{"id":"CVE-2026-89809","published":"2026-09-16T11:16:46.047","lastModified":"2026-09-17T10:17:04.047","description":"In the Linux kernel, the following vulnerability has been resolved:\n\ndrm/amdkfd: fix scope of mqd_mgr dereference in pqm_debugfs_mqds\n\nReading /sys/kernel/debug/kfd/mqds while a process holds an active KFD\nqueue triggers a NULL pointer dereference because the for loop that\ncalls mqd_mgr->debugfs_show_mqd() is incorrectly placed outside the\nif (pqn->q) block that initializes mqd_mgr.\n\nThe queue list can contain entries where pqn->q is NULL (kernel queues\nwhere only pqn->kq is valid). In the original code:\n\n  if (pqn->q) {\n      ...\n      mqd_mgr = q->device->dqm->mqd_mgrs[mqd_type];\n      size = mqd_mgr->mqd_stride(...);\n  }\n\n  for (xcc = 0; xcc < num_xccs; xcc++) {  // WRONG: outside if block\n      mqd = q->mqd + size * xcc;\n      r = mqd_mgr->debugfs_show_mqd(m, mqd);\n  }\n\nWhen iterating over a queue node where pqn->q is NULL:\n1. The if (pqn->q) block is skipped\n2. mqd_mgr remains uninitialized (NULL from declaration)\n3. The for loop executes anyway\n4. mqd_mgr->debugfs_show_mqd(m, mqd) dereferences NULL\n\nThe crash manifests as:\n\n  BUG: kernel NULL pointer dereference, address: 0000000000000000\n  #PF: supervisor instruction fetch in kernel mode\n  RIP: 0010:0x0\n  Call Trace:\n   pqm_debugfs_mqds+0x10c/0x1d0 [amdgpu]\n   kfd_debugfs_mqds_by_process+0x9b/0x110 [amdgpu]\n   seq_read_iter+0x132/0x4b0\n   ...\n\nFix by moving the for loop inside the if (pqn->q) block, so mqd_mgr\nand related variables are only used when properly initialized.\n\n(cherry picked from commit 8bfe29d5c798940f797aa24135d2734c3ffce9de)","cvssScore":null,"cvssSeverity":null,"cvssVector":null,"cwes":[],"vendors":[],"products":[],"references":[{"url":"https://git.kernel.org/stable/c/012a026bae0212952b423a842b7e2c0bf21f8e7a","tags":[]},{"url":"https://git.kernel.org/stable/c/58e866711b234571b0ce342c43d153a4786b32b8","tags":[]}],"exploitRefs":[],"hasPoc":false,"ai":null}