Messages in this thread | | | Date | Tue, 12 Oct 1999 18:56:08 +0200 | From | Petr Vandrovec <> | Subject | NMI deadlock detection works |
| |
unfortunately... I just listen some mp3 songs played through es1938 from alsa 0.4.1b on 2.3.20. I decided to start some real work, so I typed
umount /asd
where /asd was dead ncpfs filesystem (server died...). Unfortunately, music start repeating last second (64KB?) and after few seconds I got: NMI Watchdog detected lockup on CPU1, registers: EIP: 0010:[<d288da71>] EFLAGS: 00000002 EAX: 1 EBX: cfad4244 ECX: 2 EDX: 0 ESI: 86 EDI: cfcafc04 EBP: 0 ESP: c8fafde4 (all hexa) DS: 18 ES: 18 SS: 18 Process umount (pid: 18965, stackpage = c8faf000) Stack: 10101 0 0 30002 1 d288b627 cfad4244 0 1 d29220b4 cfad4244 cfcafc04 c08f8000 d29221e8 cfad4244 CallTrace: c0157c22 : kfree_skbmem + x (after call to kmem_cache_free) c0115757 : timer_bh + 291 / 951 (just after return from __global_sti) c010afa5 : handle_IRQ_event + 81 (return from call intproc) c010b1bb : do_IRQ (just after return from handle_IRQ_event) c0109a2c : ret_from_intr + 0 c013bed1 : invalidate_list + 117 / 139 c013bf1c : invalidate_inodes + 52 / 111 (just after call to invalidate_list(inode_in_use)) c013174d : do_umount + 225 / 283 (call to ^) c0131828 : umount_dev + 160 / 291 (call to ^) c0131975 : sys_umount + 201 / 263 (call to ^) c01319c0 : sys_oldumount c0109990 : system_call + 52 Code: f6 83 bc 00 00 00 00 01 75 f7 e9 c9 da ff ff f6 86 bc 00 00 00 console shuts up...
I searched my module binaries and I found only one sequence of bytes in 'code'. It was in snd-es1938.o: testb $0x1,0xbc(%ebx) jne [line_above] jmp <snd_solo_trigger+20>
So just to let you know, I'll bother alsa developers after I find some way how to send bugreport through email and not through WWW interface. But I do not understand how EIP matches with stack trace. Or is it EIP from CPU1 and stack trace from CPU0? (*) I do not know, what was CPU0 doing when this happenned. Is it possible to get other CPU state through IPI or detect that another CPU NMI-oopsed, so do NMI-oops too (at worst at next tick...). Best regards, Petr Vandrovec vandrove@vc.cvut.cz
(*) I asked these question already about month ago, but no-one replied. But my oopses EIP do not match stack since I have SMP board.
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/
| |