OCZ SSD problems on Linux due to all_generic_ide

[eran]

Hi everybody,

I'm posting about a *solved* problem I've had using OCZ SSD on the Fit-PC2 together with Debian GNU/Linux 5 (Lenny).

It worked perfectly well with an ordinary WD hard-drive. It always booted with all_generic_ide as a kernel parameter. I've replicated the root partition of the WD drive to the new SSD (adjusting partition size, filesystem size using resize2fs and installed bootloader).
However, when booting from the drive, I got:

[ 1.956493] hda: OCZ-VERTEX, ATA DISK drive
...
[ 4.326058] hda: max request size: 512KiB

[ 4.326058] hda: set_multmode: status=0x51 { DriveReady SeekComplete Error }
[ 4.326058] hda: set_multmode: error=0x04 { DriveStatusError }
[ 4.326058] ide: failed opcode was: 0x27

[ 4.331088] hda: 62533296 sectors (32017 MB) w/32767KiB Cache, CHS=16383/255/63
...

Some time later on in the boot process:
[ 5.116976] hda: task_no_data_intr: status=0x51 { DriveReady SeekComplete Error }
[ 5.117253] hda: task_no_data_intr: error=0x10 { SectorIdNotFound },
LBAsect=204674250571776, sector=18446744073709551615

It looks (to me, anyway) like the kernel thinks there are way too many sectors than there actually are. I ignored this initially, but quickly the kernel tried writing to one of those non-existing sectors, resulting in a mass of error messages in the system log. fsck found one orphan inode - supporting this theory.

I tried various BIOS settings to no avail. Then I removed the all_generic_ide kernel parameter - now the disk was detected as a SCSI device (sda) but the boot process stuck. Changing the root argument of the kernel command line to /dev/sda1 did the trick - the system booted without any of these problems.

Note: It's possible that the bios should have "disk access mode" set to "Other" instead of "DOS".

In the hope it helps somebody,
Eran