{"id":"CVE-2025-22094","state":"PUBLISHED","title":"powerpc/perf: Fix ref-counting on the PMU 'vpa_pmu'","description":"In the Linux kernel, the following vulnerability has been resolved:\n\npowerpc/perf: Fix ref-counting on the PMU 'vpa_pmu'\n\nCommit 176cda0619b6 (\"powerpc/perf: Add perf interface to expose vpa\ncounters\") introduced 'vpa_pmu' to expose Book3s-HV nested APIv2 provided\nL1<->L2 context switch latency counters to L1 user-space via\nperf-events. However the newly introduced PMU named 'vpa_pmu' doesn't\nassign ownership of the PMU to the module 'vpa_pmu'. Consequently the\nmodule 'vpa_pmu' can be unloaded while one of the perf-events are still\nactive, which can lead to kernel oops and panic of the form below on a\nPseries-LPAR:\n\nBUG: Kernel NULL pointer dereference on read at 0x00000058\n<snip>\n NIP [c000000000506cb8] event_sched_out+0x40/0x258\n LR [c00000000050e8a4] __perf_remove_from_context+0x7c/0x2b0\n Call Trace:\n [c00000025fc3fc30] [c00000025f8457a8] 0xc00000025f8457a8 (unreliable)\n [c00000025fc3fc80] [fffffffffffffee0] 0xfffffffffffffee0\n [c00000025fc3fcd0] [c000000000501e70] event_function+0xa8/0x120\n<snip>\n Kernel panic - not syncing: Aiee, killing interrupt handler!\n\nFix this by adding the module ownership to 'vpa_pmu' so that the module\n'vpa_pmu' is ref-counted and prevented from being unloaded when perf-events\nare initialized.","publishedAt":"2025-04-16T14:12:45.512Z","updatedAt":"2026-08-05T11:56:45.289Z","severity":{"score":7.3,"label":"High","version":"3.1","vector":"CVSS:3.1/AV:L/AC:L/PR:L/UI:R/S:U/C:H/I:H/A:H","source":"Linux"},"assigner":"Linux","products":["Linux Linux"],"affected":[{"vendor":"Linux","product":"Linux","defaultStatus":"unaffected","versions":[{"version":"176cda0619b6c17a553625f6e2fcbc3981ad667d","status":"affected","lessThan":"70ea7c5189197c6f5acdcfd8a2651be2c41e2faa","versionType":"git"},{"version":"176cda0619b6c17a553625f6e2fcbc3981ad667d","status":"affected","lessThan":"6cf045b51e2c5721db7e55305f09ee32741e00f9","versionType":"git"},{"version":"176cda0619b6c17a553625f6e2fcbc3981ad667d","status":"affected","lessThan":"ff99d5b6a246715f2257123cdf6c4a29cb33aa78","versionType":"git"}],"repo":"https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git","programFiles":["arch/powerpc/perf/vpa-pmu.c"]},{"vendor":"Linux","product":"Linux","defaultStatus":"affected","versions":[{"version":"6.13","status":"affected"},{"version":"0","status":"unaffected","lessThan":"6.13","versionType":"semver"},{"version":"6.13.11","status":"unaffected","lessThanOrEqual":"6.13.*","versionType":"semver"},{"version":"6.14.2","status":"unaffected","lessThanOrEqual":"6.14.*","versionType":"semver"},{"version":"6.15","status":"unaffected","lessThanOrEqual":"*","versionType":"original_commit_for_fix"}],"repo":"https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git","programFiles":["arch/powerpc/perf/vpa-pmu.c"]}],"references":["https://git.kernel.org/stable/c/70ea7c5189197c6f5acdcfd8a2651be2c41e2faa","https://git.kernel.org/stable/c/6cf045b51e2c5721db7e55305f09ee32741e00f9","https://git.kernel.org/stable/c/ff99d5b6a246715f2257123cdf6c4a29cb33aa78"],"weaknesses":[],"ingestedAt":"2026-10-29T20:08:39.238Z","source":{"repository":"CVEProject/cvelistV5","path":"cves/2025/22xxx/CVE-2025-22094.json","commit":"556c9d3a8cd17e218bb8e452ba7cd619a384ff47"},"raw":{"dataType":"CVE_RECORD","dataVersion":"5.2","cveMetadata":{"cveId":"CVE-2025-22094","state":"PUBLISHED","datePublished":"2025-04-16T14:12:45.512Z","dateUpdated":"2026-08-05T11:56:45.289Z","assignerShortName":"Linux","assignerOrgId":"416baaa9-dc9f-4396-8d5f-8c081fb06d67","dateReserved":"2024-12-29T08:45:45.818Z"},"containers":{"cna":{"title":"powerpc/perf: Fix ref-counting on the PMU 'vpa_pmu'","descriptions":[{"lang":"en","value":"In the Linux kernel, the following vulnerability has been resolved:\n\npowerpc/perf: Fix ref-counting on the PMU 'vpa_pmu'\n\nCommit 176cda0619b6 (\"powerpc/perf: Add perf interface to expose vpa\ncounters\") introduced 'vpa_pmu' to expose Book3s-HV nested APIv2 provided\nL1<->L2 context switch latency counters to L1 user-space via\nperf-events. However the newly introduced PMU named 'vpa_pmu' doesn't\nassign ownership of the PMU to the module 'vpa_pmu'. Consequently the\nmodule 'vpa_pmu' can be unloaded while one of the perf-events are still\nactive, which can lead to kernel oops and panic of the form below on a\nPseries-LPAR:\n\nBUG: Kernel NULL pointer dereference on read at 0x00000058\n<snip>\n NIP [c000000000506cb8] event_sched_out+0x40/0x258\n LR [c00000000050e8a4] __perf_remove_from_context+0x7c/0x2b0\n Call Trace:\n [c00000025fc3fc30] [c00000025f8457a8] 0xc00000025f8457a8 (unreliable)\n [c00000025fc3fc80] [fffffffffffffee0] 0xfffffffffffffee0\n [c00000025fc3fcd0] [c000000000501e70] event_function+0xa8/0x120\n<snip>\n Kernel panic - not syncing: Aiee, killing interrupt handler!\n\nFix this by adding the module ownership to 'vpa_pmu' so that the module\n'vpa_pmu' is ref-counted and prevented from being unloaded when perf-events\nare initialized."}],"affected":[{"vendor":"Linux","product":"Linux","defaultStatus":"unaffected","versions":[{"version":"176cda0619b6c17a553625f6e2fcbc3981ad667d","status":"affected","lessThan":"70ea7c5189197c6f5acdcfd8a2651be2c41e2faa","versionType":"git"},{"version":"176cda0619b6c17a553625f6e2fcbc3981ad667d","status":"affected","lessThan":"6cf045b51e2c5721db7e55305f09ee32741e00f9","versionType":"git"},{"version":"176cda0619b6c17a553625f6e2fcbc3981ad667d","status":"affected","lessThan":"ff99d5b6a246715f2257123cdf6c4a29cb33aa78","versionType":"git"}],"repo":"https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git","programFiles":["arch/powerpc/perf/vpa-pmu.c"]},{"vendor":"Linux","product":"Linux","defaultStatus":"affected","versions":[{"version":"6.13","status":"affected"},{"version":"0","status":"unaffected","lessThan":"6.13","versionType":"semver"},{"version":"6.13.11","status":"unaffected","lessThanOrEqual":"6.13.*","versionType":"semver"},{"version":"6.14.2","status":"unaffected","lessThanOrEqual":"6.14.*","versionType":"semver"},{"version":"6.15","status":"unaffected","lessThanOrEqual":"*","versionType":"original_commit_for_fix"}],"repo":"https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git","programFiles":["arch/powerpc/perf/vpa-pmu.c"]}],"references":[{"url":"https://git.kernel.org/stable/c/70ea7c5189197c6f5acdcfd8a2651be2c41e2faa"},{"url":"https://git.kernel.org/stable/c/6cf045b51e2c5721db7e55305f09ee32741e00f9"},{"url":"https://git.kernel.org/stable/c/ff99d5b6a246715f2257123cdf6c4a29cb33aa78"}],"metrics":[{"cvssV3_1":{"version":"3.1","vectorString":"CVSS:3.1/AV:L/AC:L/PR:L/UI:R/S:U/C:H/I:H/A:H","baseScore":7.3,"baseSeverity":"HIGH"},"scenarios":[{"lang":"en","value":"AV:L - The vulnerability is reached entirely through local syscalls — `perf_event_open(2)` to bind an event to the `vpa_pmu` PMU and `delete_module(2)` to unload the module. There is no network or remote-peer input path to this powerpc perf driver.\nAC:L - The attacker deterministically sets up the condition by holding a `vpa_pmu` perf event open indefinitely; there is no race to win and no memory layout beyond the attacker's influence, so any subsequent module unload reliably hits a live dangling event.\nPR:L - The missing module reference is taken (or rather, not taken) on behalf of an ordinary local user calling `perf_event_open` — no capability is checked in `vpa_pmu_event_init()`, and the `perf_allow_kernel()` gate is open to unprivileged users whenever `kernel.perf_event_paranoid <= 1`, a routine setting on the profiling/KVM-on-PowerVM hosts where this PMU is deployed.\nUI:R - Triggering the use-after-free requires the module to actually be unloaded, and `delete_module(2)` demands `CAP_SYS_MODULE` in the initial user namespace, which the low-privileged attacker cannot obtain via user namespaces — so an administrator's `rmmod vpa_pmu` (routine after a perf measurement session) is the required second-party action.\nS:U - The freed `struct pmu` and module text belong to the same kernel that the attacker's process already runs under, so impact stays within one security authority. Although `vpa_pmu` relates to KVM-on-PowerVM, the driver only runs on the L1 LPAR and the corruption never crosses the guest/host boundary.\nC:H - This is a use-after-free of a `struct pmu` living in vfree'd module memory, including a stale `pmu->pmu_disable_count` percpu pointer that is read on every event schedule; reclaiming that memory (e.g. by loading another module) turns the dangling reads into a kernel memory disclosure primitive.\nI:H - The dangling `event->pmu` still carries `->add`, `->del` and `->read` function pointers into freed module text, so the perf core makes indirect calls through attacker-influenceable reclaimed memory — a control-flow hijack primitive, not merely stale data.\nA:H - The commit documents the concrete outcome: a NULL pointer dereference in `event_sched_out()` reached from `__perf_remove_from_context()`, escalating to `Kernel panic - not syncing: Aiee, killing interrupt handler!` — a full system crash."}]}],"providerMetadata":{"shortName":"Linux","orgId":"416baaa9-dc9f-4396-8d5f-8c081fb06d67","dateUpdated":"2026-08-05T11:56:45.289Z"},"cpeApplicability":[{"nodes":[{"operator":"OR","negate":false,"cpeMatch":[{"vulnerable":true,"criteria":"cpe:2.3:o:linux:linux_kernel:*:*:*:*:*:*:*:*","versionStartIncluding":"6.13","versionEndExcluding":"6.13.11"},{"vulnerable":true,"criteria":"cpe:2.3:o:linux:linux_kernel:*:*:*:*:*:*:*:*","versionStartIncluding":"6.13","versionEndExcluding":"6.14.2"},{"vulnerable":true,"criteria":"cpe:2.3:o:linux:linux_kernel:*:*:*:*:*:*:*:*","versionStartIncluding":"6.13","versionEndExcluding":"6.15"}]}]}],"x_generator":{"engine":"bippy-1.2.0"}}}}}