A while back a problem was reported about an interface just blocking and not
transmitting any more packets, i replyed to the message to say i have had
the same problem but it had only happend once or twice, this was with kernel
2.0.30 ax25-utils-2.1.42a and ax25-modules-14c.
Now i have upgraded to kernel 2.0.31 with ax25-modules-14d, utils not
changed.
Now this failing interface problem is showing up on a daily basis, it
happens when someone connects to node on port scc0 and telnets to another
host on scc3. It seems to happen when the Tx queue creeps up over the 5000
bytes mark, i was at the console when it happend and saw what was going on.
scc0 = 1200 baud, ouch,
scc3 = 4800 baud, so the buffers do sometimes fill up very quickly.
In syslog the following entry:
ax25d[112]: accept error Invalid argument, closing socket on port SCC0
last message repeated 2 times
When i did ifconfig scc0 down, the following was enterd into syslog.
kernel: z8530drv: oops, deallocated only 2 of 12 rxbuffers
This must be from scc0 as the other interfaces have different buffer
settings.
The z8530drv error, has been seen in my logs before in many previous kernel
versions and driver versions, but has never realy caused any problems, so i
am not saying that the z8530drv is causing this particular problem.
Also i have exchanged mails over it with Joerg in the past.
Is anyone else having this sort of problem,? kernels concerned are 2.0.30 /
31, i have another machine running 2.0.29 there are (seem) to be no problems
there, it has 3 scc interfaces 2 terminals and 3 stn's via eth0, so compared
to my system here at home it is much busyier but has no buffer problems.
I did have it happen in 2.0.27 once tho'.
The only noticable change from 2.0.30 to 2.0.31 is that i am getting
RxOverruns on scc0 which i rarely had in 2.0.30. I noticed BEFORE the
interface locked up this morning that there were no RxOverruns, after the
problem there were 7, and 1 TxError.
Buffers from scc0 are.
Buffer size: 1024
rx buffers : 12
tx buffers : 12
MTU from scc3 is 512 and the scc0 interface is 256.
Can it be contributed to scc card systems only, or is there someone who has
had this sort of thing happen using a TNC?.
Would the other person who reported this problem first please look into your
logs and see if you have any error messages which could help here.!!
Can the 2 error messages in the log be in anyway connected?.
Joerg, Terry, Jonathan comments please ??.
Nothing has changed in my configuration t1 t2 timing, etc.
Z8530drv version 2.4c
Nettools 1-33 with Terry's rose patch applied.
Both modutils-14d and Terry's rose patch went in cleanly, no warnings seen
when compiling either kernel, modules or nettools.
-- Regards Richard, 73. inet xfw.enlkz@tcsn.net