{"id":"CVE-2026-98509","title":"drm/virtio: fix memory leak of fence event on execbuffer failure","summary":"In the Linux kernel, the following vulnerability has been resolved:\n\ndrm/virtio: fix memory leak of fence event on execbuffer failure\n\nvirtio_gpu_execbuffer_ioctl() reserves a DRM event with\ndrm_event_reserve_init() when VIRTGPU_EXECBUF_…","severity":"none","vendor":"Linux","product":"Linux","affected":["Linux >= cd7f5ca33585918febe5e2f6dc090a21cfa775b0 < 43bfe2b60c149bf6effda25f245a00509061fa15","Linux >= cd7f5ca33585918febe5e2f6dc090a21cfa775b0 < 4e333e2336138f4659b1d7a0371c499f80428612","Linux >= cd7f5ca33585918febe5e2f6dc090a21cfa775b0 < 8fa78900fd24247277c03dc7b8a2864cb12d4340","Linux >= cd7f5ca33585918febe5e2f6dc090a21cfa775b0 < ae583668f16d41fcc1b54f45848c1b6ac1d6669a","Linux >= cd7f5ca33585918febe5e2f6dc090a21cfa775b0 < b74aad23d99b279bb34d135795f39a6d8ecdc075","Linux 5.16"],"published":"2026-10-18","updated":"2026-10-18","sourceUpdated":"2026-10-18T15:17:19.352Z","source":"CVEORG","sourceUrl":"https://www.cve.org/CVERecord?id=CVE-2026-98509","references":[{"url":"https://git.kernel.org/stable/c/43bfe2b60c149bf6effda25f245a00509061fa15"},{"url":"https://git.kernel.org/stable/c/4e333e2336138f4659b1d7a0371c499f80428612"},{"url":"https://git.kernel.org/stable/c/8fa78900fd24247277c03dc7b8a2864cb12d4340"},{"url":"https://git.kernel.org/stable/c/ae583668f16d41fcc1b54f45848c1b6ac1d6669a"},{"url":"https://git.kernel.org/stable/c/b74aad23d99b279bb34d135795f39a6d8ecdc075"}],"tags":["cve.org"],"ingestedAt":"2026-10-18T15:45:12.286Z","slug":"CVE-2026-98509","body":"## Overview\n\nIn the Linux kernel, the following vulnerability has been resolved:\n\ndrm/virtio: fix memory leak of fence event on execbuffer failure\n\nvirtio_gpu_execbuffer_ioctl() reserves a DRM event with\ndrm_event_reserve_init() when VIRTGPU_EXECBUF_RING_IDX selects a ring that\nuserspace has enabled polling for. virtio_gpu_init_submit() does this\nbefore the BO handles, the command buffer, the syncobj arrays and the\nin-fence are processed, so every later error path runs with the event\nalready pending, including plain argument validation failures such as an\ninvalid bo_handle or an in-syncobj that carries no fence.\n\nOn those paths, virtio_gpu_cleanup_submit() drops the out-fence without\ncancelling the event. The fence is freed without ever having been emitted,\ntaking the only driver-side pointer to the event with it. Closing the DRM\nfile does not help. drm_events_release() unlinks pending events but\ndeliberately leaves the freeing to the driver's later drm_send_event(),\nwhich never runs for an orphaned event, so the allocation is leaked\npermanently.\n\nFound when fuzzing the virtio driver with Syzkaller:\n\n\tBUG: memory leak\n\tunreferenced object 0xffff88802c176e80 (size 96):\n\tcomm \"syz.1.367\", pid 10561, jiffies 4294960122\n\thex dump (first 32 bytes):\n\t\t00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00  ................\n\t\tc8 6e 17 2c 80 88 ff ff 00 00 00 00 00 00 00 00  .n.,............\n\tbacktrace (crc e1973c6b):\n\t\tkmemleak_alloc_recursive include/linux/kmemleak.h:44 [inline]\n\t\tslab_post_alloc_hook mm/slub.c:4597 [inline]\n\t\tslab_alloc_node mm/slub.c:4917 [inline]\n\t\t__kmalloc_cache_noprof+0x49d/0x6f0 mm/slub.c:5485\n\t\t_kmalloc_noprof include/linux/slab.h:988 [inline]\n\t\t_kzalloc_noprof include/linux/slab.h:1309 [inline]\n\t\tvirtio_gpu_fence_event_create drivers/gpu/drm/virtio/virtgpu_submit.c:282 [inline]\n\t\tvirtio_gpu_init_submit drivers/gpu/drm/virtio/virtgpu_submit.c:398 [inline]\n\t\tvirtio_gpu_execbuffer_ioctl+0xbbf/0x1aa0 drivers/gpu/drm/virtio/virtgpu_submit.c:505\n\t\tdrm_ioctl_kernel+0x1f4/0x3e0 drivers/gpu/drm/drm_ioctl.c:817\n\t\tdrm_ioctl+0x5f4/0xc70 drivers/gpu/drm/drm_ioctl.c:914\n\t\tvfs_ioctl fs/ioctl.c:51 [inline]\n\t\t__do_sys_ioctl fs/ioctl.c:597 [inline]\n\t\t__se_sys_ioctl fs/ioctl.c:583 [inline]\n\t\t__x64_sys_ioctl+0x18e/0x210 fs/ioctl.c:583\n\t\tdo_syscall_x64 arch/x86/entry/syscall_64.c:63 [inline]\n\t\tdo_syscall_64+0x116/0x800 arch/x86/entry/syscall_64.c:94\n\t\tentry_SYSCALL_64_after_hwframe+0x77/0x7f\n\nFix by cancelling and freeing the DRM event on the execbuffer error path\nbefore dropping the fence. Clear the fence's event pointer after\ncancellation so it does not retain a dangling pointer.\n\n## Affected\n\n- `Linux >= cd7f5ca33585918febe5e2f6dc090a21cfa775b0 < 43bfe2b60c149bf6effda25f245a00509061fa15`\n- `Linux >= cd7f5ca33585918febe5e2f6dc090a21cfa775b0 < 4e333e2336138f4659b1d7a0371c499f80428612`\n- `Linux >= cd7f5ca33585918febe5e2f6dc090a21cfa775b0 < 8fa78900fd24247277c03dc7b8a2864cb12d4340`\n- `Linux >= cd7f5ca33585918febe5e2f6dc090a21cfa775b0 < ae583668f16d41fcc1b54f45848c1b6ac1d6669a`\n- `Linux >= cd7f5ca33585918febe5e2f6dc090a21cfa775b0 < b74aad23d99b279bb34d135795f39a6d8ecdc075`\n- `Linux 5.16`\n\n## Remediation\n\nRefer to the linked advisories for vendor-supplied fixes and affected version ranges.","depth":"sunlit","depthScore":3,"depthScoreParts":{"impact":2.8,"likelihood":0,"exploitation":0,"ransomware":0},"changes":[]}