CVE-2026-99797: 
Linux Kernel vulnerability analysis and mitigation

Overview

CVE-2026-99797 is a deadlock vulnerability in the Linux kernel's io_uring subsystem, specifically in the SQPOLL (Submission Queue Poll) thread cancellation path. An exiting SQPOLL thread holds sqd->lock across cond_resched() and sleep operations in io_uring_cancel_generic(), which can block other threads indefinitely when zero-copy send notifications remain pending. The vulnerability affects Linux kernel versions starting from 6.0 and was disclosed on November 20, 2026. It carries a CVSS v3.1 base score of 5.5 (Medium), with availability impact only (GitHub Advisory, Red Hat).

Technical details

The root cause is classified as CWE-412 (Unrestricted Externally Accessible Lock): the SQPOLL thread holds sqd->lock while sleeping in io_uring_cancel_generic(), waiting for all in-flight I/O to complete. Because zero-copy send notifications cannot be canceled and remain pending until the peer reads data, the lock can be held indefinitely. Functions such as io_ring_exit_work(), IORING_REGISTER_IOWQ_AFF, and IOWQ_MAX_WORKERS all require this lock, so with 64 io_uring rings sharing a single SQPOLL thread, every exit slot blocks on the lock and no ring can be freed. The fix drops the lock around cond_resched() and the sleep — mirroring the behavior of the thread's main loop — while still holding the lock for each walk of ctx_list (GitHub Advisory).

Impact

A local user with low privileges who can create and exit multiple io_uring instances can trigger a deadlock that prevents the SQPOLL thread from terminating and prevents all associated rings from being freed, resulting in resource exhaustion and a system hang condition. The impact is limited to availability (denial of service); there is no confidentiality or integrity impact. In multi-tenant or shared-kernel environments, this could affect all users on the system (GitHub Advisory, Red Hat).

Exploitability

There is no public proof-of-concept exploit and no evidence of in-the-wild exploitation as of the disclosure date (GitHub Advisory). The EPSS score is 0.0, indicating very low probability of exploitation in the near term. The vulnerability is not listed in the CISA Known Exploited Vulnerabilities catalog. Exploitation requires local access and the ability to create io_uring instances with SQPOLL threads, limiting the attack surface to authenticated local users.

Exploitation steps

  1. Gain local access: Obtain a local user account on a vulnerable Linux system (kernel >= 6.0, unpatched).
  2. Create multiple io_uring instances with SQPOLL: Write a program that creates 64 or more io_uring rings configured with IORING_SETUP_SQPOLL, causing them to share a single SQPOLL thread.
  3. Initiate zero-copy send operations: Submit zero-copy send requests (e.g., using IORING_OP_SEND_ZC) to a socket whose peer does not read data, causing send notifications to remain pending and non-cancelable.
  4. Trigger SQPOLL thread exit: Begin exiting the io_uring instances, causing the SQPOLL thread to call io_uring_cancel_generic() while holding sqd->lock.
  5. Induce deadlock: With the lock held indefinitely (waiting for zero-copy notifications), all other rings attempting to exit block on sqd->lock, and io_ring_exit_work() cannot park the thread — resulting in resource exhaustion and system hang (GitHub Advisory).

Indicators of compromise

  • Logs: Kernel logs (dmesg) showing hung task warnings or RCU stall messages related to io_uring or SQPOLL threads; messages referencing io_uring_cancel_generic or io_ring_exit_work blocking.
  • Process: Zombie or unkillable processes associated with io_uring SQPOLL threads; ps or /proc showing threads stuck in uninterruptible sleep (D state) related to io_uring.
  • File System: Unusually high number of open io_uring file descriptors visible in /proc/<pid>/fd; /proc/sys/kernel/io_uring_disabled checks if io_uring has been administratively restricted.
  • Network: Sockets with zero-copy send operations pending indefinitely, observable via ss -e or netstat showing sockets in unusual states with no peer reads.

Mitigation and workarounds

Apply the upstream kernel patches that modify io_uring/cancel.c to drop sqd->lock around cond_resched() and sleep operations. Fixed versions include Linux kernel 6.12.118, 6.18.64, 7.3.5, and 7.4-rc1 and later. The specific fix commits are 9ebfcc60793a, 96ba960a6099, 682a8804640b, and cb0306c52ffa (GitHub Advisory). As a workaround where patching is not immediately possible, administrators can restrict io_uring access via sysctl kernel.io_uring_disabled=1 (disables io_uring system-wide) or use seccomp/LSM policies to limit which processes can invoke io_uring_setup with IORING_SETUP_SQPOLL.

Additional resources

Linux Distribution fix status

Fix availability across major Linux distributions and their releases.

Debian

Fixed

bookworm

linux

Affected

sid

linux: 7.3.5-1

Fixed

trixie

linux

Affected

Ubuntu

Unknown

bionic (esm-infra)

linux

Unknown

bionic (fips-updates)

linux-fips

Unknown

bionic (fips)

linux-fips

Unknown

devel

linux

Unknown

focal (esm-infra)

linux

Unknown

focal (fips-updates)

linux-fips

Unknown

focal (fips)

linux-fips

Unknown

jammy

linux

Unknown

RHEL / CentOS

Affected

OpenShift

openshift/ose-rhel-coreos-9

Affected

RHEL 8

Not Affected

RHEL 9

kernel.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