> Eyal Lebedinsky wrote:
>
> I now changed my test and not only hammered the raid, but also hammered
> hda directly. Withing a few minutes I got a DMA timeout of hdb, both
> hda and hdb were reset, the raid went into degraded mode, and I needed
> to power the machine down/up before hdb was properly recognised. I
> re-added hdb to the raid, which reconstructed it and brought it online.
>
> In other words there surely is something wrong here. This setup is not
usable.
>
> I will next
> - flash the latest BIOS
> - build latest kernel (2.4.4-ac17 last I checked)
> and it this is unstable then I will have to assume a mobo replacement is
required.
>
> So, can anyone in the know give a straight answer to:
>
> Is the KT7-RAID compatible with Linux kernel 2.4?
It ought to work, as long as you don't expect to use the RAID features of
the hardware, because they require support from a binary driver, Abit made
available for Win. There's a general problem with I/O in Linux that only 1
bit of error information is returned back up, so any error will cause RAID
to take that bit of the RAID offline, even if it's actually a soft error.
Discussions have already started http://www.osdn.com/conferences/kernel/ on
improving this in future in 2.5.
Something that has helped ppl during thrash tests with the HPT366, is not to
use the highest UDMA modes, but UDMA mode3, or UDMA mode2. As your
configuration IIRC had disks on their own channels, it probably wouldn't
affect performance as their transfer rates won't reach UDMA mode4. There
are black lists in the driver code, to handle known bad combinations of
controllers, drives and UDMA modes.
The HPT366 has always been a problematic beast, it's quite entertaining to
see all the special case stuff Andre Hedrik had to put in to handle it, and
account for reported issues with some IBM drives. I think that DMA timeout
messages has been reported before, http://marc.theaimsgroup.com have a
searchable archive of this mailing list.
The HPT370 I don't know much about, I believe that Abit, lost a lot of
credibility in the Linux world over the BP6 and the Gentus distro probably
didn't help matters, there's been very little KT7 & KT7-RAID traffic on this
list. This may be because the boards have been stable and generally
forgiving. Call me a cynic if you like, but I doubt that somehow, I would
have at least expected there to have been queries, about RH6.2 not booting
due Athlon's lack of the PIII processor ID, and how to enable the HPT370
with 2.2.16, 2.2.17 etc. if the board was widely used for Linux. Any board
with an HPT366 on it, is probably best avoided if you prize stability.
Perhaps you should try your board and tests with a 2.2.19 release, as it may
prove more solid than 2.4 at present. Hope this helps even if it doesn't
directly answer your quesion,
Rob
--
=- To unsubscribe, email [EMAIL PROTECTED] with the -=
=- body of "unsubscribe linux-abit". -=
