diff options
Diffstat (limited to 'runsc/boot')
| -rw-r--r-- | runsc/boot/BUILD | 3 | ||||
| -rw-r--r-- | runsc/boot/config.go | 336 | ||||
| -rw-r--r-- | runsc/boot/controller.go | 3 | ||||
| -rw-r--r-- | runsc/boot/fs.go | 41 | ||||
| -rw-r--r-- | runsc/boot/fs_test.go | 11 | ||||
| -rw-r--r-- | runsc/boot/loader.go | 19 | ||||
| -rw-r--r-- | runsc/boot/loader_test.go | 7 | ||||
| -rw-r--r-- | runsc/boot/network.go | 45 | ||||
| -rw-r--r-- | runsc/boot/strace.go | 3 | ||||
| -rw-r--r-- | runsc/boot/vfs.go | 21 |
10 files changed, 62 insertions, 427 deletions
diff --git a/runsc/boot/BUILD b/runsc/boot/BUILD index 9f52438c2..040f6a72d 100644 --- a/runsc/boot/BUILD +++ b/runsc/boot/BUILD @@ -8,7 +8,6 @@ go_library( "compat.go", "compat_amd64.go", "compat_arm64.go", - "config.go", "controller.go", "debug.go", "events.go", @@ -105,6 +104,7 @@ go_library( "//runsc/boot/filter", "//runsc/boot/platforms", "//runsc/boot/pprof", + "//runsc/config", "//runsc/specutils", "@com_github_golang_protobuf//proto:go_default_library", "@com_github_opencontainers_runtime_spec//specs-go:go_default_library", @@ -131,6 +131,7 @@ go_test( "//pkg/sentry/vfs", "//pkg/sync", "//pkg/unet", + "//runsc/config", "//runsc/fsgofer", "@com_github_opencontainers_runtime_spec//specs-go:go_default_library", "@org_golang_x_sys//unix:go_default_library", diff --git a/runsc/boot/config.go b/runsc/boot/config.go deleted file mode 100644 index 80da8b3e6..000000000 --- a/runsc/boot/config.go +++ /dev/null @@ -1,336 +0,0 @@ -// 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 boot - -import ( - "fmt" - "strconv" - "strings" - - "gvisor.dev/gvisor/pkg/refs" - "gvisor.dev/gvisor/pkg/sentry/watchdog" -) - -// FileAccessType tells how the filesystem is accessed. -type FileAccessType int - -const ( - // FileAccessShared sends IO requests to a Gofer process that validates the - // requests and forwards them to the host. - FileAccessShared FileAccessType = iota - - // FileAccessExclusive is the same as FileAccessShared, but enables - // extra caching for improved performance. It should only be used if - // the sandbox has exclusive access to the filesystem. - FileAccessExclusive -) - -// MakeFileAccessType converts type from string. -func MakeFileAccessType(s string) (FileAccessType, error) { - switch s { - case "shared": - return FileAccessShared, nil - case "exclusive": - return FileAccessExclusive, nil - default: - return 0, fmt.Errorf("invalid file access type %q", s) - } -} - -func (f FileAccessType) String() string { - switch f { - case FileAccessShared: - return "shared" - case FileAccessExclusive: - return "exclusive" - default: - return fmt.Sprintf("unknown(%d)", f) - } -} - -// NetworkType tells which network stack to use. -type NetworkType int - -const ( - // NetworkSandbox uses internal network stack, isolated from the host. - NetworkSandbox NetworkType = iota - - // NetworkHost redirects network related syscalls to the host network. - NetworkHost - - // NetworkNone sets up just loopback using netstack. - NetworkNone -) - -// MakeNetworkType converts type from string. -func MakeNetworkType(s string) (NetworkType, error) { - switch s { - case "sandbox": - return NetworkSandbox, nil - case "host": - return NetworkHost, nil - case "none": - return NetworkNone, nil - default: - return 0, fmt.Errorf("invalid network type %q", s) - } -} - -func (n NetworkType) String() string { - switch n { - case NetworkSandbox: - return "sandbox" - case NetworkHost: - return "host" - case NetworkNone: - return "none" - default: - return fmt.Sprintf("unknown(%d)", n) - } -} - -// MakeWatchdogAction converts type from string. -func MakeWatchdogAction(s string) (watchdog.Action, error) { - switch strings.ToLower(s) { - case "log", "logwarning": - return watchdog.LogWarning, nil - case "panic": - return watchdog.Panic, nil - default: - return 0, fmt.Errorf("invalid watchdog action %q", s) - } -} - -// MakeRefsLeakMode converts type from string. -func MakeRefsLeakMode(s string) (refs.LeakMode, error) { - switch strings.ToLower(s) { - case "disabled": - return refs.NoLeakChecking, nil - case "log-names": - return refs.LeaksLogWarning, nil - case "log-traces": - return refs.LeaksLogTraces, nil - default: - return 0, fmt.Errorf("invalid refs leakmode %q", s) - } -} - -func refsLeakModeToString(mode refs.LeakMode) string { - switch mode { - // If not set, default it to disabled. - case refs.UninitializedLeakChecking, refs.NoLeakChecking: - return "disabled" - case refs.LeaksLogWarning: - return "log-names" - case refs.LeaksLogTraces: - return "log-traces" - default: - panic(fmt.Sprintf("Invalid leakmode: %d", mode)) - } -} - -// Config holds configuration that is not part of the runtime spec. -type Config struct { - // RootDir is the runtime root directory. - RootDir string - - // Debug indicates that debug logging should be enabled. - Debug bool - - // LogFilename is the filename to log to, if not empty. - LogFilename string - - // LogFormat is the log format. - LogFormat string - - // DebugLog is the path to log debug information to, if not empty. - DebugLog string - - // PanicLog is the path to log GO's runtime messages, if not empty. - PanicLog string - - // DebugLogFormat is the log format for debug. - DebugLogFormat string - - // FileAccess indicates how the filesystem is accessed. - FileAccess FileAccessType - - // Overlay is whether to wrap the root filesystem in an overlay. - Overlay bool - - // FSGoferHostUDS enables the gofer to mount a host UDS. - FSGoferHostUDS bool - - // Network indicates what type of network to use. - Network NetworkType - - // EnableRaw indicates whether raw sockets should be enabled. Raw - // sockets are disabled by stripping CAP_NET_RAW from the list of - // capabilities. - EnableRaw bool - - // HardwareGSO indicates that hardware segmentation offload is enabled. - HardwareGSO bool - - // SoftwareGSO indicates that software segmentation offload is enabled. - SoftwareGSO bool - - // TXChecksumOffload indicates that TX Checksum Offload is enabled. - TXChecksumOffload bool - - // RXChecksumOffload indicates that RX Checksum Offload is enabled. - RXChecksumOffload bool - - // QDisc indicates the type of queuening discipline to use by default - // for non-loopback interfaces. - QDisc QueueingDiscipline - - // LogPackets indicates that all network packets should be logged. - LogPackets bool - - // Platform is the platform to run on. - Platform string - - // Strace indicates that strace should be enabled. - Strace bool - - // StraceSyscalls is the set of syscalls to trace. If StraceEnable is - // true and this list is empty, then all syscalls will be traced. - StraceSyscalls []string - - // StraceLogSize is the max size of data blobs to display. - StraceLogSize uint - - // DisableSeccomp indicates whether seccomp syscall filters should be - // disabled. Pardon the double negation, but default to enabled is important. - DisableSeccomp bool - - // WatchdogAction sets what action the watchdog takes when triggered. - WatchdogAction watchdog.Action - - // PanicSignal registers signal handling that panics. Usually set to - // SIGUSR2(12) to troubleshoot hangs. -1 disables it. - PanicSignal int - - // ProfileEnable is set to prepare the sandbox to be profiled. - ProfileEnable bool - - // RestoreFile is the path to the saved container image - RestoreFile string - - // NumNetworkChannels controls the number of AF_PACKET sockets that map - // to the same underlying network device. This allows netstack to better - // scale for high throughput use cases. - NumNetworkChannels int - - // Rootless allows the sandbox to be started with a user that is not root. - // Defense is depth measures are weaker with rootless. Specifically, the - // sandbox and Gofer process run as root inside a user namespace with root - // mapped to the caller's user. - Rootless bool - - // AlsoLogToStderr allows to send log messages to stderr. - AlsoLogToStderr bool - - // ReferenceLeakMode sets reference leak check mode - ReferenceLeakMode refs.LeakMode - - // OverlayfsStaleRead instructs the sandbox to assume that the root mount - // is on a Linux overlayfs mount, which does not necessarily preserve - // coherence between read-only and subsequent writable file descriptors - // representing the "same" file. - OverlayfsStaleRead bool - - // TestOnlyAllowRunAsCurrentUserWithoutChroot should only be used in - // tests. It allows runsc to start the sandbox process as the current - // user, and without chrooting the sandbox process. This can be - // necessary in test environments that have limited capabilities. - TestOnlyAllowRunAsCurrentUserWithoutChroot bool - - // TestOnlyTestNameEnv should only be used in tests. It looks up for the - // test name in the container environment variables and adds it to the debug - // log file name. This is done to help identify the log with the test when - // multiple tests are run in parallel, since there is no way to pass - // parameters to the runtime from docker. - TestOnlyTestNameEnv string - - // CPUNumFromQuota sets CPU number count to available CPU quota, using - // least integer value greater than or equal to quota. - // - // E.g. 0.2 CPU quota will result in 1, and 1.9 in 2. - CPUNumFromQuota bool - - // Enables VFS2 (not plumbled through yet). - VFS2 bool - - // Enables FUSE usage (not plumbled through yet). - FUSE bool -} - -// ToFlags returns a slice of flags that correspond to the given Config. -func (c *Config) ToFlags() []string { - f := []string{ - "--root=" + c.RootDir, - "--debug=" + strconv.FormatBool(c.Debug), - "--log=" + c.LogFilename, - "--log-format=" + c.LogFormat, - "--debug-log=" + c.DebugLog, - "--panic-log=" + c.PanicLog, - "--debug-log-format=" + c.DebugLogFormat, - "--file-access=" + c.FileAccess.String(), - "--overlay=" + strconv.FormatBool(c.Overlay), - "--fsgofer-host-uds=" + strconv.FormatBool(c.FSGoferHostUDS), - "--network=" + c.Network.String(), - "--log-packets=" + strconv.FormatBool(c.LogPackets), - "--platform=" + c.Platform, - "--strace=" + strconv.FormatBool(c.Strace), - "--strace-syscalls=" + strings.Join(c.StraceSyscalls, ","), - "--strace-log-size=" + strconv.Itoa(int(c.StraceLogSize)), - "--watchdog-action=" + c.WatchdogAction.String(), - "--panic-signal=" + strconv.Itoa(c.PanicSignal), - "--profile=" + strconv.FormatBool(c.ProfileEnable), - "--net-raw=" + strconv.FormatBool(c.EnableRaw), - "--num-network-channels=" + strconv.Itoa(c.NumNetworkChannels), - "--rootless=" + strconv.FormatBool(c.Rootless), - "--alsologtostderr=" + strconv.FormatBool(c.AlsoLogToStderr), - "--ref-leak-mode=" + refsLeakModeToString(c.ReferenceLeakMode), - "--gso=" + strconv.FormatBool(c.HardwareGSO), - "--software-gso=" + strconv.FormatBool(c.SoftwareGSO), - "--rx-checksum-offload=" + strconv.FormatBool(c.RXChecksumOffload), - "--tx-checksum-offload=" + strconv.FormatBool(c.TXChecksumOffload), - "--overlayfs-stale-read=" + strconv.FormatBool(c.OverlayfsStaleRead), - "--qdisc=" + c.QDisc.String(), - } - if c.CPUNumFromQuota { - f = append(f, "--cpu-num-from-quota") - } - // Only include these if set since it is never to be used by users. - if c.TestOnlyAllowRunAsCurrentUserWithoutChroot { - f = append(f, "--TESTONLY-unsafe-nonroot=true") - } - if len(c.TestOnlyTestNameEnv) != 0 { - f = append(f, "--TESTONLY-test-name-env="+c.TestOnlyTestNameEnv) - } - - if c.VFS2 { - f = append(f, "--vfs2=true") - } - - if c.FUSE { - f = append(f, "--fuse=true") - } - - return f -} diff --git a/runsc/boot/controller.go b/runsc/boot/controller.go index 626a3816e..68a2b45cf 100644 --- a/runsc/boot/controller.go +++ b/runsc/boot/controller.go @@ -33,6 +33,7 @@ import ( "gvisor.dev/gvisor/pkg/tcpip/stack" "gvisor.dev/gvisor/pkg/urpc" "gvisor.dev/gvisor/runsc/boot/pprof" + "gvisor.dev/gvisor/runsc/config" "gvisor.dev/gvisor/runsc/specutils" ) @@ -220,7 +221,7 @@ type StartArgs struct { Spec *specs.Spec // Config is the runsc-specific configuration for the sandbox. - Conf *Config + Conf *config.Config // CID is the ID of the container to start. CID string diff --git a/runsc/boot/fs.go b/runsc/boot/fs.go index 9dd5b0184..163265afe 100644 --- a/runsc/boot/fs.go +++ b/runsc/boot/fs.go @@ -48,6 +48,7 @@ import ( "gvisor.dev/gvisor/pkg/sentry/kernel" "gvisor.dev/gvisor/pkg/sentry/kernel/auth" "gvisor.dev/gvisor/pkg/syserror" + "gvisor.dev/gvisor/runsc/config" "gvisor.dev/gvisor/runsc/specutils" ) @@ -66,7 +67,7 @@ const ( // tmpfs has some extra supported options that we must pass through. var tmpfsAllowedData = []string{"mode", "uid", "gid"} -func addOverlay(ctx context.Context, conf *Config, lower *fs.Inode, name string, lowerFlags fs.MountSourceFlags) (*fs.Inode, error) { +func addOverlay(ctx context.Context, conf *config.Config, lower *fs.Inode, name string, lowerFlags fs.MountSourceFlags) (*fs.Inode, error) { // Upper layer uses the same flags as lower, but it must be read-write. upperFlags := lowerFlags upperFlags.ReadOnly = false @@ -156,7 +157,7 @@ func compileMounts(spec *specs.Spec) []specs.Mount { } // p9MountData creates a slice of p9 mount data. -func p9MountData(fd int, fa FileAccessType, vfs2 bool) []string { +func p9MountData(fd int, fa config.FileAccessType, vfs2 bool) []string { opts := []string{ "trans=fd", "rfdno=" + strconv.Itoa(fd), @@ -167,7 +168,7 @@ func p9MountData(fd int, fa FileAccessType, vfs2 bool) []string { // enablement. opts = append(opts, "privateunixsocket=true") } - if fa == FileAccessShared { + if fa == config.FileAccessShared { opts = append(opts, "cache=remote_revalidating") } return opts @@ -281,7 +282,7 @@ func subtargets(root string, mnts []specs.Mount) []string { return targets } -func setupContainerFS(ctx context.Context, conf *Config, mntr *containerMounter, procArgs *kernel.CreateProcessArgs) error { +func setupContainerFS(ctx context.Context, conf *config.Config, mntr *containerMounter, procArgs *kernel.CreateProcessArgs) error { if conf.VFS2 { return setupContainerVFS2(ctx, conf, mntr, procArgs) } @@ -468,11 +469,11 @@ func (m *mountHint) checkCompatible(mount specs.Mount) error { return nil } -func (m *mountHint) fileAccessType() FileAccessType { +func (m *mountHint) fileAccessType() config.FileAccessType { if m.share == container { - return FileAccessExclusive + return config.FileAccessExclusive } - return FileAccessShared + return config.FileAccessShared } func filterUnsupportedOptions(mount specs.Mount) []string { @@ -576,7 +577,7 @@ func newContainerMounter(spec *specs.Spec, goferFDs []int, k *kernel.Kernel, hin // processHints processes annotations that container hints about how volumes // should be mounted (e.g. a volume shared between containers). It must be // called for the root container only. -func (c *containerMounter) processHints(conf *Config, creds *auth.Credentials) error { +func (c *containerMounter) processHints(conf *config.Config, creds *auth.Credentials) error { if conf.VFS2 { return c.processHintsVFS2(conf, creds) } @@ -600,7 +601,7 @@ func (c *containerMounter) processHints(conf *Config, creds *auth.Credentials) e // setupFS is used to set up the file system for all containers. This is the // main entry point method, with most of the other being internal only. It // returns the mount namespace that is created for the container. -func (c *containerMounter) setupFS(conf *Config, procArgs *kernel.CreateProcessArgs) (*fs.MountNamespace, error) { +func (c *containerMounter) setupFS(conf *config.Config, procArgs *kernel.CreateProcessArgs) (*fs.MountNamespace, error) { log.Infof("Configuring container's file system") // Create context with root credentials to mount the filesystem (the current @@ -626,7 +627,7 @@ func (c *containerMounter) setupFS(conf *Config, procArgs *kernel.CreateProcessA return mns, nil } -func (c *containerMounter) createMountNamespace(ctx context.Context, conf *Config) (*fs.MountNamespace, error) { +func (c *containerMounter) createMountNamespace(ctx context.Context, conf *config.Config) (*fs.MountNamespace, error) { rootInode, err := c.createRootMount(ctx, conf) if err != nil { return nil, fmt.Errorf("creating filesystem for container: %v", err) @@ -638,7 +639,7 @@ func (c *containerMounter) createMountNamespace(ctx context.Context, conf *Confi return mns, nil } -func (c *containerMounter) mountSubmounts(ctx context.Context, conf *Config, mns *fs.MountNamespace) error { +func (c *containerMounter) mountSubmounts(ctx context.Context, conf *config.Config, mns *fs.MountNamespace) error { root := mns.Root() defer root.DecRef(ctx) @@ -674,7 +675,7 @@ func (c *containerMounter) checkDispenser() error { // mountSharedMaster mounts the master of a volume that is shared among // containers in a pod. It returns the root mount's inode. -func (c *containerMounter) mountSharedMaster(ctx context.Context, conf *Config, hint *mountHint) (*fs.Inode, error) { +func (c *containerMounter) mountSharedMaster(ctx context.Context, conf *config.Config, hint *mountHint) (*fs.Inode, error) { // Map mount type to filesystem name, and parse out the options that we are // capable of dealing with. fsName, opts, useOverlay, err := c.getMountNameAndOptions(conf, hint.mount) @@ -714,7 +715,7 @@ func (c *containerMounter) mountSharedMaster(ctx context.Context, conf *Config, } // createRootMount creates the root filesystem. -func (c *containerMounter) createRootMount(ctx context.Context, conf *Config) (*fs.Inode, error) { +func (c *containerMounter) createRootMount(ctx context.Context, conf *config.Config) (*fs.Inode, error) { // First construct the filesystem from the spec.Root. mf := fs.MountSourceFlags{ReadOnly: c.root.Readonly || conf.Overlay} @@ -759,7 +760,7 @@ func (c *containerMounter) createRootMount(ctx context.Context, conf *Config) (* // getMountNameAndOptions retrieves the fsName, opts, and useOverlay values // used for mounts. -func (c *containerMounter) getMountNameAndOptions(conf *Config, m specs.Mount) (string, []string, bool, error) { +func (c *containerMounter) getMountNameAndOptions(conf *config.Config, m specs.Mount) (string, []string, bool, error) { var ( fsName string opts []string @@ -793,19 +794,19 @@ func (c *containerMounter) getMountNameAndOptions(conf *Config, m specs.Mount) ( return fsName, opts, useOverlay, nil } -func (c *containerMounter) getMountAccessType(mount specs.Mount) FileAccessType { +func (c *containerMounter) getMountAccessType(mount specs.Mount) config.FileAccessType { if hint := c.hints.findMount(mount); hint != nil { return hint.fileAccessType() } // Non-root bind mounts are always shared if no hints were provided. - return FileAccessShared + return config.FileAccessShared } // mountSubmount mounts volumes inside the container's root. Because mounts may // be readonly, a lower ramfs overlay is added to create the mount point dir. // Another overlay is added with tmpfs on top if Config.Overlay is true. // 'm.Destination' must be an absolute path with '..' and symlinks resolved. -func (c *containerMounter) mountSubmount(ctx context.Context, conf *Config, mns *fs.MountNamespace, root *fs.Dirent, m specs.Mount) error { +func (c *containerMounter) mountSubmount(ctx context.Context, conf *config.Config, mns *fs.MountNamespace, root *fs.Dirent, m specs.Mount) error { // Map mount type to filesystem name, and parse out the options that we are // capable of dealing with. fsName, opts, useOverlay, err := c.getMountNameAndOptions(conf, m) @@ -904,7 +905,7 @@ func (c *containerMounter) mountSharedSubmount(ctx context.Context, mns *fs.Moun // addRestoreMount adds a mount to the MountSources map used for restoring a // checkpointed container. -func (c *containerMounter) addRestoreMount(conf *Config, renv *fs.RestoreEnvironment, m specs.Mount) error { +func (c *containerMounter) addRestoreMount(conf *config.Config, renv *fs.RestoreEnvironment, m specs.Mount) error { fsName, opts, useOverlay, err := c.getMountNameAndOptions(conf, m) if err != nil { return err @@ -929,7 +930,7 @@ func (c *containerMounter) addRestoreMount(conf *Config, renv *fs.RestoreEnviron // createRestoreEnvironment builds a fs.RestoreEnvironment called renv by adding // the mounts to the environment. -func (c *containerMounter) createRestoreEnvironment(conf *Config) (*fs.RestoreEnvironment, error) { +func (c *containerMounter) createRestoreEnvironment(conf *config.Config) (*fs.RestoreEnvironment, error) { renv := &fs.RestoreEnvironment{ MountSources: make(map[string][]fs.MountArgs), } @@ -984,7 +985,7 @@ func (c *containerMounter) createRestoreEnvironment(conf *Config) (*fs.RestoreEn // // Note that when there are submounts inside of '/tmp', directories for the // mount points must be present, making '/tmp' not empty anymore. -func (c *containerMounter) mountTmp(ctx context.Context, conf *Config, mns *fs.MountNamespace, root *fs.Dirent) error { +func (c *containerMounter) mountTmp(ctx context.Context, conf *config.Config, mns *fs.MountNamespace, root *fs.Dirent) error { for _, m := range c.mounts { if filepath.Clean(m.Destination) == "/tmp" { log.Debugf("Explict %q mount found, skipping internal tmpfs, mount: %+v", "/tmp", m) diff --git a/runsc/boot/fs_test.go b/runsc/boot/fs_test.go index 912037075..e986231e5 100644 --- a/runsc/boot/fs_test.go +++ b/runsc/boot/fs_test.go @@ -20,6 +20,7 @@ import ( "testing" specs "github.com/opencontainers/runtime-spec/specs-go" + "gvisor.dev/gvisor/runsc/config" ) func TestPodMountHintsHappy(t *testing.T) { @@ -196,7 +197,7 @@ func TestGetMountAccessType(t *testing.T) { for _, tst := range []struct { name string annotations map[string]string - want FileAccessType + want config.FileAccessType }{ { name: "container=exclusive", @@ -205,7 +206,7 @@ func TestGetMountAccessType(t *testing.T) { MountPrefix + "mount1.type": "bind", MountPrefix + "mount1.share": "container", }, - want: FileAccessExclusive, + want: config.FileAccessExclusive, }, { name: "pod=shared", @@ -214,7 +215,7 @@ func TestGetMountAccessType(t *testing.T) { MountPrefix + "mount1.type": "bind", MountPrefix + "mount1.share": "pod", }, - want: FileAccessShared, + want: config.FileAccessShared, }, { name: "shared=shared", @@ -223,7 +224,7 @@ func TestGetMountAccessType(t *testing.T) { MountPrefix + "mount1.type": "bind", MountPrefix + "mount1.share": "shared", }, - want: FileAccessShared, + want: config.FileAccessShared, }, { name: "default=shared", @@ -232,7 +233,7 @@ func TestGetMountAccessType(t *testing.T) { MountPrefix + "mount1.type": "bind", MountPrefix + "mount1.share": "container", }, - want: FileAccessShared, + want: config.FileAccessShared, }, } { t.Run(tst.name, func(t *testing.T) { diff --git a/runsc/boot/loader.go b/runsc/boot/loader.go index 40c6f99fd..e8ea5093b 100644 --- a/runsc/boot/loader.go +++ b/runsc/boot/loader.go @@ -67,6 +67,7 @@ import ( "gvisor.dev/gvisor/runsc/boot/filter" _ "gvisor.dev/gvisor/runsc/boot/platforms" // register all platforms. "gvisor.dev/gvisor/runsc/boot/pprof" + "gvisor.dev/gvisor/runsc/config" "gvisor.dev/gvisor/runsc/specutils" // Include supported socket providers. @@ -79,7 +80,7 @@ import ( ) type containerInfo struct { - conf *Config + conf *config.Config // spec is the base configuration for the root container. spec *specs.Spec @@ -165,7 +166,7 @@ type Args struct { // Spec is the sandbox specification. Spec *specs.Spec // Conf is the system configuration. - Conf *Config + Conf *config.Config // ControllerFD is the FD to the URPC controller. The Loader takes ownership // of this FD and may close it at any time. ControllerFD int @@ -471,7 +472,7 @@ func (l *Loader) Destroy() { } } -func createPlatform(conf *Config, deviceFile *os.File) (platform.Platform, error) { +func createPlatform(conf *config.Config, deviceFile *os.File) (platform.Platform, error) { p, err := platform.Lookup(conf.Platform) if err != nil { panic(fmt.Sprintf("invalid platform %v: %v", conf.Platform, err)) @@ -504,7 +505,7 @@ func (l *Loader) installSeccompFilters() error { } else { opts := filter.Options{ Platform: l.k.Platform, - HostNetwork: l.root.conf.Network == NetworkHost, + HostNetwork: l.root.conf.Network == config.NetworkHost, ProfileEnable: l.root.conf.ProfileEnable, ControllerFD: l.ctrl.srv.FD(), } @@ -531,7 +532,7 @@ func (l *Loader) Run() error { } func (l *Loader) run() error { - if l.root.conf.Network == NetworkHost { + if l.root.conf.Network == config.NetworkHost { // Delay host network configuration to this point because network namespace // is configured after the loader is created and before Run() is called. log.Debugf("Configuring host network") @@ -629,7 +630,7 @@ func (l *Loader) createContainer(cid string) error { // startContainer starts a child container. It returns the thread group ID of // the newly created process. Caller owns 'files' and may close them after // this method returns. -func (l *Loader) startContainer(spec *specs.Spec, conf *Config, cid string, files []*os.File) error { +func (l *Loader) startContainer(spec *specs.Spec, conf *config.Config, cid string, files []*os.File) error { // Create capabilities. caps, err := specutils.Capabilities(conf.EnableRaw, spec.Process.Capabilities) if err != nil { @@ -1017,17 +1018,17 @@ func (l *Loader) WaitExit() kernel.ExitStatus { return l.k.GlobalInit().ExitStatus() } -func newRootNetworkNamespace(conf *Config, clock tcpip.Clock, uniqueID stack.UniqueID) (*inet.Namespace, error) { +func newRootNetworkNamespace(conf *config.Config, clock tcpip.Clock, uniqueID stack.UniqueID) (*inet.Namespace, error) { // Create an empty network stack because the network namespace may be empty at // this point. Netns is configured before Run() is called. Netstack is // configured using a control uRPC message. Host network is configured inside // Run(). switch conf.Network { - case NetworkHost: + case config.NetworkHost: // No network namespacing support for hostinet yet, hence creator is nil. return inet.NewRootNamespace(hostinet.NewStack(), nil), nil - case NetworkNone, NetworkSandbox: + case config.NetworkNone, config.NetworkSandbox: s, err := newEmptySandboxNetworkStack(clock, uniqueID) if err != nil { return nil, err diff --git a/runsc/boot/loader_test.go b/runsc/boot/loader_test.go index aa3fdf96c..03cbaec33 100644 --- a/runsc/boot/loader_test.go +++ b/runsc/boot/loader_test.go @@ -34,6 +34,7 @@ import ( "gvisor.dev/gvisor/pkg/sentry/vfs" "gvisor.dev/gvisor/pkg/sync" "gvisor.dev/gvisor/pkg/unet" + "gvisor.dev/gvisor/runsc/config" "gvisor.dev/gvisor/runsc/fsgofer" ) @@ -45,10 +46,10 @@ func init() { } } -func testConfig() *Config { - return &Config{ +func testConfig() *config.Config { + return &config.Config{ RootDir: "unused_root_dir", - Network: NetworkNone, + Network: config.NetworkNone, DisableSeccomp: true, Platform: "ptrace", } diff --git a/runsc/boot/network.go b/runsc/boot/network.go index 4e1fa7665..988573640 100644 --- a/runsc/boot/network.go +++ b/runsc/boot/network.go @@ -33,6 +33,7 @@ import ( "gvisor.dev/gvisor/pkg/tcpip/network/ipv6" "gvisor.dev/gvisor/pkg/tcpip/stack" "gvisor.dev/gvisor/pkg/urpc" + "gvisor.dev/gvisor/runsc/config" ) var ( @@ -78,44 +79,6 @@ type DefaultRoute struct { Name string } -// QueueingDiscipline is used to specify the kind of Queueing Discipline to -// apply for a give FDBasedLink. -type QueueingDiscipline int - -const ( - // QDiscNone disables any queueing for the underlying FD. - QDiscNone QueueingDiscipline = iota - - // QDiscFIFO applies a simple fifo based queue to the underlying - // FD. - QDiscFIFO -) - -// MakeQueueingDiscipline if possible the equivalent QueuingDiscipline for s -// else returns an error. -func MakeQueueingDiscipline(s string) (QueueingDiscipline, error) { - switch s { - case "none": - return QDiscNone, nil - case "fifo": - return QDiscFIFO, nil - default: - return 0, fmt.Errorf("unsupported qdisc specified: %q", s) - } -} - -// String implements fmt.Stringer. -func (q QueueingDiscipline) String() string { - switch q { - case QDiscNone: - return "none" - case QDiscFIFO: - return "fifo" - default: - panic(fmt.Sprintf("Invalid queueing discipline: %d", q)) - } -} - // FDBasedLink configures an fd-based link. type FDBasedLink struct { Name string @@ -127,7 +90,7 @@ type FDBasedLink struct { TXChecksumOffload bool RXChecksumOffload bool LinkAddress net.HardwareAddr - QDisc QueueingDiscipline + QDisc config.QueueingDiscipline // NumChannels controls how many underlying FD's are to be used to // create this endpoint. @@ -247,8 +210,8 @@ func (n *Network) CreateLinksAndRoutes(args *CreateLinksAndRoutesArgs, _ *struct } switch link.QDisc { - case QDiscNone: - case QDiscFIFO: + case config.QDiscNone: + case config.QDiscFIFO: log.Infof("Enabling FIFO QDisc on %q", link.Name) linkEP = fifo.New(linkEP, runtime.GOMAXPROCS(0), 1000) } diff --git a/runsc/boot/strace.go b/runsc/boot/strace.go index fbfd3b07c..176981f74 100644 --- a/runsc/boot/strace.go +++ b/runsc/boot/strace.go @@ -16,9 +16,10 @@ package boot import ( "gvisor.dev/gvisor/pkg/sentry/strace" + "gvisor.dev/gvisor/runsc/config" ) -func enableStrace(conf *Config) error { +func enableStrace(conf *config.Config) error { // We must initialize even if strace is not enabled. strace.Initialize() diff --git a/runsc/boot/vfs.go b/runsc/boot/vfs.go index 08dce8b6c..3da7a64f0 100644 --- a/runsc/boot/vfs.go +++ b/runsc/boot/vfs.go @@ -42,6 +42,7 @@ import ( "gvisor.dev/gvisor/pkg/sentry/kernel/auth" "gvisor.dev/gvisor/pkg/sentry/vfs" "gvisor.dev/gvisor/pkg/syserror" + "gvisor.dev/gvisor/runsc/config" ) func registerFilesystems(k *kernel.Kernel) error { @@ -133,7 +134,7 @@ func registerFilesystems(k *kernel.Kernel) error { return nil } -func setupContainerVFS2(ctx context.Context, conf *Config, mntr *containerMounter, procArgs *kernel.CreateProcessArgs) error { +func setupContainerVFS2(ctx context.Context, conf *config.Config, mntr *containerMounter, procArgs *kernel.CreateProcessArgs) error { mns, err := mntr.setupVFS2(ctx, conf, procArgs) if err != nil { return fmt.Errorf("failed to setupFS: %w", err) @@ -149,7 +150,7 @@ func setupContainerVFS2(ctx context.Context, conf *Config, mntr *containerMounte return nil } -func (c *containerMounter) setupVFS2(ctx context.Context, conf *Config, procArgs *kernel.CreateProcessArgs) (*vfs.MountNamespace, error) { +func (c *containerMounter) setupVFS2(ctx context.Context, conf *config.Config, procArgs *kernel.CreateProcessArgs) (*vfs.MountNamespace, error) { log.Infof("Configuring container's file system with VFS2") // Create context with root credentials to mount the filesystem (the current @@ -175,7 +176,7 @@ func (c *containerMounter) setupVFS2(ctx context.Context, conf *Config, procArgs return mns, nil } -func (c *containerMounter) createMountNamespaceVFS2(ctx context.Context, conf *Config, creds *auth.Credentials) (*vfs.MountNamespace, error) { +func (c *containerMounter) createMountNamespaceVFS2(ctx context.Context, conf *config.Config, creds *auth.Credentials) (*vfs.MountNamespace, error) { fd := c.fds.remove() opts := p9MountData(fd, conf.FileAccess, true /* vfs2 */) @@ -196,7 +197,7 @@ func (c *containerMounter) createMountNamespaceVFS2(ctx context.Context, conf *C return mns, nil } -func (c *containerMounter) mountSubmountsVFS2(ctx context.Context, conf *Config, mns *vfs.MountNamespace, creds *auth.Credentials) error { +func (c *containerMounter) mountSubmountsVFS2(ctx context.Context, conf *config.Config, mns *vfs.MountNamespace, creds *auth.Credentials) error { mounts, err := c.prepareMountsVFS2() if err != nil { return err @@ -256,7 +257,7 @@ func (c *containerMounter) prepareMountsVFS2() ([]mountAndFD, error) { return mounts, nil } -func (c *containerMounter) mountSubmountVFS2(ctx context.Context, conf *Config, mns *vfs.MountNamespace, creds *auth.Credentials, submount *mountAndFD) error { +func (c *containerMounter) mountSubmountVFS2(ctx context.Context, conf *config.Config, mns *vfs.MountNamespace, creds *auth.Credentials, submount *mountAndFD) error { root := mns.Root() defer root.DecRef(ctx) target := &vfs.PathOperation{ @@ -285,7 +286,7 @@ func (c *containerMounter) mountSubmountVFS2(ctx context.Context, conf *Config, // getMountNameAndOptionsVFS2 retrieves the fsName, opts, and useOverlay values // used for mounts. -func (c *containerMounter) getMountNameAndOptionsVFS2(conf *Config, m *mountAndFD) (string, *vfs.MountOptions, error) { +func (c *containerMounter) getMountNameAndOptionsVFS2(conf *config.Config, m *mountAndFD) (string, *vfs.MountOptions, error) { fsName := m.Type var data []string @@ -383,7 +384,7 @@ func (c *containerMounter) makeSyntheticMount(ctx context.Context, currentPath s // // Note that when there are submounts inside of '/tmp', directories for the // mount points must be present, making '/tmp' not empty anymore. -func (c *containerMounter) mountTmpVFS2(ctx context.Context, conf *Config, creds *auth.Credentials, mns *vfs.MountNamespace) error { +func (c *containerMounter) mountTmpVFS2(ctx context.Context, conf *config.Config, creds *auth.Credentials, mns *vfs.MountNamespace) error { for _, m := range c.mounts { // m.Destination has been cleaned, so it's to use equality here. if m.Destination == "/tmp" { @@ -448,7 +449,7 @@ func (c *containerMounter) mountTmpVFS2(ctx context.Context, conf *Config, creds // processHintsVFS2 processes annotations that container hints about how volumes // should be mounted (e.g. a volume shared between containers). It must be // called for the root container only. -func (c *containerMounter) processHintsVFS2(conf *Config, creds *auth.Credentials) error { +func (c *containerMounter) processHintsVFS2(conf *config.Config, creds *auth.Credentials) error { ctx := c.k.SupervisorContext() for _, hint := range c.hints.mounts { // TODO(b/142076984): Only support tmpfs for now. Bind mounts require a @@ -469,7 +470,7 @@ func (c *containerMounter) processHintsVFS2(conf *Config, creds *auth.Credential // mountSharedMasterVFS2 mounts the master of a volume that is shared among // containers in a pod. -func (c *containerMounter) mountSharedMasterVFS2(ctx context.Context, conf *Config, hint *mountHint, creds *auth.Credentials) (*vfs.Mount, error) { +func (c *containerMounter) mountSharedMasterVFS2(ctx context.Context, conf *config.Config, hint *mountHint, creds *auth.Credentials) (*vfs.Mount, error) { // Map mount type to filesystem name, and parse out the options that we are // capable of dealing with. mntFD := &mountAndFD{Mount: hint.mount} @@ -485,7 +486,7 @@ func (c *containerMounter) mountSharedMasterVFS2(ctx context.Context, conf *Conf // mountSharedSubmount binds mount to a previously mounted volume that is shared // among containers in the same pod. -func (c *containerMounter) mountSharedSubmountVFS2(ctx context.Context, conf *Config, mns *vfs.MountNamespace, creds *auth.Credentials, mount specs.Mount, source *mountHint) error { +func (c *containerMounter) mountSharedSubmountVFS2(ctx context.Context, conf *config.Config, mns *vfs.MountNamespace, creds *auth.Credentials, mount specs.Mount, source *mountHint) error { if err := source.checkCompatible(mount); err != nil { return err } |