This is the mail archive of the ecos-discuss@sources.redhat.com mailing list for the eCos project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]

Re: Re: Ignoring packet error...


On Tue, 7 Nov 2000, Jonathan Larmour wrote:

> "richard.panton" wrote:
> > 
> >   When I first connected the Assabet board to my PC using the Intel
> > serial/parallel cable, I got these errors. I could see output from the
> > RedBoot monitor, but nothing I typed had any effect, and running GDB gave
> > the 'Ignoring packet error' messages.
> > 
> >   It turned out that the serial cable is a three wire cable - ie. does not
> > provide handshaking signals - and that for some reason, Linux ALWAYS uses
> > some form of handshaking (I think DSR/DTR handshaking).
> [ sorry for reviving this old topic ]
> 
> Did you check using a terminal emulator like minicom that there definitely
> was no handshaking? I know the default is to do handshaking but it can be
> set off. minicom is useful because it makes the current setting obvious.
> 
> stty would probably do as well of course if you are sufficiently versed in
> that.

I've just done some more experiments.

The Assabet serial cable has pins 2, 3, and 5 connected (ie, RxD, TxD, and
GND). I've added connectors between 1 and 4 and 8 (ie. DCD, DTR, and CTS,
all connected together - NOTE all references to pins and connectors are at
the Linux end of the cable).

With this cable, under Linux, I ran gdb. Everything worked. I removed the
additional connectors (without exiting gdb). Everything still worked!!!

I rebooted to WindowsNT (without power-down), and ran HyperTerminal. The
RedBoot prompt appeared, and I could type commands.

I rebooted to Linux (again without power-down). Running "cu -s 38400
--nostop -l ttyS0" and reseting the Assabet board produced the RedBoot
startup banner, but pressing Enter (or any other key) had no result.
Similarly, gdb could not connect to the board (and "stty -a < /dev/ttyS0"
hangs without producing output).

I powered down the PC, and then booted Linux. Still no response from cu or
gdb. (or stty, although interestingly, whether "stty -a" works seems to
depend upon what was run last/is running).

I connected pins 4 and 8 (DTR - CTS!!), and cu began working (i.e.,
pressing Enter gave a new RedBoot prompt). stty produced results. However,
gdb could still not connect.

I connected DCD - DTR - RTS. cu and gdb now work.

Connecting all these pins together may not be an optimal solution, but it
does work (FOR MY SERIAL HARDWARE, DRIVING A THREE WIRE LINK). And no, I
don't have a proper explanation for WHY it doesn't work without these
connectors. However, I came across pretty much the same symptoms in the
mid-90's when writing a terminal emulator for EPOC devices
(http://www.iota.demon.co.uk/psion/hermes/hermes.html), and attempting to
log in to my Linux box; I had to modify a cable (tying DCD active at the
Linux end) to get past the login prompt. [Since then, new getty versions
have come along that allow me to log on with my terminal emulator without
cable hacks].



Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]