Re: bomb and force shutdown again

[ASANO Masahiro <masano@tnes.nec.co.jp>]

Hi,

> The disk error handling of the recent xfs cvs kernel (20020323, 2.4.18)
> seems broken.

Additional information:

 CVS 2002-02-24   ok
 CVS 2002-03-13   stalled after fs shutdown
 CVS 2002-03-23   stalled after fs shutdown

> I tested xfs with BOMB patch I previously posted to this list.
> After force shutdown by the bomb, the system was hanged and I could
> not enter KDB at my single CPU box.

On SMP box, I could use KDB.
# but I don't have serial console...

[0]kdb> ps
(snip)
0xc1326000 00000008 00000001 1 001 run 0xc1326370 pagebuf_daemon
(snip)
0xc3c0c000 00008958 00000842 1 000 run 0xc3c0c370*bash
[0]kdb> bt
   flush_tlb_others             <- loop "while (flush_cpumask) ;"
   flush_tlb_page
   do_wp_page
   handle_mm_fault
   change_termios
   error_code
Interrupt registers:
(snip)
Interrupt from user space, end of kernel trace
[0]kdb> btp 8
   _text_lock_page_buf          <- maybe from pagebuf_rele()
   pagebuf_iodone
   xfs_bioerror
   pagebuf_daemon
   kernel_thread
[0]kdb> cpu
Currently on cpu 0
Available cpus: 0,1

** reboot and try again.

> ps
(snip)
0xc1322000 00000008 00000001 1 000 run 0xc1322370 pagebuf_daemon
(snip)
0xc9d40000 00000887 00000886 1 001 run 0xc9d40370*bash
> bt
   flush_tlb_others
   flush_tlb_mm
   copy_mm
   do_fork
   sys_fork
   system_call
> btp 8
   _text_lock_page_buf
   pagebuf_iodone
   xfs_bioerror
   xfs_bdstrat
   pagebuf_daemon
   kernel_thread
> cpu
Currently on cpu 1
Available cpus: 0,1


It looks like page_buf_t issue, but I could not find what led to this.
Thanks in advance.
--
Masano