ath9k: AR9380 PAPRD changes to improve tx performance

ath9k: AR9380 PAPRD changes to improve tx performance

ath9k_hw: fix PA predistortion HT40 mask

The commit 'ath9k_hw: Disable PAPRD for rates with low Tx power' changed
the code that sets the PAPRD rate masks to use only either the HT20 mask
or the HT40 mask. This is wrong, as the hardware can still use HT20 rates
even when configured for HT40, and the operating channel mode does not
affect PAPRD operation.
The register for the HT40 rate mask is applied as a mask on top of the
other registers to selectively disable PAPRD for specific rates on HT40
packets only.
This patch changes the code back to the old behavior which matches the
intended use of these registers. While with current cards this should not
make any practical difference (according to Atheros, the HT20 and HT40
mask should always be equal), it is more correct that way, and maybe
the HT40 mask will be used for some rare corner cases in the future.

Cc: Vasanthakumar Thiagarajan <vasanth@atheros.com>;
Signed-off-by: Felix Fietkau <nbd@openwrt.org>;

ath9k_hw: Configure appropriate Tx power when PAPRD fails

Target Tx power available in eeprom is for PAPRD. If PAPRD
fails, paprd scale factor needs to be detected from this
target tx power.

Signed-off-by: Vasanthakumar Thiagarajan <vasanth@atheros.com>;

ath9k_hw: Disable PAPRD for rates with low Tx power

When the drop in Tx power for a particular mcs rate exceeds
the paprd scale factor, paprd may not work properly. Disable
paprd for any such rates.

Signed-off-by: Vasanthakumar Thiagarajan <vasanth@atheros.com>;

ath9k_hw: Add a helper to get paprd scale factor

Signed-off-by: Vasanthakumar Thiagarajan <vasanth@atheros.com>;

ath9k_hw: Move get_streams() to hw.h

This helper can be used in multiple places. Also make
it inline returning u8.

Signed-off-by: Vasanthakumar Thiagarajan <vasanth@atheros.com>;

ath9k_hw: Remove unnecessary Rx IQ cal register configuration in ar9003_hw_tx_iq_cal()

Signed-off-by: Vasanthakumar Thiagarajan <vasanth@atheros.com>;

ath9k_hw: Remove delay during regwrite of analog shift registers

This is not needed for AR9003.

Signed-off-by: Vasanthakumar Thiagarajan <vasanth@atheros.com>;

ath9k_hw: fix PA predistortion training power selection

The EEPROM contains scale factors for the tx power, which define
the range of allowable difference between target power and training
power. If the difference is too big, PA predistortion cannot be used.
For 2.4 GHz there is only one scale factor, for 5 GHz there are
three, depending on the specific frequency range.

Signed-off-by: Felix Fietkau <nbd@openwrt.org>;

ath9k_hw: fix the PA predistortion rate mask

The EEPROM PAPRD rate mask fields only contain mask values for actual
rates in the low 25 bits. The upper bits are reserved for tx power
scale values. Add the proper mask definitions and use them before
writing the values to the register.

Signed-off-by: Felix Fietkau <nbd@openwrt.org>;

ath9k: fix PA predistortion thermal measurement handling

To be able to measure the thermal values correctly for PAPRD, we need
to send training frames before setting up the gain table for the measurement,
and then again afterwards for the actual training.

For further improvement, send training frames at MCS0 instead of 54 MBit/s
legacy. That way we can use the No-ACK flag for the transmission, which
speeds up PAPRD training in general, as the hardware won't have to
retransmit and wait for ACK timeout (was previously set to 4 * 6
transmission attempts).

Signed-off-by: Felix Fietkau <nbd@openwrt.org>;

ath9k: remove bfs_paprd_timestamp from struct ath_buf_state

Change-Id: I3958195e689735f6f7e70381b91869376309b083
Signed-off-by: Felix Fietkau <nbd@openwrt.org>;
Signed-off-by: John W. Linville <linville@tuxdriver.com>;

BUG=chromium-os:10211
TEST=matfunc+secmat (no regressions)

Committed: http://chrome-svn/viewvc/chromeos?view=rev&revision=b9c519b

[Sam Leffler, 2011-01-05 22:54:06]

This replaces 6023005.  It includes a change to ar9003_paprd_enable to fix the
set_txpower call to match the calling convention in the current code base.

Test so far show no change in performance.  Preparing for commit so we can track
upstream 9382 support.

[Paul Stewart, 2011-01-06 00:30:41]

LGTM