2005-05-26 15:32:07

by Meelis Roos

[permalink] [raw]
Subject: ide-cd problem in 2.6.12-rc5 + todays snapshot

Background: I have a Sony CDU5211 CD drive with Intel D815EEA2 mainboard
(ICH2 IDE in 815 chipset). Since 2.4.21 timeframe IDE DMA for this CD
drive is broken (see my post
http://www.ussg.iu.edu/hypermail/linux/kernel/0410.3/0480.html). This
happens on at least 2 identical machines. This is the first problem
(that I have learned to live with).

Now, since ide-cd dma is broken, the first access to cd always gets DMA
timeout and turns off DMA, then it works. I have hddtemp installed and
it probes for drives on boot. In 2.6.12 (and I think I tested pristine
2.6.12-rc5 too) the cd works as before - dma timeout+disable on first
access (by hddtemp).

Now, in 2.6.12-rc5 + todays git snapshot, it does not work any more. I
suspect the DMA alignment change.

In 2.6.12-rc2 the dmesg from hddtemp was

hdc: DMA disabled
hdc: drive_cmd: status=0x51 { DriveReady SeekComplete Error }
hdc: drive_cmd: error=0x04 { AbortedCommand }
ide: failed opcode was: 0xb0

In todays snapshot, the dmesg is

hdc: DMA interrupt recovery
hdc: lost interrupt
hdc: status timeout: status=0xd0 { Busy }
ide: failed opcode was: unknown
hdc: DMA disabled
hdc: drive not ready for command
hdc: ATAPI reset complete
cdrom_pc_intr, write: dev hdc: flags = REQ_STARTED REQ_PC REQ_QUIET
sector 0, nr/cnr 0/0
bio 00000000, biotail 00000000, buffer 00000000, data 00000000, len 0
cdb: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
hdc: cdrom_pc_intr: The drive appears confused (ireason = 0x02)
hdc: lost interrupt
cdrom_pc_intr, write: dev hdc: flags = REQ_STARTED REQ_PC REQ_FAILED REQ_QUIET
sector 0, nr/cnr 0/0
bio 00000000, biotail 00000000, buffer 00000000, data 00000000, len 0
cdb: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
hdc: cdrom_pc_intr: The drive appears confused (ireason = 0x02)
hdc: lost interrupt
cdrom_pc_intr, write: dev hdc: flags = REQ_STARTED REQ_PC REQ_FAILED REQ_QUIET
sector 0, nr/cnr 0/0
bio 00000000, biotail 00000000, buffer 00000000, data 00000000, len 0
cdb: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
hdc: cdrom_pc_intr: The drive appears confused (ireason = 0x02)
hdc: lost interrupt
cdrom_pc_intr, write: dev hdc: flags = REQ_STARTED REQ_PC REQ_FAILED REQ_QUIET
sector 0, nr/cnr 0/0
bio 00000000, biotail 00000000, buffer 00000000, data 00000000, len 0
cdb: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
hdc: cdrom_pc_intr: The drive appears confused (ireason = 0x02)

and so on ad infimum.

The messages are very similar to my earlier reported problems that were
fixed:
http://www.ussg.iu.edu/hypermail/linux/kernel/0312.3/1003.html
http://www.ussg.iu.edu/hypermail/linux/kernel/0402.1/0459.html


--
Meelis Roos ([email protected])


2005-05-26 16:03:10

by Alan

[permalink] [raw]
Subject: Re: ide-cd problem in 2.6.12-rc5 + todays snapshot

On Iau, 2005-05-26 at 16:31, Meelis Roos wrote:
> Background: I have a Sony CDU5211 CD drive with Intel D815EEA2 mainboard
> (ICH2 IDE in 815 chipset). Since 2.4.21 timeframe IDE DMA for this CD
> drive is broken (see my post
> http://www.ussg.iu.edu/hypermail/linux/kernel/0410.3/0480.html). This
> happens on at least 2 identical machines. This is the first problem
> (that I have learned to live with).
>
> Now, since ide-cd dma is broken, the first access to cd always gets DMA
> timeout and turns off DMA, then it works. I have hddtemp installed and
> it probes for drives on boot. In 2.6.12 (and I think I tested pristine
> 2.6.12-rc5 too) the cd works as before - dma timeout+disable on first
> access (by hddtemp).

Ok that one is very different to the end of media bugs.

2005-05-26 17:20:23

by Jens Axboe

[permalink] [raw]
Subject: Re: ide-cd problem in 2.6.12-rc5 + todays snapshot

On Thu, May 26 2005, Meelis Roos wrote:
> Background: I have a Sony CDU5211 CD drive with Intel D815EEA2 mainboard
> (ICH2 IDE in 815 chipset). Since 2.4.21 timeframe IDE DMA for this CD
> drive is broken (see my post
> http://www.ussg.iu.edu/hypermail/linux/kernel/0410.3/0480.html). This
> happens on at least 2 identical machines. This is the first problem
> (that I have learned to live with).
>
> Now, since ide-cd dma is broken, the first access to cd always gets DMA
> timeout and turns off DMA, then it works. I have hddtemp installed and
> it probes for drives on boot. In 2.6.12 (and I think I tested pristine
> 2.6.12-rc5 too) the cd works as before - dma timeout+disable on first
> access (by hddtemp).
>
> Now, in 2.6.12-rc5 + todays git snapshot, it does not work any more. I
> suspect the DMA alignment change.

It must be, thanks for reporting this so quickly. Linus, can you exclude
that patch again? Rather miserably slow burning for some, than broken
hardware for others.

Seems we do need finer granularity setting of alignment/length
restrictions.

--
Jens Axboe

2005-05-27 06:21:08

by Jens Axboe

[permalink] [raw]
Subject: Re: ide-cd problem in 2.6.12-rc5 + todays snapshot

On Thu, May 26 2005, Jens Axboe wrote:
> On Thu, May 26 2005, Meelis Roos wrote:
> > Background: I have a Sony CDU5211 CD drive with Intel D815EEA2 mainboard
> > (ICH2 IDE in 815 chipset). Since 2.4.21 timeframe IDE DMA for this CD
> > drive is broken (see my post
> > http://www.ussg.iu.edu/hypermail/linux/kernel/0410.3/0480.html). This
> > happens on at least 2 identical machines. This is the first problem
> > (that I have learned to live with).
> >
> > Now, since ide-cd dma is broken, the first access to cd always gets DMA
> > timeout and turns off DMA, then it works. I have hddtemp installed and
> > it probes for drives on boot. In 2.6.12 (and I think I tested pristine
> > 2.6.12-rc5 too) the cd works as before - dma timeout+disable on first
> > access (by hddtemp).
> >
> > Now, in 2.6.12-rc5 + todays git snapshot, it does not work any more. I
> > suspect the DMA alignment change.
>
> It must be, thanks for reporting this so quickly. Linus, can you exclude
> that patch again? Rather miserably slow burning for some, than broken
> hardware for others.
>
> Seems we do need finer granularity setting of alignment/length
> restrictions.

BTW Meelis, any chance you can change that 3 into a 15 for testing
purposes?

--
Jens Axboe

2005-05-27 07:08:39

by Meelis Roos

[permalink] [raw]
Subject: Re: ide-cd problem in 2.6.12-rc5 + todays snapshot

>> Seems we do need finer granularity setting of alignment/length
>> restrictions.
>
> BTW Meelis, any chance you can change that 3 into a 15 for testing
> purposes?

15 seems to work fine.

--
Meelis Roos ([email protected])