moby/daemon/exec.go
Cory Snider d222bf097c daemon: reload runtimes w/o breaking containers
The existing runtimes reload logic went to great lengths to replace the
directory containing runtime wrapper scripts as atomically as possible
within the limitations of the Linux filesystem ABI. Trouble is,
atomically swapping the wrapper scripts directory solves the wrong
problem! The runtime configuration is "locked in" when a container is
started, including the path to the runC binary. If a container is
started with a runtime which requires a daemon-managed wrapper script
and then the daemon is reloaded with a config which no longer requires
the wrapper script (i.e. some args -> no args, or the runtime is dropped
from the config), that container would become unmanageable. Any attempts
to stop, exec or otherwise perform lifecycle management operations on
the container are likely to fail due to the wrapper script no longer
existing at its original path.

Atomically swapping the wrapper scripts is also incompatible with the
read-copy-update paradigm for reloading configuration. A handler in the
daemon could retain a reference to the pre-reload configuration for an
indeterminate amount of time after the daemon configuration has been
reloaded and updated. It is possible for the daemon to attempt to start
a container using a deleted wrapper script if a request to run a
container races a reload.

Solve the problem of deleting referenced wrapper scripts by ensuring
that all wrapper scripts are *immutable* for the lifetime of the daemon
process. Any given runtime wrapper script must always exist with the
same contents, no matter how many times the daemon config is reloaded,
or what changes are made to the config. This is accomplished by using
everyone's favourite design pattern: content-addressable storage. Each
wrapper script file name is suffixed with the SHA-256 digest of its
contents to (probabilistically) guarantee immutability without needing
any concurrency control. Stale runtime wrapper scripts are only cleaned
up on the next daemon restart.

Split the derived runtimes configuration from the user-supplied
configuration to have a place to store derived state without mutating
the user-supplied configuration or exposing daemon internals in API
struct types. Hold the derived state and the user-supplied configuration
in a single struct value so that they can be updated as an atomic unit.

Signed-off-by: Cory Snider <csnider@mirantis.com>
2023-06-01 14:45:25 -04:00

357 lines
9.8 KiB
Go

package daemon // import "github.com/docker/docker/daemon"
import (
"context"
"encoding/json"
"fmt"
"io"
"runtime"
"strings"
"time"
"github.com/containerd/containerd"
"github.com/docker/docker/api/types"
containertypes "github.com/docker/docker/api/types/container"
"github.com/docker/docker/api/types/strslice"
"github.com/docker/docker/container"
"github.com/docker/docker/container/stream"
"github.com/docker/docker/errdefs"
"github.com/docker/docker/pkg/pools"
"github.com/moby/sys/signal"
"github.com/moby/term"
specs "github.com/opencontainers/runtime-spec/specs-go"
"github.com/pkg/errors"
"github.com/sirupsen/logrus"
)
func (daemon *Daemon) registerExecCommand(container *container.Container, config *container.ExecConfig) {
// Storing execs in container in order to kill them gracefully whenever the container is stopped or removed.
container.ExecCommands.Add(config.ID, config)
// Storing execs in daemon for easy access via Engine API.
daemon.execCommands.Add(config.ID, config)
}
// ExecExists looks up the exec instance and returns a bool if it exists or not.
// It will also return the error produced by `getConfig`
func (daemon *Daemon) ExecExists(name string) (bool, error) {
if _, err := daemon.getExecConfig(name); err != nil {
return false, err
}
return true, nil
}
// getExecConfig looks up the exec instance by name. If the container associated
// with the exec instance is stopped or paused, it will return an error.
func (daemon *Daemon) getExecConfig(name string) (*container.ExecConfig, error) {
ec := daemon.execCommands.Get(name)
if ec == nil {
return nil, errExecNotFound(name)
}
// If the exec is found but its container is not in the daemon's list of
// containers then it must have been deleted, in which case instead of
// saying the container isn't running, we should return a 404 so that
// the user sees the same error now that they will after the
// 5 minute clean-up loop is run which erases old/dead execs.
ctr := daemon.containers.Get(ec.Container.ID)
if ctr == nil {
return nil, containerNotFound(name)
}
if !ctr.IsRunning() {
return nil, errNotRunning(ctr.ID)
}
if ctr.IsPaused() {
return nil, errExecPaused(ctr.ID)
}
if ctr.IsRestarting() {
return nil, errContainerIsRestarting(ctr.ID)
}
return ec, nil
}
func (daemon *Daemon) unregisterExecCommand(container *container.Container, execConfig *container.ExecConfig) {
container.ExecCommands.Delete(execConfig.ID)
daemon.execCommands.Delete(execConfig.ID)
}
func (daemon *Daemon) getActiveContainer(name string) (*container.Container, error) {
ctr, err := daemon.GetContainer(name)
if err != nil {
return nil, err
}
if !ctr.IsRunning() {
return nil, errNotRunning(ctr.ID)
}
if ctr.IsPaused() {
return nil, errExecPaused(name)
}
if ctr.IsRestarting() {
return nil, errContainerIsRestarting(ctr.ID)
}
return ctr, nil
}
// ContainerExecCreate sets up an exec in a running container.
func (daemon *Daemon) ContainerExecCreate(name string, config *types.ExecConfig) (string, error) {
cntr, err := daemon.getActiveContainer(name)
if err != nil {
return "", err
}
cmd := strslice.StrSlice(config.Cmd)
entrypoint, args := daemon.getEntrypointAndArgs(strslice.StrSlice{}, cmd)
keys := []byte{}
if config.DetachKeys != "" {
keys, err = term.ToBytes(config.DetachKeys)
if err != nil {
err = fmt.Errorf("Invalid escape keys (%s) provided", config.DetachKeys)
return "", err
}
}
execConfig := container.NewExecConfig(cntr)
execConfig.OpenStdin = config.AttachStdin
execConfig.OpenStdout = config.AttachStdout
execConfig.OpenStderr = config.AttachStderr
execConfig.DetachKeys = keys
execConfig.Entrypoint = entrypoint
execConfig.Args = args
execConfig.Tty = config.Tty
execConfig.ConsoleSize = config.ConsoleSize
execConfig.Privileged = config.Privileged
execConfig.User = config.User
execConfig.WorkingDir = config.WorkingDir
linkedEnv, err := daemon.setupLinkedContainers(cntr)
if err != nil {
return "", err
}
execConfig.Env = container.ReplaceOrAppendEnvValues(cntr.CreateDaemonEnvironment(config.Tty, linkedEnv), config.Env)
if len(execConfig.User) == 0 {
execConfig.User = cntr.Config.User
}
if len(execConfig.WorkingDir) == 0 {
execConfig.WorkingDir = cntr.Config.WorkingDir
}
daemon.registerExecCommand(cntr, execConfig)
attributes := map[string]string{
"execID": execConfig.ID,
}
daemon.LogContainerEventWithAttributes(cntr, "exec_create: "+execConfig.Entrypoint+" "+strings.Join(execConfig.Args, " "), attributes)
return execConfig.ID, nil
}
// ContainerExecStart starts a previously set up exec instance. The
// std streams are set up.
// If ctx is cancelled, the process is terminated.
func (daemon *Daemon) ContainerExecStart(ctx context.Context, name string, options containertypes.ExecStartOptions) (err error) {
var (
cStdin io.ReadCloser
cStdout, cStderr io.Writer
)
ec, err := daemon.getExecConfig(name)
if err != nil {
return err
}
ec.Lock()
if ec.ExitCode != nil {
ec.Unlock()
err := fmt.Errorf("Error: Exec command %s has already run", ec.ID)
return errdefs.Conflict(err)
}
if ec.Running {
ec.Unlock()
return errdefs.Conflict(fmt.Errorf("Error: Exec command %s is already running", ec.ID))
}
ec.Running = true
ec.Unlock()
logrus.Debugf("starting exec command %s in container %s", ec.ID, ec.Container.ID)
attributes := map[string]string{
"execID": ec.ID,
}
daemon.LogContainerEventWithAttributes(ec.Container, "exec_start: "+ec.Entrypoint+" "+strings.Join(ec.Args, " "), attributes)
defer func() {
if err != nil {
ec.Lock()
ec.Container.ExecCommands.Delete(ec.ID)
ec.Running = false
exitCode := 126
ec.ExitCode = &exitCode
if err := ec.CloseStreams(); err != nil {
logrus.Errorf("failed to cleanup exec %s streams: %s", ec.Container.ID, err)
}
ec.Unlock()
}
}()
if ec.OpenStdin && options.Stdin != nil {
r, w := io.Pipe()
go func() {
defer w.Close()
defer logrus.Debug("Closing buffered stdin pipe")
pools.Copy(w, options.Stdin)
}()
cStdin = r
}
if ec.OpenStdout {
cStdout = options.Stdout
}
if ec.OpenStderr {
cStderr = options.Stderr
}
if ec.OpenStdin {
ec.StreamConfig.NewInputPipes()
} else {
ec.StreamConfig.NewNopInputPipe()
}
p := &specs.Process{}
if runtime.GOOS != "windows" {
ctr, err := daemon.containerdCli.LoadContainer(ctx, ec.Container.ID)
if err != nil {
return err
}
md, err := ctr.Info(ctx, containerd.WithoutRefreshedMetadata)
if err != nil {
return err
}
spec := specs.Spec{Process: p}
if err := json.Unmarshal(md.Spec.GetValue(), &spec); err != nil {
return err
}
}
p.Args = append([]string{ec.Entrypoint}, ec.Args...)
p.Env = ec.Env
p.Cwd = ec.WorkingDir
p.Terminal = ec.Tty
consoleSize := options.ConsoleSize
// If size isn't specified for start, use the one provided for create
if consoleSize == nil {
consoleSize = ec.ConsoleSize
}
if p.Terminal && consoleSize != nil {
p.ConsoleSize = &specs.Box{
Height: consoleSize[0],
Width: consoleSize[1],
}
}
if p.Cwd == "" {
p.Cwd = "/"
}
daemonCfg := &daemon.config().Config
if err := daemon.execSetPlatformOpt(ctx, daemonCfg, ec, p); err != nil {
return err
}
attachConfig := stream.AttachConfig{
TTY: ec.Tty,
UseStdin: cStdin != nil,
UseStdout: cStdout != nil,
UseStderr: cStderr != nil,
Stdin: cStdin,
Stdout: cStdout,
Stderr: cStderr,
DetachKeys: ec.DetachKeys,
CloseStdin: true,
}
ec.StreamConfig.AttachStreams(&attachConfig)
// using context.Background() so that attachErr does not race ctx.Done().
copyCtx, cancel := context.WithCancel(context.Background())
defer cancel()
attachErr := ec.StreamConfig.CopyStreams(copyCtx, &attachConfig)
ec.Container.Lock()
tsk, err := ec.Container.GetRunningTask()
ec.Container.Unlock()
if err != nil {
return err
}
// Synchronize with libcontainerd event loop
ec.Lock()
ec.Process, err = tsk.Exec(ctx, ec.ID, p, cStdin != nil, ec.InitializeStdio)
// the exec context should be ready, or error happened.
// close the chan to notify readiness
close(ec.Started)
if err != nil {
defer ec.Unlock()
return setExitCodeFromError(ec.SetExitCode, err)
}
ec.Unlock()
select {
case <-ctx.Done():
log := logrus.
WithField("container", ec.Container.ID).
WithField("exec", ec.ID)
log.Debug("Sending KILL signal to container process")
sigCtx, cancelFunc := context.WithTimeout(context.Background(), 30*time.Second)
defer cancelFunc()
err := ec.Process.Kill(sigCtx, signal.SignalMap["KILL"])
if err != nil {
log.WithError(err).Error("Could not send KILL signal to container process")
}
return ctx.Err()
case err := <-attachErr:
if err != nil {
if _, ok := err.(term.EscapeError); !ok {
return errdefs.System(errors.Wrap(err, "exec attach failed"))
}
attributes := map[string]string{
"execID": ec.ID,
}
daemon.LogContainerEventWithAttributes(ec.Container, "exec_detach", attributes)
}
}
return nil
}
// execCommandGC runs a ticker to clean up the daemon references
// of exec configs that are no longer part of the container.
func (daemon *Daemon) execCommandGC() {
for range time.Tick(5 * time.Minute) {
var (
cleaned int
liveExecCommands = daemon.containerExecIds()
)
for id, config := range daemon.execCommands.Commands() {
if config.CanRemove {
cleaned++
daemon.execCommands.Delete(id)
} else {
if _, exists := liveExecCommands[id]; !exists {
config.CanRemove = true
}
}
}
if cleaned > 0 {
logrus.Debugf("clean %d unused exec commands", cleaned)
}
}
}
// containerExecIds returns a list of all the current exec ids that are in use
// and running inside a container.
func (daemon *Daemon) containerExecIds() map[string]struct{} {
ids := map[string]struct{}{}
for _, c := range daemon.containers.List() {
for _, id := range c.ExecCommands.List() {
ids[id] = struct{}{}
}
}
return ids
}