{"dataType":"CVE_RECORD","dataVersion":"5.2","cveMetadata":{"cveId":"CVE-2026-99592","assignerOrgId":"416baaa9-dc9f-4396-8d5f-8c081fb06d67","state":"PUBLISHED","assignerShortName":"Linux","dateReserved":"2026-09-25T10:25:14.423Z","datePublished":"2026-11-20T05:49:20.304Z","dateUpdated":"2026-11-22T11:09:10.902Z"},"containers":{"cna":{"providerMetadata":{"orgId":"416baaa9-dc9f-4396-8d5f-8c081fb06d67","shortName":"Linux","dateUpdated":"2026-11-22T11:09:10.902Z"},"descriptions":[{"lang":"en","value":"In the Linux kernel, the following vulnerability has been resolved:\n\nx86/mm: Fix global ASID handed out to two processes\n\nOn AMD CPUs with INVLPGB, two live processes can get the same MAX_ASID_AVAILABLE\nglobal ASID value. The two then hit each other's TLB entries, so each can read\nor corrupt the other's memory.\n\nThe global ASID space is managed lazily. When a global ASID is freed, its bit is\nset in the global_asid_freed bitmap. Once the free ASID search reaches the end\nof the global_asid_used bitmap, a global flush is done to really reclaim the\nlazily freed global ASIDs.\n\nHowever, the current code can assign the same global ASID to two different\nprocesses, if one process is assigned MAX_ASID_AVAILABLE - 1, another process\nMAX_ASID_AVAILABLE, and both stick around (keeping their global ASID numbers in\nuse) until the search wraps around again.  At that point a third process could\nalso get assigned MAX_ASID_AVAILABLE.\n\n    the highest ASID stays in use, so it is not handed out again\n      last_global_asid stays below it, so no reset\n        the clear bits run out\n          find_next_zero_bit() returns MAX_ASID_AVAILABLE\n            global_asid_available > 0, so the failure is not caught\n              MAX_ASID_AVAILABLE is handed out as an ASID\n\nA search only looks at the bits below MAX_ASID_AVAILABLE, so it cannot see that\nthis value is in use. The next failed search hands it out again, even if the\nfirst process still holds it.\n\nThe fix used here is to reset the global ASID space after a failed search, and\nto search again. If that search fails too, no global ASID is handed out.\n\nThe test for !global_asid_available is removed, because use_global_asid() only\ncalls allocate_global_asid() if global_asid_available > 0. The remaining test is\nfor a condition that should be impossible.\n\n  [ riel: simplify code, massage changelog and comment ]"}],"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 - allocate_global_asid() runs only from use_global_asid() after consider_global_asid(), and that is called from flush_tlb_mm_range() when sys_munmap's unmap_region() reaches tlb_finish_mmu() with another CPU in mm_cpumask. The value that collides is the bitmap search result, not a field from any remote protocol.\nAC:L - The attacker fills and lazily frees global ASIDs by creating and exiting mms that mm_active_cpus_exceeds() accepts, repeating munmap until consider_global_asid() passes its pid/jiffies sample. After one live mm holds MAX_ASID_AVAILABLE, a later find_next_zero_bit(global_asid_used, MAX_ASID_AVAILABLE, last_global_asid) that fails while global_asid_available is still positive assigns that same value to a second mm.\nPR:L - sys_munmap and clone have no capability check, and use_global_asid() does not look at creds or the user namespace. An unprivileged caller that pins four of its own threads and shoots down its own pages reaches allocate_global_asid() whenever X86_FEATURE_INVLPGB is set.\nUI:N - The attacker creates the threads, issues the munmap calls, and exits the mms that set global_asid_freed. consider_global_asid() runs inside that same caller's flush_tlb_mm_range().\nS:U - Both colliding mms are host processes. switch_mm_irqs_off() installs the same kern_pcid(mm->context.global_asid) via build_cr3_noflush(), which aliases their TLB entries inside the kernel's own process isolation domain.\nC:H - choose_new_asid() returns the shared global ASID with need_flush clear, so load_new_mm_cr3() writes build_cr3_noflush(). A user load then hits the other mm's cached translation for that VA, and every page that process fills into the shared PCID is readable.\nI:H - A user store consumes the same non-flushed TLB entry. When the cached PTE is writable, the store updates the other mm's physical page, including its stack, code, or a MAP_SHARED file page.\nA:H - Writing the other mm's TLB-cached code or stack page kills that process. Leaving the first holder alive and repeating the failed search in allocate_global_asid() puts further mms on MAX_ASID_AVAILABLE, so each new collision can be crashed the same way."}]}],"affected":[{"product":"Linux","vendor":"Linux","defaultStatus":"unaffected","repo":"https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git","programFiles":["arch/x86/mm/tlb.c"],"versions":[{"version":"d504d1247e369e82c30588b296cc45a85a1ecc12","lessThan":"25bf240e1c755aaef670680a6de03f9f561c86f5","status":"affected","versionType":"git"},{"version":"d504d1247e369e82c30588b296cc45a85a1ecc12","lessThan":"02938c10d9e5c7631c8dbd6c48517805c1edf0a6","status":"affected","versionType":"git"},{"version":"d504d1247e369e82c30588b296cc45a85a1ecc12","lessThan":"5cdc3746fa163fb4e8d05fa3cabf46bc6aa7df75","status":"affected","versionType":"git"}]},{"product":"Linux","vendor":"Linux","defaultStatus":"affected","repo":"https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git","programFiles":["arch/x86/mm/tlb.c"],"versions":[{"version":"6.15","status":"affected"},{"version":"0","lessThan":"6.15","status":"unaffected","versionType":"semver"},{"version":"6.18.64","lessThanOrEqual":"6.18.*","status":"unaffected","versionType":"semver"},{"version":"7.3.5","lessThanOrEqual":"7.3.*","status":"unaffected","versionType":"semver"},{"version":"7.4-rc2","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":"6.15","versionEndExcluding":"6.18.64"},{"vulnerable":true,"criteria":"cpe:2.3:o:linux:linux_kernel:*:*:*:*:*:*:*:*","versionStartIncluding":"6.15","versionEndExcluding":"7.3.5"},{"vulnerable":true,"criteria":"cpe:2.3:o:linux:linux_kernel:*:*:*:*:*:*:*:*","versionStartIncluding":"6.15","versionEndExcluding":"7.4-rc2"}]}]}],"references":[{"url":"https://git.kernel.org/stable/c/25bf240e1c755aaef670680a6de03f9f561c86f5"},{"url":"https://git.kernel.org/stable/c/02938c10d9e5c7631c8dbd6c48517805c1edf0a6"},{"url":"https://git.kernel.org/stable/c/5cdc3746fa163fb4e8d05fa3cabf46bc6aa7df75"}],"title":"x86/mm: Fix global ASID handed out to two processes","x_generator":{"engine":"bippy-1.2.0"}}}}