ide-cd vs. DMA

ide-cd vs. DMA

[Benjamin Herrenschmidt]

Hi Jens !

Yet another case of ide-cd bogosly switching DMA of on ATAPI errors...
can this be fixed one day ? :)

The error this time is

hdb: command error: status=0x51 { DriveReady SeekComplete Error }
hdb: command error: error=0x54 { AbortedCommand LastFailedSense=0x05 }
ide: failed opcode was: unknown
end_request: I/O error, dev hdb, sector 42872

Ben.

Re: ide-cd vs. DMA

[Karim Yaghmour]

Benjamin Herrenschmidt wrote:
> hdb: command error: status=0x51 { DriveReady SeekComplete Error }
> hdb: command error: error=0x54 { AbortedCommand LastFailedSense=0x05 }
> ide: failed opcode was: unknown
> end_request: I/O error, dev hdb, sector 42872

Got plenty of these an old Dell Optiplex GX1 (PIII-450) with
vanilla FC3. ... you've got to wonder when the kernel says there
are bad sectors on a CD (?) and then they disappear with:
hdparm -d0 /dev/hdc

Karim
-- 
Author, Speaker, Developer, Consultant
Pushing Embedded and Real-Time Linux Systems Beyond the Limits
http://www.opersys.com || karim@opersys.com || 1-866-677-4546

Re: ide-cd vs. DMA

[Benjamin Herrenschmidt]

On Mon, 2005-05-23 at 23:27 -0400, Karim Yaghmour wrote:
> Benjamin Herrenschmidt wrote:
> > hdb: command error: status=0x51 { DriveReady SeekComplete Error }
> > hdb: command error: error=0x54 { AbortedCommand LastFailedSense=0x05 }
> > ide: failed opcode was: unknown
> > end_request: I/O error, dev hdb, sector 42872
> 
> Got plenty of these an old Dell Optiplex GX1 (PIII-450) with
> vanilla FC3. ... you've got to wonder when the kernel says there
> are bad sectors on a CD (?) and then they disappear with:
> hdparm -d0 /dev/hdc

Well, not sure what's wrong here, but ATAPI errors shouldn't normally
result in stopping DMA. We may want to just blacklist your drive rather
than having this stupid fallback. In this case, I suspect it's
CSS/region issue with a DVD.

Ben.

Re: ide-cd vs. DMA

[Karim Yaghmour]

Benjamin Herrenschmidt wrote:
> Well, not sure what's wrong here, but ATAPI errors shouldn't normally
> result in stopping DMA. We may want to just blacklist your drive rather
> than having this stupid fallback. In this case, I suspect it's
> CSS/region issue with a DVD.

Here's a little bit more info:

Here's from dmesg:
hdc: SAMSUNG SC-140B, ATAPI CD/DVD-ROM drive
hdc: ATAPI 40X CD-ROM drive, 128kB Cache, UDMA(33)

hdparm -i /dev/hdc:

/dev/hdc:

 Model=SAMSUNG SC-140B, FwRev=d005, SerialNo=
 Config={ SpinMotCtl Removeable DTR<=5Mbs DTR>10Mbs nonMagnetic }
 RawCHS=0/0/0, TrkSize=0, SectSize=0, ECCbytes=0
 BuffType=unknown, BuffSize=0kB, MaxMultSect=0
 (maybe): CurCHS=0/0/0, CurSects=0, LBA=yes, LBAsects=0
 IORDY=on/off, tPIO={min:120,w/IORDY:120}, tDMA={min:120,rec:120}
 PIO modes:  pio0 pio1 pio2 pio3 pio4
 DMA modes:  sdma0 sdma1 sdma2 mdma0 mdma1 mdma2
 UDMA modes: udma0 udma1 *udma2
 AdvancedPM=no

Here's what happens on the first mount attempt:
hdc: DMA interrupt recovery
hdc: lost interrupt
hdc: status timeout: status=0xd0 { Busy }
hdc: status timeout: error=0x00
hdc: DMA disabled
hdc: drive not ready for command
hdc: ATAPI reset complete

Now, if I'm stuborn and re-enable DMA using a "hdparm -d1 /dev/hdc" and then
try again, now I get:
hdc: media error (bad sector): status=0x51 { DriveReady SeekComplete Error }
hdc: media error (bad sector): error=0x34
ide: failed opcode was 100
end_request: I/O error, dev hdc, sector 16

The last error being repeated ad-nauseam for every "sector" on the
disk. Again, note that this is a CD drive, not a hard disk.

Here's from lspci:
00:00.0 Host bridge: Intel Corp. 440BX/ZX/DX - 82443BX/ZX/DX Host bridge (rev 03)
00:01.0 PCI bridge: Intel Corp. 440BX/ZX/DX - 82443BX/ZX/DX AGP bridge (rev 03)
00:07.0 ISA bridge: Intel Corp. 82371AB/EB/MB PIIX4 ISA (rev 02)
00:07.1 IDE interface: Intel Corp. 82371AB/EB/MB PIIX4 IDE (rev 01)
00:07.2 USB Controller: Intel Corp. 82371AB/EB/MB PIIX4 USB (rev 01)
00:07.3 Bridge: Intel Corp. 82371AB/EB/MB PIIX4 ACPI (rev 02)
00:11.0 Ethernet controller: 3Com Corporation 3c905B 100BaseTX [Cyclone] (rev 24)
01:00.0 VGA compatible controller: ATI Technologies Inc 3D Rage Pro AGP 1X/2X (rev 5c)

In attempting to isolate the problem, I ran into a spurious issue with
another drive I have in that machine:
hda: dma_timer_expiry: dma status == 0x61
hda: DMA timeout error
hda: dma timeout error: status=0x58 { DriveReady SeekComplete DataRequest }
hda: dma_timer_expiry: dma status == 0x61
hda: DMA timeout error
hda: dma timeout error: status=0x58 { DriveReady SeekComplete DataRequest }
hda: dma_timer_expiry: dma status == 0x61
hda: DMA timeout error
hda: dma timeout error: status=0x58 { DriveReady SeekComplete DataRequest }
hda: dma_timer_expiry: dma status == 0x61
hda: DMA timeout error
hda: dma timeout error: status=0x58 { DriveReady SeekComplete DataRequest }
hda: DMA disabled

Try as I may, however, I haven't been able to reproduce this problem with
hda (from dmesg: hda: WDC AC22500L, ATA DISK drive). It's worthy pointing
out that the machine came with a drive on which it was found that there
were actual bad sectors (tested on another machine.) ... the thought of
a buggy controller came to mind, but (though this may be possible), I've
never heard about a bad controller generating bad sectors ...

Karim
-- 
Author, Speaker, Developer, Consultant
Pushing Embedded and Real-Time Linux Systems Beyond the Limits
http://www.opersys.com || karim@opersys.com || 1-866-677-4546

Re: ide-cd vs. DMA

[Benjamin Herrenschmidt]

On Mon, 2005-05-23 at 23:59 -0400, Karim Yaghmour wrote:
> Benjamin Herrenschmidt wrote:
> > Well, not sure what's wrong here, but ATAPI errors shouldn't normally
> > result in stopping DMA. We may want to just blacklist your drive rather
> > than having this stupid fallback. In this case, I suspect it's
> > CSS/region issue with a DVD.
> 
> Here's a little bit more info:
> 
> Here's from dmesg:
> hdc: SAMSUNG SC-140B, ATAPI CD/DVD-ROM drive
> hdc: ATAPI 40X CD-ROM drive, 128kB Cache, UDMA(33)

  .../...

That's a lot of problems ... I would blame HW here, either a bad cable,
bad master/slave jumpers or a bad drive in the chain... or maybe as you
suggested, a problem with the controller.

Ben.

Re: ide-cd vs. DMA

[Kurt Wall]

On Monday 23 May 2005 23:24, Benjamin Herrenschmidt enlightened us 
thusly:
> On Mon, 2005-05-23 at 23:27 -0400, Karim Yaghmour wrote:
> > Benjamin Herrenschmidt wrote:
> > > hdb: command error: status=0x51 { DriveReady SeekComplete Error }
> > > hdb: command error: error=0x54 { AbortedCommand
> > > LastFailedSense=0x05 } ide: failed opcode was: unknown
> > > end_request: I/O error, dev hdb, sector 42872
> >
> > Got plenty of these an old Dell Optiplex GX1 (PIII-450) with
> > vanilla FC3. ... you've got to wonder when the kernel says there
> > are bad sectors on a CD (?) and then they disappear with:
> > hdparm -d0 /dev/hdc
>
> Well, not sure what's wrong here, but ATAPI errors shouldn't normally
> result in stopping DMA. We may want to just blacklist your drive
> rather than having this stupid fallback. In this case, I suspect it's
> CSS/region issue with a DVD.

I see the same thing here on a plain vanilla CD-ROM (pardon the 
unsightly wrapping):

May 23 21:52:34 luther kernel: hdc: packet command error: status=0x51 
{ DriveReady SeekComplete Error }
May 23 21:52:34 luther kernel: hdc: packet command error: error=0x54 
{ AbortedCommand LastFailedSense=0x05 }
May 23 23:12:37 luther kernel: hdc: packet command error: status=0x51 
{ DriveReady SeekComplete Error }
May 23 23:12:37 luther kernel: hdc: packet command error: error=0x54 
{ AbortedCommand LastFailedSense=0x05 }


Kurt

Re: ide-cd vs. DMA

[Lennart Sorensen]

On Tue, May 24, 2005 at 09:31:46PM -0400, Kurt Wall wrote:
> I see the same thing here on a plain vanilla CD-ROM (pardon the 
> unsightly wrapping):
> 
> May 23 21:52:34 luther kernel: hdc: packet command error: status=0x51 
> { DriveReady SeekComplete Error }
> May 23 21:52:34 luther kernel: hdc: packet command error: error=0x54 
> { AbortedCommand LastFailedSense=0x05 }
> May 23 23:12:37 luther kernel: hdc: packet command error: status=0x51 
> { DriveReady SeekComplete Error }
> May 23 23:12:37 luther kernel: hdc: packet command error: error=0x54 
> { AbortedCommand LastFailedSense=0x05 }

Looks a lot like the MultiMode problem that the kernel has an option for
dealing with.  I remember needing that on a 440BX based system in the
past (and the Dell GX1 is such a system too).

Len Sorensen

Re: ide-cd vs. DMA

[Alan Cox]

On Maw, 2005-05-24 at 04:24, Benjamin Herrenschmidt wrote:
> Well, not sure what's wrong here, but ATAPI errors shouldn't normally
> result in stopping DMA. We may want to just blacklist your drive rather
> than having this stupid fallback. In this case, I suspect it's
> CSS/region issue with a DVD.

The fallback should only be triggering on a DMA CRC error. Turning DMA
off or using ide-scsi works (as generally does using the -ac kernel
tree) because that handles partial completions correctly while readahead
caused errors break the base ide-cd driver badly.

I would suggest avoiding the ide-cd driver in the base kernel for most
purposes, at least until that bug and the "rmmod cache flush" bug are
fixed.

Alan

RE: ide-cd vs. DMA

[Chad Kitching]

Are you using hdparm -k1 to keep your settings over a reset?  If you're 
not, then this behaviour is really by-design.

-----Original Message-----
From: Benjamin Herrenschmidt
Sent: May 23, 2005 10:25 PM
Subject: Re: ide-cd vs. DMA

Well, not sure what's wrong here, but ATAPI errors shouldn't normally
result in stopping DMA. We may want to just blacklist your drive rather
than having this stupid fallback. In this case, I suspect it's
CSS/region issue with a DVD.

Ben.

RE: ide-cd vs. DMA

[Benjamin Herrenschmidt]

On Mon, 2005-05-23 at 23:26 -0500, Chad Kitching wrote:
> Are you using hdparm -k1 to keep your settings over a reset?  If you're 
> not, then this behaviour is really by-design.

And is broken. You can't expect users to play with hdparm and it's quite
common to have things like CSS or region errors on a DVD, taht shouldn't
turn your DMA off on the CD.

Damn, we are in 2005 folks !

Ben;