From: Paul Kelly (longword at domain esatclear.ie)
Date: Mon 09 Sep 2002 - 19:58:05 IST
Niall O Broin wrote:
> So if YOU (which I know includes
> Vince Cunniffe and Paul Kelly but I'm sure there are more) have connected
> non-Eircom supplied equipment to an Eircom ADSL line, please let me know
> what it was and if there were any "issues" with using it.
The web site sounds like a good idea. I have a Conexant chipset "HCF"
PCI ADSL modem. The only issues I've had with it were of my own doing,
and I've only fixed that today so I'd want to leave it a week at least
before passing final judgement. All of the hard bits of ADSL are done on
the card, with the host OS taking care of everything above the basic ATM
cells - including SAR (turning the tiny 53 byte ATM packets into more
It's made by these people: http://www.amigo.com.tw/ and currently sold
under the "Dabs Value" brand (watch the "Mfr code") at
http://www.dabs.com/ for £30 ex VAT - compares well to Eircom's €145 USB
modem. The card is also reported to work well with BT's service in the
UK under Linux. BT use the more sensible PPPoA rather than PPPoE.
The card driver interfaces with br2684 module support in the kernel
(2.4.19 is a Good Idea here) to give you an ethernet bridge which
appears as an ifconfig'able device nas0. On top of that I'm using kernel
PPPoE (requires a combination of a recent kernel, ppp package from CVS,
and a recent rp-pppoe package for a pppd plugin). The driver CD contains
the Linux driver and a handful of support packages. It's designed for
Redhat 7.1, but the only important bit is the card driver and that's
supplied as a .tar.gz as well as a .src.rpm.
My initial problems were in a machine running Red Hat (null) beta
distribution. I suspect it was somehow down to gcc 3.2 - not necessarily
because of it miscompiling the code, or the code being wrong. A large
chunk of the driver is supplied as a precompiled object module -
compiled with gcc 2.x - while the other half is provided as source code.
I think it may have interfered with the SAR. Symptoms were occasional
ATM cell errors causing dropped PPPoE packets. The larger the packet,
the more likely it was to get damaged - up to 40% loss with standard
1500 byte PPPoE packets. TCP hates those kind of odds, and I was getting
barely more than 20kBytes/sec.
The card is now installed in a Red Hat 7.3 dual PPro machine. No packet
loss. 105kBytes/sec grabbing a kernel from ftp.esat.net. Standard ICMP
pings are solidly 60ms to the far end of ppp0, and ten more to get to
This archive was generated by hypermail 2.1.6 : Thu 06 Feb 2003 - 13:18:44 GMT