Hi, (x86_64 / Xeon / perf-2.6.32-358.6.2)
The "-a" flag with perf record (and most friends except top) on RHEL 6.1 and 6.3 seems to cause a kernel panic on Intel Xeon- 12 and 24 core systems (only ones I tried so far). After about 1 second workstation completely freezes and then after maybe 10 seconds I see a panic. Since I realize I am probably using an ancient version of perf I am more trying to solicit some speculation or gather whether this is already a known issue with this particular software/hardware, backtrace combination? In the back trace the RIP points to __get_user_pages_fast, but I am unsure how to go further. Neither the perf-user archives or the git commit logs seem too helpful to me about this function as a keyword to search for. (http://article.gmane.org/gmane.linux.kernel.perf.user/665/match=__get_user_pages_fast -returns a thread but on PPC not x86. http://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/log/?qt=grep&q=__get_user_pages_fast I can't tell if any of the commit messages would fix this. However, I don't know what, "this" actually is or what's broken really. This is a back trace (of the 3 cores the RIP is in __get_user_pages_fast, different process every time. crash > log | grep BUG "BUG: unable to handle kernel NULL pointer dereference at 0000000000000050" crash> bt PID: 7558 TASK: ffff8801993614c0 CPU: 1 COMMAND: "fprintd" #0 [ffff88020bbef720] machine_kexec at ffffffff810310cb #1 [ffff88020bbef780] crash_kexec at ffffffff810b6312 #2 [ffff88020bbef850] oops_end at ffffffff814de190 #3 [ffff88020bbef880] no_context at ffffffff81040c9b #4 [ffff88020bbef8d0] __bad_area_nosemaphore at ffffffff81040f25 #5 [ffff88020bbef920] bad_area_nosemaphore at ffffffff81040ff3 #6 [ffff88020bbef930] __do_page_fault at ffffffff810416cd #7 [ffff88020bbefa50] do_page_fault at ffffffff814e017e #8 [ffff88020bbefa80] page_fault at ffffffff814dd525 [exception RIP: __get_user_pages_fast+156] RIP: ffffffff810464dc RSP: ffff88020bbefb38 RFLAGS: 00010046 RAX: 0000000000000006 RBX: 0000000000000000 RCX: ffff88020bbeffd8 RDX: 0000000000000000 RSI: 0000000000000000 RDI: 0000000000000000 RBP: ffff88020bbefb98 R8: ffff880028220000 R9: ffff88020bbefcf8 R10: ffff88020bbefde8 R11: 0000000000000001 R12: 0000000000000000 R13: 0000000000001000 R14: 0000008000000000 R15: 0000000000000fff ORIG_RAX: ffffffffffffffff CS: 0010 SS: 0018 #9 [ffff88020bbefba0] copy_from_user_nmi at ffffffff8101ae31 #10 [ffff88020bbefc00] perf_callchain_user at ffffffff8101b008 #11 [ffff88020bbefc50] perf_prepare_sample at ffffffff81106420 #12 [ffff88020bbefca0] __perf_event_overflow at ffffffff81108b38 #13 [ffff88020bbefd40] perf_swevent_overflow at ffffffff81108cd4 #14 [ffff88020bbefd80] do_perf_sw_event at ffffffff81108ee0 #15 [ffff88020bbefde0] perf_tp_event at ffffffff81108f94 #16 [ffff88020bbefe60] ftrace_profile_templ_sched_process_template at ffffffff8104d66a #17 [ffff88020bbefeb0] ftrace_profile_sched_process_exit at ffffffff8104d6d3 #18 [ffff88020bbefec0] do_exit at ffffffff8106c34d #19 [ffff88020bbeff40] do_group_exit at ffffffff8106c4b8 #20 [ffff88020bbeff70] sys_exit_group at ffffffff8106c547 #21 [ffff88020bbeff80] system_call_fastpath at ffffffff8100b172 RIP: 000000324eeabd98 RSP: 00007fff4b3a55c8 RFLAGS: 00010246 RAX: 00000000000000e7 RBX: ffffffff8100b172 RCX: 0000000000000000 RDX: 0000000000000000 RSI: 000000000000003c RDI: 0000000000000000 RBP: 0000000000000000 R8: 00000000000000e7 R9: ffffffffffffffa8 R10: 000000324f18e7e8 R11: 0000000000000246 R12: ffffffff8106c547 R13: ffff88020bbeff78 R14: 00000032504e4970 R15: 00000000020a1750 ORIG_RAX: 00000000000000e7 CS: 0033 SS: 002b "0xffffffff810464dc <__get_user_pages_fast+156>: add 0x50(%rdx),%r12" << what is this doing? > disassemble __get_user_pages_fast: 0xffffffff81046440 <__get_user_pages_fast+0>: push %rbp 0xffffffff81046441 <__get_user_pages_fast+1>: mov %rsp,%rbp 0xffffffff81046444 <__get_user_pages_fast+4>: push %r15 0xffffffff81046446 <__get_user_pages_fast+6>: push %r14 0xffffffff81046448 <__get_user_pages_fast+8>: push %r13 0xffffffff8104644a <__get_user_pages_fast+10>: push %r12 0xffffffff8104644c <__get_user_pages_fast+12>: push %rbx 0xffffffff8104644d <__get_user_pages_fast+13>: sub $0x38,%rsp 0xffffffff81046451 <__get_user_pages_fast+17>: nopl 0x0(%rax,%rax,1) 0xffffffff81046456 <__get_user_pages_fast+22>: mov %gs:0xcc00,%rax 0xffffffff8104645f <__get_user_pages_fast+31>: and $0xfffffffffffff000,%rdi 0xffffffff81046466 <__get_user_pages_fast+38>: movslq %esi,%r13 0xffffffff81046469 <__get_user_pages_fast+41>: mov %rcx,-0x50(%rbp) 0xffffffff8104646d <__get_user_pages_fast+45>: mov %edx,-0x44(%rbp) 0xffffffff81046470 <__get_user_pages_fast+48>: shl $0xc,%r13 0xffffffff81046474 <__get_user_pages_fast+52>: mov 0x480(%rax),%rdx 0xffffffff8104647b <__get_user_pages_fast+59>: mov %gs:0xcc08,%rcx 0xffffffff81046484 <__get_user_pages_fast+68>: movl $0x0,-0x34(%rbp) 0xffffffff8104648b <__get_user_pages_fast+75>: mov %rdi,%rax 0xffffffff8104648e <__get_user_pages_fast+78>: add %r13,%rax 0xffffffff81046491 <__get_user_pages_fast+81>: sbb %rbx,%rbx 0xffffffff81046494 <__get_user_pages_fast+84>: cmp %rax,-0x1fb8(%rcx) 0xffffffff8104649b <__get_user_pages_fast+91>: sbb $0x0,%rbx 0xffffffff8104649f <__get_user_pages_fast+95>: test %rbx,%rbx 0xffffffff810464a2 <__get_user_pages_fast+98>: jne 0xffffffff8104654d <__get_user_pages_fast+269> 0xffffffff810464a8 <__get_user_pages_fast+104>: add %rdi,%r13 0xffffffff810464ab <__get_user_pages_fast+107>: pushfq 0xffffffff810464ac <__get_user_pages_fast+108>: pop %rax 0xffffffff810464ad <__get_user_pages_fast+109>: nopl 0x0(%rax,%rax,1) 0xffffffff810464b2 <__get_user_pages_fast+114>: mov %rax,-0x58(%rbp) 0xffffffff810464b6 <__get_user_pages_fast+118>: cli 0xffffffff810464b7 <__get_user_pages_fast+119>: nopw 0x0(%rax,%rax,1) 0xffffffff810464bd <__get_user_pages_fast+125>: mov %rdi,%r12 0xffffffff810464c0 <__get_user_pages_fast+128>: lea -0x1(%r13),%r15 0xffffffff810464c4 <__get_user_pages_fast+132>: mov %rdi,%rsi 0xffffffff810464c7 <__get_user_pages_fast+135>: shr $0x24,%r12 0xffffffff810464cb <__get_user_pages_fast+139>: mov $0x8000000000,%r14 0xffffffff810464d5 <__get_user_pages_fast+149>: and $0xff8,%r12d 0xffffffff810464dc <__get_user_pages_fast+156>: add 0x50(%rdx),%r12 0xffffffff810464e0 <__get_user_pages_fast+160>: jmp 0xffffffff810464ef <__get_user_pages_fast+175> 0xffffffff810464e2 <__get_user_pages_fast+162>: nopw 0x0(%rax,%rax,1) 0xffffffff810464e8 <__get_user_pages_fast+168>: add $0x8,%r12 0xffffffff810464ec <__get_user_pages_fast+172>: mov %rbx,%rsi 0xffffffff810464ef <__get_user_pages_fast+175>: lea (%rsi,%r14,1),%rbx 0xffffffff810464f3 <__get_user_pages_fast+179>: mov $0xffffff8000000000,%rax 0xffffffff810464fd <__get_user_pages_fast+189>: mov (%r12),%rdi 0xffffffff81046501 <__get_user_pages_fast+193>: and %rax,%rbx 0xffffffff81046504 <__get_user_pages_fast+196>: lea -0x1(%rbx),%rax 0xffffffff81046508 <__get_user_pages_fast+200>: cmp %r15,%rax 0xffffffff8104650b <__get_user_pages_fast+203>: cmovae %r13,%rbx 0xffffffff8104650f <__get_user_pages_fast+207>: test %rdi,%rdi 0xffffffff81046512 <__get_user_pages_fast+210>: je 0xffffffff81046530 <__get_user_pages_fast+240> 0xffffffff81046514 <__get_user_pages_fast+212>: mov -0x50(%rbp),%r8 0xffffffff81046518 <__get_user_pages_fast+216>: mov -0x44(%rbp),%ecx 0xffffffff8104651b <__get_user_pages_fast+219>: lea -0x34(%rbp),%r9 0xffffffff8104651f <__get_user_pages_fast+223>: mov %rbx,%rdx 0xffffffff81046522 <__get_user_pages_fast+226>: callq 0xffffffff810460b0 <gup_pud_range> 0xffffffff81046527 <__get_user_pages_fast+231>: test %eax,%eax 0xffffffff81046529 <__get_user_pages_fast+233>: je 0xffffffff81046530 <__get_user_pages_fast+240> 0xffffffff8104652b <__get_user_pages_fast+235>: cmp %r13,%rbx 0xffffffff8104652e <__get_user_pages_fast+238>: jne 0xffffffff810464e8 <__get_user_pages_fast+168> 0xffffffff81046530 <__get_user_pages_fast+240>: mov -0x58(%rbp),%rdi 0xffffffff81046534 <__get_user_pages_fast+244>: push %rdi 0xffffffff81046535 <__get_user_pages_fast+245>: popfq 0xffffffff81046536 <__get_user_pages_fast+246>: nopl 0x0(%rax,%rax,1) 0xffffffff8104653b <__get_user_pages_fast+251>: mov -0x34(%rbp),%eax 0xffffffff8104653e <__get_user_pages_fast+254>: add $0x38,%rsp 0xffffffff81046542 <__get_user_pages_fast+258>: pop %rbx 0xffffffff81046543 <__get_user_pages_fast+259>: pop %r12 0xffffffff81046545 <__get_user_pages_fast+261>: pop %r13 0xffffffff81046547 <__get_user_pages_fast+263>: pop %r14 0xffffffff81046549 <__get_user_pages_fast+265>: pop %r15 0xffffffff8104654b <__get_user_pages_fast+267>: leaveq 0xffffffff8104654c <__get_user_pages_fast+268>: retq 0xffffffff8104654d <__get_user_pages_fast+269>: xor %eax,%eax 0xffffffff8104654f <__get_user_pages_fast+271>: jmp 0xffffffff8104653e <__get_user_pages_fast+254> End of assembler dump. crash> > whatis __get_user_pages_fast int __get_user_pages_fast(long unsigned int, int, int, struct page **); Thank you, Dennis O. -- To unsubscribe from this list: send the line "unsubscribe linux-perf-users" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html