From sape@iq.rulez.org Sun Jan 16 12:59:27 2000
Date: Wed, 12 Jan 2000 15:44:54 +0100 (CET)
From: Sasi Peter <sape@iq.rulez.org>
To: linux-kernel@vger.rutgers.edu
Subject: Re: Bug report (fwd)

Hi@

Here is a bug report for IDE extended twice, with updates.
I forward it here, because I thought I sent it to Andre, but now I think,
maybe I can better reach him here.

--  SaPE

Peter Sasi <sape@sch.hu>

---------- Forwarded message ----------
Subject: Re: Bug report

On Tue, 11 Jan 2000, Sasi Peter wrote:
> On Tue, 11 Jan 2000, Sasi Peter wrote:
> > This is supposed to be a bug report.
> > The setup:
> > disk 1: WDC AC420400D
> > disk 2: WDC WD68BA
> > disk 3: WDC AC420400D
> > disk 4: 32x Sony ATAPI CDROM
> > Abit BH6 MB, 256MB RAM, Celeron 400 (slotA), Abit HotRod
> > Yesterday:
> > Redhat 2.2.12-20 kernel (raid 0.90)
> > disks: 1 is hda, 2 is hdb, 3 is hdc, 4 is hdd
> > ribbon0 on ide0, ribbon1 on ide1, both 40 wire
> > hdd often reports media change, although no activity is present, and the
> > drive is empty.
> > hdc does not work 'tuned up': hdparm -d1 -c1 is reliable, but -u1 -m8
> > makes it have same DMA timeouts.
> > I suspected, that the ide1 cable was faulty
> > Today:
> > stock 2.2.14+raid0145-19990824-2.2.14+ide.2.2.14.20000108(use DMA if
> > avail., PIIX tuning enabled)
> > disks: 3 is hda, 4 is hdd, 2 is hde and 1 is hdg
> > 3 is udma33, 2 and 1 are udma66
> > ribbon0 on ide0, ribbon1 on ide1
> > two 80 wire (come with the HotRod) ribbons on ide2 and ide3
> > now hdg shows:
> > Jan 11 08:28:26 iq kernel: hdg: timeout waiting for DMA
> > Jan 11 08:28:26 iq kernel: hdg: irq timeout: status=0x58 { DriveReady
> > SeekComplete DataRequest }
> > Jan 11 08:28:26 iq kernel: hdg: DMA disabled
> > Jan 11 08:28:26 iq kernel: ide3: reset: success
> > If I reenable DMA for it, it has the same error in seconds, and DMA
> > disables again.
> One more think to report here:
> 
> I tried to use ide0=dma ide2=dma ide3=dma boot options, and thig got even
> worse:
> Even hde started to get a DMA timeout, and after that error message the
> box locked hard.

Even worse: with your 0111 patch it frrezes hard without a printed error,
just at the beggining fscking the volumes not unmounted clean. Currently I
have hdparm -d0 for all drives at the beginning of sysinit, so I can only
operate w/o dma, so the server function is onoperational.

Can you help w/ advices/some patch?

> > Otherwise the system is under considerable multimedia load (static video
> > and audio mpeg streams from the raid0 on the 3 drives), ~1TB/4day out on
> > my 100Mbps ethernet adapter.
> > 
> > One more weird thing: during the change once I booted the new kernel (DMA
> > and tuning enabled) with the old layout, all the disks on PIIX, and it did
> > an fsck (because of solely the mount count), and it just failed with a
> > short read on the last partition from hdc. Rebooting with the old kernel
> > it did the fsck again, and it passed without complaints.
> > 
> > What part of my system do you think causes all the mess?

Thanks in advance.

--  SaPE

Peter Sasi <sape@sch.hu>