{"api_version":"1","generated_at":"2026-10-01T12:24:04+00:00","cve":"CVE-2026-97496","urls":{"html":"https://cve.report/CVE-2026-97496","api":"https://cve.report/api/cve/CVE-2026-97496.json","docs":"https://cve.report/api","cve_org":"https://www.cve.org/CVERecord?id=CVE-2026-97496","nvd":"https://nvd.nist.gov/vuln/detail/CVE-2026-97496"},"summary":{"title":"drm/amdkfd: Fix OOB memory exposure in get_wave_state()","description":"In the Linux kernel, the following vulnerability has been resolved:\n\ndrm/amdkfd: Fix OOB memory exposure in get_wave_state()\n\nThe get_wave_state() function for v9 trusts cp_hqd_cntl_stack_size and\ncp_hqd_cntl_stack_offset values read directly from the MQD, which are\nwritten by GPU microcode and fully attacker-controlled on the\nCRIU-restore path (via AMDKFD_IOC_RESTORE_PROCESS with H3).\n\nthis leads to an unbounded copy_to_user() that can leak adjacent\nGTT/kernel memory. If offset > size, integer underflow produces a ~4 GiB\nread length, if size is set to 1 MiB against a 4 KiB allocation, we leak\n1 MiB of adjacent kernel memory (other queues' MQDs, ring buffers, KASLR\npointers).\n\nFix by clamping both cp_hqd_cntl_stack_size to the actual allocated\nbuffer size (q->ctl_stack_size) and cp_hqd_cntl_stack_offset to the\nclamped size before performing arithmetic and copy_to_user().\n\nThis ensures we never read beyond the allocated kernel BO regardless of\nattacker-supplied MQD field values.","state":"PUBLISHED","assigner":"Linux","published_at":"2026-09-24 17:17:27","updated_at":"2026-09-28 06:16:35"},"problem_types":[],"metrics":[{"version":"3.1","source":"416baaa9-dc9f-4396-8d5f-8c081fb06d67","type":"Secondary","score":"7.1","severity":"HIGH","vector":"CVSS:3.1/AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:N/A:H","data":{"version":"3.1","vectorString":"CVSS:3.1/AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:N/A:H","baseScore":7.1,"baseSeverity":"HIGH","attackVector":"LOCAL","attackComplexity":"LOW","privilegesRequired":"LOW","userInteraction":"NONE","scope":"UNCHANGED","confidentialityImpact":"HIGH","integrityImpact":"NONE","availabilityImpact":"HIGH"}},{"version":"3.1","source":"CNA","type":"DECLARED","score":"7.1","severity":"HIGH","vector":"CVSS:3.1/AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:N/A:H","data":{"baseScore":7.1,"baseSeverity":"HIGH","vectorString":"CVSS:3.1/AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:N/A:H","version":"3.1"}}],"references":[{"url":"https://git.kernel.org/stable/c/d9183d974ddd5f09d029beaf359275ac20d4d5fe","name":"https://git.kernel.org/stable/c/d9183d974ddd5f09d029beaf359275ac20d4d5fe","refsource":"416baaa9-dc9f-4396-8d5f-8c081fb06d67","tags":[],"title":"","mime":"","httpstatus":"","archivestatus":"0"},{"url":"https://git.kernel.org/stable/c/7ef144458f48d5589e36f1b3d83e83db2e5c5ba5","name":"https://git.kernel.org/stable/c/7ef144458f48d5589e36f1b3d83e83db2e5c5ba5","refsource":"416baaa9-dc9f-4396-8d5f-8c081fb06d67","tags":[],"title":"","mime":"","httpstatus":"","archivestatus":"0"},{"url":"https://git.kernel.org/stable/c/ec646686613d8ab05b282d1463a7baa49fd6b83b","name":"https://git.kernel.org/stable/c/ec646686613d8ab05b282d1463a7baa49fd6b83b","refsource":"416baaa9-dc9f-4396-8d5f-8c081fb06d67","tags":[],"title":"","mime":"","httpstatus":"","archivestatus":"0"},{"url":"https://www.cve.org/CVERecord?id=CVE-2026-97496","name":"CVE Program record","refsource":"CVE.ORG","tags":["canonical"]},{"url":"https://nvd.nist.gov/vuln/detail/CVE-2026-97496","name":"NVD vulnerability detail","refsource":"NVD","tags":["canonical","analysis"]}],"affected":[{"source":"CNA","vendor":"Linux","product":"Linux","version":"affected 42c6c48214b726c30918e8dc80e2168607d13ae4 d9183d974ddd5f09d029beaf359275ac20d4d5fe git","platforms":[]},{"source":"CNA","vendor":"Linux","product":"Linux","version":"affected 42c6c48214b726c30918e8dc80e2168607d13ae4 ec646686613d8ab05b282d1463a7baa49fd6b83b git","platforms":[]},{"source":"CNA","vendor":"Linux","product":"Linux","version":"affected 42c6c48214b726c30918e8dc80e2168607d13ae4 7ef144458f48d5589e36f1b3d83e83db2e5c5ba5 git","platforms":[]},{"source":"CNA","vendor":"Linux","product":"Linux","version":"affected 5.18","platforms":[]},{"source":"CNA","vendor":"Linux","product":"Linux","version":"unaffected 5.18 semver","platforms":[]},{"source":"CNA","vendor":"Linux","product":"Linux","version":"unaffected 6.12.111 6.12.* semver","platforms":[]},{"source":"CNA","vendor":"Linux","product":"Linux","version":"unaffected 6.18.53 6.18.* semver","platforms":[]},{"source":"CNA","vendor":"Linux","product":"Linux","version":"unaffected 7.2 * original_commit_for_fix","platforms":[]}],"timeline":[],"solutions":[],"workarounds":[],"exploits":[],"credits":[],"nvd_cpes":[],"vendor_comments":[],"enrichments":{"kev":null,"epss":{"cve_year":"2026","cve_id":"97496","cve":"CVE-2026-97496","epss":"0.001240000","percentile":"0.018190000","score_date":"2026-09-29","updated_at":"2026-09-30 00:11:23"},"legacy_qids":[]},"source_records":{"cve_program":{"containers":{"cna":{"affected":[{"defaultStatus":"unaffected","product":"Linux","programFiles":["drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v9.c"],"repo":"https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git","vendor":"Linux","versions":[{"lessThan":"d9183d974ddd5f09d029beaf359275ac20d4d5fe","status":"affected","version":"42c6c48214b726c30918e8dc80e2168607d13ae4","versionType":"git"},{"lessThan":"ec646686613d8ab05b282d1463a7baa49fd6b83b","status":"affected","version":"42c6c48214b726c30918e8dc80e2168607d13ae4","versionType":"git"},{"lessThan":"7ef144458f48d5589e36f1b3d83e83db2e5c5ba5","status":"affected","version":"42c6c48214b726c30918e8dc80e2168607d13ae4","versionType":"git"}]},{"defaultStatus":"affected","product":"Linux","programFiles":["drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v9.c"],"repo":"https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git","vendor":"Linux","versions":[{"status":"affected","version":"5.18"},{"lessThan":"5.18","status":"unaffected","version":"0","versionType":"semver"},{"lessThanOrEqual":"6.12.*","status":"unaffected","version":"6.12.111","versionType":"semver"},{"lessThanOrEqual":"6.18.*","status":"unaffected","version":"6.18.53","versionType":"semver"},{"lessThanOrEqual":"*","status":"unaffected","version":"7.2","versionType":"original_commit_for_fix"}]}],"cpeApplicability":[{"nodes":[{"cpeMatch":[{"criteria":"cpe:2.3:o:linux:linux_kernel:*:*:*:*:*:*:*:*","versionEndExcluding":"6.12.111","versionStartIncluding":"5.18","vulnerable":true},{"criteria":"cpe:2.3:o:linux:linux_kernel:*:*:*:*:*:*:*:*","versionEndExcluding":"6.18.53","versionStartIncluding":"5.18","vulnerable":true},{"criteria":"cpe:2.3:o:linux:linux_kernel:*:*:*:*:*:*:*:*","versionEndExcluding":"7.2","versionStartIncluding":"5.18","vulnerable":true}],"negate":false,"operator":"OR"}]}],"descriptions":[{"lang":"en","value":"In the Linux kernel, the following vulnerability has been resolved:\n\ndrm/amdkfd: Fix OOB memory exposure in get_wave_state()\n\nThe get_wave_state() function for v9 trusts cp_hqd_cntl_stack_size and\ncp_hqd_cntl_stack_offset values read directly from the MQD, which are\nwritten by GPU microcode and fully attacker-controlled on the\nCRIU-restore path (via AMDKFD_IOC_RESTORE_PROCESS with H3).\n\nthis leads to an unbounded copy_to_user() that can leak adjacent\nGTT/kernel memory. If offset > size, integer underflow produces a ~4 GiB\nread length, if size is set to 1 MiB against a 4 KiB allocation, we leak\n1 MiB of adjacent kernel memory (other queues' MQDs, ring buffers, KASLR\npointers).\n\nFix by clamping both cp_hqd_cntl_stack_size to the actual allocated\nbuffer size (q->ctl_stack_size) and cp_hqd_cntl_stack_offset to the\nclamped size before performing arithmetic and copy_to_user().\n\nThis ensures we never read beyond the allocated kernel BO regardless of\nattacker-supplied MQD field values."}],"metrics":[{"cvssV3_1":{"baseScore":7.1,"baseSeverity":"HIGH","vectorString":"CVSS:3.1/AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:N/A:H","version":"3.1"},"scenarios":[{"lang":"en","value":"AV:L - Attacker-controlled cp_hqd_cntl_stack_size/offset enter via AMDKFD_IOC_CRIU_OP KFD_CRIU_OP_RESTORE (kfd_ioctl_criu → criu_restore → kfd_criu_restore_queue → restore_mqd memcpy of the user MQD), then v9 get_wave_state() runs from kfd_ioctl_get_queue_wave_state (AMDKFD_IOC_GET_QUEUE_WAVE_STATE) on /dev/kfd; no remote protocol carries those fields.\nAC:L - restore_mqd() installs the crafted MQD then sets qp->is_active=0, so AMDKFD_IOC_GET_QUEUE_WAVE_STATE deterministically passes dqm get_wave_state()'s !is_active and cwsr_enabled checks (cwsr_enable defaults to 1) and v9 get_wave_state() copy_to_user()s using unclamped size-offset; no race or firmware quirk is required.\nPR:L - AMDKFD_IOC_GET_QUEUE_WAVE_STATE has ioctl flags 0, but the MQD size/offset that make get_wave_state() over-read are attacker-controlled only after KFD_CRIU_OP_RESTORE; kfd_ioctl() admits that op only with capable(CAP_CHECKPOINT_RESTORE) or capable(CAP_SYS_ADMIN), a delegated non-root cap routinely granted to CRIU/container checkpoint services.\nUI:N - The attacker opens /dev/kfd and issues KFD_CRIU_OP_RESTORE then AMDKFD_IOC_GET_QUEUE_WAVE_STATE against its own KFD process; no other user must mount a filesystem, open a file, or otherwise interact.\nS:U - The unbounded copy_to_user() from the kernel MQD control-stack BO in v9 get_wave_state() discloses host kernel/GTT memory inside the same kernel security authority; no VM, IOMMU, or sandbox boundary is crossed.\nC:H - Pre-fix v9 get_wave_state() copy_to_user()s (cp_hqd_cntl_stack_size - cp_hqd_cntl_stack_offset) bytes from mqd+AMDGPU_GPU_PAGE_SIZE without clamping to q->ctl_stack_size, so a restored 1MiB size against a 4KiB allocate_mqd() BO (or offset>size underflow to ~4GiB) dumps adjacent GTT/kernel memory including other MQDs and KASLR pointers.\nI:N - The defective operation is copy_to_user() from the kernel control-stack BO to the caller's ctl_stack_address; restore_mqd() writes only sizeof(struct v9_mqd) plus ctl_stack_size into the buffer allocate_mqd() sized from q->ctl_stack_size, so this bug does not modify kernel memory out of bounds.\nA:H - A restored cp_hqd_cntl_stack_size of ~1MiB or an offset>size underflow to ~4GiB makes get_wave_state() copy_to_user() walk off the small kmap/vmap GTT/VRAM BO into unmapped kernel VA (and can trip CONFIG_HARDENED_USERCOPY), producing a kernel oops."}]}],"providerMetadata":{"dateUpdated":"2026-09-28T05:30:29.056Z","orgId":"416baaa9-dc9f-4396-8d5f-8c081fb06d67","shortName":"Linux"},"references":[{"url":"https://git.kernel.org/stable/c/d9183d974ddd5f09d029beaf359275ac20d4d5fe"},{"url":"https://git.kernel.org/stable/c/ec646686613d8ab05b282d1463a7baa49fd6b83b"},{"url":"https://git.kernel.org/stable/c/7ef144458f48d5589e36f1b3d83e83db2e5c5ba5"}],"title":"drm/amdkfd: Fix OOB memory exposure in get_wave_state()","x_generator":{"engine":"bippy-1.2.0"}}},"cveMetadata":{"assignerOrgId":"416baaa9-dc9f-4396-8d5f-8c081fb06d67","assignerShortName":"Linux","cveId":"CVE-2026-97496","datePublished":"2026-09-24T16:04:43.625Z","dateReserved":"2026-09-24T16:01:01.150Z","dateUpdated":"2026-09-28T05:30:29.056Z","state":"PUBLISHED"},"dataType":"CVE_RECORD","dataVersion":"5.2"},"nvd":{"publishedDate":"2026-09-24 17:17:27","lastModifiedDate":"2026-09-28 06:16:35","problem_types":[],"metrics":{"cvssMetricV31":[{"source":"416baaa9-dc9f-4396-8d5f-8c081fb06d67","type":"Secondary","cvssData":{"version":"3.1","vectorString":"CVSS:3.1/AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:N/A:H","baseScore":7.1,"baseSeverity":"HIGH","attackVector":"LOCAL","attackComplexity":"LOW","privilegesRequired":"LOW","userInteraction":"NONE","scope":"UNCHANGED","confidentialityImpact":"HIGH","integrityImpact":"NONE","availabilityImpact":"HIGH"},"exploitabilityScore":1.8,"impactScore":5.2}]},"configurations":[]},"legacy_mitre":{"record":{"CveYear":"2026","CveId":"97496","Ordinal":"1","Title":"drm/amdkfd: Fix OOB memory exposure in get_wave_state()","CVE":"CVE-2026-97496","Year":"2026"},"notes":[{"CveYear":"2026","CveId":"97496","Ordinal":"1","NoteData":"In the Linux kernel, the following vulnerability has been resolved:\n\ndrm/amdkfd: Fix OOB memory exposure in get_wave_state()\n\nThe get_wave_state() function for v9 trusts cp_hqd_cntl_stack_size and\ncp_hqd_cntl_stack_offset values read directly from the MQD, which are\nwritten by GPU microcode and fully attacker-controlled on the\nCRIU-restore path (via AMDKFD_IOC_RESTORE_PROCESS with H3).\n\nthis leads to an unbounded copy_to_user() that can leak adjacent\nGTT/kernel memory. If offset > size, integer underflow produces a ~4 GiB\nread length, if size is set to 1 MiB against a 4 KiB allocation, we leak\n1 MiB of adjacent kernel memory (other queues' MQDs, ring buffers, KASLR\npointers).\n\nFix by clamping both cp_hqd_cntl_stack_size to the actual allocated\nbuffer size (q->ctl_stack_size) and cp_hqd_cntl_stack_offset to the\nclamped size before performing arithmetic and copy_to_user().\n\nThis ensures we never read beyond the allocated kernel BO regardless of\nattacker-supplied MQD field values.","Type":"Description","Title":"drm/amdkfd: Fix OOB memory exposure in get_wave_state()"}]}}}