issues with the recent ath commits (r227740, r227651, ...)

[Adrian Chadd]

On 22 November 2011 06:53, Alexander Best <arundel at freebsd.org> wrote:

> ok. after > 45 minutes uptime i can confirm that the patch fixes the issues i
> experienced beforehand. unless i didn't get a boot-up-panic, my system locked
> up after < 5 minutes.

Sweet, I'll commit that to -HEAD now.

> however using your patch i noticed the following console output:
>
> ath0: ath_start: sc_inreset_cnt > 0; bailing
>
> ... only once, however. i don't believe i've seen this one ever beforehand.

That's intentional. That means that a frame was queued for TX whilst
the hardware was busy being reset.
It may be because I'm now processing the frames on an interface reset,
rather than just deleting them.
But on SMP machines it's quite plausible that a reset on one thread
would run in parallel with if_start on another thread, causing these
to interfere.

My eventual aim is to have TX/RX properly suspended before a reset
occurs, so these messages will go away (and their appearance will only
be when there's a real problem.) But that's going to take time.

Thanks for getting back to me about this issue!


Adrian