Files
runc/libcontainer/configs/config.go
T
Aleksa Sarai 7c71a22705 rootfs: remove --no-mount-fallback and finally fix MS_REMOUNT
The original reasoning for this option was to avoid having mount options
be overwritten by runc. However, adding command-line arguments has
historically been a bad idea because it forces strict-runc-compatible
OCI runtimes to copy out-of-spec features directly from runc and these
flags are usually quite difficult to enable by users when using runc
through several layers of engines and orchestrators.

A far more preferable solution is to have a heuristic which detects
whether copying the original mount's mount options would override an
explicit mount option specified by the user. In this case, we should
return an error. You only end up in this path in the userns case, if you
have a bind-mount source with locked flags.

During the course of writing this patch, I discovered that several
aspects of our handling of flags for bind-mounts left much to be
desired. We have completely botched the handling of explicitly cleared
flags since commit 97f5ee4e6a ("Only remount if requested flags differ
from current"), with our behaviour only becoming increasingly more weird
with 50105de1d8 ("Fix failure with rw bind mount of a ro fuse") and
da780e4d27 ("Fix bind mounts of filesystems with certain options
set"). In short, we would only clear flags explicitly request by the
user purely by chance, in ways that it really should've been reported to
us by now. The most egregious is that mounts explicitly marked "rw" were
actually mounted "ro" if the bind-mount source was "ro" and no other
special flags were included. In addition, our handling of atime was
completely broken -- mostly due to how subtle the semantics of atime are
on Linux.

Unfortunately, while the runtime-spec requires us to implement
mount(8)'s behaviour, several aspects of the util-linux mount(8)'s
behaviour are broken and thus copying them makes little sense. Since the
runtime-spec behaviour for this case (should mount options for a "bind"
mount use the "mount --bind -o ..." or "mount --bind -o remount,..."
semantics? Is the fallback code we have for userns actually
spec-compliant?) and the mount(8) behaviour (see [1]) are not
well-defined, this commit simply fixes the most obvious aspects of the
behaviour that are broken while keeping the current spirit of the
implementation.

NOTE: The handling of atime in the base case is left for a future PR to
deal with. This means that the atime of the source mount will be
silently left alone unless the fallback path needs to be taken, and any
flags not explicitly set will be cleared in the base case. Whether we
should always be operating as "mount --bind -o remount,..." (where we
default to the original mount source flags) is a topic for a separate PR
and (probably) associated runtime-spec PR.

So, to resolve this:

* We store which flags were explicitly requested to be cleared by the
  user, so that we can detect whether the userns fallback path would end
  up setting a flag the user explicitly wished to clear. If so, we
  return an error because we couldn't fulfil the configuration settings.

* Revert 97f5ee4e6a ("Only remount if requested flags differ from
  current"), as missing flags do not mean we can skip MS_REMOUNT (in
  fact, missing flags are how you indicate a flag needs to be cleared
  with mount(2)). The original purpose of the patch was to fix the
  userns issue, but as mentioned above the correct mechanism is to do a
  fallback mount that copies the lockable flags from statfs(2).

* Improve handling of atime in the fallback case by:
    - Correctly handling the returned flags in statfs(2).
    - Implement the MNT_LOCK_ATIME checks in our code to ensure we
      produce errors rather than silently producing incorrect atime
      mounts.

* Improve the tests so we correctly detect all of these contingencies,
  including a general "bind-mount atime handling" test to ensure that
  the behaviour described here is accurate.

This change also inlines the remount() function -- it was only ever used
for the bind-mount remount case, and its behaviour is very bind-mount
specific.

[1]: https://github.com/util-linux/util-linux/issues/2433

Reverts: 97f5ee4e6a ("Only remount if requested flags differ from current")
Fixes: 50105de1d8 ("Fix failure with rw bind mount of a ro fuse")
Fixes: da780e4d27 ("Fix bind mounts of filesystems with certain options set")
Signed-off-by: Aleksa Sarai <cyphar@cyphar.com>
2023-10-24 17:28:25 +11:00

495 lines
15 KiB
Go

package configs
import (
"bytes"
"encoding/json"
"fmt"
"os/exec"
"time"
"github.com/sirupsen/logrus"
"golang.org/x/sys/unix"
"github.com/opencontainers/runc/libcontainer/devices"
"github.com/opencontainers/runtime-spec/specs-go"
)
type Rlimit struct {
Type int `json:"type"`
Hard uint64 `json:"hard"`
Soft uint64 `json:"soft"`
}
// IDMap represents UID/GID Mappings for User Namespaces.
type IDMap struct {
ContainerID int `json:"container_id"`
HostID int `json:"host_id"`
Size int `json:"size"`
}
// Seccomp represents syscall restrictions
// By default, only the native architecture of the kernel is allowed to be used
// for syscalls. Additional architectures can be added by specifying them in
// Architectures.
type Seccomp struct {
DefaultAction Action `json:"default_action"`
Architectures []string `json:"architectures"`
Flags []specs.LinuxSeccompFlag `json:"flags"`
Syscalls []*Syscall `json:"syscalls"`
DefaultErrnoRet *uint `json:"default_errno_ret"`
ListenerPath string `json:"listener_path,omitempty"`
ListenerMetadata string `json:"listener_metadata,omitempty"`
}
// Action is taken upon rule match in Seccomp
type Action int
const (
Kill Action = iota + 1
Errno
Trap
Allow
Trace
Log
Notify
KillThread
KillProcess
)
// Operator is a comparison operator to be used when matching syscall arguments in Seccomp
type Operator int
const (
EqualTo Operator = iota + 1
NotEqualTo
GreaterThan
GreaterThanOrEqualTo
LessThan
LessThanOrEqualTo
MaskEqualTo
)
// Arg is a rule to match a specific syscall argument in Seccomp
type Arg struct {
Index uint `json:"index"`
Value uint64 `json:"value"`
ValueTwo uint64 `json:"value_two"`
Op Operator `json:"op"`
}
// Syscall is a rule to match a syscall in Seccomp
type Syscall struct {
Name string `json:"name"`
Action Action `json:"action"`
ErrnoRet *uint `json:"errnoRet"`
Args []*Arg `json:"args"`
}
// Config defines configuration options for executing a process inside a contained environment.
type Config struct {
// NoPivotRoot will use MS_MOVE and a chroot to jail the process into the container's rootfs
// This is a common option when the container is running in ramdisk
NoPivotRoot bool `json:"no_pivot_root"`
// ParentDeathSignal specifies the signal that is sent to the container's process in the case
// that the parent process dies.
ParentDeathSignal int `json:"parent_death_signal"`
// Path to a directory containing the container's root filesystem.
Rootfs string `json:"rootfs"`
// Umask is the umask to use inside of the container.
Umask *uint32 `json:"umask"`
// Readonlyfs will remount the container's rootfs as readonly where only externally mounted
// bind mounts are writtable.
Readonlyfs bool `json:"readonlyfs"`
// Specifies the mount propagation flags to be applied to /.
RootPropagation int `json:"rootPropagation"`
// Mounts specify additional source and destination paths that will be mounted inside the container's
// rootfs and mount namespace if specified
Mounts []*Mount `json:"mounts"`
// The device nodes that should be automatically created within the container upon container start. Note, make sure that the node is marked as allowed in the cgroup as well!
Devices []*devices.Device `json:"devices"`
MountLabel string `json:"mount_label"`
// Hostname optionally sets the container's hostname if provided
Hostname string `json:"hostname"`
// Domainname optionally sets the container's domainname if provided
Domainname string `json:"domainname"`
// Namespaces specifies the container's namespaces that it should setup when cloning the init process
// If a namespace is not provided that namespace is shared from the container's parent process
Namespaces Namespaces `json:"namespaces"`
// Capabilities specify the capabilities to keep when executing the process inside the container
// All capabilities not specified will be dropped from the processes capability mask
Capabilities *Capabilities `json:"capabilities"`
// Networks specifies the container's network setup to be created
Networks []*Network `json:"networks"`
// Routes can be specified to create entries in the route table as the container is started
Routes []*Route `json:"routes"`
// Cgroups specifies specific cgroup settings for the various subsystems that the container is
// placed into to limit the resources the container has available
Cgroups *Cgroup `json:"cgroups"`
// AppArmorProfile specifies the profile to apply to the process running in the container and is
// change at the time the process is execed
AppArmorProfile string `json:"apparmor_profile,omitempty"`
// ProcessLabel specifies the label to apply to the process running in the container. It is
// commonly used by selinux
ProcessLabel string `json:"process_label,omitempty"`
// Rlimits specifies the resource limits, such as max open files, to set in the container
// If Rlimits are not set, the container will inherit rlimits from the parent process
Rlimits []Rlimit `json:"rlimits,omitempty"`
// OomScoreAdj specifies the adjustment to be made by the kernel when calculating oom scores
// for a process. Valid values are between the range [-1000, '1000'], where processes with
// higher scores are preferred for being killed. If it is unset then we don't touch the current
// value.
// More information about kernel oom score calculation here: https://lwn.net/Articles/317814/
OomScoreAdj *int `json:"oom_score_adj,omitempty"`
// UIDMappings is an array of User ID mappings for User Namespaces
UIDMappings []IDMap `json:"uid_mappings"`
// GIDMappings is an array of Group ID mappings for User Namespaces
GIDMappings []IDMap `json:"gid_mappings"`
// MaskPaths specifies paths within the container's rootfs to mask over with a bind
// mount pointing to /dev/null as to prevent reads of the file.
MaskPaths []string `json:"mask_paths"`
// ReadonlyPaths specifies paths within the container's rootfs to remount as read-only
// so that these files prevent any writes.
ReadonlyPaths []string `json:"readonly_paths"`
// Sysctl is a map of properties and their values. It is the equivalent of using
// sysctl -w my.property.name value in Linux.
Sysctl map[string]string `json:"sysctl"`
// Seccomp allows actions to be taken whenever a syscall is made within the container.
// A number of rules are given, each having an action to be taken if a syscall matches it.
// A default action to be taken if no rules match is also given.
Seccomp *Seccomp `json:"seccomp"`
// NoNewPrivileges controls whether processes in the container can gain additional privileges.
NoNewPrivileges bool `json:"no_new_privileges,omitempty"`
// Hooks are a collection of actions to perform at various container lifecycle events.
// CommandHooks are serialized to JSON, but other hooks are not.
Hooks Hooks
// Version is the version of opencontainer specification that is supported.
Version string `json:"version"`
// Labels are user defined metadata that is stored in the config and populated on the state
Labels []string `json:"labels"`
// NoNewKeyring will not allocated a new session keyring for the container. It will use the
// callers keyring in this case.
NoNewKeyring bool `json:"no_new_keyring"`
// IntelRdt specifies settings for Intel RDT group that the container is placed into
// to limit the resources (e.g., L3 cache, memory bandwidth) the container has available
IntelRdt *IntelRdt `json:"intel_rdt,omitempty"`
// RootlessEUID is set when the runc was launched with non-zero EUID.
// Note that RootlessEUID is set to false when launched with EUID=0 in userns.
// When RootlessEUID is set, runc creates a new userns for the container.
// (config.json needs to contain userns settings)
RootlessEUID bool `json:"rootless_euid,omitempty"`
// RootlessCgroups is set when unlikely to have the full access to cgroups.
// When RootlessCgroups is set, cgroups errors are ignored.
RootlessCgroups bool `json:"rootless_cgroups,omitempty"`
// TimeOffsets specifies the offset for supporting time namespaces.
TimeOffsets map[string]specs.LinuxTimeOffset `json:"time_offsets,omitempty"`
// Scheduler represents the scheduling attributes for a process.
Scheduler *Scheduler `json:"scheduler,omitempty"`
}
// Scheduler is based on the Linux sched_setattr(2) syscall.
type Scheduler = specs.Scheduler
// ToSchedAttr is to convert *configs.Scheduler to *unix.SchedAttr.
func ToSchedAttr(scheduler *Scheduler) (*unix.SchedAttr, error) {
var policy uint32
switch scheduler.Policy {
case specs.SchedOther:
policy = 0
case specs.SchedFIFO:
policy = 1
case specs.SchedRR:
policy = 2
case specs.SchedBatch:
policy = 3
case specs.SchedISO:
policy = 4
case specs.SchedIdle:
policy = 5
case specs.SchedDeadline:
policy = 6
default:
return nil, fmt.Errorf("invalid scheduler policy: %s", scheduler.Policy)
}
var flags uint64
for _, flag := range scheduler.Flags {
switch flag {
case specs.SchedFlagResetOnFork:
flags |= 0x01
case specs.SchedFlagReclaim:
flags |= 0x02
case specs.SchedFlagDLOverrun:
flags |= 0x04
case specs.SchedFlagKeepPolicy:
flags |= 0x08
case specs.SchedFlagKeepParams:
flags |= 0x10
case specs.SchedFlagUtilClampMin:
flags |= 0x20
case specs.SchedFlagUtilClampMax:
flags |= 0x40
default:
return nil, fmt.Errorf("invalid scheduler flag: %s", flag)
}
}
return &unix.SchedAttr{
Size: unix.SizeofSchedAttr,
Policy: policy,
Flags: flags,
Nice: scheduler.Nice,
Priority: uint32(scheduler.Priority),
Runtime: scheduler.Runtime,
Deadline: scheduler.Deadline,
Period: scheduler.Period,
}, nil
}
type (
HookName string
HookList []Hook
Hooks map[HookName]HookList
)
const (
// Prestart commands are executed after the container namespaces are created,
// but before the user supplied command is executed from init.
// Note: This hook is now deprecated
// Prestart commands are called in the Runtime namespace.
Prestart HookName = "prestart"
// CreateRuntime commands MUST be called as part of the create operation after
// the runtime environment has been created but before the pivot_root has been executed.
// CreateRuntime is called immediately after the deprecated Prestart hook.
// CreateRuntime commands are called in the Runtime Namespace.
CreateRuntime HookName = "createRuntime"
// CreateContainer commands MUST be called as part of the create operation after
// the runtime environment has been created but before the pivot_root has been executed.
// CreateContainer commands are called in the Container namespace.
CreateContainer HookName = "createContainer"
// StartContainer commands MUST be called as part of the start operation and before
// the container process is started.
// StartContainer commands are called in the Container namespace.
StartContainer HookName = "startContainer"
// Poststart commands are executed after the container init process starts.
// Poststart commands are called in the Runtime Namespace.
Poststart HookName = "poststart"
// Poststop commands are executed after the container init process exits.
// Poststop commands are called in the Runtime Namespace.
Poststop HookName = "poststop"
)
// KnownHookNames returns the known hook names.
// Used by `runc features`.
func KnownHookNames() []string {
return []string{
string(Prestart), // deprecated
string(CreateRuntime),
string(CreateContainer),
string(StartContainer),
string(Poststart),
string(Poststop),
}
}
type Capabilities struct {
// Bounding is the set of capabilities checked by the kernel.
Bounding []string
// Effective is the set of capabilities checked by the kernel.
Effective []string
// Inheritable is the capabilities preserved across execve.
Inheritable []string
// Permitted is the limiting superset for effective capabilities.
Permitted []string
// Ambient is the ambient set of capabilities that are kept.
Ambient []string
}
// Deprecated: use (Hooks).Run instead.
func (hooks HookList) RunHooks(state *specs.State) error {
for i, h := range hooks {
if err := h.Run(state); err != nil {
return fmt.Errorf("error running hook #%d: %w", i, err)
}
}
return nil
}
func (hooks *Hooks) UnmarshalJSON(b []byte) error {
var state map[HookName][]CommandHook
if err := json.Unmarshal(b, &state); err != nil {
return err
}
*hooks = Hooks{}
for n, commandHooks := range state {
if len(commandHooks) == 0 {
continue
}
(*hooks)[n] = HookList{}
for _, h := range commandHooks {
(*hooks)[n] = append((*hooks)[n], h)
}
}
return nil
}
func (hooks *Hooks) MarshalJSON() ([]byte, error) {
serialize := func(hooks []Hook) (serializableHooks []CommandHook) {
for _, hook := range hooks {
switch chook := hook.(type) {
case CommandHook:
serializableHooks = append(serializableHooks, chook)
default:
logrus.Warnf("cannot serialize hook of type %T, skipping", hook)
}
}
return serializableHooks
}
return json.Marshal(map[string]interface{}{
"prestart": serialize((*hooks)[Prestart]),
"createRuntime": serialize((*hooks)[CreateRuntime]),
"createContainer": serialize((*hooks)[CreateContainer]),
"startContainer": serialize((*hooks)[StartContainer]),
"poststart": serialize((*hooks)[Poststart]),
"poststop": serialize((*hooks)[Poststop]),
})
}
// Run executes all hooks for the given hook name.
func (hooks Hooks) Run(name HookName, state *specs.State) error {
list := hooks[name]
for i, h := range list {
if err := h.Run(state); err != nil {
return fmt.Errorf("error running %s hook #%d: %w", name, i, err)
}
}
return nil
}
type Hook interface {
// Run executes the hook with the provided state.
Run(*specs.State) error
}
// NewFunctionHook will call the provided function when the hook is run.
func NewFunctionHook(f func(*specs.State) error) FuncHook {
return FuncHook{
run: f,
}
}
type FuncHook struct {
run func(*specs.State) error
}
func (f FuncHook) Run(s *specs.State) error {
return f.run(s)
}
type Command struct {
Path string `json:"path"`
Args []string `json:"args"`
Env []string `json:"env"`
Dir string `json:"dir"`
Timeout *time.Duration `json:"timeout"`
}
// NewCommandHook will execute the provided command when the hook is run.
func NewCommandHook(cmd Command) CommandHook {
return CommandHook{
Command: cmd,
}
}
type CommandHook struct {
Command
}
func (c Command) Run(s *specs.State) error {
b, err := json.Marshal(s)
if err != nil {
return err
}
var stdout, stderr bytes.Buffer
cmd := exec.Cmd{
Path: c.Path,
Args: c.Args,
Env: c.Env,
Stdin: bytes.NewReader(b),
Stdout: &stdout,
Stderr: &stderr,
}
if err := cmd.Start(); err != nil {
return err
}
errC := make(chan error, 1)
go func() {
err := cmd.Wait()
if err != nil {
err = fmt.Errorf("%w, stdout: %s, stderr: %s", err, stdout.String(), stderr.String())
}
errC <- err
}()
var timerCh <-chan time.Time
if c.Timeout != nil {
timer := time.NewTimer(*c.Timeout)
defer timer.Stop()
timerCh = timer.C
}
select {
case err := <-errC:
return err
case <-timerCh:
_ = cmd.Process.Kill()
<-errC
return fmt.Errorf("hook ran past specified timeout of %.1fs", c.Timeout.Seconds())
}
}