On Sat, Aug 28, 2010 at 12:13 AM, Luis R. Rodriguez
<lrodriguez@xxxxxxxxxxx> wrote:
> @@ -520,23 +521,38 @@ static int ieee80211_scan_state_decision(struct ieee80211_local *local,
> /*
> * we're on the operating channel currently, let's
> * leave that channel now to scan another one
> + * unless the driver has other wait time constraints
> */
> - local->next_scan_state = SCAN_LEAVE_OPER_CHANNEL;
> + if (drv_sw_scan_wait_constraints(local) &&
> + local->scan_oper_chan_waits < 2) {
> + local->next_scan_state = SCAN_DECISION;
> + local->scan_oper_chan_waits++;
> + } else {
> + local->next_scan_state = SCAN_LEAVE_OPER_CHANNEL;
> + local->scan_oper_chan_waits = 0;
> + }
> }
>
> - *next_delay = 0;
> + if (local->next_scan_state == SCAN_DECISION)
> + *next_delay = HZ / 5;
This needs some good review as well. As its implemented here we'll
essentially stop all TX except null data frames and probe requests as
local->scanning is set by SCAN_SW_SCANNING through
__ieee80211_start_scan(). One possible alternative is to do the
drv_sw_scan_wait_constraints() check prior to calling the sw scan and
fail the __ieee80211_start_scan() call if we need to wait. I didn't
think this was the best thing to do at first though because the user
will just get a scan busy error from userspace. It seemed more
rational to queue the request in since we just have to wait until the
driver clears its wait constraints -- or we wait too long (just a loop
counter now on the operating channel).
> @@ -544,8 +560,26 @@ static void ieee80211_scan_state_leave_oper_channel(struct ieee80211_local *loca
> else
> *next_delay = HZ / 10;
>
> + /*
> + * If we're not getting ACKs for our nullfuncs we're likely in bad
> + * shape so we should not care about missed data as we can't even
> + * hear the AP. We want to help roam away if we can so just go
> + * ahead and scan. Try getting the ACK just 3 times.
> + */
> + r = drv_sw_scan_wait_constraints(local);
> + if (r == -EAGAIN) {
> + local->scan_nullfunc_retries++;
> + *next_delay = HZ / 10;
After some review I believe we need to set here
local->offchannel_ps_enabled otherwise we won't resend the nullfunc.
Also we may want to reconsider the amount of time we wait for a resend
as in that duration we will be only letting through nullfuncs and
probe requests.
Luis
--
To unsubscribe from this list: send the line "unsubscribe linux-wireless" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
