CVE-2026-98981 (GCVE-0-2026-98981)
Vulnerability from cvelistv5
Published
2026-10-23 12:19
Modified
2026-10-30 12:40
Severity ?
VLAI Severity ?
EPSS score ?
Summary
In the Linux kernel, the following vulnerability has been resolved:
namespace: unhash a dentry before detaching the mounts on it
unlink(), rmdir() and rename() remove the entry from the filesystem,
call detach_mounts() on the entry with the inode locked and then call
d_delete() once the inode is unlocked.
The thing is that between detach_mounts() and d_delete() the dentry is
still hashed and positive. But after detach_mounts() nothing covers the
dentry anymore. A lookup that finds this dentry in the dcache can
uncover what the mounts hid.
That's a problem when unlinking files or directories that are
mountpoints in other mount namespaces. Everybody who had the underlying
entry covered can race the detach_mount() call until d_delete() has run.
The race window isn't all that small. It encompasses namespace_unlock()
with a full synchronize_rcu_expedited() grace period and the
inode_unlock() of the entry.
In my experiments three walkers that kept trying read an overmounted
file in 11938 of the 16140 unlinks that removed its mountpoint from a
bind mount of the filesystem within a minute.
Fun fact, d_invalidate() has the same ordering problem but gets it
right. It unhashes the dentry first and detaches the mounts afterwards.
Let's do the same in __detach_mounts():
- A lookup that hasn't found the dentry yet misses it in the dcache and
waits for the directory lock that the caller holds until the name is
gone for good.
- A lockless lookup that found it already fails the mount_lock check
while the dentry still counts as a mountpoint and the d_seq check once
it doesn't, and retries.
This fixes the lockless path. We still need to fix the reference count
lookup in a follow-up patch.
So d_drop() the dentry. The dentry stays positive and held but can't be
found anymore. Then proceed with the detach and unlink.
Reproducer:
The reads were counted with the program below, three walkers for 60 s
in a VM with 4 CPUs. The race window is stretched with the debug patch
pasted here. fs.detach_race_walk_us sleeps in lookup_fast() once it
found a mountpoint dentry and before its mounts are crossed.
fs.detach_race_unlink_us sleeps in vfs_unlink() between detach_mounts()
and d_delete().
// SPDX-License-Identifier: GPL-2.0
/*
* unlink_covered: unlink a file that is a mountpoint in a detached copy of
* its mount, against lookups that open it through that copy.
*
* A is a tmpfs with the file f1 (content SECRET_F) and the plain file
* MARK_A. A' is an open_tree(OPEN_TREE_CLONE) copy of A with MARK_A bound
* on A'/f1, held through an O_PATH fd on its root once the tree fd is
* closed. Walkers open f1 through that fd while the driver unlinks A/f1,
* where nothing is mounted on it. A walker may read MARK_A or get ENOENT.
* A read of SECRET_F is a hit: the name was found after its mount was
* gone. The lockless walks use openat2(RESOLVE_CACHED).
*
* usage: unlink_covered [-t seconds] [-w walkers]
*/
#ifndef _GNU_SOURCE
#define _GNU_SOURCE
#endif
#include <errno.h>
#include <fcntl.h>
#include <pthread.h>
#include <stdatomic.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <unistd.h>
#include <sys/mount.h>
#include <sys/stat.h>
#include <sys/syscall.h>
#include <linux/openat2.h>
#ifndef __NR_open_tree
#define __NR_open_tree 428
#endif
#ifndef __NR_move_mount
#define __NR_move_mount 429
#endif
#ifndef __NR_openat2
#define __NR_openat2 437
#endif
#ifndef OPEN_TREE_CLONE
#define OPEN_TREE_CLONE 1
#endif
#ifndef OPEN_TREE_CLOEXEC
#define OPEN_TREE_CLOEXEC O_CLOEXEC
#endif
#ifndef MOVE_MOUNT_F_EMPTY_PATH
#define MOVE_MOUNT_F_EMPTY_PATH 0x00000004
#endif
#define WORK "/tmp/uc"
#define ADIR WORK "/A"
static int duration = 60, nwalkers = 3;
static atomic_int stop, writer_waiting;
static pthread_rwlock_t cur_lock = PTHREAD_RWLOCK_INITIALIZER;
static int cur_fd = -1; /* the root of A', -1 wh
---truncated---
References
Impacted products
| Vendor | Product | Version | ||
|---|---|---|---|---|
| Linux | Linux |
Version: 8ed936b5671bfb33d89bc60bdcc7cf0470ba52fe Version: 8ed936b5671bfb33d89bc60bdcc7cf0470ba52fe Version: 8ed936b5671bfb33d89bc60bdcc7cf0470ba52fe Version: 8ed936b5671bfb33d89bc60bdcc7cf0470ba52fe Version: 8ed936b5671bfb33d89bc60bdcc7cf0470ba52fe Version: 8ed936b5671bfb33d89bc60bdcc7cf0470ba52fe Version: 8ed936b5671bfb33d89bc60bdcc7cf0470ba52fe Version: 8ed936b5671bfb33d89bc60bdcc7cf0470ba52fe |
||
{
"containers": {
"cna": {
"affected": [
{
"defaultStatus": "unaffected",
"product": "Linux",
"programFiles": [
"fs/namespace.c"
],
"repo": "https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git",
"vendor": "Linux",
"versions": [
{
"lessThan": "b3fa3811ffffa130586c4ff8f7299b7fc69216d5",
"status": "affected",
"version": "8ed936b5671bfb33d89bc60bdcc7cf0470ba52fe",
"versionType": "git"
},
{
"lessThan": "97cc290c57ea04337fd33912d780367add41525c",
"status": "affected",
"version": "8ed936b5671bfb33d89bc60bdcc7cf0470ba52fe",
"versionType": "git"
},
{
"lessThan": "bb917ec497f6099c6cb9375596e085a5f6b52ff2",
"status": "affected",
"version": "8ed936b5671bfb33d89bc60bdcc7cf0470ba52fe",
"versionType": "git"
},
{
"lessThan": "db90eecbde319527aa9480c039d32a0f5fe80574",
"status": "affected",
"version": "8ed936b5671bfb33d89bc60bdcc7cf0470ba52fe",
"versionType": "git"
},
{
"lessThan": "8d71bfc9b3fc15d7103e82ea9dc5057541866e46",
"status": "affected",
"version": "8ed936b5671bfb33d89bc60bdcc7cf0470ba52fe",
"versionType": "git"
},
{
"lessThan": "d29cd99db3db74cf3447be74d25740981789d461",
"status": "affected",
"version": "8ed936b5671bfb33d89bc60bdcc7cf0470ba52fe",
"versionType": "git"
},
{
"lessThan": "8300de4aa5d09061bfcf2af5b52c0cdba3873185",
"status": "affected",
"version": "8ed936b5671bfb33d89bc60bdcc7cf0470ba52fe",
"versionType": "git"
},
{
"lessThan": "528b383dc35d1e5caf64ddf5d35f983ae543b0d0",
"status": "affected",
"version": "8ed936b5671bfb33d89bc60bdcc7cf0470ba52fe",
"versionType": "git"
}
]
},
{
"defaultStatus": "affected",
"product": "Linux",
"programFiles": [
"fs/namespace.c"
],
"repo": "https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git",
"vendor": "Linux",
"versions": [
{
"status": "affected",
"version": "3.18"
},
{
"lessThan": "3.18",
"status": "unaffected",
"version": "0",
"versionType": "semver"
},
{
"lessThanOrEqual": "5.10.*",
"status": "unaffected",
"version": "5.10.273",
"versionType": "semver"
},
{
"lessThanOrEqual": "5.15.*",
"status": "unaffected",
"version": "5.15.224",
"versionType": "semver"
},
{
"lessThanOrEqual": "6.1.*",
"status": "unaffected",
"version": "6.1.191",
"versionType": "semver"
},
{
"lessThanOrEqual": "6.6.*",
"status": "unaffected",
"version": "6.6.160",
"versionType": "semver"
},
{
"lessThanOrEqual": "6.12.*",
"status": "unaffected",
"version": "6.12.114",
"versionType": "semver"
},
{
"lessThanOrEqual": "6.18.*",
"status": "unaffected",
"version": "6.18.57",
"versionType": "semver"
},
{
"lessThanOrEqual": "7.2.*",
"status": "unaffected",
"version": "7.2.11",
"versionType": "semver"
},
{
"lessThanOrEqual": "*",
"status": "unaffected",
"version": "7.3",
"versionType": "original_commit_for_fix"
}
]
}
],
"cpeApplicability": [
{
"nodes": [
{
"cpeMatch": [
{
"criteria": "cpe:2.3:o:linux:linux_kernel:*:*:*:*:*:*:*:*",
"versionEndExcluding": "5.10.273",
"versionStartIncluding": "3.18",
"vulnerable": true
},
{
"criteria": "cpe:2.3:o:linux:linux_kernel:*:*:*:*:*:*:*:*",
"versionEndExcluding": "5.15.224",
"versionStartIncluding": "3.18",
"vulnerable": true
},
{
"criteria": "cpe:2.3:o:linux:linux_kernel:*:*:*:*:*:*:*:*",
"versionEndExcluding": "6.1.191",
"versionStartIncluding": "3.18",
"vulnerable": true
},
{
"criteria": "cpe:2.3:o:linux:linux_kernel:*:*:*:*:*:*:*:*",
"versionEndExcluding": "6.6.160",
"versionStartIncluding": "3.18",
"vulnerable": true
},
{
"criteria": "cpe:2.3:o:linux:linux_kernel:*:*:*:*:*:*:*:*",
"versionEndExcluding": "6.12.114",
"versionStartIncluding": "3.18",
"vulnerable": true
},
{
"criteria": "cpe:2.3:o:linux:linux_kernel:*:*:*:*:*:*:*:*",
"versionEndExcluding": "6.18.57",
"versionStartIncluding": "3.18",
"vulnerable": true
},
{
"criteria": "cpe:2.3:o:linux:linux_kernel:*:*:*:*:*:*:*:*",
"versionEndExcluding": "7.2.11",
"versionStartIncluding": "3.18",
"vulnerable": true
},
{
"criteria": "cpe:2.3:o:linux:linux_kernel:*:*:*:*:*:*:*:*",
"versionEndExcluding": "7.3",
"versionStartIncluding": "3.18",
"vulnerable": true
}
],
"negate": false,
"operator": "OR"
}
]
}
],
"descriptions": [
{
"lang": "en",
"value": "In the Linux kernel, the following vulnerability has been resolved:\n\nnamespace: unhash a dentry before detaching the mounts on it\n\nunlink(), rmdir() and rename() remove the entry from the filesystem,\ncall detach_mounts() on the entry with the inode locked and then call\nd_delete() once the inode is unlocked.\n\nThe thing is that between detach_mounts() and d_delete() the dentry is\nstill hashed and positive. But after detach_mounts() nothing covers the\ndentry anymore. A lookup that finds this dentry in the dcache can\nuncover what the mounts hid.\n\nThat\u0027s a problem when unlinking files or directories that are\nmountpoints in other mount namespaces. Everybody who had the underlying\nentry covered can race the detach_mount() call until d_delete() has run.\n\nThe race window isn\u0027t all that small. It encompasses namespace_unlock()\nwith a full synchronize_rcu_expedited() grace period and the\ninode_unlock() of the entry.\n\nIn my experiments three walkers that kept trying read an overmounted\nfile in 11938 of the 16140 unlinks that removed its mountpoint from a\nbind mount of the filesystem within a minute.\n\nFun fact, d_invalidate() has the same ordering problem but gets it\nright. It unhashes the dentry first and detaches the mounts afterwards.\nLet\u0027s do the same in __detach_mounts():\n\n- A lookup that hasn\u0027t found the dentry yet misses it in the dcache and\n waits for the directory lock that the caller holds until the name is\n gone for good.\n\n- A lockless lookup that found it already fails the mount_lock check\n while the dentry still counts as a mountpoint and the d_seq check once\n it doesn\u0027t, and retries.\n\nThis fixes the lockless path. We still need to fix the reference count\nlookup in a follow-up patch.\n\nSo d_drop() the dentry. The dentry stays positive and held but can\u0027t be\nfound anymore. Then proceed with the detach and unlink.\n\nReproducer:\n\nThe reads were counted with the program below, three walkers for 60 s\nin a VM with 4 CPUs. The race window is stretched with the debug patch\npasted here. fs.detach_race_walk_us sleeps in lookup_fast() once it\nfound a mountpoint dentry and before its mounts are crossed.\nfs.detach_race_unlink_us sleeps in vfs_unlink() between detach_mounts()\nand d_delete().\n\n // SPDX-License-Identifier: GPL-2.0\n /*\n * unlink_covered: unlink a file that is a mountpoint in a detached copy of\n * its mount, against lookups that open it through that copy.\n *\n * A is a tmpfs with the file f1 (content SECRET_F) and the plain file\n * MARK_A. A\u0027 is an open_tree(OPEN_TREE_CLONE) copy of A with MARK_A bound\n * on A\u0027/f1, held through an O_PATH fd on its root once the tree fd is\n * closed. Walkers open f1 through that fd while the driver unlinks A/f1,\n * where nothing is mounted on it. A walker may read MARK_A or get ENOENT.\n * A read of SECRET_F is a hit: the name was found after its mount was\n * gone. The lockless walks use openat2(RESOLVE_CACHED).\n *\n * usage: unlink_covered [-t seconds] [-w walkers]\n */\n #ifndef _GNU_SOURCE\n #define _GNU_SOURCE\n #endif\n #include \u003cerrno.h\u003e\n #include \u003cfcntl.h\u003e\n #include \u003cpthread.h\u003e\n #include \u003cstdatomic.h\u003e\n #include \u003cstdio.h\u003e\n #include \u003cstdlib.h\u003e\n #include \u003cstring.h\u003e\n #include \u003cunistd.h\u003e\n #include \u003csys/mount.h\u003e\n #include \u003csys/stat.h\u003e\n #include \u003csys/syscall.h\u003e\n #include \u003clinux/openat2.h\u003e\n\n #ifndef __NR_open_tree\n #define __NR_open_tree 428\n #endif\n #ifndef __NR_move_mount\n #define __NR_move_mount 429\n #endif\n #ifndef __NR_openat2\n #define __NR_openat2 437\n #endif\n #ifndef OPEN_TREE_CLONE\n #define OPEN_TREE_CLONE 1\n #endif\n #ifndef OPEN_TREE_CLOEXEC\n #define OPEN_TREE_CLOEXEC O_CLOEXEC\n #endif\n #ifndef MOVE_MOUNT_F_EMPTY_PATH\n #define MOVE_MOUNT_F_EMPTY_PATH 0x00000004\n #endif\n\n #define WORK \"/tmp/uc\"\n #define ADIR WORK \"/A\"\n\n static int duration = 60, nwalkers = 3;\n static atomic_int stop, writer_waiting;\n static pthread_rwlock_t cur_lock = PTHREAD_RWLOCK_INITIALIZER;\n static int cur_fd = -1;\t\t/* the root of A\u0027, -1 wh\n---truncated---"
}
],
"metrics": [
{
"cvssV3_1": {
"baseScore": 8.8,
"baseSeverity": "HIGH",
"vectorString": "CVSS:3.1/AV:L/AC:L/PR:L/UI:N/S:C/C:H/I:H/A:H",
"version": "3.1"
},
"scenarios": [
{
"lang": "en",
"value": "AV:L - The trigger uses local unlinkat(), rmdir(), or renameat2() into vfs_unlink(), vfs_rmdir(), or vfs_rename(), which call the patched __detach_mounts(), while a concurrent openat2() path walk races the detach; no network data reaches this path.\nAC:L - The attacker controls both sides by unlinking the underlying entry while concurrent lookup_fast() walkers open it through the covered mount; the reproducer obtained 11,938 disclosures in 16,140 unlinks.\nPR:L - may_delete_dentry() requires ordinary write and execute permission on the parent, while may_mount() uses CAP_SYS_ADMIN in the mount namespace\u0027s user namespace; an unprivileged user can obtain that capability with a new user/mount namespace and mount tmpfs because it has FS_USERNS_MOUNT.\nUI:N - The attacker creates the mount arrangement and directly drives both the removal syscall and concurrent openat2() calls; no separate victim action is required.\nS:C - __detach_mounts() removes mounts attached to the shared dentry across namespaces, allowing a walk in the covered A\u0027 namespace to reach the underlying A object. This crosses the mount-namespace isolation boundary the covering mount enforced.\nC:H - The stale positive dentry accepted by lookup_fast() and step_into() can be opened and read after its covering mount disappears; the reproducer reads the complete SECRET_F contents rather than a bounded metadata fragment.\nI:H - The incorrectly uncovered dentry proceeds to do_open(), where may_open() permits O_WRONLY under normal inode permissions and O_TRUNC reaches handle_truncate(); thus a mount-hidden writable object, or the target of an uncovered symlink, can be modified or erased.\nA:H - A stale covered symlink can reach pick_link() and resolve a persistent target, after which do_open() can invoke handle_truncate() for O_TRUNC. Erasing a mount-hidden critical file can cause complete loss of the dependent resource or service."
}
]
}
],
"providerMetadata": {
"dateUpdated": "2026-10-30T12:40:41.226Z",
"orgId": "416baaa9-dc9f-4396-8d5f-8c081fb06d67",
"shortName": "Linux"
},
"references": [
{
"url": "https://git.kernel.org/stable/c/b3fa3811ffffa130586c4ff8f7299b7fc69216d5"
},
{
"url": "https://git.kernel.org/stable/c/97cc290c57ea04337fd33912d780367add41525c"
},
{
"url": "https://git.kernel.org/stable/c/bb917ec497f6099c6cb9375596e085a5f6b52ff2"
},
{
"url": "https://git.kernel.org/stable/c/db90eecbde319527aa9480c039d32a0f5fe80574"
},
{
"url": "https://git.kernel.org/stable/c/8d71bfc9b3fc15d7103e82ea9dc5057541866e46"
},
{
"url": "https://git.kernel.org/stable/c/d29cd99db3db74cf3447be74d25740981789d461"
},
{
"url": "https://git.kernel.org/stable/c/8300de4aa5d09061bfcf2af5b52c0cdba3873185"
},
{
"url": "https://git.kernel.org/stable/c/528b383dc35d1e5caf64ddf5d35f983ae543b0d0"
}
],
"title": "namespace: unhash a dentry before detaching the mounts on it",
"x_generator": {
"engine": "bippy-1.2.0"
}
}
},
"cveMetadata": {
"assignerOrgId": "416baaa9-dc9f-4396-8d5f-8c081fb06d67",
"assignerShortName": "Linux",
"cveId": "CVE-2026-98981",
"datePublished": "2026-10-23T12:19:58.507Z",
"dateReserved": "2026-09-25T10:25:14.384Z",
"dateUpdated": "2026-10-30T12:40:41.226Z",
"state": "PUBLISHED"
},
"dataType": "CVE_RECORD",
"dataVersion": "5.2"
}
Loading…
Loading…
Sightings
| Author | Source | Type | Date |
|---|
Nomenclature
- Seen: The vulnerability was mentioned, discussed, or seen somewhere by the user.
- Confirmed: The vulnerability is confirmed from an analyst perspective.
- Published Proof of Concept: A public proof of concept is available for this vulnerability.
- Exploited: This vulnerability was exploited and seen by the user reporting the sighting.
- Patched: This vulnerability was successfully patched by the user reporting the sighting.
- Not exploited: This vulnerability was not exploited or seen by the user reporting the sighting.
- Not confirmed: The user expresses doubt about the veracity of the vulnerability.
- Not patched: This vulnerability was not successfully patched by the user reporting the sighting.
Loading…
Loading…