David Vasek wrote:
quoted text
> Out of curiosity, does the same happen if you dd from /dev/rwd0d?
>
> As Matthew Szudzik pointed out, dd is failing when it attempts to read 
> (2^28)th sector of the current device you are reading from. Up to, 
> including, 2^28-1 everything is ok.
>
> Regards,
> David
>
>

I made an error in my last post.  I said the "problem" sector was in 
/tmp on wd0d, but it was actually in /home on wd0e.  With that in mind, 
I tried two tests.

First, filling up /home.  The only result was the expected reaction of 
apps using /home to find it full, but no I/O error from dd while filling 
it up.  I was thinking that my previous system hangs had to do with a 
read or a write to that particular sector during normal system use, but 
I guess not.

The second was your suggestion, and interestingly, it produces the error.

Partition "e" starts at 43 022 070, the "problem" is at 268 435 455, so 
we'll skip 225 413 380 to start just before that spot:

# dd if=/dev/rwd0e of=/dev/null skip=225413380
dd: /dev/rwd0e: Input/output error
5+0 records in
5+0 records out
2560 bytes transferred in 4.084 secs (627 bytes/sec)

Doing the same thing with bs=1024:

# dd if=/dev/rwd0e of=/dev/null skip=112706690 bs=1024
^C164347+0 records in
164347+0 records out
168291328 bytes transferred in 15.241 secs (11041848 bytes/sec)
(I aborted it)

I've managed to figure out:

1) there's nothing wrong with that actual sector on the drive.
2) it's related to _this_ particular make/model.  (The 500GB Western 
Digital was fine.)
3) it's not a problem with dd.
4) there is no difference between -stable and -current for this.
5) using a bs other than 512 in dd has no problem.

Of course, there is no proof my previous hangs have anything to do with 
this.  I haven't had the system lock up in the 5 days I've been using 
this drive, so that doesn't really mean anything vs. no hangs in say, 30 
days.

-- 

-RSM

http://www.erratic.ca