{"id":"CVE-2026-93275","published":"2026-09-24T16:17:24.667","lastModified":"2026-09-24T16:17:24.667","description":"In the Linux kernel, the following vulnerability has been resolved:\n\nperf/x86/intel/pt: Fix stop/start with no update\n\nIf pt_event_stop() is called without PERF_EF_UPDATE flag, then\nperf_aux_output_end() is not called.  A subsequent call to pt_event_start()\nwill call perf_aux_output_begin() again which violates the rule against\nnesting and triggers a WARNING in perf_aux_output_begin().\n\nOriginally, pt_event_stop() was never called without PERF_EF_UPDATE,\nbecause the only code paths to do so are from event overflow, and Intel PT\ndoes not do that.\n\nHowever the introduction of group throttling by commit 9734e25fbf5ae\n(\"perf: Fix the throttle logic for a group\") meant that an Intel PT event\ncould be throttled if it was part of a group.  Throttling calls PMU\n ->stop() / ->start() callbacks without flags.\n\nAn example is when AUX area sampling is used. The following commands\nhit the issue:\n\n  echo 10000 > /proc/sys/kernel/perf_event_max_sample_rate\n\n  perf record -F32000 --aux-sample -e '{intel_pt//u,cycles:u}' \\\n  -- bash -c 'for i in `seq 1 100000` ; do true ; done'\n\nUse PERF_HES_UPTODATE to track whether perf_aux_output_begin() and\nperf_aux_output_end() are balanced.  A cleared PERF_HES_UPTODATE bit\nindicates that an AUX output context is still open.\n\nAmend pt_event_start() / pt_event_stop() accordingly so that begin/end\nstay balanced:\n\n - In non-snapshot mode, stop() always closes the buffer (the buffer may\n   have run out of space, and that accounting is done by the update), so\n   a following start() opens a fresh one as before.\n\n - In snapshot/overwrite mode, stop() without PERF_EF_UPDATE leaves the\n   buffer open so that pt_event_snapshot_aux() can still copy from it,\n   and start() then only re-enables tracing instead of calling\n   perf_aux_output_begin() again.\n\nNote that pt_event_del() calls pt_event_stop() with PERF_EF_UPDATE flag set\n(as is required by the documentation), so a final call to\nperf_aux_output_end() is assured.","cvssScore":null,"cvssSeverity":null,"cvssVector":null,"cwes":[],"vendors":[],"products":[],"references":[{"url":"https://git.kernel.org/stable/c/2e17bf3a469a41457a3bc31b1f8fd66b6ce94a6d","tags":[]},{"url":"https://git.kernel.org/stable/c/304844b5545f444d523286da8228a1cbc46ffffa","tags":[]},{"url":"https://git.kernel.org/stable/c/9b919fb5512f256e9e0604faf960b2bb98aab454","tags":[]},{"url":"https://git.kernel.org/stable/c/cab5015ee3f25f789b4a368f2abfa65536576b78","tags":[]}],"exploitRefs":[],"hasPoc":false,"ai":null}