From:   Thomas Graichen <news-innominate.list.sgi.xfs@innominate.de>
Reply-To: Thomas Graichen <thomas.graichen@innominate.com>
Subject: Re: kio on non-x86
Date:   2 Mar 2001 15:06:09 GMT
Organization: innominate AG, Berlin, Germany
Lines:  54
Distribution: local
Message-ID: <news2mail-97oct1$d7c$1@mate.bln.innominate.de>
References: <news-innominate.list.sgi.xfs@innominate.de> <200103021430.f22EUL929467@jen.americas.sgi.com>
User-Agent: tin/1.4.4-20000803 ("Vet for the Insane") (UNIX) (Linux/2.4.2-XFS (i686))
To:     linux-xfs@oss.sgi.com
Sender: owner-linux-xfs@oss.sgi.com
Precedence: bulk

Steve Lord <lord@sgi.com> wrote:
>> just updated the kernel on my ppc machine yesterday - worked all fine
>> so far ... that i tried the kio and kiocluster options on the ppc too
>> and ran into problems: as soon as i write to the fs it detects an io
>> error and shuts down ... thus a few questions:

> I presume the same devices worked OK without kio, did they work OK after
> this shutdown?

just did very short testings (will do more this evening) but i was
able to umount/mount them afterwards and xfs_repair was close to
happy with them - but more details later

without kio XFS on ppc seems to be as stable as on i386 due to my
testings (i am runing fine XFS root - never had any problems so
far)

>> * am i right in assuming that the kio stuff does not contain any
>>   arch specific parts so it should - in principle - work also on
>>   the pcc which would leave endian or gcc issues as a possible
>>   reason for the problem?

> I think this is true, there are no os specific components, where you using
> ide or scsi? The ide kio code is specific to having dma turned on. So which
> combination was this?

it's ide and dma turned on

>> * what is the state of XFS on sparc (also big endian)? - if it
>>   works there: do the kio and kiocluster options work on sparc?

> I would have to leave this to the list to answer - and by the way, there
> is no difference between these two options now, kiocluster will be going
> away I suspect.

yes i was remembering this somehow from the list - ok - it's the
same behaviour for both anyway :-)

>> * does anyone have an idea off-mind what might be the reson for
>>   this? ... i'll turn on all the debugging options and see if i
>>   get any more information out of it this evening i think

> Chait may have some suggestions, but the specific hardware combination is
> the most interesting part of the equation right now.

thanks - as said i waill (and can) do more testing this evening

t

-- 
thomas.graichen@innominate.com
                                                         innominate AG
                                                  the linux architects
tel: +49-30-308806-13   fax: -77             http://www.innominate.com