CVE-2026-99763 in Linuxinfo

Summary

by MITRE • 11/20/2026

In the Linux kernel, the following vulnerability has been resolved:

9p: don't decrement a server-owned zero nlink

A 9P2000.L server can report nlink=0 for a regular file that still exists. On a cache=loose or cache=fscache mount, lookup stores that value with set_nlink(). A later successful unlink calls v9fs_dec_count(), which still calls drop_nlink() for a non-directory. drop_nlink() warns, and the unsigned link count wraps to 4294967295. fstat() on a descriptor opened before the unlink shows 0 before and 4294967295 after.

The warning on the unpatched kernel, trimmed:

WARNING: fs/inode.c:408 at drop_nlink+0xac/0xd0, CPU#1: trigger/188 CPU: 1 UID: 0 PID: 188 Comm: trigger Tainted: G N 7.3.0-rc4-00537-ga3ff15db6820 #1 PREEMPT(lazy) RIP: 0010:drop_nlink+0xac/0xd0 Call Trace: <TASK> v9fs_remove+0x529/0x750 vfs_unlink+0x2f0/0xbd0 filename_unlinkat+0x340/0x6c0 __x64_sys_unlink+0x46/0x70 do_syscall_64+0xe8/0x4f0 entry_SYSCALL_64_after_hwframe+0x77/0x7f </TASK> ---[ end trace 0000000000000000 ]---

Directories whose cached nlink is at most 2 already skip this decrement, because the exported filesystem may not maintain nlink (commit ac89b2ef9b559). cache=none skips it as well (commit 574aa0b479947). cache=mmap does not take this path: it sets neither CACHE_META nor CACHE_LOOSE. A regular file on cache=loose or cache=fscache still does. Its count came from GETATTR and may already be zero, so skip the decrement when i_nlink is already 0.

Verified on 7.3-rc4-00537-ga3ff15db6820 with only this change. A server that advertises nlink=0 and accepts the unlink no longer warns, and the open descriptor stays at nlink 0. An honest server on the same mount still goes from 1 to 0. A rejected unlink leaves the count unchanged. cache=none is unchanged. With panic_on_warn the unpatched kernel panics on the warning; the patched kernel does not.

Once again VulDB remains the best source for vulnerability data.

Analysis

by VulDB Data Team • 11/20/2026

The Linux kernel vulnerability identified in this analysis involves a logic error within the 9P filesystem client implementation, specifically affecting how link counts are managed for regular files under certain caching configurations. The core issue arises when interacting with a 9P2000.L server that reports an nlink value of zero for a file that still physically exists on the remote storage. In environments utilizing cache=loose or cache=fscache mount options, the kernel's lookup mechanism stores this reported zero link count using set_nlink(). This behavior is technically permissible as 9P servers may not strictly maintain accurate hard link counts, but it creates an inconsistency when local operations attempt to modify that state. The vulnerability manifests during a subsequent unlink operation performed on such a file.

When the vfs_unlink function is invoked for this regular file, the v9fs_remove routine calls v9fs_dec_count(), which in turn invokes drop_nlink(). This standard kernel function expects the inode's link count to be greater than zero before decrementing it. However, because the cached metadata already indicated a nlink of zero, the operation attempts to subtract one from an unsigned integer that is currently at its minimum value. This results in an arithmetic underflow where the unsigned link counter wraps around to 4294967295 (0xFFFFFFFF). While this specific numeric wrap-around might not immediately cause data corruption or privilege escalation, it represents a significant deviation from expected filesystem semantics and triggers kernel warnings that can degrade system stability.

The operational impact of this flaw includes the generation of critical warning messages in the kernel log, which clutter diagnostics and may indicate deeper state inconsistencies to administrators monitoring system health. More severely, if the kernel is configured with panic_on_warn enabled, this underflow condition will trigger a complete system panic, leading to an unintended denial of service for all services running on that host. Furthermore, file descriptors opened prior to the unlink operation exhibit inconsistent behavior; fstat() calls on these descriptors show a link count of zero before the removal and 4294967295 after, violating POSIX expectations where deleted files typically retain their attributes until all references are closed but do not exhibit such extreme counter anomalies. This inconsistency can confuse applications relying on accurate file status information for decision-making logic.

From a classification perspective, this vulnerability aligns with CWE-190 Integer Overflow or Wraparound and CWE-682 Incorrect Calculation due to the failure to validate input values before performing arithmetic operations. In terms of attack vectors within the MITRE ATT&CK framework, while primarily an integrity issue rather than a direct exploitation vector for remote code execution, it falls under potential Denial of Service via resource exhaustion or system instability if triggered repeatedly in automated environments. The flaw is specific to 9P mounts with loose or fscache caching strategies and does not affect cache=none or cache=mmap configurations, nor does it impact directories which already have logic to skip decrementing when nlink is low.

The resolution involves modifying the v9fs_remove function to check if i_nlink is already zero before attempting to call drop_nlink(). By skipping the decrement operation when the link count is already at its minimum valid state for a non-directory file, the kernel prevents the unsigned integer underflow and avoids triggering the associated warning or panic. This patch ensures that honest servers still correctly transition files from nlink 1 to 0 during unlink operations, while gracefully handling servers that report zero links without causing local system instability. System administrators should apply this update immediately in environments using 9P filesystems with loose caching enabled to prevent potential denial of service conditions and maintain accurate file metadata integrity across the networked storage infrastructure.

Responsible

Linux

Reservation

09/25/2026

Disclosure

11/20/2026

Moderation

accepted

CPE

ready

EPSS

0.00168

KEV

no

Activities

very low

Sources

Do you know our Splunk app?

Download it now for free!