[illumos-Developer] WEBREV: 434 kernel crash dump on comstar iscsit

[Alexander Stetsenko]

On 02/17/2011 08:59 PM, Dan McDonald wrote:
> On Thu, Feb 17, 2011 at 07:22:27PM +0300, Alexander Stetsenko wrote:
>> This webrev seems to be stale.
>> Please, review it again.
>> http://cr.illumos.org/view/2y5oj5em/
> You've certainly made this code MUCH easier to read (by factoring out
> stmf_svc_timeout()).
>
> You did make one other change that may surprise users, unless the old
> behavior is part of the bug: the stmf_svc_queue() function now literally is a
> queue -->  you append new ones at the end.  Formerly, this was a stack (new
> ones placed at the front of the line).  It seems like a good fix to me,
> though.
Thanks.

Actually it was not a stack, there was 2 queues req and preq.
You right, requests were added at the front, but when you drain one 
queue (removing packets from front)
and  add to another queue (how it was in old code) they will be in 
reverse order :) right?

I think it would be a bug when new request was processed before any 
other pending request, wasn`t it?

> I say ship it.
>
> Dan