[tech] Re: [committee] 24 hour warning: This Week's Committee Meeting Agenda

[Bernard Blackham]

On Wed, Jun 16, 2004 at 12:13:40PM +0800, Davyd Madeley wrote:
> > It was already setup (i2c-proc doesnt exist in 2.6 though).
> > sensors was installed in /usr/local/bin, taking priority in the path
> > and that version didn't look in /sys.
> 
> Can someone remove the /usr/local version then at some point?

I've chmodded it -x for now. If we ever go back to a 2.4 kernel then
we'll probably need it again. (Is that why it was there?)

> > They also do bad things(tm) wrt interrupt handlers (see dmesg).
> 
> Interesting. It's the same driver I'm running, but I am not getting
> nasty going on. Even though the machine has APIC turned off, we are
> nowhere near running out of interrupts.
> [davyd at pitch davyd]$ cat /proc/interrupts
>            CPU0
>   0:   57925629          XT-PIC  timer
>   1:      36745          XT-PIC  i8042
>   2:          0          XT-PIC  cascade
>   9:          0          XT-PIC  acpi
>  11:    4603365          XT-PIC  NVidia nForce, nvidia
>  15:    8356849          XT-PIC  eth0, ohci_hcd, ohci_hcd
> 
> I guess this is a cute side-effect of not having any disks in there.

The badness is coming from the fact that the nVidia drivers are
calling functions from within interrupts that they're not supposed
to. I don't see how running out of interrupts or being diskless
correspond to this, but then again, I can't explain why we
pitch/velvet them and you don't :) They don't seem to be harmful yet
anyway.

> It would be nice to get some mrtg style monitoring on the machines.
> Perhaps we can look at that later on in the week.

6 days and I can get something going.

Bernard.

-- 
 Bernard Blackham <bernard at blackham dot com dot au>