On Mon, Apr 30, 2001 at 02:06:46PM -0700, Peter Losher wrote:
> On Mon, 30 Apr 2001, Manuel Bouyer wrote:
> 
> > On Sun, Apr 29, 2001 at 03:08:04PM -0500, Dave Huang wrote:
> > > On Sun, 29 Apr 2001, Charles M. Hannum wrote:
> > > > I have some documentation on this chip, but no hardware, and no time to
> > > > do anything about it myself.  If someone else in interested in it...
> > >
> > > Does the siop driver not work with it? At least it claims to support the
> > > 53c1010 :)
> >
> > siop has some kind of support for it; only in Ultra2wide for now
> > (and only in -current).
> > However I've got reports of commands timeout under heavy load, and no
> > time to look at this for now. I suspect turning of tagged queuing would
> > workaround the problem ...
> 
> FWIW, here is the bug report I sent to Manuel:
> 
> -=-
> On Wed, 18 Apr 2001, Manuel Bouyer wrote:
> 
> > It's already in -current; could you try this ?
> > Note that it should only run at U2W (80MB/s) for now
> 
> It detects the card just fine; however under heavy I/O (like checking out
> a CVS tree or untarring the packages tree) it cracks:
> 
> -=-
> Apr 20 06:58:53 web1 /netbsd: root file system type: ffs
> Apr 20 06:58:53 web1 /netbsd: DMA IRQ: Illegal instruction dma fifo
> empty, DSP=0x3e0 DSA=0xf058: last msg_in=0x0 status=0xff
> Apr 20 06:58:53 web1 /netbsd: siop1: scsi bus reset
> Apr 20 06:58:53 web1 /netbsd: cmd 0xc0876000 (target 0:0) in reset list
> Apr 20 06:58:53 web1 /netbsd: cmd 0xc0876000 (status 2) about to be
> processed
> Apr 20 06:58:53 web1 /netbsd: siop1: target 0 using 16bit transfers
> Apr 20 06:58:53 web1 /netbsd: siop1: target 0 now synchronous at 40.0Mhz,
> offset 31
> -=-
> 
> It does this 5-6 times over several minutes then freezes...  Any ideas?

  I've had what I think is a similar problem with a 53c810.  Manuel suggested
turning off tagged queuing, and since I did that (about 1.5 months ago) the
problem hasn't recurred.  I was seeing command timeouts every couple of days,
with sequences of them resulting in an unrecoverable freeze about once a week.
A typical signature was:

Feb  2 23:42:15 x /netbsd: sd0(siop0:0:0): command timeout
Feb  2 23:42:15 x /netbsd: siop0: scsi bus reset
Feb  2 23:42:15 x /netbsd: cmd 0xc0427a00 (target 0:0) in reset list
Feb  2 23:42:15 x /netbsd: cmd 0xc04278c0 (target 0:0) in reset list
Feb  2 23:42:15 x /netbsd: cmd 0xc0427a80 (target 0:0) in reset list
Feb  2 23:42:15 x /netbsd: cmd 0xc0427980 (target 0:0) in reset list
Feb  2 23:42:15 x /netbsd: cmd 0xc0427a00 (status 2) about to be processed   
Feb  2 23:42:15 x /netbsd: cmd 0xc04278c0 (status 2) about to be processed
Feb  2 23:42:15 x /netbsd: cmd 0xc0427a80 (status 2) about to be processed   
Feb  2 23:42:15 x /netbsd: cmd 0xc0427980 (status 2) about to be processed
Feb  2 23:42:15 x /netbsd: siop0: target 0 now synchronous at 10.0Mhz, offset 8

--Jim