Re: [PATCH] sd.c fixes applied incorectly to 2.4.19-preXX

[James Bottomley <James.Bottomley@SteelEye.com>]

gibbs@scsiguy.com said:
> In my opinion, the only way to really fix the Linux SCSI layer is to
> rip it out and start over using prior art from systems with functional
> I/O subsystems as a guide to building something enterprise worthy.  As
> soon as the community becomes serious about such an endevor, so will
> I. 8-) 

Not perhaps for ripping it up and starting again.  However, there are more 
ways than one to skin a cat.  If you can outline a set of principles, it may 
be possible to move current development towards them.  How about just the five 
worst things about the current subsystem and how we could address them?

> If you insist on doing this, lets at least preserve the current
> behavior:

> 	/*
> 	 * FIXME: this driver treats deferred errors are fatal for
[...]
> 	if (scsi_sense_valid(SCpnt)
> 	 && scsi_sense_key(SCpnt) == SKEY_CURRENT_ERROR) {

Excellent!  It's a lot more readable for people not versed in sense buffer 
layout, thanks!

James


> I don't believe there are definitions to support the second part of
> the if clause, but there could be.