From: Hans Grobler (hxwww.utnhssoj@promise.com)
Date: Tue Oct 10 2000 - 09:55:31 EEST
On Mon, 9 Oct 2000, Richard Adams wrote:
> On Mon, 09 Oct 2000, Niall Parker wrote about , Re: Status of drivers pi2.c pt.c:
>
> > I just switched one of our rf routers (using a PI2 card) from 2.0.37 with
> > the pi driver to 2.2.13 with the dmascc driver. While it seems to run OK
> > on a P75, the new stuff didn't seem to work as well as the pi driver on
> > a couple of 386's ... (lost packets due to Tx overuns). This may not be
> > the dmascc drivers fault per se though, as I recall trying both the pi and
> > dmascc drivers earlier with a 2.0 kernel and they seemed equivalent.
>
> I have several machines running a PA0HZP compatable scc using the normall
> scc.c driver from Joerg DL1BKE, i can say that the slower the PC is the
> better they work.
>
> I even have a P2 200 with 64 megs which has a SCSi drive, i "still get"
> many Rxover and Txunderuns. I must say i have 2 ethernet cards as well,
> which could well contribute to the problem
>
> When i used a 486/dx2 66 with 16megs of ram the amounts were "much smaller".
> Another problem is the amount of corrupt callsigns which appier in the
> mheard list, in that machine i only has one ethernet card. However i can
> live with it all as the systems i have keep running and "seldumly have
> problems".
We have, amongst other, an original PA0HZP here and don't see particularly
high Rxover and Txunderuns. The machines I've used are: Pentium 60,
Pentium 233 and a Celeron 333. The second one had IDE and the others
SCSI. All the machines have a single 10BaseT network card. Kernel 2.2.x.
Be that as it may, it does not surprise me that there are variations in
packet loss characteristics between the various scc drivers and cards (and
due to the machines they are used in). The scc IC's are (at least in my
opinion) particularly sensitive to interrupt latencies and I/O sequencing.
As a side note, there also seems to be anomalies with the 2.4.x kernels
and I'm still trying to get a handle on. For example, under certain
conditions, the scc.c driver will cause an OOPS and kill the machine. This
seems to be related to the new threading of the network layers.
-- Hans
-
To unsubscribe from this list: send the line "unsubscribe linux-hams" in
the body of a message to pzvm.mfyvkillqs@roskapostia.tunk.net
This archive was generated by hypermail 2b30 : Tue Oct 10 2000 - 09:59:59 EEST