{"dataType":"CVE_RECORD","dataVersion":"5.2","cveMetadata":{"cveId":"CVE-2026-99800","assignerOrgId":"416baaa9-dc9f-4396-8d5f-8c081fb06d67","state":"PUBLISHED","assignerShortName":"Linux","dateReserved":"2026-09-25T10:25:14.435Z","datePublished":"2026-11-20T07:02:22.819Z","dateUpdated":"2026-11-22T11:10:13.035Z"},"containers":{"cna":{"providerMetadata":{"orgId":"416baaa9-dc9f-4396-8d5f-8c081fb06d67","shortName":"Linux","dateUpdated":"2026-11-22T11:10:13.035Z"},"descriptions":[{"lang":"en","value":"In the Linux kernel, the following vulnerability has been resolved:\n\nio_uring: free the task context after a grace period\n\nA task other than the owner can free a request, io_ring_exit_work()\nrunning the local task_work of a DEFER_TASKRUN ring for example, or\nio_do_iopoll() reaping completions. io_put_task() then takes the else\nbranch:\n\n    percpu_counter_sub(&tctx->inflight, 1);\n    if (unlikely(atomic_read(&tctx->in_cancel)))\n            wake_up(&tctx->wait);\n    put_task_struct(tctx->task);\n\nThe first line is what lets the owner leave io_uring_cancel_generic(),\nand on exec it frees the tctx right after. Nothing holds that off for\nthe remaining three lines, the task reference only covers a tctx freed\nalong with the task_struct. It used to work out because\nio_uring_del_tctx_node() took the uring_lock that the freeing side holds\nacross io_put_task(), until that moved to tctx_lock.\n\nFree the tctx with kfree_rcu(), and enter an RCU read section in\nio_put_task() before the count is dropped."}],"metrics":[{"cvssV3_1":{"version":"3.1","vectorString":"CVSS:3.1/AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H","baseScore":7.8,"baseSeverity":"HIGH"},"scenarios":[{"lang":"en","value":"AV:L - io_put_task() is reached from io_do_iopoll() via io_uring_enter(), or from io_ring_exit_work() after close() of a DEFER_TASKRUN ring, and the owner frees the tctx in io_uring_free_tctx() from execve()->begin_new_exec()->io_uring_task_cancel(). Nothing in that race is a field parsed from a remote protocol message.\nAC:L - The same user drives both sides: execve() makes io_uring_cancel_generic() call io_uring_free_tctx() once percpu_counter_sub() zeroes tctx->inflight, while that user's io_uring_enter() iopoll or the iou_wq exit worker is still in io_put_task(). io_uring_del_tctx_node() takes tctx_lock, so the uring_lock held across the put no longer blocks the free, and the race can be retried.\nPR:L - io_uring_setup() calls io_uring_allowed(), which succeeds for any task when sysctl_io_uring_disabled is 0, the static default. IORING_SETUP_DEFER_TASKRUN and the execve() path through begin_new_exec() do not check CAP_SYS_ADMIN or any other capability.\nUI:N - The attacker itself calls io_uring_setup(), io_uring_enter() or close(), and execve() on its own tasks. No other user has to open a file, mount an image, or accept a prompt.\nS:U - The freed object is the caller's own kernel io_uring_task. The resulting memory corruption escalates that local user inside the same kernel and does not cross a KVM guest-to-host or IOMMU boundary.\nC:H - io_uring_free_tctx() kfree's the tctx while io_put_task() still executes atomic_read(&tctx->in_cancel), wake_up(&tctx->wait), and put_task_struct(tctx->task). Reclaiming that slab supplies the task pointer, and put_task_struct() then reads task->usage and, if it hits zero, the rest of the object in __put_task_struct().\nI:H - wake_up(&tctx->wait) locks and walks the wait_queue_head embedded in the freed io_uring_task, calling waiter->func if the reclaimed list is non-empty, and put_task_struct() does refcount_dec_and_test() on whatever pointer now sits in tctx->task, invoking __put_task_struct() when the count hits zero.\nA:H - io_put_task() still takes the spinlock in tctx->wait and dereferences tctx->task after io_uring_free_tctx() has returned the slab, so a poisoned or reused io_uring_task oopses or panics the kernel."}]}],"affected":[{"product":"Linux","vendor":"Linux","defaultStatus":"unaffected","repo":"https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git","programFiles":["include/linux/io_uring_types.h","io_uring/io_uring.c","io_uring/tctx.c"],"versions":[{"version":"5623eb1ed035f01dfa620366a82b667545b10c82","lessThan":"953174dcb862a7a364efdb65e178dcd3aea5562e","status":"affected","versionType":"git"},{"version":"5623eb1ed035f01dfa620366a82b667545b10c82","lessThan":"9f5315d3f4368590c9fe06d4313b7247470bcf15","status":"affected","versionType":"git"}]},{"product":"Linux","vendor":"Linux","defaultStatus":"affected","repo":"https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git","programFiles":["include/linux/io_uring_types.h","io_uring/io_uring.c","io_uring/tctx.c"],"versions":[{"version":"6.19","status":"affected"},{"version":"0","lessThan":"6.19","status":"unaffected","versionType":"semver"},{"version":"7.3.2","lessThanOrEqual":"7.3.*","status":"unaffected","versionType":"semver"},{"version":"7.4-rc1","lessThanOrEqual":"*","status":"unaffected","versionType":"original_commit_for_fix"}]}],"cpeApplicability":[{"nodes":[{"operator":"OR","negate":false,"cpeMatch":[{"vulnerable":true,"criteria":"cpe:2.3:o:linux:linux_kernel:*:*:*:*:*:*:*:*","versionStartIncluding":"6.19","versionEndExcluding":"7.3.2"},{"vulnerable":true,"criteria":"cpe:2.3:o:linux:linux_kernel:*:*:*:*:*:*:*:*","versionStartIncluding":"6.19","versionEndExcluding":"7.4-rc1"}]}]}],"references":[{"url":"https://git.kernel.org/stable/c/953174dcb862a7a364efdb65e178dcd3aea5562e"},{"url":"https://git.kernel.org/stable/c/9f5315d3f4368590c9fe06d4313b7247470bcf15"}],"title":"io_uring: free the task context after a grace period","x_generator":{"engine":"bippy-1.2.0"}}}}