2003-02-14 12:03:13

by Michael Westermann

[permalink] [raw]
Subject: strangely serial Problem


We have a strangely serial Problem.

I have a box with Kernel 2.2.23 and Suse and a box with
Debian with kernel 2.2.23. Old IDE-Disk's without DMA.
Both boxes are connect with crossober Cable.

On the reading box run's a "find /[!p]* -type f -exec grep "balffasel" {} \;
to the same time.

I have tested with:

stty -F /dev/ttyS0 0:0:800008bf:0:3:1c:7f:15:4:0:1:0:0:0:1a:0:12:f:17:16:0:0:0:0

38400 baud

cat /dev/ttyS0 > testdat

on the reading box

for i in `seq 1 500`; do \
printf "%04d 0123456789012345678901234567890123456789012345678901234567890\n" $i
> /dev/ttyS0 ; done

on the writing box

When the debian-Box send the characters, to get lost characters.
When the Suse-Box send's it's ok. i

I have tested only with a runnig bash "lili: linux init=/bin/bash"
the same Problem. On both boxes are the same kernel, the same
serial parabeters, both boxes work with FIFO.

I have tested the kernel 2.2.16...23 and 2.4.2 2.4.17 2.4.20, only
the 2.4.17 has the same problem on debian.

I think that is a interrupt Probelem,but what is the causale?
what the difference.

When I use the FIFO, the UART read 8 char's to put a Interrupt.
The time for 8 char's is 2 ms. After this time we have a overrun,
but I thin that is a long time witout a Interrupt.


Michael Westermann