Hard drive failure when running realtime kernel

ajk48n · 10-12-2008, 11:08 PM

I recently compiled a realtime patch onto a kernel, and when restarting into the new kernel, I seem to get read errors on my hard drive.

I have 1 SATA drive with no RAID, and I have been running 2.6.26 for months now will no hardware problems. However, after recompiling the kernel with the real time patch found in http://www.kernel.org/pub/linux/kernel/projects/rt/, my system freezes up completely within the first few minutes.

On one instant, I got an error from X, saying that my file system was read-only.

Here is the relevant info from dmesg:

Code:

Oct 12 22:31:43 debian kernel: [  405.369226]          res 40/00:00:00:00:00/00:00:00:00:00/00 Emask 0x4 (timeout)
Oct 12 22:31:43 debian kernel: [  405.369236]          res 40/00:00:00:00:00/00:00:00:00:00/00 Emask 0x4 (timeout)
Oct 12 22:31:43 debian kernel: [  405.369245]          res 40/00:00:00:00:00/00:00:00:00:00/00 Emask 0x4 (timeout)
Oct 12 22:31:43 debian kernel: [  405.369256] ata2: hard resetting link
Oct 12 22:31:44 debian kernel: [  405.684768] ata2: SATA link up 1.5 Gbps (SStatus 113 SControl 300)
Oct 12 22:31:44 debian kernel: [  405.748884] ata2.00: failed to IDENTIFY (INIT_DEV_PARAMS failed, err_mask=0x80)
Oct 12 22:31:44 debian kernel: [  405.748894] ata2: failed to recover some devices, retrying in 5 secs
Oct 12 22:31:49 debian kernel: [  410.749414] ata2: hard resetting link
Oct 12 22:31:49 debian kernel: [  411.064956] ata2: SATA link up 1.5 Gbps (SStatus 113 SControl 300)
Oct 12 22:31:49 debian kernel: [  411.128295] ata2.00: failed to IDENTIFY (INIT_DEV_PARAMS failed, err_mask=0x80)
Oct 12 22:31:49 debian kernel: [  411.128303] ata2: failed to recover some devices, retrying in 5 secs
Oct 12 22:31:54 debian kernel: [  416.128791] ata2: hard resetting link
Oct 12 22:31:55 debian kernel: [  416.443953] ata2: SATA link up 1.5 Gbps (SStatus 113 SControl 300)
Oct 12 22:31:55 debian kernel: [  416.507987] ata2.00: failed to IDENTIFY (INIT_DEV_PARAMS failed, err_mask=0x80)
Oct 12 22:31:55 debian kernel: [  416.507995] ata2.00: disabled
Oct 12 22:31:55 debian kernel: [  417.008049] ata2: EH complete
Oct 12 22:31:55 debian kernel: [  417.008069] sd 1:0:0:0: [sda] Result: hostbyte=DID_BAD_TARGET driverbyte=DRIVER_OK,SUG     GEST_OK
Oct 12 22:31:55 debian kernel: [  417.008082] lost page write due to I/O error on sda1
Oct 12 22:31:55 debian kernel: [  417.008094] sd 1:0:0:0: [sda] Result: hostbyte=DID_BAD_TARGET driverbyte=DRIVER_OK,SUG     GEST_OK

And then the last line is repeated until the end.

This would seem to be a hardware problem, however it has never happened in the regular kernel. The drive is a year and a half old, and is a Seagate Barracuda 7200.10

Is it possible that something in the realtime kernel is writing more frequently to the hard drive, which is them coming upon a bad sector?

Any help or suggestions are appreciated.

jb_get · 10-13-2008, 01:41 AM

Instead of wasting time trying to figure out what is wrong I would just reinstall the whole things, save you time and headaches.

ajk48n · 10-13-2008, 01:52 AM

Not quite sure what reinstalling would help. I've tried recompiling many versions of the real time kernel, all with the same result.

salasi · 10-13-2008, 02:25 AM

This sequence

Code:

Oct 12 22:31:44 debian kernel: [  405.748884] ata2.00: failed to IDENTIFY (INIT_DEV_PARAMS failed, err_mask=0x80)
Oct 12 22:31:44 debian kernel: [  405.748894] ata2: failed to recover some devices, retrying in 5 secs
Oct 12 22:31:49 debian kernel: [  410.749414] ata2: hard resetting link
Oct 12 22:31:49 debian kernel: [  411.064956] ata2: SATA link up 1.5 Gbps (SStatus 113 SControl 300)
Oct 12 22:31:49 debian kernel: [  411.128295] ata2.00: failed to IDENTIFY (INIT_DEV_PARAMS failed, err_mask=0x80)
Oct 12 22:31:49 debian kernel: [  411.128303] ata2: failed to recover some devices, retrying in 5 secs
Oct 12 22:31:54 debian kernel: [  416.128791] ata2: hard resetting link
Oct 12 22:31:55 debian kernel: [  416.443953] ata2: SATA link up 1.5 Gbps (SStatus 113 SControl 300)
Oct 12 22:31:55 debian kernel: [  416.507987] ata2.00: failed to IDENTIFY (INIT_DEV_PARAMS failed, err_mask=0x80)
Oct 12 22:31:55 debian kernel: [  416.507995] ata2.00: disabled

doesn't sound too promising; having finished up with ata2.00 disabled, that you cannot write to the file system doesn't really seem all that surprising. It seems obvious (but maybe wrong???) that initialisation has failed and if that occurs consistently, it won't be because there more frequent writes to the disk or something. It just doesn't work.

I have to ask; Is RT necessary for your application, or even advantageous? Wouldn't you be better off just not using RT?

ajk48n · 10-13-2008, 08:36 AM

I'm working with midi, so real time is pretty much mandatory. Any ideas why the errors would only occur with the new kernel? I couldn't find any additions in the new .config that deal with SATA

ajk48n · 10-20-2008, 07:39 PM

Anyone have any ideas?

Possibly something related to the speed of input/output to the drive?