Bug 7294 - ali15x3 does not allow DMA mode access on sparc64
Summary: ali15x3 does not allow DMA mode access on sparc64
Status: REJECTED INSUFFICIENT_DATA
Alias: None
Product: IO/Storage
Classification: Unclassified
Component: IDE (show other bugs)
Hardware: i386 Linux
: P2 normal
Assignee: Bartlomiej Zolnierkiewicz
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2006-10-09 16:38 UTC by Zachary Shaver
Modified: 2007-01-02 12:05 UTC (History)
2 users (show)

See Also:
Kernel Version: 2.6.15
Subsystem:
Regression: ---
Bisected commit-id:


Attachments

Description Zachary Shaver 2006-10-09 16:38:43 UTC
Most recent kernel where this bug did not occur:
Distribution: Ubuntu Linux Dapper Drake 6.06 LTS sparc64 / Gentoo Linux
Hardware Environment: Sun Blade 100
Software Environment: 
Problem Description: The ALI15X3 IDE controller cannot access drives in DMA mode
with presend driver drivers/ide/pci/alim15x3.c , as a result performance on the
affected sparc64 systems is very minimal. 

At least on my Sun Blade 100, and i have seen reports other places on the
internet, that without passing the argument ide=nodma to the kernel... this box
can't even boot due to this bug.

This was supposed to be fixed back in 2003? Apparently the below link (see
below) has a proposed solution for 2.4 kernel tree, but was this fix ever
applied to 2.6?

http://www.ussg.iu.edu/hypermail/linux/kernel/0307.2/0620.html

Steps to reproduce: Install any major linux distribution on a Sun Blade 100 (or
any sparc64) system with the ALI15x3 IDE controller.
Comment 1 Zachary Shaver 2006-10-09 16:39:57 UTC
The machine is VERY, VERY slow with only PIO mode access...
Comment 2 Zachary Shaver 2006-10-11 08:55:51 UTC
Let me clarify, when accessing a drive using the ali15x3 driver in sparc64,
severe data corruption occurs to the point that a system cannot boot under dma.
It appears it CAN access the drive, but if you actually try to do anything it
pukes because it keeps corrupting the data.
Comment 3 Alan 2006-10-11 09:42:01 UTC
The ALi documentation I posess does not deal with its interface to Sparc systems
only ALi chipsets.
Comment 4 Zachary Shaver 2006-10-14 00:26:30 UTC
Seems to be working fine in freebsd, i dunno if that's any help though because 
i'm not a developer.
Comment 5 Alan 2006-12-01 12:22:43 UTC
Bartlomiej is now back so I'm transferring this bug to the ide maintainer. If
the new pata_ali driver also shows the same problems please let me know however.
Comment 6 Bartlomiej Zolnierkiewicz 2006-12-16 15:34:52 UTC
According to follow-up email
http://www.ussg.iu.edu/hypermail/linux/kernel/0307.2/0812.html
the fix (in this or other form) was ACK-ed by Alan and applied to 2.5
but I don't know whether this is indeed happened.

Does booting with "ide0=autotune ide1=autotune" (this is example valid for
ide0/1 only) kernel parameters help?
Comment 7 Zachary Shaver 2006-12-28 18:42:48 UTC
I will install gentoo this week on the sun blade 100 with the autotune 
argument and follow up with this bug.
Comment 8 Zachary Shaver 2007-01-01 17:35:51 UTC
i changed a drive and the problem seems to have corrected itself
it was a maxtor drive, perhaps corrupting the ide channel?
anyways... what was really weird is that freebsd worked with it just fine but 
when it was plugged into linux it puked
anyways... that drive will plague me no more.

So either 1 or another thing happened.... either the drive was cooked and by 
replacing it i solved the problem... OR... the version of ubuntu linux i 
installed does not have this issue.

Linux ubuntu 2.6.15-27-sparc64 #1 Fri Dec 8 17:53:06 UTC 2006 sparc64 GNU/Linux
Comment 9 Bartlomiej Zolnierkiewicz 2007-01-02 12:05:45 UTC
OK, thanks for reporting.

Note You need to log in before you can comment on or make changes to this bug.