[ltp] Filesystem choice?

[Tod Harter]

On Wednesday 06 November 2002 09:17 am, Michael Selway wrote:
> Tilmann Singer writes:
>  > * Richard <rn214@hermes.cam.ac.uk> [20021106 12:21]:
>  > > My A22p used to fail about once per month to resume. It now fails
>  > > about 1 time in 3 (It seems that the HDD doesn't spin up). So I don't
>  > > let it suspend anymore - I just turn off the screen (fn-F3) at night.
>  >
>  > Sounds like the errors I've seen on my A20p after upgrading the kernel
>  > to 2.4.19. I remember reading somewhere that this is a known error in
>  > that kernel, went back to 2.4.18 and everything was fine again.
>
> When I upgraded from redhat 7.2 (2.4.not-much) to redhat 7.3
> (2.4.18), the crash-on-resume problem got much much worse, like 1
> time in 2 failures or so.  Ironically, I'd only upgraded in the
> hope of fixing the occasional crashes.  I aleviated the problem
> back down to one-or-two crashes per month by shutting down my
> permanently-inserted pcmcia modem card before suspending (by using
> the apmd/apm-proxy stuff in redhat).  I've tried adding all sorts
> of exclusions to the pcmcia config file, but to no effect.
>
> Indeed, I don't use any pcmcia cards now, and I still get the
> occasional crashes.
>
> I agree with Richard that it seems like the problem is that the
> hdd doesn't spin up.  I played with reseting the IDE bus using an
> obscure ioctl() call, but that crashed the machine even more
> reliably(!).  I wondered if use of the CD/DVD drive was related,
> but I've had crashes when the drive has not been used since the
> last reboot.  I share richard's thoughts about using alt-sysrq to
> get the ide driver to reset things.  Maybe I should take a look at
> this if I can find some time.
>
> I tried setting the APM_ALLOW_INTS kernel parameter to
> no-allow-ints.  It's a "well known fact" that thinkpads don't work
> with this setting, so well-known that there's code in
> arch/i386/kernel/dmi_scan.c which forces allow-ints if you're on a
> thinkpad.  I modified this code to allow me to play with the
> parameter, and for a while I thought I'd got it.  But I now
> believe this parameter setting makes no difference to my machine
> at all.
>
> My latest stab is to turn off interrupts and dma from the disc at
> suspend time, as suggested again by the redhat /etc/sysconfig/apmd
> stuff.  I've been up for an inconclusive 6 days so far...
>
> I have a long-held theory the problems might be related to NFS,
> but I don't have much evidence for this.  I use NFS a lot so it's
> harder to test this hypothesis.
>
> Incidentally, I always shutdown using "apm --suspend", I wonder if
> that's an issue...?
>

In my case I'm using vanilla MDK 9.0, on an A20p but I had pretty much 
identical results with MDK 8.1 (I skipped 8.2). My symptoms are that roughly 
1 out of 3 resumes I come back up with X completely crapped, just garbage on 
the screen and the machine locked solid. I don't use any PC cards, so its not 
related to that, it just seems that the machine fails to reinit SOMETHING. I 
suspect it might be related to whether or not the network is up or not, but I 
have also told apmd to shut that down on suspend (as well as shutting down 
sound, which seems to be mandatory on my machine). Way back when I was 
running MDK 8.0 on this same machine and using kernel 2.4.9 it DID seem like 
it was a lot more reliable, but I never had 100% of resumes work, and its 
always been the same symptom when it crashes.

BTW, I believe alt-sysreq is a response thats hard-coded into the kernel. 
There are a number of things you can do with it, like dump to a serial port, 
kill init, etc. Its one of those linux features I so rarely need to use I 
can't recall off the top of my head what the key combos are, but I do know 
that you can customize it, I just can't recall for the life of me the details 
at the moment, and too busy to go look it up!

> Michael.