ide-cd and bad sectors
Hi,
I use 2.6.12 vanilla kernel.
Since some time when a disk has bad sector because it is damaged or overburned, the Linux kernel seem too hang for this device :
the process reading is unkillable (even with SIGKILL).
the kernel seem to try to read the next sectors, but always failed.
The only way to stop the kernel trying to read the disc is to reboot or don't lock the drive and eject it, with the button on the device.
When hard-ejecting it, there are something strange in the log : the kernel failed for 508 sector (is there a so big cache for disk request ? If this is the case that explain why the kernel seem too hang as it need 1-2s per bad sector). Also at the end of the log [1] you could see that the 2 last message report error for the first sectors that produce the error. Is that normal : I thought this sector already produced an "media error (bad sector)" and should be skipped ?
Also I was wondering if all the sector that ide-cd failed to read are bad sector, or if ide-cd failed to put the drive in a consistent state for reading the next sector after corrupted one.
IIRC on 2.4 kernel there wasn't such problem, I even managed to recover some damaged disk...
thanks
Matthieu
[1]
Jul 2 15:10:13 localhost kernel: hdc: media error (bad sector): status=0x51 { D
riveReady SeekComplete Error }
Jul 2 15:10:13 localhost kernel: hdc: media error (bad sector): error=0x30 { La
stFailedSense=0x03 }
Jul 2 15:10:13 localhost kernel: ide: failed opcode was: unknown
Jul 2 15:10:13 localhost kernel: end_request: I/O error, dev hdc, sector 119095
12
Jul 2 15:10:13 localhost kernel: printk: 105 messages suppressed.
Jul 2 15:10:13 localhost kernel: Buffer I/O error on device hdc, logical block
2977378
Jul 2 15:10:17 localhost kernel: hdc: media error (bad sector): status=0x51 { D
riveReady SeekComplete Error }
Jul 2 15:10:17 localhost kernel: hdc: media error (bad sector): error=0x30 { La
stFailedSense=0x03 }
Jul 2 15:10:17 localhost kernel: ide: failed opcode was: unknown
Jul 2 15:10:17 localhost kernel: end_request: I/O error, dev hdc, sector 119095
16
[...]
Jul 2 15:12:03 localhost kernel: hdc: tray open
Jul 2 15:12:03 localhost kernel: end_request: I/O error, dev hdc, sector 119100
16
Jul 2 15:12:03 localhost kernel: hdc: tray open
Jul 2 15:12:03 localhost kernel: end_request: I/O error, dev hdc, sector 119100
20
Jul 2 15:12:03 localhost kernel: hdc: tray open
Jul 2 15:12:03 localhost kernel: end_request: I/O error, dev hdc, sector 119095
12
Jul 2 15:12:03 localhost kernel: hdc: tray open
Jul 2 15:12:03 localhost kernel: end_request: I/O error, dev hdc, sector 11909516
Quoted 3 lines · matthieu castet
Also I was wondering if all the sector that ide-cd failed to read are bad sector, or if ide-cd failed to put the drive in a consistent state for reading the next sector after corrupted one.
ide-cd wrongly errors all the sectors around an error, ide-scsi gets it right if the IDE firmware does. I sent Bartlomiej patches to fix that and I believe he accepted them
Hi Alan,
Quoted 9 lines · Alan Cox
Quoted 1 line · matthieu castet
Also I was wondering if all the sector that ide-cd failed to read are bad sector, or if ide-cd failed to put the drive in a consistent state for reading the next sector after corrupted one.
ide-cd wrongly errors all the sectors around an error, ide-scsi gets it right if the IDE firmware does. I sent Bartlomiej patches to fix that and I believe he accepted them
thanks,
they don't seem to be in his tree : http://www.kernel.org/git/?p=linux%2Fkernel%2Fgit%2Fbart%2Fide-2.6.git&a=search&s=ide-cd :(
Matthieu
Quoted 14 lines · matthieu castet
Hi Alan,
Quoted 10 lines · Alan Cox
Quoted 1 line · matthieu castet
Also I was wondering if all the sector that ide-cd failed to read are bad sector, or if ide-cd failed to put the drive in a consistent state for reading the next sector after corrupted one.
ide-cd wrongly errors all the sectors around an error, ide-scsi gets it right if the IDE firmware does. I sent Bartlomiej patches to fix that and I believe he accepted them
Quoted 5 lines
thanks,
they don't seem to be in his tree : http://www.kernel.org/git/?p=linux%2Fkernel%2Fgit%2Fbart%2Fide-2.6.git&a=search&s=ide-cd :(
Matthieu
Alan, maybe you could send them again, it would be nice to have a useable ide-cd capability which doesn't prevent reading the good sectors on a CD.
Quoted 2 lines · matthieu castet
IIRC on 2.4 kernel there wasn't such problem, I even managed to recover some damaged disk...
Oh on some 2.4 the kernel would retry each sector so many times that recovering undamaged sectors from a CD would have taken longer than the MTBF for a CDROM drive :)