Subject: Re: generic HBA error messages on 1.6beta2
To: Jason R Thorpe <thorpej@wasabisystems.com>
From: Matthew Jacob <mjacob@feral.com>
List: port-alpha
Date: 07/10/2002 17:17:00
Note that hte problem exists in 1.5.X.

Why do you think that this is wrong?

If I get either EAGAIN or ENOMEM from dma mapping, I return an internal
CMD_EAGAIN. This normally sets the error to XS_RESOURCE_SHORTAGE and does a
channel freeze and calls scsipi_done, which is correct. The setting to
HBA_BOTCH is in case this is a *POLLED* command, where it would be correct to
fail the command.

This is better than what it was doing before (which was nothing). Why is this
in error and what is your basis for saying this rev 'broke' things?

If you have help to offer great. If you're swamped, don't fire from the hip-
it's just annoying.

-matt


On Wed, 10 Jul 2002, Jason R Thorpe wrote:

> On Wed, Jul 10, 2002 at 03:58:58PM -0700, Matthew Jacob wrote:
> 
>  > Jason- read the thread. The patches I offered for him to try already have
>  > this.
> 
> Sorry, I've been swamped, and haven't had much time to keep up.
> 
> In any case:
> 
> revision 1.80
> date: 2002/05/17 19:05:08;  author: mjacob;  state: Exp;  lines: +5 -4
> Slight touchup on last change- make sure HBA_BOTCH is set (just in case
> somebody at a different level does the wrong thing) on ENOMEM/EAGAIN returns
> from loading DMA.
> 
> This seems to be the revision that broke it.  Back out rev. 1.80 and try
> again.
> 
> -- 
>         -- Jason R. Thorpe <thorpej@wasabisystems.com>
>