[Bug 30153] IBM X31 freeze under X, shortly after boot

[Olivier Cortes]

Public bug report changed:
https://launchpad.net/malone/bugs/30153

Comment:
more info: it could be related to the ATA / DMA code in the kernel : i
use this line :

hdparm -c 1 -d 1 -A 1 -a 1024 -m 16 -u 1 -X 69 /dev/hda

at boot.
On warty / hoary / breezy i never had a problem with this command on this hardware. With last 2.6.15, the command fails, and dmesg reports :
hda: dma_intr: status=0x58 { DriveReady SeekComplete DataRequest }
[4302783.866000] ide: failed opcode was: unknown
[4302783.931000] hda: set_drive_speed_status: status=0x58 { DriveReady SeekComplete DataRequest }
[4302803.981000] hda: dma_timer_expiry: dma status == 0x00
[4302803.981000] hda: DMA timeout retry
[4302803.981000] hda: timeout waiting for DMA
[4302803.981000] hda: status error: status=0x58 { DriveReady SeekComplete DataRequest }
[4302803.981000] ide: failed opcode was: unknown
[4302803.982000] hda: drive not ready for command

and hdparm reports using_dma = 0. restarting the command multiple times can lead to DMA to be activated in some cases, but not all.
when the machine freeze it always lead to hard data corruption, so i suppose theses two facts could be related.

another long smart test shows no errors, and smartcl continues to report
"no errors loggued" so i think it is related with ata / dma handling
somewhere.

info on my disk can be found at http://deep-ocean.net/files/x31/ (smartctl-a_hda.txt).
i ran another suite of tests to be sure this week-end : reformated the machine another time and installed breezy on it :
-> no problems with 686 kernel.
Upgrading to dapper leads to freeze + data corruption with 686 2.6.15 after a reboot. using the 386 2.6.15-16 kernel confirms "no freeze" (the machine has been running many hours without freezing).

Any advice or hint you could give me is welcome. I can recompile a
kernel or whatever, but i don't know what to do to narrow the freeze
further...

Anyway i continue trying to use the machine, rebooting the machine from
time to time with 686 kernel, reformating when i get data corruption,
dist-upgrading to dapper, in case anything new could help us to find the
cause of the problem.

thanks in advance for your time.