Re: Serial port mis-detected?

["Denis A. Doroshenko" <d.doroshenko@omnitel.net>]

the problem you descibe is very difficult to happen (according to source
code of the pccom driver). to be detected as ti16750, uart should report
about working fifos (i.e. claim it's 16550A) and should also report about
64 byte fifos (texas instruments extention). the on thing which is really
doubtfull for me is line #544:
                bus_space_write_1(iot, ioh, com_lcr, lcr | LCR_DLAB);
i wonder if DLAB is needed to enable 64 byte fifos! on my 2.5-patched
pccom it works with _real_ ti16750 with this line commented.

anyways, unfortunatelly openbsd's pccom driver doesn't support more
flexible work with uarts (for example there's no possiblility to
setup fifo settings, they fifos will be set to 0 -- no fifos -- for
speeds <= 1200 bps, for higher speeds half of fifos will be used).

On Fri, Mar 31, 2000 at 09:54:10PM -0600, Michael S. Keller wrote:
> Host has a multi-I/O card (2 com, 1 parallel, IDE, floppy) with second
> serial port disabled.
> 
> Host has another multi-I/O card (2 com, 1 parallel, game) with only
> second serial port enabled. UART socket for this port contains a
> 16550A. Kernel detects it as a "ti16750". When I configure PPP at
> 38400bps on cua00 (the 16450), it connects fine. When I change only to
> cua01 (the 16550A) and move the modem, it won't work. Forget 115200bps.
> 
> This modem worked at 115200bps on this very serial port in my Linux
> gateway, so I have little reason to suspect the hardware.
> 
> Any easy way to tell the kernel that this UART has a 16-byte FIFO, not
> a 64-byte?

-- 
Denis A. Doroshenko
Omnitel Ltd., Sevcenkos 25, Vilnius 2600, Lithuania
mailto:d.doroshenko@omnitel.net