Did "pmcstat -S instructions -T" on a 10-stable (ish) sandybridge xeon box and it paniced. Nov 3 13:51:11 <kern.crit> vxs162401 kernel: panic: [pmc,4186] cpu 3 didn't find a sample to collect Nov 3 13:51:11 <kern.crit> vxs162401 kernel: cpuid = 3 Nov 3 13:51:11 <kern.crit> vxs162401 kernel: KDB: stack backtrace: Nov 3 13:51:11 <kern.crit> vxs162401 kernel: db_trace_self_wrapper() at db_trace_self_wrapper+0x2b/frame 0xfffffe1 06dd0d840 Nov 3 13:51:11 <kern.crit> vxs162401 kernel: kdb_backtrace() at kdb_backtrace+0x39/frame 0xfffffe106dd0d8f0 Nov 3 13:51:11 <kern.crit> vxs162401 kernel: vpanic() at vpanic+0x126/frame 0xfffffe106dd0d930 Nov 3 13:51:11 <kern.crit> vxs162401 kernel: kassert_panic() at kassert_panic+0x136/frame 0xfffffe106dd0d9a0 Nov 3 13:51:11 <kern.crit> vxs162401 kernel: pmc_capture_user_callchain() at pmc_capture_user_callchain+0x124/fram e 0xfffffe106dd0d9f0 Nov 3 13:51:11 <kern.crit> vxs162401 kernel: pmc_hook_handler() at pmc_hook_handler+0x6bf/frame 0xfffffe106dd0dab0 Nov 3 13:51:11 <kern.crit> vxs162401 kernel: outofnmi() at outofnmi+0x23/frame 0xffff5be0 Nov 3 13:51:11 <kern.crit> vxs162401 kernel: Uptime: 4d14h51m21s From chat: 14:38 <@davide> hiren: I see nothing in the code preventing the thread to be rescheduled into another cpu from a first look 14:38 <@adrian> *burp* 14:38 <@davide> at the end of pmc_capture_user_callchain() you call sched_unpin() but there's no matching sched_pin() call Need to investigate more.
Hiren, is this PR still valid?