On Mon, 17 Feb 2020, David Given wrote: > On Mon, 17 Feb 2020 at 00:30, Paul Osmialowski <pawelo@xxxxxxxxxxx> wrote: > I'm using null-modem cable I've made myself and been using it for years > (it corsses-over Rx/Tx and hardware control lines). It works nicely with > Telix running under FreeDOS on my XT and minicom on 'big' Linux, at speed > 115200, which is the top speed for 8250 chip on the XT side. > > > Exactly which 8250 is it? Some versions had major bugs which needed non-backwards-compatible workarounds in the BIOS. If the ELKS driver doesn't have these workarounds, which is likely if it's been targeted at later versions of the 8250 which didn't have the bugs, > then it's likely not to work very well. I also believe (heard from Random Person On The Internet) that driving an 8250 at high speed is difficult due to the very small buffer, and you need very good interrupt performance, which I don't know if ELKS has. > > Information on the different 8250s is here: https://www.freebsd.org/doc/en_US.ISO8859-1/articles/serial-uart/index.html > In this particular case (Amstrad PC 2086) it's on-board Amstrad 40049 UART (the same chip as used for a serial port in Amstrad Portable PC models), I don't know any more details, except that all DOS programs (e.g. Norton SI) report it as 8250. Yet at my old home I have Turbo XT with old, long 8-bit ISA I/O card having two regular 8250's on it and as far as I remember, it was the first time I experienced losing bits while communicating ELKS and Linux running on a bigger PC.
