[TAG] Flaky system

[Neil Youngman]

On Wednesday 19 Oct 2005 13:33, Thomas Adam wrote:
> --- Neil Youngman <n.youngman at ntlworld.com> wrote:
> > My main system has come over all flaky this morning
> >
> > Symptoms:
> > ifup won't bring up eth0 (No networking problems in win98)
> > On boot it generates a bunch of messages about hda9 not having a
> > valid
> > FAT filesystem and not having a valid NTFS filesystem.
>
> So what have you got in /etc/fstab?  What does /proc/mounts report?

/etc/fstab currently has 

...
# Dynamic entries
...
/dev/hda9 /mnt/hda9 ext3 noauto,users,exec 0 0

It seems to be auto-detected and this time those errors didn't appear, so that 
problem seems to be cleared up, at least for now. I'm not sure if that's a 
result of dunning fsck.ext2 on it.

> > On one boot it had lost the system time
>
> That's not uncommon.  Windows always assumes the system clock is set to
> local time, yet most Linux distros (Debian is symptomatic of this)
> assumes that the system clock is set to UTC time.  If you're using
> Debian, then changing the value of:

As it's worked correctly for around a year, I think the configuration is 
unlikely to be the issue. 

> > On another boot I had to reset the HDD details in the BIOS to get it
> > to boot

I forgot to mention that the BIOS boot order had changed.

Taking all 3 things together, I guess the CMOS battery is a likely culprit.

> > On the most recent boot it complained about unknown filesystem type
> > 'supermount'
>
> I hope you're not referring to the nasty, and definitely-deprecated
> version of 'supermount' that attemps to poll the drive to see if a disk
> has changed.   If it is, remove it.  You don't need it.  But again, you
> must look at your /etc/fstab file.

neil ~ 18:26:05 505 > grep supermount /etc/fstab
none            /mnt/floppy supermount 
dev=/dev/fd0,fs=auto,--,iocharset=iso8859-15,sync,codepage=850,umask=0 0 0
neil ~ 19:22:04 506 > 

Ooh, nasty. 

> > The thing I most want to sort out is the network not coming up. I've
> > looked for anything that's changed in that area. None of the relevant
> > config files and binaries seem to have changed for months. The
> > network
> > card works in Windows, so I guess the card itself is OK.
>
> One would assume so, yes.  Diagnosing this sort of issue, is best done
> manually.  The first thing you should do is check that your card is
> supported by the kernel.  That also includes that you have the correct
> modules.  If you run:
>
> ``
> % grep -i 8139 /boot/config-$(uname -r)
> ''

> The line that you're probably most interested in is this one:
>
> ``
> CONFIG_8139TOO=m

Yes. it's a module. For some reason I couldn't find that module when I first 
looked, but it is there.

> If it's a module, somewhere in your init sequence, the line:
>
> ``
> modprobe 8139too
> ''
>
> .. happens.  For Debian, the file /etc/{modules,modprobe.conf} is used
> -- the former file, is used for 2.4.X kernels, whilst the latter is
> used for 2.6.X.

Both of those are pretty empty. As the base system was Mepis, I guess meauto 
does most of the work of detecting and loading up modules I can't find any 
documentation for meauto. 

> > The card is a cheap RTL8139 IIRC, which has worked for years. lsmod
> > doesn't show an rtl8139 module loaded. Is there an easy way to tell
> > if
> > the kernel has support built in?
>
> See above.  The module name is '8139too'.
>
> Having assured yourself that support for that card is in the kernel,
> albeit via a module, or as an in-built resource (in which case, you
> generally won't have to do anything with it), then the next stage is to
> try and configure it.

I skipped most of that and went straight to a simple 'modprobe 8139too', whcih 
worked first time. I then did 'ifup eth0' and 'pump' and it all works. The 
question now is why isn't this happening automatically any more. 

I'm sure I can get it working, but I guess I'll ask around on the Mepis forums 
before doing anything drastic.

Thanks for that, Thomas. Some very useful pointers in there.

Neil