Bob, to identify the problem I had a printk in ath5k_hw_register_timeout printing the remaining loops. Normally it took only very few loops for the register to change to the target value (loops remaining in the range of 20000..19980). To me this indicates that something went seriously wrong and just doubling the time to 40000 loops was pointless (if the normal range was about 20 loops and it doesn't finish in 20000 it won't finish in 40000). That's why I didn't even try this.
However, I made the following observations: 1) The situation is recoverable without drastic measures: just restarting hostapd and reconnecting the client is enough to get things going again (this, BTW, is were my assumption about an open reference came from). 2) Sometimes I saw a successful reset after a failed one, however there was no traffic. >From these observations I came to the ath5k_reset_wake function. It gets regularly called from the calibration tasklet and on error it leaves the queues locked down. So instead of considering a failed reset as fatal I introduced a retry with spectacular success. The code is now running continuously for over 24 hours and I intend to let it over the weekend. -- Regards Joerg --- Bob Copeland <m...@bobcopeland.com> schrieb am Fr, 19.3.2010: > Von: Bob Copeland <m...@bobcopeland.com> > Betreff: Re: [ath5k-devel] ATH5K in AP mode and "ath5k phy0: failed to warm > reset the MAC Chip" errors > An: "Joerg Pommnitz" <pommn...@yahoo.com> > CC: ath5k-devel@lists.ath5k.org > Datum: Freitag, 19. März, 2010 12:58 Uhr > On Fri, Mar 19, 2010 at 01:58:22AM > -0700, Joerg Pommnitz wrote: > > Bob, > > I have cooked up a workaround myself. With the > following patch applied > > I have been running iperf with good performance for > over 20 hours now > > without interruption). > > > > And yes, the debug log was triggered quite often: > > So, I guess from that it took longer than 10 ms to wake the > chip up. > You might also see if increasing the loop counter in > ath5k_hw_set_power() > has the same effect (and for what values). > > If you get a chance I would like to know if my workarounds > also work. > I'm trying to figure out the underlying problem with this > and it would > be a nice clue. Not all hardware can replicate these > problems. > > -- > Bob Copeland %% www.bobcopeland.com > > __________________________________________________ Do You Yahoo!? Sie sind Spam leid? Yahoo! Mail verfügt über einen herausragenden Schutz gegen Massenmails. http://mail.yahoo.com _______________________________________________ ath5k-devel mailing list ath5k-devel@lists.ath5k.org https://lists.ath5k.org/mailman/listinfo/ath5k-devel