Well i have done some experimenting by way of setting an external squelch
to the point that it is constantly flikkering, trying to simulate soft dcd,
while the amount of RxErrors keeps increasing like an interface with a soft
dcd does, the interface does not go to sleep, its a 4k8 ifc.
This might indicate that the problem is not caused by the soft dcd,??.
>
> My 9600 baud port does not seem to be affected, I was pondering if it was thst
> the 9600 baud p[ort uses external dcd.
I have had a case of a 4k8 port going to sleep ONCE and once only, this was
way back when i used 2.0.27, indicating that the problem has always! been
there.
>
> Finding a beter solution to the problem in the software driver woiuld be
> better, as we have tewo different cards then the problem seems to lie with the
> linux driver.
Well as to what is the cause i would be more careful as to say it lies with
the driver, there is a known problem with ax25/sockets, and i am wondering
if this "might have something to do with this particular problem".
The following text is taken from the Documentaion directory of the
pre-2.0.34 kernel source tree.
Socket crashes
A possible socket layer crash with AX.25/NetROM/ROSE/X.25 has
been fixed in 2.0.34
Unfortunatly the "fix" introduced an even greater problem, as to what is
being done about it i dont know, i havent been following the pre-34 pacthes
lately due to lack of time.
>
> Do you also notice a high level of rx errors oin both 1200 bad 9600 baud
> ports????
The level of RxErrors will keep increasing with soft_dcd because every spek
of noise makes an interrupt thereby increasing RxErrors dramacticly.
I have several types of modems all using different chip types for the
hardware squelch, it seems to me that a MC1458 is the best chiose. On a 4k8
interface with a MC1458,
Sccstat shows.
Received : 1866
RxErrors : 2714
On another 4k8 ifc using a TL084.
Received : 113009
RxErrors : 1919647
But this interface also has some 1k2 traffic so that just shows up as
RxErrors, confusing the issue somewhat.
>
> |
> | My situation is a little different, it might stay up for
> | weeks on end,
> | othertimes it dissapiers twice a day.
>
> Mine also varies in thime from an hout to a couple of days
I have made one observation which may be of interest to Joerg.
When my "go to sleep" interface is handling IP traffic (routing packets
from 1k2 to 9k6) it seems never to fail, it always goes to sleep when;
pa3xxx connects with ax25 at 1200 baud to node which is called by ax25d,
he then connects either to a mailbox somewhere futher via the 9k6 interface
or joins the convers server, both of which send him large amounts of data.
I have noticed that when i have to retransmit packets to him and the
Txqueue stays high (i have seen it at 12k) then out of the blue the
interface goes to sleep.
It never seems to go to sleep when;
pa3xxx connects via ax25 on the 9k6 interface which also has ax25d open node
he then simulates the same as mentioned above, there seems to be no problem
this way round.
I wonder if that can help Joreg???.
> |
> | I have mentioned this problem here several times, however
> | there has been
> | little responce.
>
> Maybee we can get a few ideas running.....
Who knows.
> I am using the 2.4c driver.....
So am i.
> 73 Peter VK3AVE
73
-- Regards Richard. fmlab@mx.dy.fi