CVE-2026-99255: 
Linux Kernel vulnerability analysis and mitigation

Overview

CVE-2026-99255 is a race condition vulnerability in the Linux kernel's io_uring/io-wq subsystem, specifically in the handling of linked request cancellations during task exit. When an io-wq worker completes a request that has a linked follow-up request, there is a brief window during state transition where the linked request is invisible to the cancellation logic in io_uring_cancel_generic(). This can cause an exiting task to sleep indefinitely, becoming unkillable. The vulnerability affects Linux kernel versions from 5.5 onward, with fixes available in stable releases 6.1.191, 6.6.161, 6.12.115, 6.18.60, 7.2.14, 7.3.2, and 7.4-rc1. It carries a CVSS v3.1 base score of 4.7 (Medium) (GitHub Advisory, Microsoft MSRC).

Technical details

The root cause is a time-of-check/time-of-use (TOCTOU) race condition (related to CWE-362: Concurrent Execution Using Shared Resource with Improper Synchronization) in io_worker_handle_work(). When io_wq_free_work() calls io_req_find_next() to extract the linked request and then io_free_req() to free the current request, the linked request enters a transient state: it is no longer attached to the completed request, not yet assigned as worker->cur_work, and not yet on any parking list. If io_uring_cancel_generic() scans for cancellable requests during this exact window, it finds nothing and puts the task to sleep. The worker then proceeds to execute the linked request — which IO_WQ_BIT_EXIT does not prevent for bounded work — potentially parking it without triggering any wakeup. The fix splits io_wq_free_work() into two phases, deferring the freeing of the original request until the linked request is safely assigned as current work or placed in a pending queue (GitHub Advisory).

Impact

Successful exploitation results in a denial of service: a local attacker can cause a task (or entire process) to hang indefinitely and become unkillable, as the exiting task sleeps waiting for a wakeup that never arrives. There is no confidentiality or integrity impact — the vulnerability is purely an availability issue. In the worst case, if the affected task is the main process thread, the process cannot be terminated even with SIGKILL, requiring a system reboot to recover (GitHub Advisory, Feedly).

Exploitability

There is no public proof-of-concept exploit and no evidence of in-the-wild exploitation as of the time of reporting. The EPSS score is approximately 0.197% (9th percentile), indicating a low near-term exploitation probability. Exploitation requires local access with low privileges and high attack complexity due to the narrow race condition window — the original reporter noted that with four linked requests per round, only 5–8 of 10,000 rounds triggered the hang on a 4-CPU KVM system. The vulnerability is not listed in the CISA Known Exploited Vulnerabilities catalog (GitHub Advisory, Microsoft MSRC).

Exploitation steps

  1. Gain local access: Obtain a low-privileged local user account on a vulnerable Linux system running kernel version 5.5 or later without the patch applied.
  2. Craft io_uring linked requests: Construct an io_uring submission queue with linked operations — for example, a NOP request with IOSQE_ASYNC | IOSQE_IO_LINK followed by a FUTEX_WAIT with FUTEX2_PRIVATE — designed to create a linked chain that will be processed by an io-wq worker.
  3. Trigger task exit during the race window: Submit the linked requests and then exit the thread or process (within approximately 63 microseconds of submission) to trigger io_uring_cancel_generic() during the brief window when the linked request is in transition between worker states.
  4. Induce task hang: If the race condition is won, the cancellation pass finds no pending request, the task sleeps, the worker executes and parks the linked request without completing or freeing it, and no wakeup is issued — leaving the task permanently hung and unkillable.
  5. Repeat for reliability: Due to the narrow race window, repeat the sequence multiple times (e.g., thousands of rounds) to reliably trigger the hang (GitHub Advisory).

Indicators of compromise

  • Process: Zombie or unkillable processes that do not respond to SIGKILL; processes stuck in D (uninterruptible sleep) state indefinitely as shown by ps aux or top.
  • Logs: Kernel logs (dmesg) may show io_uring-related warnings or hung task messages (e.g., INFO: task <name>:<pid> blocked for more than 120 seconds) associated with io_uring/io-wq subsystem.
  • System Behavior: System requiring reboot to clear hung processes; repeated failed attempts to kill a specific process via standard signals.

Mitigation and workarounds

Apply the upstream Linux kernel patches that resolve the race condition by splitting io_wq_free_work() into two phases. Fixed versions include: 6.1.191, 6.6.161, 6.12.115, 6.18.60, 7.2.14, 7.3.2, and 7.4-rc1. Specific stable-tree commits are available at the kernel.org stable repository (e.g., 98470f164286, d39ebc7a583d, ebe8e97fb99d, 3e02b8df0e9f, 92e20e9c72bd, 59e45f44c3ce, 436d48a2ed13). Microsoft's Azure Linux 3 kernel package azl3_kernel_6.6.159.1-1 is listed as affected; users should update to a patched kernel package. Amazon Linux 2023 users should apply the update referenced in ALAS2023-2026-3448. No configuration-based workaround is available; patching is the only remediation (GitHub Advisory, Microsoft MSRC).

Community reactions

The vulnerability was disclosed as part of the Linux kernel CVE announcement process via the kernel mailing list and was included in Microsoft's November 2026 Patch Tuesday, which addressed 1,235 flaws in total. Coverage by BleepingComputer noted the breadth of the November 2026 Patch Tuesday release. No notable individual researcher commentary or significant social media discussion specific to this CVE has been observed, consistent with its medium severity and low exploitability rating.

Additional resources

Linux Distribution fix status

Fix availability across major Linux distributions and their releases.

Debian

Fixed

bookworm

linux-6.12: 6.12.115-2~deb12u1

Fixed

sid

linux: 7.2.14-1

Fixed

trixie

linux: 6.12.115-1

Fixed

Ubuntu

Fixed

bionic

linux

Not Affected

bionic (esm-infra)

linux

Not Affected

bionic (fips-updates)

linux-fips

Not Affected

bionic (fips)

linux-fips

Not Affected

devel

linux

Affected

focal

linux

Not Affected

focal (esm-infra)

linux-hwe-5.15

Affected

focal (fips-updates)

linux-fips

Not Affected

RHEL / CentOS

Affected

OpenShift

openshift/ose-rhel-coreos-9

Affected

RHEL 8

Not Affected

RHEL 9

kernel-rt.src

Affected

RHEL 10

kernel.src

Affected

Source: This report was generated using AI

Related Linux Kernel vulnerabilities:

CVE ID

Severity

Score

Technologies

Component name

CISA KEV exploit

Has fix

Published date

CVE-2026-99804MEDIUM5.1
  • Linux Kernel logoLinux Kernel
  • linux-aws-6.14
NoYesNov 20, 2026
CVE-2026-99799MEDIUM5.1
  • Linux Kernel logoLinux Kernel
  • linux-hwe-5.4
NoYesNov 20, 2026
CVE-2026-99797MEDIUM5.1
  • Linux Kernel logoLinux Kernel
  • linux-oracle
NoYesNov 20, 2026
CVE-2026-99796MEDIUM5.1
  • Linux Kernel logoLinux Kernel
  • kernel-kmap-internal
NoYesNov 20, 2026
CVE-2026-99798MEDIUM4.3
  • Linux Kernel logoLinux Kernel
  • kernel-64k-devel
NoYesNov 20, 2026

Free Vulnerability Assessment

Benchmark your Cloud Security Posture

Evaluate your cloud security practices across 9 security domains to benchmark your risk level and identify gaps in your defenses.

Request assessment

Additional Wiz resources

Get a personalized demo

Ready to see Wiz in action?

"Best User Experience I have ever seen, provides full visibility to cloud workloads."
David EstlickCISO
"Wiz provides a single pane of glass to see what is going on in our cloud environments."
Adam FletcherChief Security Officer
"We know that if Wiz identifies something as critical, it actually is."
Greg PoniatowskiHead of Threat and Vulnerability Management