update on inode problem
kerry at gotss.net
Sun Nov 10 08:59:55 EST 2002
On Fri, Nov 08, 2002 at 12:32:29AM -0800, Ralph W. Reid wrote:
> Igor Gueths staggered into view and mumbled:
> >Hi all. Well after digging into this problem a bit further, I figured out
> >that the 1:0 reported in the errors about ext2_read_inode were refering to
> >/dev/ram0 according to /usr/src/linux/Documentation/devices.txt. That
> Hmmm. I wonder if this is an indication of a bit of bad RAM. I have
> heard that Linux is not very tolerant of bad memory, but RAM problems
> are rare enough that I have never heard exactly how they manifest
> themselves. If all else fails (and I do mean _all_ else), you might
Ram problems are quite regular. Unless you have ECC ram;
it is not unheard of to have a few cells fail in the millions
of cells in the average computer's sdram/dram/rdram etc.
True; memory has become more reliable;
but as we run it at faster clock rates; the margin
for error deminishes.
I had 1 failing bit in 3 sticks totalling 768 megabytes of sdram.
The error would quietly corrupt a bit in 768 megabytes whenever that location got used.
The error took 30 kernel compiles or 2-3 days to manifest itself
in an md5 sum failure or segmentation fault etc.
Testing all the memory in the machine with t he memory test at
http://www.memtest86.com (you can enable a serial console at compile time)
found the error reliably in 40-50 minutes.
A switching of the sticks and watching the failing address tracked the module at fault.
I removed and tested t he memory again. No errors.
I then send the 256-meg stick back for replacement under Waranty.
Now I have 768-meg of working memory and audio files or compresssed files
do not become corrupt. I used to test the error
by bzipping and unbzipping an iso image 10 times or more and checking for errors in the data.
This was done on a backup of the image since it would reliably corrupt the
> try swapping memory module locations on your mother board to see if
> this affects the situation at all. RAM failures are very rare, but
> they can happen. The simple RAM test performed during system boot
> can miss some memory errors, but this is pretty rare as well. I hope
> you can get your system working soon.
> Have a _great_ day!
> Ralph. N6BNO. Wisdom comes from central processing, not from I/O.
> rreid at sunset.net http://personalweb.sunset.net/~rreid
> Opinions herein are either mine or they are flame bait.
> y = x ^ LOG_B (x, y)
> Speakup mailing list
> Speakup at braille.uwo.ca
Kerry Hoath: kerry at gotss.net kerry at gotss.eu.org or kerry at gotss.spice.net.au
ICQ: 8226547 msn: kerry at gotss.net Yahoo: kerryhoath at yahoo.com.au
More information about the Speakup