[Prism54-devel] GW3887 STALL endpoint problem

[Jean-Baptiste Note]

Hello Sebastien, dear list,

Sebastien B <sebastien.b at swissinfo.org> said :
>> My WG111v2 stalls endpoint 0x81 right after the MGMT readback. The last
>> MGMT transfer is OK, the STALL concerns the URB which comes right after.
> This seems to be a FreeBSD USB bug.

Glad to hear this !

> With Hans Petter Selasky's alternative USB driver, it works. But I still don't 
> know why the Linux driver crashes...
> Monitor mode seems OK (Adhoc beacons are the only kind of 802.11 traffic I can 
> generate in the area where I live, so this is not a good test...). By the 
> way, JB, why do you set chan_m1 to P54U_TX_CONTROL_CHANNEL_MAGIC1_SCAN and	
> chan_m2 to P54U_TX_CONTROL_CHANNEL_MAGIC2_SCAN when entering monitor mode, 
> instead of leaving them to P54U_TX_CONTROL_CHANNEL_MAGIC1_TX and 
> P54U_TX_CONTROL_CHANNEL_MAGIC2_TX like previous driver releases did ? 

Good question, i'm glad you asked ! well, no reason, really. Didn't even
know something had changed. Thanks for the report =]

> Doing so causes the WG111 to send back oversized bogus data frames (I
> haven't tried with the Gigaset nor attempted to dump those frames for
> now, I'd like to know why you changed this first...)

The oversized bogus frames are a sign that something went askew with the
device (same as interrupt in version1 devices). They're always generated
when you don't speak right to it, and i think they may be a memory dump.

By the way, we should do something about them, at least recognize the
error condition. Do you find any way of recognizing the first one of
those frames for instance, so that we can generate a "device is
completely out to lunch" event (maybe triggering reinitialization,
etc...) ? I didn't think of doing it earlier, but it would have saved me
a great deal of disk space :)

> Another small thing, the LED control packets' meaning is reversed on GW3887s. 
> When you want to turn the LEDs off, you need to tell the device to turn them 
> on, and vice versa.

:( Thanks. Adding to "TOFIX".

I'll go back to "consolidate the code and protocol" as soon as the pci
integration is finished. For now piling up the patches :/

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