Subject: Re: ncr53c810 driver
To: Hal Murray <murray@pa.dec.com>
From: Manuel Bouyer <bouyer@antioche.lip6.fr>
List: port-alpha
Date: 06/10/2000 17:56:17
On Fri, Jun 09, 2000 at 11:54:36PM -0700, Hal Murray wrote:
> 
> > 'ncr' driver is dead.  long live 'siop'.. :)
> 
> I think I'm missing the big picture.  If ncr is buggy and siop is 
> the true way, why is ncr still used in so many of the Alpha config 
> files?  In particular, INSTALL and GENERIC. 
> 
>   siop is used in FRAU-FARBISSINA and YEAH-BABY
>   ncr is used in ALPHA, BABYLON, ENIWETOK GENERIC, INSTALL, MINI-ME, and RAWHIDE
> 
> 
> I tried to jump on the bandwagon and make a kernel that used siop 
> rather than ncr and this is what I got.  Is there something obvious 
> I did wrong?  (I just changed both places in my config file where 
> ncr occurred to siop.) 
> 
> 
> probe(siop0:6:0): command timeout

No error messages before this line ?

> siop0: scsi bus reset
> cmd 0xfffffe00000a47e8 (target 6) in reset list
> cmd 0xfffffe00000a47e8 about to be processed
> siop0: unhandled scsi interrupt, sist0=0x0 sist1=0x10 sstat1=0x8 DSA=0x829bce60
> DSP=0x58
> siop0: scsi bus reset
> siop0: unhandled scsi interrupt, sist0=0x40 sist1=0x10 sstat1=0xe DSA=0x829bcf00
>  DSP=0x20
> probe(siop0:6:0): command timeout
> siop0: scsi bus reset
> cmd 0xfffffe00000a4898 (target 6) in reset list
> cmd 0xfffffe00000a4898 about to be processed
> siop0: unhandled scsi interrupt, sist0=0x0 sist1=0x10 sstat1=0x8 DSA=0x829bcfa0
> DSP=0x58
> siop0: scsi bus reset
> siop0: unhandled scsi interrupt, sist0=0x40 sist1=0x10 sstat1=0xe DSA=0x829bd040
>  DSP=0x20

This doesn't make much sense. It seems the chip generated an interrupt;
the ISTAT register indicate it's an "SCSI interrupt". The interrupt reason
should be in sist0 (and eventually in sist1), but sist0 doesn't contains a
valid reason (0x40 is "command complete" and should't generate an interrupt).
Also, 0x10 for sist1 isn't valid (bits 3 to 7 are marked reserved in my docs).

What king of controller is it  (dmesg lines please) ?
What is on target 6 ? Does is successfully probe other devices before
target 6 ? Also, could you check in the SRM that the scsi controller is
*not* on target 6 ?

--
Manuel Bouyer <bouyer@antioche.eu.org>
--