{"id":"CVE-2026-89812","published":"2026-09-16T11:16:46.380","lastModified":"2026-09-17T10:17:04.390","description":"In the Linux kernel, the following vulnerability has been resolved:\n\ndrm/amdgpu: force complete the MES ring fences on reset\n\nThe MES scheduler ring has no drm scheduler (no_scheduler = true), so it is\nskipped by the force-completion loop in amdgpu_device_pre_asic_reset(). It uses\na polling fence whose hw value lives in wb (GTT) memory and survives a MODE1\nreset, while fence_drv.sync_seq keeps advancing for every packet.\n\nWhen the reset is triggered because MES itself stopped responding, the\ntimed-out packets advance sync_seq past the last hw fence value MES wrote.\nAfter resume the first MES submission polls forever on a seq that is never\nwritten back, failing the resume and wedging the box on a second reset:\n\n  amdgpu: MES ring buffer is full.\n  amdgpu: *ERROR* ring gfx_0.0.0 test failed (-110)\n  amdgpu: resume of IP block <gfx_v11_0> failed -110\n  amdgpu: GPU reset end with ret = -110\n\nForce complete the MES scheduler ring fences together with the scheduler rings\nso their hw fence is realigned to sync_seq.\n\nv2: cover all XCCs (one scheduler ring each), not just mes.ring[0].","cvssScore":null,"cvssSeverity":null,"cvssVector":null,"cwes":[],"vendors":[],"products":[],"references":[{"url":"https://git.kernel.org/stable/c/48dc279c3010ac8f91b1845b2abb3a1e9943a0f5","tags":[]},{"url":"https://git.kernel.org/stable/c/ebe7542553d008e2285e0fbba0c63f0295a7e694","tags":[]}],"exploitRefs":[],"hasPoc":false,"ai":null}