In the OPEN_TREE_NAMESPACE path vfs_open_tree() resolves a path via
filename_lookup() without holding namespace_lock. Between the lookup
and create_new_namespace() acquiring namespace_lock via
LOCK_MOUNT_EXACT_COPY() another thread can unmount the mount, setting
mnt->mnt_ns to NULL.
When create_new_namespace() then checks !mnt->mnt_ns it incorrectly
takes the swap-and-mntget path that was designed for fsmount()'s
detached mounts. This reuses a mount whose mnt_mp_list is in an
inconsistent state from the concurrent unmount, causing a general
protection fault in __umount_mnt() -> hlist_del_init(&mnt->mnt_mp_list)
during namespace teardown.
Remove the !mnt->mnt_ns special case entirely. Instead, always
duplicate the mount:
- For OPEN_TREE_NAMESPACE use __do_loopback() which will properly
clone the mount or reject it via may_copy_tree() if it was
unmounted in the race window.
- For fsmount() use clone_mnt() directly (via the new MOUNT_COPY_NEW
flag) since the mount is freshly created by vfs_create_mount() and
not in any namespace so __do_loopback()'s IS_MNT_UNBINDABLE,
may_copy_tree, and __has_locked_children checks don't apply.
Reported-by: syzbot+e4470cc28308f2081ec8@syzkaller.appspotmail.com
Signed-off-by: Christian Brauner <brauner@kernel.org>
Allow MOVE_MOUNT_BENEATH to target the caller's rootfs. When the target
of a mount-beneath operation is the caller's root mount, verify that:
(1) The caller is located at the root of the mount, as enforced by
path_mounted() in do_lock_mount().
(2) Propagation from the parent mount would not overmount the target,
to avoid propagating beneath the rootfs of other mount namespaces.
The root-switching is decomposed into individually atomic, locally-scoped
steps: mount-beneath inserts the new root under the old one, chroot(".")
switches the caller's root, and umount2(".", MNT_DETACH) removes the old
root. Since each step only modifies the caller's own state, this avoids
cross-namespace vulnerabilities and inherent fork/unshare/setns races
that a chroot_fs_refs()-based approach would have.
Userspace can use the following workflow to switch roots:
fd_tree = open_tree(-EBADF, "/newroot",
OPEN_TREE_CLONE | OPEN_TREE_CLOEXEC);
fchdir(fd_tree);
move_mount(fd_tree, "", AT_FDCWD, "/",
MOVE_MOUNT_BENEATH | MOVE_MOUNT_F_EMPTY_PATH);
chroot(".");
umount2(".", MNT_DETACH);
Link: https://patch.msgid.link/20260224-work-mount-beneath-rootfs-v1-2-8c58bf08488f@kernel.org
Signed-off-by: Christian Brauner <brauner@kernel.org>
When performing a mount-beneath operation the target mount can often be
locked:
unshare(CLONE_NEWUSER | CLONE_NEWNS);
mount --beneath -t tmpfs tmpfs /proc
will fail because the procfs mount on /proc became locked when the mount
namespace was created from the parent mount namespace. Same logic for:
unshare(CLONE_NEWUSER | CLONE_NEWNS);
mount --beneath -t tmpfs tmpfs /
MNT_LOCKED is raised to prevent an unprivileged mount namespace from
revealing whatever is under a given mount. To replace the rootfs we need
to handle that case though.
We can simply transfer the locked mount property from the top mount to
the mount beneath. The new mount we mounted beneath the top mount takes
over the job of the top mount in protecting the parent mount from being
revealed. This leaves us free to allow the top mount to be unmounted.
This also works during mount propagation and also works for the
non-MOVE_MOUNT_BENEATH case:
(1) move_mount(MOVE_MOUNT_BENEATH): @source_mnt->overmount always NULL
(2) move_mount(): @source_mnt->overmount maybe !NULL
For (1) can_move_mount_beneath() rejects overmounted @source_mnt (We
could allow this but whatever it's not really a use-case and it's fugly
to move an overmounted mount stack around. What are you even doing? So
let's keep that restriction.
For (2) we can have @source_mnt overmounted (Someone overmounted us
while we locked the target mount.). Both are fine. @source_mnt will be
mounted on whatever @q was mounted on and @q will be mounted on the top
of the @source_mnt mount stack. Even in such cases we can unlock @q and
lock @source_mnt if @q was locked.
This effectively makes mount propagation useful in cases where a mount
namespace has a locked mount somewhere and we propagate a new mount
beneath it but the mount namespace could never get at it because the old
top mount remains locked. Again, we just let the newly propagated mount
take over the protection and unlock the top mount.
Link: https://patch.msgid.link/20260224-work-mount-beneath-rootfs-v1-1-8c58bf08488f@kernel.org
Signed-off-by: Christian Brauner <brauner@kernel.org>
Add support for creating a mount namespace that contains only a copy of
the root mount from the caller's mount namespace, with none of the
child mounts. This is useful for containers and sandboxes that want to
start with a minimal mount table and populate it from scratch rather
than inheriting and then tearing down the full mount tree.
Two new flags are introduced:
- CLONE_EMPTY_MNTNS for clone3(), using the 64-bit flag space.
- UNSHARE_EMPTY_MNTNS for unshare(), reusing the
CLONE_PARENT_SETTID bit which has no meaning for unshare.
Both flags imply CLONE_NEWNS. For the unshare path,
UNSHARE_EMPTY_MNTNS is converted to CLONE_EMPTY_MNTNS in
unshare_nsproxy_namespaces() before it reaches copy_mnt_ns(), so the
mount namespace code only needs to handle a single flag.
In copy_mnt_ns(), when CLONE_EMPTY_MNTNS is set, clone_mnt() is used
instead of copy_tree() to clone only the root mount. The caller's root
and working directory are both reset to the root dentry of the new
mount.
The cleanup variables are changed from vfsmount pointers with
__free(mntput) to struct path with __free(path_put) because the empty
mount namespace path needs to release both mount and dentry references
when replacing the caller's root and pwd. In the normal (non-empty)
path only the mount component is set, and dput(NULL) is a no-op so
path_put remains correct there as well.
Link: https://patch.msgid.link/20260306-work-empty-mntns-consolidated-v1-1-6eb30529bbb0@kernel.org
Signed-off-by: Christian Brauner <brauner@kernel.org>
Add FSMOUNT_NAMESPACE flag to fsmount() that creates a new mount
namespace with the newly created filesystem attached to a copy of the
real rootfs. This returns a namespace file descriptor instead of an
O_PATH mount fd, similar to how OPEN_TREE_NAMESPACE works for open_tree().
This allows creating a new filesystem and immediately placing it in a
new mount namespace in a single operation, which is useful for container
runtimes and other namespace-based isolation mechanisms.
The rootfs mount is created before copying the real rootfs for the new
namespace meaning that the mount namespace id for the mount of the root
of the namespace is bigger than the child mounted on top of it. We've
never explicitly given the guarantee for such ordering and I doubt
anyone relies on it. Accepting that lets us avoid copying the mount
again and also avoids having to massage may_copy_tree() to grant an
exception for fsmount->mnt->mnt_ns being NULL.
Link: https://patch.msgid.link/20260122-work-fsmount-namespace-v1-3-5ef0a886e646@kernel.org
Signed-off-by: Christian Brauner <brauner@kernel.org>
Pull vfs fixes from Christian Brauner:
- Fix an uninitialized variable in file_getattr().
The flags_valid field wasn't initialized before calling
vfs_fileattr_get(), triggering KMSAN uninit-value reports in fuse
- Fix writeback wakeup and logging timeouts when DETECT_HUNG_TASK is
not enabled.
sysctl_hung_task_timeout_secs is 0 in that case causing spurious
"waiting for writeback completion for more than 1 seconds" warnings
- Fix a null-ptr-deref in do_statmount() when the mount is internal
- Add missing kernel-doc description for the @private parameter in
iomap_readahead()
- Fix mount namespace creation to hold namespace_sem across the mount
copy in create_new_namespace().
The previous drop-and-reacquire pattern was fragile and failed to
clean up mount propagation links if the real rootfs was a shared or
dependent mount
- Fix /proc mount iteration where m->index wasn't updated when
m->show() overflows, causing a restart to repeatedly show the same
mount entry in a rapidly expanding mount table
- Return EFSCORRUPTED instead of ENOSPC in minix_new_inode() when the
inode number is out of range
- Fix unshare(2) when CLONE_NEWNS is set and current->fs isn't shared.
copy_mnt_ns() received the live fs_struct so if a subsequent
namespace creation failed the rollback would leave pwd and root
pointing to detached mounts. Always allocate a new fs_struct when
CLONE_NEWNS is requested
- fserror bug fixes:
- Remove the unused fsnotify_sb_error() helper now that all callers
have been converted to fserror_report_metadata
- Fix a lockdep splat in fserror_report() where igrab() takes
inode::i_lock which can be held in IRQ context.
Replace igrab() with a direct i_count bump since filesystems
should not report inodes that are about to be freed or not yet
exposed
- Handle error pointer in procfs for try_lookup_noperm()
- Fix an integer overflow in ep_loop_check_proc() where recursive calls
returning INT_MAX would overflow when +1 is added, breaking the
recursion depth check
- Fix a misleading break in pidfs
* tag 'vfs-7.0-rc2.fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/vfs/vfs:
pidfs: avoid misleading break
eventpoll: Fix integer overflow in ep_loop_check_proc()
proc: Fix pointer error dereference
fserror: fix lockdep complaint when igrabbing inode
fsnotify: drop unused helper
unshare: fix unshare_fs() handling
minix: Correct errno in minix_new_inode
namespace: fix proc mount iteration
mount: hold namespace_sem across copy in create_new_namespace()
iomap: Describe @private in iomap_readahead()
statmount: Fix the null-ptr-deref in do_statmount()
writeback: Fix wakeup and logging timeouts for !DETECT_HUNG_TASK
fs: init flags_valid before calling vfs_fileattr_get
This was done entirely with mindless brute force, using
git grep -l '\<k[vmz]*alloc_objs*(.*, GFP_KERNEL)' |
xargs sed -i 's/\(alloc_objs*(.*\), GFP_KERNEL)/\1)/'
to convert the new alloc_obj() users that had a simple GFP_KERNEL
argument to just drop that argument.
Note that due to the extreme simplicity of the scripting, any slightly
more complex cases spread over multiple lines would not be triggered:
they definitely exist, but this covers the vast bulk of the cases, and
the resulting diff is also then easier to check automatically.
For the same reason the 'flex' versions will be done as a separate
conversion.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This is the result of running the Coccinelle script from
scripts/coccinelle/api/kmalloc_objs.cocci. The script is designed to
avoid scalar types (which need careful case-by-case checking), and
instead replace kmalloc-family calls that allocate struct or union
object instances:
Single allocations: kmalloc(sizeof(TYPE), ...)
are replaced with: kmalloc_obj(TYPE, ...)
Array allocations: kmalloc_array(COUNT, sizeof(TYPE), ...)
are replaced with: kmalloc_objs(TYPE, COUNT, ...)
Flex array allocations: kmalloc(struct_size(PTR, FAM, COUNT), ...)
are replaced with: kmalloc_flex(*PTR, FAM, COUNT, ...)
(where TYPE may also be *VAR)
The resulting allocations no longer return "void *", instead returning
"TYPE *".
Signed-off-by: Kees Cook <kees@kernel.org>
The m->index isn't updated when m->show() overflows and retains its
value before the current mount causing a restart to start at the same
value. If that happens in short order to due a quickly expanding mount
table this would cause the same mount to be shown again and again.
Ensure that *pos always equals the mount id of the mount that was
returned by start/next. On restart after overflow mnt_find_id_at(*pos)
finds the exact mount. This should avoid duplicates, avoid skips and
should handle concurrent modification just fine.
Cc: <stable@vger.kernel.org>
Fixed: 2eea9ce431 ("mounts: keep list of mounts in an rbtree")
Link: https://patch.msgid.link/20260129-geleckt-treuhand-4bb940acacd9@brauner
Signed-off-by: Christian Brauner <brauner@kernel.org>
Fix an oversight when creating a new mount namespace. If someone had the
bright idea to make the real rootfs a shared or dependent mount and it
is later copied the copy will become a peer of the old real rootfs mount
or a dependent mount of it. The namespace semaphore is dropped and we
use mount lock exact to lock the new real root mount. If that fails or
the subsequent do_loopback() fails we rely on the copy of the real root
mount to be cleaned up by path_put(). The problem is that this doesn't
deal with mount propagation and will leave the mounts linked in the
propagation lists.
When creating a new mount namespace create_new_namespace() first
acquires namespace_sem to clone the nullfs root, drops it, then
reacquires it via LOCK_MOUNT_EXACT which takes inode_lock first to
respect the inode_lock -> namespace_sem lock ordering. This
drop-and-reacquire pattern is fragile and was the source of the
propagation cleanup bug fixed in the preceding commit.
Extend lock_mount_exact() with a copy_mount mode that clones the mount
under the locks atomically. When copy_mount is true, path_overmounted()
is skipped since we're copying the mount, not mounting on top of it -
the nullfs root always has rootfs mounted on top so the check would
always fail. If clone_mnt() fails after get_mountpoint() has pinned the
mountpoint, __unlock_mount() is used to properly unpin the mountpoint
and release both locks.
This allows create_new_namespace() to use LOCK_MOUNT_EXACT_COPY which
takes inode_lock and namespace_sem once and holds them throughout the
clone and subsequent mount operations, eliminating the
drop-and-reacquire pattern entirely.
Reported-by: syzbot+a89f9434fb5a001ccd58@syzkaller.appspotmail.com
Fixes: 9b8a0ba682 ("mount: add OPEN_TREE_NAMESPACE") # mainline only
Link: https://lore.kernel.org/699047f6.050a0220.2757fb.0024.GAE@google.com
Signed-off-by: Christian Brauner <brauner@kernel.org>
Pull vfs 'struct filename' updates from Al Viro:
"[Mostly] sanitize struct filename handling"
* tag 'pull-filename' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs: (68 commits)
sysfs(2): fs_index() argument is _not_ a pathname
alpha: switch osf_mount() to strndup_user()
ksmbd: use CLASS(filename_kernel)
mqueue: switch to CLASS(filename)
user_statfs(): switch to CLASS(filename)
statx: switch to CLASS(filename_maybe_null)
quotactl_block(): switch to CLASS(filename)
chroot(2): switch to CLASS(filename)
move_mount(2): switch to CLASS(filename_maybe_null)
namei.c: switch user pathname imports to CLASS(filename{,_flags})
namei.c: convert getname_kernel() callers to CLASS(filename_kernel)
do_f{chmod,chown,access}at(): use CLASS(filename_uflags)
do_readlinkat(): switch to CLASS(filename_flags)
do_sys_truncate(): switch to CLASS(filename)
do_utimes_path(): switch to CLASS(filename_uflags)
chdir(2): unspaghettify a bit...
do_fchownat(): unspaghettify a bit...
fspick(2): use CLASS(filename_flags)
name_to_handle_at(): use CLASS(filename_uflags)
vfs_open_tree(): use CLASS(filename_uflags)
...
Pull misc vfs updates from Christian Brauner:
"This contains a mix of VFS cleanups, performance improvements, API
fixes, documentation, and a deprecation notice.
Scalability and performance:
- Rework pid allocation to only take pidmap_lock once instead of
twice during alloc_pid(), improving thread creation/teardown
throughput by 10-16% depending on false-sharing luck. Pad the
namespace refcount to reduce false-sharing
- Track file lock presence via a flag in ->i_opflags instead of
reading ->i_flctx, avoiding false-sharing with ->i_readcount on
open/close hot paths. Measured 4-16% improvement on 24-core
open-in-a-loop benchmarks
- Use a consume fence in locks_inode_context() to match the
store-release/load-consume idiom, eliminating a hardware fence on
some architectures
- Annotate cdev_lock with __cacheline_aligned_in_smp to prevent
false-sharing
- Remove a redundant DCACHE_MANAGED_DENTRY check in
__follow_mount_rcu() that never fires since the caller already
verifies it, eliminating a 100% mispredicted branch
- Fix a 100% mispredicted likely() in devcgroup_inode_permission()
that became wrong after a prior code reorder
Bug fixes and correctness:
- Make insert_inode_locked() wait for inode destruction instead of
skipping, fixing a corner case where two matching inodes could
exist in the hash
- Move f_mode initialization before file_ref_init() in alloc_file()
to respect the SLAB_TYPESAFE_BY_RCU ordering contract
- Add a WARN_ON_ONCE guard in try_to_free_buffers() for folios with
no buffers attached, preventing a null pointer dereference when
AS_RELEASE_ALWAYS is set but no release_folio op exists
- Fix select restart_block to store end_time as timespec64, avoiding
truncation of tv_sec on 32-bit architectures
- Make dump_inode() use get_kernel_nofault() to safely access inode
and superblock fields, matching the dump_mapping() pattern
API modernization:
- Make posix_acl_to_xattr() allocate the buffer internally since
every single caller was doing it anyway. Reduces boilerplate and
unnecessary error checking across ~15 filesystems
- Replace deprecated simple_strtoul() with kstrtoul() for the
ihash_entries, dhash_entries, mhash_entries, and mphash_entries
boot parameters, adding proper error handling
- Convert chardev code to use guard(mutex) and __free(kfree) cleanup
patterns
- Replace min_t() with min() or umin() in VFS code to avoid silently
truncating unsigned long to unsigned int
- Gate LOOKUP_RCU assertions behind CONFIG_DEBUG_VFS since callers
already check the flag
Deprecation:
- Begin deprecating legacy BSD process accounting (acct(2)). The
interface has numerous footguns and better alternatives exist
(eBPF)
Documentation:
- Fix and complete kernel-doc for struct export_operations, removing
duplicated documentation between ReST and source
- Fix kernel-doc warnings for __start_dirop() and ilookup5_nowait()
Testing:
- Add a kunit test for initramfs cpio handling of entries with
filesize > PATH_MAX
Misc:
- Add missing <linux/init_task.h> include in fs_struct.c"
* tag 'vfs-7.0-rc1.misc' of git://git.kernel.org/pub/scm/linux/kernel/git/vfs/vfs: (28 commits)
posix_acl: make posix_acl_to_xattr() alloc the buffer
fs: make insert_inode_locked() wait for inode destruction
initramfs_test: kunit test for cpio.filesize > PATH_MAX
fs: improve dump_inode() to safely access inode fields
fs: add <linux/init_task.h> for 'init_fs'
docs: exportfs: Use source code struct documentation
fs: move initializing f_mode before file_ref_init()
exportfs: Complete kernel-doc for struct export_operations
exportfs: Mark struct export_operations functions at kernel-doc
exportfs: Fix kernel-doc output for get_name()
acct(2): begin the deprecation of legacy BSD process accounting
device_cgroup: remove branch hint after code refactor
VFS: fix __start_dirop() kernel-doc warnings
fs: Describe @isnew parameter in ilookup5_nowait()
fs/namei: Remove redundant DCACHE_MANAGED_DENTRY check in __follow_mount_rcu
fs: only assert on LOOKUP_RCU when built with CONFIG_DEBUG_VFS
select: store end_time as timespec64 in restart block
chardev: Switch to guard(mutex) and __free(kfree)
namespace: Replace simple_strtoul with kstrtoul to parse boot params
dcache: Replace simple_strtoul with kstrtoul in set_dhash_entries
...
Pull vfs mount updates from Christian Brauner:
- statmount: accept fd as a parameter
Extend struct mnt_id_req with a file descriptor field and a new
STATMOUNT_BY_FD flag. When set, statmount() returns mount information
for the mount the fd resides on — including detached mounts
(unmounted via umount2(MNT_DETACH)).
For detached mounts the STATMOUNT_MNT_POINT and STATMOUNT_MNT_NS_ID
mask bits are cleared since neither is meaningful. The capability
check is skipped for STATMOUNT_BY_FD since holding an fd already
implies prior access to the mount and equivalent information is
available through fstatfs() and /proc/pid/mountinfo without
privilege. Includes comprehensive selftests covering both attached
and detached mount cases.
- fs: Remove internal old mount API code (1 patch)
Now that every in-tree filesystem has been converted to the new
mount API, remove all the legacy shim code in fs_context.c that
handled unconverted filesystems. This deletes ~280 lines including
legacy_init_fs_context(), the legacy_fs_context struct, and
associated wrappers. The mount(2) syscall path for userspace remains
untouched. Documentation references to the legacy callbacks are
cleaned up.
- mount: add OPEN_TREE_NAMESPACE to open_tree()
Container runtimes currently use CLONE_NEWNS to copy the caller's
entire mount namespace — only to then pivot_root() and recursively
unmount everything they just copied. With large mount tables and
thousands of parallel container launches this creates significant
contention on the namespace semaphore.
OPEN_TREE_NAMESPACE copies only the specified mount tree (like
OPEN_TREE_CLONE) but returns a mount namespace fd instead of a
detached mount fd. The new namespace contains the copied tree mounted
on top of a clone of the real rootfs.
This functions as a combined unshare(CLONE_NEWNS) + pivot_root() in a
single syscall. Works with user namespaces: an unshare(CLONE_NEWUSER)
followed by OPEN_TREE_NAMESPACE creates a mount namespace owned by
the new user namespace. Mount namespace file mounts are excluded from
the copy to prevent cycles. Includes ~1000 lines of selftests"
* tag 'vfs-7.0-rc1.namespace' of git://git.kernel.org/pub/scm/linux/kernel/git/vfs/vfs:
selftests/open_tree: add OPEN_TREE_NAMESPACE tests
mount: add OPEN_TREE_NAMESPACE
fs: Remove internal old mount API code
selftests: statmount: tests for STATMOUNT_BY_FD
statmount: accept fd as a parameter
statmount: permission check should return EPERM
When creating containers the setup usually involves using CLONE_NEWNS
via clone3() or unshare(). This copies the caller's complete mount
namespace. The runtime will also assemble a new rootfs and then use
pivot_root() to switch the old mount tree with the new rootfs. Afterward
it will recursively umount the old mount tree thereby getting rid of all
mounts.
On a basic system here where the mount table isn't particularly large
this still copies about 30 mounts. Copying all of these mounts only to
get rid of them later is pretty wasteful.
This is exacerbated if intermediary mount namespaces are used that only
exist for a very short amount of time and are immediately destroyed
again causing a ton of mounts to be copied and destroyed needlessly.
With a large mount table and a system where thousands or ten-thousands
of containers are spawned in parallel this quickly becomes a bottleneck
increasing contention on the semaphore.
Extend open_tree() with a new OPEN_TREE_NAMESPACE flag. Similar to
OPEN_TREE_CLONE only the indicated mount tree is copied. Instead of
returning a file descriptor referring to that mount tree
OPEN_TREE_NAMESPACE will cause open_tree() to return a file descriptor
to a new mount namespace. In that new mount namespace the copied mount
tree has been mounted on top of a copy of the real rootfs.
The caller can setns() into that mount namespace and perform any
additionally required setup such as move_mount() detached mounts in
there.
This allows OPEN_TREE_NAMESPACE to function as a combined
unshare(CLONE_NEWNS) and pivot_root().
A caller may for example choose to create an extremely minimal rootfs:
fd_mntns = open_tree(-EBADF, "/var/lib/containers/wootwoot", OPEN_TREE_NAMESPACE);
This will create a mount namespace where "wootwoot" has become the
rootfs mounted on top of the real rootfs. The caller can now setns()
into this new mount namespace and assemble additional mounts.
This also works with user namespaces:
unshare(CLONE_NEWUSER);
fd_mntns = open_tree(-EBADF, "/var/lib/containers/wootwoot", OPEN_TREE_NAMESPACE);
which creates a new mount namespace owned by the earlier created user
namespace with "wootwoot" as the rootfs mounted on top of the real
rootfs.
Link: https://patch.msgid.link/20251229-work-empty-namespace-v1-1-bfb24c7b061f@kernel.org
Tested-by: Jeff Layton <jlayton@kernel.org>
Reviewed-by: Aleksa Sarai <cyphar@cyphar.com>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Suggested-by: Christian Brauner <brauner@kernel.org>
Suggested-by: Aleksa Sarai <cyphar@cyphar.com>
Signed-off-by: Christian Brauner <brauner@kernel.org>
Remove the "nullfs_rootfs" boot parameter and simply always use nullfs.
The mutable rootfs will be mounted on top of it. Systems that don't use
pivot_root() to pivot away from the real rootfs will have an additional
mount stick around but that shouldn't be a problem at all. If it is
we'll rever this commit.
This also simplifies the boot process and removes the need for the
traditional switch_root workarounds.
Suggested-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Christian Brauner <brauner@kernel.org>
Currently pivot_root() doesn't work on the real rootfs because it
cannot be unmounted. Userspace has to do a recursive removal of the
initramfs contents manually before continuing the boot.
Really all we want from the real rootfs is to serve as the parent mount
for anything that is actually useful such as the tmpfs or ramfs for
initramfs unpacking or the rootfs itself. There's no need for the real
rootfs to actually be anything meaningful or useful. Add a immutable
rootfs called "nullfs" that can be selected via the "nullfs_rootfs"
kernel command line option.
The kernel will mount a tmpfs/ramfs on top of it, unpack the initramfs
and fire up userspace which mounts the rootfs and can then just do:
chdir(rootfs);
pivot_root(".", ".");
umount2(".", MNT_DETACH);
and be done with it. (Ofc, userspace can also choose to retain the
initramfs contents by using something like pivot_root(".", "/initramfs")
without unmounting it.)
Technically this also means that the rootfs mount in unprivileged
namespaces doesn't need to become MNT_LOCKED anymore as it's guaranteed
that the immutable rootfs remains permanently empty so there cannot be
anything revealed by unmounting the covering mount.
In the future this will also allow us to create completely empty mount
namespaces without risking to leak anything.
systemd already handles this all correctly as it tries to pivot_root()
first and falls back to MS_MOVE only when that fails.
This goes back to various discussion in previous years and a LPC 2024
presentation about this very topic.
Link: https://patch.msgid.link/20260112-work-immutable-rootfs-v2-3-88dd1c34a204@kernel.org
Signed-off-by: Christian Brauner <brauner@kernel.org>
Christian Brauner
Linus Torvalds
Kees Cook
Qing WangandChristian Brauner
Al Viro