I'm looking to see if there is any interest in adding AR5513 support
to the ath5k HAL. I have one of the devices myself and would love to
add support to the HAL for this device(It's the only missing one, no?)

I did some preliminary mmio traces on the binary ndiswrappered
drivers, and it looks like the device is nothing more than a glorified
AR5212 with two radios, and the eeprom functions changed. (EEPROM
functions like the one in ath9k, with 0x10000 marking EEPROM busy,
shared status/command register)
I've done some driver development before, but nothing for the 2.6
series kernel, hopefully I can catch myself up on how everything
works. I also hope that by running mmio traces I have not nulled my
legal right to write code, hopefully somebody with more experience on
the subject can fill me in.

I did some quick patching of the driver and dropped in a new EEPROM
function to make it work, a few IDs to identify the card, and the
existing code recognizes the card properly beyond that. However, I
have not been able to get the LEDs to function, or the card to
properly transmit or receive packets. So it looks like I still have
some work to do towards writing support myself.

It appears that all of the tx queues and rx queues behave
properly(fill and drain like normal), but nothing is getting to the
mac/network layer. Which leads me to believe that the buffers may be
wrapped in some protocol(The AR5005VL is designed for embedded devices
such as DVR boxes and streaming media. It has a full MIPS32 R3000(I
think?) cpu and likely its own firmware/operating system)

So in adding support, the following things will likely have to be
done: change the data type of registers from u16 to u32(the device
makes use of a few registers that require it.), fill in
missing/incorrect eeprom reading functions, correct eeprom reading
functions for eeprom version 0x4009(the cards A band data doesn't
appear to be properly detected), add the eeprom shim to read eeprom
for the specific card, work out why TX/RX isn't working properly,
anything else involved with why TX/RX doesn't work.

Obviously I'm going to have to do a lot of that myself, as I don't
know of anyone else who has one of the cards, but in the event that I
am legally unable to produce patches or code, I'm fully willing to
document my findings to allow somebody else to write the required
code, in the event there is interest in this.
_______________________________________________
ath5k-devel mailing list
ath5k-devel@lists.ath5k.org
https://lists.ath5k.org/mailman/listinfo/ath5k-devel

Reply via email to