xfs
[Top] [All Lists]

Re: XFS mount, dma problem

To: "linux" <dahouet@xxxxxxx>, "Mihai RUSU" <dizzy@xxxxxxxxx>
Subject: Re: XFS mount, dma problem
From: Seth Mos <knuffie@xxxxxxxxx>
Date: Mon, 24 Dec 2001 00:21:48 +0100
Cc: linux-xfs@xxxxxxxxxxx
In-reply-to: <200112231312560614.02600378@mail.gmx.de>
References: <Pine.LNX.4.33.0112231356040.14369-100000@ahriman.bucharest.roedu.net> <Pine.LNX.4.33.0112231356040.14369-100000@ahriman.bucharest.roedu.net>
Sender: owner-linux-xfs@xxxxxxxxxxx
At 13:12 23-12-2001 +0100, linux wrote:
Le 23/12/2001 à 13:58 Mihai RUSU a écrit :

>have you tried using a rh 7.2 default kernel ? without XFS? im pretty sure
>it will generate the same messages

i can't make any test without XFS
because my system is full XFS.


>your kernel is heving problems enabling dma, thats NOT a XFS problem >try to compile your own kernel with support for your I/O chipset >and then try use hdparm to see what problems you have with dma

i use the same .config for the two version of kernel
i think there's two problems on two drives :

The 2.4.5 kernel had a lot of VIA chipset fixes but there were some more fixes that went into later kernels that solved most of the bugs including disk corruption. The error is not fatal, it is just switching to a lower speed to prevent problems.


This is not XFS related at all but a generic kernel thing. A lot of people were bitten by VIA chipset bugs and earlier kernels then 2.4.5 had DMA disabled per default on all VIA chipsets to prevent fs corruption.

The error also indicates CRC errors to which I recommend checking your cables. I got rid of those errors on a production machine by replacing the UDMA IDE cables with new ones.

If you don't have a secondary slave (hdd) it is normal for that port to disable DMA.

Cheers

--
Seth
Every program has two purposes one for which
it was written and another for which it wasn't
I use the last kind.


<Prev in Thread] Current Thread [Next in Thread>