{ "schema_version": "1.6.1", "id": "RPI-CVE-2026-72095", "modified": "2026-10-09T11:39:44Z", "upstream": [ "CVE-2026-72095" ], "published": "2026-08-15T06:04:33Z", "summary": "dma-fence: Make dma_fence_dedup_array() robust against 0-count input", "details": "In the Linux kernel, the following vulnerability has been resolved:\n\ndma-fence: Make dma_fence_dedup_array() robust against 0-count input\n\ndma_fence_dedup_array() returns 1 when called with num_fences == 0:\nthe for-loop body never executes, j stays at 0, and the final\n`return ++j` yields 1. This contradicts both the kernel-doc (\"Return:\nNumber of unique fences remaining in the array\") and the natural\nexpectation that 0 input gives 0 output.\n\nThe caller __dma_fence_unwrap_merge() bails out via the\n`if (count == 0 || count == 1)` fast path and so is save.\n\nBut amdgpu_userq_wait_*() could reach the dedup call with a zero local\ncount and dereference an uninitialized fence slot in the array.\n\nMake the contract match the documentation by returning 0 early. This\nalso skips an unnecessary sort() call on an empty array.", "severity": [ { "type": "CVSS_V3", "score": "CVSS:3.1/AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H" } ], "affected": [ { "package": { "ecosystem": "Raspberry Pi OS:13", "name": "linux" }, "versions": [ "1:6.18.29-1+rpt1", "1:6.18.33-1+rpt1", "1:6.18.34-1+rpt1", "1:6.18.39-1+rpt1" ], "ecosystem_specific": { "binaries": [ { "binary_name": "linux-image-6.18.29+rpt-rpi-2712", "binary_version": "1:6.18.29-1+rpt1" }, { "binary_name": "linux-image-6.18.29+rpt-rpi-v6", "binary_version": "1:6.18.29-1+rpt1" }, { "binary_name": "linux-image-6.18.29+rpt-rpi-v7", "binary_version": "1:6.18.29-1+rpt1" }, { "binary_name": "linux-image-6.18.29+rpt-rpi-v8", "binary_version": "1:6.18.29-1+rpt1" }, { "binary_name": "linux-image-6.18.29+rpt-rpi-v8-rt", "binary_version": "1:6.18.29-1+rpt1" }, { "binary_name": "linux-image-6.18.33+rpt-rpi-2712", "binary_version": "1:6.18.33-1+rpt1" }, { "binary_name": "linux-image-6.18.33+rpt-rpi-v6", "binary_version": "1:6.18.33-1+rpt1" }, { "binary_name": "linux-image-6.18.33+rpt-rpi-v7", "binary_version": "1:6.18.33-1+rpt1" }, { "binary_name": "linux-image-6.18.33+rpt-rpi-v8", "binary_version": "1:6.18.33-1+rpt1" }, { "binary_name": "linux-image-6.18.33+rpt-rpi-v8-rt", "binary_version": "1:6.18.33-1+rpt1" }, { "binary_name": "linux-image-6.18.34+rpt-rpi-2712", "binary_version": "1:6.18.34-1+rpt1" }, { "binary_name": "linux-image-6.18.34+rpt-rpi-v6", "binary_version": "1:6.18.34-1+rpt1" }, { "binary_name": "linux-image-6.18.34+rpt-rpi-v7", "binary_version": "1:6.18.34-1+rpt1" }, { "binary_name": "linux-image-6.18.34+rpt-rpi-v8", "binary_version": "1:6.18.34-1+rpt1" }, { "binary_name": "linux-image-6.18.34+rpt-rpi-v8-rt", "binary_version": "1:6.18.34-1+rpt1" }, { "binary_name": "linux-image-6.18.39+rpt-rpi-2712", "binary_version": "1:6.18.39-1+rpt1" }, { "binary_name": "linux-image-6.18.39+rpt-rpi-v6", "binary_version": "1:6.18.39-1+rpt1" }, { "binary_name": "linux-image-6.18.39+rpt-rpi-v7", "binary_version": "1:6.18.39-1+rpt1" }, { "binary_name": "linux-image-6.18.39+rpt-rpi-v8", "binary_version": "1:6.18.39-1+rpt1" }, { "binary_name": "linux-image-6.18.39+rpt-rpi-v8-rt", "binary_version": "1:6.18.39-1+rpt1" } ] }, "database_specific": { "per_version": { "1:6.18.29-1+rpt1": { "linux_commit": "66318c5b5cd8145a662303af12ed39ac07eb27cc", "upstream_version": "6.18.29", "assessment": "commit ancestry (strak rule)", "upstream_fixed": "6.18.40" }, "1:6.18.33-1+rpt1": { "linux_commit": "95b85bebbedcaedfa7ca79116ed38b7376fba412", "upstream_version": "6.18.33", "assessment": "commit ancestry (strak rule)", "upstream_fixed": "6.18.40" }, "1:6.18.34-1+rpt1": { "linux_commit": "c8c7494100e99ee05b11aaa4f0588a223a63d1af", "upstream_version": "6.18.34", "assessment": "commit ancestry (strak rule)", "upstream_fixed": "6.18.40" }, "1:6.18.39-1+rpt1": { "linux_commit": "60ea684a8ace97bb0db1a16e20753bdd6ab371ff", "upstream_version": "6.18.39", "assessment": "commit ancestry (strak rule)", "upstream_fixed": "6.18.40" } } } } ], "references": [ { "type": "ADVISORY", "url": "https://www.cve.org/CVERecord?id=CVE-2026-72095" }, { "type": "FIX", "url": "https://git.kernel.org/stable/c/7aa8f3dba53422465dbe1be8dbb7240304462bb2" }, { "type": "FIX", "url": "https://git.kernel.org/stable/c/e2d9a2ea178a5da0b4a6693e8ebca5c7fc4d7051" }, { "type": "FIX", "url": "https://git.kernel.org/stable/c/77a9298741f8f9e8b963c977f5582ab21c6d3427" } ], "database_specific": { "source": "https://git.kernel.org/pub/scm/linux/security/vulns.git", "upstream_modified": "2026-08-17T05:50:24Z", "severity": { "source": "kernel CNA", "cvss_v3_score": 7.8, "cvss_v3_severity": "High", "scenario": "AV:L - The bug is reached only via the AMDGPU_USERQ_WAIT DRM ioctl on /dev/dri/renderD*, requiring local access to an AMD GPU render node; it is not reachable from network protocols or remote packet handling.\nAC:L - An attacker with render-node access can reliably trigger the bug by creating a userq, calling AMDGPU_USERQ_WAIT with num_fences>0 and zero syncobj/BO handles so local num_fences stays 0 and dma_fence_dedup_array() incorrectly returns 1.\nPR:L - Exploitation requires only an authenticated DRM client on a render node (DRM_AUTH|DRM_RENDER_ALLOW), which unprivileged users commonly obtain via the render group on workstations, cloud GPU instances, and shared compute nodes.\nUI:N - No victim interaction is required; the attacker drives the vulnerable AMDGPU_USERQ_WAIT ioctl directly after creating their own user queue.\nS:U - Impact remains within the host kernel security domain (local privilege escalation/DoS on the GPU server); it does not cross VM, container, or IOMMU boundaries.\nC:H - When num_fences is 0 the dedup helper returns 1, causing iteration over uninitialized kmalloc slots whose stale contents are then dereferenced via to_amdgpu_userq_fence()/dma_fence_wait(), enabling kernel memory reads and pointer disclosure.\nI:H - The uninitialized fences[0] pointer is passed into dma_fence_enable_sw_signaling() and dma_fence_wait(), which invoke fence ops and locking on attacker-influenced addresses, creating a plausible path to control-flow hijacking via heap grooming.\nA:H - Dereferencing an uninitialized or attacker-influenced dma_fence pointer in kernel context can cause an oops, BUG, or panic, producing a repeatable local denial of service on affected AMDGPU userq systems." } } }