Skip site navigation (1)Skip section navigation (2)
Date:      Thu, 22 Jul 1999 12:41:43 -0700 (PDT)
From:      Julian Elischer <julian@whistle.com>
To:        Dag-Erling Smorgrav <des@FreeBSD.org>
Cc:        cvs-committers@FreeBSD.org, cvs-all@FreeBSD.org
Subject:   Re: cvs commit: src/sys/pci ide_pci.c
Message-ID:  <Pine.BSF.3.95.990722123505.21011B-100000@current1.whistle.com>
In-Reply-To: <199907221134.EAA03813@freefall.freebsd.org>

next in thread | previous in thread | raw e-mail | index | archive | help
ERK!
I found it..
-       firstpage = DMA_PG_SZ - ((uintptr_t)vaddr & (DMA_PG_SZ));
+       firstpage = DMA_PG_SZ - ((uintptr_t)vaddr & (DMA_PG_SZ - 1));

gives the same result for page alligned transfers, but is definitly wrong
for unalligned transfers..
we must be doing all alligned transfers!

(I have a rack of 4 machines doing load testing with the bad code and all
working fine!)  at a guess, the failure must have been on an access to the
raw device.  and.... AHA! the Cx5530 can not do accesses not alligned to
16 byte boundaries, so we are not testing them.. they get done by PIO, and
not this code.  this explains why we can do heavy testing on it and not
see this bug

DUH!

pass pointy hat please....

julian
will fix in -current and remerge with fix..


On Thu, 22 Jul 1999, Dag-Erling Smorgrav wrote:

> des         1999/07/22 04:34:13 PDT
> 
>   Modified files:        (Branch: RELENG_3)
>     sys/pci              ide_pci.c 
>   Log:
>   Back out previous commit so IDE works again.
>   Whatever happened to testing before MFC?
>   
>   Revision  Changes    Path
>   1.28.2.2  +99 -124   src/sys/pci/ide_pci.c
> 
> 
> 



To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe cvs-all" in the body of the message




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?Pine.BSF.3.95.990722123505.21011B-100000>