CVE-2026-99183
UnknownSummary
In the Linux kernel accel/rocket driver, sched_to_core() searches only up to rdev->num_cores, which is decremented when a core is unbound. Unbinding a non-last core leaves later cores outside the search, so sched_to_core() returns NULL for a core that is still bound and running jobs. Neither rocket_job_run() nor rocket_job_timedout() checks the result, leading to NULL pointer dereference (KASAN: null-ptr-deref).
Risk Assessment
Unbinding a middle NPU core (e.g., on RK3588) while clients are active causes kernel faults and potential system crash. This can lead to denial of service or platform instability.
Recommendation
Update the Linux kernel to a version containing the fix that searches all allocated core slots. Avoid unbinding individual NPU cores while clients are active until the patch is deployed.
Other vulnerabilities in Linux kernel
See all- CVE-2026-58942High
In do_fault_around of memory.c, there is a possible use after free due to a logic error in the code. This could lead to local escalation of privilege with no additional execution privileges needed.
- CVE-2026-58847Unknown
In kvm_vfio_pviommu_attach of vfio.c, there is a possible double free due to a logic error in the code. This could lead to local escalation of privilege with no additional execution privileges needed. User interaction is not needed for exploitation.
- CVE-2026-58829High
In fuse_backing_mmap of backing.c, there is a possible memory corruption due to a use after free. This could lead to local escalation of privilege with no additional execution privileges needed.
- CVE-2026-58828High
In multiple functions of base.c, there is a possible use-after-free due to a race condition. This could lead to local escalation of privilege with no additional execution privileges needed.
- CVE-2026-58827High
In multiple functions of dma-buf.c, there is a possible use after free due to a race condition. This could lead to local escalation of privilege with no additional execution privileges needed.
- CVE-2026-58826High
In ksys_unshare of fork.c, there is a possible memory corruption due to a use after free. This could lead to local escalation of privilege with no additional execution privileges needed.
- CVE-2026-58825High
In multiple functions of fork.c, there is a possible use after free due to a logic error in the code. This could lead to local escalation of privilege with no additional execution privileges needed.
- CVE-2026-99242Unknown
In the Linux kernel, a race between fallocate and mmap reads in NTFS was fixed. Fallocate only took invalidate_lock for punch hole, collapse range and insert range operations, and during standard allocation new clusters were mapped before being zeroed, allowing concurrent mmap page faults to read uninitialized disk data.
- CVE-2026-99241Unknown
In the Linux kernel, a missing lock in ntfs_filemap_page_mkwrite() was fixed; it called iomap_page_mkwrite() without holding mapping->invalidate_lock. A concurrent truncate or fallocate could be invalidating pagecache and rewriting the runlist while the write fault mapped blocks and dirtied the folio.
- CVE-2026-99240Unknown
In the Linux kernel, the DAMON subsystem (mm/damon/core) has a division-by-zero bug when a DAMOS quota goal has a target_value of zero. The fix moves handling of this corner case to the core layer by adding validation in damos_commit_quota_goal(), which now returns an error for zero target values.
Original NVD description (English source)
In the Linux kernel, the following vulnerability has been resolved: accel/rocket: search every core slot when looking up a scheduler sched_to_core() walks rdev->cores[] up to rdev->num_cores, and rocket_remove() decrements num_cores for every core it removes. Unbind a core that is not the last one and the cores behind it fall outside the search, so sched_to_core() returns NULL for a core that is still bound and still running jobs. Neither caller checks the result: rocket_job_run(): rocket_fence_create(core), core->dev rocket_job_timedout(): dev_err(core->dev, "NPU job timed out") Unbinding the middle core of the three on an RK3588 while three clients are submitting to all of them faults twice, once from the surviving core's job queue and once from its reset work: KASAN: null-ptr-deref in range [0x0000000000000220-0x0000000000000227] Workqueue: fdad0000.npu drm_sched_run_job_work [gpu_sched] pc : rocket_job_run+0x234/0x838 [rocket] Call trace: rocket_job_run+0x234/0x838 [rocket] drm_sched_run_job_work+0x2cc/0xad8 [gpu_sched] process_one_work+0x640/0x14f0 KASAN: null-ptr-deref in range [0x0000000000000000-0x0000000000000007] Workqueue: rocket-reset-2 drm_sched_job_timedout [gpu_sched] pc : rocket_job_timedout+0xf0/0x1e0 [rocket] Call trace: rocket_job_timedout+0xf0/0x1e0 [rocket] drm_sched_job_timedout+0x188/0x6a0 [gpu_sched] Both are the third core: the workqueue names are its device and its core->index, and it was left at slot 2 while num_cores had dropped to 2. Search all the slots that were allocated, the way find_core_for_dev() now does. A core that is still bound is then found, and the two callers get the pointer they already assume they have. This does not make unbinding one core out of several safe. An open client keeps an entity pointing at the scheduler of the core that went away: drm_sched reports it as not ready for every job that lands on it, and the client waits in dma_fence_default_wait for a fence that will never signal. Stopping the NULL dereference is what belongs in a fix; the rest wants more thought.
Vulnerability data from NVD (NIST) · CISA KEV · EPSS