Re: ppp/diald networking dying under heavy IDE (or CPU?) load

["Daniel B." <REMOVEdanielTHIS@fgm.com>]

Jens Simmoleit wrote:

> Hi Daniel :-)
>
>
> > Is there a known reason why heavy IDE load (or possibly heavy CPU
> > load in copying disk to disk) would cause network connections over
> > ppp and diald to die?
>
>
> Could it be by any chance that the NIC and the ide controller and also the
> serial port controller are hooked up on the same IRQ? 

At least according to /proc/interrupts, the serial port doesn't seem to
be sharing interrupts:

# more /proc/interrupts
           CPU0       CPU1
  0:    1924074          0          XT-PIC  timer
  1:      13204          0          XT-PIC  keyboard
  2:          0          0          XT-PIC  cascade
  3:     490726          0          XT-PIC  serial
  5:          0          0          XT-PIC  eth0
  8:          3          0          XT-PIC  rtc
 10:          0          0          XT-PIC  CMI8738-MC6
 11:      53222          0          XT-PIC  ide2, ide3
 12:     125384          0          XT-PIC  PS/2 Mouse
 14:    3696354          0          XT-PIC  ide0
 15:     121555          0          XT-PIC  ide1
NMI:          0          0
LOC:    1922900    1924053
ERR:       3287
MIS:          0
#

I was thinking that it was only disks on my ide0 and ide1 controllers
(ports?) (for which I had disabled DMA to avoid horrible file system
corruption) that were involved with network problems, but I need to see
if things differ for disks on my ide2 and ide3 controllers (a Promise
Ultra100/TX2 card).

I am getting occasional "kernel: PPP: VJ decompression error" in my
logs.  Might that result from dropped characters from a buffer
overrun (e.g., if the CPU is too busy doing PIO disk I/O to service
the serial port frequently enough)?


Daniel