Help with testing the AP+Client Bug

[Eric]

Now that IS interesting.... look at the image you posted! The MAC address changed! That's why it thinks it's a different vendor.

It's good to know I'm not crazy... I had it right the first time. v15465 is killing it.

Now, to figure out which patch is causing the problems... (there are a bunch of patches that changed in 15465). The different MAC is good clue though.

ispyisail: I'm a bit puzzled by the report that 13794 is causing problems on reboot but 13761 is not -- the only change made there is in one of the configuration scripts and it should have absolutely no effect. Are you sure this fail-on-reboot happens reliably between in 13794 but not 13761?

[ispyisail]

ispyisail: I'm a bit puzzled by the report that 13794 is causing problems on reboot but 13761 is not -- the only change made there is in one of the configuration scripts and it should have absolutely no effect. Are you sure this fail-on-reboot happens reliably between in 13794 but not 13761?

I've done more tests

The problem with 13794 is repeatable with the meraki mini but the fault does not show up on my DIR-300.

Looks like it might be a meraki mini specific fault?

[DoesItMatter]

The Fonera 2200, 2201, and 2201+ are all similar as far as functionality goes, and they are also similar to the DIR-300 and the Airlink AR-430W. I suspect the Meraki's are different, especially being that they need a different flash build than the Fons/Airlinks/Dir-300's

Might be fighting multiple issues.

Does the 15464 work ok for you IspyIsail?

[ispyisail]

Does the 15464 work ok for you IspyIsail?

I got my MR3201a out for this test

I found some strange results

Will need to do more tests

[ispyisail]

I can confirm that 15464 works well om the MR3201a

[DoesItMatter]

Nice!

So it does look like 15465 is the problem build.
Or at least the build that causes the most problems and
breaks the AP+Client mode.

I'm currently running 1.0.4 and it's running fine for me,
but I don't normally use the AP+Client mode myself.

If a new test build pops up, I'll be ready to flash it to my Fonera!

[Eric]

Someone named Marek Lindner just submitted a patch to the OpenWrt developers mailing list that may fix the AP+Client bug that's been causing so many problems.

There's one part in particular that I think is important:

Code: Select all

-               if (sc->sc_beacon_cal && (jiffies > sc->sc_lastcal + (ath_calinterval * HZ))) {
+               if (!sc->sc_beacon_cal && (jiffies > sc->sc_lastcal + (ath_calinterval * HZ))) {
                        sc->sc_cal_ch.expires = jiffies + msecs_to_jiffies(10);
                        add_timer(&sc->sc_cal_ch);
                }

I find this code particularly interesting because between r15464 and r15465, the add_timer line was added. So I bet before r15465, the timer causing the race condition was never called and there wasn't a problem. After r15465 the timer gets called, and hence the problems that start then. Then, this patch makes sure the code only gets called at the appropriate time (since the original if statement is wrong), solving the problem in the appropriate way.

This is conjecture of course.. I still need to try it. But it looks like we may have a solution!

[Eric]

The potential solution mentioned in the previous post didn't work, but I believe I have found one that does. If some of you could help by testing the 1.0.14 firmware, it would be greatly appreciated.

[ispyisail]

Tested with MR3201a

Works well

[ispyisail]

Still working and I must say really well