[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: What was wrong with this sequence?

Stephen C. Tweedie:
> Hi,
> On Thu, Oct 18, 2001 at 08:17:23PM -0400, John Ruttenberg wrote:
> > I thought I understood things, but I guess not.  I recently configured a new
> > system for a colleague as follows:
> > 
> >     1. Redhat 7.1 install Dell Inspiron 8100 - three partitions on 30G - /,
> >        /boot, swap
> >     2. Boot up
> >     3. configure, make, make install of linux-2.4.12-ac3 (with approriate lilo
> >        changes) lilo (but no reboot until 8)
> >     4. rpm -U mount-2.11g-5.i386.rpm (from rawhide)
> >     5. configure, make install e2fsprogs-1.25
> >     6. tune2fs -j /dev/hda1 (boot) /dev/hda6 (/)
> >     7. substitute ext3 for ext2 for bot filesystems in /etc/fstab
> >     8. reboot
> >     9. 2.4.12-ac3 kernel boots, but fsck detects massive problems and we end
> >        up with a ton of stuff in /lost+found and enough stuff seems maybe
> >        broken that we think we should start from scratch
> Why did fsck get invoked?  It will always tell you why it is starting
> a full scan (eg. "Too many mounts since last check, fsck forced" or
> similar messages). 

Sorry, we don't remember the answer.

> After a normal controlled reboot, the full fsck should not be
> triggered, unless the previous (ext2) filesystem had already detected
> some filesystem errors and recorded the need for a fsck in the
> filesystem's superblock.

Come to think of it, we did run dumpe2fs before tune2fs and it did say
something about the state that looked wrong.  At the time I confused it with
the "not clean" message, perhaps "... with errors".  Woody may remember
exactly what it said.

> Either way, this sounds much more like an ide-level problem than a
> fs-level one.  So far all of the similar problem reports I've seen
> have been with certain IBM 20GB laptop disks, in three cases with an
> inspiron 8100 involved.  What disk is your system using?  The more
> examples I get, the more we may be able to narrow down the
> drivers/disks/kernels which are causing the problem.

It has a 30gb drive.  /proc/ide/hda/model/ has HITACHI_DK23CA-30.

This makes some sense because the redhat 7.1 installation was very difficult.
The install kept crashing in the middle.  I found (but cannot find again) a
redhat bug about it and ended up starting the install with "linux ide=nodma"
on the lilo command line.  I found another scrap of information about this
that said the cd/dvd drive cannot use dma.  Anyway, after step 2 we were
running the standard redhat kernel without "ide=nodma".  So perhaps that was
the problem.

So we probably need to add ide=nodma to the lilo boot options during the install.

[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]