HG2F09 mystery USB-to-ethernet adapter fix

2013-03-03 Thread Chuck Guzis
Lately, a bunch of cheap Chinese USB-to-Ethernet dongles have been 
making their appearance in various parts of the world by a Chinese 
vendor.  Often these can be gotten for around USD$2 or less. They're 
frequently referred to as HG2f09 adapters.

The VID/PID is 066b/20f9, which would lead one to think that the maker 
is Linksys--except the PID doesn't appear in any Linksys registry.   So 
we've got a counterfeit.  (Why pay good money to IF-USB when you can 
just borrow a VID?  I've seen the same thing in cheap USB flash drives. )

A bit of probing shows the operative device is the Asix AX88772B (and 
has been verified by others).

The pragmatic approach would be to reprogram the serial configuration 
EEPROM in this thing to match something better known, say, the Linksys 
USB200, but Asix Taiwan is not forthcoming with their programming 
utility, citing confidential, restricted to verified customers.I'm 
not inclined to tear my hair out trying to write a utility for a $2 part 
(there are some pretty good hints in the Asix datasheet).

Mine arrived with a mini-CD containing Windows drivers (uncertified, of 
course) and Linux source (no good for OpenBSD).

At any rate, a stopgap solution for me was to simply add the following 
line to the 5.2 USB if_axe.c axe_devs[] structure:

{ { USB_VENDOR_LINKSYS, 0x20f9}, AX772 | AX772B},  // Fake Linksys 

It's an ugly hack, but it seems to work just fine.  I have a bit of a 
dilemma tagging the code as if this really were a Linksys-badged 
device.  I'll leave the symbolic definitions to those official 
custodians of OpenBSD source for a cleaner version, should the need for 
this arise.

You can see the extent of the problem, just by searching the web for 

Perhaps there should be some sort of USB VID/PID aliasing capability 
to avoid having to rebuild the kernel for this sort of thing.

Submitted for whatever it's worth...

Chuck Guzis
Sydex, Inc.

Re: goodbye to some isa devices

2013-03-27 Thread Chuck Guzis

On 03/27/2013 01:01 PM, Creamy wrote:
Or, more realistically, perhaps you could just choose to maintain the 
-patch branch of a particular version that was of interest to you. For 
example, if we stopped supporting 486 in 6.0, by way of example, what 
is to stop you taking 6.0 and maintaining a -patch branch of it for 
ever more, backporting any new security and other important patches? 
Frankly, that would probably benefit the community much more than 
trying to keep the main distribution working on ancient kit forever 
more. Please don't put too much weight on a comment which was said 
quite casually as a small part of a much wider discussion. 

That's probably the best approach--as long as basic things such as 
networking protocols don't change too much, I can deal with the 
build-from-source-branch issue.

You can sort of see this business of deprecation creeping in, even 
though no broad consensus seems to be behind it.  For example, the 
current Linux X86 kernel apparently does not support some VIA IDE 
controllers (IIRC, VIA 8237?), so my Via Esther thin clients won't boot 
using it (OpenBSD runs fine, however).  So my hat's off to the community 
for keeping what it does keep.