Files
runc/libcontainer/cgroups/file.go
T
Aleksa Sarai 8e8b136c49 tree-wide: use /proc/thread-self for thread-local state
With the idmap work, we will have a tainted Go thread in our
thread-group that has a different mount namespace to the other threads.
It seems that (due to some bad luck) the Go scheduler tends to make this
thread the thread-group leader in our tests, which results in very
baffling failures where /proc/self/mountinfo produces gibberish results.

In order to avoid this, switch to using /proc/thread-self for everything
that is thread-local. This primarily includes switching all file
descriptor paths (CLONE_FS), all of the places that check the current
cgroup (technically we never will run a single runc thread in a separate
cgroup, but better to be safe than sorry), and the aforementioned
mountinfo code. We don't need to do anything for the following because
the results we need aren't thread-local:

 * Checks that certain namespaces are supported by stat(2)ing
   /proc/self/ns/...

 * /proc/self/exe and /proc/self/cmdline are not thread-local.

 * While threads can be in different cgroups, we do not do this for the
   runc binary (or libcontainer) and thus we do not need to switch to
   the thread-local version of /proc/self/cgroups.

 * All of the CLONE_NEWUSER files are not thread-local because you
   cannot set the usernamespace of a single thread (setns(CLONE_NEWUSER)
   is blocked for multi-threaded programs).

Note that we have to use runtime.LockOSThread when we have an open
handle to a tid-specific procfs file that we are operating on multiple
times. Go can reschedule us such that we are running on a different
thread and then kill the original thread (causing -ENOENT or similarly
confusing errors). This is not strictly necessary for most usages of
/proc/thread-self (such as using /proc/thread-self/fd/$n directly) since
only operating on the actual inodes associated with the tid requires
this locking, but because of the pre-3.17 fallback for CentOS, we have
to do this in most cases.

In addition, CentOS's kernel is too old for /proc/thread-self, which
requires us to emulate it -- however in rootfs_linux.go, we are in the
container pid namespace but /proc is the host's procfs. This leads to
the incredibly frustrating situation where there is no way (on pre-4.1
Linux) to figure out which /proc/self/task/... entry refers to the
current tid. We can just use /proc/self in this case.

Yes this is all pretty ugly. I also wish it wasn't necessary.

Signed-off-by: Aleksa Sarai <cyphar@cyphar.com>
2023-12-14 11:36:41 +11:00

182 lines
4.8 KiB
Go

package cgroups
import (
"bytes"
"errors"
"fmt"
"os"
"path"
"strconv"
"strings"
"sync"
"github.com/opencontainers/runc/libcontainer/utils"
"github.com/sirupsen/logrus"
"golang.org/x/sys/unix"
)
// OpenFile opens a cgroup file in a given dir with given flags.
// It is supposed to be used for cgroup files only, and returns
// an error if the file is not a cgroup file.
//
// Arguments dir and file are joined together to form an absolute path
// to a file being opened.
func OpenFile(dir, file string, flags int) (*os.File, error) {
if dir == "" {
return nil, fmt.Errorf("no directory specified for %s", file)
}
return openFile(dir, file, flags)
}
// ReadFile reads data from a cgroup file in dir.
// It is supposed to be used for cgroup files only.
func ReadFile(dir, file string) (string, error) {
fd, err := OpenFile(dir, file, unix.O_RDONLY)
if err != nil {
return "", err
}
defer fd.Close()
var buf bytes.Buffer
_, err = buf.ReadFrom(fd)
return buf.String(), err
}
// WriteFile writes data to a cgroup file in dir.
// It is supposed to be used for cgroup files only.
func WriteFile(dir, file, data string) error {
fd, err := OpenFile(dir, file, unix.O_WRONLY)
if err != nil {
return err
}
defer fd.Close()
if _, err := fd.WriteString(data); err != nil {
// Having data in the error message helps in debugging.
return fmt.Errorf("failed to write %q: %w", data, err)
}
return nil
}
const (
cgroupfsDir = "/sys/fs/cgroup"
cgroupfsPrefix = cgroupfsDir + "/"
)
var (
// TestMode is set to true by unit tests that need "fake" cgroupfs.
TestMode bool
cgroupFd int = -1
prepOnce sync.Once
prepErr error
resolveFlags uint64
)
func prepareOpenat2() error {
prepOnce.Do(func() {
fd, err := unix.Openat2(-1, cgroupfsDir, &unix.OpenHow{
Flags: unix.O_DIRECTORY | unix.O_PATH,
})
if err != nil {
prepErr = &os.PathError{Op: "openat2", Path: cgroupfsDir, Err: err}
if err != unix.ENOSYS {
logrus.Warnf("falling back to securejoin: %s", prepErr)
} else {
logrus.Debug("openat2 not available, falling back to securejoin")
}
return
}
var st unix.Statfs_t
if err = unix.Fstatfs(fd, &st); err != nil {
prepErr = &os.PathError{Op: "statfs", Path: cgroupfsDir, Err: err}
logrus.Warnf("falling back to securejoin: %s", prepErr)
return
}
cgroupFd = fd
resolveFlags = unix.RESOLVE_BENEATH | unix.RESOLVE_NO_MAGICLINKS
if st.Type == unix.CGROUP2_SUPER_MAGIC {
// cgroupv2 has a single mountpoint and no "cpu,cpuacct" symlinks
resolveFlags |= unix.RESOLVE_NO_XDEV | unix.RESOLVE_NO_SYMLINKS
}
})
return prepErr
}
func openFile(dir, file string, flags int) (*os.File, error) {
mode := os.FileMode(0)
if TestMode && flags&os.O_WRONLY != 0 {
// "emulate" cgroup fs for unit tests
flags |= os.O_TRUNC | os.O_CREATE
mode = 0o600
}
path := path.Join(dir, utils.CleanPath(file))
if prepareOpenat2() != nil {
return openFallback(path, flags, mode)
}
relPath := strings.TrimPrefix(path, cgroupfsPrefix)
if len(relPath) == len(path) { // non-standard path, old system?
return openFallback(path, flags, mode)
}
fd, err := unix.Openat2(cgroupFd, relPath,
&unix.OpenHow{
Resolve: resolveFlags,
Flags: uint64(flags) | unix.O_CLOEXEC,
Mode: uint64(mode),
})
if err != nil {
err = &os.PathError{Op: "openat2", Path: path, Err: err}
// Check if cgroupFd is still opened to cgroupfsDir
// (happens when this package is incorrectly used
// across the chroot/pivot_root/mntns boundary, or
// when /sys/fs/cgroup is remounted).
//
// TODO: if such usage will ever be common, amend this
// to reopen cgroupFd and retry openat2.
fdPath, closer := utils.ProcThreadSelf("fd/" + strconv.Itoa(cgroupFd))
defer closer()
fdDest, _ := os.Readlink(fdPath)
if fdDest != cgroupfsDir {
// Wrap the error so it is clear that cgroupFd
// is opened to an unexpected/wrong directory.
err = fmt.Errorf("cgroupFd %d unexpectedly opened to %s != %s: %w",
cgroupFd, fdDest, cgroupfsDir, err)
}
return nil, err
}
return os.NewFile(uintptr(fd), path), nil
}
var errNotCgroupfs = errors.New("not a cgroup file")
// Can be changed by unit tests.
var openFallback = openAndCheck
// openAndCheck is used when openat2(2) is not available. It checks the opened
// file is on cgroupfs, returning an error otherwise.
func openAndCheck(path string, flags int, mode os.FileMode) (*os.File, error) {
fd, err := os.OpenFile(path, flags, mode)
if err != nil {
return nil, err
}
if TestMode {
return fd, nil
}
// Check this is a cgroupfs file.
var st unix.Statfs_t
if err := unix.Fstatfs(int(fd.Fd()), &st); err != nil {
_ = fd.Close()
return nil, &os.PathError{Op: "statfs", Path: path, Err: err}
}
if st.Type != unix.CGROUP_SUPER_MAGIC && st.Type != unix.CGROUP2_SUPER_MAGIC {
_ = fd.Close()
return nil, &os.PathError{Op: "open", Path: path, Err: errNotCgroupfs}
}
return fd, nil
}