CVE-2026-99796: 
Linux Kernel vulnerability analysis and mitigation

Overview

CVE-2026-99796 is a Denial of Service vulnerability in the Linux kernel's io_uring/poll subsystem caused by inefficient async poll cancellation logic. When IORING_ASYNC_CANCEL_ALL is used, io_poll_cancel() cancels only one request per call and restarts each lookup from the beginning of the hash table, resulting in O(K×N/2) complexity when cancelling K of N pending polls. This allows a local attacker to hold the uring_lock for extended periods, triggering soft lockup watchdog warnings and freezing CPU execution. The vulnerability was published on November 20, 2026, and affects Linux kernel versions from 5.19 up to the fixed releases. The CVSS estimate is Medium severity (Red Hat Bugzilla, Red Hat CVE).

Technical details

The root cause is classified as CWE-1050 (Excessive Platform Resource Consumption within a Loop). In the vulnerable code path, __io_async_cancel() repeatedly calls io_try_cancel() → io_poll_cancel(), which cancels one request per invocation and restarts the hash table walk from the beginning each time. Because cancelled requests remain in the hash until their task work runs, each subsequent lookup traverses all previously cancelled entries, producing quadratic time complexity. A concrete example: cancelling 80,000 polls on a single eventfd via CANCEL_FD|ALL takes 17.06 seconds with the uring_lock held, versus 0.02 seconds after the fix. The kernel watchdog fires a soft lockup after ~26 seconds, as shown in the stack trace: io_cancel_req_match → io_poll_cancel → io_try_cancel → __io_async_cancel (Red Hat Bugzilla).

Impact

Successful exploitation causes a CPU soft lockup and effective denial of service on the affected system, freezing kernel execution for the duration of the cancellation loop. The impact is limited to availability — there is no evidence of confidentiality or integrity compromise. Any local user with permission to invoke io_uring operations can trigger this condition, making multi-tenant Linux environments (e.g., container hosts, shared servers) particularly at risk (Red Hat CVE, Red Hat Bugzilla).

Exploitability

There is no public proof-of-concept exploit and no evidence of in-the-wild exploitation as of the disclosure date. The vulnerability requires local access to the system and the ability to invoke io_uring operations with IORING_ASYNC_CANCEL_ALL on a large number of pending polls. No threat actor attribution has been reported. The EPSS score is 0.0, and the vulnerability is not listed in the CISA Known Exploited Vulnerabilities catalog (Red Hat CVE).

Exploitation steps

  1. Gain local access: Obtain a local user account on a Linux system running a vulnerable kernel (5.19 through versions prior to 6.12.118, 6.18.64, 7.3.5, or 7.4-rc2).
  2. Create a large number of io_uring poll requests: Use the io_uring API to register approximately 80,000 or more poll operations targeting a single file descriptor (e.g., an eventfd), building up a large hash table of pending poll requests.
  3. Issue a bulk cancellation: Submit an IORING_OP_ASYNC_CANCEL operation with the IORING_ASYNC_CANCEL_ALL and IORING_ASYNC_CANCEL_FD flags set, targeting the eventfd used in step 2.
  4. Trigger the soft lockup: The kernel enters the quadratic cancellation loop, holding uring_lock while walking the hash table O(K×N/2) times. With 80,000 polls, this takes ~17 seconds, causing the soft lockup watchdog to fire after ~26 seconds and effectively freezing the CPU (Red Hat Bugzilla).

Indicators of compromise

  • Logs: Kernel log (dmesg / /var/log/kern.log) entries containing watchdog: BUG: soft lockup - CPU#N stuck for 26s! with a call trace referencing io_cancel_req_match, io_poll_cancel, io_try_cancel, and __io_async_cancel.
  • Process Behavior: A process repeatedly calling io_uring_enter() with cancellation operations that hold the CPU for abnormally long periods; high system CPU usage attributed to a single process using io_uring.
  • Kernel Trace: Stack traces in crash dumps or live kernel traces showing the call chain __io_async_cancel+0x41/0x110 → io_try_cancel+0x7b/0xf0 → io_poll_cancel+0x73/0x170 → io_cancel_req_match+0x19/0xa0 (Red Hat Bugzilla).

Mitigation and workarounds

Update the Linux kernel to a fixed version: 6.12.118, 6.18.64, 7.3.5, or 7.4-rc2 or later, which contain the patch that rewrites io_poll_cancel() to walk the hash once and cancel all matching requests in a single pass. As an immediate workaround, restrict unprivileged user access to io_uring operations by setting kernel.io_uring_disabled=1 (disables io_uring entirely) or kernel.io_uring_group to limit access to trusted users. Patch commits are available in the stable kernel tree (Kernel Patch 1, Kernel Patch 2, Red Hat CVE).

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-2026-99798HIGH7.8
  • Linux Kernel logoLinux Kernel
  • kernel-zfcpdump
NoYesNov 20, 2026
CVE-2026-99804MEDIUM5.5
  • Linux Kernel logoLinux Kernel
  • kernel-rt-64k-debug-modules-internal
NoYesNov 20, 2026
CVE-2026-99799MEDIUM5.5
  • Linux Kernel logoLinux Kernel
  • kernel-64k-debug-modules-partner
NoYesNov 20, 2026
CVE-2026-99797MEDIUM5.5
  • Linux Kernel logoLinux Kernel
  • kernel-zfcpdump-devel-matched
NoYesNov 20, 2026
CVE-2026-99796MEDIUM5.5
  • Linux Kernel logoLinux Kernel
  • kernel-rt-debug-modules-partner
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