> 
> Of course, it isn't necessarily the same problem, I'm just pointing out that
> we had a problem that looked like this and really looked like it wasn't
> hardware (for the same reasons you listed) but it was, after all.
> 

I hate to be so verbose, but I just thought I'd throw one more factoid out
there, in case it makes anyone say, 'Ahhh yes':  

All of the illegal blocks are actually ASCII data from the files I was
accessing at the time:

raid0_map bug: hash->zone0==NULL for block 808464440
raid0_map bug: hash->zone0==NULL for block 171521844
raid0_map bug: hash->zone0==NULL for block 959524912

808464440 -> 0x30303038 = "0008"
171521844 -> 0x0A393734 = "\n974"
959524912 -> 0x39313030 = "9100"

The large files were all 'call records' containing only ASCII digit
characters and linefeeds.  Subsequent accesses to the files caused no
problems.

David


-- 
/==============================\
| David Mansfield              |
| [EMAIL PROTECTED]             |
\==============================/

Reply via email to