PROBLEM: kernel 2.4 pdc202xx and kernel 2.6 pdc202xx_old Broken

PROBLEM: kernel 2.4 pdc202xx and kernel 2.6 pdc202xx_old Broken

[joram.agten]

Hello

please put me in CC

I'm using a pdc20262 promise utra66 controller to manage 4 HD's, all 30GB
and I put them in a software /dev/md/0 raid5 configuration

recently I upgraded my kernel to linux-2.6.11-gentoo-r4 and also
linux-2.6.11-gentoo-r5
and the raid array would go offline with dma timeouts when putting load to
it

Apr  9 23:30:15 vennen hde: DMA timeout error
Apr  9 23:30:15 vennen hde: dma timeout error: status=0x58 { DriveReady
SeekComplete DataRequest }
Apr  9 23:30:15 vennen 
Apr  9 23:30:15 vennen ide: failed opcode was: unknown
Apr  9 23:30:15 vennen hdh: dma_timer_expiry: dma status == 0x62
Apr  9 23:30:15 vennen hdh: DMA timeout error
Apr  9 23:30:15 vennen hdh: dma timeout error: status=0x58 { DriveReady
SeekComplete DataRequest }
Apr  9 23:30:15 vennen 
Apr  9 23:30:15 vennen ide: failed opcode was: unknown
Apr  9 23:30:15 vennen hdg: status timeout: status=0xd0 { Busy }
Apr  9 23:30:15 vennen 
Apr  9 23:30:15 vennen ide: failed opcode was: unknown
Apr  9 23:30:15 vennen hdg: DMA disabled
Apr  9 23:30:15 vennen PDC202XX: Secondary channel reset.
Apr  9 23:30:15 vennen PDC202XX: Primary channel reset.

in the beginning it was only with hde, so I suspected the disk, put it on
another controller, run a complete diskcheck on it (maxtor dos tool), no
problems found.
I attach it to the onboard motherboard controller and put it back in the
raid array
but again the raid array would fail, this time another disk.

If I boot again, and reinitialize the array, so that it starts building
again, it would fail quite fast
when not putting any load on the array it would just run fine

since I had installed the whole system from scratch I didn't know the exact
version of the previous kernel, so I went back to linux-2.4.20-gentoo-r33,
an here everything works
I also remember trying some more recent 2.4 kernel, but I do not recall the
exact version, there the same problem would manifest.

after some googling I found this thread: PDC202XX_OLD broken
http://www.ussg.iu.edu/hypermail/linux/kernel/0412.0/1277.html

this is almost the exact same behaviour
but in the thread it does not say if his problem is solved or not
the suggestion of  Bartlomiej Zolnierkiewicz  is present in my 2.6 kernel
images

same info is at http://users.pandora.be/gemma.joram/pdc202xx/pdc202xx.html
I also have sysinfo
http://users.pandora.be/gemma.joram/pdc202xx/pdc202xx_sysinfo.html
and log http://users.pandora.be/gemma.joram/pdc202xx/pdc202xx_log.html

hope this is enough.
I'l happily try more if you need more info.

greetings
joram

Re: PROBLEM: kernel 2.4 pdc202xx and kernel 2.6 pdc202xx_old Broken

[Bartlomiej Zolnierkiewicz]

Hi,

On Apr 10, 2005 2:12 PM, joram.agten@pandora.be <joram.agten@pandora.be> wrote:
> Hello
> 
> please put me in CC
> 
> I'm using a pdc20262 promise utra66 controller to manage 4 HD's, all 30GB
> and I put them in a software /dev/md/0 raid5 configuration
> 
> recently I upgraded my kernel to linux-2.6.11-gentoo-r4 and also
> linux-2.6.11-gentoo-r5
> and the raid array would go offline with dma timeouts when putting load to
> it
> 
> Apr  9 23:30:15 vennen hde: DMA timeout error
> Apr  9 23:30:15 vennen hde: dma timeout error: status=0x58 { DriveReady
> SeekComplete DataRequest }
> Apr  9 23:30:15 vennen
> Apr  9 23:30:15 vennen ide: failed opcode was: unknown
> Apr  9 23:30:15 vennen hdh: dma_timer_expiry: dma status == 0x62
> Apr  9 23:30:15 vennen hdh: DMA timeout error
> Apr  9 23:30:15 vennen hdh: dma timeout error: status=0x58 { DriveReady
> SeekComplete DataRequest }
> Apr  9 23:30:15 vennen
> Apr  9 23:30:15 vennen ide: failed opcode was: unknown
> Apr  9 23:30:15 vennen hdg: status timeout: status=0xd0 { Busy }
> Apr  9 23:30:15 vennen
> Apr  9 23:30:15 vennen ide: failed opcode was: unknown
> Apr  9 23:30:15 vennen hdg: DMA disabled
> Apr  9 23:30:15 vennen PDC202XX: Secondary channel reset.
> Apr  9 23:30:15 vennen PDC202XX: Primary channel reset.

This is a buggy expiry handler but it gets triggered as a result of
earlier OOPS:

Apr  9 23:30:01 vennen Call Trace:
Apr  9 23:30:01 vennen [] _spin_lock+0x16/0x90
Apr  9 23:30:01 vennen [] _spin_unlock+0xd/0x30
Apr  9 23:30:01 vennen [] do_truncate+0x4a/0x70
Apr  9 23:30:01 vennen [] _spin_unlock_irqrestore+0xf/0x30
Apr  9 23:30:01 vennen [] do_coredump+0x1fd/0x216
Apr  9 23:30:01 vennen [] __dequeue_signal+0xf5/0x1b0
Apr  9 23:30:01 vennen [] dequeue_signal+0x35/0xd0
Apr  9 23:30:01 vennen [] get_signal_to_deliver+0x229/0x320
Apr  9 23:30:01 vennen [] do_signal+0x9b/0x130
Apr  9 23:30:01 vennen [] sys_rt_sigaction+0xaa/0xc0
Apr  9 23:30:01 vennen [] do_page_fault+0x0/0x5d5
Apr  9 23:30:01 vennen [] do_notify_resume+0x37/0x3c
Apr  9 23:30:01 vennen [] work_notifysig+0x13/0x15

which doesn't seem to be related to IDE in any way.

It would be useful if you pinpoint issue further to a specific kernel version.

Bartlomiej