Re: satalink0 error froze netbsd

[Eduardo Horvath <eeh%NetBSD.org@localhost>]

On Sat, 14 Jun 2014, Darren Reed wrote:

> An attempt to copy netbsd.gdb from /usr/src/sys/... to /var/crash
> produced this:
> 
> satalink0:0:0: bus-master DMA error: status=0x66
> wd0a: DMA error reading fsbn 34031136 of 34031136-34031167 (wd0 bn
> 34031136; cn 33761 tn 0 sn 48), retrying
> Stopped in pid 167.1 (cp) at    netbsd:cpu_Debugger+0x4:        nop
> db{1}> bt
> intr_list_handler(107678e18, 0, e0017ed0, 145e498, 1117a20, 146f5c0) at
> netbsd:intr_list_handler+0x24
> sparc_interrupt(0, 1408000, 147e000, 8021, 147ef70, 1468c68) at
> netbsd:sparc_interrupt+0x278
> db{1}> c
> Stopped in pid 167.1 (cp) at    netbsd:cpu_Debugger+0x4:        nop
> 
> and left the machine unresponsive except to BREAK on the console.
> 
> No further messages from satalink0 were seen. wd0 is, in this case, SSD
> rather than a HDD.

Not surprising.   Most drivers don't recover from DMA errors properly.  
They need to drain the operation and resubmit it, which most of them don't 
do.  The other option would be to panic the machine on a DMA error, but 
that would prevent drivers that do correct error recovery from continuing 
on.

Eduardo