Linux-Hams archive - July 1997: Re: KPC3+ problem and serial chip

Re: KPC3+ problem and serial chip

Barry McLarnon VE3JF (ghlf.cbyvol@aisinusa.com)
Sat, 5 Jul 1997 03:33:34 +0000 (GMT)


Albert Woodhull <urrholsj@osl.no> wrote:
> A while back I asked for help on a problem I have had with the KPC3-Plus
> TNC (but not the older KPC3) in KISS mode. I still haven't solved that
> one, nor has Kantronics answered my query to them, but I received a reply
> that has me wondering about another possible problem.
>
> One writer suggested I would have a problem with either Windows or Linux
> and a KISS TNC if I do not have a 16550AF serial interface chip. I have
> noticed on my listen screen that sometimes I get packets that aare missing
> a character or two. The characters are not garbage, they seem to have
> dropped out. This would be consistent with a serial buffer problem.
> This is different from the problem I had earlier, although I suppose it
> could be related -- I suppose there are bytes that could cause something
> that looks like a wild buffer overrun problem if they dropped out.
>
> My Linux dmesg and an old Microsoft MSD tell me only that I have a 16550A.
> Question number one is, can someone confirm that an ordinary 16550A is not
> good enough? Question number 2 is, can the Linux kernel detect the difference
> between the A and the AF (I am now using a 2.0.30 kernel)? Question 3 is,
> is there anything I can tweak in the kernel code to compensate?

AFAIK, the 16550A and 16550AF are functionally identical. The original 16550
had a non-functioning FIFO. National Semi added the 'A' after the bug was fixed.
I think they added the 'F' later just to emphasize that the chip now had a functioning
FIFO... that's my theory, anyway.

Barry

--
  Barry McLarnon  VE3JF/VA3TCP   |  Internet: zrtrvg.tlxyibrz@goanna.net.au
  Ottawa Amateur Radio Club      |  AMPRnet:  terhi.victor@logonet.com
  Packet Working Group           |  Web:      http://hydra.carleton.ca