From owner-freebsd-current  Wed Mar 15 10:44:56 1995
Return-Path: current-owner
Received: (from majordom@localhost) by freefall.cdrom.com (8.6.10/8.6.6) id KAA07080 for current-outgoing; Wed, 15 Mar 1995 10:44:56 -0800
Received: from gndrsh.aac.dev.com (gndrsh.aac.dev.com [198.145.92.241]) by freefall.cdrom.com (8.6.10/8.6.6) with ESMTP id KAA07071 for <current@FreeBSD.org>; Wed, 15 Mar 1995 10:44:52 -0800
Received: (from rgrimes@localhost) by gndrsh.aac.dev.com (8.6.8/8.6.6) id KAA13156; Wed, 15 Mar 1995 10:31:13 -0800
From: "Rodney W. Grimes" <rgrimes@gndrsh.aac.dev.com>
Message-Id: <199503151831.KAA13156@gndrsh.aac.dev.com>
Subject: Re: scsi_sense
To: dufault@hda.com (Peter Dufault)
Date: Wed, 15 Mar 1995 10:31:13 -0800 (PST)
Cc: charnier@lirmm.fr, current@FreeBSD.org
In-Reply-To: <199503151812.NAA00750@hda.com> from "Peter Dufault" at Mar 15, 95 01:12:14 pm
X-Mailer: ELM [version 2.4 PL23]
Content-Type: text
Content-Length: 1404      
Sender: current-owner@FreeBSD.org
Precedence: bulk

> 
> Philippe Charnier writes:
> > 
> > Hello,
> > 
> > 
> > from scsi_sence.c (in src/sys/scsi)
> > 
> >         {0x26, 0x03, "Threshold parameters not supported" },
> >         {0x46, 0x00, "Unsuccessful soft reset" },
> > #endif /* NO_SCSI_SENSE */
> >         {0xff, 0xff, 0 },
> > };
> > 
> > char *scsi_sense_desc(int asc, int ascq)
> > {
> >         int i;
> >         for (i = 0; i < sizeof(tab) / sizeof(tab[0]); i++)
> >                 if (tab[i].asc == asc && tab[i].ascq == ascq)
> >                         return tab[i].desc;
> > 
> >         return "no available sense description";
> > }
> > 
> > Why this code don't use {0xff, 0xff, 0 } as a end of list?
> 
> Because we know the table size.

I agree with Peter here, the way he has coded it is more efficent.

But I do have a concert about the 0xff, 0xff, 0 at the end of the
table.  What happens if (and don't say no device should ever return
this value you never know what might go wrong) a device returns
asc=0xff, ascq=0xff.  We end up returning a null pointer, this is
not good :-(.

How about:
	{0xff, 0xff, "Sense values are invalid (0xFF, 0xFF)"},

That way if this boundary condition is ever seen atleast we do
return a pointer to a string that does tell what is wrong!


-- 
Rod Grimes                                      rgrimes@gndrsh.aac.dev.com
Accurate Automation Company                   Custom computers for FreeBSD