Hi there,

I am running Debian3.1 with kernel 2.6.8-2-386 on a Dell Laptop. I am using it as a remotely accessed always-on server. It seems to run quite well, but sometimes after a few months, I get a "dma_timer_expiry: dma status == 0x20" error the file system (reiserFS) resets itself into readonly mode. I get all sorts of Input/Output errors and I can't even reboot the box.

Is this showing that the disk is bad? It seems not to be as it is working fine for months before getting the DMA errors. Is there anything I can configure to avoid these problems? Is there a kernel modification, or fstab/hdparm command, or other filesystem that I should use that will prevent the drive going into read-only mode?

Thank you for any help!

root@gateway:/etc# cat debian_version
cat: debian_version: Input/output error

from syslog:

Jul 1 13:55:54 localhost kernel: hda: dma_timer_expiry: dma status == 0x20
Jul 1 13:56:59 localhost syslogd: /var/log/kern.log: Input/output error
Jul 1 13:56:59 localhost kernel: hda: DMA timeout retry
Jul 1 13:56:59 localhost kernel: hda: timeout waiting for DMA
Jul 1 13:56:59 localhost kernel: hda: status timeout: status=0xd0 { Busy }
Jul 1 13:56:59 localhost kernel:
Jul 1 13:56:59 localhost kernel: hda: drive not ready for command
Jul 1 13:56:59 localhost kernel: ide0: reset timed-out, status=0xd0
Jul 1 13:56:59 localhost kernel: hda: status timeout: status=0xd0 { Busy }
Jul 1 13:56:59 localhost kernel:
Jul 1 13:56:59 localhost kernel: hda: drive not ready for command
Jul 1 13:56:59 localhost kernel: ide0: reset timed-out, status=0xd0
Jul 1 13:56:59 localhost kernel: end_request: I/O error, dev hda, sector 13187688
Jul 1 13:56:59 localhost kernel: end_request: I/O error, dev hda, sector 270432
Jul 1 13:56:59 localhost kernel: Buffer I/O error on device hda2, logical block 1674
Jul 1 13:56:59 localhost kernel: lost page write due to I/O error on hda2
Jul 1 13:56:59 localhost kernel: end_request: I/O error, dev hda, sector 270440
Jul 1 13:56:59 localhost kernel: Buffer I/O error on device hda2, logical block 1675
Jul 1 13:56:59 localhost kernel: lost page write due to I/O error on hda2
Jul 1 13:56:59 localhost kernel: end_request: I/O error, dev hda, sector 270448
Jul 1 13:56:59 localhost kernel: Buffer I/O error on device hda2, logical block 1676
Jul 1 13:56:59 localhost kernel: lost page write due to I/O error on hda2
Jul 1 13:56:59 localhost kernel: REISERFS: abort (device hda2): Journal write error in flush_commit_list
Jul 1 13:56:59 localhost kernel: REISERFS: Aborting journal for filesystem on hda2
Jul 1 13:56:59 localhost kernel: ReiserFS: hda2: warning: clm-6006: writing inode 688 on readonly FS
Jul 1 13:56:59 localhost last message repeated 11 times
Jul 1 13:56:59 localhost kernel: ReiserFS: hda2: warning: clm-6006: writing inode 1726 on readonly FS
Jul 1 13:56:59 localhost last message repeated 5 times
Jul 1 13:56:59 localhost kernel: ReiserFS: hda2: warning: clm-6006: writing inode 688 on readonly FS
Jul 1 13:56:59 localhost last message repeated 5 times
Jul 1 13:56:59 localhost kernel: ReiserFS: hda2: warning: clm-6006: writing inode 1726 on readonly FS
Jul 1 13:56:59 localhost last message repeated 5 times
Jul 1 13:56:59 localhost kernel: ReiserFS: hda2: warning: clm-6006: writing inode 688 on readonly FS
Jul 1 13:56:59 localhost last message repeated 11 times
Jul 1 13:56:59 localhost kernel: ReiserFS: hda2: warning: clm-6006: writing inode 1726 on readonly FS
Jul 1 13:56:59 localhost root@gateway:/var/log#

That is a weird message to get. I think, probably, it has something to do with the laptop's power saving features, such as suspend to disk, which linux can't adequately handle.

if dma is the error, then perhaps you can just turn it off by 'hdparm -d0 /dev/hda'. but i'm warning you. your hd's problems might only get worse with this.

btw how did you access the syslog messages anyway?

It seems as though there was a bug submitted to debian as someone else was having problems with their drives "going to sleep". He was using ext3, which was also freaking out and going into read-only mode.

(won't let me post URLs yet as I am too new to this forum so check out bug 335538 in debian bugs if you are interested)

I'll have to double check the server the next time I have physical access (server in Canada, I'm in the UK) but I think that I have all the BIOS-related power saving turned off.

I'll check hdparm, but I think that the guy who raised the bug also tried this to no effect. It is funny how the server works for weeks and months fine without any problems, but then it gets this problem

I was thinking of changing filesystems the next time I have physical access to this laptop gateway/server, but he was using ext3 and he ended up ro as well!

konsolebox,

I am able to log into the gateway/server still via SSH. Most of the commands I cannot run as I get the following I/O errors:

root@gateway:~# ping
-su: /bin/ping: Input/output error

For some reason, it allows me to su to root and cat syslog where I find these messages.

Any way to avoid these problems? In the bug report he says that his drives "go to sleep" and then they can only be restarted with a hard reset (power cycle in my case).

I guess I had better look at a more stable server/gatway solution.

don't you think it's also possible that the power supply unit is having problems when things starts to heat-up? not really connected though

Could be a heat problem with the powersupply as well as something powering down the drive and then the kernel freaking out when it can't access the drive.

I think that I will retire the laptop and be moving to a solid-state solution (NLSU2) and connecting an external HD for storage (if necessary).