Re: Total lockups using ext3

Re: Total lockups using ext3

[Anthony Campbell]

On 01 Feb 2002, Stephen C. Tweedie wrote:
> Hi,
> 
> On Fri, Feb 01, 2002 at 06:10:48PM +0000, Anthony Campbell wrote:
>  
> > They always occur when I am online, downloading. The moden stops, the
> > hard disk light is on, and no key works. They never happen with 2.2.20
> > or with earlier (i.e. pre-ext3) 2.4 kernels, and they never occur on my
> > laptop with any kernel, including those using ext3.
> > 
> > Since inactivating ext3 (by changing /etc/fstab) the lockups have ceased
> > to occur.
> > 
> > I therefore conclude that the lockups are due to ext3. I can manage
> > without it but would like to use it if possible. Any suggestions please?
> 
> Doesn't sound like a filesystem problem --- if the hd remains on, it
> sounds very much as if you've got something going wrong in the disk
> stack, either in the ide driver or the disk itself giving up.  The
> fact that it always happens when the modem is on also suggests that
> this is not a filesystem problem.
> 

> Try reproducing it with a readable console enabled --- either run in
> text console mode (no X), or set up a serial console and use another
> machine to capture the output.  "ttywatch" is excellent for monitoring
> serial consoles.
> 
> Cheers,
>  Stephen
> 

I doubt it's the disk; it works perfectly without ext3, even downloading
a complete kernel tarball. I'll try the text console later (though
probably at the cost of yet another reboot...)

Anthony

-- 
Anthony Campbell - running Linux GNU/Debian (Windows-free zone)
For an electronic book (The Assassins of Alamut), skeptical 
essays, and over 150 book reviews, go to: http://www.acampbell.org.uk/

Our planet is a lonely speck in the great enveloping cosmic dark. In our
obscurity, in all this vastness, there is no hint that help will come
from elsewhere to save us from ourselves. [Carl Sagan]

Re: Total lockups using ext3

[Anthony Campbell]

On 01 Feb 2002, Andrew Morton wrote:
> "Stephen C. Tweedie" wrote:
> > 
> > The fact that it always happens when the modem is on also suggests that
> > this is not a filesystem problem.
> 
> mm.. There are known scheduling-inside-spinlock bugs in the PPP
> driver.  Paul is working on these at present.  This will cause
> deadlocks on SMP.  Anthony seems to be running uniprocessor, but
> bad things could still happen.
> 
> It'd be interesting to enable the local-apic-on-uniprocessor
> and io-apic-on-uniprocessor options, then boot with the
> `nmi_watchdog=1' LILO option, see if you can get an NMI oops
> backtrace.
> 
> 

I discovered recently that, at least in 2.4.17, I had included SMP in
the configuration by mistake (in fact, I think it happens by default).
Eliminating it seems to have solved the lockups, or at least made them
much less frequent.

Please cc any replies to me as I'm not on the list at present.

Anthony


-- 
Anthony Campbell - running Linux GNU/Debian (Windows-free zone)
For an electronic book (The Assassins of Alamut), skeptical 
essays, and over 150 book reviews, go to: http://www.acampbell.org.uk/

Our planet is a lonely speck in the great enveloping cosmic dark. In our
obscurity, in all this vastness, there is no hint that help will come
from elsewhere to save us from ourselves. [Carl Sagan]