writer{Next: tw}}
bl := &BasicLogger{
Emitter: e,
Level: Debug,
diff --git a/pkg/p9/client.go b/pkg/p9/client.go
index a6f493b82..71e944c30 100644
--- a/pkg/p9/client.go
+++ b/pkg/p9/client.go
@@ -174,7 +174,7 @@ func NewClient(socket *unet.Socket, messageSize uint32, version string) (*Client
// our sendRecv function to use that functionality. Otherwise,
// we stick to sendRecvLegacy.
rversion := Rversion{}
- err := c.sendRecvLegacy(&Tversion{
+ _, err := c.sendRecvLegacy(&Tversion{
Version: versionString(requested),
MSize: messageSize,
}, &rversion)
@@ -219,11 +219,11 @@ func NewClient(socket *unet.Socket, messageSize uint32, version string) (*Client
c.sendRecv = c.sendRecvChannel
} else {
// Channel setup failed; fallback.
- c.sendRecv = c.sendRecvLegacy
+ c.sendRecv = c.sendRecvLegacySyscallErr
}
} else {
// No channels available: use the legacy mechanism.
- c.sendRecv = c.sendRecvLegacy
+ c.sendRecv = c.sendRecvLegacySyscallErr
}
// Ensure that the socket and channels are closed when the socket is shut
@@ -305,7 +305,7 @@ func (c *Client) openChannel(id int) error {
)
// Open the data channel.
- if err := c.sendRecvLegacy(&Tchannel{
+ if _, err := c.sendRecvLegacy(&Tchannel{
ID: uint32(id),
Control: 0,
}, &rchannel0); err != nil {
@@ -319,7 +319,7 @@ func (c *Client) openChannel(id int) error {
defer rchannel0.FilePayload().Close()
// Open the channel for file descriptors.
- if err := c.sendRecvLegacy(&Tchannel{
+ if _, err := c.sendRecvLegacy(&Tchannel{
ID: uint32(id),
Control: 1,
}, &rchannel1); err != nil {
@@ -431,13 +431,28 @@ func (c *Client) waitAndRecv(done chan error) error {
}
}
+// sendRecvLegacySyscallErr is a wrapper for sendRecvLegacy that converts all
+// non-syscall errors to EIO.
+func (c *Client) sendRecvLegacySyscallErr(t message, r message) error {
+ received, err := c.sendRecvLegacy(t, r)
+ if !received {
+ log.Warningf("p9.Client.sendRecvChannel: %v", err)
+ return syscall.EIO
+ }
+ return err
+}
+
// sendRecvLegacy performs a roundtrip message exchange.
//
+// sendRecvLegacy returns true if a message was received. This allows us to
+// differentiate between failed receives and successful receives where the
+// response was an error message.
+//
// This is called by internal functions.
-func (c *Client) sendRecvLegacy(t message, r message) error {
+func (c *Client) sendRecvLegacy(t message, r message) (bool, error) {
tag, ok := c.tagPool.Get()
if !ok {
- return ErrOutOfTags
+ return false, ErrOutOfTags
}
defer c.tagPool.Put(tag)
@@ -457,12 +472,12 @@ func (c *Client) sendRecvLegacy(t message, r message) error {
err := send(c.socket, Tag(tag), t)
c.sendMu.Unlock()
if err != nil {
- return err
+ return false, err
}
// Co-ordinate with other receivers.
if err := c.waitAndRecv(resp.done); err != nil {
- return err
+ return false, err
}
// Is it an error message?
@@ -470,14 +485,14 @@ func (c *Client) sendRecvLegacy(t message, r message) error {
// For convenience, we transform these directly
// into errors. Handlers need not handle this case.
if rlerr, ok := resp.r.(*Rlerror); ok {
- return syscall.Errno(rlerr.Error)
+ return true, syscall.Errno(rlerr.Error)
}
// At this point, we know it matches.
//
// Per recv call above, we will only allow a type
// match (and give our r) or an instance of Rlerror.
- return nil
+ return true, nil
}
// sendRecvChannel uses channels to send a message.
@@ -486,7 +501,7 @@ func (c *Client) sendRecvChannel(t message, r message) error {
c.channelsMu.Lock()
if len(c.availableChannels) == 0 {
c.channelsMu.Unlock()
- return c.sendRecvLegacy(t, r)
+ return c.sendRecvLegacySyscallErr(t, r)
}
idx := len(c.availableChannels) - 1
ch := c.availableChannels[idx]
@@ -526,7 +541,11 @@ func (c *Client) sendRecvChannel(t message, r message) error {
}
// Parse the server's response.
- _, retErr := ch.recv(r, rsz)
+ resp, retErr := ch.recv(r, rsz)
+ if resp == nil {
+ log.Warningf("p9.Client.sendRecvChannel: p9.channel.recv: %v", retErr)
+ retErr = syscall.EIO
+ }
// Release the channel.
c.channelsMu.Lock()
diff --git a/pkg/p9/client_test.go b/pkg/p9/client_test.go
index 29a0afadf..c757583e0 100644
--- a/pkg/p9/client_test.go
+++ b/pkg/p9/client_test.go
@@ -96,7 +96,12 @@ func benchmarkSendRecv(b *testing.B, fn func(c *Client) func(message, message) e
}
func BenchmarkSendRecvLegacy(b *testing.B) {
- benchmarkSendRecv(b, func(c *Client) func(message, message) error { return c.sendRecvLegacy })
+ benchmarkSendRecv(b, func(c *Client) func(message, message) error {
+ return func(t message, r message) error {
+ _, err := c.sendRecvLegacy(t, r)
+ return err
+ }
+ })
}
func BenchmarkSendRecvChannel(b *testing.B) {
diff --git a/pkg/p9/handlers.go b/pkg/p9/handlers.go
index a8b714cf5..1db5797dd 100644
--- a/pkg/p9/handlers.go
+++ b/pkg/p9/handlers.go
@@ -48,6 +48,8 @@ func ExtractErrno(err error) syscall.Errno {
return ExtractErrno(e.Err)
case *os.SyscallError:
return ExtractErrno(e.Err)
+ case *os.LinkError:
+ return ExtractErrno(e.Err)
}
// Default case.
diff --git a/pkg/p9/transport_flipcall.go b/pkg/p9/transport_flipcall.go
index a0d274f3b..38038abdf 100644
--- a/pkg/p9/transport_flipcall.go
+++ b/pkg/p9/transport_flipcall.go
@@ -236,7 +236,7 @@ func (ch *channel) recv(r message, rsz uint32) (message, error) {
// Convert errors appropriately; see above.
if rlerr, ok := r.(*Rlerror); ok {
- return nil, syscall.Errno(rlerr.Error)
+ return r, syscall.Errno(rlerr.Error)
}
return r, nil
diff --git a/pkg/safecopy/memcpy_amd64.s b/pkg/safecopy/memcpy_amd64.s
index 129691d68..00b46c18f 100644
--- a/pkg/safecopy/memcpy_amd64.s
+++ b/pkg/safecopy/memcpy_amd64.s
@@ -55,15 +55,9 @@ TEXT ·memcpy(SB), NOSPLIT, $0-36
MOVQ from+8(FP), SI
MOVQ n+16(FP), BX
- // REP instructions have a high startup cost, so we handle small sizes
- // with some straightline code. The REP MOVSQ instruction is really fast
- // for large sizes. The cutover is approximately 2K.
tail:
- // move_129through256 or smaller work whether or not the source and the
- // destination memory regions overlap because they load all data into
- // registers before writing it back. move_256through2048 on the other
- // hand can be used only when the memory regions don't overlap or the copy
- // direction is forward.
+ // BSR+branch table make almost all memmove/memclr benchmarks worse. Not
+ // worth doing.
TESTQ BX, BX
JEQ move_0
CMPQ BX, $2
@@ -83,31 +77,45 @@ tail:
JBE move_65through128
CMPQ BX, $256
JBE move_129through256
- // TODO: use branch table and BSR to make this just a single dispatch
-/*
- * forward copy loop
- */
- CMPQ BX, $2048
- JLS move_256through2048
-
- // Check alignment
- MOVL SI, AX
- ORL DI, AX
- TESTL $7, AX
- JEQ fwdBy8
-
- // Do 1 byte at a time
- MOVQ BX, CX
- REP; MOVSB
- RET
-
-fwdBy8:
- // Do 8 bytes at a time
- MOVQ BX, CX
- SHRQ $3, CX
- ANDQ $7, BX
- REP; MOVSQ
+move_257plus:
+ SUBQ $256, BX
+ MOVOU (SI), X0
+ MOVOU X0, (DI)
+ MOVOU 16(SI), X1
+ MOVOU X1, 16(DI)
+ MOVOU 32(SI), X2
+ MOVOU X2, 32(DI)
+ MOVOU 48(SI), X3
+ MOVOU X3, 48(DI)
+ MOVOU 64(SI), X4
+ MOVOU X4, 64(DI)
+ MOVOU 80(SI), X5
+ MOVOU X5, 80(DI)
+ MOVOU 96(SI), X6
+ MOVOU X6, 96(DI)
+ MOVOU 112(SI), X7
+ MOVOU X7, 112(DI)
+ MOVOU 128(SI), X8
+ MOVOU X8, 128(DI)
+ MOVOU 144(SI), X9
+ MOVOU X9, 144(DI)
+ MOVOU 160(SI), X10
+ MOVOU X10, 160(DI)
+ MOVOU 176(SI), X11
+ MOVOU X11, 176(DI)
+ MOVOU 192(SI), X12
+ MOVOU X12, 192(DI)
+ MOVOU 208(SI), X13
+ MOVOU X13, 208(DI)
+ MOVOU 224(SI), X14
+ MOVOU X14, 224(DI)
+ MOVOU 240(SI), X15
+ MOVOU X15, 240(DI)
+ CMPQ BX, $256
+ LEAQ 256(SI), SI
+ LEAQ 256(DI), DI
+ JGE move_257plus
JMP tail
move_1or2:
@@ -209,42 +217,3 @@ move_129through256:
MOVOU -16(SI)(BX*1), X15
MOVOU X15, -16(DI)(BX*1)
RET
-move_256through2048:
- SUBQ $256, BX
- MOVOU (SI), X0
- MOVOU X0, (DI)
- MOVOU 16(SI), X1
- MOVOU X1, 16(DI)
- MOVOU 32(SI), X2
- MOVOU X2, 32(DI)
- MOVOU 48(SI), X3
- MOVOU X3, 48(DI)
- MOVOU 64(SI), X4
- MOVOU X4, 64(DI)
- MOVOU 80(SI), X5
- MOVOU X5, 80(DI)
- MOVOU 96(SI), X6
- MOVOU X6, 96(DI)
- MOVOU 112(SI), X7
- MOVOU X7, 112(DI)
- MOVOU 128(SI), X8
- MOVOU X8, 128(DI)
- MOVOU 144(SI), X9
- MOVOU X9, 144(DI)
- MOVOU 160(SI), X10
- MOVOU X10, 160(DI)
- MOVOU 176(SI), X11
- MOVOU X11, 176(DI)
- MOVOU 192(SI), X12
- MOVOU X12, 192(DI)
- MOVOU 208(SI), X13
- MOVOU X13, 208(DI)
- MOVOU 224(SI), X14
- MOVOU X14, 224(DI)
- MOVOU 240(SI), X15
- MOVOU X15, 240(DI)
- CMPQ BX, $256
- LEAQ 256(SI), SI
- LEAQ 256(DI), DI
- JGE move_256through2048
- JMP tail
diff --git a/pkg/segment/test/segment_test.go b/pkg/segment/test/segment_test.go
index f19a005f3..97b16c158 100644
--- a/pkg/segment/test/segment_test.go
+++ b/pkg/segment/test/segment_test.go
@@ -63,7 +63,7 @@ func checkSet(s *Set, expectedSegments int) error {
return fmt.Errorf("incorrect order: key %d (segment %d) >= key %d (segment %d)", prev, nrSegments-1, next, nrSegments)
}
if got, want := seg.Value(), seg.Start()+valueOffset; got != want {
- return fmt.Errorf("segment %d has key %d, value %d (expected %d)", nrSegments, seg.Start, got, want)
+ return fmt.Errorf("segment %d has key %d, value %d (expected %d)", nrSegments, seg.Start(), got, want)
}
prev = next
havePrev = true
diff --git a/pkg/sentry/arch/arch.go b/pkg/sentry/arch/arch.go
index 1d11cc472..a903d031c 100644
--- a/pkg/sentry/arch/arch.go
+++ b/pkg/sentry/arch/arch.go
@@ -88,6 +88,9 @@ type Context interface {
// SyscallNo returns the syscall number.
SyscallNo() uintptr
+ // SyscallSaveOrig save orignal register value.
+ SyscallSaveOrig()
+
// SyscallArgs returns the syscall arguments in an array.
SyscallArgs() SyscallArguments
diff --git a/pkg/sentry/arch/stack.go b/pkg/sentry/arch/stack.go
index 09bceabc9..1108fa0bd 100644
--- a/pkg/sentry/arch/stack.go
+++ b/pkg/sentry/arch/stack.go
@@ -97,7 +97,6 @@ func (s *Stack) Push(vals ...interface{}) (usermem.Addr, error) {
if c < 0 {
return 0, fmt.Errorf("bad binary.Size for %T", v)
}
- // TODO(b/38173783): Use a real context.Context.
n, err := usermem.CopyObjectOut(context.Background(), s.IO, s.Bottom-usermem.Addr(c), norm, usermem.IOOpts{})
if err != nil || c != n {
return 0, err
@@ -121,11 +120,9 @@ func (s *Stack) Pop(vals ...interface{}) (usermem.Addr, error) {
var err error
if isVaddr {
value := s.Arch.Native(uintptr(0))
- // TODO(b/38173783): Use a real context.Context.
n, err = usermem.CopyObjectIn(context.Background(), s.IO, s.Bottom, value, usermem.IOOpts{})
*vaddr = usermem.Addr(s.Arch.Value(value))
} else {
- // TODO(b/38173783): Use a real context.Context.
n, err = usermem.CopyObjectIn(context.Background(), s.IO, s.Bottom, v, usermem.IOOpts{})
}
if err != nil {
diff --git a/pkg/sentry/arch/syscalls_amd64.go b/pkg/sentry/arch/syscalls_amd64.go
index 8b4f23007..3859f41ee 100644
--- a/pkg/sentry/arch/syscalls_amd64.go
+++ b/pkg/sentry/arch/syscalls_amd64.go
@@ -18,6 +18,13 @@ package arch
const restartSyscallNr = uintptr(219)
+// SyscallSaveOrig save the value of the register which is clobbered in
+// syscall handler(doSyscall()).
+//
+// Noop on x86.
+func (c *context64) SyscallSaveOrig() {
+}
+
// SyscallNo returns the syscall number according to the 64-bit convention.
func (c *context64) SyscallNo() uintptr {
return uintptr(c.Regs.Orig_rax)
diff --git a/pkg/sentry/arch/syscalls_arm64.go b/pkg/sentry/arch/syscalls_arm64.go
index dc13b6124..92d062513 100644
--- a/pkg/sentry/arch/syscalls_arm64.go
+++ b/pkg/sentry/arch/syscalls_arm64.go
@@ -18,6 +18,17 @@ package arch
const restartSyscallNr = uintptr(128)
+// SyscallSaveOrig save the value of the register R0 which is clobbered in
+// syscall handler(doSyscall()).
+//
+// In linux, at the entry of the syscall handler(el0_svc_common()), value of R0
+// is saved to the pt_regs.orig_x0 in kernel code. But currently, the orig_x0
+// was not accessible to the user space application, so we have to do the same
+// operation in the sentry code to save the R0 value into the App context.
+func (c *context64) SyscallSaveOrig() {
+ c.OrigR0 = c.Regs.Regs[0]
+}
+
// SyscallNo returns the syscall number according to the 64-bit convention.
func (c *context64) SyscallNo() uintptr {
return uintptr(c.Regs.Regs[8])
@@ -40,7 +51,7 @@ func (c *context64) SyscallNo() uintptr {
// R30: the link register.
func (c *context64) SyscallArgs() SyscallArguments {
return SyscallArguments{
- SyscallArgument{Value: uintptr(c.Regs.Regs[0])},
+ SyscallArgument{Value: uintptr(c.OrigR0)},
SyscallArgument{Value: uintptr(c.Regs.Regs[1])},
SyscallArgument{Value: uintptr(c.Regs.Regs[2])},
SyscallArgument{Value: uintptr(c.Regs.Regs[3])},
diff --git a/pkg/sentry/contexttest/contexttest.go b/pkg/sentry/contexttest/contexttest.go
index 031fc64ec..8e5658c7a 100644
--- a/pkg/sentry/contexttest/contexttest.go
+++ b/pkg/sentry/contexttest/contexttest.go
@@ -97,7 +97,7 @@ type hostClock struct {
}
// Now implements ktime.Clock.Now.
-func (hostClock) Now() ktime.Time {
+func (*hostClock) Now() ktime.Time {
return ktime.FromNanoseconds(time.Now().UnixNano())
}
@@ -127,7 +127,7 @@ func (t *TestContext) Value(key interface{}) interface{} {
case uniqueid.CtxInotifyCookie:
return atomic.AddUint32(&lastInotifyCookie, 1)
case ktime.CtxRealtimeClock:
- return hostClock{}
+ return &hostClock{}
default:
if val, ok := t.otherValues[key]; ok {
return val
diff --git a/pkg/sentry/fs/dirent.go b/pkg/sentry/fs/dirent.go
index 0266a5287..65be12175 100644
--- a/pkg/sentry/fs/dirent.go
+++ b/pkg/sentry/fs/dirent.go
@@ -312,9 +312,9 @@ func (d *Dirent) SyncAll(ctx context.Context) {
// There is nothing to sync for a read-only filesystem.
if !d.Inode.MountSource.Flags.ReadOnly {
- // FIXME(b/34856369): This should be a mount traversal, not a
- // Dirent traversal, because some Inodes that need to be synced
- // may no longer be reachable by name (after sys_unlink).
+ // NOTE(b/34856369): This should be a mount traversal, not a Dirent
+ // traversal, because some Inodes that need to be synced may no longer
+ // be reachable by name (after sys_unlink).
//
// Write out metadata, dirty page cached pages, and sync disk/remote
// caches.
diff --git a/pkg/sentry/fs/fdpipe/pipe_test.go b/pkg/sentry/fs/fdpipe/pipe_test.go
index 5aff0cc95..a0082ecca 100644
--- a/pkg/sentry/fs/fdpipe/pipe_test.go
+++ b/pkg/sentry/fs/fdpipe/pipe_test.go
@@ -119,7 +119,7 @@ func TestNewPipe(t *testing.T) {
continue
}
if flags := p.flags; test.flags != flags {
- t.Errorf("%s: got file flags %s, want %s", test.desc, flags, test.flags)
+ t.Errorf("%s: got file flags %v, want %v", test.desc, flags, test.flags)
continue
}
if len(test.readAheadBuffer) != len(p.readAheadBuffer) {
@@ -136,7 +136,7 @@ func TestNewPipe(t *testing.T) {
continue
}
if !fdnotifier.HasFD(int32(f.FD())) {
- t.Errorf("%s: pipe fd %d is not registered for events", test.desc, f.FD)
+ t.Errorf("%s: pipe fd %d is not registered for events", test.desc, f.FD())
}
}
}
diff --git a/pkg/sentry/fs/gofer/file_state.go b/pkg/sentry/fs/gofer/file_state.go
index ff96b28ba..edd6576aa 100644
--- a/pkg/sentry/fs/gofer/file_state.go
+++ b/pkg/sentry/fs/gofer/file_state.go
@@ -34,7 +34,6 @@ func (f *fileOperations) afterLoad() {
flags := f.flags
flags.Truncate = false
- // TODO(b/38173783): Context is not plumbed to save/restore.
f.handles, err = f.inodeOperations.fileState.getHandles(context.Background(), flags, f.inodeOperations.cachingInodeOps)
if err != nil {
return fmt.Errorf("failed to re-open handle: %v", err)
diff --git a/pkg/sentry/fs/gofer/handles.go b/pkg/sentry/fs/gofer/handles.go
index 9f7c3e89f..fc14249be 100644
--- a/pkg/sentry/fs/gofer/handles.go
+++ b/pkg/sentry/fs/gofer/handles.go
@@ -57,7 +57,6 @@ func (h *handles) DecRef() {
}
}
}
- // FIXME(b/38173783): Context is not plumbed here.
if err := h.File.close(context.Background()); err != nil {
log.Warningf("error closing p9 file: %v", err)
}
diff --git a/pkg/sentry/fs/gofer/inode.go b/pkg/sentry/fs/gofer/inode.go
index 1c934981b..a016c896e 100644
--- a/pkg/sentry/fs/gofer/inode.go
+++ b/pkg/sentry/fs/gofer/inode.go
@@ -273,7 +273,7 @@ func (i *inodeFileState) recreateReadHandles(ctx context.Context, writer *handle
// operations on the old will see the new data. Then, make the new handle take
// ownereship of the old FD and mark the old readHandle to not close the FD
// when done.
- if err := syscall.Dup3(h.Host.FD(), i.readHandles.Host.FD(), 0); err != nil {
+ if err := syscall.Dup3(h.Host.FD(), i.readHandles.Host.FD(), syscall.O_CLOEXEC); err != nil {
return err
}
@@ -710,13 +710,10 @@ func init() {
}
// AddLink implements InodeOperations.AddLink, but is currently a noop.
-// FIXME(b/63117438): Remove this from InodeOperations altogether.
func (*inodeOperations) AddLink() {}
// DropLink implements InodeOperations.DropLink, but is currently a noop.
-// FIXME(b/63117438): Remove this from InodeOperations altogether.
func (*inodeOperations) DropLink() {}
// NotifyStatusChange implements fs.InodeOperations.NotifyStatusChange.
-// FIXME(b/63117438): Remove this from InodeOperations altogether.
func (i *inodeOperations) NotifyStatusChange(ctx context.Context) {}
diff --git a/pkg/sentry/fs/gofer/inode_state.go b/pkg/sentry/fs/gofer/inode_state.go
index 238f7804c..a3402e343 100644
--- a/pkg/sentry/fs/gofer/inode_state.go
+++ b/pkg/sentry/fs/gofer/inode_state.go
@@ -123,7 +123,6 @@ func (i *inodeFileState) afterLoad() {
// beforeSave.
return fmt.Errorf("failed to find path for inode number %d. Device %s contains %s", i.sattr.InodeID, i.s.connID, fs.InodeMappings(i.s.inodeMappings))
}
- // TODO(b/38173783): Context is not plumbed to save/restore.
ctx := &dummyClockContext{context.Background()}
_, i.file, err = i.s.attach.walk(ctx, splitAbsolutePath(name))
diff --git a/pkg/sentry/fs/gofer/session_state.go b/pkg/sentry/fs/gofer/session_state.go
index 111da59f9..2d398b753 100644
--- a/pkg/sentry/fs/gofer/session_state.go
+++ b/pkg/sentry/fs/gofer/session_state.go
@@ -104,7 +104,6 @@ func (s *session) afterLoad() {
// If private unix sockets are enabled, create and fill the session's endpoint
// maps.
if opts.privateunixsocket {
- // TODO(b/38173783): Context is not plumbed to save/restore.
ctx := &dummyClockContext{context.Background()}
if err = s.restoreEndpointMaps(ctx); err != nil {
diff --git a/pkg/sentry/fs/gofer/util.go b/pkg/sentry/fs/gofer/util.go
index 2d8d3a2ea..47a6c69bf 100644
--- a/pkg/sentry/fs/gofer/util.go
+++ b/pkg/sentry/fs/gofer/util.go
@@ -20,17 +20,29 @@ import (
"gvisor.dev/gvisor/pkg/context"
"gvisor.dev/gvisor/pkg/p9"
"gvisor.dev/gvisor/pkg/sentry/fs"
+ ktime "gvisor.dev/gvisor/pkg/sentry/kernel/time"
)
func utimes(ctx context.Context, file contextFile, ts fs.TimeSpec) error {
if ts.ATimeOmit && ts.MTimeOmit {
return nil
}
+
+ // Replace requests to use the "system time" with the current time to
+ // ensure that timestamps remain consistent with the remote
+ // filesystem.
+ now := ktime.NowFromContext(ctx)
+ if ts.ATimeSetSystemTime {
+ ts.ATime = now
+ }
+ if ts.MTimeSetSystemTime {
+ ts.MTime = now
+ }
mask := p9.SetAttrMask{
ATime: !ts.ATimeOmit,
- ATimeNotSystemTime: !ts.ATimeSetSystemTime,
+ ATimeNotSystemTime: true,
MTime: !ts.MTimeOmit,
- MTimeNotSystemTime: !ts.MTimeSetSystemTime,
+ MTimeNotSystemTime: true,
}
as, ans := ts.ATime.Unix()
ms, mns := ts.MTime.Unix()
diff --git a/pkg/sentry/fs/host/inode.go b/pkg/sentry/fs/host/inode.go
index 1da3c0a17..62f1246aa 100644
--- a/pkg/sentry/fs/host/inode.go
+++ b/pkg/sentry/fs/host/inode.go
@@ -397,15 +397,12 @@ func (i *inodeOperations) StatFS(context.Context) (fs.Info, error) {
}
// AddLink implements fs.InodeOperations.AddLink.
-// FIXME(b/63117438): Remove this from InodeOperations altogether.
func (i *inodeOperations) AddLink() {}
// DropLink implements fs.InodeOperations.DropLink.
-// FIXME(b/63117438): Remove this from InodeOperations altogether.
func (i *inodeOperations) DropLink() {}
// NotifyStatusChange implements fs.InodeOperations.NotifyStatusChange.
-// FIXME(b/63117438): Remove this from InodeOperations altogether.
func (i *inodeOperations) NotifyStatusChange(ctx context.Context) {}
// readdirAll returns all of the directory entries in i.
diff --git a/pkg/sentry/fs/host/socket_test.go b/pkg/sentry/fs/host/socket_test.go
index eb4afe520..affdbcacb 100644
--- a/pkg/sentry/fs/host/socket_test.go
+++ b/pkg/sentry/fs/host/socket_test.go
@@ -199,14 +199,14 @@ func TestListen(t *testing.T) {
}
func TestPasscred(t *testing.T) {
- e := ConnectedEndpoint{}
+ e := &ConnectedEndpoint{}
if got, want := e.Passcred(), false; got != want {
t.Errorf("Got %#v.Passcred() = %t, want = %t", e, got, want)
}
}
func TestGetLocalAddress(t *testing.T) {
- e := ConnectedEndpoint{path: "foo"}
+ e := &ConnectedEndpoint{path: "foo"}
want := tcpip.FullAddress{Addr: tcpip.Address("foo")}
if got, err := e.GetLocalAddress(); err != nil || got != want {
t.Errorf("Got %#v.GetLocalAddress() = %#v, %v, want = %#v, %v", e, got, err, want, nil)
@@ -214,7 +214,7 @@ func TestGetLocalAddress(t *testing.T) {
}
func TestQueuedSize(t *testing.T) {
- e := ConnectedEndpoint{}
+ e := &ConnectedEndpoint{}
tests := []struct {
name string
f func() int64
diff --git a/pkg/sentry/fs/inode.go b/pkg/sentry/fs/inode.go
index 55fb71c16..a34fbc946 100644
--- a/pkg/sentry/fs/inode.go
+++ b/pkg/sentry/fs/inode.go
@@ -102,7 +102,6 @@ func (i *Inode) DecRef() {
// destroy releases the Inode and releases the msrc reference taken.
func (i *Inode) destroy() {
- // FIXME(b/38173783): Context is not plumbed here.
ctx := context.Background()
if err := i.WriteOut(ctx); err != nil {
// FIXME(b/65209558): Mark as warning again once noatime is
@@ -397,8 +396,6 @@ func (i *Inode) Getlink(ctx context.Context) (*Dirent, error) {
// AddLink calls i.InodeOperations.AddLink.
func (i *Inode) AddLink() {
if i.overlay != nil {
- // FIXME(b/63117438): Remove this from InodeOperations altogether.
- //
// This interface is only used by ramfs to update metadata of
// children. These filesystems should _never_ have overlay
// Inodes cached as children. So explicitly disallow this
diff --git a/pkg/sentry/fs/proc/sys_net.go b/pkg/sentry/fs/proc/sys_net.go
index d4c4b533d..702fdd392 100644
--- a/pkg/sentry/fs/proc/sys_net.go
+++ b/pkg/sentry/fs/proc/sys_net.go
@@ -80,7 +80,7 @@ func newTCPMemInode(ctx context.Context, msrc *fs.MountSource, s inet.Stack, dir
}
// Truncate implements fs.InodeOperations.Truncate.
-func (tcpMemInode) Truncate(context.Context, *fs.Inode, int64) error {
+func (*tcpMemInode) Truncate(context.Context, *fs.Inode, int64) error {
return nil
}
@@ -196,7 +196,7 @@ func newTCPSackInode(ctx context.Context, msrc *fs.MountSource, s inet.Stack) *f
}
// Truncate implements fs.InodeOperations.Truncate.
-func (tcpSack) Truncate(context.Context, *fs.Inode, int64) error {
+func (*tcpSack) Truncate(context.Context, *fs.Inode, int64) error {
return nil
}
diff --git a/pkg/sentry/fs/proc/task.go b/pkg/sentry/fs/proc/task.go
index d6c5dd2c1..4d42eac83 100644
--- a/pkg/sentry/fs/proc/task.go
+++ b/pkg/sentry/fs/proc/task.go
@@ -57,6 +57,16 @@ func getTaskMM(t *kernel.Task) (*mm.MemoryManager, error) {
return m, nil
}
+func checkTaskState(t *kernel.Task) error {
+ switch t.ExitState() {
+ case kernel.TaskExitZombie:
+ return syserror.EACCES
+ case kernel.TaskExitDead:
+ return syserror.ESRCH
+ }
+ return nil
+}
+
// taskDir represents a task-level directory.
//
// +stateify savable
@@ -254,11 +264,12 @@ func newExe(t *kernel.Task, msrc *fs.MountSource) *fs.Inode {
}
func (e *exe) executable() (file fsbridge.File, err error) {
+ if err := checkTaskState(e.t); err != nil {
+ return nil, err
+ }
e.t.WithMuLocked(func(t *kernel.Task) {
mm := t.MemoryManager()
if mm == nil {
- // TODO(b/34851096): Check shouldn't allow Readlink once the
- // Task is zombied.
err = syserror.EACCES
return
}
@@ -268,7 +279,7 @@ func (e *exe) executable() (file fsbridge.File, err error) {
// (with locks held).
file = mm.Executable()
if file == nil {
- err = syserror.ENOENT
+ err = syserror.ESRCH
}
})
return
@@ -313,11 +324,22 @@ func newNamespaceSymlink(t *kernel.Task, msrc *fs.MountSource, name string) *fs.
return newProcInode(t, n, msrc, fs.Symlink, t)
}
+// Readlink reads the symlink value.
+func (n *namespaceSymlink) Readlink(ctx context.Context, inode *fs.Inode) (string, error) {
+ if err := checkTaskState(n.t); err != nil {
+ return "", err
+ }
+ return n.Symlink.Readlink(ctx, inode)
+}
+
// Getlink implements fs.InodeOperations.Getlink.
func (n *namespaceSymlink) Getlink(ctx context.Context, inode *fs.Inode) (*fs.Dirent, error) {
if !kernel.ContextCanTrace(ctx, n.t, false) {
return nil, syserror.EACCES
}
+ if err := checkTaskState(n.t); err != nil {
+ return nil, err
+ }
// Create a new regular file to fake the namespace file.
iops := fsutil.NewNoReadWriteFileInode(ctx, fs.RootOwner, fs.FilePermsFromMode(0777), linux.PROC_SUPER_MAGIC)
diff --git a/pkg/sentry/fs/tmpfs/fs.go b/pkg/sentry/fs/tmpfs/fs.go
index d5be56c3f..bc117ca6a 100644
--- a/pkg/sentry/fs/tmpfs/fs.go
+++ b/pkg/sentry/fs/tmpfs/fs.go
@@ -44,9 +44,6 @@ const (
// lookup.
cacheRevalidate = "revalidate"
- // TODO(edahlgren/mpratt): support a tmpfs size limit.
- // size = "size"
-
// Permissions that exceed modeMask will be rejected.
modeMask = 01777
diff --git a/pkg/sentry/fsimpl/ext/filesystem.go b/pkg/sentry/fsimpl/ext/filesystem.go
index 48eaccdbc..afea58f65 100644
--- a/pkg/sentry/fsimpl/ext/filesystem.go
+++ b/pkg/sentry/fsimpl/ext/filesystem.go
@@ -476,7 +476,7 @@ func (fs *filesystem) BoundEndpointAt(ctx context.Context, rp *vfs.ResolvingPath
}
// ListxattrAt implements vfs.FilesystemImpl.ListxattrAt.
-func (fs *filesystem) ListxattrAt(ctx context.Context, rp *vfs.ResolvingPath) ([]string, error) {
+func (fs *filesystem) ListxattrAt(ctx context.Context, rp *vfs.ResolvingPath, size uint64) ([]string, error) {
_, _, err := fs.walk(rp, false)
if err != nil {
return nil, err
@@ -485,7 +485,7 @@ func (fs *filesystem) ListxattrAt(ctx context.Context, rp *vfs.ResolvingPath) ([
}
// GetxattrAt implements vfs.FilesystemImpl.GetxattrAt.
-func (fs *filesystem) GetxattrAt(ctx context.Context, rp *vfs.ResolvingPath, name string) (string, error) {
+func (fs *filesystem) GetxattrAt(ctx context.Context, rp *vfs.ResolvingPath, opts vfs.GetxattrOptions) (string, error) {
_, _, err := fs.walk(rp, false)
if err != nil {
return "", err
diff --git a/pkg/sentry/fsimpl/gofer/filesystem.go b/pkg/sentry/fsimpl/gofer/filesystem.go
index 137260898..cd744bf5e 100644
--- a/pkg/sentry/fsimpl/gofer/filesystem.go
+++ b/pkg/sentry/fsimpl/gofer/filesystem.go
@@ -1080,7 +1080,7 @@ func (fs *filesystem) BoundEndpointAt(ctx context.Context, rp *vfs.ResolvingPath
}
// ListxattrAt implements vfs.FilesystemImpl.ListxattrAt.
-func (fs *filesystem) ListxattrAt(ctx context.Context, rp *vfs.ResolvingPath) ([]string, error) {
+func (fs *filesystem) ListxattrAt(ctx context.Context, rp *vfs.ResolvingPath, size uint64) ([]string, error) {
var ds *[]*dentry
fs.renameMu.RLock()
defer fs.renameMuRUnlockAndCheckCaching(&ds)
@@ -1088,11 +1088,11 @@ func (fs *filesystem) ListxattrAt(ctx context.Context, rp *vfs.ResolvingPath) ([
if err != nil {
return nil, err
}
- return d.listxattr(ctx)
+ return d.listxattr(ctx, rp.Credentials(), size)
}
// GetxattrAt implements vfs.FilesystemImpl.GetxattrAt.
-func (fs *filesystem) GetxattrAt(ctx context.Context, rp *vfs.ResolvingPath, name string) (string, error) {
+func (fs *filesystem) GetxattrAt(ctx context.Context, rp *vfs.ResolvingPath, opts vfs.GetxattrOptions) (string, error) {
var ds *[]*dentry
fs.renameMu.RLock()
defer fs.renameMuRUnlockAndCheckCaching(&ds)
@@ -1100,7 +1100,7 @@ func (fs *filesystem) GetxattrAt(ctx context.Context, rp *vfs.ResolvingPath, nam
if err != nil {
return "", err
}
- return d.getxattr(ctx, name)
+ return d.getxattr(ctx, rp.Credentials(), &opts)
}
// SetxattrAt implements vfs.FilesystemImpl.SetxattrAt.
@@ -1112,7 +1112,7 @@ func (fs *filesystem) SetxattrAt(ctx context.Context, rp *vfs.ResolvingPath, opt
if err != nil {
return err
}
- return d.setxattr(ctx, &opts)
+ return d.setxattr(ctx, rp.Credentials(), &opts)
}
// RemovexattrAt implements vfs.FilesystemImpl.RemovexattrAt.
@@ -1124,7 +1124,7 @@ func (fs *filesystem) RemovexattrAt(ctx context.Context, rp *vfs.ResolvingPath,
if err != nil {
return err
}
- return d.removexattr(ctx, name)
+ return d.removexattr(ctx, rp.Credentials(), name)
}
// PrependPath implements vfs.FilesystemImpl.PrependPath.
diff --git a/pkg/sentry/fsimpl/gofer/gofer.go b/pkg/sentry/fsimpl/gofer/gofer.go
index 20edaf643..2485cdb53 100644
--- a/pkg/sentry/fsimpl/gofer/gofer.go
+++ b/pkg/sentry/fsimpl/gofer/gofer.go
@@ -34,6 +34,7 @@ package gofer
import (
"fmt"
"strconv"
+ "strings"
"sync"
"sync/atomic"
"syscall"
@@ -1024,21 +1025,50 @@ func (d *dentry) setDeleted() {
atomic.StoreUint32(&d.deleted, 1)
}
-func (d *dentry) listxattr(ctx context.Context) ([]string, error) {
- return nil, syserror.ENOTSUP
+// We only support xattrs prefixed with "user." (see b/148380782). Currently,
+// there is no need to expose any other xattrs through a gofer.
+func (d *dentry) listxattr(ctx context.Context, creds *auth.Credentials, size uint64) ([]string, error) {
+ xattrMap, err := d.file.listXattr(ctx, size)
+ if err != nil {
+ return nil, err
+ }
+ xattrs := make([]string, 0, len(xattrMap))
+ for x := range xattrMap {
+ if strings.HasPrefix(x, linux.XATTR_USER_PREFIX) {
+ xattrs = append(xattrs, x)
+ }
+ }
+ return xattrs, nil
}
-func (d *dentry) getxattr(ctx context.Context, name string) (string, error) {
- // TODO(jamieliu): add vfs.GetxattrOptions.Size
- return d.file.getXattr(ctx, name, linux.XATTR_SIZE_MAX)
+func (d *dentry) getxattr(ctx context.Context, creds *auth.Credentials, opts *vfs.GetxattrOptions) (string, error) {
+ if err := d.checkPermissions(creds, vfs.MayRead); err != nil {
+ return "", err
+ }
+ if !strings.HasPrefix(opts.Name, linux.XATTR_USER_PREFIX) {
+ return "", syserror.EOPNOTSUPP
+ }
+ return d.file.getXattr(ctx, opts.Name, opts.Size)
}
-func (d *dentry) setxattr(ctx context.Context, opts *vfs.SetxattrOptions) error {
+func (d *dentry) setxattr(ctx context.Context, creds *auth.Credentials, opts *vfs.SetxattrOptions) error {
+ if err := d.checkPermissions(creds, vfs.MayWrite); err != nil {
+ return err
+ }
+ if !strings.HasPrefix(opts.Name, linux.XATTR_USER_PREFIX) {
+ return syserror.EOPNOTSUPP
+ }
return d.file.setXattr(ctx, opts.Name, opts.Value, opts.Flags)
}
-func (d *dentry) removexattr(ctx context.Context, name string) error {
- return syserror.ENOTSUP
+func (d *dentry) removexattr(ctx context.Context, creds *auth.Credentials, name string) error {
+ if err := d.checkPermissions(creds, vfs.MayWrite); err != nil {
+ return err
+ }
+ if !strings.HasPrefix(name, linux.XATTR_USER_PREFIX) {
+ return syserror.EOPNOTSUPP
+ }
+ return d.file.removeXattr(ctx, name)
}
// Preconditions: d.isRegularFile() || d.isDirectory().
@@ -1089,7 +1119,7 @@ func (d *dentry) ensureSharedHandle(ctx context.Context, read, write, trunc bool
// description, but this doesn't matter since they refer to the
// same file (unless d.fs.opts.overlayfsStaleRead is true,
// which we handle separately).
- if err := syscall.Dup3(int(h.fd), int(d.handle.fd), 0); err != nil {
+ if err := syscall.Dup3(int(h.fd), int(d.handle.fd), syscall.O_CLOEXEC); err != nil {
d.handleMu.Unlock()
ctx.Warningf("gofer.dentry.ensureSharedHandle: failed to dup fd %d to fd %d: %v", h.fd, d.handle.fd, err)
h.close(ctx)
@@ -1189,21 +1219,21 @@ func (fd *fileDescription) SetStat(ctx context.Context, opts vfs.SetStatOptions)
}
// Listxattr implements vfs.FileDescriptionImpl.Listxattr.
-func (fd *fileDescription) Listxattr(ctx context.Context) ([]string, error) {
- return fd.dentry().listxattr(ctx)
+func (fd *fileDescription) Listxattr(ctx context.Context, size uint64) ([]string, error) {
+ return fd.dentry().listxattr(ctx, auth.CredentialsFromContext(ctx), size)
}
// Getxattr implements vfs.FileDescriptionImpl.Getxattr.
-func (fd *fileDescription) Getxattr(ctx context.Context, name string) (string, error) {
- return fd.dentry().getxattr(ctx, name)
+func (fd *fileDescription) Getxattr(ctx context.Context, opts vfs.GetxattrOptions) (string, error) {
+ return fd.dentry().getxattr(ctx, auth.CredentialsFromContext(ctx), &opts)
}
// Setxattr implements vfs.FileDescriptionImpl.Setxattr.
func (fd *fileDescription) Setxattr(ctx context.Context, opts vfs.SetxattrOptions) error {
- return fd.dentry().setxattr(ctx, &opts)
+ return fd.dentry().setxattr(ctx, auth.CredentialsFromContext(ctx), &opts)
}
// Removexattr implements vfs.FileDescriptionImpl.Removexattr.
func (fd *fileDescription) Removexattr(ctx context.Context, name string) error {
- return fd.dentry().removexattr(ctx, name)
+ return fd.dentry().removexattr(ctx, auth.CredentialsFromContext(ctx), name)
}
diff --git a/pkg/sentry/fsimpl/gofer/p9file.go b/pkg/sentry/fsimpl/gofer/p9file.go
index 755ac2985..87f0b877f 100644
--- a/pkg/sentry/fsimpl/gofer/p9file.go
+++ b/pkg/sentry/fsimpl/gofer/p9file.go
@@ -85,6 +85,13 @@ func (f p9file) setAttr(ctx context.Context, valid p9.SetAttrMask, attr p9.SetAt
return err
}
+func (f p9file) listXattr(ctx context.Context, size uint64) (map[string]struct{}, error) {
+ ctx.UninterruptibleSleepStart(false)
+ xattrs, err := f.file.ListXattr(size)
+ ctx.UninterruptibleSleepFinish(false)
+ return xattrs, err
+}
+
func (f p9file) getXattr(ctx context.Context, name string, size uint64) (string, error) {
ctx.UninterruptibleSleepStart(false)
val, err := f.file.GetXattr(name, size)
@@ -99,6 +106,13 @@ func (f p9file) setXattr(ctx context.Context, name, value string, flags uint32)
return err
}
+func (f p9file) removeXattr(ctx context.Context, name string) error {
+ ctx.UninterruptibleSleepStart(false)
+ err := f.file.RemoveXattr(name)
+ ctx.UninterruptibleSleepFinish(false)
+ return err
+}
+
func (f p9file) allocate(ctx context.Context, mode p9.AllocateMode, offset, length uint64) error {
ctx.UninterruptibleSleepStart(false)
err := f.file.Allocate(mode, offset, length)
diff --git a/pkg/sentry/fsimpl/kernfs/filesystem.go b/pkg/sentry/fsimpl/kernfs/filesystem.go
index 16a3c18ae..baf81b4db 100644
--- a/pkg/sentry/fsimpl/kernfs/filesystem.go
+++ b/pkg/sentry/fsimpl/kernfs/filesystem.go
@@ -682,7 +682,7 @@ func (fs *Filesystem) StatFSAt(ctx context.Context, rp *vfs.ResolvingPath) (linu
if err != nil {
return linux.Statfs{}, err
}
- // TODO: actually implement statfs
+ // TODO(gvisor.dev/issue/1193): actually implement statfs.
return linux.Statfs{}, syserror.ENOSYS
}
@@ -763,7 +763,7 @@ func (fs *Filesystem) BoundEndpointAt(ctx context.Context, rp *vfs.ResolvingPath
}
// ListxattrAt implements vfs.FilesystemImpl.ListxattrAt.
-func (fs *Filesystem) ListxattrAt(ctx context.Context, rp *vfs.ResolvingPath) ([]string, error) {
+func (fs *Filesystem) ListxattrAt(ctx context.Context, rp *vfs.ResolvingPath, size uint64) ([]string, error) {
fs.mu.RLock()
_, _, err := fs.walkExistingLocked(ctx, rp)
fs.mu.RUnlock()
@@ -776,7 +776,7 @@ func (fs *Filesystem) ListxattrAt(ctx context.Context, rp *vfs.ResolvingPath) ([
}
// GetxattrAt implements vfs.FilesystemImpl.GetxattrAt.
-func (fs *Filesystem) GetxattrAt(ctx context.Context, rp *vfs.ResolvingPath, name string) (string, error) {
+func (fs *Filesystem) GetxattrAt(ctx context.Context, rp *vfs.ResolvingPath, opts vfs.GetxattrOptions) (string, error) {
fs.mu.RLock()
_, _, err := fs.walkExistingLocked(ctx, rp)
fs.mu.RUnlock()
diff --git a/pkg/sentry/fsimpl/proc/task.go b/pkg/sentry/fsimpl/proc/task.go
index aee2a4392..888afc0fd 100644
--- a/pkg/sentry/fsimpl/proc/task.go
+++ b/pkg/sentry/fsimpl/proc/task.go
@@ -214,22 +214,6 @@ func newIO(t *kernel.Task, isThreadGroup bool) *ioData {
return &ioData{ioUsage: t}
}
-func newNamespaceSymlink(task *kernel.Task, ino uint64, ns string) *kernfs.Dentry {
- // Namespace symlinks should contain the namespace name and the inode number
- // for the namespace instance, so for example user:[123456]. We currently fake
- // the inode number by sticking the symlink inode in its place.
- target := fmt.Sprintf("%s:[%d]", ns, ino)
-
- inode := &kernfs.StaticSymlink{}
- // Note: credentials are overridden by taskOwnedInode.
- inode.Init(task.Credentials(), ino, target)
-
- taskInode := &taskOwnedInode{Inode: inode, owner: task}
- d := &kernfs.Dentry{}
- d.Init(taskInode)
- return d
-}
-
// newCgroupData creates inode that shows cgroup information.
// From man 7 cgroups: "For each cgroup hierarchy of which the process is a
// member, there is one entry containing three colon-separated fields:
diff --git a/pkg/sentry/fsimpl/proc/task_fds.go b/pkg/sentry/fsimpl/proc/task_fds.go
index 9c8656b28..046265eca 100644
--- a/pkg/sentry/fsimpl/proc/task_fds.go
+++ b/pkg/sentry/fsimpl/proc/task_fds.go
@@ -30,34 +30,35 @@ import (
"gvisor.dev/gvisor/pkg/syserror"
)
-type fdDir struct {
- inoGen InoGenerator
- task *kernel.Task
-
- // When produceSymlinks is set, dirents produces for the FDs are reported
- // as symlink. Otherwise, they are reported as regular files.
- produceSymlink bool
-}
-
-func (i *fdDir) lookup(name string) (*vfs.FileDescription, kernel.FDFlags, error) {
- fd, err := strconv.ParseUint(name, 10, 64)
- if err != nil {
- return nil, kernel.FDFlags{}, syserror.ENOENT
- }
-
+func getTaskFD(t *kernel.Task, fd int32) (*vfs.FileDescription, kernel.FDFlags) {
var (
file *vfs.FileDescription
flags kernel.FDFlags
)
- i.task.WithMuLocked(func(t *kernel.Task) {
- if fdTable := t.FDTable(); fdTable != nil {
- file, flags = fdTable.GetVFS2(int32(fd))
+ t.WithMuLocked(func(t *kernel.Task) {
+ if fdt := t.FDTable(); fdt != nil {
+ file, flags = fdt.GetVFS2(fd)
}
})
+ return file, flags
+}
+
+func taskFDExists(t *kernel.Task, fd int32) bool {
+ file, _ := getTaskFD(t, fd)
if file == nil {
- return nil, kernel.FDFlags{}, syserror.ENOENT
+ return false
}
- return file, flags, nil
+ file.DecRef()
+ return true
+}
+
+type fdDir struct {
+ inoGen InoGenerator
+ task *kernel.Task
+
+ // When produceSymlinks is set, dirents produces for the FDs are reported
+ // as symlink. Otherwise, they are reported as regular files.
+ produceSymlink bool
}
// IterDirents implements kernfs.inodeDynamicLookup.
@@ -128,11 +129,15 @@ func newFDDirInode(task *kernel.Task, inoGen InoGenerator) *kernfs.Dentry {
// Lookup implements kernfs.inodeDynamicLookup.
func (i *fdDirInode) Lookup(ctx context.Context, name string) (*vfs.Dentry, error) {
- file, _, err := i.lookup(name)
+ fdInt, err := strconv.ParseInt(name, 10, 32)
if err != nil {
- return nil, err
+ return nil, syserror.ENOENT
+ }
+ fd := int32(fdInt)
+ if !taskFDExists(i.task, fd) {
+ return nil, syserror.ENOENT
}
- taskDentry := newFDSymlink(i.task.Credentials(), file, i.inoGen.NextIno())
+ taskDentry := newFDSymlink(i.task, fd, i.inoGen.NextIno())
return taskDentry.VFSDentry(), nil
}
@@ -169,19 +174,22 @@ func (i *fdDirInode) CheckPermissions(ctx context.Context, creds *auth.Credentia
//
// +stateify savable
type fdSymlink struct {
- refs.AtomicRefCount
kernfs.InodeAttrs
+ kernfs.InodeNoopRefCount
kernfs.InodeSymlink
- file *vfs.FileDescription
+ task *kernel.Task
+ fd int32
}
var _ kernfs.Inode = (*fdSymlink)(nil)
-func newFDSymlink(creds *auth.Credentials, file *vfs.FileDescription, ino uint64) *kernfs.Dentry {
- file.IncRef()
- inode := &fdSymlink{file: file}
- inode.Init(creds, ino, linux.ModeSymlink|0777)
+func newFDSymlink(task *kernel.Task, fd int32, ino uint64) *kernfs.Dentry {
+ inode := &fdSymlink{
+ task: task,
+ fd: fd,
+ }
+ inode.Init(task.Credentials(), ino, linux.ModeSymlink|0777)
d := &kernfs.Dentry{}
d.Init(inode)
@@ -189,29 +197,27 @@ func newFDSymlink(creds *auth.Credentials, file *vfs.FileDescription, ino uint64
}
func (s *fdSymlink) Readlink(ctx context.Context) (string, error) {
+ file, _ := getTaskFD(s.task, s.fd)
+ if file == nil {
+ return "", syserror.ENOENT
+ }
+ defer file.DecRef()
root := vfs.RootFromContext(ctx)
defer root.DecRef()
-
- vfsObj := s.file.VirtualDentry().Mount().Filesystem().VirtualFilesystem()
- return vfsObj.PathnameWithDeleted(ctx, root, s.file.VirtualDentry())
+ return s.task.Kernel().VFS().PathnameWithDeleted(ctx, root, file.VirtualDentry())
}
func (s *fdSymlink) Getlink(ctx context.Context) (vfs.VirtualDentry, string, error) {
- vd := s.file.VirtualDentry()
+ file, _ := getTaskFD(s.task, s.fd)
+ if file == nil {
+ return vfs.VirtualDentry{}, "", syserror.ENOENT
+ }
+ defer file.DecRef()
+ vd := file.VirtualDentry()
vd.IncRef()
return vd, "", nil
}
-func (s *fdSymlink) DecRef() {
- s.AtomicRefCount.DecRefWithDestructor(func() {
- s.Destroy()
- })
-}
-
-func (s *fdSymlink) Destroy() {
- s.file.DecRef()
-}
-
// fdInfoDirInode represents the inode for /proc/[pid]/fdinfo directory.
//
// +stateify savable
@@ -244,12 +250,18 @@ func newFDInfoDirInode(task *kernel.Task, inoGen InoGenerator) *kernfs.Dentry {
// Lookup implements kernfs.inodeDynamicLookup.
func (i *fdInfoDirInode) Lookup(ctx context.Context, name string) (*vfs.Dentry, error) {
- file, flags, err := i.lookup(name)
+ fdInt, err := strconv.ParseInt(name, 10, 32)
if err != nil {
- return nil, err
+ return nil, syserror.ENOENT
+ }
+ fd := int32(fdInt)
+ if !taskFDExists(i.task, fd) {
+ return nil, syserror.ENOENT
+ }
+ data := &fdInfoData{
+ task: i.task,
+ fd: fd,
}
-
- data := &fdInfoData{file: file, flags: flags}
dentry := newTaskOwnedFile(i.task, i.inoGen.NextIno(), 0444, data)
return dentry.VFSDentry(), nil
}
@@ -268,26 +280,23 @@ type fdInfoData struct {
kernfs.DynamicBytesFile
refs.AtomicRefCount
- file *vfs.FileDescription
- flags kernel.FDFlags
+ task *kernel.Task
+ fd int32
}
var _ dynamicInode = (*fdInfoData)(nil)
-func (d *fdInfoData) DecRef() {
- d.AtomicRefCount.DecRefWithDestructor(d.destroy)
-}
-
-func (d *fdInfoData) destroy() {
- d.file.DecRef()
-}
-
// Generate implements vfs.DynamicBytesSource.Generate.
func (d *fdInfoData) Generate(ctx context.Context, buf *bytes.Buffer) error {
+ file, descriptorFlags := getTaskFD(d.task, d.fd)
+ if file == nil {
+ return syserror.ENOENT
+ }
+ defer file.DecRef()
// TODO(b/121266871): Include pos, locks, and other data. For now we only
// have flags.
// See https://www.kernel.org/doc/Documentation/filesystems/proc.txt
- flags := uint(d.file.StatusFlags()) | d.flags.ToLinuxFileFlags()
+ flags := uint(file.StatusFlags()) | descriptorFlags.ToLinuxFileFlags()
fmt.Fprintf(buf, "flags:\t0%o\n", flags)
return nil
}
diff --git a/pkg/sentry/fsimpl/proc/task_files.go b/pkg/sentry/fsimpl/proc/task_files.go
index 88ea6a6d8..2c6f8bdfc 100644
--- a/pkg/sentry/fsimpl/proc/task_files.go
+++ b/pkg/sentry/fsimpl/proc/task_files.go
@@ -64,6 +64,16 @@ func getMMIncRef(task *kernel.Task) (*mm.MemoryManager, error) {
return m, nil
}
+func checkTaskState(t *kernel.Task) error {
+ switch t.ExitState() {
+ case kernel.TaskExitZombie:
+ return syserror.EACCES
+ case kernel.TaskExitDead:
+ return syserror.ESRCH
+ }
+ return nil
+}
+
type bufferWriter struct {
buf *bytes.Buffer
}
@@ -628,11 +638,13 @@ func (s *exeSymlink) Getlink(ctx context.Context) (vfs.VirtualDentry, string, er
}
func (s *exeSymlink) executable() (file fsbridge.File, err error) {
+ if err := checkTaskState(s.task); err != nil {
+ return nil, err
+ }
+
s.task.WithMuLocked(func(t *kernel.Task) {
mm := t.MemoryManager()
if mm == nil {
- // TODO(b/34851096): Check shouldn't allow Readlink once the
- // Task is zombied.
err = syserror.EACCES
return
}
@@ -642,7 +654,7 @@ func (s *exeSymlink) executable() (file fsbridge.File, err error) {
// (with locks held).
file = mm.Executable()
if file == nil {
- err = syserror.ENOENT
+ err = syserror.ESRCH
}
})
return
@@ -709,3 +721,41 @@ func (i *mountsData) Generate(ctx context.Context, buf *bytes.Buffer) error {
i.task.Kernel().VFS().GenerateProcMounts(ctx, rootDir, buf)
return nil
}
+
+type namespaceSymlink struct {
+ kernfs.StaticSymlink
+
+ task *kernel.Task
+}
+
+func newNamespaceSymlink(task *kernel.Task, ino uint64, ns string) *kernfs.Dentry {
+ // Namespace symlinks should contain the namespace name and the inode number
+ // for the namespace instance, so for example user:[123456]. We currently fake
+ // the inode number by sticking the symlink inode in its place.
+ target := fmt.Sprintf("%s:[%d]", ns, ino)
+
+ inode := &namespaceSymlink{task: task}
+ // Note: credentials are overridden by taskOwnedInode.
+ inode.Init(task.Credentials(), ino, target)
+
+ taskInode := &taskOwnedInode{Inode: inode, owner: task}
+ d := &kernfs.Dentry{}
+ d.Init(taskInode)
+ return d
+}
+
+// Readlink implements Inode.
+func (s *namespaceSymlink) Readlink(ctx context.Context) (string, error) {
+ if err := checkTaskState(s.task); err != nil {
+ return "", err
+ }
+ return s.StaticSymlink.Readlink(ctx)
+}
+
+// Getlink implements Inode.Getlink.
+func (s *namespaceSymlink) Getlink(ctx context.Context) (vfs.VirtualDentry, string, error) {
+ if err := checkTaskState(s.task); err != nil {
+ return vfs.VirtualDentry{}, "", err
+ }
+ return s.StaticSymlink.Getlink(ctx)
+}
diff --git a/pkg/sentry/fsimpl/proc/task_net.go b/pkg/sentry/fsimpl/proc/task_net.go
index 6b2a77328..6595fcee6 100644
--- a/pkg/sentry/fsimpl/proc/task_net.go
+++ b/pkg/sentry/fsimpl/proc/task_net.go
@@ -688,9 +688,9 @@ func (d *netSnmpData) Generate(ctx context.Context, buf *bytes.Buffer) error {
if line.prefix == "Tcp" {
tcp := stat.(*inet.StatSNMPTCP)
// "Tcp" needs special processing because MaxConn is signed. RFC 2012.
- fmt.Sprintf("%s: %s %d %s\n", line.prefix, sprintSlice(tcp[:3]), int64(tcp[3]), sprintSlice(tcp[4:]))
+ fmt.Fprintf(buf, "%s: %s %d %s\n", line.prefix, sprintSlice(tcp[:3]), int64(tcp[3]), sprintSlice(tcp[4:]))
} else {
- fmt.Sprintf("%s: %s\n", line.prefix, sprintSlice(toSlice(stat)))
+ fmt.Fprintf(buf, "%s: %s\n", line.prefix, sprintSlice(toSlice(stat)))
}
}
return nil
diff --git a/pkg/sentry/fsimpl/tmpfs/BUILD b/pkg/sentry/fsimpl/tmpfs/BUILD
index f2ac23c88..4e6cd3491 100644
--- a/pkg/sentry/fsimpl/tmpfs/BUILD
+++ b/pkg/sentry/fsimpl/tmpfs/BUILD
@@ -51,6 +51,7 @@ go_library(
"//pkg/sentry/usage",
"//pkg/sentry/vfs",
"//pkg/sentry/vfs/lock",
+ "//pkg/sentry/vfs/memxattr",
"//pkg/sync",
"//pkg/syserror",
"//pkg/usermem",
diff --git a/pkg/sentry/fsimpl/tmpfs/filesystem.go b/pkg/sentry/fsimpl/tmpfs/filesystem.go
index 5339d7072..f4d50d64f 100644
--- a/pkg/sentry/fsimpl/tmpfs/filesystem.go
+++ b/pkg/sentry/fsimpl/tmpfs/filesystem.go
@@ -696,51 +696,47 @@ func (fs *filesystem) BoundEndpointAt(ctx context.Context, rp *vfs.ResolvingPath
}
// ListxattrAt implements vfs.FilesystemImpl.ListxattrAt.
-func (fs *filesystem) ListxattrAt(ctx context.Context, rp *vfs.ResolvingPath) ([]string, error) {
+func (fs *filesystem) ListxattrAt(ctx context.Context, rp *vfs.ResolvingPath, size uint64) ([]string, error) {
fs.mu.RLock()
defer fs.mu.RUnlock()
- _, err := resolveLocked(rp)
+ d, err := resolveLocked(rp)
if err != nil {
return nil, err
}
- // TODO(b/127675828): support extended attributes
- return nil, syserror.ENOTSUP
+ return d.inode.listxattr(size)
}
// GetxattrAt implements vfs.FilesystemImpl.GetxattrAt.
-func (fs *filesystem) GetxattrAt(ctx context.Context, rp *vfs.ResolvingPath, name string) (string, error) {
+func (fs *filesystem) GetxattrAt(ctx context.Context, rp *vfs.ResolvingPath, opts vfs.GetxattrOptions) (string, error) {
fs.mu.RLock()
defer fs.mu.RUnlock()
- _, err := resolveLocked(rp)
+ d, err := resolveLocked(rp)
if err != nil {
return "", err
}
- // TODO(b/127675828): support extended attributes
- return "", syserror.ENOTSUP
+ return d.inode.getxattr(rp.Credentials(), &opts)
}
// SetxattrAt implements vfs.FilesystemImpl.SetxattrAt.
func (fs *filesystem) SetxattrAt(ctx context.Context, rp *vfs.ResolvingPath, opts vfs.SetxattrOptions) error {
fs.mu.RLock()
defer fs.mu.RUnlock()
- _, err := resolveLocked(rp)
+ d, err := resolveLocked(rp)
if err != nil {
return err
}
- // TODO(b/127675828): support extended attributes
- return syserror.ENOTSUP
+ return d.inode.setxattr(rp.Credentials(), &opts)
}
// RemovexattrAt implements vfs.FilesystemImpl.RemovexattrAt.
func (fs *filesystem) RemovexattrAt(ctx context.Context, rp *vfs.ResolvingPath, name string) error {
fs.mu.RLock()
defer fs.mu.RUnlock()
- _, err := resolveLocked(rp)
+ d, err := resolveLocked(rp)
if err != nil {
return err
}
- // TODO(b/127675828): support extended attributes
- return syserror.ENOTSUP
+ return d.inode.removexattr(rp.Credentials(), name)
}
// PrependPath implements vfs.FilesystemImpl.PrependPath.
diff --git a/pkg/sentry/fsimpl/tmpfs/stat_test.go b/pkg/sentry/fsimpl/tmpfs/stat_test.go
index 3e02e7190..d4f59ee5b 100644
--- a/pkg/sentry/fsimpl/tmpfs/stat_test.go
+++ b/pkg/sentry/fsimpl/tmpfs/stat_test.go
@@ -140,7 +140,7 @@ func TestSetStatAtime(t *testing.T) {
Mask: 0,
Atime: linux.NsecToStatxTimestamp(100),
}}); err != nil {
- t.Errorf("SetStat atime without mask failed: %v")
+ t.Errorf("SetStat atime without mask failed: %v", err)
}
// Atime should be unchanged.
if gotStat, err := fd.Stat(ctx, allStatOptions); err != nil {
@@ -155,7 +155,7 @@ func TestSetStatAtime(t *testing.T) {
Atime: linux.NsecToStatxTimestamp(100),
}
if err := fd.SetStat(ctx, vfs.SetStatOptions{Stat: setStat}); err != nil {
- t.Errorf("SetStat atime with mask failed: %v")
+ t.Errorf("SetStat atime with mask failed: %v", err)
}
if gotStat, err := fd.Stat(ctx, allStatOptions); err != nil {
t.Errorf("Stat got error: %v", err)
@@ -205,7 +205,7 @@ func TestSetStat(t *testing.T) {
Mask: 0,
Atime: linux.NsecToStatxTimestamp(100),
}}); err != nil {
- t.Errorf("SetStat atime without mask failed: %v")
+ t.Errorf("SetStat atime without mask failed: %v", err)
}
// Atime should be unchanged.
if gotStat, err := fd.Stat(ctx, allStatOptions); err != nil {
@@ -220,7 +220,7 @@ func TestSetStat(t *testing.T) {
Atime: linux.NsecToStatxTimestamp(100),
}
if err := fd.SetStat(ctx, vfs.SetStatOptions{Stat: setStat}); err != nil {
- t.Errorf("SetStat atime with mask failed: %v")
+ t.Errorf("SetStat atime with mask failed: %v", err)
}
if gotStat, err := fd.Stat(ctx, allStatOptions); err != nil {
t.Errorf("Stat got error: %v", err)
diff --git a/pkg/sentry/fsimpl/tmpfs/tmpfs.go b/pkg/sentry/fsimpl/tmpfs/tmpfs.go
index 654e788e3..9fa8637d5 100644
--- a/pkg/sentry/fsimpl/tmpfs/tmpfs.go
+++ b/pkg/sentry/fsimpl/tmpfs/tmpfs.go
@@ -27,6 +27,7 @@ package tmpfs
import (
"fmt"
"math"
+ "strings"
"sync/atomic"
"gvisor.dev/gvisor/pkg/abi/linux"
@@ -37,6 +38,7 @@ import (
"gvisor.dev/gvisor/pkg/sentry/pgalloc"
"gvisor.dev/gvisor/pkg/sentry/vfs"
"gvisor.dev/gvisor/pkg/sentry/vfs/lock"
+ "gvisor.dev/gvisor/pkg/sentry/vfs/memxattr"
"gvisor.dev/gvisor/pkg/sync"
"gvisor.dev/gvisor/pkg/syserror"
)
@@ -186,6 +188,11 @@ type inode struct {
// filesystem.RmdirAt() drops the reference.
refs int64
+ // xattrs implements extended attributes.
+ //
+ // TODO(b/148380782): Support xattrs other than user.*
+ xattrs memxattr.SimpleExtendedAttributes
+
// Inode metadata. Writing multiple fields atomically requires holding
// mu, othewise atomic operations can be used.
mu sync.Mutex
@@ -535,6 +542,56 @@ func (i *inode) touchCMtimeLocked() {
atomic.StoreInt64(&i.ctime, now)
}
+func (i *inode) listxattr(size uint64) ([]string, error) {
+ return i.xattrs.Listxattr(size)
+}
+
+func (i *inode) getxattr(creds *auth.Credentials, opts *vfs.GetxattrOptions) (string, error) {
+ if err := i.checkPermissions(creds, vfs.MayRead); err != nil {
+ return "", err
+ }
+ if !strings.HasPrefix(opts.Name, linux.XATTR_USER_PREFIX) {
+ return "", syserror.EOPNOTSUPP
+ }
+ if !i.userXattrSupported() {
+ return "", syserror.ENODATA
+ }
+ return i.xattrs.Getxattr(opts)
+}
+
+func (i *inode) setxattr(creds *auth.Credentials, opts *vfs.SetxattrOptions) error {
+ if err := i.checkPermissions(creds, vfs.MayWrite); err != nil {
+ return err
+ }
+ if !strings.HasPrefix(opts.Name, linux.XATTR_USER_PREFIX) {
+ return syserror.EOPNOTSUPP
+ }
+ if !i.userXattrSupported() {
+ return syserror.EPERM
+ }
+ return i.xattrs.Setxattr(opts)
+}
+
+func (i *inode) removexattr(creds *auth.Credentials, name string) error {
+ if err := i.checkPermissions(creds, vfs.MayWrite); err != nil {
+ return err
+ }
+ if !strings.HasPrefix(name, linux.XATTR_USER_PREFIX) {
+ return syserror.EOPNOTSUPP
+ }
+ if !i.userXattrSupported() {
+ return syserror.EPERM
+ }
+ return i.xattrs.Removexattr(name)
+}
+
+// Extended attributes in the user.* namespace are only supported for regular
+// files and directories.
+func (i *inode) userXattrSupported() bool {
+ filetype := linux.S_IFMT & atomic.LoadUint32(&i.mode)
+ return filetype == linux.S_IFREG || filetype == linux.S_IFDIR
+}
+
// fileDescription is embedded by tmpfs implementations of
// vfs.FileDescriptionImpl.
type fileDescription struct {
@@ -562,3 +619,23 @@ func (fd *fileDescription) SetStat(ctx context.Context, opts vfs.SetStatOptions)
creds := auth.CredentialsFromContext(ctx)
return fd.inode().setStat(ctx, creds, &opts.Stat)
}
+
+// Listxattr implements vfs.FileDescriptionImpl.Listxattr.
+func (fd *fileDescription) Listxattr(ctx context.Context, size uint64) ([]string, error) {
+ return fd.inode().listxattr(size)
+}
+
+// Getxattr implements vfs.FileDescriptionImpl.Getxattr.
+func (fd *fileDescription) Getxattr(ctx context.Context, opts vfs.GetxattrOptions) (string, error) {
+ return fd.inode().getxattr(auth.CredentialsFromContext(ctx), &opts)
+}
+
+// Setxattr implements vfs.FileDescriptionImpl.Setxattr.
+func (fd *fileDescription) Setxattr(ctx context.Context, opts vfs.SetxattrOptions) error {
+ return fd.inode().setxattr(auth.CredentialsFromContext(ctx), &opts)
+}
+
+// Removexattr implements vfs.FileDescriptionImpl.Removexattr.
+func (fd *fileDescription) Removexattr(ctx context.Context, name string) error {
+ return fd.inode().removexattr(auth.CredentialsFromContext(ctx), name)
+}
diff --git a/pkg/sentry/kernel/fd_table.go b/pkg/sentry/kernel/fd_table.go
index d09d97825..ed40b5303 100644
--- a/pkg/sentry/kernel/fd_table.go
+++ b/pkg/sentry/kernel/fd_table.go
@@ -307,6 +307,61 @@ func (f *FDTable) NewFDs(ctx context.Context, fd int32, files []*fs.File, flags
return fds, nil
}
+// NewFDsVFS2 allocates new FDs guaranteed to be the lowest number available
+// greater than or equal to the fd parameter. All files will share the set
+// flags. Success is guaranteed to be all or none.
+func (f *FDTable) NewFDsVFS2(ctx context.Context, fd int32, files []*vfs.FileDescription, flags FDFlags) (fds []int32, err error) {
+ if fd < 0 {
+ // Don't accept negative FDs.
+ return nil, syscall.EINVAL
+ }
+
+ // Default limit.
+ end := int32(math.MaxInt32)
+
+ // Ensure we don't get past the provided limit.
+ if limitSet := limits.FromContext(ctx); limitSet != nil {
+ lim := limitSet.Get(limits.NumberOfFiles)
+ if lim.Cur != limits.Infinity {
+ end = int32(lim.Cur)
+ }
+ if fd >= end {
+ return nil, syscall.EMFILE
+ }
+ }
+
+ f.mu.Lock()
+ defer f.mu.Unlock()
+
+ // From f.next to find available fd.
+ if fd < f.next {
+ fd = f.next
+ }
+
+ // Install all entries.
+ for i := fd; i < end && len(fds) < len(files); i++ {
+ if d, _, _ := f.getVFS2(i); d == nil {
+ f.setVFS2(i, files[len(fds)], flags) // Set the descriptor.
+ fds = append(fds, i) // Record the file descriptor.
+ }
+ }
+
+ // Failure? Unwind existing FDs.
+ if len(fds) < len(files) {
+ for _, i := range fds {
+ f.setVFS2(i, nil, FDFlags{}) // Zap entry.
+ }
+ return nil, syscall.EMFILE
+ }
+
+ if fd == f.next {
+ // Update next search start position.
+ f.next = fds[len(fds)-1] + 1
+ }
+
+ return fds, nil
+}
+
// NewFDVFS2 allocates a file descriptor greater than or equal to minfd for
// the given file description. If it succeeds, it takes a reference on file.
func (f *FDTable) NewFDVFS2(ctx context.Context, minfd int32, file *vfs.FileDescription, flags FDFlags) (int32, error) {
diff --git a/pkg/sentry/kernel/ptrace.go b/pkg/sentry/kernel/ptrace.go
index 35ad97d5d..e23e796ef 100644
--- a/pkg/sentry/kernel/ptrace.go
+++ b/pkg/sentry/kernel/ptrace.go
@@ -184,7 +184,6 @@ func (t *Task) CanTrace(target *Task, attach bool) bool {
if targetCreds.PermittedCaps&^callerCreds.PermittedCaps != 0 {
return false
}
- // TODO: Yama LSM
return true
}
diff --git a/pkg/sentry/kernel/shm/shm.go b/pkg/sentry/kernel/shm/shm.go
index 208569057..f66cfcc7f 100644
--- a/pkg/sentry/kernel/shm/shm.go
+++ b/pkg/sentry/kernel/shm/shm.go
@@ -461,7 +461,7 @@ func (s *Shm) AddMapping(ctx context.Context, _ memmap.MappingSpace, _ usermem.A
func (s *Shm) RemoveMapping(ctx context.Context, _ memmap.MappingSpace, _ usermem.AddrRange, _ uint64, _ bool) {
s.mu.Lock()
defer s.mu.Unlock()
- // TODO(b/38173783): RemoveMapping may be called during task exit, when ctx
+ // RemoveMapping may be called during task exit, when ctx
// is context.Background. Gracefully handle missing clocks. Failing to
// update the detach time in these cases is ok, since no one can observe the
// omission.
diff --git a/pkg/sentry/kernel/syscalls.go b/pkg/sentry/kernel/syscalls.go
index 93c4fe969..2e3565747 100644
--- a/pkg/sentry/kernel/syscalls.go
+++ b/pkg/sentry/kernel/syscalls.go
@@ -209,65 +209,61 @@ type Stracer interface {
// SyscallEnter is called on syscall entry.
//
// The returned private data is passed to SyscallExit.
- //
- // TODO(gvisor.dev/issue/155): remove kernel imports from the strace
- // package so that the type can be used directly.
SyscallEnter(t *Task, sysno uintptr, args arch.SyscallArguments, flags uint32) interface{}
// SyscallExit is called on syscall exit.
SyscallExit(context interface{}, t *Task, sysno, rval uintptr, err error)
}
-// SyscallTable is a lookup table of system calls. Critically, a SyscallTable
-// is *immutable*. In order to make supporting suspend and resume sane, they
-// must be uniquely registered and may not change during operation.
+// SyscallTable is a lookup table of system calls.
//
-// +stateify savable
+// Note that a SyscallTable is not savable directly. Instead, they are saved as
+// an OS/Arch pair and lookup happens again on restore.
type SyscallTable struct {
// OS is the operating system that this syscall table implements.
- OS abi.OS `state:"wait"`
+ OS abi.OS
// Arch is the architecture that this syscall table targets.
- Arch arch.Arch `state:"wait"`
+ Arch arch.Arch
// The OS version that this syscall table implements.
- Version Version `state:"manual"`
+ Version Version
// AuditNumber is a numeric constant that represents the syscall table. If
// non-zero, auditNumber must be one of the AUDIT_ARCH_* values defined by
// linux/audit.h.
- AuditNumber uint32 `state:"manual"`
+ AuditNumber uint32
// Table is the collection of functions.
- Table map[uintptr]Syscall `state:"manual"`
+ Table map[uintptr]Syscall
// lookup is a fixed-size array that holds the syscalls (indexed by
// their numbers). It is used for fast look ups.
- lookup []SyscallFn `state:"manual"`
+ lookup []SyscallFn
// Emulate is a collection of instruction addresses to emulate. The
// keys are addresses, and the values are system call numbers.
- Emulate map[usermem.Addr]uintptr `state:"manual"`
+ Emulate map[usermem.Addr]uintptr
// The function to call in case of a missing system call.
- Missing MissingFn `state:"manual"`
+ Missing MissingFn
// Stracer traces this syscall table.
- Stracer Stracer `state:"manual"`
+ Stracer Stracer
// External is used to handle an external callback.
- External func(*Kernel) `state:"manual"`
+ External func(*Kernel)
// ExternalFilterBefore is called before External is called before the syscall is executed.
// External is not called if it returns false.
- ExternalFilterBefore func(*Task, uintptr, arch.SyscallArguments) bool `state:"manual"`
+ ExternalFilterBefore func(*Task, uintptr, arch.SyscallArguments) bool
// ExternalFilterAfter is called before External is called after the syscall is executed.
// External is not called if it returns false.
- ExternalFilterAfter func(*Task, uintptr, arch.SyscallArguments) bool `state:"manual"`
+ ExternalFilterAfter func(*Task, uintptr, arch.SyscallArguments) bool
// FeatureEnable stores the strace and one-shot enable bits.
- FeatureEnable SyscallFlagsTable `state:"manual"`
+ FeatureEnable SyscallFlagsTable
}
// allSyscallTables contains all known tables.
diff --git a/pkg/sentry/kernel/syscalls_state.go b/pkg/sentry/kernel/syscalls_state.go
index 00358326b..90f890495 100644
--- a/pkg/sentry/kernel/syscalls_state.go
+++ b/pkg/sentry/kernel/syscalls_state.go
@@ -14,16 +14,34 @@
package kernel
-import "fmt"
+import (
+ "fmt"
-// afterLoad is invoked by stateify.
-func (s *SyscallTable) afterLoad() {
- otherTable, ok := LookupSyscallTable(s.OS, s.Arch)
- if !ok {
- // Couldn't find a reference?
- panic(fmt.Sprintf("syscall table not found for OS %v Arch %v", s.OS, s.Arch))
+ "gvisor.dev/gvisor/pkg/abi"
+ "gvisor.dev/gvisor/pkg/sentry/arch"
+)
+
+// syscallTableInfo is used to reload the SyscallTable.
+//
+// +stateify savable
+type syscallTableInfo struct {
+ OS abi.OS
+ Arch arch.Arch
+}
+
+// saveSt saves the SyscallTable.
+func (tc *TaskContext) saveSt() syscallTableInfo {
+ return syscallTableInfo{
+ OS: tc.st.OS,
+ Arch: tc.st.Arch,
}
+}
- // Copy the table.
- *s = *otherTable
+// loadSt loads the SyscallTable.
+func (tc *TaskContext) loadSt(sti syscallTableInfo) {
+ st, ok := LookupSyscallTable(sti.OS, sti.Arch)
+ if !ok {
+ panic(fmt.Sprintf("syscall table not found for OS %v, Arch %v", sti.OS, sti.Arch))
+ }
+ tc.st = st // Save the table reference.
}
diff --git a/pkg/sentry/kernel/task.go b/pkg/sentry/kernel/task.go
index d6546735e..e5d133d6c 100644
--- a/pkg/sentry/kernel/task.go
+++ b/pkg/sentry/kernel/task.go
@@ -777,6 +777,15 @@ func (t *Task) NewFDs(fd int32, files []*fs.File, flags FDFlags) ([]int32, error
return t.fdTable.NewFDs(t, fd, files, flags)
}
+// NewFDsVFS2 is a convenience wrapper for t.FDTable().NewFDsVFS2.
+//
+// This automatically passes the task as the context.
+//
+// Precondition: same as FDTable.
+func (t *Task) NewFDsVFS2(fd int32, files []*vfs.FileDescription, flags FDFlags) ([]int32, error) {
+ return t.fdTable.NewFDsVFS2(t, fd, files, flags)
+}
+
// NewFDFrom is a convenience wrapper for t.FDTable().NewFDs with a single file.
//
// This automatically passes the task as the context.
diff --git a/pkg/sentry/kernel/task_context.go b/pkg/sentry/kernel/task_context.go
index 0158b1788..9fa528384 100644
--- a/pkg/sentry/kernel/task_context.go
+++ b/pkg/sentry/kernel/task_context.go
@@ -49,7 +49,7 @@ type TaskContext struct {
fu *futex.Manager
// st is the task's syscall table.
- st *SyscallTable
+ st *SyscallTable `state:".(syscallTableInfo)"`
}
// release releases all resources held by the TaskContext. release is called by
@@ -58,7 +58,6 @@ func (tc *TaskContext) release() {
// Nil out pointers so that if the task is saved after release, it doesn't
// follow the pointers to possibly now-invalid objects.
if tc.MemoryManager != nil {
- // TODO(b/38173783)
tc.MemoryManager.DecUsers(context.Background())
tc.MemoryManager = nil
}
diff --git a/pkg/sentry/kernel/task_identity.go b/pkg/sentry/kernel/task_identity.go
index ce3e6ef28..0325967e4 100644
--- a/pkg/sentry/kernel/task_identity.go
+++ b/pkg/sentry/kernel/task_identity.go
@@ -455,7 +455,7 @@ func (t *Task) SetKeepCaps(k bool) {
t.creds.Store(creds)
}
-// updateCredsForExec updates t.creds to reflect an execve().
+// updateCredsForExecLocked updates t.creds to reflect an execve().
//
// NOTE(b/30815691): We currently do not implement privileged executables
// (set-user/group-ID bits and file capabilities). This allows us to make a lot
diff --git a/pkg/sentry/kernel/task_signals.go b/pkg/sentry/kernel/task_signals.go
index 8802db142..6aa798346 100644
--- a/pkg/sentry/kernel/task_signals.go
+++ b/pkg/sentry/kernel/task_signals.go
@@ -513,8 +513,6 @@ func (t *Task) canReceiveSignalLocked(sig linux.Signal) bool {
if t.stop != nil {
return false
}
- // - TODO(b/38173783): No special case for when t is also the sending task,
- // because the identity of the sender is unknown.
// - Do not choose tasks that have already been interrupted, as they may be
// busy handling another signal.
if len(t.interruptChan) != 0 {
diff --git a/pkg/sentry/kernel/task_syscall.go b/pkg/sentry/kernel/task_syscall.go
index d555d69a8..3d7a734ef 100644
--- a/pkg/sentry/kernel/task_syscall.go
+++ b/pkg/sentry/kernel/task_syscall.go
@@ -194,6 +194,19 @@ func (t *Task) executeSyscall(sysno uintptr, args arch.SyscallArguments) (rval u
//
// The syscall path is very hot; avoid defer.
func (t *Task) doSyscall() taskRunState {
+ // Save value of the register which is clobbered in the following
+ // t.Arch().SetReturn(-ENOSYS) operation. This is dedicated to arm64.
+ //
+ // On x86, register rax was shared by syscall number and return
+ // value, and at the entry of the syscall handler, the rax was
+ // saved to regs.orig_rax which was exposed to user space.
+ // But on arm64, syscall number was passed through X8, and the X0
+ // was shared by the first syscall argument and return value. The
+ // X0 was saved to regs.orig_x0 which was not exposed to user space.
+ // So we have to do the same operation here to save the X0 value
+ // into the task context.
+ t.Arch().SyscallSaveOrig()
+
sysno := t.Arch().SyscallNo()
args := t.Arch().SyscallArgs()
@@ -269,6 +282,7 @@ func (*runSyscallAfterSyscallEnterStop) execute(t *Task) taskRunState {
return (*runSyscallExit)(nil)
}
args := t.Arch().SyscallArgs()
+
return t.doSyscallInvoke(sysno, args)
}
diff --git a/pkg/sentry/kernel/time/time.go b/pkg/sentry/kernel/time/time.go
index 706de83ef..e959700f2 100644
--- a/pkg/sentry/kernel/time/time.go
+++ b/pkg/sentry/kernel/time/time.go
@@ -245,7 +245,7 @@ type Clock interface {
type WallRateClock struct{}
// WallTimeUntil implements Clock.WallTimeUntil.
-func (WallRateClock) WallTimeUntil(t, now Time) time.Duration {
+func (*WallRateClock) WallTimeUntil(t, now Time) time.Duration {
return t.Sub(now)
}
@@ -254,16 +254,16 @@ func (WallRateClock) WallTimeUntil(t, now Time) time.Duration {
type NoClockEvents struct{}
// Readiness implements waiter.Waitable.Readiness.
-func (NoClockEvents) Readiness(mask waiter.EventMask) waiter.EventMask {
+func (*NoClockEvents) Readiness(mask waiter.EventMask) waiter.EventMask {
return 0
}
// EventRegister implements waiter.Waitable.EventRegister.
-func (NoClockEvents) EventRegister(e *waiter.Entry, mask waiter.EventMask) {
+func (*NoClockEvents) EventRegister(e *waiter.Entry, mask waiter.EventMask) {
}
// EventUnregister implements waiter.Waitable.EventUnregister.
-func (NoClockEvents) EventUnregister(e *waiter.Entry) {
+func (*NoClockEvents) EventUnregister(e *waiter.Entry) {
}
// ClockEventsQueue implements waiter.Waitable by wrapping waiter.Queue and
@@ -273,7 +273,7 @@ type ClockEventsQueue struct {
}
// Readiness implements waiter.Waitable.Readiness.
-func (ClockEventsQueue) Readiness(mask waiter.EventMask) waiter.EventMask {
+func (*ClockEventsQueue) Readiness(mask waiter.EventMask) waiter.EventMask {
return 0
}
diff --git a/pkg/sentry/mm/address_space.go b/pkg/sentry/mm/address_space.go
index 0332fc71c..5c667117c 100644
--- a/pkg/sentry/mm/address_space.go
+++ b/pkg/sentry/mm/address_space.go
@@ -201,8 +201,10 @@ func (mm *MemoryManager) mapASLocked(pseg pmaIterator, ar usermem.AddrRange, pre
if pma.needCOW {
perms.Write = false
}
- if err := mm.as.MapFile(pmaMapAR.Start, pma.file, pseg.fileRangeOf(pmaMapAR), perms, precommit); err != nil {
- return err
+ if perms.Any() { // MapFile precondition
+ if err := mm.as.MapFile(pmaMapAR.Start, pma.file, pseg.fileRangeOf(pmaMapAR), perms, precommit); err != nil {
+ return err
+ }
}
pseg = pseg.NextSegment()
}
diff --git a/pkg/sentry/mm/aio_context.go b/pkg/sentry/mm/aio_context.go
index cb29d94b0..379148903 100644
--- a/pkg/sentry/mm/aio_context.go
+++ b/pkg/sentry/mm/aio_context.go
@@ -59,25 +59,27 @@ func (a *aioManager) newAIOContext(events uint32, id uint64) bool {
}
a.contexts[id] = &AIOContext{
- done: make(chan struct{}, 1),
+ requestReady: make(chan struct{}, 1),
maxOutstanding: events,
}
return true
}
-// destroyAIOContext destroys an asynchronous I/O context.
+// destroyAIOContext destroys an asynchronous I/O context. It doesn't wait for
+// for pending requests to complete. Returns the destroyed AIOContext so it can
+// be drained.
//
-// False is returned if the context does not exist.
-func (a *aioManager) destroyAIOContext(id uint64) bool {
+// Nil is returned if the context does not exist.
+func (a *aioManager) destroyAIOContext(id uint64) *AIOContext {
a.mu.Lock()
defer a.mu.Unlock()
ctx, ok := a.contexts[id]
if !ok {
- return false
+ return nil
}
delete(a.contexts, id)
ctx.destroy()
- return true
+ return ctx
}
// lookupAIOContext looks up the given context.
@@ -102,8 +104,8 @@ type ioResult struct {
//
// +stateify savable
type AIOContext struct {
- // done is the notification channel used for all requests.
- done chan struct{} `state:"nosave"`
+ // requestReady is the notification channel used for all requests.
+ requestReady chan struct{} `state:"nosave"`
// mu protects below.
mu sync.Mutex `state:"nosave"`
@@ -129,8 +131,14 @@ func (ctx *AIOContext) destroy() {
ctx.mu.Lock()
defer ctx.mu.Unlock()
ctx.dead = true
- if ctx.outstanding == 0 {
- close(ctx.done)
+ ctx.checkForDone()
+}
+
+// Preconditions: ctx.mu must be held by caller.
+func (ctx *AIOContext) checkForDone() {
+ if ctx.dead && ctx.outstanding == 0 {
+ close(ctx.requestReady)
+ ctx.requestReady = nil
}
}
@@ -154,11 +162,12 @@ func (ctx *AIOContext) PopRequest() (interface{}, bool) {
// Is there anything ready?
if e := ctx.results.Front(); e != nil {
- ctx.results.Remove(e)
- ctx.outstanding--
- if ctx.outstanding == 0 && ctx.dead {
- close(ctx.done)
+ if ctx.outstanding == 0 {
+ panic("AIOContext outstanding is going negative")
}
+ ctx.outstanding--
+ ctx.results.Remove(e)
+ ctx.checkForDone()
return e.data, true
}
return nil, false
@@ -172,26 +181,58 @@ func (ctx *AIOContext) FinishRequest(data interface{}) {
// Push to the list and notify opportunistically. The channel notify
// here is guaranteed to be safe because outstanding must be non-zero.
- // The done channel is only closed when outstanding reaches zero.
+ // The requestReady channel is only closed when outstanding reaches zero.
ctx.results.PushBack(&ioResult{data: data})
select {
- case ctx.done <- struct{}{}:
+ case ctx.requestReady <- struct{}{}:
default:
}
}
// WaitChannel returns a channel that is notified when an AIO request is
-// completed.
-//
-// The boolean return value indicates whether or not the context is active.
-func (ctx *AIOContext) WaitChannel() (chan struct{}, bool) {
+// completed. Returns nil if the context is destroyed and there are no more
+// outstanding requests.
+func (ctx *AIOContext) WaitChannel() chan struct{} {
ctx.mu.Lock()
defer ctx.mu.Unlock()
- if ctx.outstanding == 0 && ctx.dead {
- return nil, false
+ return ctx.requestReady
+}
+
+// Dead returns true if the context has been destroyed.
+func (ctx *AIOContext) Dead() bool {
+ ctx.mu.Lock()
+ defer ctx.mu.Unlock()
+ return ctx.dead
+}
+
+// CancelPendingRequest forgets about a request that hasn't yet completed.
+func (ctx *AIOContext) CancelPendingRequest() {
+ ctx.mu.Lock()
+ defer ctx.mu.Unlock()
+
+ if ctx.outstanding == 0 {
+ panic("AIOContext outstanding is going negative")
}
- return ctx.done, true
+ ctx.outstanding--
+ ctx.checkForDone()
+}
+
+// Drain drops all completed requests. Pending requests remain untouched.
+func (ctx *AIOContext) Drain() {
+ ctx.mu.Lock()
+ defer ctx.mu.Unlock()
+
+ if ctx.outstanding == 0 {
+ return
+ }
+ size := uint32(ctx.results.Len())
+ if ctx.outstanding < size {
+ panic("AIOContext outstanding is going negative")
+ }
+ ctx.outstanding -= size
+ ctx.results.Reset()
+ ctx.checkForDone()
}
// aioMappable implements memmap.MappingIdentity and memmap.Mappable for AIO
@@ -332,9 +373,9 @@ func (mm *MemoryManager) NewAIOContext(ctx context.Context, events uint32) (uint
Length: aioRingBufferSize,
MappingIdentity: m,
Mappable: m,
- // TODO(fvoznika): Linux does "do_mmap_pgoff(..., PROT_READ |
- // PROT_WRITE, ...)" in fs/aio.c:aio_setup_ring(); why do we make this
- // mapping read-only?
+ // Linux uses "do_mmap_pgoff(..., PROT_READ | PROT_WRITE, ...)" in
+ // fs/aio.c:aio_setup_ring(). Since we don't implement AIO_RING_MAGIC,
+ // user mode should not write to this page.
Perms: usermem.Read,
MaxPerms: usermem.Read,
})
@@ -349,11 +390,11 @@ func (mm *MemoryManager) NewAIOContext(ctx context.Context, events uint32) (uint
return id, nil
}
-// DestroyAIOContext destroys an asynchronous I/O context. It returns false if
-// the context does not exist.
-func (mm *MemoryManager) DestroyAIOContext(ctx context.Context, id uint64) bool {
+// DestroyAIOContext destroys an asynchronous I/O context. It returns the
+// destroyed context. nil if the context does not exist.
+func (mm *MemoryManager) DestroyAIOContext(ctx context.Context, id uint64) *AIOContext {
if _, ok := mm.LookupAIOContext(ctx, id); !ok {
- return false
+ return nil
}
// Only unmaps after it assured that the address is a valid aio context to
diff --git a/pkg/sentry/mm/aio_context_state.go b/pkg/sentry/mm/aio_context_state.go
index c37fc9f7b..3dabac1af 100644
--- a/pkg/sentry/mm/aio_context_state.go
+++ b/pkg/sentry/mm/aio_context_state.go
@@ -16,5 +16,5 @@ package mm
// afterLoad is invoked by stateify.
func (a *AIOContext) afterLoad() {
- a.done = make(chan struct{}, 1)
+ a.requestReady = make(chan struct{}, 1)
}
diff --git a/pkg/sentry/platform/kvm/kvm_arm64.go b/pkg/sentry/platform/kvm/kvm_arm64.go
index 79045651e..716198712 100644
--- a/pkg/sentry/platform/kvm/kvm_arm64.go
+++ b/pkg/sentry/platform/kvm/kvm_arm64.go
@@ -18,6 +18,8 @@ package kvm
import (
"syscall"
+
+ "gvisor.dev/gvisor/pkg/sentry/platform/ring0"
)
type kvmOneReg struct {
@@ -46,6 +48,6 @@ type userRegs struct {
func updateGlobalOnce(fd int) error {
physicalInit()
err := updateSystemValues(int(fd))
- updateVectorTable()
+ ring0.Init()
return err
}
diff --git a/pkg/sentry/platform/kvm/machine_arm64_unsafe.go b/pkg/sentry/platform/kvm/machine_arm64_unsafe.go
index b531f2f85..3b35858ae 100644
--- a/pkg/sentry/platform/kvm/machine_arm64_unsafe.go
+++ b/pkg/sentry/platform/kvm/machine_arm64_unsafe.go
@@ -48,69 +48,6 @@ func (m *machine) initArchState() error {
return nil
}
-func getPageWithReflect(p uintptr) []byte {
- return (*(*[0xFFFFFF]byte)(unsafe.Pointer(p & ^uintptr(syscall.Getpagesize()-1))))[:syscall.Getpagesize()]
-}
-
-// Work around: move ring0.Vectors() into a specific address with 11-bits alignment.
-//
-// According to the design documentation of Arm64,
-// the start address of exception vector table should be 11-bits aligned.
-// Please see the code in linux kernel as reference: arch/arm64/kernel/entry.S
-// But, we can't align a function's start address to a specific address by using golang.
-// We have raised this question in golang community:
-// https://groups.google.com/forum/m/#!topic/golang-dev/RPj90l5x86I
-// This function will be removed when golang supports this feature.
-//
-// There are 2 jobs were implemented in this function:
-// 1, move the start address of exception vector table into the specific address.
-// 2, modify the offset of each instruction.
-func updateVectorTable() {
- fromLocation := reflect.ValueOf(ring0.Vectors).Pointer()
- offset := fromLocation & (1<<11 - 1)
- if offset != 0 {
- offset = 1<<11 - offset
- }
-
- toLocation := fromLocation + offset
- page := getPageWithReflect(toLocation)
- if err := syscall.Mprotect(page, syscall.PROT_READ|syscall.PROT_WRITE|syscall.PROT_EXEC); err != nil {
- panic(err)
- }
-
- page = getPageWithReflect(toLocation + 4096)
- if err := syscall.Mprotect(page, syscall.PROT_READ|syscall.PROT_WRITE|syscall.PROT_EXEC); err != nil {
- panic(err)
- }
-
- // Move exception-vector-table into the specific address.
- var entry *uint32
- var entryFrom *uint32
- for i := 1; i <= 0x800; i++ {
- entry = (*uint32)(unsafe.Pointer(toLocation + 0x800 - uintptr(i)))
- entryFrom = (*uint32)(unsafe.Pointer(fromLocation + 0x800 - uintptr(i)))
- *entry = *entryFrom
- }
-
- // The offset from the address of each unconditionally branch is changed.
- // We should modify the offset of each instruction.
- nums := []uint32{0x0, 0x80, 0x100, 0x180, 0x200, 0x280, 0x300, 0x380, 0x400, 0x480, 0x500, 0x580, 0x600, 0x680, 0x700, 0x780}
- for _, num := range nums {
- entry = (*uint32)(unsafe.Pointer(toLocation + uintptr(num)))
- *entry = *entry - (uint32)(offset/4)
- }
-
- page = getPageWithReflect(toLocation)
- if err := syscall.Mprotect(page, syscall.PROT_READ|syscall.PROT_EXEC); err != nil {
- panic(err)
- }
-
- page = getPageWithReflect(toLocation + 4096)
- if err := syscall.Mprotect(page, syscall.PROT_READ|syscall.PROT_EXEC); err != nil {
- panic(err)
- }
-}
-
// initArchState initializes architecture-specific state.
func (c *vCPU) initArchState() error {
var (
diff --git a/pkg/sentry/platform/ring0/BUILD b/pkg/sentry/platform/ring0/BUILD
index 934b6fbcd..b69520030 100644
--- a/pkg/sentry/platform/ring0/BUILD
+++ b/pkg/sentry/platform/ring0/BUILD
@@ -72,11 +72,13 @@ go_library(
"lib_amd64.s",
"lib_arm64.go",
"lib_arm64.s",
+ "lib_arm64_unsafe.go",
"ring0.go",
],
visibility = ["//pkg/sentry:internal"],
deps = [
"//pkg/cpuid",
+ "//pkg/safecopy",
"//pkg/sentry/platform/ring0/pagetables",
"//pkg/usermem",
],
diff --git a/pkg/sentry/platform/ring0/lib_arm64.go b/pkg/sentry/platform/ring0/lib_arm64.go
index af075aae4..242b9305c 100644
--- a/pkg/sentry/platform/ring0/lib_arm64.go
+++ b/pkg/sentry/platform/ring0/lib_arm64.go
@@ -37,3 +37,10 @@ func SaveVRegs(*byte)
// LoadVRegs loads V0-V31 registers.
func LoadVRegs(*byte)
+
+// Init sets function pointers based on architectural features.
+//
+// This must be called prior to using ring0.
+func Init() {
+ rewriteVectors()
+}
diff --git a/pkg/sentry/platform/ring0/lib_arm64_unsafe.go b/pkg/sentry/platform/ring0/lib_arm64_unsafe.go
new file mode 100644
index 000000000..c05166fea
--- /dev/null
+++ b/pkg/sentry/platform/ring0/lib_arm64_unsafe.go
@@ -0,0 +1,108 @@
+// 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.
+
+// +build arm64
+
+package ring0
+
+import (
+ "reflect"
+ "syscall"
+ "unsafe"
+
+ "gvisor.dev/gvisor/pkg/safecopy"
+ "gvisor.dev/gvisor/pkg/usermem"
+)
+
+const (
+ nopInstruction = 0xd503201f
+ instSize = unsafe.Sizeof(uint32(0))
+ vectorsRawLen = 0x800
+)
+
+func unsafeSlice(addr uintptr, length int) (slice []uint32) {
+ hdr := (*reflect.SliceHeader)(unsafe.Pointer(&slice))
+ hdr.Data = addr
+ hdr.Len = length / int(instSize)
+ hdr.Cap = length / int(instSize)
+ return slice
+}
+
+// Work around: move ring0.Vectors() into a specific address with 11-bits alignment.
+//
+// According to the design documentation of Arm64,
+// the start address of exception vector table should be 11-bits aligned.
+// Please see the code in linux kernel as reference: arch/arm64/kernel/entry.S
+// But, we can't align a function's start address to a specific address by using golang.
+// We have raised this question in golang community:
+// https://groups.google.com/forum/m/#!topic/golang-dev/RPj90l5x86I
+// This function will be removed when golang supports this feature.
+//
+// There are 2 jobs were implemented in this function:
+// 1, move the start address of exception vector table into the specific address.
+// 2, modify the offset of each instruction.
+func rewriteVectors() {
+ vectorsBegin := reflect.ValueOf(Vectors).Pointer()
+
+ // The exception-vector-table is required to be 11-bits aligned.
+ // And the size is 0x800.
+ // Please see the documentation as reference:
+ // https://developer.arm.com/docs/100933/0100/aarch64-exception-vector-table
+ //
+ // But, golang does not allow to set a function's address to a specific value.
+ // So, for gvisor, I defined the size of exception-vector-table as 4K,
+ // filled the 2nd 2K part with NOP-s.
+ // So that, I can safely move the 1st 2K part into the address with 11-bits alignment.
+ //
+ // So, the prerequisite for this function to work correctly is:
+ // vectorsSafeLen >= 0x1000
+ // vectorsRawLen = 0x800
+ vectorsSafeLen := int(safecopy.FindEndAddress(vectorsBegin) - vectorsBegin)
+ if vectorsSafeLen < 2*vectorsRawLen {
+ panic("Can't update vectors")
+ }
+
+ vectorsSafeTable := unsafeSlice(vectorsBegin, vectorsSafeLen) // Now a []uint32
+ vectorsRawLen32 := vectorsRawLen / int(instSize)
+
+ offset := vectorsBegin & (1<<11 - 1)
+ if offset != 0 {
+ offset = 1<<11 - offset
+ }
+
+ pageBegin := (vectorsBegin + offset) & ^uintptr(usermem.PageSize-1)
+
+ _, _, errno := syscall.Syscall(syscall.SYS_MPROTECT, uintptr(pageBegin), uintptr(usermem.PageSize), uintptr(syscall.PROT_READ|syscall.PROT_WRITE|syscall.PROT_EXEC))
+ if errno != 0 {
+ panic(errno.Error())
+ }
+
+ offset = offset / instSize // By index, not bytes.
+ // Move exception-vector-table into the specific address, should uses memmove here.
+ for i := 1; i <= vectorsRawLen32; i++ {
+ vectorsSafeTable[int(offset)+vectorsRawLen32-i] = vectorsSafeTable[vectorsRawLen32-i]
+ }
+
+ // Adjust branch since instruction was moved forward.
+ for i := 0; i < vectorsRawLen32; i++ {
+ if vectorsSafeTable[int(offset)+i] != nopInstruction {
+ vectorsSafeTable[int(offset)+i] -= uint32(offset)
+ }
+ }
+
+ _, _, errno = syscall.Syscall(syscall.SYS_MPROTECT, uintptr(pageBegin), uintptr(usermem.PageSize), uintptr(syscall.PROT_READ|syscall.PROT_EXEC))
+ if errno != 0 {
+ panic(errno.Error())
+ }
+}
diff --git a/pkg/sentry/socket/netstack/netstack.go b/pkg/sentry/socket/netstack/netstack.go
index 5d0085462..7ac38764d 100644
--- a/pkg/sentry/socket/netstack/netstack.go
+++ b/pkg/sentry/socket/netstack/netstack.go
@@ -300,7 +300,7 @@ type SocketOperations struct {
// New creates a new endpoint socket.
func New(t *kernel.Task, family int, skType linux.SockType, protocol int, queue *waiter.Queue, endpoint tcpip.Endpoint) (*fs.File, *syserr.Error) {
if skType == linux.SOCK_STREAM {
- if err := endpoint.SetSockOptInt(tcpip.DelayOption, 1); err != nil {
+ if err := endpoint.SetSockOptBool(tcpip.DelayOption, true); err != nil {
return nil, syserr.TranslateNetstackError(err)
}
}
@@ -535,7 +535,7 @@ func (s *SocketOperations) Write(ctx context.Context, _ *fs.File, src usermem.IO
}
if resCh != nil {
- t := ctx.(*kernel.Task)
+ t := kernel.TaskFromContext(ctx)
if err := t.Block(resCh); err != nil {
return 0, syserr.FromError(err).ToError()
}
@@ -608,7 +608,7 @@ func (s *SocketOperations) ReadFrom(ctx context.Context, _ *fs.File, r io.Reader
}
if resCh != nil {
- t := ctx.(*kernel.Task)
+ t := kernel.TaskFromContext(ctx)
if err := t.Block(resCh); err != nil {
return 0, syserr.FromError(err).ToError()
}
@@ -965,6 +965,13 @@ func GetSockOpt(t *kernel.Task, s socket.SocketOps, ep commonEndpoint, family in
return nil, syserr.ErrProtocolNotAvailable
}
+func boolToInt32(v bool) int32 {
+ if v {
+ return 1
+ }
+ return 0
+}
+
// getSockOptSocket implements GetSockOpt when level is SOL_SOCKET.
func getSockOptSocket(t *kernel.Task, s socket.SocketOps, ep commonEndpoint, family int, skType linux.SockType, name, outLen int) (interface{}, *syserr.Error) {
// TODO(b/124056281): Stop rejecting short optLen values in getsockopt.
@@ -998,12 +1005,11 @@ func getSockOptSocket(t *kernel.Task, s socket.SocketOps, ep commonEndpoint, fam
return nil, syserr.ErrInvalidArgument
}
- var v tcpip.PasscredOption
- if err := ep.GetSockOpt(&v); err != nil {
+ v, err := ep.GetSockOptBool(tcpip.PasscredOption)
+ if err != nil {
return nil, syserr.TranslateNetstackError(err)
}
-
- return int32(v), nil
+ return boolToInt32(v), nil
case linux.SO_SNDBUF:
if outLen < sizeOfInt32 {
@@ -1042,24 +1048,22 @@ func getSockOptSocket(t *kernel.Task, s socket.SocketOps, ep commonEndpoint, fam
return nil, syserr.ErrInvalidArgument
}
- var v tcpip.ReuseAddressOption
- if err := ep.GetSockOpt(&v); err != nil {
+ v, err := ep.GetSockOptBool(tcpip.ReuseAddressOption)
+ if err != nil {
return nil, syserr.TranslateNetstackError(err)
}
-
- return int32(v), nil
+ return boolToInt32(v), nil
case linux.SO_REUSEPORT:
if outLen < sizeOfInt32 {
return nil, syserr.ErrInvalidArgument
}
- var v tcpip.ReusePortOption
- if err := ep.GetSockOpt(&v); err != nil {
+ v, err := ep.GetSockOptBool(tcpip.ReusePortOption)
+ if err != nil {
return nil, syserr.TranslateNetstackError(err)
}
-
- return int32(v), nil
+ return boolToInt32(v), nil
case linux.SO_BINDTODEVICE:
var v tcpip.BindToDeviceOption
@@ -1089,24 +1093,22 @@ func getSockOptSocket(t *kernel.Task, s socket.SocketOps, ep commonEndpoint, fam
return nil, syserr.ErrInvalidArgument
}
- var v tcpip.BroadcastOption
- if err := ep.GetSockOpt(&v); err != nil {
+ v, err := ep.GetSockOptBool(tcpip.BroadcastOption)
+ if err != nil {
return nil, syserr.TranslateNetstackError(err)
}
-
- return int32(v), nil
+ return boolToInt32(v), nil
case linux.SO_KEEPALIVE:
if outLen < sizeOfInt32 {
return nil, syserr.ErrInvalidArgument
}
- var v tcpip.KeepaliveEnabledOption
- if err := ep.GetSockOpt(&v); err != nil {
+ v, err := ep.GetSockOptBool(tcpip.KeepaliveEnabledOption)
+ if err != nil {
return nil, syserr.TranslateNetstackError(err)
}
-
- return int32(v), nil
+ return boolToInt32(v), nil
case linux.SO_LINGER:
if outLen < linux.SizeOfLinger {
@@ -1156,47 +1158,41 @@ func getSockOptTCP(t *kernel.Task, ep commonEndpoint, name, outLen int) (interfa
return nil, syserr.ErrInvalidArgument
}
- v, err := ep.GetSockOptInt(tcpip.DelayOption)
+ v, err := ep.GetSockOptBool(tcpip.DelayOption)
if err != nil {
return nil, syserr.TranslateNetstackError(err)
}
-
- if v == 0 {
- return int32(1), nil
- }
- return int32(0), nil
+ return boolToInt32(!v), nil
case linux.TCP_CORK:
if outLen < sizeOfInt32 {
return nil, syserr.ErrInvalidArgument
}
- var v tcpip.CorkOption
- if err := ep.GetSockOpt(&v); err != nil {
+ v, err := ep.GetSockOptBool(tcpip.CorkOption)
+ if err != nil {
return nil, syserr.TranslateNetstackError(err)
}
-
- return int32(v), nil
+ return boolToInt32(v), nil
case linux.TCP_QUICKACK:
if outLen < sizeOfInt32 {
return nil, syserr.ErrInvalidArgument
}
- var v tcpip.QuickAckOption
- if err := ep.GetSockOpt(&v); err != nil {
+ v, err := ep.GetSockOptBool(tcpip.QuickAckOption)
+ if err != nil {
return nil, syserr.TranslateNetstackError(err)
}
-
- return int32(v), nil
+ return boolToInt32(v), nil
case linux.TCP_MAXSEG:
if outLen < sizeOfInt32 {
return nil, syserr.ErrInvalidArgument
}
- var v tcpip.MaxSegOption
- if err := ep.GetSockOpt(&v); err != nil {
+ v, err := ep.GetSockOptInt(tcpip.MaxSegOption)
+ if err != nil {
return nil, syserr.TranslateNetstackError(err)
}
@@ -1328,11 +1324,7 @@ func getSockOptIPv6(t *kernel.Task, ep commonEndpoint, name, outLen int) (interf
if err != nil {
return nil, syserr.TranslateNetstackError(err)
}
- var o int32
- if v {
- o = 1
- }
- return o, nil
+ return boolToInt32(v), nil
case linux.IPV6_PATHMTU:
t.Kernel().EmitUnimplementedEvent(t)
@@ -1342,8 +1334,8 @@ func getSockOptIPv6(t *kernel.Task, ep commonEndpoint, name, outLen int) (interf
if outLen == 0 {
return make([]byte, 0), nil
}
- var v tcpip.IPv6TrafficClassOption
- if err := ep.GetSockOpt(&v); err != nil {
+ v, err := ep.GetSockOptInt(tcpip.IPv6TrafficClassOption)
+ if err != nil {
return nil, syserr.TranslateNetstackError(err)
}
@@ -1365,12 +1357,7 @@ func getSockOptIPv6(t *kernel.Task, ep commonEndpoint, name, outLen int) (interf
if err != nil {
return nil, syserr.TranslateNetstackError(err)
}
-
- var o int32
- if v {
- o = 1
- }
- return o, nil
+ return boolToInt32(v), nil
default:
emitUnimplementedEventIPv6(t, name)
@@ -1386,8 +1373,8 @@ func getSockOptIP(t *kernel.Task, ep commonEndpoint, name, outLen int, family in
return nil, syserr.ErrInvalidArgument
}
- var v tcpip.TTLOption
- if err := ep.GetSockOpt(&v); err != nil {
+ v, err := ep.GetSockOptInt(tcpip.TTLOption)
+ if err != nil {
return nil, syserr.TranslateNetstackError(err)
}
@@ -1403,8 +1390,8 @@ func getSockOptIP(t *kernel.Task, ep commonEndpoint, name, outLen int, family in
return nil, syserr.ErrInvalidArgument
}
- var v tcpip.MulticastTTLOption
- if err := ep.GetSockOpt(&v); err != nil {
+ v, err := ep.GetSockOptInt(tcpip.MulticastTTLOption)
+ if err != nil {
return nil, syserr.TranslateNetstackError(err)
}
@@ -1429,23 +1416,19 @@ func getSockOptIP(t *kernel.Task, ep commonEndpoint, name, outLen int, family in
return nil, syserr.ErrInvalidArgument
}
- var v tcpip.MulticastLoopOption
- if err := ep.GetSockOpt(&v); err != nil {
+ v, err := ep.GetSockOptBool(tcpip.MulticastLoopOption)
+ if err != nil {
return nil, syserr.TranslateNetstackError(err)
}
-
- if v {
- return int32(1), nil
- }
- return int32(0), nil
+ return boolToInt32(v), nil
case linux.IP_TOS:
// Length handling for parity with Linux.
if outLen == 0 {
return []byte(nil), nil
}
- var v tcpip.IPv4TOSOption
- if err := ep.GetSockOpt(&v); err != nil {
+ v, err := ep.GetSockOptInt(tcpip.IPv4TOSOption)
+ if err != nil {
return nil, syserr.TranslateNetstackError(err)
}
if outLen < sizeOfInt32 {
@@ -1462,11 +1445,7 @@ func getSockOptIP(t *kernel.Task, ep commonEndpoint, name, outLen int, family in
if err != nil {
return nil, syserr.TranslateNetstackError(err)
}
- var o int32
- if v {
- o = 1
- }
- return o, nil
+ return boolToInt32(v), nil
case linux.IP_PKTINFO:
if outLen < sizeOfInt32 {
@@ -1477,11 +1456,7 @@ func getSockOptIP(t *kernel.Task, ep commonEndpoint, name, outLen int, family in
if err != nil {
return nil, syserr.TranslateNetstackError(err)
}
- var o int32
- if v {
- o = 1
- }
- return o, nil
+ return boolToInt32(v), nil
default:
emitUnimplementedEventIP(t, name)
@@ -1592,7 +1567,7 @@ func setSockOptSocket(t *kernel.Task, s socket.SocketOps, ep commonEndpoint, nam
}
v := usermem.ByteOrder.Uint32(optVal)
- return syserr.TranslateNetstackError(ep.SetSockOpt(tcpip.ReuseAddressOption(v)))
+ return syserr.TranslateNetstackError(ep.SetSockOptBool(tcpip.ReuseAddressOption, v != 0))
case linux.SO_REUSEPORT:
if len(optVal) < sizeOfInt32 {
@@ -1600,7 +1575,7 @@ func setSockOptSocket(t *kernel.Task, s socket.SocketOps, ep commonEndpoint, nam
}
v := usermem.ByteOrder.Uint32(optVal)
- return syserr.TranslateNetstackError(ep.SetSockOpt(tcpip.ReusePortOption(v)))
+ return syserr.TranslateNetstackError(ep.SetSockOptBool(tcpip.ReusePortOption, v != 0))
case linux.SO_BINDTODEVICE:
n := bytes.IndexByte(optVal, 0)
@@ -1628,7 +1603,7 @@ func setSockOptSocket(t *kernel.Task, s socket.SocketOps, ep commonEndpoint, nam
}
v := usermem.ByteOrder.Uint32(optVal)
- return syserr.TranslateNetstackError(ep.SetSockOpt(tcpip.BroadcastOption(v)))
+ return syserr.TranslateNetstackError(ep.SetSockOptBool(tcpip.BroadcastOption, v != 0))
case linux.SO_PASSCRED:
if len(optVal) < sizeOfInt32 {
@@ -1636,7 +1611,7 @@ func setSockOptSocket(t *kernel.Task, s socket.SocketOps, ep commonEndpoint, nam
}
v := usermem.ByteOrder.Uint32(optVal)
- return syserr.TranslateNetstackError(ep.SetSockOpt(tcpip.PasscredOption(v)))
+ return syserr.TranslateNetstackError(ep.SetSockOptBool(tcpip.PasscredOption, v != 0))
case linux.SO_KEEPALIVE:
if len(optVal) < sizeOfInt32 {
@@ -1644,7 +1619,7 @@ func setSockOptSocket(t *kernel.Task, s socket.SocketOps, ep commonEndpoint, nam
}
v := usermem.ByteOrder.Uint32(optVal)
- return syserr.TranslateNetstackError(ep.SetSockOpt(tcpip.KeepaliveEnabledOption(v)))
+ return syserr.TranslateNetstackError(ep.SetSockOptBool(tcpip.KeepaliveEnabledOption, v != 0))
case linux.SO_SNDTIMEO:
if len(optVal) < linux.SizeOfTimeval {
@@ -1716,11 +1691,7 @@ func setSockOptTCP(t *kernel.Task, ep commonEndpoint, name int, optVal []byte) *
}
v := usermem.ByteOrder.Uint32(optVal)
- var o int
- if v == 0 {
- o = 1
- }
- return syserr.TranslateNetstackError(ep.SetSockOptInt(tcpip.DelayOption, o))
+ return syserr.TranslateNetstackError(ep.SetSockOptBool(tcpip.DelayOption, v == 0))
case linux.TCP_CORK:
if len(optVal) < sizeOfInt32 {
@@ -1728,7 +1699,7 @@ func setSockOptTCP(t *kernel.Task, ep commonEndpoint, name int, optVal []byte) *
}
v := usermem.ByteOrder.Uint32(optVal)
- return syserr.TranslateNetstackError(ep.SetSockOpt(tcpip.CorkOption(v)))
+ return syserr.TranslateNetstackError(ep.SetSockOptBool(tcpip.CorkOption, v != 0))
case linux.TCP_QUICKACK:
if len(optVal) < sizeOfInt32 {
@@ -1736,7 +1707,7 @@ func setSockOptTCP(t *kernel.Task, ep commonEndpoint, name int, optVal []byte) *
}
v := usermem.ByteOrder.Uint32(optVal)
- return syserr.TranslateNetstackError(ep.SetSockOpt(tcpip.QuickAckOption(v)))
+ return syserr.TranslateNetstackError(ep.SetSockOptBool(tcpip.QuickAckOption, v != 0))
case linux.TCP_MAXSEG:
if len(optVal) < sizeOfInt32 {
@@ -1744,7 +1715,7 @@ func setSockOptTCP(t *kernel.Task, ep commonEndpoint, name int, optVal []byte) *
}
v := usermem.ByteOrder.Uint32(optVal)
- return syserr.TranslateNetstackError(ep.SetSockOpt(tcpip.MaxSegOption(v)))
+ return syserr.TranslateNetstackError(ep.SetSockOptInt(tcpip.MaxSegOption, int(v)))
case linux.TCP_KEEPIDLE:
if len(optVal) < sizeOfInt32 {
@@ -1855,7 +1826,7 @@ func setSockOptIPv6(t *kernel.Task, ep commonEndpoint, name int, optVal []byte)
if v == -1 {
v = 0
}
- return syserr.TranslateNetstackError(ep.SetSockOpt(tcpip.IPv6TrafficClassOption(v)))
+ return syserr.TranslateNetstackError(ep.SetSockOptInt(tcpip.IPv6TrafficClassOption, int(v)))
case linux.IPV6_RECVTCLASS:
v, err := parseIntOrChar(optVal)
@@ -1940,7 +1911,7 @@ func setSockOptIP(t *kernel.Task, ep commonEndpoint, name int, optVal []byte) *s
if v < 0 || v > 255 {
return syserr.ErrInvalidArgument
}
- return syserr.TranslateNetstackError(ep.SetSockOpt(tcpip.MulticastTTLOption(v)))
+ return syserr.TranslateNetstackError(ep.SetSockOptInt(tcpip.MulticastTTLOption, int(v)))
case linux.IP_ADD_MEMBERSHIP:
req, err := copyInMulticastRequest(optVal, false /* allowAddr */)
@@ -1987,9 +1958,7 @@ func setSockOptIP(t *kernel.Task, ep commonEndpoint, name int, optVal []byte) *s
return err
}
- return syserr.TranslateNetstackError(ep.SetSockOpt(
- tcpip.MulticastLoopOption(v != 0),
- ))
+ return syserr.TranslateNetstackError(ep.SetSockOptBool(tcpip.MulticastLoopOption, v != 0))
case linux.MCAST_JOIN_GROUP:
// FIXME(b/124219304): Implement MCAST_JOIN_GROUP.
@@ -2008,7 +1977,7 @@ func setSockOptIP(t *kernel.Task, ep commonEndpoint, name int, optVal []byte) *s
} else if v < 1 || v > 255 {
return syserr.ErrInvalidArgument
}
- return syserr.TranslateNetstackError(ep.SetSockOpt(tcpip.TTLOption(v)))
+ return syserr.TranslateNetstackError(ep.SetSockOptInt(tcpip.TTLOption, int(v)))
case linux.IP_TOS:
if len(optVal) == 0 {
@@ -2018,7 +1987,7 @@ func setSockOptIP(t *kernel.Task, ep commonEndpoint, name int, optVal []byte) *s
if err != nil {
return err
}
- return syserr.TranslateNetstackError(ep.SetSockOpt(tcpip.IPv4TOSOption(v)))
+ return syserr.TranslateNetstackError(ep.SetSockOptInt(tcpip.IPv4TOSOption, int(v)))
case linux.IP_RECVTOS:
v, err := parseIntOrChar(optVal)
diff --git a/pkg/sentry/socket/unix/transport/BUILD b/pkg/sentry/socket/unix/transport/BUILD
index 74bcd6300..c708b6030 100644
--- a/pkg/sentry/socket/unix/transport/BUILD
+++ b/pkg/sentry/socket/unix/transport/BUILD
@@ -30,6 +30,7 @@ go_library(
"//pkg/abi/linux",
"//pkg/context",
"//pkg/ilist",
+ "//pkg/log",
"//pkg/refs",
"//pkg/sync",
"//pkg/syserr",
diff --git a/pkg/sentry/socket/unix/transport/unix.go b/pkg/sentry/socket/unix/transport/unix.go
index 2ef654235..1f3880cc5 100644
--- a/pkg/sentry/socket/unix/transport/unix.go
+++ b/pkg/sentry/socket/unix/transport/unix.go
@@ -20,6 +20,7 @@ import (
"gvisor.dev/gvisor/pkg/abi/linux"
"gvisor.dev/gvisor/pkg/context"
+ "gvisor.dev/gvisor/pkg/log"
"gvisor.dev/gvisor/pkg/sync"
"gvisor.dev/gvisor/pkg/syserr"
"gvisor.dev/gvisor/pkg/tcpip"
@@ -838,24 +839,45 @@ func (e *baseEndpoint) SendMsg(ctx context.Context, data [][]byte, c ControlMess
// SetSockOpt sets a socket option. Currently not supported.
func (e *baseEndpoint) SetSockOpt(opt interface{}) *tcpip.Error {
- switch v := opt.(type) {
- case tcpip.PasscredOption:
- e.setPasscred(v != 0)
- return nil
- }
return nil
}
func (e *baseEndpoint) SetSockOptBool(opt tcpip.SockOptBool, v bool) *tcpip.Error {
+ switch opt {
+ case tcpip.BroadcastOption:
+ case tcpip.PasscredOption:
+ e.setPasscred(v)
+ case tcpip.ReuseAddressOption:
+ default:
+ log.Warningf("Unsupported socket option: %d", opt)
+ return tcpip.ErrUnknownProtocolOption
+ }
return nil
}
func (e *baseEndpoint) SetSockOptInt(opt tcpip.SockOptInt, v int) *tcpip.Error {
+ switch opt {
+ case tcpip.SendBufferSizeOption:
+ case tcpip.ReceiveBufferSizeOption:
+ default:
+ log.Warningf("Unsupported socket option: %d", opt)
+ return tcpip.ErrUnknownProtocolOption
+ }
return nil
}
func (e *baseEndpoint) GetSockOptBool(opt tcpip.SockOptBool) (bool, *tcpip.Error) {
- return false, tcpip.ErrUnknownProtocolOption
+ switch opt {
+ case tcpip.KeepaliveEnabledOption:
+ return false, nil
+
+ case tcpip.PasscredOption:
+ return e.Passcred(), nil
+
+ default:
+ log.Warningf("Unsupported socket option: %d", opt)
+ return false, tcpip.ErrUnknownProtocolOption
+ }
}
func (e *baseEndpoint) GetSockOptInt(opt tcpip.SockOptInt) (int, *tcpip.Error) {
@@ -914,29 +936,19 @@ func (e *baseEndpoint) GetSockOptInt(opt tcpip.SockOptInt) (int, *tcpip.Error) {
return int(v), nil
default:
+ log.Warningf("Unsupported socket option: %d", opt)
return -1, tcpip.ErrUnknownProtocolOption
}
}
// GetSockOpt implements tcpip.Endpoint.GetSockOpt.
func (e *baseEndpoint) GetSockOpt(opt interface{}) *tcpip.Error {
- switch o := opt.(type) {
+ switch opt.(type) {
case tcpip.ErrorOption:
return nil
- case *tcpip.PasscredOption:
- if e.Passcred() {
- *o = tcpip.PasscredOption(1)
- } else {
- *o = tcpip.PasscredOption(0)
- }
- return nil
-
- case *tcpip.KeepaliveEnabledOption:
- *o = 0
- return nil
-
default:
+ log.Warningf("Unsupported socket option: %T", opt)
return tcpip.ErrUnknownProtocolOption
}
}
diff --git a/pkg/sentry/strace/strace.go b/pkg/sentry/strace/strace.go
index 77655558e..b94c4fbf5 100644
--- a/pkg/sentry/strace/strace.go
+++ b/pkg/sentry/strace/strace.go
@@ -778,9 +778,6 @@ func (s SyscallMap) Name(sysno uintptr) string {
//
// N.B. This is not in an init function because we can't be sure all syscall
// tables are registered with the kernel when init runs.
-//
-// TODO(gvisor.dev/issue/155): remove kernel package dependencies from this
-// package and have the kernel package self-initialize all syscall tables.
func Initialize() {
for _, table := range kernel.SyscallTables() {
// Is this known?
diff --git a/pkg/sentry/syscalls/linux/sys_aio.go b/pkg/sentry/syscalls/linux/sys_aio.go
index b401978db..38cbeba5a 100644
--- a/pkg/sentry/syscalls/linux/sys_aio.go
+++ b/pkg/sentry/syscalls/linux/sys_aio.go
@@ -114,14 +114,28 @@ func IoSetup(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Sysca
func IoDestroy(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.SyscallControl, error) {
id := args[0].Uint64()
- // Destroy the given context.
- if !t.MemoryManager().DestroyAIOContext(t, id) {
+ ctx := t.MemoryManager().DestroyAIOContext(t, id)
+ if ctx == nil {
// Does not exist.
return 0, nil, syserror.EINVAL
}
- // FIXME(fvoznika): Linux blocks until all AIO to the destroyed context is
- // done.
- return 0, nil, nil
+
+ // Drain completed requests amd wait for pending requests until there are no
+ // more.
+ for {
+ ctx.Drain()
+
+ ch := ctx.WaitChannel()
+ if ch == nil {
+ // No more requests, we're done.
+ return 0, nil, nil
+ }
+ // The task cannot be interrupted during the wait. Equivalent to
+ // TASK_UNINTERRUPTIBLE in Linux.
+ t.UninterruptibleSleepStart(true /* deactivate */)
+ <-ch
+ t.UninterruptibleSleepFinish(true /* activate */)
+ }
}
// IoGetevents implements linux syscall io_getevents(2).
@@ -200,13 +214,13 @@ func IoGetevents(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.S
func waitForRequest(ctx *mm.AIOContext, t *kernel.Task, haveDeadline bool, deadline ktime.Time) (interface{}, error) {
for {
if v, ok := ctx.PopRequest(); ok {
- // Request was readly available. Just return it.
+ // Request was readily available. Just return it.
return v, nil
}
// Need to wait for request completion.
- done, active := ctx.WaitChannel()
- if !active {
+ done := ctx.WaitChannel()
+ if done == nil {
// Context has been destroyed.
return nil, syserror.EINVAL
}
@@ -248,6 +262,10 @@ func memoryFor(t *kernel.Task, cb *ioCallback) (usermem.IOSequence, error) {
}
func performCallback(t *kernel.Task, file *fs.File, cbAddr usermem.Addr, cb *ioCallback, ioseq usermem.IOSequence, ctx *mm.AIOContext, eventFile *fs.File) {
+ if ctx.Dead() {
+ ctx.CancelPendingRequest()
+ return
+ }
ev := &ioEvent{
Data: cb.Data,
Obj: uint64(cbAddr),
diff --git a/pkg/sentry/syscalls/linux/sys_prctl.go b/pkg/sentry/syscalls/linux/sys_prctl.go
index 9c6728530..f92bf8096 100644
--- a/pkg/sentry/syscalls/linux/sys_prctl.go
+++ b/pkg/sentry/syscalls/linux/sys_prctl.go
@@ -161,8 +161,8 @@ func Prctl(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Syscall
if args[1].Int() != 1 || args[2].Int() != 0 || args[3].Int() != 0 || args[4].Int() != 0 {
return 0, nil, syserror.EINVAL
}
- // no_new_privs is assumed to always be set. See
- // kernel.Task.updateCredsForExec.
+ // PR_SET_NO_NEW_PRIVS is assumed to always be set.
+ // See kernel.Task.updateCredsForExecLocked.
return 0, nil, nil
case linux.PR_GET_NO_NEW_PRIVS:
diff --git a/pkg/sentry/syscalls/linux/sys_read.go b/pkg/sentry/syscalls/linux/sys_read.go
index 78a2cb750..071b4bacc 100644
--- a/pkg/sentry/syscalls/linux/sys_read.go
+++ b/pkg/sentry/syscalls/linux/sys_read.go
@@ -96,8 +96,8 @@ func Readahead(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Sys
return 0, nil, syserror.EINVAL
}
- // Check that the offset is legitimate.
- if offset < 0 {
+ // Check that the offset is legitimate and does not overflow.
+ if offset < 0 || offset+int64(size) < 0 {
return 0, nil, syserror.EINVAL
}
@@ -120,8 +120,8 @@ func Pread64(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Sysca
}
defer file.DecRef()
- // Check that the offset is legitimate.
- if offset < 0 {
+ // Check that the offset is legitimate and does not overflow.
+ if offset < 0 || offset+int64(size) < 0 {
return 0, nil, syserror.EINVAL
}
diff --git a/pkg/sentry/syscalls/linux/sys_socket.go b/pkg/sentry/syscalls/linux/sys_socket.go
index 2919228d0..61b2576ac 100644
--- a/pkg/sentry/syscalls/linux/sys_socket.go
+++ b/pkg/sentry/syscalls/linux/sys_socket.go
@@ -31,6 +31,8 @@ import (
"gvisor.dev/gvisor/pkg/usermem"
)
+// LINT.IfChange
+
// minListenBacklog is the minimum reasonable backlog for listening sockets.
const minListenBacklog = 8
@@ -244,7 +246,10 @@ func SocketPair(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Sy
// Copy the file descriptors out.
if _, err := t.CopyOut(socks, fds); err != nil {
- // Note that we don't close files here; see pipe(2) also.
+ for _, fd := range fds {
+ _, file := t.FDTable().Remove(fd)
+ file.DecRef()
+ }
return 0, nil, err
}
@@ -1128,3 +1133,5 @@ func SendTo(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Syscal
n, err := sendTo(t, fd, bufPtr, bufLen, flags, namePtr, nameLen)
return n, nil, err
}
+
+// LINT.ThenChange(./vfs2/socket.go)
diff --git a/pkg/sentry/syscalls/linux/sys_splice.go b/pkg/sentry/syscalls/linux/sys_splice.go
index fd642834b..df0d0f461 100644
--- a/pkg/sentry/syscalls/linux/sys_splice.go
+++ b/pkg/sentry/syscalls/linux/sys_splice.go
@@ -16,6 +16,7 @@ package linux
import (
"gvisor.dev/gvisor/pkg/abi/linux"
+ "gvisor.dev/gvisor/pkg/log"
"gvisor.dev/gvisor/pkg/sentry/arch"
"gvisor.dev/gvisor/pkg/sentry/fs"
"gvisor.dev/gvisor/pkg/sentry/kernel"
@@ -25,10 +26,15 @@ import (
// doSplice implements a blocking splice operation.
func doSplice(t *kernel.Task, outFile, inFile *fs.File, opts fs.SpliceOpts, nonBlocking bool) (int64, error) {
- if opts.Length < 0 || opts.SrcStart < 0 || opts.DstStart < 0 {
+ log.Infof("NLAC: doSplice opts: %+v", opts)
+ if opts.Length < 0 || opts.SrcStart < 0 || opts.DstStart < 0 || (opts.SrcStart+opts.Length < 0) {
return 0, syserror.EINVAL
}
+ if opts.Length > int64(kernel.MAX_RW_COUNT) {
+ opts.Length = int64(kernel.MAX_RW_COUNT)
+ }
+
var (
total int64
n int64
diff --git a/pkg/sentry/syscalls/linux/sys_write.go b/pkg/sentry/syscalls/linux/sys_write.go
index 506ee54ce..6ec0de96e 100644
--- a/pkg/sentry/syscalls/linux/sys_write.go
+++ b/pkg/sentry/syscalls/linux/sys_write.go
@@ -87,8 +87,8 @@ func Pwrite64(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Sysc
}
defer file.DecRef()
- // Check that the offset is legitimate.
- if offset < 0 {
+ // Check that the offset is legitimate and does not overflow.
+ if offset < 0 || offset+int64(size) < 0 {
return 0, nil, syserror.EINVAL
}
diff --git a/pkg/sentry/syscalls/linux/vfs2/BUILD b/pkg/sentry/syscalls/linux/vfs2/BUILD
index 0004e60d9..b32abfe59 100644
--- a/pkg/sentry/syscalls/linux/vfs2/BUILD
+++ b/pkg/sentry/syscalls/linux/vfs2/BUILD
@@ -21,6 +21,7 @@ go_library(
"poll.go",
"read_write.go",
"setstat.go",
+ "socket.go",
"stat.go",
"stat_amd64.go",
"stat_arm64.go",
@@ -32,6 +33,7 @@ go_library(
visibility = ["//:sandbox"],
deps = [
"//pkg/abi/linux",
+ "//pkg/binary",
"//pkg/bits",
"//pkg/fspath",
"//pkg/gohacks",
@@ -43,10 +45,14 @@ go_library(
"//pkg/sentry/limits",
"//pkg/sentry/loader",
"//pkg/sentry/memmap",
+ "//pkg/sentry/socket",
+ "//pkg/sentry/socket/control",
+ "//pkg/sentry/socket/unix/transport",
"//pkg/sentry/syscalls",
"//pkg/sentry/syscalls/linux",
"//pkg/sentry/vfs",
"//pkg/sync",
+ "//pkg/syserr",
"//pkg/syserror",
"//pkg/usermem",
"//pkg/waiter",
diff --git a/pkg/sentry/syscalls/linux/vfs2/getdents.go b/pkg/sentry/syscalls/linux/vfs2/getdents.go
index a61cc5059..62e98817d 100644
--- a/pkg/sentry/syscalls/linux/vfs2/getdents.go
+++ b/pkg/sentry/syscalls/linux/vfs2/getdents.go
@@ -97,6 +97,7 @@ func (cb *getdentsCallback) Handle(dirent vfs.Dirent) error {
// char d_name[]; /* Filename (null-terminated) */
// };
size := 8 + 8 + 2 + 1 + 1 + len(dirent.Name)
+ size = (size + 7) &^ 7 // round up to multiple of 8
if size > cb.remaining {
return syserror.EINVAL
}
@@ -106,7 +107,12 @@ func (cb *getdentsCallback) Handle(dirent vfs.Dirent) error {
usermem.ByteOrder.PutUint16(buf[16:18], uint16(size))
buf[18] = dirent.Type
copy(buf[19:], dirent.Name)
- buf[size-1] = 0 // NUL terminator
+ // Zero out all remaining bytes in buf, including the NUL terminator
+ // after dirent.Name.
+ bufTail := buf[19+len(dirent.Name):]
+ for i := range bufTail {
+ bufTail[i] = 0
+ }
} else {
// struct linux_dirent {
// unsigned long d_ino; /* Inode number */
@@ -125,6 +131,7 @@ func (cb *getdentsCallback) Handle(dirent vfs.Dirent) error {
panic(fmt.Sprintf("unsupported sizeof(unsigned long): %d", cb.t.Arch().Width()))
}
size := 8 + 8 + 2 + 1 + 1 + 1 + len(dirent.Name)
+ size = (size + 7) &^ 7 // round up to multiple of sizeof(long)
if size > cb.remaining {
return syserror.EINVAL
}
@@ -133,9 +140,14 @@ func (cb *getdentsCallback) Handle(dirent vfs.Dirent) error {
usermem.ByteOrder.PutUint64(buf[8:16], uint64(dirent.NextOff))
usermem.ByteOrder.PutUint16(buf[16:18], uint16(size))
copy(buf[18:], dirent.Name)
- buf[size-3] = 0 // NUL terminator
- buf[size-2] = 0 // zero padding byte
- buf[size-1] = dirent.Type
+ // Zero out all remaining bytes in buf, including the NUL terminator
+ // after dirent.Name and the zero padding byte between the name and
+ // dirent type.
+ bufTail := buf[18+len(dirent.Name):]
+ for i := range bufTail {
+ bufTail[i] = 0
+ }
+ bufTail[2] = dirent.Type
}
n, err := cb.t.CopyOutBytes(cb.addr, buf)
if err != nil {
diff --git a/pkg/sentry/syscalls/linux/vfs2/linux64_override_amd64.go b/pkg/sentry/syscalls/linux/vfs2/linux64_override_amd64.go
index 63febc2f7..645e0bcb8 100644
--- a/pkg/sentry/syscalls/linux/vfs2/linux64_override_amd64.go
+++ b/pkg/sentry/syscalls/linux/vfs2/linux64_override_amd64.go
@@ -44,21 +44,22 @@ func Override(table map[uintptr]kernel.Syscall) {
table[32] = syscalls.Supported("dup", Dup)
table[33] = syscalls.Supported("dup2", Dup2)
delete(table, 40) // sendfile
- delete(table, 41) // socket
- delete(table, 42) // connect
- delete(table, 43) // accept
- delete(table, 44) // sendto
- delete(table, 45) // recvfrom
- delete(table, 46) // sendmsg
- delete(table, 47) // recvmsg
- delete(table, 48) // shutdown
- delete(table, 49) // bind
- delete(table, 50) // listen
- delete(table, 51) // getsockname
- delete(table, 52) // getpeername
- delete(table, 53) // socketpair
- delete(table, 54) // setsockopt
- delete(table, 55) // getsockopt
+ // TODO(gvisor.dev/issue/1485): Port all socket variants to VFS2.
+ table[41] = syscalls.PartiallySupported("socket", Socket, "In process of porting socket syscalls to VFS2.", nil)
+ table[42] = syscalls.PartiallySupported("connect", Connect, "In process of porting socket syscalls to VFS2.", nil)
+ table[43] = syscalls.PartiallySupported("accept", Accept, "In process of porting socket syscalls to VFS2.", nil)
+ table[44] = syscalls.PartiallySupported("sendto", SendTo, "In process of porting socket syscalls to VFS2.", nil)
+ table[45] = syscalls.PartiallySupported("recvfrom", RecvFrom, "In process of porting socket syscalls to VFS2.", nil)
+ table[46] = syscalls.PartiallySupported("sendmsg", SendMsg, "In process of porting socket syscalls to VFS2.", nil)
+ table[47] = syscalls.PartiallySupported("recvmsg", RecvMsg, "In process of porting socket syscalls to VFS2.", nil)
+ table[48] = syscalls.PartiallySupported("shutdown", Shutdown, "In process of porting socket syscalls to VFS2.", nil)
+ table[49] = syscalls.PartiallySupported("bind", Bind, "In process of porting socket syscalls to VFS2.", nil)
+ table[50] = syscalls.PartiallySupported("listen", Listen, "In process of porting socket syscalls to VFS2.", nil)
+ table[51] = syscalls.PartiallySupported("getsockname", GetSockName, "In process of porting socket syscalls to VFS2.", nil)
+ table[52] = syscalls.PartiallySupported("getpeername", GetPeerName, "In process of porting socket syscalls to VFS2.", nil)
+ table[53] = syscalls.PartiallySupported("socketpair", SocketPair, "In process of porting socket syscalls to VFS2.", nil)
+ table[54] = syscalls.PartiallySupported("getsockopt", GetSockOpt, "In process of porting socket syscalls to VFS2.", nil)
+ table[55] = syscalls.PartiallySupported("setsockopt", SetSockOpt, "In process of porting socket syscalls to VFS2.", nil)
table[59] = syscalls.Supported("execve", Execve)
table[72] = syscalls.Supported("fcntl", Fcntl)
delete(table, 73) // flock
@@ -144,7 +145,8 @@ func Override(table map[uintptr]kernel.Syscall) {
delete(table, 285) // fallocate
table[286] = syscalls.Supported("timerfd_settime", TimerfdSettime)
table[287] = syscalls.Supported("timerfd_gettime", TimerfdGettime)
- delete(table, 288) // accept4
+ // TODO(gvisor.dev/issue/1485): Port all socket variants to VFS2.
+ table[288] = syscalls.PartiallySupported("accept4", Accept4, "In process of porting socket syscalls to VFS2.", nil)
delete(table, 289) // signalfd4
delete(table, 290) // eventfd2
table[291] = syscalls.Supported("epoll_create1", EpollCreate1)
@@ -153,9 +155,11 @@ func Override(table map[uintptr]kernel.Syscall) {
delete(table, 294) // inotify_init1
table[295] = syscalls.Supported("preadv", Preadv)
table[296] = syscalls.Supported("pwritev", Pwritev)
- delete(table, 299) // recvmmsg
+ // TODO(gvisor.dev/issue/1485): Port all socket variants to VFS2.
+ table[299] = syscalls.PartiallySupported("recvmmsg", RecvMMsg, "In process of porting socket syscalls to VFS2.", nil)
table[306] = syscalls.Supported("syncfs", Syncfs)
- delete(table, 307) // sendmmsg
+ // TODO(gvisor.dev/issue/1485): Port all socket variants to VFS2.
+ table[307] = syscalls.PartiallySupported("sendmmsg", SendMMsg, "In process of porting socket syscalls to VFS2.", nil)
table[316] = syscalls.Supported("renameat2", Renameat2)
delete(table, 319) // memfd_create
table[322] = syscalls.Supported("execveat", Execveat)
diff --git a/pkg/sentry/syscalls/linux/vfs2/read_write.go b/pkg/sentry/syscalls/linux/vfs2/read_write.go
index 35f6308d6..898b190fd 100644
--- a/pkg/sentry/syscalls/linux/vfs2/read_write.go
+++ b/pkg/sentry/syscalls/linux/vfs2/read_write.go
@@ -130,8 +130,8 @@ func Pread64(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Sysca
}
defer file.DecRef()
- // Check that the offset is legitimate.
- if offset < 0 {
+ // Check that the offset is legitimate and does not overflow.
+ if offset < 0 || offset+int64(size) < 0 {
return 0, nil, syserror.EINVAL
}
@@ -362,8 +362,8 @@ func Pwrite64(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Sysc
}
defer file.DecRef()
- // Check that the offset is legitimate.
- if offset < 0 {
+ // Check that the offset is legitimate and does not overflow.
+ if offset < 0 || offset+int64(size) < 0 {
return 0, nil, syserror.EINVAL
}
diff --git a/pkg/sentry/syscalls/linux/vfs2/socket.go b/pkg/sentry/syscalls/linux/vfs2/socket.go
new file mode 100644
index 000000000..79a4a7ada
--- /dev/null
+++ b/pkg/sentry/syscalls/linux/vfs2/socket.go
@@ -0,0 +1,1138 @@
+// Copyright 2018 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 vfs2
+
+import (
+ "time"
+
+ "gvisor.dev/gvisor/pkg/abi/linux"
+ "gvisor.dev/gvisor/pkg/binary"
+ "gvisor.dev/gvisor/pkg/sentry/arch"
+ "gvisor.dev/gvisor/pkg/sentry/kernel"
+ ktime "gvisor.dev/gvisor/pkg/sentry/kernel/time"
+ "gvisor.dev/gvisor/pkg/sentry/socket"
+ "gvisor.dev/gvisor/pkg/sentry/socket/control"
+ "gvisor.dev/gvisor/pkg/sentry/socket/unix/transport"
+ slinux "gvisor.dev/gvisor/pkg/sentry/syscalls/linux"
+ "gvisor.dev/gvisor/pkg/sentry/vfs"
+ "gvisor.dev/gvisor/pkg/syserr"
+ "gvisor.dev/gvisor/pkg/syserror"
+ "gvisor.dev/gvisor/pkg/usermem"
+)
+
+// minListenBacklog is the minimum reasonable backlog for listening sockets.
+const minListenBacklog = 8
+
+// maxListenBacklog is the maximum allowed backlog for listening sockets.
+const maxListenBacklog = 1024
+
+// maxAddrLen is the maximum socket address length we're willing to accept.
+const maxAddrLen = 200
+
+// maxOptLen is the maximum sockopt parameter length we're willing to accept.
+const maxOptLen = 1024 * 8
+
+// maxControlLen is the maximum length of the msghdr.msg_control buffer we're
+// willing to accept. Note that this limit is smaller than Linux, which allows
+// buffers upto INT_MAX.
+const maxControlLen = 10 * 1024 * 1024
+
+// nameLenOffset is the offset from the start of the MessageHeader64 struct to
+// the NameLen field.
+const nameLenOffset = 8
+
+// controlLenOffset is the offset form the start of the MessageHeader64 struct
+// to the ControlLen field.
+const controlLenOffset = 40
+
+// flagsOffset is the offset form the start of the MessageHeader64 struct
+// to the Flags field.
+const flagsOffset = 48
+
+const sizeOfInt32 = 4
+
+// messageHeader64Len is the length of a MessageHeader64 struct.
+var messageHeader64Len = uint64(binary.Size(MessageHeader64{}))
+
+// multipleMessageHeader64Len is the length of a multipeMessageHeader64 struct.
+var multipleMessageHeader64Len = uint64(binary.Size(multipleMessageHeader64{}))
+
+// baseRecvFlags are the flags that are accepted across recvmsg(2),
+// recvmmsg(2), and recvfrom(2).
+const baseRecvFlags = linux.MSG_OOB | linux.MSG_DONTROUTE | linux.MSG_DONTWAIT | linux.MSG_NOSIGNAL | linux.MSG_WAITALL | linux.MSG_TRUNC | linux.MSG_CTRUNC
+
+// MessageHeader64 is the 64-bit representation of the msghdr struct used in
+// the recvmsg and sendmsg syscalls.
+type MessageHeader64 struct {
+ // Name is the optional pointer to a network address buffer.
+ Name uint64
+
+ // NameLen is the length of the buffer pointed to by Name.
+ NameLen uint32
+ _ uint32
+
+ // Iov is a pointer to an array of io vectors that describe the memory
+ // locations involved in the io operation.
+ Iov uint64
+
+ // IovLen is the length of the array pointed to by Iov.
+ IovLen uint64
+
+ // Control is the optional pointer to ancillary control data.
+ Control uint64
+
+ // ControlLen is the length of the data pointed to by Control.
+ ControlLen uint64
+
+ // Flags on the sent/received message.
+ Flags int32
+ _ int32
+}
+
+// multipleMessageHeader64 is the 64-bit representation of the mmsghdr struct used in
+// the recvmmsg and sendmmsg syscalls.
+type multipleMessageHeader64 struct {
+ msgHdr MessageHeader64
+ msgLen uint32
+ _ int32
+}
+
+// CopyInMessageHeader64 copies a message header from user to kernel memory.
+func CopyInMessageHeader64(t *kernel.Task, addr usermem.Addr, msg *MessageHeader64) error {
+ b := t.CopyScratchBuffer(52)
+ if _, err := t.CopyInBytes(addr, b); err != nil {
+ return err
+ }
+
+ msg.Name = usermem.ByteOrder.Uint64(b[0:])
+ msg.NameLen = usermem.ByteOrder.Uint32(b[8:])
+ msg.Iov = usermem.ByteOrder.Uint64(b[16:])
+ msg.IovLen = usermem.ByteOrder.Uint64(b[24:])
+ msg.Control = usermem.ByteOrder.Uint64(b[32:])
+ msg.ControlLen = usermem.ByteOrder.Uint64(b[40:])
+ msg.Flags = int32(usermem.ByteOrder.Uint32(b[48:]))
+
+ return nil
+}
+
+// CaptureAddress allocates memory for and copies a socket address structure
+// from the untrusted address space range.
+func CaptureAddress(t *kernel.Task, addr usermem.Addr, addrlen uint32) ([]byte, error) {
+ if addrlen > maxAddrLen {
+ return nil, syserror.EINVAL
+ }
+
+ addrBuf := make([]byte, addrlen)
+ if _, err := t.CopyInBytes(addr, addrBuf); err != nil {
+ return nil, err
+ }
+
+ return addrBuf, nil
+}
+
+// writeAddress writes a sockaddr structure and its length to an output buffer
+// in the unstrusted address space range. If the address is bigger than the
+// buffer, it is truncated.
+func writeAddress(t *kernel.Task, addr interface{}, addrLen uint32, addrPtr usermem.Addr, addrLenPtr usermem.Addr) error {
+ // Get the buffer length.
+ var bufLen uint32
+ if _, err := t.CopyIn(addrLenPtr, &bufLen); err != nil {
+ return err
+ }
+
+ if int32(bufLen) < 0 {
+ return syserror.EINVAL
+ }
+
+ // Write the length unconditionally.
+ if _, err := t.CopyOut(addrLenPtr, addrLen); err != nil {
+ return err
+ }
+
+ if addr == nil {
+ return nil
+ }
+
+ if bufLen > addrLen {
+ bufLen = addrLen
+ }
+
+ // Copy as much of the address as will fit in the buffer.
+ encodedAddr := binary.Marshal(nil, usermem.ByteOrder, addr)
+ if bufLen > uint32(len(encodedAddr)) {
+ bufLen = uint32(len(encodedAddr))
+ }
+ _, err := t.CopyOutBytes(addrPtr, encodedAddr[:int(bufLen)])
+ return err
+}
+
+// Socket implements the linux syscall socket(2).
+func Socket(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.SyscallControl, error) {
+ domain := int(args[0].Int())
+ stype := args[1].Int()
+ protocol := int(args[2].Int())
+
+ // Check and initialize the flags.
+ if stype & ^(0xf|linux.SOCK_NONBLOCK|linux.SOCK_CLOEXEC) != 0 {
+ return 0, nil, syserror.EINVAL
+ }
+
+ // Create the new socket.
+ s, e := socket.NewVFS2(t, domain, linux.SockType(stype&0xf), protocol)
+ if e != nil {
+ return 0, nil, e.ToError()
+ }
+ defer s.DecRef()
+
+ if err := s.SetStatusFlags(t, t.Credentials(), uint32(stype&linux.SOCK_NONBLOCK)); err != nil {
+ return 0, nil, err
+ }
+
+ fd, err := t.NewFDFromVFS2(0, s, kernel.FDFlags{
+ CloseOnExec: stype&linux.SOCK_CLOEXEC != 0,
+ })
+ if err != nil {
+ return 0, nil, err
+ }
+
+ return uintptr(fd), nil, nil
+}
+
+// SocketPair implements the linux syscall socketpair(2).
+func SocketPair(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.SyscallControl, error) {
+ domain := int(args[0].Int())
+ stype := args[1].Int()
+ protocol := int(args[2].Int())
+ addr := args[3].Pointer()
+
+ // Check and initialize the flags.
+ if stype & ^(0xf|linux.SOCK_NONBLOCK|linux.SOCK_CLOEXEC) != 0 {
+ return 0, nil, syserror.EINVAL
+ }
+
+ // Create the socket pair.
+ s1, s2, e := socket.PairVFS2(t, domain, linux.SockType(stype&0xf), protocol)
+ if e != nil {
+ return 0, nil, e.ToError()
+ }
+ // Adding to the FD table will cause an extra reference to be acquired.
+ defer s1.DecRef()
+ defer s2.DecRef()
+
+ nonblocking := uint32(stype & linux.SOCK_NONBLOCK)
+ if err := s1.SetStatusFlags(t, t.Credentials(), nonblocking); err != nil {
+ return 0, nil, err
+ }
+ if err := s2.SetStatusFlags(t, t.Credentials(), nonblocking); err != nil {
+ return 0, nil, err
+ }
+
+ // Create the FDs for the sockets.
+ flags := kernel.FDFlags{
+ CloseOnExec: stype&linux.SOCK_CLOEXEC != 0,
+ }
+ fds, err := t.NewFDsVFS2(0, []*vfs.FileDescription{s1, s2}, flags)
+ if err != nil {
+ return 0, nil, err
+ }
+
+ if _, err := t.CopyOut(addr, fds); err != nil {
+ for _, fd := range fds {
+ _, file := t.FDTable().Remove(fd)
+ file.DecRef()
+ }
+ return 0, nil, err
+ }
+
+ return 0, nil, nil
+}
+
+// Connect implements the linux syscall connect(2).
+func Connect(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.SyscallControl, error) {
+ fd := args[0].Int()
+ addr := args[1].Pointer()
+ addrlen := args[2].Uint()
+
+ // Get socket from the file descriptor.
+ file := t.GetFileVFS2(fd)
+ if file == nil {
+ return 0, nil, syserror.EBADF
+ }
+ defer file.DecRef()
+
+ // Extract the socket.
+ s, ok := file.Impl().(socket.SocketVFS2)
+ if !ok {
+ return 0, nil, syserror.ENOTSOCK
+ }
+
+ // Capture address and call syscall implementation.
+ a, err := CaptureAddress(t, addr, addrlen)
+ if err != nil {
+ return 0, nil, err
+ }
+
+ blocking := (file.StatusFlags() & linux.SOCK_NONBLOCK) == 0
+ return 0, nil, syserror.ConvertIntr(s.Connect(t, a, blocking).ToError(), kernel.ERESTARTSYS)
+}
+
+// accept is the implementation of the accept syscall. It is called by accept
+// and accept4 syscall handlers.
+func accept(t *kernel.Task, fd int32, addr usermem.Addr, addrLen usermem.Addr, flags int) (uintptr, error) {
+ // Check that no unsupported flags are passed in.
+ if flags & ^(linux.SOCK_NONBLOCK|linux.SOCK_CLOEXEC) != 0 {
+ return 0, syserror.EINVAL
+ }
+
+ // Get socket from the file descriptor.
+ file := t.GetFileVFS2(fd)
+ if file == nil {
+ return 0, syserror.EBADF
+ }
+ defer file.DecRef()
+
+ // Extract the socket.
+ s, ok := file.Impl().(socket.SocketVFS2)
+ if !ok {
+ return 0, syserror.ENOTSOCK
+ }
+
+ // Call the syscall implementation for this socket, then copy the
+ // output address if one is specified.
+ blocking := (file.StatusFlags() & linux.SOCK_NONBLOCK) == 0
+
+ peerRequested := addrLen != 0
+ nfd, peer, peerLen, e := s.Accept(t, peerRequested, flags, blocking)
+ if e != nil {
+ return 0, syserror.ConvertIntr(e.ToError(), kernel.ERESTARTSYS)
+ }
+ if peerRequested {
+ // NOTE(magi): Linux does not give you an error if it can't
+ // write the data back out so neither do we.
+ if err := writeAddress(t, peer, peerLen, addr, addrLen); err == syserror.EINVAL {
+ return 0, err
+ }
+ }
+ return uintptr(nfd), nil
+}
+
+// Accept4 implements the linux syscall accept4(2).
+func Accept4(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.SyscallControl, error) {
+ fd := args[0].Int()
+ addr := args[1].Pointer()
+ addrlen := args[2].Pointer()
+ flags := int(args[3].Int())
+
+ n, err := accept(t, fd, addr, addrlen, flags)
+ return n, nil, err
+}
+
+// Accept implements the linux syscall accept(2).
+func Accept(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.SyscallControl, error) {
+ fd := args[0].Int()
+ addr := args[1].Pointer()
+ addrlen := args[2].Pointer()
+
+ n, err := accept(t, fd, addr, addrlen, 0)
+ return n, nil, err
+}
+
+// Bind implements the linux syscall bind(2).
+func Bind(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.SyscallControl, error) {
+ fd := args[0].Int()
+ addr := args[1].Pointer()
+ addrlen := args[2].Uint()
+
+ // Get socket from the file descriptor.
+ file := t.GetFileVFS2(fd)
+ if file == nil {
+ return 0, nil, syserror.EBADF
+ }
+ defer file.DecRef()
+
+ // Extract the socket.
+ s, ok := file.Impl().(socket.SocketVFS2)
+ if !ok {
+ return 0, nil, syserror.ENOTSOCK
+ }
+
+ // Capture address and call syscall implementation.
+ a, err := CaptureAddress(t, addr, addrlen)
+ if err != nil {
+ return 0, nil, err
+ }
+
+ return 0, nil, s.Bind(t, a).ToError()
+}
+
+// Listen implements the linux syscall listen(2).
+func Listen(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.SyscallControl, error) {
+ fd := args[0].Int()
+ backlog := args[1].Int()
+
+ // Get socket from the file descriptor.
+ file := t.GetFileVFS2(fd)
+ if file == nil {
+ return 0, nil, syserror.EBADF
+ }
+ defer file.DecRef()
+
+ // Extract the socket.
+ s, ok := file.Impl().(socket.SocketVFS2)
+ if !ok {
+ return 0, nil, syserror.ENOTSOCK
+ }
+
+ // Per Linux, the backlog is silently capped to reasonable values.
+ if backlog <= 0 {
+ backlog = minListenBacklog
+ }
+ if backlog > maxListenBacklog {
+ backlog = maxListenBacklog
+ }
+
+ return 0, nil, s.Listen(t, int(backlog)).ToError()
+}
+
+// Shutdown implements the linux syscall shutdown(2).
+func Shutdown(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.SyscallControl, error) {
+ fd := args[0].Int()
+ how := args[1].Int()
+
+ // Get socket from the file descriptor.
+ file := t.GetFileVFS2(fd)
+ if file == nil {
+ return 0, nil, syserror.EBADF
+ }
+ defer file.DecRef()
+
+ // Extract the socket.
+ s, ok := file.Impl().(socket.SocketVFS2)
+ if !ok {
+ return 0, nil, syserror.ENOTSOCK
+ }
+
+ // Validate how, then call syscall implementation.
+ switch how {
+ case linux.SHUT_RD, linux.SHUT_WR, linux.SHUT_RDWR:
+ default:
+ return 0, nil, syserror.EINVAL
+ }
+
+ return 0, nil, s.Shutdown(t, int(how)).ToError()
+}
+
+// GetSockOpt implements the linux syscall getsockopt(2).
+func GetSockOpt(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.SyscallControl, error) {
+ fd := args[0].Int()
+ level := args[1].Int()
+ name := args[2].Int()
+ optValAddr := args[3].Pointer()
+ optLenAddr := args[4].Pointer()
+
+ // Get socket from the file descriptor.
+ file := t.GetFileVFS2(fd)
+ if file == nil {
+ return 0, nil, syserror.EBADF
+ }
+ defer file.DecRef()
+
+ // Extract the socket.
+ s, ok := file.Impl().(socket.SocketVFS2)
+ if !ok {
+ return 0, nil, syserror.ENOTSOCK
+ }
+
+ // Read the length. Reject negative values.
+ optLen := int32(0)
+ if _, err := t.CopyIn(optLenAddr, &optLen); err != nil {
+ return 0, nil, err
+ }
+ if optLen < 0 {
+ return 0, nil, syserror.EINVAL
+ }
+
+ // Call syscall implementation then copy both value and value len out.
+ v, e := getSockOpt(t, s, int(level), int(name), optValAddr, int(optLen))
+ if e != nil {
+ return 0, nil, e.ToError()
+ }
+
+ vLen := int32(binary.Size(v))
+ if _, err := t.CopyOut(optLenAddr, vLen); err != nil {
+ return 0, nil, err
+ }
+
+ if v != nil {
+ if _, err := t.CopyOut(optValAddr, v); err != nil {
+ return 0, nil, err
+ }
+ }
+
+ return 0, nil, nil
+}
+
+// getSockOpt tries to handle common socket options, or dispatches to a specific
+// socket implementation.
+func getSockOpt(t *kernel.Task, s socket.SocketVFS2, level, name int, optValAddr usermem.Addr, len int) (interface{}, *syserr.Error) {
+ if level == linux.SOL_SOCKET {
+ switch name {
+ case linux.SO_TYPE, linux.SO_DOMAIN, linux.SO_PROTOCOL:
+ if len < sizeOfInt32 {
+ return nil, syserr.ErrInvalidArgument
+ }
+ }
+
+ switch name {
+ case linux.SO_TYPE:
+ _, skType, _ := s.Type()
+ return int32(skType), nil
+ case linux.SO_DOMAIN:
+ family, _, _ := s.Type()
+ return int32(family), nil
+ case linux.SO_PROTOCOL:
+ _, _, protocol := s.Type()
+ return int32(protocol), nil
+ }
+ }
+
+ return s.GetSockOpt(t, level, name, optValAddr, len)
+}
+
+// SetSockOpt implements the linux syscall setsockopt(2).
+//
+// Note that unlike Linux, enabling SO_PASSCRED does not autobind the socket.
+func SetSockOpt(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.SyscallControl, error) {
+ fd := args[0].Int()
+ level := args[1].Int()
+ name := args[2].Int()
+ optValAddr := args[3].Pointer()
+ optLen := args[4].Int()
+
+ // Get socket from the file descriptor.
+ file := t.GetFileVFS2(fd)
+ if file == nil {
+ return 0, nil, syserror.EBADF
+ }
+ defer file.DecRef()
+
+ // Extract the socket.
+ s, ok := file.Impl().(socket.SocketVFS2)
+ if !ok {
+ return 0, nil, syserror.ENOTSOCK
+ }
+
+ if optLen < 0 {
+ return 0, nil, syserror.EINVAL
+ }
+ if optLen > maxOptLen {
+ return 0, nil, syserror.EINVAL
+ }
+ buf := t.CopyScratchBuffer(int(optLen))
+ if _, err := t.CopyIn(optValAddr, &buf); err != nil {
+ return 0, nil, err
+ }
+
+ // Call syscall implementation.
+ if err := s.SetSockOpt(t, int(level), int(name), buf); err != nil {
+ return 0, nil, err.ToError()
+ }
+
+ return 0, nil, nil
+}
+
+// GetSockName implements the linux syscall getsockname(2).
+func GetSockName(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.SyscallControl, error) {
+ fd := args[0].Int()
+ addr := args[1].Pointer()
+ addrlen := args[2].Pointer()
+
+ // Get socket from the file descriptor.
+ file := t.GetFileVFS2(fd)
+ if file == nil {
+ return 0, nil, syserror.EBADF
+ }
+ defer file.DecRef()
+
+ // Extract the socket.
+ s, ok := file.Impl().(socket.SocketVFS2)
+ if !ok {
+ return 0, nil, syserror.ENOTSOCK
+ }
+
+ // Get the socket name and copy it to the caller.
+ v, vl, err := s.GetSockName(t)
+ if err != nil {
+ return 0, nil, err.ToError()
+ }
+
+ return 0, nil, writeAddress(t, v, vl, addr, addrlen)
+}
+
+// GetPeerName implements the linux syscall getpeername(2).
+func GetPeerName(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.SyscallControl, error) {
+ fd := args[0].Int()
+ addr := args[1].Pointer()
+ addrlen := args[2].Pointer()
+
+ // Get socket from the file descriptor.
+ file := t.GetFileVFS2(fd)
+ if file == nil {
+ return 0, nil, syserror.EBADF
+ }
+ defer file.DecRef()
+
+ // Extract the socket.
+ s, ok := file.Impl().(socket.SocketVFS2)
+ if !ok {
+ return 0, nil, syserror.ENOTSOCK
+ }
+
+ // Get the socket peer name and copy it to the caller.
+ v, vl, err := s.GetPeerName(t)
+ if err != nil {
+ return 0, nil, err.ToError()
+ }
+
+ return 0, nil, writeAddress(t, v, vl, addr, addrlen)
+}
+
+// RecvMsg implements the linux syscall recvmsg(2).
+func RecvMsg(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.SyscallControl, error) {
+ fd := args[0].Int()
+ msgPtr := args[1].Pointer()
+ flags := args[2].Int()
+
+ if t.Arch().Width() != 8 {
+ // We only handle 64-bit for now.
+ return 0, nil, syserror.EINVAL
+ }
+
+ // Get socket from the file descriptor.
+ file := t.GetFileVFS2(fd)
+ if file == nil {
+ return 0, nil, syserror.EBADF
+ }
+ defer file.DecRef()
+
+ // Extract the socket.
+ s, ok := file.Impl().(socket.SocketVFS2)
+ if !ok {
+ return 0, nil, syserror.ENOTSOCK
+ }
+
+ // Reject flags that we don't handle yet.
+ if flags & ^(baseRecvFlags|linux.MSG_PEEK|linux.MSG_CMSG_CLOEXEC|linux.MSG_ERRQUEUE) != 0 {
+ return 0, nil, syserror.EINVAL
+ }
+
+ if (file.StatusFlags() & linux.SOCK_NONBLOCK) != 0 {
+ flags |= linux.MSG_DONTWAIT
+ }
+
+ var haveDeadline bool
+ var deadline ktime.Time
+ if dl := s.RecvTimeout(); dl > 0 {
+ deadline = t.Kernel().MonotonicClock().Now().Add(time.Duration(dl) * time.Nanosecond)
+ haveDeadline = true
+ } else if dl < 0 {
+ flags |= linux.MSG_DONTWAIT
+ }
+
+ n, err := recvSingleMsg(t, s, msgPtr, flags, haveDeadline, deadline)
+ return n, nil, err
+}
+
+// RecvMMsg implements the linux syscall recvmmsg(2).
+func RecvMMsg(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.SyscallControl, error) {
+ fd := args[0].Int()
+ msgPtr := args[1].Pointer()
+ vlen := args[2].Uint()
+ flags := args[3].Int()
+ toPtr := args[4].Pointer()
+
+ if t.Arch().Width() != 8 {
+ // We only handle 64-bit for now.
+ return 0, nil, syserror.EINVAL
+ }
+
+ // Reject flags that we don't handle yet.
+ if flags & ^(baseRecvFlags|linux.MSG_CMSG_CLOEXEC|linux.MSG_ERRQUEUE) != 0 {
+ return 0, nil, syserror.EINVAL
+ }
+
+ // Get socket from the file descriptor.
+ file := t.GetFileVFS2(fd)
+ if file == nil {
+ return 0, nil, syserror.EBADF
+ }
+ defer file.DecRef()
+
+ // Extract the socket.
+ s, ok := file.Impl().(socket.SocketVFS2)
+ if !ok {
+ return 0, nil, syserror.ENOTSOCK
+ }
+
+ if (file.StatusFlags() & linux.SOCK_NONBLOCK) != 0 {
+ flags |= linux.MSG_DONTWAIT
+ }
+
+ var haveDeadline bool
+ var deadline ktime.Time
+ if toPtr != 0 {
+ var ts linux.Timespec
+ if _, err := ts.CopyIn(t, toPtr); err != nil {
+ return 0, nil, err
+ }
+ if !ts.Valid() {
+ return 0, nil, syserror.EINVAL
+ }
+ deadline = t.Kernel().MonotonicClock().Now().Add(ts.ToDuration())
+ haveDeadline = true
+ }
+
+ if !haveDeadline {
+ if dl := s.RecvTimeout(); dl > 0 {
+ deadline = t.Kernel().MonotonicClock().Now().Add(time.Duration(dl) * time.Nanosecond)
+ haveDeadline = true
+ } else if dl < 0 {
+ flags |= linux.MSG_DONTWAIT
+ }
+ }
+
+ var count uint32
+ var err error
+ for i := uint64(0); i < uint64(vlen); i++ {
+ mp, ok := msgPtr.AddLength(i * multipleMessageHeader64Len)
+ if !ok {
+ return 0, nil, syserror.EFAULT
+ }
+ var n uintptr
+ if n, err = recvSingleMsg(t, s, mp, flags, haveDeadline, deadline); err != nil {
+ break
+ }
+
+ // Copy the received length to the caller.
+ lp, ok := mp.AddLength(messageHeader64Len)
+ if !ok {
+ return 0, nil, syserror.EFAULT
+ }
+ if _, err = t.CopyOut(lp, uint32(n)); err != nil {
+ break
+ }
+ count++
+ }
+
+ if count == 0 {
+ return 0, nil, err
+ }
+ return uintptr(count), nil, nil
+}
+
+func recvSingleMsg(t *kernel.Task, s socket.SocketVFS2, msgPtr usermem.Addr, flags int32, haveDeadline bool, deadline ktime.Time) (uintptr, error) {
+ // Capture the message header and io vectors.
+ var msg MessageHeader64
+ if err := CopyInMessageHeader64(t, msgPtr, &msg); err != nil {
+ return 0, err
+ }
+
+ if msg.IovLen > linux.UIO_MAXIOV {
+ return 0, syserror.EMSGSIZE
+ }
+ dst, err := t.IovecsIOSequence(usermem.Addr(msg.Iov), int(msg.IovLen), usermem.IOOpts{
+ AddressSpaceActive: true,
+ })
+ if err != nil {
+ return 0, err
+ }
+
+ // FIXME(b/63594852): Pretend we have an empty error queue.
+ if flags&linux.MSG_ERRQUEUE != 0 {
+ return 0, syserror.EAGAIN
+ }
+
+ // Fast path when no control message nor name buffers are provided.
+ if msg.ControlLen == 0 && msg.NameLen == 0 {
+ n, mflags, _, _, cms, err := s.RecvMsg(t, dst, int(flags), haveDeadline, deadline, false, 0)
+ if err != nil {
+ return 0, syserror.ConvertIntr(err.ToError(), kernel.ERESTARTSYS)
+ }
+ if !cms.Unix.Empty() {
+ mflags |= linux.MSG_CTRUNC
+ cms.Release()
+ }
+
+ if int(msg.Flags) != mflags {
+ // Copy out the flags to the caller.
+ if _, err := t.CopyOut(msgPtr+flagsOffset, int32(mflags)); err != nil {
+ return 0, err
+ }
+ }
+
+ return uintptr(n), nil
+ }
+
+ if msg.ControlLen > maxControlLen {
+ return 0, syserror.ENOBUFS
+ }
+ n, mflags, sender, senderLen, cms, e := s.RecvMsg(t, dst, int(flags), haveDeadline, deadline, msg.NameLen != 0, msg.ControlLen)
+ if e != nil {
+ return 0, syserror.ConvertIntr(e.ToError(), kernel.ERESTARTSYS)
+ }
+ defer cms.Release()
+
+ controlData := make([]byte, 0, msg.ControlLen)
+ controlData = control.PackControlMessages(t, cms, controlData)
+
+ if cr, ok := s.(transport.Credentialer); ok && cr.Passcred() {
+ creds, _ := cms.Unix.Credentials.(control.SCMCredentials)
+ controlData, mflags = control.PackCredentials(t, creds, controlData, mflags)
+ }
+
+ if cms.Unix.Rights != nil {
+ controlData, mflags = control.PackRights(t, cms.Unix.Rights.(control.SCMRights), flags&linux.MSG_CMSG_CLOEXEC != 0, controlData, mflags)
+ }
+
+ // Copy the address to the caller.
+ if msg.NameLen != 0 {
+ if err := writeAddress(t, sender, senderLen, usermem.Addr(msg.Name), usermem.Addr(msgPtr+nameLenOffset)); err != nil {
+ return 0, err
+ }
+ }
+
+ // Copy the control data to the caller.
+ if _, err := t.CopyOut(msgPtr+controlLenOffset, uint64(len(controlData))); err != nil {
+ return 0, err
+ }
+ if len(controlData) > 0 {
+ if _, err := t.CopyOut(usermem.Addr(msg.Control), controlData); err != nil {
+ return 0, err
+ }
+ }
+
+ // Copy out the flags to the caller.
+ if _, err := t.CopyOut(msgPtr+flagsOffset, int32(mflags)); err != nil {
+ return 0, err
+ }
+
+ return uintptr(n), nil
+}
+
+// recvFrom is the implementation of the recvfrom syscall. It is called by
+// recvfrom and recv syscall handlers.
+func recvFrom(t *kernel.Task, fd int32, bufPtr usermem.Addr, bufLen uint64, flags int32, namePtr usermem.Addr, nameLenPtr usermem.Addr) (uintptr, error) {
+ if int(bufLen) < 0 {
+ return 0, syserror.EINVAL
+ }
+
+ // Reject flags that we don't handle yet.
+ if flags & ^(baseRecvFlags|linux.MSG_PEEK|linux.MSG_CONFIRM) != 0 {
+ return 0, syserror.EINVAL
+ }
+
+ // Get socket from the file descriptor.
+ file := t.GetFileVFS2(fd)
+ if file == nil {
+ return 0, syserror.EBADF
+ }
+ defer file.DecRef()
+
+ // Extract the socket.
+ s, ok := file.Impl().(socket.SocketVFS2)
+ if !ok {
+ return 0, syserror.ENOTSOCK
+ }
+
+ if (file.StatusFlags() & linux.SOCK_NONBLOCK) != 0 {
+ flags |= linux.MSG_DONTWAIT
+ }
+
+ dst, err := t.SingleIOSequence(bufPtr, int(bufLen), usermem.IOOpts{
+ AddressSpaceActive: true,
+ })
+ if err != nil {
+ return 0, err
+ }
+
+ var haveDeadline bool
+ var deadline ktime.Time
+ if dl := s.RecvTimeout(); dl > 0 {
+ deadline = t.Kernel().MonotonicClock().Now().Add(time.Duration(dl) * time.Nanosecond)
+ haveDeadline = true
+ } else if dl < 0 {
+ flags |= linux.MSG_DONTWAIT
+ }
+
+ n, _, sender, senderLen, cm, e := s.RecvMsg(t, dst, int(flags), haveDeadline, deadline, nameLenPtr != 0, 0)
+ cm.Release()
+ if e != nil {
+ return 0, syserror.ConvertIntr(e.ToError(), kernel.ERESTARTSYS)
+ }
+
+ // Copy the address to the caller.
+ if nameLenPtr != 0 {
+ if err := writeAddress(t, sender, senderLen, namePtr, nameLenPtr); err != nil {
+ return 0, err
+ }
+ }
+
+ return uintptr(n), nil
+}
+
+// RecvFrom implements the linux syscall recvfrom(2).
+func RecvFrom(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.SyscallControl, error) {
+ fd := args[0].Int()
+ bufPtr := args[1].Pointer()
+ bufLen := args[2].Uint64()
+ flags := args[3].Int()
+ namePtr := args[4].Pointer()
+ nameLenPtr := args[5].Pointer()
+
+ n, err := recvFrom(t, fd, bufPtr, bufLen, flags, namePtr, nameLenPtr)
+ return n, nil, err
+}
+
+// SendMsg implements the linux syscall sendmsg(2).
+func SendMsg(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.SyscallControl, error) {
+ fd := args[0].Int()
+ msgPtr := args[1].Pointer()
+ flags := args[2].Int()
+
+ if t.Arch().Width() != 8 {
+ // We only handle 64-bit for now.
+ return 0, nil, syserror.EINVAL
+ }
+
+ // Get socket from the file descriptor.
+ file := t.GetFileVFS2(fd)
+ if file == nil {
+ return 0, nil, syserror.EBADF
+ }
+ defer file.DecRef()
+
+ // Extract the socket.
+ s, ok := file.Impl().(socket.SocketVFS2)
+ if !ok {
+ return 0, nil, syserror.ENOTSOCK
+ }
+
+ // Reject flags that we don't handle yet.
+ if flags & ^(linux.MSG_DONTWAIT|linux.MSG_EOR|linux.MSG_MORE|linux.MSG_NOSIGNAL) != 0 {
+ return 0, nil, syserror.EINVAL
+ }
+
+ if (file.StatusFlags() & linux.SOCK_NONBLOCK) != 0 {
+ flags |= linux.MSG_DONTWAIT
+ }
+
+ n, err := sendSingleMsg(t, s, file, msgPtr, flags)
+ return n, nil, err
+}
+
+// SendMMsg implements the linux syscall sendmmsg(2).
+func SendMMsg(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.SyscallControl, error) {
+ fd := args[0].Int()
+ msgPtr := args[1].Pointer()
+ vlen := args[2].Uint()
+ flags := args[3].Int()
+
+ if t.Arch().Width() != 8 {
+ // We only handle 64-bit for now.
+ return 0, nil, syserror.EINVAL
+ }
+
+ // Get socket from the file descriptor.
+ file := t.GetFileVFS2(fd)
+ if file == nil {
+ return 0, nil, syserror.EBADF
+ }
+ defer file.DecRef()
+
+ // Extract the socket.
+ s, ok := file.Impl().(socket.SocketVFS2)
+ if !ok {
+ return 0, nil, syserror.ENOTSOCK
+ }
+
+ // Reject flags that we don't handle yet.
+ if flags & ^(linux.MSG_DONTWAIT|linux.MSG_EOR|linux.MSG_MORE|linux.MSG_NOSIGNAL) != 0 {
+ return 0, nil, syserror.EINVAL
+ }
+
+ if (file.StatusFlags() & linux.SOCK_NONBLOCK) != 0 {
+ flags |= linux.MSG_DONTWAIT
+ }
+
+ var count uint32
+ var err error
+ for i := uint64(0); i < uint64(vlen); i++ {
+ mp, ok := msgPtr.AddLength(i * multipleMessageHeader64Len)
+ if !ok {
+ return 0, nil, syserror.EFAULT
+ }
+ var n uintptr
+ if n, err = sendSingleMsg(t, s, file, mp, flags); err != nil {
+ break
+ }
+
+ // Copy the received length to the caller.
+ lp, ok := mp.AddLength(messageHeader64Len)
+ if !ok {
+ return 0, nil, syserror.EFAULT
+ }
+ if _, err = t.CopyOut(lp, uint32(n)); err != nil {
+ break
+ }
+ count++
+ }
+
+ if count == 0 {
+ return 0, nil, err
+ }
+ return uintptr(count), nil, nil
+}
+
+func sendSingleMsg(t *kernel.Task, s socket.SocketVFS2, file *vfs.FileDescription, msgPtr usermem.Addr, flags int32) (uintptr, error) {
+ // Capture the message header.
+ var msg MessageHeader64
+ if err := CopyInMessageHeader64(t, msgPtr, &msg); err != nil {
+ return 0, err
+ }
+
+ var controlData []byte
+ if msg.ControlLen > 0 {
+ // Put an upper bound to prevent large allocations.
+ if msg.ControlLen > maxControlLen {
+ return 0, syserror.ENOBUFS
+ }
+ controlData = make([]byte, msg.ControlLen)
+ if _, err := t.CopyIn(usermem.Addr(msg.Control), &controlData); err != nil {
+ return 0, err
+ }
+ }
+
+ // Read the destination address if one is specified.
+ var to []byte
+ if msg.NameLen != 0 {
+ var err error
+ to, err = CaptureAddress(t, usermem.Addr(msg.Name), msg.NameLen)
+ if err != nil {
+ return 0, err
+ }
+ }
+
+ // Read data then call the sendmsg implementation.
+ if msg.IovLen > linux.UIO_MAXIOV {
+ return 0, syserror.EMSGSIZE
+ }
+ src, err := t.IovecsIOSequence(usermem.Addr(msg.Iov), int(msg.IovLen), usermem.IOOpts{
+ AddressSpaceActive: true,
+ })
+ if err != nil {
+ return 0, err
+ }
+
+ controlMessages, err := control.Parse(t, s, controlData)
+ if err != nil {
+ return 0, err
+ }
+
+ var haveDeadline bool
+ var deadline ktime.Time
+ if dl := s.SendTimeout(); dl > 0 {
+ deadline = t.Kernel().MonotonicClock().Now().Add(time.Duration(dl) * time.Nanosecond)
+ haveDeadline = true
+ } else if dl < 0 {
+ flags |= linux.MSG_DONTWAIT
+ }
+
+ // Call the syscall implementation.
+ n, e := s.SendMsg(t, src, to, int(flags), haveDeadline, deadline, controlMessages)
+ err = slinux.HandleIOErrorVFS2(t, n != 0, e.ToError(), kernel.ERESTARTSYS, "sendmsg", file)
+ if err != nil {
+ controlMessages.Release()
+ }
+ return uintptr(n), err
+}
+
+// sendTo is the implementation of the sendto syscall. It is called by sendto
+// and send syscall handlers.
+func sendTo(t *kernel.Task, fd int32, bufPtr usermem.Addr, bufLen uint64, flags int32, namePtr usermem.Addr, nameLen uint32) (uintptr, error) {
+ bl := int(bufLen)
+ if bl < 0 {
+ return 0, syserror.EINVAL
+ }
+
+ // Get socket from the file descriptor.
+ file := t.GetFileVFS2(fd)
+ if file == nil {
+ return 0, syserror.EBADF
+ }
+ defer file.DecRef()
+
+ // Extract the socket.
+ s, ok := file.Impl().(socket.SocketVFS2)
+ if !ok {
+ return 0, syserror.ENOTSOCK
+ }
+
+ if (file.StatusFlags() & linux.SOCK_NONBLOCK) != 0 {
+ flags |= linux.MSG_DONTWAIT
+ }
+
+ // Read the destination address if one is specified.
+ var to []byte
+ var err error
+ if namePtr != 0 {
+ to, err = CaptureAddress(t, namePtr, nameLen)
+ if err != nil {
+ return 0, err
+ }
+ }
+
+ src, err := t.SingleIOSequence(bufPtr, bl, usermem.IOOpts{
+ AddressSpaceActive: true,
+ })
+ if err != nil {
+ return 0, err
+ }
+
+ var haveDeadline bool
+ var deadline ktime.Time
+ if dl := s.SendTimeout(); dl > 0 {
+ deadline = t.Kernel().MonotonicClock().Now().Add(time.Duration(dl) * time.Nanosecond)
+ haveDeadline = true
+ } else if dl < 0 {
+ flags |= linux.MSG_DONTWAIT
+ }
+
+ // Call the syscall implementation.
+ n, e := s.SendMsg(t, src, to, int(flags), haveDeadline, deadline, socket.ControlMessages{Unix: control.New(t, s, nil)})
+ return uintptr(n), slinux.HandleIOErrorVFS2(t, n != 0, e.ToError(), kernel.ERESTARTSYS, "sendto", file)
+}
+
+// SendTo implements the linux syscall sendto(2).
+func SendTo(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.SyscallControl, error) {
+ fd := args[0].Int()
+ bufPtr := args[1].Pointer()
+ bufLen := args[2].Uint64()
+ flags := args[3].Int()
+ namePtr := args[4].Pointer()
+ nameLen := args[5].Uint()
+
+ n, err := sendTo(t, fd, bufPtr, bufLen, flags, namePtr, nameLen)
+ return n, nil, err
+}
diff --git a/pkg/sentry/syscalls/linux/vfs2/xattr.go b/pkg/sentry/syscalls/linux/vfs2/xattr.go
index 89e9ff4d7..af455d5c1 100644
--- a/pkg/sentry/syscalls/linux/vfs2/xattr.go
+++ b/pkg/sentry/syscalls/linux/vfs2/xattr.go
@@ -51,7 +51,7 @@ func listxattr(t *kernel.Task, args arch.SyscallArguments, shouldFollowFinalSyml
}
defer tpop.Release()
- names, err := t.Kernel().VFS().ListxattrAt(t, t.Credentials(), &tpop.pop)
+ names, err := t.Kernel().VFS().ListxattrAt(t, t.Credentials(), &tpop.pop, uint64(size))
if err != nil {
return 0, nil, err
}
@@ -74,7 +74,7 @@ func Flistxattr(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Sy
}
defer file.DecRef()
- names, err := file.Listxattr(t)
+ names, err := file.Listxattr(t, uint64(size))
if err != nil {
return 0, nil, err
}
@@ -116,7 +116,10 @@ func getxattr(t *kernel.Task, args arch.SyscallArguments, shouldFollowFinalSymli
return 0, nil, err
}
- value, err := t.Kernel().VFS().GetxattrAt(t, t.Credentials(), &tpop.pop, name)
+ value, err := t.Kernel().VFS().GetxattrAt(t, t.Credentials(), &tpop.pop, &vfs.GetxattrOptions{
+ Name: name,
+ Size: uint64(size),
+ })
if err != nil {
return 0, nil, err
}
@@ -145,7 +148,7 @@ func Fgetxattr(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Sys
return 0, nil, err
}
- value, err := file.Getxattr(t, name)
+ value, err := file.Getxattr(t, &vfs.GetxattrOptions{Name: name, Size: uint64(size)})
if err != nil {
return 0, nil, err
}
@@ -230,7 +233,7 @@ func Fsetxattr(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Sys
return 0, nil, err
}
- return 0, nil, file.Setxattr(t, vfs.SetxattrOptions{
+ return 0, nil, file.Setxattr(t, &vfs.SetxattrOptions{
Name: name,
Value: value,
Flags: uint32(flags),
diff --git a/pkg/sentry/vfs/anonfs.go b/pkg/sentry/vfs/anonfs.go
index d1f6dfb45..a64d86122 100644
--- a/pkg/sentry/vfs/anonfs.go
+++ b/pkg/sentry/vfs/anonfs.go
@@ -245,7 +245,7 @@ func (fs *anonFilesystem) BoundEndpointAt(ctx context.Context, rp *ResolvingPath
}
// ListxattrAt implements FilesystemImpl.ListxattrAt.
-func (fs *anonFilesystem) ListxattrAt(ctx context.Context, rp *ResolvingPath) ([]string, error) {
+func (fs *anonFilesystem) ListxattrAt(ctx context.Context, rp *ResolvingPath, size uint64) ([]string, error) {
if !rp.Done() {
return nil, syserror.ENOTDIR
}
@@ -253,7 +253,7 @@ func (fs *anonFilesystem) ListxattrAt(ctx context.Context, rp *ResolvingPath) ([
}
// GetxattrAt implements FilesystemImpl.GetxattrAt.
-func (fs *anonFilesystem) GetxattrAt(ctx context.Context, rp *ResolvingPath, name string) (string, error) {
+func (fs *anonFilesystem) GetxattrAt(ctx context.Context, rp *ResolvingPath, opts GetxattrOptions) (string, error) {
if !rp.Done() {
return "", syserror.ENOTDIR
}
diff --git a/pkg/sentry/vfs/file_description.go b/pkg/sentry/vfs/file_description.go
index 20c545fca..4fb9aea87 100644
--- a/pkg/sentry/vfs/file_description.go
+++ b/pkg/sentry/vfs/file_description.go
@@ -401,11 +401,11 @@ type FileDescriptionImpl interface {
Ioctl(ctx context.Context, uio usermem.IO, args arch.SyscallArguments) (uintptr, error)
// Listxattr returns all extended attribute names for the file.
- Listxattr(ctx context.Context) ([]string, error)
+ Listxattr(ctx context.Context, size uint64) ([]string, error)
// Getxattr returns the value associated with the given extended attribute
// for the file.
- Getxattr(ctx context.Context, name string) (string, error)
+ Getxattr(ctx context.Context, opts GetxattrOptions) (string, error)
// Setxattr changes the value associated with the given extended attribute
// for the file.
@@ -605,18 +605,23 @@ func (fd *FileDescription) Ioctl(ctx context.Context, uio usermem.IO, args arch.
// Listxattr returns all extended attribute names for the file represented by
// fd.
-func (fd *FileDescription) Listxattr(ctx context.Context) ([]string, error) {
+//
+// If the size of the list (including a NUL terminating byte after every entry)
+// would exceed size, ERANGE may be returned. Note that implementations
+// are free to ignore size entirely and return without error). In all cases,
+// if size is 0, the list should be returned without error, regardless of size.
+func (fd *FileDescription) Listxattr(ctx context.Context, size uint64) ([]string, error) {
if fd.opts.UseDentryMetadata {
vfsObj := fd.vd.mount.vfs
rp := vfsObj.getResolvingPath(auth.CredentialsFromContext(ctx), &PathOperation{
Root: fd.vd,
Start: fd.vd,
})
- names, err := fd.vd.mount.fs.impl.ListxattrAt(ctx, rp)
+ names, err := fd.vd.mount.fs.impl.ListxattrAt(ctx, rp, size)
vfsObj.putResolvingPath(rp)
return names, err
}
- names, err := fd.impl.Listxattr(ctx)
+ names, err := fd.impl.Listxattr(ctx, size)
if err == syserror.ENOTSUP {
// Linux doesn't actually return ENOTSUP in this case; instead,
// fs/xattr.c:vfs_listxattr() falls back to allowing the security
@@ -629,34 +634,39 @@ func (fd *FileDescription) Listxattr(ctx context.Context) ([]string, error) {
// Getxattr returns the value associated with the given extended attribute for
// the file represented by fd.
-func (fd *FileDescription) Getxattr(ctx context.Context, name string) (string, error) {
+//
+// If the size of the return value exceeds opts.Size, ERANGE may be returned
+// (note that implementations are free to ignore opts.Size entirely and return
+// without error). In all cases, if opts.Size is 0, the value should be
+// returned without error, regardless of size.
+func (fd *FileDescription) Getxattr(ctx context.Context, opts *GetxattrOptions) (string, error) {
if fd.opts.UseDentryMetadata {
vfsObj := fd.vd.mount.vfs
rp := vfsObj.getResolvingPath(auth.CredentialsFromContext(ctx), &PathOperation{
Root: fd.vd,
Start: fd.vd,
})
- val, err := fd.vd.mount.fs.impl.GetxattrAt(ctx, rp, name)
+ val, err := fd.vd.mount.fs.impl.GetxattrAt(ctx, rp, *opts)
vfsObj.putResolvingPath(rp)
return val, err
}
- return fd.impl.Getxattr(ctx, name)
+ return fd.impl.Getxattr(ctx, *opts)
}
// Setxattr changes the value associated with the given extended attribute for
// the file represented by fd.
-func (fd *FileDescription) Setxattr(ctx context.Context, opts SetxattrOptions) error {
+func (fd *FileDescription) Setxattr(ctx context.Context, opts *SetxattrOptions) error {
if fd.opts.UseDentryMetadata {
vfsObj := fd.vd.mount.vfs
rp := vfsObj.getResolvingPath(auth.CredentialsFromContext(ctx), &PathOperation{
Root: fd.vd,
Start: fd.vd,
})
- err := fd.vd.mount.fs.impl.SetxattrAt(ctx, rp, opts)
+ err := fd.vd.mount.fs.impl.SetxattrAt(ctx, rp, *opts)
vfsObj.putResolvingPath(rp)
return err
}
- return fd.impl.Setxattr(ctx, opts)
+ return fd.impl.Setxattr(ctx, *opts)
}
// Removexattr removes the given extended attribute from the file represented
diff --git a/pkg/sentry/vfs/file_description_impl_util.go b/pkg/sentry/vfs/file_description_impl_util.go
index d45e602ce..f4c111926 100644
--- a/pkg/sentry/vfs/file_description_impl_util.go
+++ b/pkg/sentry/vfs/file_description_impl_util.go
@@ -130,14 +130,14 @@ func (FileDescriptionDefaultImpl) Ioctl(ctx context.Context, uio usermem.IO, arg
// Listxattr implements FileDescriptionImpl.Listxattr analogously to
// inode_operations::listxattr == NULL in Linux.
-func (FileDescriptionDefaultImpl) Listxattr(ctx context.Context) ([]string, error) {
+func (FileDescriptionDefaultImpl) Listxattr(ctx context.Context, size uint64) ([]string, error) {
// This isn't exactly accurate; see FileDescription.Listxattr.
return nil, syserror.ENOTSUP
}
// Getxattr implements FileDescriptionImpl.Getxattr analogously to
// inode::i_opflags & IOP_XATTR == 0 in Linux.
-func (FileDescriptionDefaultImpl) Getxattr(ctx context.Context, name string) (string, error) {
+func (FileDescriptionDefaultImpl) Getxattr(ctx context.Context, opts GetxattrOptions) (string, error) {
return "", syserror.ENOTSUP
}
diff --git a/pkg/sentry/vfs/filesystem.go b/pkg/sentry/vfs/filesystem.go
index cd34782ff..a537a29d1 100644
--- a/pkg/sentry/vfs/filesystem.go
+++ b/pkg/sentry/vfs/filesystem.go
@@ -442,7 +442,13 @@ type FilesystemImpl interface {
// - If extended attributes are not supported by the filesystem,
// ListxattrAt returns nil. (See FileDescription.Listxattr for an
// explanation.)
- ListxattrAt(ctx context.Context, rp *ResolvingPath) ([]string, error)
+ //
+ // - If the size of the list (including a NUL terminating byte after every
+ // entry) would exceed size, ERANGE may be returned. Note that
+ // implementations are free to ignore size entirely and return without
+ // error). In all cases, if size is 0, the list should be returned without
+ // error, regardless of size.
+ ListxattrAt(ctx context.Context, rp *ResolvingPath, size uint64) ([]string, error)
// GetxattrAt returns the value associated with the given extended
// attribute for the file at rp.
@@ -451,7 +457,15 @@ type FilesystemImpl interface {
//
// - If extended attributes are not supported by the filesystem, GetxattrAt
// returns ENOTSUP.
- GetxattrAt(ctx context.Context, rp *ResolvingPath, name string) (string, error)
+ //
+ // - If an extended attribute named opts.Name does not exist, ENODATA is
+ // returned.
+ //
+ // - If the size of the return value exceeds opts.Size, ERANGE may be
+ // returned (note that implementations are free to ignore opts.Size entirely
+ // and return without error). In all cases, if opts.Size is 0, the value
+ // should be returned without error, regardless of size.
+ GetxattrAt(ctx context.Context, rp *ResolvingPath, opts GetxattrOptions) (string, error)
// SetxattrAt changes the value associated with the given extended
// attribute for the file at rp.
@@ -460,6 +474,10 @@ type FilesystemImpl interface {
//
// - If extended attributes are not supported by the filesystem, SetxattrAt
// returns ENOTSUP.
+ //
+ // - If XATTR_CREATE is set in opts.Flag and opts.Name already exists,
+ // EEXIST is returned. If XATTR_REPLACE is set and opts.Name does not exist,
+ // ENODATA is returned.
SetxattrAt(ctx context.Context, rp *ResolvingPath, opts SetxattrOptions) error
// RemovexattrAt removes the given extended attribute from the file at rp.
@@ -468,6 +486,8 @@ type FilesystemImpl interface {
//
// - If extended attributes are not supported by the filesystem,
// RemovexattrAt returns ENOTSUP.
+ //
+ // - If name does not exist, ENODATA is returned.
RemovexattrAt(ctx context.Context, rp *ResolvingPath, name string) error
// BoundEndpointAt returns the Unix socket endpoint bound at the path rp.
@@ -497,7 +517,7 @@ type FilesystemImpl interface {
// Preconditions: vd.Mount().Filesystem().Impl() == this FilesystemImpl.
PrependPath(ctx context.Context, vfsroot, vd VirtualDentry, b *fspath.Builder) error
- // TODO: inotify_add_watch()
+ // TODO(gvisor.dev/issue/1479): inotify_add_watch()
}
// PrependPathAtVFSRootError is returned by implementations of
diff --git a/pkg/sentry/vfs/memxattr/BUILD b/pkg/sentry/vfs/memxattr/BUILD
new file mode 100644
index 000000000..d8c4d27b9
--- /dev/null
+++ b/pkg/sentry/vfs/memxattr/BUILD
@@ -0,0 +1,15 @@
+load("//tools:defs.bzl", "go_library")
+
+package(licenses = ["notice"])
+
+go_library(
+ name = "memxattr",
+ srcs = ["xattr.go"],
+ visibility = ["//pkg/sentry:internal"],
+ deps = [
+ "//pkg/abi/linux",
+ "//pkg/sentry/vfs",
+ "//pkg/sync",
+ "//pkg/syserror",
+ ],
+)
diff --git a/pkg/sentry/vfs/memxattr/xattr.go b/pkg/sentry/vfs/memxattr/xattr.go
new file mode 100644
index 000000000..cc1e7d764
--- /dev/null
+++ b/pkg/sentry/vfs/memxattr/xattr.go
@@ -0,0 +1,102 @@
+// 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 memxattr provides a default, in-memory extended attribute
+// implementation.
+package memxattr
+
+import (
+ "gvisor.dev/gvisor/pkg/abi/linux"
+ "gvisor.dev/gvisor/pkg/sentry/vfs"
+ "gvisor.dev/gvisor/pkg/sync"
+ "gvisor.dev/gvisor/pkg/syserror"
+)
+
+// SimpleExtendedAttributes implements extended attributes using a map of
+// names to values.
+//
+// +stateify savable
+type SimpleExtendedAttributes struct {
+ // mu protects the below fields.
+ mu sync.RWMutex `state:"nosave"`
+ xattrs map[string]string
+}
+
+// Getxattr returns the value at 'name'.
+func (x *SimpleExtendedAttributes) Getxattr(opts *vfs.GetxattrOptions) (string, error) {
+ x.mu.RLock()
+ value, ok := x.xattrs[opts.Name]
+ x.mu.RUnlock()
+ if !ok {
+ return "", syserror.ENODATA
+ }
+ // Check that the size of the buffer provided in getxattr(2) is large enough
+ // to contain the value.
+ if opts.Size != 0 && uint64(len(value)) > opts.Size {
+ return "", syserror.ERANGE
+ }
+ return value, nil
+}
+
+// Setxattr sets 'value' at 'name'.
+func (x *SimpleExtendedAttributes) Setxattr(opts *vfs.SetxattrOptions) error {
+ x.mu.Lock()
+ defer x.mu.Unlock()
+ if x.xattrs == nil {
+ if opts.Flags&linux.XATTR_REPLACE != 0 {
+ return syserror.ENODATA
+ }
+ x.xattrs = make(map[string]string)
+ }
+
+ _, ok := x.xattrs[opts.Name]
+ if ok && opts.Flags&linux.XATTR_CREATE != 0 {
+ return syserror.EEXIST
+ }
+ if !ok && opts.Flags&linux.XATTR_REPLACE != 0 {
+ return syserror.ENODATA
+ }
+
+ x.xattrs[opts.Name] = opts.Value
+ return nil
+}
+
+// Listxattr returns all names in xattrs.
+func (x *SimpleExtendedAttributes) Listxattr(size uint64) ([]string, error) {
+ // Keep track of the size of the buffer needed in listxattr(2) for the list.
+ listSize := 0
+ x.mu.RLock()
+ names := make([]string, 0, len(x.xattrs))
+ for n := range x.xattrs {
+ names = append(names, n)
+ // Add one byte per null terminator.
+ listSize += len(n) + 1
+ }
+ x.mu.RUnlock()
+ if size != 0 && uint64(listSize) > size {
+ return nil, syserror.ERANGE
+ }
+ return names, nil
+}
+
+// Removexattr removes the xattr at 'name'.
+func (x *SimpleExtendedAttributes) Removexattr(name string) error {
+ x.mu.Lock()
+ defer x.mu.Unlock()
+ if _, ok := x.xattrs[name]; !ok {
+ return syserror.ENODATA
+ }
+ delete(x.xattrs, name)
+ return nil
+}
diff --git a/pkg/sentry/vfs/mount.go b/pkg/sentry/vfs/mount.go
index 1b8ecc415..f06946103 100644
--- a/pkg/sentry/vfs/mount.go
+++ b/pkg/sentry/vfs/mount.go
@@ -233,9 +233,9 @@ func (vfs *VirtualFilesystem) MountAt(ctx context.Context, creds *auth.Credentia
}
vd.dentry.mu.Lock()
}
- // TODO: Linux requires that either both the mount point and the mount root
- // are directories, or neither are, and returns ENOTDIR if this is not the
- // case.
+ // TODO(gvisor.dev/issue/1035): Linux requires that either both the mount
+ // point and the mount root are directories, or neither are, and returns
+ // ENOTDIR if this is not the case.
mntns := vd.mount.ns
mnt := newMount(vfs, fs, root, mntns, opts)
vfs.mounts.seq.BeginWrite()
@@ -274,9 +274,9 @@ func (vfs *VirtualFilesystem) UmountAt(ctx context.Context, creds *auth.Credenti
}
}
- // TODO(jamieliu): Linux special-cases umount of the caller's root, which
- // we don't implement yet (we'll just fail it since the caller holds a
- // reference on it).
+ // TODO(gvisor.dev/issue/1035): Linux special-cases umount of the caller's
+ // root, which we don't implement yet (we'll just fail it since the caller
+ // holds a reference on it).
vfs.mounts.seq.BeginWrite()
if opts.Flags&linux.MNT_DETACH == 0 {
diff --git a/pkg/sentry/vfs/mount_test.go b/pkg/sentry/vfs/mount_test.go
index 3b933468d..3335e4057 100644
--- a/pkg/sentry/vfs/mount_test.go
+++ b/pkg/sentry/vfs/mount_test.go
@@ -55,7 +55,7 @@ func TestMountTableInsertLookup(t *testing.T) {
}
}
-// TODO: concurrent lookup/insertion/removal
+// TODO(gvisor.dev/issue/1035): concurrent lookup/insertion/removal.
// must be powers of 2
var benchNumMounts = []int{1 << 2, 1 << 5, 1 << 8}
diff --git a/pkg/sentry/vfs/options.go b/pkg/sentry/vfs/options.go
index 2f04bf882..534528ce6 100644
--- a/pkg/sentry/vfs/options.go
+++ b/pkg/sentry/vfs/options.go
@@ -132,6 +132,20 @@ type SetStatOptions struct {
Stat linux.Statx
}
+// GetxattrOptions contains options to VirtualFilesystem.GetxattrAt(),
+// FilesystemImpl.GetxattrAt(), FileDescription.Getxattr(), and
+// FileDescriptionImpl.Getxattr().
+type GetxattrOptions struct {
+ // Name is the name of the extended attribute to retrieve.
+ Name string
+
+ // Size is the maximum value size that the caller will tolerate. If the value
+ // is larger than size, getxattr methods may return ERANGE, but they are also
+ // free to ignore the hint entirely (i.e. the value returned may be larger
+ // than size). All size checking is done independently at the syscall layer.
+ Size uint64
+}
+
// SetxattrOptions contains options to VirtualFilesystem.SetxattrAt(),
// FilesystemImpl.SetxattrAt(), FileDescription.Setxattr(), and
// FileDescriptionImpl.Setxattr().
diff --git a/pkg/sentry/vfs/vfs.go b/pkg/sentry/vfs/vfs.go
index 720b90d8f..f592913d5 100644
--- a/pkg/sentry/vfs/vfs.go
+++ b/pkg/sentry/vfs/vfs.go
@@ -680,10 +680,10 @@ func (vfs *VirtualFilesystem) UnlinkAt(ctx context.Context, creds *auth.Credenti
// ListxattrAt returns all extended attribute names for the file at the given
// path.
-func (vfs *VirtualFilesystem) ListxattrAt(ctx context.Context, creds *auth.Credentials, pop *PathOperation) ([]string, error) {
+func (vfs *VirtualFilesystem) ListxattrAt(ctx context.Context, creds *auth.Credentials, pop *PathOperation, size uint64) ([]string, error) {
rp := vfs.getResolvingPath(creds, pop)
for {
- names, err := rp.mount.fs.impl.ListxattrAt(ctx, rp)
+ names, err := rp.mount.fs.impl.ListxattrAt(ctx, rp, size)
if err == nil {
vfs.putResolvingPath(rp)
return names, nil
@@ -705,10 +705,10 @@ func (vfs *VirtualFilesystem) ListxattrAt(ctx context.Context, creds *auth.Crede
// GetxattrAt returns the value associated with the given extended attribute
// for the file at the given path.
-func (vfs *VirtualFilesystem) GetxattrAt(ctx context.Context, creds *auth.Credentials, pop *PathOperation, name string) (string, error) {
+func (vfs *VirtualFilesystem) GetxattrAt(ctx context.Context, creds *auth.Credentials, pop *PathOperation, opts *GetxattrOptions) (string, error) {
rp := vfs.getResolvingPath(creds, pop)
for {
- val, err := rp.mount.fs.impl.GetxattrAt(ctx, rp, name)
+ val, err := rp.mount.fs.impl.GetxattrAt(ctx, rp, *opts)
if err == nil {
vfs.putResolvingPath(rp)
return val, nil
diff --git a/pkg/state/state.go b/pkg/state/state.go
index dbe507ab4..03ae2dbb0 100644
--- a/pkg/state/state.go
+++ b/pkg/state/state.go
@@ -241,10 +241,7 @@ func Register(name string, instance interface{}, fns Fns) {
//
// This function is used by the stateify tool.
func IsZeroValue(val interface{}) bool {
- if val == nil {
- return true
- }
- return reflect.DeepEqual(val, reflect.Zero(reflect.TypeOf(val)).Interface())
+ return val == nil || reflect.ValueOf(val).Elem().IsZero()
}
// step captures one encoding / decoding step. On each step, there is up to one
diff --git a/pkg/tcpip/checker/checker.go b/pkg/tcpip/checker/checker.go
index 8dc0f7c0e..307f1b666 100644
--- a/pkg/tcpip/checker/checker.go
+++ b/pkg/tcpip/checker/checker.go
@@ -728,7 +728,7 @@ func ICMPv6Code(want byte) TransportChecker {
// message for type of ty, with potentially additional checks specified by
// checkers.
//
-// checkers may assume that a valid ICMPv6 is passed to it containing a valid
+// Checkers may assume that a valid ICMPv6 is passed to it containing a valid
// NDP message as far as the size of the message (minSize) is concerned. The
// values within the message are up to checkers to validate.
func NDP(msgType header.ICMPv6Type, minSize int, checkers ...TransportChecker) NetworkChecker {
@@ -760,9 +760,9 @@ func NDP(msgType header.ICMPv6Type, minSize int, checkers ...TransportChecker) N
// Neighbor Solicitation message (as per the raw wire format), with potentially
// additional checks specified by checkers.
//
-// checkers may assume that a valid ICMPv6 is passed to it containing a valid
-// NDPNS message as far as the size of the messages concerned. The values within
-// the message are up to checkers to validate.
+// Checkers may assume that a valid ICMPv6 is passed to it containing a valid
+// NDPNS message as far as the size of the message is concerned. The values
+// within the message are up to checkers to validate.
func NDPNS(checkers ...TransportChecker) NetworkChecker {
return NDP(header.ICMPv6NeighborSolicit, header.NDPNSMinimumSize, checkers...)
}
@@ -780,7 +780,54 @@ func NDPNSTargetAddress(want tcpip.Address) TransportChecker {
ns := header.NDPNeighborSolicit(icmp.NDPPayload())
if got := ns.TargetAddress(); got != want {
- t.Fatalf("got %T.TargetAddress = %s, want = %s", ns, got, want)
+ t.Errorf("got %T.TargetAddress() = %s, want = %s", ns, got, want)
+ }
+ }
+}
+
+// NDPNA creates a checker that checks that the packet contains a valid NDP
+// Neighbor Advertisement message (as per the raw wire format), with potentially
+// additional checks specified by checkers.
+//
+// Checkers may assume that a valid ICMPv6 is passed to it containing a valid
+// NDPNA message as far as the size of the message is concerned. The values
+// within the message are up to checkers to validate.
+func NDPNA(checkers ...TransportChecker) NetworkChecker {
+ return NDP(header.ICMPv6NeighborAdvert, header.NDPNAMinimumSize, checkers...)
+}
+
+// NDPNATargetAddress creates a checker that checks the Target Address field of
+// a header.NDPNeighborAdvert.
+//
+// The returned TransportChecker assumes that a valid ICMPv6 is passed to it
+// containing a valid NDPNA message as far as the size is concerned.
+func NDPNATargetAddress(want tcpip.Address) TransportChecker {
+ return func(t *testing.T, h header.Transport) {
+ t.Helper()
+
+ icmp := h.(header.ICMPv6)
+ na := header.NDPNeighborAdvert(icmp.NDPPayload())
+
+ if got := na.TargetAddress(); got != want {
+ t.Errorf("got %T.TargetAddress() = %s, want = %s", na, got, want)
+ }
+ }
+}
+
+// NDPNASolicitedFlag creates a checker that checks the Solicited field of
+// a header.NDPNeighborAdvert.
+//
+// The returned TransportChecker assumes that a valid ICMPv6 is passed to it
+// containing a valid NDPNA message as far as the size is concerned.
+func NDPNASolicitedFlag(want bool) TransportChecker {
+ return func(t *testing.T, h header.Transport) {
+ t.Helper()
+
+ icmp := h.(header.ICMPv6)
+ na := header.NDPNeighborAdvert(icmp.NDPPayload())
+
+ if got := na.SolicitedFlag(); got != want {
+ t.Errorf("got %T.SolicitedFlag = %t, want = %t", na, got, want)
}
}
}
@@ -819,6 +866,13 @@ func ndpOptions(t *testing.T, optsBuf header.NDPOptions, opts []header.NDPOption
} else if got, want := gotOpt.EthernetAddress(), wantOpt.EthernetAddress(); got != want {
t.Errorf("got EthernetAddress() = %s at index %d, want = %s", got, i, want)
}
+ case header.NDPTargetLinkLayerAddressOption:
+ gotOpt, ok := opt.(header.NDPTargetLinkLayerAddressOption)
+ if !ok {
+ t.Errorf("got type = %T at index = %d; want = %T", opt, i, wantOpt)
+ } else if got, want := gotOpt.EthernetAddress(), wantOpt.EthernetAddress(); got != want {
+ t.Errorf("got EthernetAddress() = %s at index %d, want = %s", got, i, want)
+ }
default:
t.Fatalf("checker not implemented for expected NDP option: %T", wantOpt)
}
@@ -831,6 +885,21 @@ func ndpOptions(t *testing.T, optsBuf header.NDPOptions, opts []header.NDPOption
}
}
+// NDPNAOptions creates a checker that checks that the packet contains the
+// provided NDP options within an NDP Neighbor Solicitation message.
+//
+// The returned TransportChecker assumes that a valid ICMPv6 is passed to it
+// containing a valid NDPNA message as far as the size is concerned.
+func NDPNAOptions(opts []header.NDPOption) TransportChecker {
+ return func(t *testing.T, h header.Transport) {
+ t.Helper()
+
+ icmp := h.(header.ICMPv6)
+ na := header.NDPNeighborAdvert(icmp.NDPPayload())
+ ndpOptions(t, na.Options(), opts)
+ }
+}
+
// NDPNSOptions creates a checker that checks that the packet contains the
// provided NDP options within an NDP Neighbor Solicitation message.
//
@@ -849,7 +918,7 @@ func NDPNSOptions(opts []header.NDPOption) TransportChecker {
// NDPRS creates a checker that checks that the packet contains a valid NDP
// Router Solicitation message (as per the raw wire format).
//
-// checkers may assume that a valid ICMPv6 is passed to it containing a valid
+// Checkers may assume that a valid ICMPv6 is passed to it containing a valid
// NDPRS as far as the size of the message is concerned. The values within the
// message are up to checkers to validate.
func NDPRS(checkers ...TransportChecker) NetworkChecker {
diff --git a/pkg/tcpip/header/eth_test.go b/pkg/tcpip/header/eth_test.go
index 7a0014ad9..14413f2ce 100644
--- a/pkg/tcpip/header/eth_test.go
+++ b/pkg/tcpip/header/eth_test.go
@@ -88,7 +88,7 @@ func TestEthernetAddressFromMulticastIPv4Address(t *testing.T) {
for _, test := range tests {
t.Run(test.name, func(t *testing.T) {
if got := EthernetAddressFromMulticastIPv4Address(test.addr); got != test.expectedLinkAddr {
- t.Fatalf("got EthernetAddressFromMulticastIPv4Address(%s) = %s, want = %s", got, test.expectedLinkAddr)
+ t.Fatalf("got EthernetAddressFromMulticastIPv4Address(%s) = %s, want = %s", test.addr, got, test.expectedLinkAddr)
}
})
}
diff --git a/pkg/tcpip/header/ndp_test.go b/pkg/tcpip/header/ndp_test.go
index 1cb9f5dc8..969f8f1e8 100644
--- a/pkg/tcpip/header/ndp_test.go
+++ b/pkg/tcpip/header/ndp_test.go
@@ -115,7 +115,7 @@ func TestNDPNeighborAdvert(t *testing.T) {
// Make sure flags got updated in the backing buffer.
if got := b[ndpNAFlagsOffset]; got != 64 {
- t.Errorf("got flags byte = %d, want = 64")
+ t.Errorf("got flags byte = %d, want = 64", got)
}
}
diff --git a/pkg/tcpip/link/fdbased/endpoint.go b/pkg/tcpip/link/fdbased/endpoint.go
index 7198742b7..b857ce9d0 100644
--- a/pkg/tcpip/link/fdbased/endpoint.go
+++ b/pkg/tcpip/link/fdbased/endpoint.go
@@ -91,7 +91,7 @@ func (p PacketDispatchMode) String() string {
case PacketMMap:
return "PacketMMap"
default:
- return fmt.Sprintf("unknown packet dispatch mode %v", p)
+ return fmt.Sprintf("unknown packet dispatch mode '%d'", p)
}
}
diff --git a/pkg/tcpip/network/arp/arp_test.go b/pkg/tcpip/network/arp/arp_test.go
index b3e239ac7..1646d9cde 100644
--- a/pkg/tcpip/network/arp/arp_test.go
+++ b/pkg/tcpip/network/arp/arp_test.go
@@ -138,7 +138,8 @@ func TestDirectRequest(t *testing.T) {
// Sleep tests are gross, but this will only potentially flake
// if there's a bug. If there is no bug this will reliably
// succeed.
- ctx, _ := context.WithTimeout(context.Background(), 100*time.Millisecond)
+ ctx, cancel := context.WithTimeout(context.Background(), 100*time.Millisecond)
+ defer cancel()
if pkt, ok := c.linkEP.ReadContext(ctx); ok {
t.Errorf("stackAddrBad: unexpected packet sent, Proto=%v", pkt.Proto)
}
diff --git a/pkg/tcpip/network/ipv6/BUILD b/pkg/tcpip/network/ipv6/BUILD
index a93a7621a..3f71fc520 100644
--- a/pkg/tcpip/network/ipv6/BUILD
+++ b/pkg/tcpip/network/ipv6/BUILD
@@ -31,6 +31,7 @@ go_test(
deps = [
"//pkg/tcpip",
"//pkg/tcpip/buffer",
+ "//pkg/tcpip/checker",
"//pkg/tcpip/header",
"//pkg/tcpip/link/channel",
"//pkg/tcpip/link/sniffer",
diff --git a/pkg/tcpip/network/ipv6/icmp.go b/pkg/tcpip/network/ipv6/icmp.go
index f91180aa3..dc0369156 100644
--- a/pkg/tcpip/network/ipv6/icmp.go
+++ b/pkg/tcpip/network/ipv6/icmp.go
@@ -138,53 +138,48 @@ func (e *endpoint) handleICMP(r *stack.Route, netHeader buffer.View, pkt stack.P
targetAddr := ns.TargetAddress()
s := r.Stack()
- rxNICID := r.NICID()
- if isTentative, err := s.IsAddrTentative(rxNICID, targetAddr); err != nil {
- // We will only get an error if rxNICID is unrecognized,
- // which should not happen. For now short-circuit this
- // packet.
+ if isTentative, err := s.IsAddrTentative(e.nicID, targetAddr); err != nil {
+ // We will only get an error if the NIC is unrecognized, which should not
+ // happen. For now, drop this packet.
//
// TODO(b/141002840): Handle this better?
return
} else if isTentative {
- // If the target address is tentative and the source
- // of the packet is a unicast (specified) address, then
- // the source of the packet is attempting to perform
- // address resolution on the target. In this case, the
- // solicitation is silently ignored, as per RFC 4862
- // section 5.4.3.
+ // If the target address is tentative and the source of the packet is a
+ // unicast (specified) address, then the source of the packet is
+ // attempting to perform address resolution on the target. In this case,
+ // the solicitation is silently ignored, as per RFC 4862 section 5.4.3.
//
- // If the target address is tentative and the source of
- // the packet is the unspecified address (::), then we
- // know another node is also performing DAD for the
- // same address (since targetAddr is tentative for us,
- // we know we are also performing DAD on it). In this
- // case we let the stack know so it can handle such a
- // scenario and do nothing further with the NDP NS.
- if iph.SourceAddress() == header.IPv6Any {
- s.DupTentativeAddrDetected(rxNICID, targetAddr)
+ // If the target address is tentative and the source of the packet is the
+ // unspecified address (::), then we know another node is also performing
+ // DAD for the same address (since the target address is tentative for us,
+ // we know we are also performing DAD on it). In this case we let the
+ // stack know so it can handle such a scenario and do nothing further with
+ // the NS.
+ if r.RemoteAddress == header.IPv6Any {
+ s.DupTentativeAddrDetected(e.nicID, targetAddr)
}
- // Do not handle neighbor solicitations targeted
- // to an address that is tentative on the received
- // NIC any further.
+ // Do not handle neighbor solicitations targeted to an address that is
+ // tentative on the NIC any further.
return
}
- // At this point we know that targetAddr is not tentative on
- // rxNICID so the packet is processed as defined in RFC 4861,
- // as per RFC 4862 section 5.4.3.
+ // At this point we know that the target address is not tentative on the NIC
+ // so the packet is processed as defined in RFC 4861, as per RFC 4862
+ // section 5.4.3.
+ // Is the NS targetting us?
if e.linkAddrCache.CheckLocalAddress(e.nicID, ProtocolNumber, targetAddr) == 0 {
- // We don't have a useful answer; the best we can do is ignore the request.
return
}
- // If the NS message has the source link layer option, update the link
- // address cache with the link address for the sender of the message.
+ // If the NS message contains the Source Link-Layer Address option, update
+ // the link address cache with the value of the option.
//
// TODO(b/148429853): Properly process the NS message and do Neighbor
// Unreachability Detection.
+ var sourceLinkAddr tcpip.LinkAddress
for {
opt, done, err := it.Next()
if err != nil {
@@ -197,22 +192,36 @@ func (e *endpoint) handleICMP(r *stack.Route, netHeader buffer.View, pkt stack.P
switch opt := opt.(type) {
case header.NDPSourceLinkLayerAddressOption:
- e.linkAddrCache.AddLinkAddress(e.nicID, r.RemoteAddress, opt.EthernetAddress())
+ // No RFCs define what to do when an NS message has multiple Source
+ // Link-Layer Address options. Since no interface can have multiple
+ // link-layer addresses, we consider such messages invalid.
+ if len(sourceLinkAddr) != 0 {
+ received.Invalid.Increment()
+ return
+ }
+
+ sourceLinkAddr = opt.EthernetAddress()
}
}
- optsSerializer := header.NDPOptionsSerializer{
- header.NDPTargetLinkLayerAddressOption(r.LocalLinkAddress[:]),
+ unspecifiedSource := r.RemoteAddress == header.IPv6Any
+
+ // As per RFC 4861 section 4.3, the Source Link-Layer Address Option MUST
+ // NOT be included when the source IP address is the unspecified address.
+ // Otherwise, on link layers that have addresses this option MUST be
+ // included in multicast solicitations and SHOULD be included in unicast
+ // solicitations.
+ if len(sourceLinkAddr) == 0 {
+ if header.IsV6MulticastAddress(r.LocalAddress) && !unspecifiedSource {
+ received.Invalid.Increment()
+ return
+ }
+ } else if unspecifiedSource {
+ received.Invalid.Increment()
+ return
+ } else {
+ e.linkAddrCache.AddLinkAddress(e.nicID, r.RemoteAddress, sourceLinkAddr)
}
- hdr := buffer.NewPrependable(int(r.MaxHeaderLength()) + header.ICMPv6NeighborAdvertMinimumSize + int(optsSerializer.Length()))
- packet := header.ICMPv6(hdr.Prepend(header.ICMPv6NeighborAdvertSize))
- packet.SetType(header.ICMPv6NeighborAdvert)
- na := header.NDPNeighborAdvert(packet.NDPPayload())
- na.SetSolicitedFlag(true)
- na.SetOverrideFlag(true)
- na.SetTargetAddress(targetAddr)
- opts := na.Options()
- opts.Serialize(optsSerializer)
// ICMPv6 Neighbor Solicit messages are always sent to
// specially crafted IPv6 multicast addresses. As a result, the
@@ -225,6 +234,40 @@ func (e *endpoint) handleICMP(r *stack.Route, netHeader buffer.View, pkt stack.P
r := r.Clone()
defer r.Release()
r.LocalAddress = targetAddr
+
+ // As per RFC 4861 section 7.2.4, if the the source of the solicitation is
+ // the unspecified address, the node MUST set the Solicited flag to zero and
+ // multicast the advertisement to the all-nodes address.
+ solicited := true
+ if unspecifiedSource {
+ solicited = false
+ r.RemoteAddress = header.IPv6AllNodesMulticastAddress
+ }
+
+ // If the NS has a source link-layer option, use the link address it
+ // specifies as the remote link address for the response instead of the
+ // source link address of the packet.
+ //
+ // TODO(#2401): As per RFC 4861 section 7.2.4 we should consult our link
+ // address cache for the right destination link address instead of manually
+ // patching the route with the remote link address if one is specified in a
+ // Source Link-Layer Address option.
+ if len(sourceLinkAddr) != 0 {
+ r.RemoteLinkAddress = sourceLinkAddr
+ }
+
+ optsSerializer := header.NDPOptionsSerializer{
+ header.NDPTargetLinkLayerAddressOption(r.LocalLinkAddress),
+ }
+ hdr := buffer.NewPrependable(int(r.MaxHeaderLength()) + header.ICMPv6NeighborAdvertMinimumSize + int(optsSerializer.Length()))
+ packet := header.ICMPv6(hdr.Prepend(header.ICMPv6NeighborAdvertSize))
+ packet.SetType(header.ICMPv6NeighborAdvert)
+ na := header.NDPNeighborAdvert(packet.NDPPayload())
+ na.SetSolicitedFlag(solicited)
+ na.SetOverrideFlag(true)
+ na.SetTargetAddress(targetAddr)
+ opts := na.Options()
+ opts.Serialize(optsSerializer)
packet.SetChecksum(header.ICMPv6Checksum(packet, r.LocalAddress, r.RemoteAddress, buffer.VectorisedView{}))
// RFC 4861 Neighbor Discovery for IP version 6 (IPv6)
diff --git a/pkg/tcpip/network/ipv6/icmp_test.go b/pkg/tcpip/network/ipv6/icmp_test.go
index bae09ed94..bd099a7f8 100644
--- a/pkg/tcpip/network/ipv6/icmp_test.go
+++ b/pkg/tcpip/network/ipv6/icmp_test.go
@@ -32,7 +32,8 @@ import (
const (
linkAddr0 = tcpip.LinkAddress("\x02\x02\x03\x04\x05\x06")
- linkAddr1 = tcpip.LinkAddress("\x0a\x0b\x0c\x0d\x0e\x0f")
+ linkAddr1 = tcpip.LinkAddress("\x0a\x0b\x0c\x0d\x0e\x0e")
+ linkAddr2 = tcpip.LinkAddress("\x0a\x0b\x0c\x0d\x0e\x0f")
)
var (
diff --git a/pkg/tcpip/network/ipv6/ndp_test.go b/pkg/tcpip/network/ipv6/ndp_test.go
index b113aaacc..8db51da96 100644
--- a/pkg/tcpip/network/ipv6/ndp_test.go
+++ b/pkg/tcpip/network/ipv6/ndp_test.go
@@ -20,6 +20,7 @@ import (
"gvisor.dev/gvisor/pkg/tcpip"
"gvisor.dev/gvisor/pkg/tcpip/buffer"
+ "gvisor.dev/gvisor/pkg/tcpip/checker"
"gvisor.dev/gvisor/pkg/tcpip/header"
"gvisor.dev/gvisor/pkg/tcpip/link/channel"
"gvisor.dev/gvisor/pkg/tcpip/stack"
@@ -173,6 +174,257 @@ func TestNeighorSolicitationWithSourceLinkLayerOption(t *testing.T) {
}
}
+func TestNeighorSolicitationResponse(t *testing.T) {
+ const nicID = 1
+ nicAddr := lladdr0
+ remoteAddr := lladdr1
+ nicAddrSNMC := header.SolicitedNodeAddr(nicAddr)
+ nicLinkAddr := linkAddr0
+ remoteLinkAddr0 := linkAddr1
+ remoteLinkAddr1 := linkAddr2
+
+ tests := []struct {
+ name string
+ nsOpts header.NDPOptionsSerializer
+ nsSrcLinkAddr tcpip.LinkAddress
+ nsSrc tcpip.Address
+ nsDst tcpip.Address
+ nsInvalid bool
+ naDstLinkAddr tcpip.LinkAddress
+ naSolicited bool
+ naSrc tcpip.Address
+ naDst tcpip.Address
+ }{
+ {
+ name: "Unspecified source to multicast destination",
+ nsOpts: nil,
+ nsSrcLinkAddr: remoteLinkAddr0,
+ nsSrc: header.IPv6Any,
+ nsDst: nicAddrSNMC,
+ nsInvalid: false,
+ naDstLinkAddr: remoteLinkAddr0,
+ naSolicited: false,
+ naSrc: nicAddr,
+ naDst: header.IPv6AllNodesMulticastAddress,
+ },
+ {
+ name: "Unspecified source with source ll option to multicast destination",
+ nsOpts: header.NDPOptionsSerializer{
+ header.NDPSourceLinkLayerAddressOption(remoteLinkAddr0[:]),
+ },
+ nsSrcLinkAddr: remoteLinkAddr0,
+ nsSrc: header.IPv6Any,
+ nsDst: nicAddrSNMC,
+ nsInvalid: true,
+ },
+ {
+ name: "Unspecified source to unicast destination",
+ nsOpts: nil,
+ nsSrcLinkAddr: remoteLinkAddr0,
+ nsSrc: header.IPv6Any,
+ nsDst: nicAddr,
+ nsInvalid: false,
+ naDstLinkAddr: remoteLinkAddr0,
+ naSolicited: false,
+ naSrc: nicAddr,
+ naDst: header.IPv6AllNodesMulticastAddress,
+ },
+ {
+ name: "Unspecified source with source ll option to unicast destination",
+ nsOpts: header.NDPOptionsSerializer{
+ header.NDPSourceLinkLayerAddressOption(remoteLinkAddr0[:]),
+ },
+ nsSrcLinkAddr: remoteLinkAddr0,
+ nsSrc: header.IPv6Any,
+ nsDst: nicAddr,
+ nsInvalid: true,
+ },
+
+ {
+ name: "Specified source with 1 source ll to multicast destination",
+ nsOpts: header.NDPOptionsSerializer{
+ header.NDPSourceLinkLayerAddressOption(remoteLinkAddr0[:]),
+ },
+ nsSrcLinkAddr: remoteLinkAddr0,
+ nsSrc: remoteAddr,
+ nsDst: nicAddrSNMC,
+ nsInvalid: false,
+ naDstLinkAddr: remoteLinkAddr0,
+ naSolicited: true,
+ naSrc: nicAddr,
+ naDst: remoteAddr,
+ },
+ {
+ name: "Specified source with 1 source ll different from route to multicast destination",
+ nsOpts: header.NDPOptionsSerializer{
+ header.NDPSourceLinkLayerAddressOption(remoteLinkAddr1[:]),
+ },
+ nsSrcLinkAddr: remoteLinkAddr0,
+ nsSrc: remoteAddr,
+ nsDst: nicAddrSNMC,
+ nsInvalid: false,
+ naDstLinkAddr: remoteLinkAddr1,
+ naSolicited: true,
+ naSrc: nicAddr,
+ naDst: remoteAddr,
+ },
+ {
+ name: "Specified source to multicast destination",
+ nsOpts: nil,
+ nsSrcLinkAddr: remoteLinkAddr0,
+ nsSrc: remoteAddr,
+ nsDst: nicAddrSNMC,
+ nsInvalid: true,
+ },
+ {
+ name: "Specified source with 2 source ll to multicast destination",
+ nsOpts: header.NDPOptionsSerializer{
+ header.NDPSourceLinkLayerAddressOption(remoteLinkAddr0[:]),
+ header.NDPSourceLinkLayerAddressOption(remoteLinkAddr1[:]),
+ },
+ nsSrcLinkAddr: remoteLinkAddr0,
+ nsSrc: remoteAddr,
+ nsDst: nicAddrSNMC,
+ nsInvalid: true,
+ },
+
+ {
+ name: "Specified source to unicast destination",
+ nsOpts: nil,
+ nsSrcLinkAddr: remoteLinkAddr0,
+ nsSrc: remoteAddr,
+ nsDst: nicAddr,
+ nsInvalid: false,
+ naDstLinkAddr: remoteLinkAddr0,
+ naSolicited: true,
+ naSrc: nicAddr,
+ naDst: remoteAddr,
+ },
+ {
+ name: "Specified source with 1 source ll to unicast destination",
+ nsOpts: header.NDPOptionsSerializer{
+ header.NDPSourceLinkLayerAddressOption(remoteLinkAddr0[:]),
+ },
+ nsSrcLinkAddr: remoteLinkAddr0,
+ nsSrc: remoteAddr,
+ nsDst: nicAddr,
+ nsInvalid: false,
+ naDstLinkAddr: remoteLinkAddr0,
+ naSolicited: true,
+ naSrc: nicAddr,
+ naDst: remoteAddr,
+ },
+ {
+ name: "Specified source with 1 source ll different from route to unicast destination",
+ nsOpts: header.NDPOptionsSerializer{
+ header.NDPSourceLinkLayerAddressOption(remoteLinkAddr1[:]),
+ },
+ nsSrcLinkAddr: remoteLinkAddr0,
+ nsSrc: remoteAddr,
+ nsDst: nicAddr,
+ nsInvalid: false,
+ naDstLinkAddr: remoteLinkAddr1,
+ naSolicited: true,
+ naSrc: nicAddr,
+ naDst: remoteAddr,
+ },
+ {
+ name: "Specified source with 2 source ll to unicast destination",
+ nsOpts: header.NDPOptionsSerializer{
+ header.NDPSourceLinkLayerAddressOption(remoteLinkAddr0[:]),
+ header.NDPSourceLinkLayerAddressOption(remoteLinkAddr1[:]),
+ },
+ nsSrcLinkAddr: remoteLinkAddr0,
+ nsSrc: remoteAddr,
+ nsDst: nicAddr,
+ nsInvalid: true,
+ },
+ }
+
+ for _, test := range tests {
+ t.Run(test.name, func(t *testing.T) {
+ s := stack.New(stack.Options{
+ NetworkProtocols: []stack.NetworkProtocol{NewProtocol()},
+ })
+ e := channel.New(1, 1280, nicLinkAddr)
+ if err := s.CreateNIC(nicID, e); err != nil {
+ t.Fatalf("CreateNIC(%d, _) = %s", nicID, err)
+ }
+ if err := s.AddAddress(nicID, ProtocolNumber, nicAddr); err != nil {
+ t.Fatalf("AddAddress(%d, %d, %s) = %s", nicID, ProtocolNumber, nicAddr, err)
+ }
+
+ ndpNSSize := header.ICMPv6NeighborSolicitMinimumSize + test.nsOpts.Length()
+ hdr := buffer.NewPrependable(header.IPv6MinimumSize + ndpNSSize)
+ pkt := header.ICMPv6(hdr.Prepend(ndpNSSize))
+ pkt.SetType(header.ICMPv6NeighborSolicit)
+ ns := header.NDPNeighborSolicit(pkt.NDPPayload())
+ ns.SetTargetAddress(nicAddr)
+ opts := ns.Options()
+ opts.Serialize(test.nsOpts)
+ pkt.SetChecksum(header.ICMPv6Checksum(pkt, test.nsSrc, test.nsDst, buffer.VectorisedView{}))
+ payloadLength := hdr.UsedLength()
+ ip := header.IPv6(hdr.Prepend(header.IPv6MinimumSize))
+ ip.Encode(&header.IPv6Fields{
+ PayloadLength: uint16(payloadLength),
+ NextHeader: uint8(header.ICMPv6ProtocolNumber),
+ HopLimit: 255,
+ SrcAddr: test.nsSrc,
+ DstAddr: test.nsDst,
+ })
+
+ invalid := s.Stats().ICMP.V6PacketsReceived.Invalid
+
+ // Invalid count should initially be 0.
+ if got := invalid.Value(); got != 0 {
+ t.Fatalf("got invalid = %d, want = 0", got)
+ }
+
+ e.InjectLinkAddr(ProtocolNumber, test.nsSrcLinkAddr, stack.PacketBuffer{
+ Data: hdr.View().ToVectorisedView(),
+ })
+
+ if test.nsInvalid {
+ if got := invalid.Value(); got != 1 {
+ t.Fatalf("got invalid = %d, want = 1", got)
+ }
+
+ if p, got := e.Read(); got {
+ t.Fatalf("unexpected response to an invalid NS = %+v", p.Pkt)
+ }
+
+ // If we expected the NS to be invalid, we have nothing else to check.
+ return
+ }
+
+ if got := invalid.Value(); got != 0 {
+ t.Fatalf("got invalid = %d, want = 0", got)
+ }
+
+ p, got := e.Read()
+ if !got {
+ t.Fatal("expected an NDP NA response")
+ }
+
+ if p.Route.RemoteLinkAddress != test.naDstLinkAddr {
+ t.Errorf("got p.Route.RemoteLinkAddress = %s, want = %s", p.Route.RemoteLinkAddress, test.naDstLinkAddr)
+ }
+
+ checker.IPv6(t, p.Pkt.Header.View(),
+ checker.SrcAddr(test.naSrc),
+ checker.DstAddr(test.naDst),
+ checker.TTL(header.NDPHopLimit),
+ checker.NDPNA(
+ checker.NDPNASolicitedFlag(test.naSolicited),
+ checker.NDPNATargetAddress(nicAddr),
+ checker.NDPNAOptions([]header.NDPOption{
+ header.NDPTargetLinkLayerAddressOption(nicLinkAddr[:]),
+ }),
+ ))
+ })
+ }
+}
+
// TestNeighorAdvertisementWithTargetLinkLayerOption tests that receiving a
// valid NDP NA message with the Target Link Layer Address option results in a
// new entry in the link address cache for the target of the message.
diff --git a/pkg/tcpip/stack/ndp_test.go b/pkg/tcpip/stack/ndp_test.go
index 27dc8baf9..acb2d4731 100644
--- a/pkg/tcpip/stack/ndp_test.go
+++ b/pkg/tcpip/stack/ndp_test.go
@@ -1959,7 +1959,7 @@ func TestAutoGenAddrDeprecateFromPI(t *testing.T) {
// addr2 is deprecated but if explicitly requested, it should be used.
fullAddr2 := tcpip.FullAddress{Addr: addr2.Address, NIC: nicID}
if got := addrForNewConnectionWithAddr(t, s, fullAddr2); got != addr2.Address {
- t.Errorf("got addrForNewConnectionWithAddr(_, _, %+v) = %s, want = %s", got, addr2.Address)
+ t.Errorf("got addrForNewConnectionWithAddr(_, _, %+v) = %s, want = %s", fullAddr2, got, addr2.Address)
}
// Another PI w/ 0 preferred lifetime should not result in a deprecation
@@ -1972,7 +1972,7 @@ func TestAutoGenAddrDeprecateFromPI(t *testing.T) {
}
expectPrimaryAddr(addr1)
if got := addrForNewConnectionWithAddr(t, s, fullAddr2); got != addr2.Address {
- t.Errorf("got addrForNewConnectionWithAddr(_, _, %+v) = %s, want = %s", got, addr2.Address)
+ t.Errorf("got addrForNewConnectionWithAddr(_, _, %+v) = %s, want = %s", fullAddr2, got, addr2.Address)
}
// Refresh lifetimes of addr generated from prefix2.
@@ -2084,7 +2084,7 @@ func TestAutoGenAddrTimerDeprecation(t *testing.T) {
// addr1 is deprecated but if explicitly requested, it should be used.
fullAddr1 := tcpip.FullAddress{Addr: addr1.Address, NIC: nicID}
if got := addrForNewConnectionWithAddr(t, s, fullAddr1); got != addr1.Address {
- t.Errorf("got addrForNewConnectionWithAddr(_, _, %+v) = %s, want = %s", got, addr1.Address)
+ t.Errorf("got addrForNewConnectionWithAddr(_, _, %+v) = %s, want = %s", fullAddr1, got, addr1.Address)
}
// Refresh valid lifetime for addr of prefix1, w/ 0 preferred lifetime to make
@@ -2097,7 +2097,7 @@ func TestAutoGenAddrTimerDeprecation(t *testing.T) {
}
expectPrimaryAddr(addr2)
if got := addrForNewConnectionWithAddr(t, s, fullAddr1); got != addr1.Address {
- t.Errorf("got addrForNewConnectionWithAddr(_, _, %+v) = %s, want = %s", got, addr1.Address)
+ t.Errorf("got addrForNewConnectionWithAddr(_, _, %+v) = %s, want = %s", fullAddr1, got, addr1.Address)
}
// Refresh lifetimes for addr of prefix1.
@@ -2121,7 +2121,7 @@ func TestAutoGenAddrTimerDeprecation(t *testing.T) {
// addr2 should be the primary endpoint now since it is not deprecated.
expectPrimaryAddr(addr2)
if got := addrForNewConnectionWithAddr(t, s, fullAddr1); got != addr1.Address {
- t.Errorf("got addrForNewConnectionWithAddr(_, _, %+v) = %s, want = %s", got, addr1.Address)
+ t.Errorf("got addrForNewConnectionWithAddr(_, _, %+v) = %s, want = %s", fullAddr1, got, addr1.Address)
}
// Wait for addr of prefix1 to be invalidated.
@@ -2564,7 +2564,7 @@ func TestAutoGenAddrAfterRemoval(t *testing.T) {
AddressWithPrefix: addr2,
}
if err := s.AddProtocolAddressWithOptions(nicID, protoAddr2, stack.FirstPrimaryEndpoint); err != nil {
- t.Fatalf("AddProtocolAddressWithOptions(%d, %+v, %d, %s) = %s", nicID, protoAddr2, stack.FirstPrimaryEndpoint, err)
+ t.Fatalf("AddProtocolAddressWithOptions(%d, %+v, %d) = %s", nicID, protoAddr2, stack.FirstPrimaryEndpoint, err)
}
// addr2 should be more preferred now since it is at the front of the primary
// list.
@@ -3483,7 +3483,8 @@ func TestRouterSolicitation(t *testing.T) {
e.Endpoint.LinkEPCapabilities |= stack.CapabilityResolutionRequired
waitForPkt := func(timeout time.Duration) {
t.Helper()
- ctx, _ := context.WithTimeout(context.Background(), timeout)
+ ctx, cancel := context.WithTimeout(context.Background(), timeout)
+ defer cancel()
p, ok := e.ReadContext(ctx)
if !ok {
t.Fatal("timed out waiting for packet")
@@ -3513,7 +3514,8 @@ func TestRouterSolicitation(t *testing.T) {
}
waitForNothing := func(timeout time.Duration) {
t.Helper()
- ctx, _ := context.WithTimeout(context.Background(), timeout)
+ ctx, cancel := context.WithTimeout(context.Background(), timeout)
+ defer cancel()
if _, ok := e.ReadContext(ctx); ok {
t.Fatal("unexpectedly got a packet")
}
diff --git a/pkg/tcpip/stack/stack_test.go b/pkg/tcpip/stack/stack_test.go
index 3f8a2a095..c7634ceb1 100644
--- a/pkg/tcpip/stack/stack_test.go
+++ b/pkg/tcpip/stack/stack_test.go
@@ -1445,19 +1445,19 @@ func TestOutgoingBroadcastWithEmptyRouteTable(t *testing.T) {
protoAddr := tcpip.ProtocolAddress{Protocol: fakeNetNumber, AddressWithPrefix: tcpip.AddressWithPrefix{header.IPv4Any, 0}}
if err := s.AddProtocolAddress(1, protoAddr); err != nil {
- t.Fatalf("AddProtocolAddress(1, %s) failed: %s", protoAddr, err)
+ t.Fatalf("AddProtocolAddress(1, %v) failed: %v", protoAddr, err)
}
r, err := s.FindRoute(1, header.IPv4Any, header.IPv4Broadcast, fakeNetNumber, false /* multicastLoop */)
if err != nil {
- t.Fatalf("FindRoute(1, %s, %s, %d) failed: %s", header.IPv4Any, header.IPv4Broadcast, fakeNetNumber, err)
+ t.Fatalf("FindRoute(1, %v, %v, %d) failed: %v", header.IPv4Any, header.IPv4Broadcast, fakeNetNumber, err)
}
if err := verifyRoute(r, stack.Route{LocalAddress: header.IPv4Any, RemoteAddress: header.IPv4Broadcast}); err != nil {
- t.Errorf("FindRoute(1, %s, %s, %d) returned unexpected Route: %s)", header.IPv4Any, header.IPv4Broadcast, fakeNetNumber, err)
+ t.Errorf("FindRoute(1, %v, %v, %d) returned unexpected Route: %v", header.IPv4Any, header.IPv4Broadcast, fakeNetNumber, err)
}
// If the NIC doesn't exist, it won't work.
if _, err := s.FindRoute(2, header.IPv4Any, header.IPv4Broadcast, fakeNetNumber, false /* multicastLoop */); err != tcpip.ErrNetworkUnreachable {
- t.Fatalf("got FindRoute(2, %s, %s, %d) = %s want = %s", header.IPv4Any, header.IPv4Broadcast, fakeNetNumber, err, tcpip.ErrNetworkUnreachable)
+ t.Fatalf("got FindRoute(2, %v, %v, %d) = %v want = %v", header.IPv4Any, header.IPv4Broadcast, fakeNetNumber, err, tcpip.ErrNetworkUnreachable)
}
}
@@ -1483,12 +1483,12 @@ func TestOutgoingBroadcastWithRouteTable(t *testing.T) {
}
nic1ProtoAddr := tcpip.ProtocolAddress{fakeNetNumber, nic1Addr}
if err := s.AddProtocolAddress(1, nic1ProtoAddr); err != nil {
- t.Fatalf("AddProtocolAddress(1, %s) failed: %s", nic1ProtoAddr, err)
+ t.Fatalf("AddProtocolAddress(1, %v) failed: %v", nic1ProtoAddr, err)
}
nic2ProtoAddr := tcpip.ProtocolAddress{fakeNetNumber, nic2Addr}
if err := s.AddProtocolAddress(2, nic2ProtoAddr); err != nil {
- t.Fatalf("AddAddress(2, %s) failed: %s", nic2ProtoAddr, err)
+ t.Fatalf("AddAddress(2, %v) failed: %v", nic2ProtoAddr, err)
}
// Set the initial route table.
@@ -1503,10 +1503,10 @@ func TestOutgoingBroadcastWithRouteTable(t *testing.T) {
// When an interface is given, the route for a broadcast goes through it.
r, err := s.FindRoute(1, nic1Addr.Address, header.IPv4Broadcast, fakeNetNumber, false /* multicastLoop */)
if err != nil {
- t.Fatalf("FindRoute(1, %s, %s, %d) failed: %s", nic1Addr.Address, header.IPv4Broadcast, fakeNetNumber, err)
+ t.Fatalf("FindRoute(1, %v, %v, %d) failed: %v", nic1Addr.Address, header.IPv4Broadcast, fakeNetNumber, err)
}
if err := verifyRoute(r, stack.Route{LocalAddress: nic1Addr.Address, RemoteAddress: header.IPv4Broadcast}); err != nil {
- t.Errorf("FindRoute(1, %s, %s, %d) returned unexpected Route: %s)", nic1Addr.Address, header.IPv4Broadcast, fakeNetNumber, err)
+ t.Errorf("FindRoute(1, %v, %v, %d) returned unexpected Route: %v", nic1Addr.Address, header.IPv4Broadcast, fakeNetNumber, err)
}
// When an interface is not given, it consults the route table.
@@ -2399,7 +2399,7 @@ func TestNICContextPreservation(t *testing.T) {
t.Fatalf("got nicinfos[%d] = _, %t, want _, true; nicinfos = %+v", id, ok, nicinfos)
}
if got, want := nicinfo.Context == test.want, true; got != want {
- t.Fatal("got nicinfo.Context == ctx = %t, want %t; nicinfo.Context = %p, ctx = %p", got, want, nicinfo.Context, test.want)
+ t.Fatalf("got nicinfo.Context == ctx = %t, want %t; nicinfo.Context = %p, ctx = %p", got, want, nicinfo.Context, test.want)
}
})
}
@@ -2768,7 +2768,7 @@ func TestNewPEBOnPromotionToPermanent(t *testing.T) {
{
subnet, err := tcpip.NewSubnet("\x00", "\x00")
if err != nil {
- t.Fatalf("NewSubnet failed:", err)
+ t.Fatalf("NewSubnet failed: %v", err)
}
s.SetRouteTable([]tcpip.Route{{Destination: subnet, Gateway: "\x00", NIC: 1}})
}
@@ -2782,11 +2782,11 @@ func TestNewPEBOnPromotionToPermanent(t *testing.T) {
// permanentExpired kind.
r, err := s.FindRoute(1, "\x01", "\x02", fakeNetNumber, false)
if err != nil {
- t.Fatal("FindRoute failed:", err)
+ t.Fatalf("FindRoute failed: %v", err)
}
defer r.Release()
if err := s.RemoveAddress(1, "\x01"); err != nil {
- t.Fatalf("RemoveAddress failed:", err)
+ t.Fatalf("RemoveAddress failed: %v", err)
}
//
@@ -2798,7 +2798,7 @@ func TestNewPEBOnPromotionToPermanent(t *testing.T) {
// Add some other address with peb set to
// FirstPrimaryEndpoint.
if err := s.AddAddressWithOptions(1, fakeNetNumber, "\x03", stack.FirstPrimaryEndpoint); err != nil {
- t.Fatal("AddAddressWithOptions failed:", err)
+ t.Fatalf("AddAddressWithOptions failed: %v", err)
}
@@ -2806,7 +2806,7 @@ func TestNewPEBOnPromotionToPermanent(t *testing.T) {
// make sure the new peb was respected.
// (The address should just be promoted now).
if err := s.AddAddressWithOptions(1, fakeNetNumber, "\x01", ps); err != nil {
- t.Fatal("AddAddressWithOptions failed:", err)
+ t.Fatalf("AddAddressWithOptions failed: %v", err)
}
var primaryAddrs []tcpip.Address
for _, pa := range s.NICInfo()[1].ProtocolAddresses {
@@ -2839,11 +2839,11 @@ func TestNewPEBOnPromotionToPermanent(t *testing.T) {
// GetMainNICAddress; else, our original address
// should be returned.
if err := s.RemoveAddress(1, "\x03"); err != nil {
- t.Fatalf("RemoveAddress failed:", err)
+ t.Fatalf("RemoveAddress failed: %v", err)
}
addr, err = s.GetMainNICAddress(1, fakeNetNumber)
if err != nil {
- t.Fatal("s.GetMainNICAddress failed:", err)
+ t.Fatalf("s.GetMainNICAddress failed: %v", err)
}
if ps == stack.NeverPrimaryEndpoint {
if want := (tcpip.AddressWithPrefix{}); addr != want {
diff --git a/pkg/tcpip/stack/transport_demuxer_test.go b/pkg/tcpip/stack/transport_demuxer_test.go
index c65b0c632..2474a7db3 100644
--- a/pkg/tcpip/stack/transport_demuxer_test.go
+++ b/pkg/tcpip/stack/transport_demuxer_test.go
@@ -206,7 +206,7 @@ func TestTransportDemuxerRegister(t *testing.T) {
// the distribution of packets received matches expectations.
func TestBindToDeviceDistribution(t *testing.T) {
type endpointSockopts struct {
- reuse int
+ reuse bool
bindToDevice tcpip.NICID
}
for _, test := range []struct {
@@ -221,11 +221,11 @@ func TestBindToDeviceDistribution(t *testing.T) {
"BindPortReuse",
// 5 endpoints that all have reuse set.
[]endpointSockopts{
- {reuse: 1, bindToDevice: 0},
- {reuse: 1, bindToDevice: 0},
- {reuse: 1, bindToDevice: 0},
- {reuse: 1, bindToDevice: 0},
- {reuse: 1, bindToDevice: 0},
+ {reuse: true, bindToDevice: 0},
+ {reuse: true, bindToDevice: 0},
+ {reuse: true, bindToDevice: 0},
+ {reuse: true, bindToDevice: 0},
+ {reuse: true, bindToDevice: 0},
},
map[tcpip.NICID][]float64{
// Injected packets on dev0 get distributed evenly.
@@ -236,9 +236,9 @@ func TestBindToDeviceDistribution(t *testing.T) {
"BindToDevice",
// 3 endpoints with various bindings.
[]endpointSockopts{
- {reuse: 0, bindToDevice: 1},
- {reuse: 0, bindToDevice: 2},
- {reuse: 0, bindToDevice: 3},
+ {reuse: false, bindToDevice: 1},
+ {reuse: false, bindToDevice: 2},
+ {reuse: false, bindToDevice: 3},
},
map[tcpip.NICID][]float64{
// Injected packets on dev0 go only to the endpoint bound to dev0.
@@ -253,12 +253,12 @@ func TestBindToDeviceDistribution(t *testing.T) {
"ReuseAndBindToDevice",
// 6 endpoints with various bindings.
[]endpointSockopts{
- {reuse: 1, bindToDevice: 1},
- {reuse: 1, bindToDevice: 1},
- {reuse: 1, bindToDevice: 2},
- {reuse: 1, bindToDevice: 2},
- {reuse: 1, bindToDevice: 2},
- {reuse: 1, bindToDevice: 0},
+ {reuse: true, bindToDevice: 1},
+ {reuse: true, bindToDevice: 1},
+ {reuse: true, bindToDevice: 2},
+ {reuse: true, bindToDevice: 2},
+ {reuse: true, bindToDevice: 2},
+ {reuse: true, bindToDevice: 0},
},
map[tcpip.NICID][]float64{
// Injected packets on dev0 get distributed among endpoints bound to
@@ -309,9 +309,8 @@ func TestBindToDeviceDistribution(t *testing.T) {
}(ep)
defer ep.Close()
- reusePortOption := tcpip.ReusePortOption(endpoint.reuse)
- if err := ep.SetSockOpt(reusePortOption); err != nil {
- t.Fatalf("SetSockOpt(%#v) on endpoint %d failed: %s", reusePortOption, i, err)
+ if err := ep.SetSockOptBool(tcpip.ReusePortOption, endpoint.reuse); err != nil {
+ t.Fatalf("SetSockOptBool(ReusePortOption, %t) on endpoint %d failed: %s", endpoint.reuse, i, err)
}
bindToDeviceOption := tcpip.BindToDeviceOption(endpoint.bindToDevice)
if err := ep.SetSockOpt(bindToDeviceOption); err != nil {
diff --git a/pkg/tcpip/tcpip.go b/pkg/tcpip/tcpip.go
index 2ef3271f1..aec7126ff 100644
--- a/pkg/tcpip/tcpip.go
+++ b/pkg/tcpip/tcpip.go
@@ -520,34 +520,90 @@ type WriteOptions struct {
type SockOptBool int
const (
+ // BroadcastOption is used by SetSockOpt/GetSockOpt to specify whether
+ // datagram sockets are allowed to send packets to a broadcast address.
+ BroadcastOption SockOptBool = iota
+
+ // CorkOption is used by SetSockOpt/GetSockOpt to specify if data should be
+ // held until segments are full by the TCP transport protocol.
+ CorkOption
+
+ // DelayOption is used by SetSockOpt/GetSockOpt to specify if data
+ // should be sent out immediately by the transport protocol. For TCP,
+ // it determines if the Nagle algorithm is on or off.
+ DelayOption
+
+ // KeepaliveEnabledOption is used by SetSockOpt/GetSockOpt to specify whether
+ // TCP keepalive is enabled for this socket.
+ KeepaliveEnabledOption
+
+ // MulticastLoopOption is used by SetSockOpt/GetSockOpt to specify whether
+ // multicast packets sent over a non-loopback interface will be looped back.
+ MulticastLoopOption
+
+ // PasscredOption is used by SetSockOpt/GetSockOpt to specify whether
+ // SCM_CREDENTIALS socket control messages are enabled.
+ //
+ // Only supported on Unix sockets.
+ PasscredOption
+
+ // QuickAckOption is stubbed out in SetSockOpt/GetSockOpt.
+ QuickAckOption
+
// ReceiveTClassOption is used by SetSockOpt/GetSockOpt to specify if the
// IPV6_TCLASS ancillary message is passed with incoming packets.
- ReceiveTClassOption SockOptBool = iota
+ ReceiveTClassOption
// ReceiveTOSOption is used by SetSockOpt/GetSockOpt to specify if the TOS
// ancillary message is passed with incoming packets.
ReceiveTOSOption
- // V6OnlyOption is used by {G,S}etSockOptBool to specify whether an IPv6
- // socket is to be restricted to sending and receiving IPv6 packets only.
- V6OnlyOption
-
// ReceiveIPPacketInfoOption is used by {G,S}etSockOptBool to specify
// if more inforamtion is provided with incoming packets such
// as interface index and address.
ReceiveIPPacketInfoOption
- // TODO(b/146901447): convert existing bool socket options to be handled via
- // Get/SetSockOptBool
+ // ReuseAddressOption is used by SetSockOpt/GetSockOpt to specify whether Bind()
+ // should allow reuse of local address.
+ ReuseAddressOption
+
+ // ReusePortOption is used by SetSockOpt/GetSockOpt to permit multiple sockets
+ // to be bound to an identical socket address.
+ ReusePortOption
+
+ // V6OnlyOption is used by {G,S}etSockOptBool to specify whether an IPv6
+ // socket is to be restricted to sending and receiving IPv6 packets only.
+ V6OnlyOption
)
// SockOptInt represents socket options which values have the int type.
type SockOptInt int
const (
+ // KeepaliveCountOption is used by SetSockOpt/GetSockOpt to specify the number
+ // of un-ACKed TCP keepalives that will be sent before the connection is
+ // closed.
+ KeepaliveCountOption SockOptInt = iota
+
+ // IPv4TOSOption is used by SetSockOpt/GetSockOpt to specify TOS
+ // for all subsequent outgoing IPv4 packets from the endpoint.
+ IPv4TOSOption
+
+ // IPv6TrafficClassOption is used by SetSockOpt/GetSockOpt to specify TOS
+ // for all subsequent outgoing IPv6 packets from the endpoint.
+ IPv6TrafficClassOption
+
+ // MaxSegOption is used by SetSockOpt/GetSockOpt to set/get the current
+ // Maximum Segment Size(MSS) value as specified using the TCP_MAXSEG option.
+ MaxSegOption
+
+ // MulticastTTLOption is used by SetSockOpt/GetSockOpt to control the default
+ // TTL value for multicast messages. The default is 1.
+ MulticastTTLOption
+
// ReceiveQueueSizeOption is used in GetSockOptInt to specify that the
// number of unread bytes in the input buffer should be returned.
- ReceiveQueueSizeOption SockOptInt = iota
+ ReceiveQueueSizeOption
// SendBufferSizeOption is used by SetSockOptInt/GetSockOptInt to
// specify the send buffer size option.
@@ -561,44 +617,21 @@ const (
// number of unread bytes in the output buffer should be returned.
SendQueueSizeOption
- // DelayOption is used by SetSockOpt/GetSockOpt to specify if data
- // should be sent out immediately by the transport protocol. For TCP,
- // it determines if the Nagle algorithm is on or off.
- DelayOption
-
- // TODO(b/137664753): convert all int socket options to be handled via
- // GetSockOptInt.
+ // TTLOption is used by SetSockOpt/GetSockOpt to control the default TTL/hop
+ // limit value for unicast messages. The default is protocol specific.
+ //
+ // A zero value indicates the default.
+ TTLOption
)
// ErrorOption is used in GetSockOpt to specify that the last error reported by
// the endpoint should be cleared and returned.
type ErrorOption struct{}
-// CorkOption is used by SetSockOpt/GetSockOpt to specify if data should be
-// held until segments are full by the TCP transport protocol.
-type CorkOption int
-
-// ReuseAddressOption is used by SetSockOpt/GetSockOpt to specify whether Bind()
-// should allow reuse of local address.
-type ReuseAddressOption int
-
-// ReusePortOption is used by SetSockOpt/GetSockOpt to permit multiple sockets
-// to be bound to an identical socket address.
-type ReusePortOption int
-
// BindToDeviceOption is used by SetSockOpt/GetSockOpt to specify that sockets
// should bind only on a specific NIC.
type BindToDeviceOption NICID
-// QuickAckOption is stubbed out in SetSockOpt/GetSockOpt.
-type QuickAckOption int
-
-// PasscredOption is used by SetSockOpt/GetSockOpt to specify whether
-// SCM_CREDENTIALS socket control messages are enabled.
-//
-// Only supported on Unix sockets.
-type PasscredOption int
-
// TCPInfoOption is used by GetSockOpt to expose TCP statistics.
//
// TODO(b/64800844): Add and populate stat fields.
@@ -607,10 +640,6 @@ type TCPInfoOption struct {
RTTVar time.Duration
}
-// KeepaliveEnabledOption is used by SetSockOpt/GetSockOpt to specify whether
-// TCP keepalive is enabled for this socket.
-type KeepaliveEnabledOption int
-
// KeepaliveIdleOption is used by SetSockOpt/GetSockOpt to specify the time a
// connection must remain idle before the first TCP keepalive packet is sent.
// Once this time is reached, KeepaliveIntervalOption is used instead.
@@ -620,11 +649,6 @@ type KeepaliveIdleOption time.Duration
// interval between sending TCP keepalive packets.
type KeepaliveIntervalOption time.Duration
-// KeepaliveCountOption is used by SetSockOpt/GetSockOpt to specify the number
-// of un-ACKed TCP keepalives that will be sent before the connection is
-// closed.
-type KeepaliveCountOption int
-
// TCPUserTimeoutOption is used by SetSockOpt/GetSockOpt to specify a user
// specified timeout for a given TCP connection.
// See: RFC5482 for details.
@@ -638,20 +662,9 @@ type CongestionControlOption string
// control algorithms.
type AvailableCongestionControlOption string
-// ModerateReceiveBufferOption allows the caller to enable/disable TCP receive
// buffer moderation.
type ModerateReceiveBufferOption bool
-// MaxSegOption is used by SetSockOpt/GetSockOpt to set/get the current
-// Maximum Segment Size(MSS) value as specified using the TCP_MAXSEG option.
-type MaxSegOption int
-
-// TTLOption is used by SetSockOpt/GetSockOpt to control the default TTL/hop
-// limit value for unicast messages. The default is protocol specific.
-//
-// A zero value indicates the default.
-type TTLOption uint8
-
// TCPLingerTimeoutOption is used by SetSockOpt/GetSockOpt to set/get the
// maximum duration for which a socket lingers in the TCP_FIN_WAIT_2 state
// before being marked closed.
@@ -668,10 +681,6 @@ type TCPTimeWaitTimeoutOption time.Duration
// for a handshake till the specified timeout until a segment with data arrives.
type TCPDeferAcceptOption time.Duration
-// MulticastTTLOption is used by SetSockOpt/GetSockOpt to control the default
-// TTL value for multicast messages. The default is 1.
-type MulticastTTLOption uint8
-
// MulticastInterfaceOption is used by SetSockOpt/GetSockOpt to specify a
// default interface for multicast.
type MulticastInterfaceOption struct {
@@ -679,10 +688,6 @@ type MulticastInterfaceOption struct {
InterfaceAddr Address
}
-// MulticastLoopOption is used by SetSockOpt/GetSockOpt to specify whether
-// multicast packets sent over a non-loopback interface will be looped back.
-type MulticastLoopOption bool
-
// MembershipOption is used by SetSockOpt/GetSockOpt as an argument to
// AddMembershipOption and RemoveMembershipOption.
type MembershipOption struct {
@@ -705,22 +710,10 @@ type RemoveMembershipOption MembershipOption
// TCP out-of-band data is delivered along with the normal in-band data.
type OutOfBandInlineOption int
-// BroadcastOption is used by SetSockOpt/GetSockOpt to specify whether
-// datagram sockets are allowed to send packets to a broadcast address.
-type BroadcastOption int
-
// DefaultTTLOption is used by stack.(*Stack).NetworkProtocolOption to specify
// a default TTL.
type DefaultTTLOption uint8
-// IPv4TOSOption is used by SetSockOpt/GetSockOpt to specify TOS
-// for all subsequent outgoing IPv4 packets from the endpoint.
-type IPv4TOSOption uint8
-
-// IPv6TrafficClassOption is used by SetSockOpt/GetSockOpt to specify TOS
-// for all subsequent outgoing IPv6 packets from the endpoint.
-type IPv6TrafficClassOption uint8
-
// IPPacketInfo is the message struture for IP_PKTINFO.
//
// +stateify savable
diff --git a/pkg/tcpip/tcpip_test.go b/pkg/tcpip/tcpip_test.go
index 8c0aacffa..1c8e2bc34 100644
--- a/pkg/tcpip/tcpip_test.go
+++ b/pkg/tcpip/tcpip_test.go
@@ -218,7 +218,7 @@ func TestAddressWithPrefixSubnet(t *testing.T) {
gotSubnet := ap.Subnet()
wantSubnet, err := NewSubnet(tt.subnetAddr, tt.subnetMask)
if err != nil {
- t.Error("NewSubnet(%q, %q) failed: %s", tt.subnetAddr, tt.subnetMask, err)
+ t.Errorf("NewSubnet(%q, %q) failed: %s", tt.subnetAddr, tt.subnetMask, err)
continue
}
if gotSubnet != wantSubnet {
diff --git a/pkg/tcpip/transport/icmp/endpoint.go b/pkg/tcpip/transport/icmp/endpoint.go
index b007302fb..3a133eef9 100644
--- a/pkg/tcpip/transport/icmp/endpoint.go
+++ b/pkg/tcpip/transport/icmp/endpoint.go
@@ -348,29 +348,37 @@ func (e *endpoint) Peek([][]byte) (int64, tcpip.ControlMessages, *tcpip.Error) {
// SetSockOpt sets a socket option.
func (e *endpoint) SetSockOpt(opt interface{}) *tcpip.Error {
- switch o := opt.(type) {
- case tcpip.TTLOption:
- e.mu.Lock()
- e.ttl = uint8(o)
- e.mu.Unlock()
- }
-
return nil
}
// SetSockOptBool sets a socket option. Currently not supported.
func (e *endpoint) SetSockOptBool(opt tcpip.SockOptBool, v bool) *tcpip.Error {
- return nil
+ return tcpip.ErrUnknownProtocolOption
}
// SetSockOptInt sets a socket option. Currently not supported.
func (e *endpoint) SetSockOptInt(opt tcpip.SockOptInt, v int) *tcpip.Error {
+ switch opt {
+ case tcpip.TTLOption:
+ e.mu.Lock()
+ e.ttl = uint8(v)
+ e.mu.Unlock()
+
+ default:
+ return tcpip.ErrUnknownProtocolOption
+ }
return nil
}
// GetSockOptBool implements tcpip.Endpoint.GetSockOptBool.
func (e *endpoint) GetSockOptBool(opt tcpip.SockOptBool) (bool, *tcpip.Error) {
- return false, tcpip.ErrUnknownProtocolOption
+ switch opt {
+ case tcpip.KeepaliveEnabledOption:
+ return false, nil
+
+ default:
+ return false, tcpip.ErrUnknownProtocolOption
+ }
}
// GetSockOptInt implements tcpip.Endpoint.GetSockOptInt.
@@ -397,26 +405,23 @@ func (e *endpoint) GetSockOptInt(opt tcpip.SockOptInt) (int, *tcpip.Error) {
e.rcvMu.Unlock()
return v, nil
+ case tcpip.TTLOption:
+ e.rcvMu.Lock()
+ v := int(e.ttl)
+ e.rcvMu.Unlock()
+ return v, nil
+
+ default:
+ return -1, tcpip.ErrUnknownProtocolOption
}
- return -1, tcpip.ErrUnknownProtocolOption
}
// GetSockOpt implements tcpip.Endpoint.GetSockOpt.
func (e *endpoint) GetSockOpt(opt interface{}) *tcpip.Error {
- switch o := opt.(type) {
+ switch opt.(type) {
case tcpip.ErrorOption:
return nil
- case *tcpip.KeepaliveEnabledOption:
- *o = 0
- return nil
-
- case *tcpip.TTLOption:
- e.rcvMu.Lock()
- *o = tcpip.TTLOption(e.ttl)
- e.rcvMu.Unlock()
- return nil
-
default:
return tcpip.ErrUnknownProtocolOption
}
diff --git a/pkg/tcpip/transport/raw/endpoint.go b/pkg/tcpip/transport/raw/endpoint.go
index 337bc1c71..eee754a5a 100644
--- a/pkg/tcpip/transport/raw/endpoint.go
+++ b/pkg/tcpip/transport/raw/endpoint.go
@@ -533,14 +533,10 @@ func (e *endpoint) SetSockOptInt(opt tcpip.SockOptInt, v int) *tcpip.Error {
// GetSockOpt implements tcpip.Endpoint.GetSockOpt.
func (e *endpoint) GetSockOpt(opt interface{}) *tcpip.Error {
- switch o := opt.(type) {
+ switch opt.(type) {
case tcpip.ErrorOption:
return nil
- case *tcpip.KeepaliveEnabledOption:
- *o = 0
- return nil
-
default:
return tcpip.ErrUnknownProtocolOption
}
@@ -548,7 +544,13 @@ func (e *endpoint) GetSockOpt(opt interface{}) *tcpip.Error {
// GetSockOptBool implements tcpip.Endpoint.GetSockOptBool.
func (e *endpoint) GetSockOptBool(opt tcpip.SockOptBool) (bool, *tcpip.Error) {
- return false, tcpip.ErrUnknownProtocolOption
+ switch opt {
+ case tcpip.KeepaliveEnabledOption:
+ return false, nil
+
+ default:
+ return false, tcpip.ErrUnknownProtocolOption
+ }
}
// GetSockOptInt implements tcpip.Endpoint.GetSockOptInt.
@@ -576,9 +578,9 @@ func (e *endpoint) GetSockOptInt(opt tcpip.SockOptInt) (int, *tcpip.Error) {
e.rcvMu.Unlock()
return v, nil
+ default:
+ return -1, tcpip.ErrUnknownProtocolOption
}
-
- return -1, tcpip.ErrUnknownProtocolOption
}
// HandlePacket implements stack.RawTransportEndpoint.HandlePacket.
diff --git a/pkg/tcpip/transport/tcp/endpoint.go b/pkg/tcpip/transport/tcp/endpoint.go
index 9b123e968..a8d443f73 100644
--- a/pkg/tcpip/transport/tcp/endpoint.go
+++ b/pkg/tcpip/transport/tcp/endpoint.go
@@ -821,7 +821,7 @@ func newEndpoint(s *stack.Stack, netProto tcpip.NetworkProtocolNumber, waiterQue
var de DelayEnabled
if err := s.TransportProtocolOption(ProtocolNumber, &de); err == nil && de {
- e.SetSockOptInt(tcpip.DelayOption, 1)
+ e.SetSockOptBool(tcpip.DelayOption, true)
}
var tcpLT tcpip.TCPLingerTimeoutOption
@@ -1409,10 +1409,60 @@ func (e *endpoint) windowCrossedACKThresholdLocked(deltaBefore int) (crossed boo
// SetSockOptBool sets a socket option.
func (e *endpoint) SetSockOptBool(opt tcpip.SockOptBool, v bool) *tcpip.Error {
- e.LockUser()
- defer e.UnlockUser()
-
switch opt {
+
+ case tcpip.BroadcastOption:
+ e.LockUser()
+ e.broadcast = v
+ e.UnlockUser()
+
+ case tcpip.CorkOption:
+ e.LockUser()
+ if !v {
+ atomic.StoreUint32(&e.cork, 0)
+
+ // Handle the corked data.
+ e.sndWaker.Assert()
+ } else {
+ atomic.StoreUint32(&e.cork, 1)
+ }
+ e.UnlockUser()
+
+ case tcpip.DelayOption:
+ if v {
+ atomic.StoreUint32(&e.delay, 1)
+ } else {
+ atomic.StoreUint32(&e.delay, 0)
+
+ // Handle delayed data.
+ e.sndWaker.Assert()
+ }
+
+ case tcpip.KeepaliveEnabledOption:
+ e.keepalive.Lock()
+ e.keepalive.enabled = v
+ e.keepalive.Unlock()
+ e.notifyProtocolGoroutine(notifyKeepaliveChanged)
+
+ case tcpip.QuickAckOption:
+ o := uint32(1)
+ if v {
+ o = 0
+ }
+ atomic.StoreUint32(&e.slowAck, o)
+
+ case tcpip.ReuseAddressOption:
+ e.LockUser()
+ e.reuseAddr = v
+ e.UnlockUser()
+ return nil
+
+ case tcpip.ReusePortOption:
+ e.LockUser()
+ e.reusePort = v
+ e.UnlockUser()
+ return nil
+
case tcpip.V6OnlyOption:
// We only recognize this option on v6 endpoints.
if e.NetProto != header.IPv6ProtocolNumber {
@@ -1424,7 +1474,11 @@ func (e *endpoint) SetSockOptBool(opt tcpip.SockOptBool, v bool) *tcpip.Error {
return tcpip.ErrInvalidEndpointState
}
+ e.LockUser()
e.v6only = v
+ e.UnlockUser()
+ default:
+ return tcpip.ErrUnknownProtocolOption
}
return nil
@@ -1432,7 +1486,40 @@ func (e *endpoint) SetSockOptBool(opt tcpip.SockOptBool, v bool) *tcpip.Error {
// SetSockOptInt sets a socket option.
func (e *endpoint) SetSockOptInt(opt tcpip.SockOptInt, v int) *tcpip.Error {
+ // Lower 2 bits represents ECN bits. RFC 3168, section 23.1
+ const inetECNMask = 3
+
switch opt {
+ case tcpip.KeepaliveCountOption:
+ e.keepalive.Lock()
+ e.keepalive.count = int(v)
+ e.keepalive.Unlock()
+ e.notifyProtocolGoroutine(notifyKeepaliveChanged)
+
+ case tcpip.IPv4TOSOption:
+ e.LockUser()
+ // TODO(gvisor.dev/issue/995): ECN is not currently supported,
+ // ignore the bits for now.
+ e.sendTOS = uint8(v) & ^uint8(inetECNMask)
+ e.UnlockUser()
+
+ case tcpip.IPv6TrafficClassOption:
+ e.LockUser()
+ // TODO(gvisor.dev/issue/995): ECN is not currently supported,
+ // ignore the bits for now.
+ e.sendTOS = uint8(v) & ^uint8(inetECNMask)
+ e.UnlockUser()
+
+ case tcpip.MaxSegOption:
+ userMSS := v
+ if userMSS < header.TCPMinimumMSS || userMSS > header.TCPMaximumMSS {
+ return tcpip.ErrInvalidOptionValue
+ }
+ e.LockUser()
+ e.userMSS = uint16(userMSS)
+ e.UnlockUser()
+ e.notifyProtocolGoroutine(notifyMSSChanged)
+
case tcpip.ReceiveBufferSizeOption:
// Make sure the receive buffer size is within the min and max
// allowed.
@@ -1483,7 +1570,6 @@ func (e *endpoint) SetSockOptInt(opt tcpip.SockOptInt, v int) *tcpip.Error {
e.rcvListMu.Unlock()
e.UnlockUser()
e.notifyProtocolGoroutine(mask)
- return nil
case tcpip.SendBufferSizeOption:
// Make sure the send buffer size is within the min and max
@@ -1502,52 +1588,21 @@ func (e *endpoint) SetSockOptInt(opt tcpip.SockOptInt, v int) *tcpip.Error {
e.sndBufMu.Lock()
e.sndBufSize = size
e.sndBufMu.Unlock()
- return nil
- case tcpip.DelayOption:
- if v == 0 {
- atomic.StoreUint32(&e.delay, 0)
-
- // Handle delayed data.
- e.sndWaker.Assert()
- } else {
- atomic.StoreUint32(&e.delay, 1)
- }
- return nil
+ case tcpip.TTLOption:
+ e.LockUser()
+ e.ttl = uint8(v)
+ e.UnlockUser()
default:
- return nil
+ return tcpip.ErrUnknownProtocolOption
}
+ return nil
}
// SetSockOpt sets a socket option.
func (e *endpoint) SetSockOpt(opt interface{}) *tcpip.Error {
- // Lower 2 bits represents ECN bits. RFC 3168, section 23.1
- const inetECNMask = 3
switch v := opt.(type) {
- case tcpip.CorkOption:
- if v == 0 {
- atomic.StoreUint32(&e.cork, 0)
-
- // Handle the corked data.
- e.sndWaker.Assert()
- } else {
- atomic.StoreUint32(&e.cork, 1)
- }
- return nil
-
- case tcpip.ReuseAddressOption:
- e.LockUser()
- e.reuseAddr = v != 0
- e.UnlockUser()
- return nil
-
- case tcpip.ReusePortOption:
- e.LockUser()
- e.reusePort = v != 0
- e.UnlockUser()
- return nil
-
case tcpip.BindToDeviceOption:
id := tcpip.NICID(v)
if id != 0 && !e.stack.HasNIC(id) {
@@ -1556,72 +1611,26 @@ func (e *endpoint) SetSockOpt(opt interface{}) *tcpip.Error {
e.LockUser()
e.bindToDevice = id
e.UnlockUser()
- return nil
-
- case tcpip.QuickAckOption:
- if v == 0 {
- atomic.StoreUint32(&e.slowAck, 1)
- } else {
- atomic.StoreUint32(&e.slowAck, 0)
- }
- return nil
-
- case tcpip.MaxSegOption:
- userMSS := v
- if userMSS < header.TCPMinimumMSS || userMSS > header.TCPMaximumMSS {
- return tcpip.ErrInvalidOptionValue
- }
- e.LockUser()
- e.userMSS = uint16(userMSS)
- e.UnlockUser()
- e.notifyProtocolGoroutine(notifyMSSChanged)
- return nil
-
- case tcpip.TTLOption:
- e.LockUser()
- e.ttl = uint8(v)
- e.UnlockUser()
- return nil
-
- case tcpip.KeepaliveEnabledOption:
- e.keepalive.Lock()
- e.keepalive.enabled = v != 0
- e.keepalive.Unlock()
- e.notifyProtocolGoroutine(notifyKeepaliveChanged)
- return nil
case tcpip.KeepaliveIdleOption:
e.keepalive.Lock()
e.keepalive.idle = time.Duration(v)
e.keepalive.Unlock()
e.notifyProtocolGoroutine(notifyKeepaliveChanged)
- return nil
case tcpip.KeepaliveIntervalOption:
e.keepalive.Lock()
e.keepalive.interval = time.Duration(v)
e.keepalive.Unlock()
e.notifyProtocolGoroutine(notifyKeepaliveChanged)
- return nil
- case tcpip.KeepaliveCountOption:
- e.keepalive.Lock()
- e.keepalive.count = int(v)
- e.keepalive.Unlock()
- e.notifyProtocolGoroutine(notifyKeepaliveChanged)
- return nil
+ case tcpip.OutOfBandInlineOption:
+ // We don't currently support disabling this option.
case tcpip.TCPUserTimeoutOption:
e.LockUser()
e.userTimeout = time.Duration(v)
e.UnlockUser()
- return nil
-
- case tcpip.BroadcastOption:
- e.LockUser()
- e.broadcast = v != 0
- e.UnlockUser()
- return nil
case tcpip.CongestionControlOption:
// Query the available cc algorithms in the stack and
@@ -1652,22 +1661,6 @@ func (e *endpoint) SetSockOpt(opt interface{}) *tcpip.Error {
// control algorithm is specified.
return tcpip.ErrNoSuchFile
- case tcpip.IPv4TOSOption:
- e.LockUser()
- // TODO(gvisor.dev/issue/995): ECN is not currently supported,
- // ignore the bits for now.
- e.sendTOS = uint8(v) & ^uint8(inetECNMask)
- e.UnlockUser()
- return nil
-
- case tcpip.IPv6TrafficClassOption:
- e.LockUser()
- // TODO(gvisor.dev/issue/995): ECN is not currently supported,
- // ignore the bits for now.
- e.sendTOS = uint8(v) & ^uint8(inetECNMask)
- e.UnlockUser()
- return nil
-
case tcpip.TCPLingerTimeoutOption:
e.LockUser()
if v < 0 {
@@ -1688,7 +1681,6 @@ func (e *endpoint) SetSockOpt(opt interface{}) *tcpip.Error {
}
e.tcpLingerTimeout = time.Duration(v)
e.UnlockUser()
- return nil
case tcpip.TCPDeferAcceptOption:
e.LockUser()
@@ -1697,11 +1689,11 @@ func (e *endpoint) SetSockOpt(opt interface{}) *tcpip.Error {
}
e.deferAccept = time.Duration(v)
e.UnlockUser()
- return nil
default:
return nil
}
+ return nil
}
// readyReceiveSize returns the number of bytes ready to be received.
@@ -1723,6 +1715,43 @@ func (e *endpoint) readyReceiveSize() (int, *tcpip.Error) {
// GetSockOptBool implements tcpip.Endpoint.GetSockOptBool.
func (e *endpoint) GetSockOptBool(opt tcpip.SockOptBool) (bool, *tcpip.Error) {
switch opt {
+ case tcpip.BroadcastOption:
+ e.LockUser()
+ v := e.broadcast
+ e.UnlockUser()
+ return v, nil
+
+ case tcpip.CorkOption:
+ return atomic.LoadUint32(&e.cork) != 0, nil
+
+ case tcpip.DelayOption:
+ return atomic.LoadUint32(&e.delay) != 0, nil
+
+ case tcpip.KeepaliveEnabledOption:
+ e.keepalive.Lock()
+ v := e.keepalive.enabled
+ e.keepalive.Unlock()
+
+ return v, nil
+
+ case tcpip.QuickAckOption:
+ v := atomic.LoadUint32(&e.slowAck) == 0
+ return v, nil
+
+ case tcpip.ReuseAddressOption:
+ e.LockUser()
+ v := e.reuseAddr
+ e.UnlockUser()
+
+ return v, nil
+
+ case tcpip.ReusePortOption:
+ e.LockUser()
+ v := e.reusePort
+ e.UnlockUser()
+
+ return v, nil
+
case tcpip.V6OnlyOption:
// We only recognize this option on v6 endpoints.
if e.NetProto != header.IPv6ProtocolNumber {
@@ -1734,14 +1763,41 @@ func (e *endpoint) GetSockOptBool(opt tcpip.SockOptBool) (bool, *tcpip.Error) {
e.UnlockUser()
return v, nil
- }
- return false, tcpip.ErrUnknownProtocolOption
+ default:
+ return false, tcpip.ErrUnknownProtocolOption
+ }
}
// GetSockOptInt implements tcpip.Endpoint.GetSockOptInt.
func (e *endpoint) GetSockOptInt(opt tcpip.SockOptInt) (int, *tcpip.Error) {
switch opt {
+ case tcpip.KeepaliveCountOption:
+ e.keepalive.Lock()
+ v := e.keepalive.count
+ e.keepalive.Unlock()
+ return v, nil
+
+ case tcpip.IPv4TOSOption:
+ e.LockUser()
+ v := int(e.sendTOS)
+ e.UnlockUser()
+ return v, nil
+
+ case tcpip.IPv6TrafficClassOption:
+ e.LockUser()
+ v := int(e.sendTOS)
+ e.UnlockUser()
+ return v, nil
+
+ case tcpip.MaxSegOption:
+ // This is just stubbed out. Linux never returns the user_mss
+ // value as it either returns the defaultMSS or returns the
+ // actual current MSS. Netstack just returns the defaultMSS
+ // always for now.
+ v := header.TCPDefaultMSS
+ return v, nil
+
case tcpip.ReceiveQueueSizeOption:
return e.readyReceiveSize()
@@ -1757,12 +1813,11 @@ func (e *endpoint) GetSockOptInt(opt tcpip.SockOptInt) (int, *tcpip.Error) {
e.rcvListMu.Unlock()
return v, nil
- case tcpip.DelayOption:
- var o int
- if v := atomic.LoadUint32(&e.delay); v != 0 {
- o = 1
- }
- return o, nil
+ case tcpip.TTLOption:
+ e.LockUser()
+ v := int(e.ttl)
+ e.UnlockUser()
+ return v, nil
default:
return -1, tcpip.ErrUnknownProtocolOption
@@ -1779,61 +1834,10 @@ func (e *endpoint) GetSockOpt(opt interface{}) *tcpip.Error {
e.lastErrorMu.Unlock()
return err
- case *tcpip.MaxSegOption:
- // This is just stubbed out. Linux never returns the user_mss
- // value as it either returns the defaultMSS or returns the
- // actual current MSS. Netstack just returns the defaultMSS
- // always for now.
- *o = header.TCPDefaultMSS
- return nil
-
- case *tcpip.CorkOption:
- *o = 0
- if v := atomic.LoadUint32(&e.cork); v != 0 {
- *o = 1
- }
- return nil
-
- case *tcpip.ReuseAddressOption:
- e.LockUser()
- v := e.reuseAddr
- e.UnlockUser()
-
- *o = 0
- if v {
- *o = 1
- }
- return nil
-
- case *tcpip.ReusePortOption:
- e.LockUser()
- v := e.reusePort
- e.UnlockUser()
-
- *o = 0
- if v {
- *o = 1
- }
- return nil
-
case *tcpip.BindToDeviceOption:
e.LockUser()
*o = tcpip.BindToDeviceOption(e.bindToDevice)
e.UnlockUser()
- return nil
-
- case *tcpip.QuickAckOption:
- *o = 1
- if v := atomic.LoadUint32(&e.slowAck); v != 0 {
- *o = 0
- }
- return nil
-
- case *tcpip.TTLOption:
- e.LockUser()
- *o = tcpip.TTLOption(e.ttl)
- e.UnlockUser()
- return nil
case *tcpip.TCPInfoOption:
*o = tcpip.TCPInfoOption{}
@@ -1846,92 +1850,45 @@ func (e *endpoint) GetSockOpt(opt interface{}) *tcpip.Error {
o.RTTVar = snd.rtt.rttvar
snd.rtt.Unlock()
}
- return nil
-
- case *tcpip.KeepaliveEnabledOption:
- e.keepalive.Lock()
- v := e.keepalive.enabled
- e.keepalive.Unlock()
-
- *o = 0
- if v {
- *o = 1
- }
- return nil
case *tcpip.KeepaliveIdleOption:
e.keepalive.Lock()
*o = tcpip.KeepaliveIdleOption(e.keepalive.idle)
e.keepalive.Unlock()
- return nil
case *tcpip.KeepaliveIntervalOption:
e.keepalive.Lock()
*o = tcpip.KeepaliveIntervalOption(e.keepalive.interval)
e.keepalive.Unlock()
- return nil
-
- case *tcpip.KeepaliveCountOption:
- e.keepalive.Lock()
- *o = tcpip.KeepaliveCountOption(e.keepalive.count)
- e.keepalive.Unlock()
- return nil
case *tcpip.TCPUserTimeoutOption:
e.LockUser()
*o = tcpip.TCPUserTimeoutOption(e.userTimeout)
e.UnlockUser()
- return nil
case *tcpip.OutOfBandInlineOption:
// We don't currently support disabling this option.
*o = 1
- return nil
-
- case *tcpip.BroadcastOption:
- e.LockUser()
- v := e.broadcast
- e.UnlockUser()
-
- *o = 0
- if v {
- *o = 1
- }
- return nil
case *tcpip.CongestionControlOption:
e.LockUser()
*o = e.cc
e.UnlockUser()
- return nil
-
- case *tcpip.IPv4TOSOption:
- e.LockUser()
- *o = tcpip.IPv4TOSOption(e.sendTOS)
- e.UnlockUser()
- return nil
-
- case *tcpip.IPv6TrafficClassOption:
- e.LockUser()
- *o = tcpip.IPv6TrafficClassOption(e.sendTOS)
- e.UnlockUser()
- return nil
case *tcpip.TCPLingerTimeoutOption:
e.LockUser()
*o = tcpip.TCPLingerTimeoutOption(e.tcpLingerTimeout)
e.UnlockUser()
- return nil
case *tcpip.TCPDeferAcceptOption:
e.LockUser()
*o = tcpip.TCPDeferAcceptOption(e.deferAccept)
e.UnlockUser()
- return nil
default:
return tcpip.ErrUnknownProtocolOption
}
+ return nil
}
// checkV4MappedLocked determines the effective network protocol and converts
diff --git a/pkg/tcpip/transport/tcp/tcp_test.go b/pkg/tcpip/transport/tcp/tcp_test.go
index ce3df7478..29301a45c 100644
--- a/pkg/tcpip/transport/tcp/tcp_test.go
+++ b/pkg/tcpip/transport/tcp/tcp_test.go
@@ -284,7 +284,7 @@ func TestTCPResetSentForACKWhenNotUsingSynCookies(t *testing.T) {
// are released instantly on Close.
tcpTW := tcpip.TCPTimeWaitTimeoutOption(1 * time.Millisecond)
if err := c.Stack().SetTransportProtocolOption(tcp.ProtocolNumber, tcpTW); err != nil {
- t.Fatalf("e.stack.SetTransportProtocolOption(%d, %s) = %s", tcp.ProtocolNumber, tcpTW, err)
+ t.Fatalf("e.stack.SetTransportProtocolOption(%d, %v) = %v", tcp.ProtocolNumber, tcpTW, err)
}
c.EP.Close()
@@ -728,7 +728,7 @@ func TestUserSuppliedMSSOnConnectV4(t *testing.T) {
const maxMSS = mtu - header.IPv4MinimumSize - header.TCPMinimumSize
tests := []struct {
name string
- setMSS uint16
+ setMSS int
expMSS uint16
}{
{
@@ -756,15 +756,14 @@ func TestUserSuppliedMSSOnConnectV4(t *testing.T) {
c.Create(-1)
// Set the MSS socket option.
- opt := tcpip.MaxSegOption(test.setMSS)
- if err := c.EP.SetSockOpt(opt); err != nil {
- t.Fatalf("SetSockOpt(%#v) failed: %s", opt, err)
+ if err := c.EP.SetSockOptInt(tcpip.MaxSegOption, test.setMSS); err != nil {
+ t.Fatalf("SetSockOptInt(MaxSegOption, %d) failed: %s", test.setMSS, err)
}
// Get expected window size.
rcvBufSize, err := c.EP.GetSockOptInt(tcpip.ReceiveBufferSizeOption)
if err != nil {
- t.Fatalf("GetSockOpt(%v) failed: %s", tcpip.ReceiveBufferSizeOption, err)
+ t.Fatalf("GetSockOptInt(ReceiveBufferSizeOption) failed: %s", err)
}
ws := tcp.FindWndScale(seqnum.Size(rcvBufSize))
@@ -818,15 +817,14 @@ func TestUserSuppliedMSSOnConnectV6(t *testing.T) {
c.CreateV6Endpoint(true)
// Set the MSS socket option.
- opt := tcpip.MaxSegOption(test.setMSS)
- if err := c.EP.SetSockOpt(opt); err != nil {
- t.Fatalf("SetSockOpt(%#v) failed: %s", opt, err)
+ if err := c.EP.SetSockOptInt(tcpip.MaxSegOption, int(test.setMSS)); err != nil {
+ t.Fatalf("SetSockOptInt(MaxSegOption, %d) failed: %s", test.setMSS, err)
}
// Get expected window size.
rcvBufSize, err := c.EP.GetSockOptInt(tcpip.ReceiveBufferSizeOption)
if err != nil {
- t.Fatalf("GetSockOpt(%v) failed: %s", tcpip.ReceiveBufferSizeOption, err)
+ t.Fatalf("GetSockOptInt(ReceiveBufferSizeOption) failed: %s", err)
}
ws := tcp.FindWndScale(seqnum.Size(rcvBufSize))
@@ -1077,17 +1075,17 @@ func TestTOSV4(t *testing.T) {
c.EP = ep
const tos = 0xC0
- if err := c.EP.SetSockOpt(tcpip.IPv4TOSOption(tos)); err != nil {
- t.Errorf("SetSockOpt(%#v) failed: %s", tcpip.IPv4TOSOption(tos), err)
+ if err := c.EP.SetSockOptInt(tcpip.IPv4TOSOption, tos); err != nil {
+ t.Errorf("SetSockOptInt(IPv4TOSOption, %d) failed: %s", tos, err)
}
- var v tcpip.IPv4TOSOption
- if err := c.EP.GetSockOpt(&v); err != nil {
- t.Errorf("GetSockopt failed: %s", err)
+ v, err := c.EP.GetSockOptInt(tcpip.IPv4TOSOption)
+ if err != nil {
+ t.Errorf("GetSockoptInt(IPv4TOSOption) failed: %s", err)
}
- if want := tcpip.IPv4TOSOption(tos); v != want {
- t.Errorf("got GetSockOpt(...) = %#v, want = %#v", v, want)
+ if v != tos {
+ t.Errorf("got GetSockOptInt(IPv4TOSOption) = %d, want = %d", v, tos)
}
testV4Connect(t, c, checker.TOS(tos, 0))
@@ -1125,17 +1123,17 @@ func TestTrafficClassV6(t *testing.T) {
c.CreateV6Endpoint(false)
const tos = 0xC0
- if err := c.EP.SetSockOpt(tcpip.IPv6TrafficClassOption(tos)); err != nil {
- t.Errorf("SetSockOpt(%#v) failed: %s", tcpip.IPv6TrafficClassOption(tos), err)
+ if err := c.EP.SetSockOptInt(tcpip.IPv6TrafficClassOption, tos); err != nil {
+ t.Errorf("SetSockOpInt(IPv6TrafficClassOption, %d) failed: %s", tos, err)
}
- var v tcpip.IPv6TrafficClassOption
- if err := c.EP.GetSockOpt(&v); err != nil {
- t.Fatalf("GetSockopt failed: %s", err)
+ v, err := c.EP.GetSockOptInt(tcpip.IPv6TrafficClassOption)
+ if err != nil {
+ t.Fatalf("GetSockoptInt(IPv6TrafficClassOption) failed: %s", err)
}
- if want := tcpip.IPv6TrafficClassOption(tos); v != want {
- t.Errorf("got GetSockOpt(...) = %#v, want = %#v", v, want)
+ if v != tos {
+ t.Errorf("got GetSockOptInt(IPv6TrafficClassOption) = %d, want = %d", v, tos)
}
// Test the connection request.
@@ -1711,7 +1709,7 @@ func TestNoWindowShrinking(t *testing.T) {
c.CreateConnected(789, 30000, 10)
if err := c.EP.SetSockOptInt(tcpip.ReceiveBufferSizeOption, 5); err != nil {
- t.Fatalf("SetSockOpt failed: %v", err)
+ t.Fatalf("SetSockOptInt(ReceiveBufferSizeOption, 5) failed: %v", err)
}
we, ch := waiter.NewChannelEntry(nil)
@@ -1984,7 +1982,7 @@ func TestScaledWindowAccept(t *testing.T) {
// Set the window size greater than the maximum non-scaled window.
if err := ep.SetSockOptInt(tcpip.ReceiveBufferSizeOption, 65535*3); err != nil {
- t.Fatalf("SetSockOpt failed failed: %v", err)
+ t.Fatalf("SetSockOptInt(ReceiveBufferSizeOption, 65535*3) failed failed: %v", err)
}
if err := ep.Bind(tcpip.FullAddress{Port: context.StackPort}); err != nil {
@@ -2057,7 +2055,7 @@ func TestNonScaledWindowAccept(t *testing.T) {
// Set the window size greater than the maximum non-scaled window.
if err := ep.SetSockOptInt(tcpip.ReceiveBufferSizeOption, 65535*3); err != nil {
- t.Fatalf("SetSockOpt failed failed: %v", err)
+ t.Fatalf("SetSockOptInt(ReceiveBufferSizeOption, 65535*3) failed failed: %v", err)
}
if err := ep.Bind(tcpip.FullAddress{Port: context.StackPort}); err != nil {
@@ -2221,10 +2219,10 @@ func TestSegmentMerging(t *testing.T) {
{
"cork",
func(ep tcpip.Endpoint) {
- ep.SetSockOpt(tcpip.CorkOption(1))
+ ep.SetSockOptBool(tcpip.CorkOption, true)
},
func(ep tcpip.Endpoint) {
- ep.SetSockOpt(tcpip.CorkOption(0))
+ ep.SetSockOptBool(tcpip.CorkOption, false)
},
},
}
@@ -2316,7 +2314,7 @@ func TestDelay(t *testing.T) {
c.CreateConnected(789, 30000, -1 /* epRcvBuf */)
- c.EP.SetSockOptInt(tcpip.DelayOption, 1)
+ c.EP.SetSockOptBool(tcpip.DelayOption, true)
var allData []byte
for i, data := range [][]byte{{0}, {1, 2, 3, 4}, {5, 6, 7}, {8, 9}, {10}, {11}} {
@@ -2364,7 +2362,7 @@ func TestUndelay(t *testing.T) {
c.CreateConnected(789, 30000, -1 /* epRcvBuf */)
- c.EP.SetSockOptInt(tcpip.DelayOption, 1)
+ c.EP.SetSockOptBool(tcpip.DelayOption, true)
allData := [][]byte{{0}, {1, 2, 3}}
for i, data := range allData {
@@ -2397,7 +2395,7 @@ func TestUndelay(t *testing.T) {
// Check that we don't get the second packet yet.
c.CheckNoPacketTimeout("delayed second packet transmitted", 100*time.Millisecond)
- c.EP.SetSockOptInt(tcpip.DelayOption, 0)
+ c.EP.SetSockOptBool(tcpip.DelayOption, false)
// Check that data is received.
second := c.GetPacket()
@@ -2434,8 +2432,8 @@ func TestMSSNotDelayed(t *testing.T) {
fn func(tcpip.Endpoint)
}{
{"no-op", func(tcpip.Endpoint) {}},
- {"delay", func(ep tcpip.Endpoint) { ep.SetSockOptInt(tcpip.DelayOption, 1) }},
- {"cork", func(ep tcpip.Endpoint) { ep.SetSockOpt(tcpip.CorkOption(1)) }},
+ {"delay", func(ep tcpip.Endpoint) { ep.SetSockOptBool(tcpip.DelayOption, true) }},
+ {"cork", func(ep tcpip.Endpoint) { ep.SetSockOptBool(tcpip.CorkOption, true) }},
}
for _, test := range tests {
@@ -2576,12 +2574,12 @@ func TestSetTTL(t *testing.T) {
t.Fatalf("NewEndpoint failed: %v", err)
}
- if err := c.EP.SetSockOpt(tcpip.TTLOption(wantTTL)); err != nil {
- t.Fatalf("SetSockOpt failed: %v", err)
+ if err := c.EP.SetSockOptInt(tcpip.TTLOption, int(wantTTL)); err != nil {
+ t.Fatalf("SetSockOptInt(TTLOption, %d) failed: %s", wantTTL, err)
}
if err := c.EP.Connect(tcpip.FullAddress{Addr: context.TestAddr, Port: context.TestPort}); err != tcpip.ErrConnectStarted {
- t.Fatalf("Unexpected return value from Connect: %v", err)
+ t.Fatalf("Unexpected return value from Connect: %s", err)
}
// Receive SYN packet.
@@ -2621,7 +2619,7 @@ func TestPassiveSendMSSLessThanMTU(t *testing.T) {
// window scaling option.
const rcvBufferSize = 0x20000
if err := ep.SetSockOptInt(tcpip.ReceiveBufferSizeOption, rcvBufferSize); err != nil {
- t.Fatalf("SetSockOpt failed failed: %v", err)
+ t.Fatalf("SetSockOptInt(ReceiveBufferSizeOption, %d) failed failed: %s", rcvBufferSize, err)
}
if err := ep.Bind(tcpip.FullAddress{Port: context.StackPort}); err != nil {
@@ -2765,7 +2763,7 @@ func TestSynOptionsOnActiveConnect(t *testing.T) {
const rcvBufferSize = 0x20000
const wndScale = 2
if err := c.EP.SetSockOptInt(tcpip.ReceiveBufferSizeOption, rcvBufferSize); err != nil {
- t.Fatalf("SetSockOpt failed failed: %v", err)
+ t.Fatalf("SetSockOptInt(ReceiveBufferSizeOption, %d) failed failed: %s", rcvBufferSize, err)
}
// Start connection attempt.
@@ -3882,26 +3880,26 @@ func TestMinMaxBufferSizes(t *testing.T) {
// Set values below the min.
if err := ep.SetSockOptInt(tcpip.ReceiveBufferSizeOption, 199); err != nil {
- t.Fatalf("GetSockOpt failed: %v", err)
+ t.Fatalf("SetSockOptInt(ReceiveBufferSizeOption, 199) failed: %s", err)
}
checkRecvBufferSize(t, ep, 200)
if err := ep.SetSockOptInt(tcpip.SendBufferSizeOption, 299); err != nil {
- t.Fatalf("GetSockOpt failed: %v", err)
+ t.Fatalf("SetSockOptInt(SendBufferSizeOption, 299) failed: %s", err)
}
checkSendBufferSize(t, ep, 300)
// Set values above the max.
if err := ep.SetSockOptInt(tcpip.ReceiveBufferSizeOption, 1+tcp.DefaultReceiveBufferSize*20); err != nil {
- t.Fatalf("GetSockOpt failed: %v", err)
+ t.Fatalf("SetSockOptInt(ReceiveBufferSizeOption) failed: %s", err)
}
checkRecvBufferSize(t, ep, tcp.DefaultReceiveBufferSize*20)
if err := ep.SetSockOptInt(tcpip.SendBufferSizeOption, 1+tcp.DefaultSendBufferSize*30); err != nil {
- t.Fatalf("GetSockOpt failed: %v", err)
+ t.Fatalf("SetSockOptInt(SendBufferSizeOption) failed: %s", err)
}
checkSendBufferSize(t, ep, tcp.DefaultSendBufferSize*30)
@@ -4147,11 +4145,11 @@ func TestConnectAvoidsBoundPorts(t *testing.T) {
case "ipv4":
case "ipv6":
if err := ep.SetSockOptBool(tcpip.V6OnlyOption, true); err != nil {
- t.Fatalf("SetSockOpt(V6OnlyOption(true)) failed: %v", err)
+ t.Fatalf("SetSockOptBool(V6OnlyOption(true)) failed: %s", err)
}
case "dual":
if err := ep.SetSockOptBool(tcpip.V6OnlyOption, false); err != nil {
- t.Fatalf("SetSockOpt(V6OnlyOption(false)) failed: %v", err)
+ t.Fatalf("SetSockOptBool(V6OnlyOption(false)) failed: %s", err)
}
default:
t.Fatalf("unknown network: '%s'", network)
@@ -4477,8 +4475,8 @@ func TestKeepalive(t *testing.T) {
const keepAliveInterval = 10 * time.Millisecond
c.EP.SetSockOpt(tcpip.KeepaliveIdleOption(10 * time.Millisecond))
c.EP.SetSockOpt(tcpip.KeepaliveIntervalOption(keepAliveInterval))
- c.EP.SetSockOpt(tcpip.KeepaliveCountOption(5))
- c.EP.SetSockOpt(tcpip.KeepaliveEnabledOption(1))
+ c.EP.SetSockOptInt(tcpip.KeepaliveCountOption, 5)
+ c.EP.SetSockOptBool(tcpip.KeepaliveEnabledOption, true)
// 5 unacked keepalives are sent. ACK each one, and check that the
// connection stays alive after 5.
@@ -5609,7 +5607,7 @@ func TestReceiveBufferAutoTuningApplicationLimited(t *testing.T) {
return
}
if w := tcp.WindowSize(); w == 0 || w > uint16(wantRcvWnd) {
- t.Errorf("expected a non-zero window: got %d, want <= wantRcvWnd", w, wantRcvWnd)
+ t.Errorf("expected a non-zero window: got %d, want <= wantRcvWnd", w)
}
},
))
@@ -5770,14 +5768,14 @@ func TestReceiveBufferAutoTuning(t *testing.T) {
func TestDelayEnabled(t *testing.T) {
c := context.New(t, defaultMTU)
defer c.Cleanup()
- checkDelayOption(t, c, false, 0) // Delay is disabled by default.
+ checkDelayOption(t, c, false, false) // Delay is disabled by default.
for _, v := range []struct {
delayEnabled tcp.DelayEnabled
- wantDelayOption int
+ wantDelayOption bool
}{
- {delayEnabled: false, wantDelayOption: 0},
- {delayEnabled: true, wantDelayOption: 1},
+ {delayEnabled: false, wantDelayOption: false},
+ {delayEnabled: true, wantDelayOption: true},
} {
c := context.New(t, defaultMTU)
defer c.Cleanup()
@@ -5788,7 +5786,7 @@ func TestDelayEnabled(t *testing.T) {
}
}
-func checkDelayOption(t *testing.T, c *context.Context, wantDelayEnabled tcp.DelayEnabled, wantDelayOption int) {
+func checkDelayOption(t *testing.T, c *context.Context, wantDelayEnabled tcp.DelayEnabled, wantDelayOption bool) {
t.Helper()
var gotDelayEnabled tcp.DelayEnabled
@@ -5803,12 +5801,12 @@ func checkDelayOption(t *testing.T, c *context.Context, wantDelayEnabled tcp.Del
if err != nil {
t.Fatalf("NewEndPoint(tcp, ipv4, new(waiter.Queue)) failed: %v", err)
}
- gotDelayOption, err := ep.GetSockOptInt(tcpip.DelayOption)
+ gotDelayOption, err := ep.GetSockOptBool(tcpip.DelayOption)
if err != nil {
- t.Fatalf("ep.GetSockOptInt(tcpip.DelayOption) failed: %v", err)
+ t.Fatalf("ep.GetSockOptBool(tcpip.DelayOption) failed: %s", err)
}
if gotDelayOption != wantDelayOption {
- t.Errorf("ep.GetSockOptInt(tcpip.DelayOption) got: %d, want: %d", gotDelayOption, wantDelayOption)
+ t.Errorf("ep.GetSockOptBool(tcpip.DelayOption) got: %t, want: %t", gotDelayOption, wantDelayOption)
}
}
@@ -6620,8 +6618,8 @@ func TestKeepaliveWithUserTimeout(t *testing.T) {
const keepAliveInterval = 10 * time.Millisecond
c.EP.SetSockOpt(tcpip.KeepaliveIdleOption(10 * time.Millisecond))
c.EP.SetSockOpt(tcpip.KeepaliveIntervalOption(keepAliveInterval))
- c.EP.SetSockOpt(tcpip.KeepaliveCountOption(10))
- c.EP.SetSockOpt(tcpip.KeepaliveEnabledOption(1))
+ c.EP.SetSockOptInt(tcpip.KeepaliveCountOption, 10)
+ c.EP.SetSockOptBool(tcpip.KeepaliveEnabledOption, true)
// Set userTimeout to be the duration for 3 keepalive probes.
userTimeout := 30 * time.Millisecond
diff --git a/pkg/tcpip/transport/tcp/testing/context/context.go b/pkg/tcpip/transport/tcp/testing/context/context.go
index d4f6bc635..431ab4e6b 100644
--- a/pkg/tcpip/transport/tcp/testing/context/context.go
+++ b/pkg/tcpip/transport/tcp/testing/context/context.go
@@ -217,7 +217,8 @@ func (c *Context) Stack() *stack.Stack {
func (c *Context) CheckNoPacketTimeout(errMsg string, wait time.Duration) {
c.t.Helper()
- ctx, _ := context.WithTimeout(context.Background(), wait)
+ ctx, cancel := context.WithTimeout(context.Background(), wait)
+ defer cancel()
if _, ok := c.linkEP.ReadContext(ctx); ok {
c.t.Fatal(errMsg)
}
@@ -235,7 +236,8 @@ func (c *Context) CheckNoPacket(errMsg string) {
func (c *Context) GetPacket() []byte {
c.t.Helper()
- ctx, _ := context.WithTimeout(context.Background(), 2*time.Second)
+ ctx, cancel := context.WithTimeout(context.Background(), 2*time.Second)
+ defer cancel()
p, ok := c.linkEP.ReadContext(ctx)
if !ok {
c.t.Fatalf("Packet wasn't written out")
@@ -486,7 +488,8 @@ func (c *Context) CreateV6Endpoint(v6only bool) {
func (c *Context) GetV6Packet() []byte {
c.t.Helper()
- ctx, _ := context.WithTimeout(context.Background(), 2*time.Second)
+ ctx, cancel := context.WithTimeout(context.Background(), 2*time.Second)
+ defer cancel()
p, ok := c.linkEP.ReadContext(ctx)
if !ok {
c.t.Fatalf("Packet wasn't written out")
diff --git a/pkg/tcpip/transport/udp/endpoint.go b/pkg/tcpip/transport/udp/endpoint.go
index 120d3baa3..492cc1fcb 100644
--- a/pkg/tcpip/transport/udp/endpoint.go
+++ b/pkg/tcpip/transport/udp/endpoint.go
@@ -501,11 +501,20 @@ func (e *endpoint) Peek([][]byte) (int64, tcpip.ControlMessages, *tcpip.Error) {
// SetSockOptBool implements tcpip.Endpoint.SetSockOptBool.
func (e *endpoint) SetSockOptBool(opt tcpip.SockOptBool, v bool) *tcpip.Error {
switch opt {
+ case tcpip.BroadcastOption:
+ e.mu.Lock()
+ e.broadcast = v
+ e.mu.Unlock()
+
+ case tcpip.MulticastLoopOption:
+ e.mu.Lock()
+ e.multicastLoop = v
+ e.mu.Unlock()
+
case tcpip.ReceiveTOSOption:
e.mu.Lock()
e.receiveTOS = v
e.mu.Unlock()
- return nil
case tcpip.ReceiveTClassOption:
// We only support this option on v6 endpoints.
@@ -516,7 +525,18 @@ func (e *endpoint) SetSockOptBool(opt tcpip.SockOptBool, v bool) *tcpip.Error {
e.mu.Lock()
e.receiveTClass = v
e.mu.Unlock()
- return nil
+
+ case tcpip.ReceiveIPPacketInfoOption:
+ e.mu.Lock()
+ e.receiveIPPacketInfo = v
+ e.mu.Unlock()
+
+ case tcpip.ReuseAddressOption:
+
+ case tcpip.ReusePortOption:
+ e.mu.Lock()
+ e.reusePort = v
+ e.mu.Unlock()
case tcpip.V6OnlyOption:
// We only recognize this option on v6 endpoints.
@@ -533,13 +553,8 @@ func (e *endpoint) SetSockOptBool(opt tcpip.SockOptBool, v bool) *tcpip.Error {
}
e.v6only = v
- return nil
-
- case tcpip.ReceiveIPPacketInfoOption:
- e.mu.Lock()
- e.receiveIPPacketInfo = v
- e.mu.Unlock()
- return nil
+ default:
+ return tcpip.ErrUnknownProtocolOption
}
return nil
@@ -547,22 +562,40 @@ func (e *endpoint) SetSockOptBool(opt tcpip.SockOptBool, v bool) *tcpip.Error {
// SetSockOptInt implements tcpip.Endpoint.SetSockOptInt.
func (e *endpoint) SetSockOptInt(opt tcpip.SockOptInt, v int) *tcpip.Error {
- return nil
-}
+ switch opt {
+ case tcpip.MulticastTTLOption:
+ e.mu.Lock()
+ e.multicastTTL = uint8(v)
+ e.mu.Unlock()
-// SetSockOpt implements tcpip.Endpoint.SetSockOpt.
-func (e *endpoint) SetSockOpt(opt interface{}) *tcpip.Error {
- switch v := opt.(type) {
case tcpip.TTLOption:
e.mu.Lock()
e.ttl = uint8(v)
e.mu.Unlock()
- case tcpip.MulticastTTLOption:
+ case tcpip.IPv4TOSOption:
e.mu.Lock()
- e.multicastTTL = uint8(v)
+ e.sendTOS = uint8(v)
+ e.mu.Unlock()
+
+ case tcpip.IPv6TrafficClassOption:
+ e.mu.Lock()
+ e.sendTOS = uint8(v)
e.mu.Unlock()
+ case tcpip.ReceiveBufferSizeOption:
+ case tcpip.SendBufferSizeOption:
+
+ default:
+ return tcpip.ErrUnknownProtocolOption
+ }
+
+ return nil
+}
+
+// SetSockOpt implements tcpip.Endpoint.SetSockOpt.
+func (e *endpoint) SetSockOpt(opt interface{}) *tcpip.Error {
+ switch v := opt.(type) {
case tcpip.MulticastInterfaceOption:
e.mu.Lock()
defer e.mu.Unlock()
@@ -686,16 +719,6 @@ func (e *endpoint) SetSockOpt(opt interface{}) *tcpip.Error {
e.multicastMemberships[memToRemoveIndex] = e.multicastMemberships[len(e.multicastMemberships)-1]
e.multicastMemberships = e.multicastMemberships[:len(e.multicastMemberships)-1]
- case tcpip.MulticastLoopOption:
- e.mu.Lock()
- e.multicastLoop = bool(v)
- e.mu.Unlock()
-
- case tcpip.ReusePortOption:
- e.mu.Lock()
- e.reusePort = v != 0
- e.mu.Unlock()
-
case tcpip.BindToDeviceOption:
id := tcpip.NICID(v)
if id != 0 && !e.stack.HasNIC(id) {
@@ -704,26 +727,6 @@ func (e *endpoint) SetSockOpt(opt interface{}) *tcpip.Error {
e.mu.Lock()
e.bindToDevice = id
e.mu.Unlock()
- return nil
-
- case tcpip.BroadcastOption:
- e.mu.Lock()
- e.broadcast = v != 0
- e.mu.Unlock()
-
- return nil
-
- case tcpip.IPv4TOSOption:
- e.mu.Lock()
- e.sendTOS = uint8(v)
- e.mu.Unlock()
- return nil
-
- case tcpip.IPv6TrafficClassOption:
- e.mu.Lock()
- e.sendTOS = uint8(v)
- e.mu.Unlock()
- return nil
}
return nil
}
@@ -731,6 +734,21 @@ func (e *endpoint) SetSockOpt(opt interface{}) *tcpip.Error {
// GetSockOptBool implements tcpip.Endpoint.GetSockOptBool.
func (e *endpoint) GetSockOptBool(opt tcpip.SockOptBool) (bool, *tcpip.Error) {
switch opt {
+ case tcpip.BroadcastOption:
+ e.mu.RLock()
+ v := e.broadcast
+ e.mu.RUnlock()
+ return v, nil
+
+ case tcpip.KeepaliveEnabledOption:
+ return false, nil
+
+ case tcpip.MulticastLoopOption:
+ e.mu.RLock()
+ v := e.multicastLoop
+ e.mu.RUnlock()
+ return v, nil
+
case tcpip.ReceiveTOSOption:
e.mu.RLock()
v := e.receiveTOS
@@ -748,6 +766,22 @@ func (e *endpoint) GetSockOptBool(opt tcpip.SockOptBool) (bool, *tcpip.Error) {
e.mu.RUnlock()
return v, nil
+ case tcpip.ReceiveIPPacketInfoOption:
+ e.mu.RLock()
+ v := e.receiveIPPacketInfo
+ e.mu.RUnlock()
+ return v, nil
+
+ case tcpip.ReuseAddressOption:
+ return false, nil
+
+ case tcpip.ReusePortOption:
+ e.mu.RLock()
+ v := e.reusePort
+ e.mu.RUnlock()
+
+ return v, nil
+
case tcpip.V6OnlyOption:
// We only recognize this option on v6 endpoints.
if e.NetProto != header.IPv6ProtocolNumber {
@@ -760,19 +794,32 @@ func (e *endpoint) GetSockOptBool(opt tcpip.SockOptBool) (bool, *tcpip.Error) {
return v, nil
- case tcpip.ReceiveIPPacketInfoOption:
- e.mu.RLock()
- v := e.receiveIPPacketInfo
- e.mu.RUnlock()
- return v, nil
+ default:
+ return false, tcpip.ErrUnknownProtocolOption
}
-
- return false, tcpip.ErrUnknownProtocolOption
}
// GetSockOptInt implements tcpip.Endpoint.GetSockOptInt.
func (e *endpoint) GetSockOptInt(opt tcpip.SockOptInt) (int, *tcpip.Error) {
switch opt {
+ case tcpip.IPv4TOSOption:
+ e.mu.RLock()
+ v := int(e.sendTOS)
+ e.mu.RUnlock()
+ return v, nil
+
+ case tcpip.IPv6TrafficClassOption:
+ e.mu.RLock()
+ v := int(e.sendTOS)
+ e.mu.RUnlock()
+ return v, nil
+
+ case tcpip.MulticastTTLOption:
+ e.mu.Lock()
+ v := int(e.multicastTTL)
+ e.mu.Unlock()
+ return v, nil
+
case tcpip.ReceiveQueueSizeOption:
v := 0
e.rcvMu.Lock()
@@ -794,29 +841,22 @@ func (e *endpoint) GetSockOptInt(opt tcpip.SockOptInt) (int, *tcpip.Error) {
v := e.rcvBufSizeMax
e.rcvMu.Unlock()
return v, nil
- }
- return -1, tcpip.ErrUnknownProtocolOption
+ case tcpip.TTLOption:
+ e.mu.Lock()
+ v := int(e.ttl)
+ e.mu.Unlock()
+ return v, nil
+
+ default:
+ return -1, tcpip.ErrUnknownProtocolOption
+ }
}
// GetSockOpt implements tcpip.Endpoint.GetSockOpt.
func (e *endpoint) GetSockOpt(opt interface{}) *tcpip.Error {
switch o := opt.(type) {
case tcpip.ErrorOption:
- return nil
-
- case *tcpip.TTLOption:
- e.mu.Lock()
- *o = tcpip.TTLOption(e.ttl)
- e.mu.Unlock()
- return nil
-
- case *tcpip.MulticastTTLOption:
- e.mu.Lock()
- *o = tcpip.MulticastTTLOption(e.multicastTTL)
- e.mu.Unlock()
- return nil
-
case *tcpip.MulticastInterfaceOption:
e.mu.Lock()
*o = tcpip.MulticastInterfaceOption{
@@ -824,67 +864,16 @@ func (e *endpoint) GetSockOpt(opt interface{}) *tcpip.Error {
e.multicastAddr,
}
e.mu.Unlock()
- return nil
-
- case *tcpip.MulticastLoopOption:
- e.mu.RLock()
- v := e.multicastLoop
- e.mu.RUnlock()
-
- *o = tcpip.MulticastLoopOption(v)
- return nil
-
- case *tcpip.ReuseAddressOption:
- *o = 0
- return nil
-
- case *tcpip.ReusePortOption:
- e.mu.RLock()
- v := e.reusePort
- e.mu.RUnlock()
-
- *o = 0
- if v {
- *o = 1
- }
- return nil
case *tcpip.BindToDeviceOption:
e.mu.RLock()
*o = tcpip.BindToDeviceOption(e.bindToDevice)
e.mu.RUnlock()
- return nil
-
- case *tcpip.KeepaliveEnabledOption:
- *o = 0
- return nil
-
- case *tcpip.BroadcastOption:
- e.mu.RLock()
- v := e.broadcast
- e.mu.RUnlock()
-
- *o = 0
- if v {
- *o = 1
- }
- return nil
-
- case *tcpip.IPv4TOSOption:
- e.mu.RLock()
- *o = tcpip.IPv4TOSOption(e.sendTOS)
- e.mu.RUnlock()
- return nil
-
- case *tcpip.IPv6TrafficClassOption:
- e.mu.RLock()
- *o = tcpip.IPv6TrafficClassOption(e.sendTOS)
- e.mu.RUnlock()
- return nil
default:
return tcpip.ErrUnknownProtocolOption
}
+ return nil
}
// sendUDP sends a UDP segment via the provided network endpoint and under the
diff --git a/pkg/tcpip/transport/udp/udp_test.go b/pkg/tcpip/transport/udp/udp_test.go
index 0905726c1..8acaa607a 100644
--- a/pkg/tcpip/transport/udp/udp_test.go
+++ b/pkg/tcpip/transport/udp/udp_test.go
@@ -343,11 +343,11 @@ func (c *testContext) createEndpointForFlow(flow testFlow) {
c.createEndpoint(flow.sockProto())
if flow.isV6Only() {
if err := c.ep.SetSockOptBool(tcpip.V6OnlyOption, true); err != nil {
- c.t.Fatalf("SetSockOpt failed: %v", err)
+ c.t.Fatalf("SetSockOptBool failed: %s", err)
}
} else if flow.isBroadcast() {
- if err := c.ep.SetSockOpt(tcpip.BroadcastOption(1)); err != nil {
- c.t.Fatal("SetSockOpt failed:", err)
+ if err := c.ep.SetSockOptBool(tcpip.BroadcastOption, true); err != nil {
+ c.t.Fatalf("SetSockOptBool failed: %s", err)
}
}
}
@@ -358,7 +358,8 @@ func (c *testContext) createEndpointForFlow(flow testFlow) {
func (c *testContext) getPacketAndVerify(flow testFlow, checkers ...checker.NetworkChecker) []byte {
c.t.Helper()
- ctx, _ := context.WithTimeout(context.Background(), 2*time.Second)
+ ctx, cancel := context.WithTimeout(context.Background(), 2*time.Second)
+ defer cancel()
p, ok := c.linkEP.ReadContext(ctx)
if !ok {
c.t.Fatalf("Packet wasn't written out")
@@ -607,7 +608,7 @@ func testReadInternal(c *testContext, flow testFlow, packetShouldBeDropped, expe
// Check the peer address.
h := flow.header4Tuple(incoming)
if addr.Addr != h.srcAddr.Addr {
- c.t.Fatalf("unexpected remote address: got %s, want %s", addr.Addr, h.srcAddr)
+ c.t.Fatalf("unexpected remote address: got %s, want %v", addr.Addr, h.srcAddr)
}
// Check the payload.
@@ -1271,8 +1272,8 @@ func TestTTL(t *testing.T) {
c.createEndpointForFlow(flow)
const multicastTTL = 42
- if err := c.ep.SetSockOpt(tcpip.MulticastTTLOption(multicastTTL)); err != nil {
- c.t.Fatalf("SetSockOpt failed: %v", err)
+ if err := c.ep.SetSockOptInt(tcpip.MulticastTTLOption, multicastTTL); err != nil {
+ c.t.Fatalf("SetSockOptInt failed: %s", err)
}
var wantTTL uint8
@@ -1311,8 +1312,8 @@ func TestSetTTL(t *testing.T) {
c.createEndpointForFlow(flow)
- if err := c.ep.SetSockOpt(tcpip.TTLOption(wantTTL)); err != nil {
- c.t.Fatalf("SetSockOpt failed: %v", err)
+ if err := c.ep.SetSockOptInt(tcpip.TTLOption, int(wantTTL)); err != nil {
+ c.t.Fatalf("SetSockOptInt(TTLOption, %d) failed: %s", wantTTL, err)
}
var p stack.NetworkProtocol
@@ -1346,25 +1347,26 @@ func TestSetTOS(t *testing.T) {
c.createEndpointForFlow(flow)
const tos = testTOS
- var v tcpip.IPv4TOSOption
- if err := c.ep.GetSockOpt(&v); err != nil {
- c.t.Errorf("GetSockopt(%T) failed: %s", v, err)
+ v, err := c.ep.GetSockOptInt(tcpip.IPv4TOSOption)
+ if err != nil {
+ c.t.Errorf("GetSockOptInt(IPv4TOSOption) failed: %s", err)
}
// Test for expected default value.
if v != 0 {
- c.t.Errorf("got GetSockOpt(%T) = 0x%x, want = 0x%x", v, v, 0)
+ c.t.Errorf("got GetSockOpt(IPv4TOSOption) = 0x%x, want = 0x%x", v, 0)
}
- if err := c.ep.SetSockOpt(tcpip.IPv4TOSOption(tos)); err != nil {
- c.t.Errorf("SetSockOpt(%T, 0x%x) failed: %s", v, tcpip.IPv4TOSOption(tos), err)
+ if err := c.ep.SetSockOptInt(tcpip.IPv4TOSOption, tos); err != nil {
+ c.t.Errorf("SetSockOptInt(IPv4TOSOption, 0x%x) failed: %s", tos, err)
}
- if err := c.ep.GetSockOpt(&v); err != nil {
- c.t.Errorf("GetSockopt(%T) failed: %s", v, err)
+ v, err = c.ep.GetSockOptInt(tcpip.IPv4TOSOption)
+ if err != nil {
+ c.t.Errorf("GetSockOptInt(IPv4TOSOption) failed: %s", err)
}
- if want := tcpip.IPv4TOSOption(tos); v != want {
- c.t.Errorf("got GetSockOpt(%T) = 0x%x, want = 0x%x", v, v, want)
+ if v != tos {
+ c.t.Errorf("got GetSockOptInt(IPv4TOSOption) = 0x%x, want = 0x%x", v, tos)
}
testWrite(c, flow, checker.TOS(tos, 0))
@@ -1381,25 +1383,26 @@ func TestSetTClass(t *testing.T) {
c.createEndpointForFlow(flow)
const tClass = testTOS
- var v tcpip.IPv6TrafficClassOption
- if err := c.ep.GetSockOpt(&v); err != nil {
- c.t.Errorf("GetSockopt(%T) failed: %s", v, err)
+ v, err := c.ep.GetSockOptInt(tcpip.IPv6TrafficClassOption)
+ if err != nil {
+ c.t.Errorf("GetSockOptInt(IPv6TrafficClassOption) failed: %s", err)
}
// Test for expected default value.
if v != 0 {
- c.t.Errorf("got GetSockOpt(%T) = 0x%x, want = 0x%x", v, v, 0)
+ c.t.Errorf("got GetSockOptInt(IPv6TrafficClassOption) = 0x%x, want = 0x%x", v, 0)
}
- if err := c.ep.SetSockOpt(tcpip.IPv6TrafficClassOption(tClass)); err != nil {
- c.t.Errorf("SetSockOpt(%T, 0x%x) failed: %s", v, tcpip.IPv6TrafficClassOption(tClass), err)
+ if err := c.ep.SetSockOptInt(tcpip.IPv6TrafficClassOption, tClass); err != nil {
+ c.t.Errorf("SetSockOptInt(IPv6TrafficClassOption, 0x%x) failed: %s", tClass, err)
}
- if err := c.ep.GetSockOpt(&v); err != nil {
- c.t.Errorf("GetSockopt(%T) failed: %s", v, err)
+ v, err = c.ep.GetSockOptInt(tcpip.IPv6TrafficClassOption)
+ if err != nil {
+ c.t.Errorf("GetSockOptInt(IPv6TrafficClassOption) failed: %s", err)
}
- if want := tcpip.IPv6TrafficClassOption(tClass); v != want {
- c.t.Errorf("got GetSockOpt(%T) = 0x%x, want = 0x%x", v, v, want)
+ if v != tClass {
+ c.t.Errorf("got GetSockOptInt(IPv6TrafficClassOption) = 0x%x, want = 0x%x", v, tClass)
}
// The header getter for TClass is called TOS, so use that checker.
@@ -1430,7 +1433,7 @@ func TestReceiveTosTClass(t *testing.T) {
// Verify that setting and reading the option works.
v, err := c.ep.GetSockOptBool(option)
if err != nil {
- c.t.Errorf("GetSockoptBool(%s) failed: %s", name, err)
+ c.t.Errorf("GetSockOptBool(%s) failed: %s", name, err)
}
// Test for expected default value.
if v != false {
@@ -1444,7 +1447,7 @@ func TestReceiveTosTClass(t *testing.T) {
got, err := c.ep.GetSockOptBool(option)
if err != nil {
- c.t.Errorf("GetSockoptBool(%s) failed: %s", name, err)
+ c.t.Errorf("GetSockOptBool(%s) failed: %s", name, err)
}
if got != want {
@@ -1563,7 +1566,8 @@ func TestV4UnknownDestination(t *testing.T) {
}
c.injectPacket(tc.flow, payload)
if !tc.icmpRequired {
- ctx, _ := context.WithTimeout(context.Background(), time.Second)
+ ctx, cancel := context.WithTimeout(context.Background(), time.Second)
+ defer cancel()
if p, ok := c.linkEP.ReadContext(ctx); ok {
t.Fatalf("unexpected packet received: %+v", p)
}
@@ -1571,7 +1575,8 @@ func TestV4UnknownDestination(t *testing.T) {
}
// ICMP required.
- ctx, _ := context.WithTimeout(context.Background(), time.Second)
+ ctx, cancel := context.WithTimeout(context.Background(), time.Second)
+ defer cancel()
p, ok := c.linkEP.ReadContext(ctx)
if !ok {
t.Fatalf("packet wasn't written out")
@@ -1639,7 +1644,8 @@ func TestV6UnknownDestination(t *testing.T) {
}
c.injectPacket(tc.flow, payload)
if !tc.icmpRequired {
- ctx, _ := context.WithTimeout(context.Background(), time.Second)
+ ctx, cancel := context.WithTimeout(context.Background(), time.Second)
+ defer cancel()
if p, ok := c.linkEP.ReadContext(ctx); ok {
t.Fatalf("unexpected packet received: %+v", p)
}
@@ -1647,7 +1653,8 @@ func TestV6UnknownDestination(t *testing.T) {
}
// ICMP required.
- ctx, _ := context.WithTimeout(context.Background(), time.Second)
+ ctx, cancel := context.WithTimeout(context.Background(), time.Second)
+ defer cancel()
p, ok := c.linkEP.ReadContext(ctx)
if !ok {
t.Fatalf("packet wasn't written out")
diff --git a/pkg/usermem/usermem.go b/pkg/usermem/usermem.go
index d2f4403b0..cd6a0ea6b 100644
--- a/pkg/usermem/usermem.go
+++ b/pkg/usermem/usermem.go
@@ -29,9 +29,6 @@ import (
)
// IO provides access to the contents of a virtual memory space.
-//
-// FIXME(b/38173783): Implementations of IO cannot expect ctx to contain any
-// meaningful data.
type IO interface {
// CopyOut copies len(src) bytes from src to the memory mapped at addr. It
// returns the number of bytes copied. If the number of bytes copied is <