{"dataType":"CVE_RECORD","dataVersion":"5.2","cveMetadata":{"cveId":"CVE-2026-68298","assignerOrgId":"416baaa9-dc9f-4396-8d5f-8c081fb06d67","state":"PUBLISHED","assignerShortName":"Linux","dateReserved":"2026-07-30T09:28:09.380Z","datePublished":"2026-08-10T12:02:32.194Z","dateUpdated":"2026-08-17T05:02:34.595Z"},"containers":{"cna":{"providerMetadata":{"orgId":"416baaa9-dc9f-4396-8d5f-8c081fb06d67","shortName":"Linux","dateUpdated":"2026-08-17T05:02:34.595Z"},"descriptions":[{"lang":"en","value":"In the Linux kernel, the following vulnerability has been resolved:\n\ndrm/xe/vm: Fix SVM leak on resv obj alloc failure in xe_vm_create()\n\nCommit 9e9787414882 (\"drm/xe/userptr: replace xe_hmm with gpusvm\") made\nxe_svm_init() unconditional in xe_vm_create() and extended it to also\ninitialize a \"simple\" gpusvm state for non-fault-mode VMs. The matching\nxe_svm_fini() call in xe_vm_close_and_put() was updated to run\nunconditionally, but the error unwind path in xe_vm_create() was not.\n\nOn the drm_gpuvm_resv_object_alloc() failure path, xe_svm_init() has\nalready succeeded but xe_svm_fini() is only called when\nXE_VM_FLAG_FAULT_MODE is set. For non-fault-mode VMs this leaves\nvm->svm.gpusvm partially initialized and leaks the resources allocated\nby drm_gpusvm_init().\n\nFor fault-mode VMs, xe_svm_init() additionally acquires the pagemap\nowner via drm_pagemap_acquire_owner() and the pagemaps via\nxe_svm_get_pagemaps(). Those resources are released by xe_svm_close(),\nnot xe_svm_fini(). On the same error path, xe_svm_close() is not\ncalled either, so fault-mode VMs leak the pagemap owner and pagemaps.\n\nFix both leaks:\n\n- Call xe_svm_fini() unconditionally on the err_svm_fini path, matching\n  the unconditional xe_svm_init() call. Move the vm->size = 0\n  assignment out of the conditional so the xe_vm_is_closed() assert in\n  xe_svm_fini() (and xe_svm_close()) holds for both modes.\n\n- Call xe_svm_close() for fault-mode VMs before xe_svm_fini(), matching\n  the ordering used in xe_vm_close_and_put().\n\n(cherry picked from commit ca2a3587d577ba764e0fe628fb676244fc33ddd4)"}],"metrics":[{"cvssV3_1":{"version":"3.1","vectorString":"CVSS:3.1/AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H","baseScore":7.8,"baseSeverity":"HIGH"},"scenarios":[{"lang":"en","value":"AV:L - Reached through the DRM_IOCTL_XE_VM_CREATE ioctl on /dev/dri/renderD*, which requires local access to the render node; there is no remote or network path into xe_vm_create().\nAC:L - The attacker induces the required drm_gpuvm_resv_object_alloc() ENOMEM by generating memory pressure while looping VM creates, and controls the second half of the bug directly by issuing another fault-mode VM create to walk the corrupted global peer list.\nPR:L - Any user with access to the render node (render/video group, typical for desktop, container, and Android GPU clients) can issue the ioctl; xe_vm_create_ioctl() performs no capability check.\nUI:N - The attacking process triggers the faulty error path entirely on its own via repeated ioctls; no action by another user is needed.\nS:U - The corruption stays within the kernel's own memory and security authority; no VM, IOMMU, or sandbox boundary is crossed.\nC:H - The freed xe_vm stays linked in the global xe_owner_list, so later drm_pagemap_acquire_owner() walks read cur_peer->owner and dereference container_of() pointers out of a reclaimable slab object, giving an attacker who sprays that object a kernel read primitive.\nI:H - The dangling list node lets list_add_tail() write through attacker-reclaimed next/prev pointers in the freed xe_vm, a classic linked-list write primitive that can be escalated to control-flow hijack and privilege escalation.\nA:H - Traversing the global owner list through a freed xe_vm reliably oopses or panics the kernel, and the leaked pagemap owner and xe_pagemap references permanently pin device memory resources."}]}],"affected":[{"product":"Linux","vendor":"Linux","defaultStatus":"unaffected","repo":"https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git","programFiles":["drivers/gpu/drm/xe/xe_vm.c"],"versions":[{"version":"9e978741488261e117bb50e5dfcf8e4080990958","lessThan":"279339aa8bdcf9db40094cf2bcbd495c53dbe817","status":"affected","versionType":"git"},{"version":"9e978741488261e117bb50e5dfcf8e4080990958","lessThan":"9ac92736030f3395d970c300eaeb59ac258a0c3e","status":"affected","versionType":"git"},{"version":"9e978741488261e117bb50e5dfcf8e4080990958","lessThan":"d2c6800ad1802bed72a6de1416536737f114f1d6","status":"affected","versionType":"git"}]},{"product":"Linux","vendor":"Linux","defaultStatus":"affected","repo":"https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git","programFiles":["drivers/gpu/drm/xe/xe_vm.c"],"versions":[{"version":"6.18","status":"affected"},{"version":"0","lessThan":"6.18","status":"unaffected","versionType":"semver"},{"version":"6.18.42","lessThanOrEqual":"6.18.*","status":"unaffected","versionType":"semver"},{"version":"7.1.6","lessThanOrEqual":"7.1.*","status":"unaffected","versionType":"semver"},{"version":"7.2","lessThanOrEqual":"*","status":"unaffected","versionType":"original_commit_for_fix"}]}],"cpeApplicability":[{"nodes":[{"operator":"OR","negate":false,"cpeMatch":[{"vulnerable":true,"criteria":"cpe:2.3:o:linux:linux_kernel:*:*:*:*:*:*:*:*","versionStartIncluding":"6.18","versionEndExcluding":"6.18.42"},{"vulnerable":true,"criteria":"cpe:2.3:o:linux:linux_kernel:*:*:*:*:*:*:*:*","versionStartIncluding":"6.18","versionEndExcluding":"7.1.6"},{"vulnerable":true,"criteria":"cpe:2.3:o:linux:linux_kernel:*:*:*:*:*:*:*:*","versionStartIncluding":"6.18","versionEndExcluding":"7.2"}]}]}],"references":[{"url":"https://git.kernel.org/stable/c/279339aa8bdcf9db40094cf2bcbd495c53dbe817"},{"url":"https://git.kernel.org/stable/c/9ac92736030f3395d970c300eaeb59ac258a0c3e"},{"url":"https://git.kernel.org/stable/c/d2c6800ad1802bed72a6de1416536737f114f1d6"}],"title":"drm/xe/vm: Fix SVM leak on resv obj alloc failure in xe_vm_create()","x_generator":{"engine":"bippy-1.2.0"}}}}