From: Niall Parker (xaz.lcndpe@umsl.edu)
Date: Fri Jan 05 2001 - 07:30:50 EET
On Thu, Jan 04, 2001 at 09:33:54AM -0500, John Ackermann wrote:
> In message <qlqrhb.smtu@mail.dy.fi>
> Tomi wrote:
> >On Tue, 2 Jan 2001, John Ackermann wrote:
> >
> >> The second problem relates to the dmascc driver which is running an
> >> Ottawa PI card. In ifconfig's output I'm seeing a lot of RX frame
> >> errors -- about 1/3 of the total number of RX packets shown. What does
> >> the "frame" error signify? Could it be a sign of either TXD or TXtail
> >> being too short on one of the stations being heard, such that the flags
> >> are being cut off? I didn't notice this error with the old
> >> motherboard, but that may just be a coincidence.
> >
> >I have tried to keep quiet and wait for some who really knows come up with
> >an answer but seems it isn't going to happen... :)
> >
> >Anyway I seem to remember quite a few reports from people with PI2
> >cards having problems with the newer motherboards. I don't remember the
> >symptoms but it was some sort of errors on rx or tx (helpful, eh?).
> >
> >Someone even came up with a possible explanation that had something to do
> >with the Z8530 DMA interface not being quite in spec for direct connection
> >to the ISA bus. It all worked just fine with the way ISA bus was
> >implemented before the age of VLB and PCI but the PCI-to-ISA bridges
> >changed all that. They work within the ISA timing spec but in way that
> >ruins communication with the Z8530 chip. Or something like that.
> >
> >All I have is just a vague recollection of discussions on some mailing
> >list long ago and I don't remember the details nor do I have any first
> >hand experience (I have some PI2 cards but they mainly just gather
> >dust...). Thought I'd mention anyway...
>
> Hi Tomi --
>
> Yes, there were problems that seemed to be related to bus timing issues
> with the faster 486 machines. When I saw the problem, it tended to be
> a total failure -- I think the TX keyed up and never unkeyed. Changing
> from an original 8530 to an 85C30 chip on the board seems to help.
>
> The problem I'm seeing may be related to that, but the more I look at
> it the more I think it is either a problem decoding from one station,
> or possibly not a real problem at all; the RX seems to be working OK,
> and although I'm sure there are some dropped packets, I don't think
> that we're actually losing 25-33%.
>
> That brings up another question -- would a frame reported as an "error"
> by ifconfig still be passed on from the driver, or is it dropped on the
> floor?
>
> 73,
> John
Hi guys,
There is definitely some problems with the DMA on the PI2 cards which
we have seen here on various occasions. Typically the system looks like
it is working but wouldn't transmit decodeable packets. For a while we
had an external reboot box ...
This was all with the original PI2 driver by Dave Perry ... I haven't
seen the same problem with the dmascc driver (I think the author of the new
driver made a comment about the original driver losing packets). The
card has a problem with DDIOW timing, delaying it could fix the Tx
problem on some motherboards. The latest revision of the card (PI2 rev B,
August 1995) includes some flipflops and a jumper block to slow down
DDIOW (we had previously used a RC kludge)
I have also noticed frame errors with the latest driver and kernel 2.2.x,
I don't think they were there when using the dmascc driver on the
2.0.x kernels but can't be to sure (only used the system for a short while)
They don't seem to interfere with operation of the system so I haven't
tracked them down further (we seem to have enough other problems keeping
the system up :(
Given the PI2 flakyness, we have been looking for other cards. Anyone had
much experience with the PCISCC card that came out recently from Germany ?
... Niall
-
To unsubscribe from this list: send the line "unsubscribe linux-hams" in
the body of a message to gwtabr.nldedtzu@iwic.net
This archive was generated by hypermail 2b30 : Fri Jan 05 2001 - 08:08:52 EET