Re: [PATCH] scsi: handle ABORTED_COMMAND on Fujitsu ETERNUS

[James Bottomley <jejb@linux.vnet.ibm.com>]

On Thu, 2018-01-18 at 11:17 +0100, Hannes Reinecke wrote:
> On 01/18/2018 03:43 AM, Martin K. Petersen wrote:
> > 
> > 
> > Martin,
> > 
> > > 
> > > You'd like to spend a precious BLIST bit for this single device
> > > which uses vendor-specific ASC/Q?
> > 
> > I really don't want string comparisons in the regular code paths.
> > Also not a fan of vendor-specific ASCs. But if you must use them,
> > please add a flag and trigger on that.
> > 
> > Since this is a bit of an unusual check condition combo, we could
> > entertain whether we should simply always ADD_TO_MLQUEUE on
> > 0xb/0xc1/0x1. I wonder what would break?
> > 
> That's the usual problem I have with vendor-specific sense codes.
> In general the risk is quite low of different vendors actually using
> the same code; I guess we can get away with just adding a debug
> message here and enable it without any vendor check.
> Then we can reconsider the whole thing once we notice several vendors
> using the same sense codes.

Murphy's law says that if we rely on Vendors to chose non-overlapping
numbers we'll end up with a clash fairly quickly ...

Can't we find a way of doing this in the device_handler?  That way we
can use vendor specific codes only when we know who the vendor is?

James