K 10
svn:author
V 3
sam
K 8
svn:date
V 27
2009-01-23T07:00:33.332776Z
K 7
svn:log
V 1954
Checkpoint gutting of HAL_CHANNEL:
o replace uses of HAL_CHANNEL by struct ieee80211_channel
o remove HAL_CHANNEL shadow state from HAL_CHANNEL_PRIVATE; instead users
  are given an ieee80211_channel and if they need private state they use
  ic_devdata to find the private channel (shrinks private chans by 12 bytes)
o pack the private channel array (one entry per frequency)
o mark 802.11 channels directly when interference is detected
o use ic_devdata to index into ani state cache; note this consolidates
  state for enumerated channels so can change behaviour (e.g. for b/g
  and turbo combinations)
o remove private mhz2->ieee conversion routines as we now get the ieee
  channel number directly from ic_ieee
o setup 11g channels with correct channel flags (used to mark them pureg
  and then futz the flags when constructing net80211 channels)
o remove conversion code in regulatory routines now that we don't do any
  conversion to/from private/internal data structures
o de-inline ath_hal_getantennareduction

Note: HAL_CHANNEL_PRIVATE temporarily shadows the frequency to validate
   lookups and facilitate noise floor processing; this will go away soon
Note: ath_hal_computetxtime needs take a channel to eliminate use of
   ah_curchan (possibly other routines too)
Note: channel interference handling should be moved to a callback so we
   can const'ify parameters and eliminate assumptions about when/how
   detection is done
Note: private channel table can shrink and ani state table should be the
   same size to avoid checking references
Note: ath_hal_checkchannel should be inlined when we stop validating lookups
Note: privFlags can go away if we move CHANNEL_IQVALID elsewhere but might
   be worth keeping as removing it won't shrink HAL_CHANNEL_PRIVATE
Note: purge direct use of ic_flags (e.g. in switch statements)

Tested in sta mode on 5210, 5211, 5416, and a wide variety of 5212 cards
in normal modes; turbo modes need testing.

END