On Tue, 30 Apr 2002, Andreas Dilger wrote:
> On Apr 30, 2002 21:54 -0400, Tom Diehl wrote:
>
> It looks like you were getting garbage from the disk before the journal
> assertion happened (i.e. the ext3 error), and the journal assertion is
> just there to save your filesystem from getting corrupted with further
> bad operations.
>
> > This is a stock 7.2 system with all revelant updates.
> > Not sure what other info to provide so if I missed something please let me
> > know.
>
> I would really recommend upgrading to the latest RH errata kernel. The
> ext3 code has had a number of bugs fixed since 2.4.9. It might also be
> related to IDE stuff, don't know.
AFAIK 2.4.9-31 is their latest errata kernel. Just checked the ftp site and
that is the latest one there, although the roumer mill would suggest this
might change shortly. I could upgrade to their beta kernel I suppose.
> > ide1: reset: success
>
> When did that reset happen? It wasn't in the syslog that you sent.
Looks like it happened just before the logs were rotated. I missed it sorry.
Here it is:
Apr 25 04:35:18 kanga kernel: hdc: timeout waiting for DMA
Apr 25 04:35:18 kanga kernel: ide_dmaproc: chipset supported ide_dma_timeout func only: 14
Apr 25 04:35:18 kanga kernel: hdc: status timeout: status=0xd0 { Busy }
Apr 25 04:35:18 kanga kernel: hdd: DMA disabled
Apr 25 04:35:18 kanga kernel: hdc: drive not ready for command
Apr 25 04:35:27 kanga kernel: ide1: reset: success
The next entry was from the syslog output I provided in the previous message.
So the reset was just a few seconds before.
Apr 25 04:35:53 kanga syslogd 1.4.1: restart.
> > EXT3-fs error (device ide1(22,65)): ext3_readdir: bad entry in directory #2665467: rec_len % 4 != 0 - offset=0, inode=762621470, rec_len=44574, name_len=110
>
> The rec_len is way out. The inode number is probably also bad, but not
> sure...
>
> > Assertion failure in journal_bmap_Rbbdc8009() at journal.c:602: "ret != 0"
> > kernel BUG at journal.c:602!
>
> Just a symptom of bad data, not the real cause. Note that I wanted to
> look at this bit of code, but that assertion is not even there anymore
> (the kernel turns the filesystem read only and just returns now).
So am I understanding you correctly that there is still no good way to tell if this
was hdwe or a software failure?
--
.............Tom "Nothing would please me more than being able to
tdiehl@rogueind.com hire ten programmers and deluge the hobby market
with good software." -- Bill Gates 1976
We are still waiting ....
