CVE-2026-80589

*** Pendiente de traducción *** In the Linux kernel, the following vulnerability has been resolved:<br /> <br /> block: stop the timeout timer when releasing a never added disk<br /> <br /> disk_release() undoes blk_mq_init_allocated_queue() for a disk whose<br /> probe failed before add_disk(), but it only calls blk_mq_exit_queue().<br /> Nothing there stops q-&gt;timeout, and that timer rolls forward: it stays<br /> pending until it next expires, not until the last request completes.<br /> So if the driver issued any I/O before adding the disk, the<br /> request_queue is freed while still linked into a timer wheel bucket.<br /> <br /> Commit 6f8191fdf41d ("block: simplify disk shutdown") dropped the<br /> blk_cleanup_queue() call that used to stop it. __del_gendisk() and<br /> blk_mq_destroy_queue() still do; only the probe failure path lost it.<br /> <br /> nvme gets there because nvme_update_ns_info() submits Report Zones or<br /> FDP io-mgmt-recv on ns-&gt;queue before the disk is added, so a later<br /> failure - a concurrent reset setting NVME_CTRL_FROZEN, or<br /> device_add_disk() failing - lands in put_disk() with the timer armed:<br /> <br /> BUG: KASAN: slab-use-after-free in detach_if_pending+0x30c/0x340<br /> Write of size 8 at addr ffff888004d71310 by task kworker/u8:2/37<br /> __timer_delete_sync+0x156/0x240 kernel/time/timer.c:1621<br /> blk_sync_queue+0x22/0x40 block/blk-core.c:222<br /> nvme_sync_queues+0x100/0x150 drivers/nvme/host/core.c:5362<br /> nvme_reset_work+0x138/0x930 drivers/nvme/host/pci.c:3264<br /> <br /> Allocated by task 34:<br /> __blk_mq_alloc_disk+0x33/0x100 block/blk-mq.c:4462<br /> nvme_alloc_ns+0x290/0x3870 drivers/nvme/host/core.c:4146<br /> <br /> Freed by task 0:<br /> blk_free_queue_rcu+0x3a/0x50 block/blk-core.c:254<br /> rcu_core+0xc10/0x1730 kernel/rcu/tree.c:2857<br /> <br /> The queue being synced there is ctrl-&gt;admin_q, only a victim sharing a<br /> timer wheel bucket with the freed queue&amp;#39;s dangling entry; other runs<br /> tripped in enqueue_timer(), __run_timers() or blk_mq_timeout_work().<br /> Failing nvme_alloc_ns() with a debug patch makes it deterministic: one<br /> leaked timer trips KASAN within seconds, while 1987 patched releases<br /> produced no splat.<br /> <br /> Stop the timer and the queue work items before blk_mq_exit_queue(), like<br /> blk_mq_destroy_queue() does.<br /> <br /> Found by FuzzNvme.