p54: fix rssi auto calibration

Ever wondered why the signal was so bad with p54 compared to madwifi, or intel?

Well, if you have revision 1 rssi calibration curve points in your EEPROM, then wonder no more.
The firmware wants a extra 1 byte padding for every curve point. But someone forgot to put
them into the EEPROM's data structure...

So now, big question: what happens when we blindly "memcpy" these data points?

Signed-off-by: Christian Lamparter <chunkeey@web.de>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

1 files changed, 19 insertions, 4 deletions

diff --git a/drivers/net/wireless/p54/p54common.h b/drivers/net/wireless/p54/p54common.h
index 9bc2a1cf4b57..a79c1a146917 100644
--- a/drivers/net/wireless/p54/p54common.h
+++ b/drivers/net/wireless/p54/p54common.h
@@ -89,6 +89,16 @@ struct pda_pa_curve_data_sample_rev1 {
         u8 data_qpsk;
         u8 data_16qam;
         u8 data_64qam;
+} __attribute__ ((packed));
+
+struct p54_pa_curve_data_sample {
+        u8 rf_power;
+        u8 pa_detector;
+        u8 data_barker;
+        u8 data_bpsk;
+        u8 data_qpsk;
+        u8 data_16qam;
+        u8 data_64qam;
         u8 padding;
 } __attribute__ ((packed));
 
@@ -212,8 +222,8 @@ struct p54_tx_control_filter {
 } __attribute__ ((packed));
 
 struct p54_tx_control_channel {
-        __le16 magic1;
-        __le16 magic2;
+        __le16 flags;
+        __le16 dwell;
         u8 padding1[20];
         struct pda_iq_autocal_entry iq_autocal;
         u8 pa_points_per_curve;
@@ -222,8 +232,13 @@ struct p54_tx_control_channel {
         u8 val_qpsk;
         u8 val_16qam;
         u8 val_64qam;
-        struct pda_pa_curve_data_sample_rev1 curve_data[0];
-        /* additional padding/data after curve_data */
+        struct pda_pa_curve_data_sample_rev1 curve_data[8];
+        u8 dup_bpsk;
+        u8 dup_qpsk;
+        u8 dup_16qam;
+        u8 dup_64qam;
+        __le16 rssical_mul;
+        __le16 rssical_add;
 } __attribute__ ((packed));
 
 struct p54_tx_control_led {