net80211: do not cancel scan when packet is sent

[avos]

Summary

Restore pre-r300383 behavior when a frame is sent:

• stop scan;
• send frame;
• when beacon arrives and our bit in TIM is not set - restart the scan.

NOTE 1:
Ideally, this should introduce new interface (ieee80211_pause_anyscan()); however, since
ieee80211_cancel_anyscan() is not used by drivers and only called by ieee80211_start_pkt()
the current patch overrides it's behavior instead.

NOTE 2:
In case, if AP will forget to reset TIM (see r295932 as an example), the scan will hang forever.

Test Plan

Tested with Intel 3945BG, STA mode, w/ data transmission:

• without patch: scan results may be truncated;
• with patch:
  • TIM is clear: net80211 will wait, until data will be transmitted (not nice when mplayer(1) is running);
  • our bit in TIM is set: 'ifconfig wlan0 scan' hangs (interruptible).

[adrian]

Hm, don't we have a general scan timeout that should prevent a scan from hanging forever?

[avos]

In D7979#165342, @adrian wrote:

Hm, don't we have a general scan timeout that should prevent a scan from hanging forever?

Well,

1. (answers your question) no, there is no 'global' scan watchdog (only few local watchodgs in wpi(4)/iwn(4)/iwm(4); however, we don't need them for software-driven scan, because it cannot hang).
2. (mine statement) 'hang' means:
   • scan request is sent; ifconfig waits for 'scan complete event';
   • outgoing traffic cancels scan; it stops (but preserves IEEE80211_FEXT_BGSCAN flag); no event is sent (with this patch; currently the event is sent anyway, but scan results may be incomplete);
   • traffic was sent; ic_lastdata settles;
3. sta_recv_mgmt() checks via contbgscan() if the scan may be restarted; if everything is okay, new scan starts;
4. scan ends; 'scan complete' event is sent (and ifconfig displays scan results).