Re: [Qemu-devel] [PATCH 07/11] blockdev: flip default cache mode from writethrough to writeback

[Paolo Bonzini]

Il 27/03/2013 16:16, Artyom Tarasenko ha scritto:
> This patch breaks shutting down of a sparc32 guest (or at least the
> Debian-4 image I have):
> 
> $ sparc-softmmu/qemu-system-sparc -M SS-5 -nographic -hda  ../disk-debian-4
> [...]
> Debian GNU/Linux 4.0 debian ttyS0
> 
> debian login: root
> Password:
> Linux debian 2.6.18-6-sparc32 #1 Tue Nov 10 00:31:37 UTC 2009 sparc
> # poweroff
> [...]
> Will now halt.
> Synchronizing SCSI cache for disk sda:
> esp0: Aborting command
> esp0: dumping state
> esp0: dma -- cond_reg<a4400010> addr<f000000b>
> esp0: SW [sreg<03> sstep<04> ireg<10>]
> esp0: HW reread [sreg<03> sstep<00> ireg<08>]
> esp0: current command [tgt<00> lun<00> pphase<MSGINDONE> cphase<CLUELESS>]
> esp0: disconnected
> esp0: Aborting command
> esp0: dumping state
> esp0: dma -- cond_reg<a4400010> addr<f000000b>
> esp0: SW [sreg<03> sstep<04> ireg<10>]
> esp0: HW reread [sreg<03> sstep<04> ireg<00>]
> esp0: current command [tgt<00> lun<00> pphase<UNISSUED> cphase<UNISSUED>]
> esp0: disconnected
> esp0: Resetting scsi bus
> esp0: SCSI bus reset interrupt
> esp0: no command in esp_handle()
> Kernel panic - not syncing: esp_handle: current_SC == penguin within 
> interrupt!
>  <0>Press Stop-A (L1-A) to return to the boot prom
> 
> 
> Without the patch, the line "Synchronizing SCSI cache for disk sda"
> doesn't come up, so the patch probably just unveils a bug somewhere
> else (esp?).

It doesn't come up because, with a writethrough cache, there is no need
to flush the cache.  The bug should be reproducible before this patch
with -drive file=../disk-debian-4,cache=writeback.

Paolo