[Prism54-devel] USB & al

[Jean-Baptiste Note]

Hello Sebastien,

Sebastien B <sebastien.b at swissinfo.org> said :
> Running the attached source code produces the following output :
> $ ./softmac 3886lmac_2.7.0.0.arm 
> Calling prism_softmac_parse_bra() at offset 0
> Calling prism_softmac_parse_bra() at offset 1
> Bus error (core dumped)
> So you should not call prism_softmac_parse_bra() repeatedly with differents 
> offsets until it works : some bogus data can crash it, and in the kernel this 
> is very bad.
> Experimentations show that the function doesn't crash because it is called 
> twice, but really because of the data it finds at offset 1.

That's okay, i start at the right offset anyways. But look at the
testcase in the arch repo, which is arguably the same, and it doesn't
segfault here. Ah, the joys of close-source tinkering.

>> So if anyone with a last-gen usb device could try this...
>
> I'll have a WG111v2 next week (thanks a lot to DG4OAQ !!). I've started a 
> driver which uses the UMAC library in usermode for easy experimentation. I'm 
> waiting for the device for testing/debugging :)
> If there is anything someone on the list wants me to try, it's OK (including 
> tinkering with the device's electronics).

Could you try the driver in the arch repo ?

>
>> I have doubts the usb version is okay (there are discrepancies between the
>> expected behaviour of the driver and what's observed in logs of the
>> devices). 
>
> What kind of discrepancies ?
> That may be because of the 2.7.0 firmware being different from the 2.5.8 we've 
> seen so far.

The descrepencies that lead to the need to write the compat layer, or
the discrepencies in version number (the code seems to refer to 3887,
the lib is for 3886, let's pray they're compatible :). Obviously the usb
code in the tarball was never compiled as-is, never tested, at least
with the softmac interface and library and firmware that's in the
original tarball.

> Not submitting the URB (and returning success) when you spot a LED control 
> packet isn't fine ?

It'll take me 1 hour at least to write, which i don't feel like spending
on this ATM. You'd also want to suppress the *previous* announcing
packet. If the above test does not convince you, i'd be happy if you
could do it (honestly, i wouldn't care, and it seems you didn't either
last week ;) ).

>>   By the way, regarding ndiswrapper, if any of its fine developers are
>> listening : all versions above the first ones -- modprobe indicates
>> 0.13pre1 here -- do not work with my old WUSB54G driver.
>
> DG4OAQ also reported problems with his D-Link-G122 and ndiswrapper > 0.12.

OK, thanks for the info

Regards,
JB

-- 
Jean-Baptiste Note
+33 (0)6 83 03 42 38
jean-baptiste.note at wanadoo.fr