Re: could someone plz explain those ext3/hard disk errors

[John Bradford]

Quote from Bill Davidsen <davidsen@xxxxxxx>:
> John Bradford wrote:
> >>now...hm, it all started when i upgraded from kernel 2.4.19 to 2.6.0
> >>in late decemeber, the system worked very fine for a week or so
> >>(having great response times!) but then all of a sudden the problems
> >>started. 2 disks died. then my gigabit network card was only able to
> >>transmit 200kb/s (but this was really a hardware problem, a new card
> >>is working fine again, well...). a week later the next disks are
> >>having problems and i have yet to RMA three disks. and now the next
> >>two disks..., i'm getting insane ;) i can't see any EXT3 error anymore
> >>*g* the next disks will be reiserfs only to see other error messages
> >>;) well, but that doesn't solve the problem of 6 disks within 2
> >>months...this is so unlikely.
> > 
> > 
> > Please read the FAQ, fix your mail application - you are sending long
> > lines, and don't break the CC list.
> > 
> > As to your problem, look at the LBA sector addresses in the error
> > message:
> > 
> > 280923064991615
> > 
> > is your drive really over 100 EB?  No...
> 
> I think we could assume that (a) he never told the kernel the disk was 
> "over 100 EB" and (b) the kernel was trying to use that LBA anyway. 
> Which could be due to either a kernel bug or memory corruption (or CPU 
> problems, but unlikely).

What I was trying to point out is that the error message is clearly
the result of a problem elsewhere.  Unless the drive firmware is
buggy, or something very strange is going on inside the drive, (bad
internal RAM or something like that), then the kernel did send a
request for a sector which is well out of range.  What caused that
request we don't know - quite possibly corruption of some filesystem
structure on the disk caused that request, but it's important to be
clear that the error message is the expected response to a request for
such a high block number, and doesn't within itself indicate a problem
with the disk.

I.E. Even though there is every chance that the drive is faulty, the
posted error message doesn't indicate a drive failiure in itself, and
you should look elsewhere.

John.
-
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/