We've been trying to set up a laptop here to use a Novatel Minstrel PCMCIA
modem (wireless Richocet network). The card shows up as a serial port
(ttySx) and accepts AT commands just like a normal modem.
It dials fine, PPP connects, gets IPs, etc just as it should. However,
any packet over about 400 bytes gets dropped on the recieve. Also, the RX
errors on ppp0 increment occasionally.
TCP connections connect (because the SYN's are small), but as soon as you
start trying to do bulk transfers (`ls -la` in an ssh window, or an HTTP
GET), the connection stalls. Pinging other hosts also works fine, except
when you do -s with a value larger than 300 or so.
It doesn't work with anything we've tried on 2.4 (changing mtu/mru, serial
port speed, etc). However, under 2.2.x, we were able to get connections to
stay running and not stall by setting the MTU on ppp0 to 120 after the ppp
comes up. As you can imagine, this makes the modem seem even slower than
it already is.
Not that its relevent, but pppstats shows 0 in the 'vjcomp' fields of
both rx and tx (as well as 'vjerr'). I've tried starting pppd with and
without 'novj' just in case. Same result.
Any ideas? The 'rx error' count going up is kind of suspicious. My
attempts at getting pppd to print more debugging output have been
futile; aparently the debug and kdebug options no longer work ('debug'
produces the LCP traffic, yes, but thats working fine).
af
Adam Fritzler wrote:
> We've been trying to set up a laptop here to use a Novatel Minstrel PCMCIA
> modem (wireless Richocet network). The card shows up as a serial port
> (ttySx) and accepts AT commands just like a normal modem.
>
I'm in the process of bringing up a Novatel Merlin now (tomorrow morning
actually). I don't know how similar it is, but they do at least use some
common hardware on the Merlin and Sage products.
> It dials fine, PPP connects, gets IPs, etc just as it should. However,
> any packet over about 400 bytes gets dropped on the recieve. Also, the RX
> errors on ppp0 increment occasionally.
>
According to http://people.freebsd.org/~nsayer/ you have to set the mru
to 576 for the Merlin. There is some fairly detailed info on
http://www.mrollins.com/newtmerlin.html about setting it up for (of all
things) a Newton including all the expected commands which implies that
you should adjust the mru setting on the card and then match that with ppp.
> TCP connections connect (because the SYN's are small), but as soon as you
> start trying to do bulk transfers (`ls -la` in an ssh window, or an HTTP
> GET), the connection stalls. Pinging other hosts also works fine, except
> when you do -s with a value larger than 300 or so.
>
> It doesn't work with anything we've tried on 2.4 (changing mtu/mru, serial
> port speed, etc). However, under 2.2.x, we were able to get connections to
> stay running and not stall by setting the MTU on ppp0 to 120 after the ppp
> comes up. As you can imagine, this makes the modem seem even slower than
> it already is.
>
> Not that its relevent, but pppstats shows 0 in the 'vjcomp' fields of
> both rx and tx (as well as 'vjerr'). I've tried starting pppd with and
> without 'novj' just in case. Same result.
>
> Any ideas? The 'rx error' count going up is kind of suspicious. My
> attempts at getting pppd to print more debugging output have been
> futile; aparently the debug and kdebug options no longer work ('debug'
> produces the LCP traffic, yes, but thats working fine).
>
> af
>
> -
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to [email protected]
> Please read the FAQ at http://www.tux.org/lkml/
--
Joe deBlaquiere
Red Hat, Inc.
307 Wynn Drive
Huntsville AL, 35805
voice : (256)-704-9200
fax : (256)-837-3839
Next time, I'll finish reading everything in my l-k mailbox before I
post...
For the archives:
See the thread entitled 'Serial device with very large buffer' started by
Alex Belits <[email protected]> for a patch to fix this
bug completly. (Thanks for the great timing, Alex.)
The modem is now working perfectly under 2.4.1 (on a Sony vaio
C1VN... very nice).
af
On Fri, 2 Feb 2001, Adam Fritzler wrote:
> Any ideas? The 'rx error' count going up is kind of suspicious. My