Re: Debugging I/O errors further?

[Roy Sigurd Karlsbakk]

> > Sorry for stressing this, but is there a way I can debug this  
> > further? it's a seagate drive connected to a sata_sil controller.  
> > I only get ext3 errors, and it fails after a while whatever I do
>
> Only idea I have is to unmount the drive ( or remount r/o ) and  
> repeatedly md5sum the block device and see if it ever fails to  
> correctly read the data, and if you get any errors in your syslog.   
> If you get no error messages in your syslog and md5sum completes  
> without error but does not get the same hash each time, then there  
> is definitely something very fubar with the hardware or deep in the  
> kernel.

md5sum has now been running in a loop for some 22 hours and completed  
11 sums of the drive (md5summing 400 gigs takes a little while). the  
md5sum is identical for each test,  and the syslog has no error  
indications. Then, starting harddisk stresstest, I get this error  
again after about an hour testing:

Nov  3 11:33:17 ganske kernel: EXT3-fs error (device sda1):  
ext3_free_blocks: Freeing blocks not in datazone - block =  
1349004846, count = 1
Nov  3 11:33:20 ganske kernel: EXT3-fs error (device sda1):  
ext3_free_blocks: Freeing blocks not in datazone - block =  
1449605700, count = 1
Nov  3 11:33:23 ganske kernel: EXT3-fs error (device sda1):  
ext3_free_blocks: Freeing blocks not in datazone - block = 629024587,  
count = 1
Nov  3 11:33:24 ganske kernel: EXT3-fs error (device sda1):  
ext3_free_blocks: Freeing blocks not in datazone - block =  
1059741014, count = 1
...

So, error only occurs on filesystem usage, not with direct  
blockdevice access.

Any ideas?

roy
--
Roy Sigurd Karlsbakk
roy@xxxxxxxxxxxxx
-------------------------------
MICROSOFT: Acronym for "Most Intelligent Customers Realise Our  
Software Only Fools Teenagers"

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/