Re: [block IO crash] Re: 2.6.39-rc5-git2 boot crashs

[Ingo Molnar <mingo@elte.hu>]

* Ingo Molnar <mingo@elte.hu> wrote:

>  - Maybe i reproduced the original crash by pure luck.
> 
>    To exclude this i'll re-test once again with the non-debug .config.

I got a different problem this time around:

Fedora Core release 6 (Zod)
Kernel 2.6.39-rc5-i486-1sys+ on an i686

mercury login: ata1: lost interrupt (Status 0x58)
ata1.00: exception Emask 0x0 SAct 0x0 SErr 0x0 action 0x6 frozen
ata1.00: failed command: WRITE DMA EXT
ata1.00: cmd 35/00:38:56:cf:a5/00:01:05:00:00/e0 tag 0 dma 159744 out
         res 40/00:00:00:00:00/00:00:00:00:00/00 Emask 0x4 (timeout)
ata1.00: status: { DRDY }
ata1: soft resetting link
ata1: nv_mode_filter: 0x3f39f&0x3f39f->0x3f39f, BIOS=0x3f000 (0xc60000c0) ACPI=0x3f01f (20:600:0x13)
ata1.00: configured for UDMA/100
ata1.00: device reported invalid CHS sector 0
ata1: EH complete
ata1: lost interrupt (Status 0x58)
ata1.00: exception Emask 0x0 SAct 0x0 SErr 0x0 action 0x6 frozen
ata1.00: failed command: WRITE DMA EXT
ata1.00: cmd 35/00:38:56:cf:a5/00:01:05:00:00/e0 tag 0 dma 159744 out
         res 40/00:00:00:00:00/00:00:00:00:00/00 Emask 0x4 (timeout)
ata1.00: status: { DRDY }
ata1: soft resetting link
ata1: nv_mode_filter: 0x3f39f&0x3f39f->0x3f39f, BIOS=0x3f000 (0xc60000c0) ACPI=0x3f01f (20:600:0x13)
ata1.00: configured for UDMA/100
ata1.00: device reported invalid CHS sector 0
ata1: EH complete

Fedora Core release 6 (Zod)
Kernel 2.6.39-rc5-i486-1sys+ on an i686

mercury login: 
Fedora Core release 6 (Zod)
Kernel 2.6.39-rc5-i486-1sys+ on an i686

mercury login: ata1: lost interrupt (Status 0x58)
ata1.00: exception Emask 0x0 SAct 0x0 SErr 0x0 action 0x6 frozen
ata1.00: failed command: WRITE DMA EXT
ata1.00: cmd 35/00:38:56:cf:a5/00:01:05:00:00/e0 tag 0 dma 159744 out
         res 40/00:00:00:00:00/00:00:00:00:00/00 Emask 0x4 (timeout)
ata1.00: status: { DRDY }
ata1: soft resetting link
ata1: nv_mode_filter: 0x3f39f&0x3f39f->0x3f39f, BIOS=0x3f000 (0xc60000c0) ACPI=0x3f01f (20:600:0x13)
ata1.00: configured for UDMA/100
ata1.00: device reported invalid CHS sector 0
ata1: EH complete

The system is dead (is not performing any IO) after that.

Note that i reported similar problems early during this merge window, in this mail:

  [sporadic crash] blk: request botched

That sha1 was 89078d572eb9ce8d4c04264b8b0ba86de0d74c8f - that already had 
lockless SLUB:

  e8c500c2b64b: Merge branch 'slub/lockless' into for-linus

I never managed to pin that down, these IO problems were very sporadic in my 
test setup and never reproducible, until today.

	Ingo