
Cloud Vulnerability DB
A community-led vulnerabilities database
CVE-2026-99799 is a denial-of-service vulnerability in the Linux kernel's io_uring SQPOLL thread management subsystem, caused by an incorrect synchronization flaw (CWE-821). When io_sq_thread_stop() attempts to stop an SQPOLL thread, it wakes the thread and waits for it to exit; however, if the thread is held in get_signal() by a SIGSTOP signal or cgroup freezer, the wake-up has no effect, causing io_ring_exit_work() to sleep indefinitely. Because iou_exit has only 64 slots, 64 closed SQPOLL rings from a stopped process can block all other rings from being freed system-wide. The vulnerability affects Linux kernel versions starting from 5.12 up to the patched commits, with fixes backported to 6.12.117, 6.18.61, 7.3.2, and 7.4-rc1. It carries an estimated CVSS severity of Medium (GitHub Advisory, Red Hat).
The root cause is an incorrect synchronization pattern (CWE-821) in io_sq_thread_stop(), which blocks waiting for the SQPOLL thread to exit after sending a wake-up signal. Because the SQPOLL thread belongs to the thread group that created the io_uring ring, a SIGSTOP or cgroup freeze traps it in get_signal(), rendering the wake-up ineffective and causing the caller to sleep indefinitely. The fix changes the stop mechanism to set a stop bit and wake the thread without waiting, allowing it to exit when next scheduled. To prevent use-after-free conditions (a secondary concern, estimated CWE-416) that arise because the sqd (shared queue descriptor) can now outlive its last ring, a dual-reference counting scheme is introduced — one reference for the rings and one for the thread — freeing the sqd only when the last reference is dropped (GitHub Advisory, Red Hat).
Successful exploitation allows a local user with the ability to stop processes (via SIGSTOP or cgroup freezer control) to exhaust the 64-slot iou_exit queue by closing SQPOLL rings of a stopped process, preventing io_uring resource cleanup for all other processes on the system. This results in a system-wide denial of service affecting availability of io_uring resources, with no direct confidentiality or integrity impact. The scope extends beyond the attacker's own process, potentially disrupting any workload relying on io_uring across the entire host (GitHub Advisory, Feedly).
There is no public proof-of-concept exploit and no evidence of in-the-wild exploitation as of the disclosure date (GitHub Advisory). Exploitation requires local access and the ability to send SIGSTOP to a process or control cgroup freezer settings, limiting the attack surface. The EPSS score is 0.0, and the vulnerability is not listed in the CISA Known Exploited Vulnerabilities catalog. No threat actor attribution has been reported.
Update to a patched Linux kernel version: 6.12.117, 6.18.61, 7.3.2, or 7.4-rc1 and later. The upstream fix commits are a85561cc4fc5, 8bfd0b631926, 8298ce6443ff, and 360395e8dbf9 (GitHub Advisory, Red Hat). As a temporary workaround where patching is not immediately possible, restricting unprivileged access to io_uring (e.g., via sysctl kernel.io_uring_disabled=1 or seccomp policies) and limiting cgroup freezer permissions to trusted administrators can reduce exposure.
Fix availability across major Linux distributions and their releases.
bookworm
linux-6.12: 6.12.117-1~deb12u1
sid
linux: 7.3.3-1
trixie
linux: 6.12.117-1
bionic (esm-infra)
linux
bionic (fips-updates)
linux-fips
bionic (fips)
linux-fips
devel
linux
focal (esm-infra)
linux
focal (fips-updates)
linux-fips
focal (fips)
linux-fips
jammy
linux
Source: This report was generated using AI
Free Vulnerability Assessment
Evaluate your cloud security practices across 9 security domains to benchmark your risk level and identify gaps in your defenses.
Get a personalized demo
"Best User Experience I have ever seen, provides full visibility to cloud workloads."
"Wiz provides a single pane of glass to see what is going on in our cloud environments."
"We know that if Wiz identifies something as critical, it actually is."