[rt2x00-users] [RFC/RFT 3/5] rt2x00: rt2800usb: rework txstatus code

[Stanislaw Gruszka]

On Mon, Mar 12, 2012 at 05:01:34PM +0100, Ronald Wahl wrote:
> On 12.03.2012 16:36, Stanislaw Gruszka wrote:
> >On Mon, Mar 12, 2012 at 03:16:20PM +0100, Ronald Wahl wrote:
> >>I'm just testing this patchset on my AT91 embedded system (USB full
> >>speed only) with RT3572 and got a TX lockup. I got a lot of "TX
> >>status report missed" messages and then suddenly
> >>
> >>   phy1 ->  rt2800usb_tx_sta_fifo_read_completed: Warning - TX status
> >>   read failed -62
> >>
> >>After that the TX path was dead - RX was still working. Cannot say
> >>yet how often this happens - I saw it just once and it did not
> >>happen again yet. I'm not sure if this was triggered by this patch
> >>but since you changed something here it is possible.
> >
> >There is some problem with usb communication on 3572, I don't know how
> >to fix it, I currently try to investigate it. Did you see only one
> >"TX status read failed" message or there was more messages like that,
> >or other USB failures?
> 
> I saw this message only once. The only different messages I saw were
> the "status report missed" messages. Interestingly there are periods
> where no such messages appear at all and other ones where these
> messages are generated very often.

Hmm, then I do not understand how stall happen. After error in
urb_status we jump to stop_reading, and then if
rt2800usb_txstatus_pending() is true, we reread TX_STA_FIFO register
again.

> I had similar TX stalls even without your patchset using the driver
> in the vanilla 3.2.5 kernel before but at this time the debugging
> was off so I'm not sure if this was related.
> 
> I ordered a bunch (~15) of different Ralink sticks - some hopefully
> with 3572 which I'm going to put in our test lab for endurance
> testing. Maybe this increases the probability of triggering it
> again.
Cool, thanks for testing.

Stanislaw