Files
moby/daemon/info.go
Derek McGowan ea11b5f3fe Move cmd/dockerd/debug to daemon/command/debug
Signed-off-by: Derek McGowan <derek@mcg.dev>
2025-06-23 14:45:02 -07:00

429 lines
14 KiB
Go

// FIXME(thaJeztah): remove once we are a module; the go:build directive prevents go from downgrading language version to go1.16:
//go:build go1.23
package daemon
import (
"context"
"fmt"
"os"
"runtime"
"strconv"
"strings"
"time"
"github.com/containerd/containerd/v2/pkg/tracing"
"github.com/containerd/log"
"github.com/docker/docker/api"
"github.com/docker/docker/api/types"
"github.com/docker/docker/api/types/system"
"github.com/docker/docker/daemon/command/debug"
"github.com/docker/docker/daemon/config"
"github.com/docker/docker/daemon/internal/filedescriptors"
"github.com/docker/docker/daemon/logger"
"github.com/docker/docker/dockerversion"
"github.com/docker/docker/internal/metrics"
"github.com/docker/docker/internal/platform"
"github.com/docker/docker/pkg/meminfo"
"github.com/docker/docker/pkg/parsers/kernel"
"github.com/docker/docker/pkg/parsers/operatingsystem"
"github.com/docker/docker/pkg/sysinfo"
"github.com/docker/docker/registry"
"github.com/opencontainers/selinux/go-selinux"
)
func doWithTrace[T any](ctx context.Context, name string, f func() T) T {
_, span := tracing.StartSpan(ctx, name)
defer span.End()
return f()
}
// SystemInfo returns information about the host server the daemon is running on.
//
// The only error this should return is due to context cancellation/deadline.
// Anything else should be logged and ignored because this is looking up
// multiple things and is often used for debugging.
// The only case valid early return is when the caller doesn't want the result anymore (ie context cancelled).
func (daemon *Daemon) SystemInfo(ctx context.Context) (*system.Info, error) {
defer metrics.StartTimer(metrics.HostInfoFunctions.WithValues("system_info"))()
sysInfo := daemon.RawSysInfo()
cfg := daemon.config()
v := &system.Info{
ID: daemon.id,
Images: daemon.imageService.CountImages(ctx),
IPv4Forwarding: !sysInfo.IPv4ForwardingDisabled,
Name: hostName(ctx),
SystemTime: time.Now().Format(time.RFC3339Nano),
LoggingDriver: daemon.defaultLogConfig.Type,
KernelVersion: kernelVersion(ctx),
OperatingSystem: operatingSystem(ctx),
OSVersion: osVersion(ctx),
IndexServerAddress: registry.IndexServer,
OSType: runtime.GOOS,
Architecture: platform.Architecture(),
RegistryConfig: doWithTrace(ctx, "registry.ServiceConfig", daemon.registryService.ServiceConfig),
NCPU: doWithTrace(ctx, "runtime.NumCPU", runtime.NumCPU),
MemTotal: memInfo(ctx).MemTotal,
GenericResources: daemon.genericResources,
DockerRootDir: cfg.Root,
Labels: cfg.Labels,
ExperimentalBuild: cfg.Experimental,
ServerVersion: dockerversion.Version,
HTTPProxy: config.MaskCredentials(getConfigOrEnv(cfg.HTTPProxy, "HTTP_PROXY", "http_proxy")),
HTTPSProxy: config.MaskCredentials(getConfigOrEnv(cfg.HTTPSProxy, "HTTPS_PROXY", "https_proxy")),
NoProxy: getConfigOrEnv(cfg.NoProxy, "NO_PROXY", "no_proxy"),
LiveRestoreEnabled: cfg.LiveRestoreEnabled,
Isolation: daemon.defaultIsolation,
CDISpecDirs: promoteNil(cfg.CDISpecDirs),
}
daemon.fillContainerStates(v)
daemon.fillDebugInfo(ctx, v)
daemon.fillContainerdInfo(v, &cfg.Config)
daemon.fillAPIInfo(v, &cfg.Config)
// Retrieve platform specific info
if err := daemon.fillPlatformInfo(ctx, v, sysInfo, cfg); err != nil {
return nil, err
}
daemon.fillDriverInfo(v)
daemon.fillPluginsInfo(ctx, v, &cfg.Config)
daemon.fillSecurityOptions(v, sysInfo, &cfg.Config)
daemon.fillLicense(v)
daemon.fillDefaultAddressPools(ctx, v, &cfg.Config)
daemon.fillFirewallInfo(v)
daemon.fillDiscoveredDevicesFromDrivers(ctx, v, &cfg.Config)
return v, nil
}
// SystemVersion returns version information about the daemon.
//
// The only error this should return is due to context cancellation/deadline.
// Anything else should be logged and ignored because this is looking up
// multiple things and is often used for debugging.
// The only case valid early return is when the caller doesn't want the result anymore (ie context cancelled).
func (daemon *Daemon) SystemVersion(ctx context.Context) (types.Version, error) {
defer metrics.StartTimer(metrics.HostInfoFunctions.WithValues("system_version"))()
kernelVer := kernelVersion(ctx)
cfg := daemon.config()
v := types.Version{
Components: []types.ComponentVersion{
{
Name: "Engine",
Version: dockerversion.Version,
Details: map[string]string{
"GitCommit": dockerversion.GitCommit,
"ApiVersion": api.DefaultVersion,
"MinAPIVersion": cfg.MinAPIVersion,
"GoVersion": runtime.Version(),
"Os": runtime.GOOS,
"Arch": runtime.GOARCH,
"BuildTime": dockerversion.BuildTime,
"KernelVersion": kernelVer,
"Experimental": strconv.FormatBool(cfg.Experimental),
},
},
},
// Populate deprecated fields for older clients
Version: dockerversion.Version,
GitCommit: dockerversion.GitCommit,
APIVersion: api.DefaultVersion,
MinAPIVersion: cfg.MinAPIVersion,
GoVersion: runtime.Version(),
Os: runtime.GOOS,
Arch: runtime.GOARCH,
BuildTime: dockerversion.BuildTime,
KernelVersion: kernelVer,
Experimental: cfg.Experimental,
}
v.Platform.Name = dockerversion.PlatformName
if err := daemon.fillPlatformVersion(ctx, &v, cfg); err != nil {
return v, err
}
return v, nil
}
func (daemon *Daemon) fillDriverInfo(v *system.Info) {
v.Driver = daemon.imageService.StorageDriver()
v.DriverStatus = daemon.imageService.LayerStoreStatus()
const warnMsg = `
WARNING: The %s storage-driver is deprecated, and will be removed in a future release.
Refer to the documentation for more information: https://docs.docker.com/go/storage-driver/`
switch v.Driver {
case "overlay":
v.Warnings = append(v.Warnings, fmt.Sprintf(warnMsg, v.Driver))
}
fillDriverWarnings(v)
}
func (daemon *Daemon) fillPluginsInfo(ctx context.Context, v *system.Info, cfg *config.Config) {
v.Plugins = system.PluginsInfo{
Volume: daemon.volumes.GetDriverList(),
Network: daemon.GetNetworkDriverList(ctx),
// The authorization plugins are returned in the order they are
// used as they constitute a request/response modification chain.
Authorization: cfg.AuthorizationPlugins,
Log: logger.ListDrivers(),
}
}
func (daemon *Daemon) fillSecurityOptions(v *system.Info, sysInfo *sysinfo.SysInfo, cfg *config.Config) {
var securityOptions []string
if sysInfo.AppArmor {
securityOptions = append(securityOptions, "name=apparmor")
}
if sysInfo.Seccomp && supportsSeccomp {
if daemon.seccompProfilePath != config.SeccompProfileDefault {
v.Warnings = append(v.Warnings, "WARNING: daemon is not using the default seccomp profile")
}
securityOptions = append(securityOptions, "name=seccomp,profile="+daemon.seccompProfilePath)
}
if selinux.GetEnabled() {
securityOptions = append(securityOptions, "name=selinux")
}
if uid, gid := daemon.idMapping.RootPair(); uid != 0 || gid != 0 {
securityOptions = append(securityOptions, "name=userns")
}
if Rootless(cfg) {
securityOptions = append(securityOptions, "name=rootless")
}
if cgroupNamespacesEnabled(sysInfo, cfg) {
securityOptions = append(securityOptions, "name=cgroupns")
}
if noNewPrivileges(cfg) {
securityOptions = append(securityOptions, "name=no-new-privileges")
}
v.SecurityOptions = securityOptions
}
func (daemon *Daemon) fillContainerStates(v *system.Info) {
cRunning, cPaused, cStopped := metrics.StateCtr.Get()
v.Containers = cRunning + cPaused + cStopped
v.ContainersPaused = cPaused
v.ContainersRunning = cRunning
v.ContainersStopped = cStopped
}
// fillDebugInfo sets the current debugging state of the daemon, and additional
// debugging information, such as the number of Go-routines, and file descriptors.
//
// Note that this currently always collects the information, but the CLI only
// prints it if the daemon has debug enabled. We should consider to either make
// this information optional (cli to request "with debugging information"), or
// only collect it if the daemon has debug enabled. For the CLI code, see
// https://github.com/docker/cli/blob/v20.10.12/cli/command/system/info.go#L239-L244
func (daemon *Daemon) fillDebugInfo(ctx context.Context, v *system.Info) {
v.Debug = debug.IsEnabled()
v.NFd = filedescriptors.GetTotalUsedFds(ctx)
v.NGoroutines = runtime.NumGoroutine()
v.NEventsListener = daemon.EventsService.SubscribersCount()
}
// fillContainerdInfo provides information about the containerd configuration
// for debugging purposes.
func (daemon *Daemon) fillContainerdInfo(v *system.Info, cfg *config.Config) {
if cfg.ContainerdAddr == "" {
return
}
v.Containerd = &system.ContainerdInfo{
Address: cfg.ContainerdAddr,
Namespaces: system.ContainerdNamespaces{
Containers: cfg.ContainerdNamespace,
Plugins: cfg.ContainerdPluginNamespace,
},
}
}
func (daemon *Daemon) fillAPIInfo(v *system.Info, cfg *config.Config) {
const warn string = `
Access to the remote API is equivalent to root access on the host. Refer
to the 'Docker daemon attack surface' section in the documentation for
more information: https://docs.docker.com/go/attack-surface/`
if cfg.CorsHeaders != "" {
v.Warnings = append(v.Warnings, `DEPRECATED: The "api-cors-header" config parameter and the dockerd "--api-cors-header" option will be removed in the next release. Use a reverse proxy if you need CORS headers.`)
}
for _, host := range cfg.Hosts {
// cnf.Hosts is normalized during startup, so should always have a scheme/proto
proto, addr, _ := strings.Cut(host, "://")
if proto != "tcp" {
continue
}
const removal = "In future versions this will be a hard failure preventing the daemon from starting! Learn more at: https://docs.docker.com/go/api-security/"
if cfg.TLS == nil || !*cfg.TLS {
v.Warnings = append(v.Warnings, fmt.Sprintf("[DEPRECATION NOTICE]: API is accessible on http://%s without encryption.%s\n%s", addr, warn, removal))
continue
}
if cfg.TLSVerify == nil || !*cfg.TLSVerify {
v.Warnings = append(v.Warnings, fmt.Sprintf("[DEPRECATION NOTICE]: API is accessible on https://%s without TLS client verification.%s\n%s", addr, warn, removal))
continue
}
}
}
func (daemon *Daemon) fillDefaultAddressPools(ctx context.Context, v *system.Info, cfg *config.Config) {
_, span := tracing.StartSpan(ctx, "fillDefaultAddressPools")
defer span.End()
for _, pool := range cfg.DefaultAddressPools.Value() {
v.DefaultAddressPools = append(v.DefaultAddressPools, system.NetworkAddressPool{
Base: pool.Base.String(),
Size: pool.Size,
})
}
}
func (daemon *Daemon) fillFirewallInfo(v *system.Info) {
if daemon.netController == nil {
return
}
v.FirewallBackend = daemon.netController.FirewallBackend()
}
func hostName(ctx context.Context) string {
ctx, span := tracing.StartSpan(ctx, "hostName")
defer span.End()
hn, err := os.Hostname()
if err != nil {
log.G(ctx).WithError(err).Warn("Could not get hostname")
return ""
}
return hn
}
func kernelVersion(ctx context.Context) string {
ctx, span := tracing.StartSpan(ctx, "kernelVersion")
defer span.End()
var ver string
if kv, err := kernel.GetKernelVersion(); err != nil {
log.G(ctx).WithError(err).Warn("Could not get kernel version")
} else {
ver = kv.String()
}
return ver
}
func memInfo(ctx context.Context) *meminfo.Memory {
ctx, span := tracing.StartSpan(ctx, "memInfo")
defer span.End()
mi, err := meminfo.Read()
if err != nil {
log.G(ctx).WithError(err).Error("Could not read system memory info")
return &meminfo.Memory{}
}
return mi
}
func operatingSystem(ctx context.Context) (operatingSystem string) {
ctx, span := tracing.StartSpan(ctx, "operatingSystem")
defer span.End()
defer metrics.StartTimer(metrics.HostInfoFunctions.WithValues("operating_system"))()
if s, err := operatingsystem.GetOperatingSystem(); err != nil {
log.G(ctx).WithError(err).Warn("Could not get operating system name")
} else {
operatingSystem = s
}
if inContainer, err := operatingsystem.IsContainerized(); err != nil {
log.G(ctx).WithError(err).Error("Could not determine if daemon is containerized")
operatingSystem += " (error determining if containerized)"
} else if inContainer {
operatingSystem += " (containerized)"
}
return operatingSystem
}
func osVersion(ctx context.Context) (version string) {
ctx, span := tracing.StartSpan(ctx, "osVersion")
defer span.End()
defer metrics.StartTimer(metrics.HostInfoFunctions.WithValues("os_version"))()
version, err := operatingsystem.GetOperatingSystemVersion()
if err != nil {
log.G(ctx).WithError(err).Warn("Could not get operating system version")
return ""
}
return version
}
func getEnvAny(names ...string) string {
for _, n := range names {
if val := os.Getenv(n); val != "" {
return val
}
}
return ""
}
func getConfigOrEnv(config string, env ...string) string {
if config != "" {
return config
}
return getEnvAny(env...)
}
// promoteNil converts a nil slice to an empty slice of that type.
// A non-nil slice is returned as is.
func promoteNil[S ~[]E, E any](s S) S {
if s == nil {
return S{}
}
return s
}
// fillDiscoveredDevicesFromDrivers iterates over registered device drivers
// and calls their ListDevices method (if available) to populate system info.
func (daemon *Daemon) fillDiscoveredDevicesFromDrivers(ctx context.Context, v *system.Info, cfg *config.Config) {
ctx, span := tracing.StartSpan(ctx, "daemon.fillDiscoveredDevicesFromDrivers")
defer span.End()
// Make sure v.DiscoveredDevices is initialized to an empty slice instead of nil.
// This ensures that the JSON output is always a valid array, even if no devices are discovered.
v.DiscoveredDevices = []system.DeviceInfo{}
for driverName, driver := range deviceDrivers {
if driver.ListDevices == nil {
log.G(ctx).WithField("driver", driverName).Trace("Device driver does not implement ListDevices method.")
continue
}
ls, err := driver.ListDevices(ctx, cfg)
if err != nil {
log.G(ctx).WithFields(log.Fields{
"driver": driverName,
"error": err,
}).Warn("Failed to list devices for driver")
v.Warnings = append(v.Warnings, fmt.Sprintf("Failed to list devices from driver '%s': %v", driverName, err))
continue
}
if len(ls.Warnings) > 0 {
v.Warnings = append(v.Warnings, ls.Warnings...)
}
for _, device := range ls.Devices {
device.Source = driverName
v.DiscoveredDevices = append(v.DiscoveredDevices, device)
}
}
}