linux/drivers/net/wireless/ath/ath9k/ar9003_calib.c

1126 lines
30 KiB
C
Raw Normal View History

/*
* Copyright (c) 2010-2011 Atheros Communications Inc.
*
* Permission to use, copy, modify, and/or distribute this software for any
* purpose with or without fee is hereby granted, provided that the above
* copyright notice and this permission notice appear in all copies.
*
* THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
* WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
* MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
* ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
* ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
* OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
*/
#include "hw.h"
#include "hw-ops.h"
#include "ar9003_phy.h"
#include "ar9003_rtt.h"
#include "ar9003_mci.h"
#define MAX_MEASUREMENT MAX_IQCAL_MEASUREMENT
ath9k_hw: Fix Tx IQ Calibration hang issue in AR9003 chips On AR9003 chips, doing three IQ calibrations will possibly cause chip in stuck state. In noisy environment, chip could receive a packet during the middle of three calibrations and it causes the conflict of HW access and the eventual failure. It also causes IQ calibration outliers which results in poor Tx EVM. The IQ Cal procedure is after resetting the chip, run IQ cal 3 times per each cal cycle and find the two closest readings and average of two. The advantage of running Tx IQ cal more than once is that we can compare calibration results for the same gain setting over multiple iterations. Most of the cases the IQ failures were observed after first pass. For the AR9485 and later chips, Tx IQ Calibration is performed along with AGC cal. But for pre-AR9485 chips, Tx IQ cal HW has to be separated from the rest of calibration HW to avoid chip hang. After all calibrations are done in HW, we can start SW post-processing. By doing this way, we minimize the SW difference among all chips. The order of calibration (run IQ cal before other calibration) is also needed to avoid chip hang for chips before AR9485. This issue was originally observed with AR9382. During the issue kernel log was filled with following message ath: timeout (100000 us) on reg 0xa640: 0x00000001 & 0x00000001 != 0x00000000 ath: timeout (100000 us) on reg 0xa2c4: 0x00158dd9 & 0x00000001 != 0x00000000 ath: Unable to reset channel (2412 MHz), reset status -5 ath: Unable to set channel Signed-off-by: Rajkumar Manoharan <rmanoharan@atheros.com> Signed-off-by: John W. Linville <linville@tuxdriver.com>
2011-04-25 00:04:39 +08:00
#define MAX_MAG_DELTA 11
#define MAX_PHS_DELTA 10
struct coeff {
ath9k_hw: Fix Tx IQ Calibration hang issue in AR9003 chips On AR9003 chips, doing three IQ calibrations will possibly cause chip in stuck state. In noisy environment, chip could receive a packet during the middle of three calibrations and it causes the conflict of HW access and the eventual failure. It also causes IQ calibration outliers which results in poor Tx EVM. The IQ Cal procedure is after resetting the chip, run IQ cal 3 times per each cal cycle and find the two closest readings and average of two. The advantage of running Tx IQ cal more than once is that we can compare calibration results for the same gain setting over multiple iterations. Most of the cases the IQ failures were observed after first pass. For the AR9485 and later chips, Tx IQ Calibration is performed along with AGC cal. But for pre-AR9485 chips, Tx IQ cal HW has to be separated from the rest of calibration HW to avoid chip hang. After all calibrations are done in HW, we can start SW post-processing. By doing this way, we minimize the SW difference among all chips. The order of calibration (run IQ cal before other calibration) is also needed to avoid chip hang for chips before AR9485. This issue was originally observed with AR9382. During the issue kernel log was filled with following message ath: timeout (100000 us) on reg 0xa640: 0x00000001 & 0x00000001 != 0x00000000 ath: timeout (100000 us) on reg 0xa2c4: 0x00158dd9 & 0x00000001 != 0x00000000 ath: Unable to reset channel (2412 MHz), reset status -5 ath: Unable to set channel Signed-off-by: Rajkumar Manoharan <rmanoharan@atheros.com> Signed-off-by: John W. Linville <linville@tuxdriver.com>
2011-04-25 00:04:39 +08:00
int mag_coeff[AR9300_MAX_CHAINS][MAX_MEASUREMENT];
int phs_coeff[AR9300_MAX_CHAINS][MAX_MEASUREMENT];
int iqc_coeff[2];
};
enum ar9003_cal_types {
IQ_MISMATCH_CAL = BIT(0),
TEMP_COMP_CAL = BIT(1),
};
static void ar9003_hw_setup_calibration(struct ath_hw *ah,
struct ath9k_cal_list *currCal)
{
struct ath_common *common = ath9k_hw_common(ah);
/* Select calibration to run */
switch (currCal->calData->calType) {
case IQ_MISMATCH_CAL:
/*
* Start calibration with
* 2^(INIT_IQCAL_LOG_COUNT_MAX+1) samples
*/
REG_RMW_FIELD(ah, AR_PHY_TIMING4,
AR_PHY_TIMING4_IQCAL_LOG_COUNT_MAX,
currCal->calData->calCountMax);
REG_WRITE(ah, AR_PHY_CALMODE, AR_PHY_CALMODE_IQ);
ath_dbg(common, CALIBRATE,
"starting IQ Mismatch Calibration\n");
/* Kick-off cal */
REG_SET_BIT(ah, AR_PHY_TIMING4, AR_PHY_TIMING4_DO_CAL);
break;
case TEMP_COMP_CAL:
REG_RMW_FIELD(ah, AR_PHY_65NM_CH0_THERM,
AR_PHY_65NM_CH0_THERM_LOCAL, 1);
REG_RMW_FIELD(ah, AR_PHY_65NM_CH0_THERM,
AR_PHY_65NM_CH0_THERM_START, 1);
ath_dbg(common, CALIBRATE,
"starting Temperature Compensation Calibration\n");
break;
}
}
/*
* Generic calibration routine.
* Recalibrate the lower PHY chips to account for temperature/environment
* changes.
*/
static bool ar9003_hw_per_calibration(struct ath_hw *ah,
struct ath9k_channel *ichan,
u8 rxchainmask,
struct ath9k_cal_list *currCal)
{
struct ath9k_hw_cal_data *caldata = ah->caldata;
/* Cal is assumed not done until explicitly set below */
bool iscaldone = false;
/* Calibration in progress. */
if (currCal->calState == CAL_RUNNING) {
/* Check to see if it has finished. */
if (!(REG_READ(ah, AR_PHY_TIMING4) & AR_PHY_TIMING4_DO_CAL)) {
/*
* Accumulate cal measures for active chains
*/
currCal->calData->calCollect(ah);
ah->cal_samples++;
if (ah->cal_samples >=
currCal->calData->calNumSamples) {
unsigned int i, numChains = 0;
for (i = 0; i < AR9300_MAX_CHAINS; i++) {
if (rxchainmask & (1 << i))
numChains++;
}
/*
* Process accumulated data
*/
currCal->calData->calPostProc(ah, numChains);
/* Calibration has finished. */
caldata->CalValid |= currCal->calData->calType;
currCal->calState = CAL_DONE;
iscaldone = true;
} else {
/*
* Set-up collection of another sub-sample until we
* get desired number
*/
ar9003_hw_setup_calibration(ah, currCal);
}
}
} else if (!(caldata->CalValid & currCal->calData->calType)) {
/* If current cal is marked invalid in channel, kick it off */
ath9k_hw_reset_calibration(ah, currCal);
}
return iscaldone;
}
static bool ar9003_hw_calibrate(struct ath_hw *ah,
struct ath9k_channel *chan,
u8 rxchainmask,
bool longcal)
{
bool iscaldone = true;
struct ath9k_cal_list *currCal = ah->cal_list_curr;
/*
* For given calibration:
* 1. Call generic cal routine
* 2. When this cal is done (isCalDone) if we have more cals waiting
* (eg after reset), mask this to upper layers by not propagating
* isCalDone if it is set to TRUE.
* Instead, change isCalDone to FALSE and setup the waiting cal(s)
* to be run.
*/
if (currCal &&
(currCal->calState == CAL_RUNNING ||
currCal->calState == CAL_WAITING)) {
iscaldone = ar9003_hw_per_calibration(ah, chan,
rxchainmask, currCal);
if (iscaldone) {
ah->cal_list_curr = currCal = currCal->calNext;
if (currCal->calState == CAL_WAITING) {
iscaldone = false;
ath9k_hw_reset_calibration(ah, currCal);
}
}
}
/* Do NF cal only at longer intervals */
if (longcal) {
/*
* Get the value from the previous NF cal and update
* history buffer.
*/
ath9k_hw_getnf(ah, chan);
/*
* Load the NF from history buffer of the current channel.
* NF is slow time-variant, so it is OK to use a historical
* value.
*/
ath9k_hw_loadnf(ah, ah->curchan);
/* start NF calibration, without updating BB NF register */
ath9k_hw_start_nfcal(ah, false);
}
return iscaldone;
}
static void ar9003_hw_iqcal_collect(struct ath_hw *ah)
{
int i;
/* Accumulate IQ cal measures for active chains */
for (i = 0; i < AR5416_MAX_CHAINS; i++) {
if (ah->txchainmask & BIT(i)) {
ah->totalPowerMeasI[i] +=
REG_READ(ah, AR_PHY_CAL_MEAS_0(i));
ah->totalPowerMeasQ[i] +=
REG_READ(ah, AR_PHY_CAL_MEAS_1(i));
ah->totalIqCorrMeas[i] +=
(int32_t) REG_READ(ah, AR_PHY_CAL_MEAS_2(i));
ath_dbg(ath9k_hw_common(ah), CALIBRATE,
"%d: Chn %d pmi=0x%08x;pmq=0x%08x;iqcm=0x%08x;\n",
ah->cal_samples, i, ah->totalPowerMeasI[i],
ah->totalPowerMeasQ[i],
ah->totalIqCorrMeas[i]);
}
}
}
static void ar9003_hw_iqcalibrate(struct ath_hw *ah, u8 numChains)
{
struct ath_common *common = ath9k_hw_common(ah);
u32 powerMeasQ, powerMeasI, iqCorrMeas;
u32 qCoffDenom, iCoffDenom;
int32_t qCoff, iCoff;
int iqCorrNeg, i;
ath9k: Use static const Using static const generally increases object text and decreases data size. It also generally decreases overall object size. text data bss dec hex filename 11161 56 2136 13353 3429 drivers/net/wireless/ath/ath9k/ar9003_paprd.o.new 11167 56 2136 13359 342f drivers/net/wireless/ath/ath9k/ar9003_paprd.o.old 15428 56 3056 18540 486c drivers/net/wireless/ath/ath9k/eeprom_4k.o.old 15451 56 3056 18563 4883 drivers/net/wireless/ath/ath9k/eeprom_4k.o.new 14087 56 2560 16703 413f drivers/net/wireless/ath/ath9k/eeprom_9287.o.old 14036 56 2560 16652 410c drivers/net/wireless/ath/ath9k/eeprom_9287.o.new 10041 56 2384 12481 30c1 drivers/net/wireless/ath/ath9k/ani.o.new 10088 56 2384 12528 30f0 drivers/net/wireless/ath/ath9k/ani.o.old 9316 1580 2304 13200 3390 drivers/net/wireless/ath/ath9k/htc_drv_init.o.new 9316 1580 2304 13200 3390 drivers/net/wireless/ath/ath9k/htc_drv_init.o.old 16483 56 3432 19971 4e03 drivers/net/wireless/ath/ath9k/ar9003_phy.o.new 16517 56 3432 20005 4e25 drivers/net/wireless/ath/ath9k/ar9003_phy.o.old 18221 104 2960 21285 5325 drivers/net/wireless/ath/ath9k/rc.o.old 18203 104 2960 21267 5313 drivers/net/wireless/ath/ath9k/rc.o.new 19985 56 4288 24329 5f09 drivers/net/wireless/ath/ath9k/eeprom_def.o.new 20040 56 4288 24384 5f40 drivers/net/wireless/ath/ath9k/eeprom_def.o.old 23997 56 4984 29037 716d drivers/net/wireless/ath/ath9k/ar5008_phy.o.old 23846 56 4984 28886 70d6 drivers/net/wireless/ath/ath9k/ar5008_phy.o.new 24285 56 3184 27525 6b85 drivers/net/wireless/ath/ath9k/ar9003_eeprom.o.old 24101 56 3184 27341 6acd drivers/net/wireless/ath/ath9k/ar9003_eeprom.o.new 6834 56 1032 7922 1ef2 drivers/net/wireless/ath/ath9k/ar9002_phy.o.old 6780 56 1032 7868 1ebc drivers/net/wireless/ath/ath9k/ar9002_phy.o.new 36211 64 8624 44899 af63 drivers/net/wireless/ath/ath9k/hw.o.new 36401 64 8624 45089 b021 drivers/net/wireless/ath/ath9k/hw.o.old 9281 56 1496 10833 2a51 drivers/net/wireless/ath/ath9k/ar9003_calib.o.old 9150 56 1496 10702 29ce drivers/net/wireless/ath/ath9k/ar9003_calib.o.new Use ARRAY_SIZE instead of a magic number. Signed-off-by: Joe Perches <joe@perches.com> Signed-off-by: John W. Linville <linville@tuxdriver.com>
2010-11-21 10:38:53 +08:00
static const u_int32_t offset_array[3] = {
AR_PHY_RX_IQCAL_CORR_B0,
AR_PHY_RX_IQCAL_CORR_B1,
AR_PHY_RX_IQCAL_CORR_B2,
};
for (i = 0; i < numChains; i++) {
powerMeasI = ah->totalPowerMeasI[i];
powerMeasQ = ah->totalPowerMeasQ[i];
iqCorrMeas = ah->totalIqCorrMeas[i];
ath_dbg(common, CALIBRATE,
"Starting IQ Cal and Correction for Chain %d\n", i);
ath_dbg(common, CALIBRATE,
"Original: Chn %d iq_corr_meas = 0x%08x\n",
i, ah->totalIqCorrMeas[i]);
iqCorrNeg = 0;
if (iqCorrMeas > 0x80000000) {
iqCorrMeas = (0xffffffff - iqCorrMeas) + 1;
iqCorrNeg = 1;
}
ath_dbg(common, CALIBRATE, "Chn %d pwr_meas_i = 0x%08x\n",
i, powerMeasI);
ath_dbg(common, CALIBRATE, "Chn %d pwr_meas_q = 0x%08x\n",
i, powerMeasQ);
ath_dbg(common, CALIBRATE, "iqCorrNeg is 0x%08x\n", iqCorrNeg);
iCoffDenom = (powerMeasI / 2 + powerMeasQ / 2) / 256;
qCoffDenom = powerMeasQ / 64;
if ((iCoffDenom != 0) && (qCoffDenom != 0)) {
iCoff = iqCorrMeas / iCoffDenom;
qCoff = powerMeasI / qCoffDenom - 64;
ath_dbg(common, CALIBRATE, "Chn %d iCoff = 0x%08x\n",
i, iCoff);
ath_dbg(common, CALIBRATE, "Chn %d qCoff = 0x%08x\n",
i, qCoff);
/* Force bounds on iCoff */
if (iCoff >= 63)
iCoff = 63;
else if (iCoff <= -63)
iCoff = -63;
/* Negate iCoff if iqCorrNeg == 0 */
if (iqCorrNeg == 0x0)
iCoff = -iCoff;
/* Force bounds on qCoff */
if (qCoff >= 63)
qCoff = 63;
else if (qCoff <= -63)
qCoff = -63;
iCoff = iCoff & 0x7f;
qCoff = qCoff & 0x7f;
ath_dbg(common, CALIBRATE,
"Chn %d : iCoff = 0x%x qCoff = 0x%x\n",
i, iCoff, qCoff);
ath_dbg(common, CALIBRATE,
"Register offset (0x%04x) before update = 0x%x\n",
offset_array[i],
REG_READ(ah, offset_array[i]));
REG_RMW_FIELD(ah, offset_array[i],
AR_PHY_RX_IQCAL_CORR_IQCORR_Q_I_COFF,
iCoff);
REG_RMW_FIELD(ah, offset_array[i],
AR_PHY_RX_IQCAL_CORR_IQCORR_Q_Q_COFF,
qCoff);
ath_dbg(common, CALIBRATE,
"Register offset (0x%04x) QI COFF (bitfields 0x%08x) after update = 0x%x\n",
offset_array[i],
AR_PHY_RX_IQCAL_CORR_IQCORR_Q_I_COFF,
REG_READ(ah, offset_array[i]));
ath_dbg(common, CALIBRATE,
"Register offset (0x%04x) QQ COFF (bitfields 0x%08x) after update = 0x%x\n",
offset_array[i],
AR_PHY_RX_IQCAL_CORR_IQCORR_Q_Q_COFF,
REG_READ(ah, offset_array[i]));
ath_dbg(common, CALIBRATE,
"IQ Cal and Correction done for Chain %d\n", i);
}
}
REG_SET_BIT(ah, AR_PHY_RX_IQCAL_CORR_B0,
AR_PHY_RX_IQCAL_CORR_IQCORR_ENABLE);
ath_dbg(common, CALIBRATE,
"IQ Cal and Correction (offset 0x%04x) enabled (bit position 0x%08x). New Value 0x%08x\n",
(unsigned) (AR_PHY_RX_IQCAL_CORR_B0),
AR_PHY_RX_IQCAL_CORR_IQCORR_ENABLE,
REG_READ(ah, AR_PHY_RX_IQCAL_CORR_B0));
}
static const struct ath9k_percal_data iq_cal_single_sample = {
IQ_MISMATCH_CAL,
MIN_CAL_SAMPLES,
PER_MAX_LOG_COUNT,
ar9003_hw_iqcal_collect,
ar9003_hw_iqcalibrate
};
static void ar9003_hw_init_cal_settings(struct ath_hw *ah)
{
ah->iq_caldata.calData = &iq_cal_single_sample;
}
/*
* solve 4x4 linear equation used in loopback iq cal.
*/
static bool ar9003_hw_solve_iq_cal(struct ath_hw *ah,
s32 sin_2phi_1,
s32 cos_2phi_1,
s32 sin_2phi_2,
s32 cos_2phi_2,
s32 mag_a0_d0,
s32 phs_a0_d0,
s32 mag_a1_d0,
s32 phs_a1_d0,
s32 solved_eq[])
{
s32 f1 = cos_2phi_1 - cos_2phi_2,
f3 = sin_2phi_1 - sin_2phi_2,
f2;
s32 mag_tx, phs_tx, mag_rx, phs_rx;
const s32 result_shift = 1 << 15;
struct ath_common *common = ath9k_hw_common(ah);
f2 = (f1 * f1 + f3 * f3) / result_shift;
if (!f2) {
ath_dbg(common, CALIBRATE, "Divide by 0\n");
return false;
}
/* mag mismatch, tx */
mag_tx = f1 * (mag_a0_d0 - mag_a1_d0) + f3 * (phs_a0_d0 - phs_a1_d0);
/* phs mismatch, tx */
phs_tx = f3 * (-mag_a0_d0 + mag_a1_d0) + f1 * (phs_a0_d0 - phs_a1_d0);
mag_tx = (mag_tx / f2);
phs_tx = (phs_tx / f2);
/* mag mismatch, rx */
mag_rx = mag_a0_d0 - (cos_2phi_1 * mag_tx + sin_2phi_1 * phs_tx) /
result_shift;
/* phs mismatch, rx */
phs_rx = phs_a0_d0 + (sin_2phi_1 * mag_tx - cos_2phi_1 * phs_tx) /
result_shift;
solved_eq[0] = mag_tx;
solved_eq[1] = phs_tx;
solved_eq[2] = mag_rx;
solved_eq[3] = phs_rx;
return true;
}
static s32 ar9003_hw_find_mag_approx(struct ath_hw *ah, s32 in_re, s32 in_im)
{
s32 abs_i = abs(in_re),
abs_q = abs(in_im),
max_abs, min_abs;
if (abs_i > abs_q) {
max_abs = abs_i;
min_abs = abs_q;
} else {
max_abs = abs_q;
min_abs = abs_i;
}
return max_abs - (max_abs / 32) + (min_abs / 8) + (min_abs / 4);
}
#define DELPT 32
static bool ar9003_hw_calc_iq_corr(struct ath_hw *ah,
s32 chain_idx,
const s32 iq_res[],
s32 iqc_coeff[])
{
s32 i2_m_q2_a0_d0, i2_p_q2_a0_d0, iq_corr_a0_d0,
i2_m_q2_a0_d1, i2_p_q2_a0_d1, iq_corr_a0_d1,
i2_m_q2_a1_d0, i2_p_q2_a1_d0, iq_corr_a1_d0,
i2_m_q2_a1_d1, i2_p_q2_a1_d1, iq_corr_a1_d1;
s32 mag_a0_d0, mag_a1_d0, mag_a0_d1, mag_a1_d1,
phs_a0_d0, phs_a1_d0, phs_a0_d1, phs_a1_d1,
sin_2phi_1, cos_2phi_1,
sin_2phi_2, cos_2phi_2;
s32 mag_tx, phs_tx, mag_rx, phs_rx;
s32 solved_eq[4], mag_corr_tx, phs_corr_tx, mag_corr_rx, phs_corr_rx,
q_q_coff, q_i_coff;
const s32 res_scale = 1 << 15;
const s32 delpt_shift = 1 << 8;
s32 mag1, mag2;
struct ath_common *common = ath9k_hw_common(ah);
i2_m_q2_a0_d0 = iq_res[0] & 0xfff;
i2_p_q2_a0_d0 = (iq_res[0] >> 12) & 0xfff;
iq_corr_a0_d0 = ((iq_res[0] >> 24) & 0xff) + ((iq_res[1] & 0xf) << 8);
if (i2_m_q2_a0_d0 > 0x800)
i2_m_q2_a0_d0 = -((0xfff - i2_m_q2_a0_d0) + 1);
if (i2_p_q2_a0_d0 > 0x800)
i2_p_q2_a0_d0 = -((0xfff - i2_p_q2_a0_d0) + 1);
if (iq_corr_a0_d0 > 0x800)
iq_corr_a0_d0 = -((0xfff - iq_corr_a0_d0) + 1);
i2_m_q2_a0_d1 = (iq_res[1] >> 4) & 0xfff;
i2_p_q2_a0_d1 = (iq_res[2] & 0xfff);
iq_corr_a0_d1 = (iq_res[2] >> 12) & 0xfff;
if (i2_m_q2_a0_d1 > 0x800)
i2_m_q2_a0_d1 = -((0xfff - i2_m_q2_a0_d1) + 1);
if (i2_p_q2_a0_d1 > 0x800)
i2_p_q2_a0_d1 = -((0xfff - i2_p_q2_a0_d1) + 1);
if (iq_corr_a0_d1 > 0x800)
iq_corr_a0_d1 = -((0xfff - iq_corr_a0_d1) + 1);
i2_m_q2_a1_d0 = ((iq_res[2] >> 24) & 0xff) + ((iq_res[3] & 0xf) << 8);
i2_p_q2_a1_d0 = (iq_res[3] >> 4) & 0xfff;
iq_corr_a1_d0 = iq_res[4] & 0xfff;
if (i2_m_q2_a1_d0 > 0x800)
i2_m_q2_a1_d0 = -((0xfff - i2_m_q2_a1_d0) + 1);
if (i2_p_q2_a1_d0 > 0x800)
i2_p_q2_a1_d0 = -((0xfff - i2_p_q2_a1_d0) + 1);
if (iq_corr_a1_d0 > 0x800)
iq_corr_a1_d0 = -((0xfff - iq_corr_a1_d0) + 1);
i2_m_q2_a1_d1 = (iq_res[4] >> 12) & 0xfff;
i2_p_q2_a1_d1 = ((iq_res[4] >> 24) & 0xff) + ((iq_res[5] & 0xf) << 8);
iq_corr_a1_d1 = (iq_res[5] >> 4) & 0xfff;
if (i2_m_q2_a1_d1 > 0x800)
i2_m_q2_a1_d1 = -((0xfff - i2_m_q2_a1_d1) + 1);
if (i2_p_q2_a1_d1 > 0x800)
i2_p_q2_a1_d1 = -((0xfff - i2_p_q2_a1_d1) + 1);
if (iq_corr_a1_d1 > 0x800)
iq_corr_a1_d1 = -((0xfff - iq_corr_a1_d1) + 1);
if ((i2_p_q2_a0_d0 == 0) || (i2_p_q2_a0_d1 == 0) ||
(i2_p_q2_a1_d0 == 0) || (i2_p_q2_a1_d1 == 0)) {
ath_dbg(common, CALIBRATE,
"Divide by 0:\n"
"a0_d0=%d\n"
"a0_d1=%d\n"
"a2_d0=%d\n"
"a1_d1=%d\n",
i2_p_q2_a0_d0, i2_p_q2_a0_d1,
i2_p_q2_a1_d0, i2_p_q2_a1_d1);
return false;
}
mag_a0_d0 = (i2_m_q2_a0_d0 * res_scale) / i2_p_q2_a0_d0;
phs_a0_d0 = (iq_corr_a0_d0 * res_scale) / i2_p_q2_a0_d0;
mag_a0_d1 = (i2_m_q2_a0_d1 * res_scale) / i2_p_q2_a0_d1;
phs_a0_d1 = (iq_corr_a0_d1 * res_scale) / i2_p_q2_a0_d1;
mag_a1_d0 = (i2_m_q2_a1_d0 * res_scale) / i2_p_q2_a1_d0;
phs_a1_d0 = (iq_corr_a1_d0 * res_scale) / i2_p_q2_a1_d0;
mag_a1_d1 = (i2_m_q2_a1_d1 * res_scale) / i2_p_q2_a1_d1;
phs_a1_d1 = (iq_corr_a1_d1 * res_scale) / i2_p_q2_a1_d1;
/* w/o analog phase shift */
sin_2phi_1 = (((mag_a0_d0 - mag_a0_d1) * delpt_shift) / DELPT);
/* w/o analog phase shift */
cos_2phi_1 = (((phs_a0_d1 - phs_a0_d0) * delpt_shift) / DELPT);
/* w/ analog phase shift */
sin_2phi_2 = (((mag_a1_d0 - mag_a1_d1) * delpt_shift) / DELPT);
/* w/ analog phase shift */
cos_2phi_2 = (((phs_a1_d1 - phs_a1_d0) * delpt_shift) / DELPT);
/*
* force sin^2 + cos^2 = 1;
* find magnitude by approximation
*/
mag1 = ar9003_hw_find_mag_approx(ah, cos_2phi_1, sin_2phi_1);
mag2 = ar9003_hw_find_mag_approx(ah, cos_2phi_2, sin_2phi_2);
if ((mag1 == 0) || (mag2 == 0)) {
ath_dbg(common, CALIBRATE, "Divide by 0: mag1=%d, mag2=%d\n",
mag1, mag2);
return false;
}
/* normalization sin and cos by mag */
sin_2phi_1 = (sin_2phi_1 * res_scale / mag1);
cos_2phi_1 = (cos_2phi_1 * res_scale / mag1);
sin_2phi_2 = (sin_2phi_2 * res_scale / mag2);
cos_2phi_2 = (cos_2phi_2 * res_scale / mag2);
/* calculate IQ mismatch */
if (!ar9003_hw_solve_iq_cal(ah,
sin_2phi_1, cos_2phi_1,
sin_2phi_2, cos_2phi_2,
mag_a0_d0, phs_a0_d0,
mag_a1_d0,
phs_a1_d0, solved_eq)) {
ath_dbg(common, CALIBRATE,
"Call to ar9003_hw_solve_iq_cal() failed\n");
return false;
}
mag_tx = solved_eq[0];
phs_tx = solved_eq[1];
mag_rx = solved_eq[2];
phs_rx = solved_eq[3];
ath_dbg(common, CALIBRATE,
"chain %d: mag mismatch=%d phase mismatch=%d\n",
chain_idx, mag_tx/res_scale, phs_tx/res_scale);
if (res_scale == mag_tx) {
ath_dbg(common, CALIBRATE,
"Divide by 0: mag_tx=%d, res_scale=%d\n",
mag_tx, res_scale);
return false;
}
/* calculate and quantize Tx IQ correction factor */
mag_corr_tx = (mag_tx * res_scale) / (res_scale - mag_tx);
phs_corr_tx = -phs_tx;
q_q_coff = (mag_corr_tx * 128 / res_scale);
q_i_coff = (phs_corr_tx * 256 / res_scale);
ath_dbg(common, CALIBRATE, "tx chain %d: mag corr=%d phase corr=%d\n",
chain_idx, q_q_coff, q_i_coff);
if (q_i_coff < -63)
q_i_coff = -63;
if (q_i_coff > 63)
q_i_coff = 63;
if (q_q_coff < -63)
q_q_coff = -63;
if (q_q_coff > 63)
q_q_coff = 63;
iqc_coeff[0] = (q_q_coff * 128) + q_i_coff;
ath_dbg(common, CALIBRATE, "tx chain %d: iq corr coeff=%x\n",
chain_idx, iqc_coeff[0]);
if (-mag_rx == res_scale) {
ath_dbg(common, CALIBRATE,
"Divide by 0: mag_rx=%d, res_scale=%d\n",
mag_rx, res_scale);
return false;
}
/* calculate and quantize Rx IQ correction factors */
mag_corr_rx = (-mag_rx * res_scale) / (res_scale + mag_rx);
phs_corr_rx = -phs_rx;
q_q_coff = (mag_corr_rx * 128 / res_scale);
q_i_coff = (phs_corr_rx * 256 / res_scale);
ath_dbg(common, CALIBRATE, "rx chain %d: mag corr=%d phase corr=%d\n",
chain_idx, q_q_coff, q_i_coff);
if (q_i_coff < -63)
q_i_coff = -63;
if (q_i_coff > 63)
q_i_coff = 63;
if (q_q_coff < -63)
q_q_coff = -63;
if (q_q_coff > 63)
q_q_coff = 63;
iqc_coeff[1] = (q_q_coff * 128) + q_i_coff;
ath_dbg(common, CALIBRATE, "rx chain %d: iq corr coeff=%x\n",
chain_idx, iqc_coeff[1]);
return true;
}
ath9k_hw: Fix Tx IQ Calibration hang issue in AR9003 chips On AR9003 chips, doing three IQ calibrations will possibly cause chip in stuck state. In noisy environment, chip could receive a packet during the middle of three calibrations and it causes the conflict of HW access and the eventual failure. It also causes IQ calibration outliers which results in poor Tx EVM. The IQ Cal procedure is after resetting the chip, run IQ cal 3 times per each cal cycle and find the two closest readings and average of two. The advantage of running Tx IQ cal more than once is that we can compare calibration results for the same gain setting over multiple iterations. Most of the cases the IQ failures were observed after first pass. For the AR9485 and later chips, Tx IQ Calibration is performed along with AGC cal. But for pre-AR9485 chips, Tx IQ cal HW has to be separated from the rest of calibration HW to avoid chip hang. After all calibrations are done in HW, we can start SW post-processing. By doing this way, we minimize the SW difference among all chips. The order of calibration (run IQ cal before other calibration) is also needed to avoid chip hang for chips before AR9485. This issue was originally observed with AR9382. During the issue kernel log was filled with following message ath: timeout (100000 us) on reg 0xa640: 0x00000001 & 0x00000001 != 0x00000000 ath: timeout (100000 us) on reg 0xa2c4: 0x00158dd9 & 0x00000001 != 0x00000000 ath: Unable to reset channel (2412 MHz), reset status -5 ath: Unable to set channel Signed-off-by: Rajkumar Manoharan <rmanoharan@atheros.com> Signed-off-by: John W. Linville <linville@tuxdriver.com>
2011-04-25 00:04:39 +08:00
static void ar9003_hw_detect_outlier(int *mp_coeff, int nmeasurement,
int max_delta)
{
ath9k_hw: Fix Tx IQ Calibration hang issue in AR9003 chips On AR9003 chips, doing three IQ calibrations will possibly cause chip in stuck state. In noisy environment, chip could receive a packet during the middle of three calibrations and it causes the conflict of HW access and the eventual failure. It also causes IQ calibration outliers which results in poor Tx EVM. The IQ Cal procedure is after resetting the chip, run IQ cal 3 times per each cal cycle and find the two closest readings and average of two. The advantage of running Tx IQ cal more than once is that we can compare calibration results for the same gain setting over multiple iterations. Most of the cases the IQ failures were observed after first pass. For the AR9485 and later chips, Tx IQ Calibration is performed along with AGC cal. But for pre-AR9485 chips, Tx IQ cal HW has to be separated from the rest of calibration HW to avoid chip hang. After all calibrations are done in HW, we can start SW post-processing. By doing this way, we minimize the SW difference among all chips. The order of calibration (run IQ cal before other calibration) is also needed to avoid chip hang for chips before AR9485. This issue was originally observed with AR9382. During the issue kernel log was filled with following message ath: timeout (100000 us) on reg 0xa640: 0x00000001 & 0x00000001 != 0x00000000 ath: timeout (100000 us) on reg 0xa2c4: 0x00158dd9 & 0x00000001 != 0x00000000 ath: Unable to reset channel (2412 MHz), reset status -5 ath: Unable to set channel Signed-off-by: Rajkumar Manoharan <rmanoharan@atheros.com> Signed-off-by: John W. Linville <linville@tuxdriver.com>
2011-04-25 00:04:39 +08:00
int mp_max = -64, max_idx = 0;
int mp_min = 63, min_idx = 0;
int mp_avg = 0, i, outlier_idx = 0, mp_count = 0;
ath9k_hw: Fix Tx IQ Calibration hang issue in AR9003 chips On AR9003 chips, doing three IQ calibrations will possibly cause chip in stuck state. In noisy environment, chip could receive a packet during the middle of three calibrations and it causes the conflict of HW access and the eventual failure. It also causes IQ calibration outliers which results in poor Tx EVM. The IQ Cal procedure is after resetting the chip, run IQ cal 3 times per each cal cycle and find the two closest readings and average of two. The advantage of running Tx IQ cal more than once is that we can compare calibration results for the same gain setting over multiple iterations. Most of the cases the IQ failures were observed after first pass. For the AR9485 and later chips, Tx IQ Calibration is performed along with AGC cal. But for pre-AR9485 chips, Tx IQ cal HW has to be separated from the rest of calibration HW to avoid chip hang. After all calibrations are done in HW, we can start SW post-processing. By doing this way, we minimize the SW difference among all chips. The order of calibration (run IQ cal before other calibration) is also needed to avoid chip hang for chips before AR9485. This issue was originally observed with AR9382. During the issue kernel log was filled with following message ath: timeout (100000 us) on reg 0xa640: 0x00000001 & 0x00000001 != 0x00000000 ath: timeout (100000 us) on reg 0xa2c4: 0x00158dd9 & 0x00000001 != 0x00000000 ath: Unable to reset channel (2412 MHz), reset status -5 ath: Unable to set channel Signed-off-by: Rajkumar Manoharan <rmanoharan@atheros.com> Signed-off-by: John W. Linville <linville@tuxdriver.com>
2011-04-25 00:04:39 +08:00
/* find min/max mismatch across all calibrated gains */
for (i = 0; i < nmeasurement; i++) {
if (mp_coeff[i] > mp_max) {
mp_max = mp_coeff[i];
max_idx = i;
} else if (mp_coeff[i] < mp_min) {
mp_min = mp_coeff[i];
min_idx = i;
}
}
ath9k_hw: Fix Tx IQ Calibration hang issue in AR9003 chips On AR9003 chips, doing three IQ calibrations will possibly cause chip in stuck state. In noisy environment, chip could receive a packet during the middle of three calibrations and it causes the conflict of HW access and the eventual failure. It also causes IQ calibration outliers which results in poor Tx EVM. The IQ Cal procedure is after resetting the chip, run IQ cal 3 times per each cal cycle and find the two closest readings and average of two. The advantage of running Tx IQ cal more than once is that we can compare calibration results for the same gain setting over multiple iterations. Most of the cases the IQ failures were observed after first pass. For the AR9485 and later chips, Tx IQ Calibration is performed along with AGC cal. But for pre-AR9485 chips, Tx IQ cal HW has to be separated from the rest of calibration HW to avoid chip hang. After all calibrations are done in HW, we can start SW post-processing. By doing this way, we minimize the SW difference among all chips. The order of calibration (run IQ cal before other calibration) is also needed to avoid chip hang for chips before AR9485. This issue was originally observed with AR9382. During the issue kernel log was filled with following message ath: timeout (100000 us) on reg 0xa640: 0x00000001 & 0x00000001 != 0x00000000 ath: timeout (100000 us) on reg 0xa2c4: 0x00158dd9 & 0x00000001 != 0x00000000 ath: Unable to reset channel (2412 MHz), reset status -5 ath: Unable to set channel Signed-off-by: Rajkumar Manoharan <rmanoharan@atheros.com> Signed-off-by: John W. Linville <linville@tuxdriver.com>
2011-04-25 00:04:39 +08:00
/* find average (exclude max abs value) */
for (i = 0; i < nmeasurement; i++) {
if ((abs(mp_coeff[i]) < abs(mp_max)) ||
(abs(mp_coeff[i]) < abs(mp_min))) {
ath9k_hw: Fix Tx IQ Calibration hang issue in AR9003 chips On AR9003 chips, doing three IQ calibrations will possibly cause chip in stuck state. In noisy environment, chip could receive a packet during the middle of three calibrations and it causes the conflict of HW access and the eventual failure. It also causes IQ calibration outliers which results in poor Tx EVM. The IQ Cal procedure is after resetting the chip, run IQ cal 3 times per each cal cycle and find the two closest readings and average of two. The advantage of running Tx IQ cal more than once is that we can compare calibration results for the same gain setting over multiple iterations. Most of the cases the IQ failures were observed after first pass. For the AR9485 and later chips, Tx IQ Calibration is performed along with AGC cal. But for pre-AR9485 chips, Tx IQ cal HW has to be separated from the rest of calibration HW to avoid chip hang. After all calibrations are done in HW, we can start SW post-processing. By doing this way, we minimize the SW difference among all chips. The order of calibration (run IQ cal before other calibration) is also needed to avoid chip hang for chips before AR9485. This issue was originally observed with AR9382. During the issue kernel log was filled with following message ath: timeout (100000 us) on reg 0xa640: 0x00000001 & 0x00000001 != 0x00000000 ath: timeout (100000 us) on reg 0xa2c4: 0x00158dd9 & 0x00000001 != 0x00000000 ath: Unable to reset channel (2412 MHz), reset status -5 ath: Unable to set channel Signed-off-by: Rajkumar Manoharan <rmanoharan@atheros.com> Signed-off-by: John W. Linville <linville@tuxdriver.com>
2011-04-25 00:04:39 +08:00
mp_avg += mp_coeff[i];
mp_count++;
}
ath9k_hw: Fix Tx IQ Calibration hang issue in AR9003 chips On AR9003 chips, doing three IQ calibrations will possibly cause chip in stuck state. In noisy environment, chip could receive a packet during the middle of three calibrations and it causes the conflict of HW access and the eventual failure. It also causes IQ calibration outliers which results in poor Tx EVM. The IQ Cal procedure is after resetting the chip, run IQ cal 3 times per each cal cycle and find the two closest readings and average of two. The advantage of running Tx IQ cal more than once is that we can compare calibration results for the same gain setting over multiple iterations. Most of the cases the IQ failures were observed after first pass. For the AR9485 and later chips, Tx IQ Calibration is performed along with AGC cal. But for pre-AR9485 chips, Tx IQ cal HW has to be separated from the rest of calibration HW to avoid chip hang. After all calibrations are done in HW, we can start SW post-processing. By doing this way, we minimize the SW difference among all chips. The order of calibration (run IQ cal before other calibration) is also needed to avoid chip hang for chips before AR9485. This issue was originally observed with AR9382. During the issue kernel log was filled with following message ath: timeout (100000 us) on reg 0xa640: 0x00000001 & 0x00000001 != 0x00000000 ath: timeout (100000 us) on reg 0xa2c4: 0x00158dd9 & 0x00000001 != 0x00000000 ath: Unable to reset channel (2412 MHz), reset status -5 ath: Unable to set channel Signed-off-by: Rajkumar Manoharan <rmanoharan@atheros.com> Signed-off-by: John W. Linville <linville@tuxdriver.com>
2011-04-25 00:04:39 +08:00
}
/*
* finding mean magnitude/phase if possible, otherwise
* just use the last value as the mean
*/
if (mp_count)
mp_avg /= mp_count;
else
mp_avg = mp_coeff[nmeasurement - 1];
ath9k_hw: Fix Tx IQ Calibration hang issue in AR9003 chips On AR9003 chips, doing three IQ calibrations will possibly cause chip in stuck state. In noisy environment, chip could receive a packet during the middle of three calibrations and it causes the conflict of HW access and the eventual failure. It also causes IQ calibration outliers which results in poor Tx EVM. The IQ Cal procedure is after resetting the chip, run IQ cal 3 times per each cal cycle and find the two closest readings and average of two. The advantage of running Tx IQ cal more than once is that we can compare calibration results for the same gain setting over multiple iterations. Most of the cases the IQ failures were observed after first pass. For the AR9485 and later chips, Tx IQ Calibration is performed along with AGC cal. But for pre-AR9485 chips, Tx IQ cal HW has to be separated from the rest of calibration HW to avoid chip hang. After all calibrations are done in HW, we can start SW post-processing. By doing this way, we minimize the SW difference among all chips. The order of calibration (run IQ cal before other calibration) is also needed to avoid chip hang for chips before AR9485. This issue was originally observed with AR9382. During the issue kernel log was filled with following message ath: timeout (100000 us) on reg 0xa640: 0x00000001 & 0x00000001 != 0x00000000 ath: timeout (100000 us) on reg 0xa2c4: 0x00158dd9 & 0x00000001 != 0x00000000 ath: Unable to reset channel (2412 MHz), reset status -5 ath: Unable to set channel Signed-off-by: Rajkumar Manoharan <rmanoharan@atheros.com> Signed-off-by: John W. Linville <linville@tuxdriver.com>
2011-04-25 00:04:39 +08:00
/* detect outlier */
if (abs(mp_max - mp_min) > max_delta) {
if (abs(mp_max - mp_avg) > abs(mp_min - mp_avg))
outlier_idx = max_idx;
else
outlier_idx = min_idx;
mp_coeff[outlier_idx] = mp_avg;
ath9k_hw: Fix Tx IQ Calibration hang issue in AR9003 chips On AR9003 chips, doing three IQ calibrations will possibly cause chip in stuck state. In noisy environment, chip could receive a packet during the middle of three calibrations and it causes the conflict of HW access and the eventual failure. It also causes IQ calibration outliers which results in poor Tx EVM. The IQ Cal procedure is after resetting the chip, run IQ cal 3 times per each cal cycle and find the two closest readings and average of two. The advantage of running Tx IQ cal more than once is that we can compare calibration results for the same gain setting over multiple iterations. Most of the cases the IQ failures were observed after first pass. For the AR9485 and later chips, Tx IQ Calibration is performed along with AGC cal. But for pre-AR9485 chips, Tx IQ cal HW has to be separated from the rest of calibration HW to avoid chip hang. After all calibrations are done in HW, we can start SW post-processing. By doing this way, we minimize the SW difference among all chips. The order of calibration (run IQ cal before other calibration) is also needed to avoid chip hang for chips before AR9485. This issue was originally observed with AR9382. During the issue kernel log was filled with following message ath: timeout (100000 us) on reg 0xa640: 0x00000001 & 0x00000001 != 0x00000000 ath: timeout (100000 us) on reg 0xa2c4: 0x00158dd9 & 0x00000001 != 0x00000000 ath: Unable to reset channel (2412 MHz), reset status -5 ath: Unable to set channel Signed-off-by: Rajkumar Manoharan <rmanoharan@atheros.com> Signed-off-by: John W. Linville <linville@tuxdriver.com>
2011-04-25 00:04:39 +08:00
}
}
static void ar9003_hw_tx_iqcal_load_avg_2_passes(struct ath_hw *ah,
u8 num_chains,
struct coeff *coeff,
bool is_reusable)
{
int i, im, nmeasurement;
u32 tx_corr_coeff[MAX_MEASUREMENT][AR9300_MAX_CHAINS];
struct ath9k_hw_cal_data *caldata = ah->caldata;
memset(tx_corr_coeff, 0, sizeof(tx_corr_coeff));
for (i = 0; i < MAX_MEASUREMENT / 2; i++) {
tx_corr_coeff[i * 2][0] = tx_corr_coeff[(i * 2) + 1][0] =
AR_PHY_TX_IQCAL_CORR_COEFF_B0(i);
if (!AR_SREV_9485(ah)) {
tx_corr_coeff[i * 2][1] =
tx_corr_coeff[(i * 2) + 1][1] =
AR_PHY_TX_IQCAL_CORR_COEFF_B1(i);
tx_corr_coeff[i * 2][2] =
tx_corr_coeff[(i * 2) + 1][2] =
AR_PHY_TX_IQCAL_CORR_COEFF_B2(i);
}
}
/* Load the average of 2 passes */
for (i = 0; i < num_chains; i++) {
ath9k_hw: Fix Tx IQ Calibration hang issue in AR9003 chips On AR9003 chips, doing three IQ calibrations will possibly cause chip in stuck state. In noisy environment, chip could receive a packet during the middle of three calibrations and it causes the conflict of HW access and the eventual failure. It also causes IQ calibration outliers which results in poor Tx EVM. The IQ Cal procedure is after resetting the chip, run IQ cal 3 times per each cal cycle and find the two closest readings and average of two. The advantage of running Tx IQ cal more than once is that we can compare calibration results for the same gain setting over multiple iterations. Most of the cases the IQ failures were observed after first pass. For the AR9485 and later chips, Tx IQ Calibration is performed along with AGC cal. But for pre-AR9485 chips, Tx IQ cal HW has to be separated from the rest of calibration HW to avoid chip hang. After all calibrations are done in HW, we can start SW post-processing. By doing this way, we minimize the SW difference among all chips. The order of calibration (run IQ cal before other calibration) is also needed to avoid chip hang for chips before AR9485. This issue was originally observed with AR9382. During the issue kernel log was filled with following message ath: timeout (100000 us) on reg 0xa640: 0x00000001 & 0x00000001 != 0x00000000 ath: timeout (100000 us) on reg 0xa2c4: 0x00158dd9 & 0x00000001 != 0x00000000 ath: Unable to reset channel (2412 MHz), reset status -5 ath: Unable to set channel Signed-off-by: Rajkumar Manoharan <rmanoharan@atheros.com> Signed-off-by: John W. Linville <linville@tuxdriver.com>
2011-04-25 00:04:39 +08:00
nmeasurement = REG_READ_FIELD(ah,
AR_PHY_TX_IQCAL_STATUS_B0,
AR_PHY_CALIBRATED_GAINS_0);
if (nmeasurement > MAX_MEASUREMENT)
nmeasurement = MAX_MEASUREMENT;
ath9k_hw: Fix Tx IQ Calibration hang issue in AR9003 chips On AR9003 chips, doing three IQ calibrations will possibly cause chip in stuck state. In noisy environment, chip could receive a packet during the middle of three calibrations and it causes the conflict of HW access and the eventual failure. It also causes IQ calibration outliers which results in poor Tx EVM. The IQ Cal procedure is after resetting the chip, run IQ cal 3 times per each cal cycle and find the two closest readings and average of two. The advantage of running Tx IQ cal more than once is that we can compare calibration results for the same gain setting over multiple iterations. Most of the cases the IQ failures were observed after first pass. For the AR9485 and later chips, Tx IQ Calibration is performed along with AGC cal. But for pre-AR9485 chips, Tx IQ cal HW has to be separated from the rest of calibration HW to avoid chip hang. After all calibrations are done in HW, we can start SW post-processing. By doing this way, we minimize the SW difference among all chips. The order of calibration (run IQ cal before other calibration) is also needed to avoid chip hang for chips before AR9485. This issue was originally observed with AR9382. During the issue kernel log was filled with following message ath: timeout (100000 us) on reg 0xa640: 0x00000001 & 0x00000001 != 0x00000000 ath: timeout (100000 us) on reg 0xa2c4: 0x00158dd9 & 0x00000001 != 0x00000000 ath: Unable to reset channel (2412 MHz), reset status -5 ath: Unable to set channel Signed-off-by: Rajkumar Manoharan <rmanoharan@atheros.com> Signed-off-by: John W. Linville <linville@tuxdriver.com>
2011-04-25 00:04:39 +08:00
/* detect outlier only if nmeasurement > 1 */
if (nmeasurement > 1) {
/* Detect magnitude outlier */
ar9003_hw_detect_outlier(coeff->mag_coeff[i],
nmeasurement, MAX_MAG_DELTA);
ath9k_hw: Fix Tx IQ Calibration hang issue in AR9003 chips On AR9003 chips, doing three IQ calibrations will possibly cause chip in stuck state. In noisy environment, chip could receive a packet during the middle of three calibrations and it causes the conflict of HW access and the eventual failure. It also causes IQ calibration outliers which results in poor Tx EVM. The IQ Cal procedure is after resetting the chip, run IQ cal 3 times per each cal cycle and find the two closest readings and average of two. The advantage of running Tx IQ cal more than once is that we can compare calibration results for the same gain setting over multiple iterations. Most of the cases the IQ failures were observed after first pass. For the AR9485 and later chips, Tx IQ Calibration is performed along with AGC cal. But for pre-AR9485 chips, Tx IQ cal HW has to be separated from the rest of calibration HW to avoid chip hang. After all calibrations are done in HW, we can start SW post-processing. By doing this way, we minimize the SW difference among all chips. The order of calibration (run IQ cal before other calibration) is also needed to avoid chip hang for chips before AR9485. This issue was originally observed with AR9382. During the issue kernel log was filled with following message ath: timeout (100000 us) on reg 0xa640: 0x00000001 & 0x00000001 != 0x00000000 ath: timeout (100000 us) on reg 0xa2c4: 0x00158dd9 & 0x00000001 != 0x00000000 ath: Unable to reset channel (2412 MHz), reset status -5 ath: Unable to set channel Signed-off-by: Rajkumar Manoharan <rmanoharan@atheros.com> Signed-off-by: John W. Linville <linville@tuxdriver.com>
2011-04-25 00:04:39 +08:00
/* Detect phase outlier */
ar9003_hw_detect_outlier(coeff->phs_coeff[i],
nmeasurement, MAX_PHS_DELTA);
}
ath9k_hw: Fix Tx IQ Calibration hang issue in AR9003 chips On AR9003 chips, doing three IQ calibrations will possibly cause chip in stuck state. In noisy environment, chip could receive a packet during the middle of three calibrations and it causes the conflict of HW access and the eventual failure. It also causes IQ calibration outliers which results in poor Tx EVM. The IQ Cal procedure is after resetting the chip, run IQ cal 3 times per each cal cycle and find the two closest readings and average of two. The advantage of running Tx IQ cal more than once is that we can compare calibration results for the same gain setting over multiple iterations. Most of the cases the IQ failures were observed after first pass. For the AR9485 and later chips, Tx IQ Calibration is performed along with AGC cal. But for pre-AR9485 chips, Tx IQ cal HW has to be separated from the rest of calibration HW to avoid chip hang. After all calibrations are done in HW, we can start SW post-processing. By doing this way, we minimize the SW difference among all chips. The order of calibration (run IQ cal before other calibration) is also needed to avoid chip hang for chips before AR9485. This issue was originally observed with AR9382. During the issue kernel log was filled with following message ath: timeout (100000 us) on reg 0xa640: 0x00000001 & 0x00000001 != 0x00000000 ath: timeout (100000 us) on reg 0xa2c4: 0x00158dd9 & 0x00000001 != 0x00000000 ath: Unable to reset channel (2412 MHz), reset status -5 ath: Unable to set channel Signed-off-by: Rajkumar Manoharan <rmanoharan@atheros.com> Signed-off-by: John W. Linville <linville@tuxdriver.com>
2011-04-25 00:04:39 +08:00
for (im = 0; im < nmeasurement; im++) {
coeff->iqc_coeff[0] = (coeff->mag_coeff[i][im] & 0x7f) |
((coeff->phs_coeff[i][im] & 0x7f) << 7);
if ((im % 2) == 0)
REG_RMW_FIELD(ah, tx_corr_coeff[im][i],
AR_PHY_TX_IQCAL_CORR_COEFF_00_COEFF_TABLE,
coeff->iqc_coeff[0]);
else
REG_RMW_FIELD(ah, tx_corr_coeff[im][i],
AR_PHY_TX_IQCAL_CORR_COEFF_01_COEFF_TABLE,
coeff->iqc_coeff[0]);
if (caldata)
caldata->tx_corr_coeff[im][i] =
coeff->iqc_coeff[0];
}
if (caldata)
caldata->num_measures[i] = nmeasurement;
}
REG_RMW_FIELD(ah, AR_PHY_TX_IQCAL_CONTROL_3,
AR_PHY_TX_IQCAL_CONTROL_3_IQCORR_EN, 0x1);
REG_RMW_FIELD(ah, AR_PHY_RX_IQCAL_CORR_B0,
AR_PHY_RX_IQCAL_CORR_B0_LOOPBACK_IQCORR_EN, 0x1);
if (caldata)
caldata->done_txiqcal_once = is_reusable;
return;
}
ath9k_hw: Fix Tx IQ Calibration hang issue in AR9003 chips On AR9003 chips, doing three IQ calibrations will possibly cause chip in stuck state. In noisy environment, chip could receive a packet during the middle of three calibrations and it causes the conflict of HW access and the eventual failure. It also causes IQ calibration outliers which results in poor Tx EVM. The IQ Cal procedure is after resetting the chip, run IQ cal 3 times per each cal cycle and find the two closest readings and average of two. The advantage of running Tx IQ cal more than once is that we can compare calibration results for the same gain setting over multiple iterations. Most of the cases the IQ failures were observed after first pass. For the AR9485 and later chips, Tx IQ Calibration is performed along with AGC cal. But for pre-AR9485 chips, Tx IQ cal HW has to be separated from the rest of calibration HW to avoid chip hang. After all calibrations are done in HW, we can start SW post-processing. By doing this way, we minimize the SW difference among all chips. The order of calibration (run IQ cal before other calibration) is also needed to avoid chip hang for chips before AR9485. This issue was originally observed with AR9382. During the issue kernel log was filled with following message ath: timeout (100000 us) on reg 0xa640: 0x00000001 & 0x00000001 != 0x00000000 ath: timeout (100000 us) on reg 0xa2c4: 0x00158dd9 & 0x00000001 != 0x00000000 ath: Unable to reset channel (2412 MHz), reset status -5 ath: Unable to set channel Signed-off-by: Rajkumar Manoharan <rmanoharan@atheros.com> Signed-off-by: John W. Linville <linville@tuxdriver.com>
2011-04-25 00:04:39 +08:00
static bool ar9003_hw_tx_iq_cal_run(struct ath_hw *ah)
{
struct ath_common *common = ath9k_hw_common(ah);
u8 tx_gain_forced;
tx_gain_forced = REG_READ_FIELD(ah, AR_PHY_TX_FORCED_GAIN,
AR_PHY_TXGAIN_FORCE);
if (tx_gain_forced)
REG_RMW_FIELD(ah, AR_PHY_TX_FORCED_GAIN,
AR_PHY_TXGAIN_FORCE, 0);
ath9k_hw: Fix Tx IQ Calibration hang issue in AR9003 chips On AR9003 chips, doing three IQ calibrations will possibly cause chip in stuck state. In noisy environment, chip could receive a packet during the middle of three calibrations and it causes the conflict of HW access and the eventual failure. It also causes IQ calibration outliers which results in poor Tx EVM. The IQ Cal procedure is after resetting the chip, run IQ cal 3 times per each cal cycle and find the two closest readings and average of two. The advantage of running Tx IQ cal more than once is that we can compare calibration results for the same gain setting over multiple iterations. Most of the cases the IQ failures were observed after first pass. For the AR9485 and later chips, Tx IQ Calibration is performed along with AGC cal. But for pre-AR9485 chips, Tx IQ cal HW has to be separated from the rest of calibration HW to avoid chip hang. After all calibrations are done in HW, we can start SW post-processing. By doing this way, we minimize the SW difference among all chips. The order of calibration (run IQ cal before other calibration) is also needed to avoid chip hang for chips before AR9485. This issue was originally observed with AR9382. During the issue kernel log was filled with following message ath: timeout (100000 us) on reg 0xa640: 0x00000001 & 0x00000001 != 0x00000000 ath: timeout (100000 us) on reg 0xa2c4: 0x00158dd9 & 0x00000001 != 0x00000000 ath: Unable to reset channel (2412 MHz), reset status -5 ath: Unable to set channel Signed-off-by: Rajkumar Manoharan <rmanoharan@atheros.com> Signed-off-by: John W. Linville <linville@tuxdriver.com>
2011-04-25 00:04:39 +08:00
REG_RMW_FIELD(ah, AR_PHY_TX_IQCAL_START,
AR_PHY_TX_IQCAL_START_DO_CAL, 1);
if (!ath9k_hw_wait(ah, AR_PHY_TX_IQCAL_START,
AR_PHY_TX_IQCAL_START_DO_CAL, 0,
AH_WAIT_TIMEOUT)) {
ath_dbg(common, CALIBRATE, "Tx IQ Cal is not completed\n");
ath9k_hw: Fix Tx IQ Calibration hang issue in AR9003 chips On AR9003 chips, doing three IQ calibrations will possibly cause chip in stuck state. In noisy environment, chip could receive a packet during the middle of three calibrations and it causes the conflict of HW access and the eventual failure. It also causes IQ calibration outliers which results in poor Tx EVM. The IQ Cal procedure is after resetting the chip, run IQ cal 3 times per each cal cycle and find the two closest readings and average of two. The advantage of running Tx IQ cal more than once is that we can compare calibration results for the same gain setting over multiple iterations. Most of the cases the IQ failures were observed after first pass. For the AR9485 and later chips, Tx IQ Calibration is performed along with AGC cal. But for pre-AR9485 chips, Tx IQ cal HW has to be separated from the rest of calibration HW to avoid chip hang. After all calibrations are done in HW, we can start SW post-processing. By doing this way, we minimize the SW difference among all chips. The order of calibration (run IQ cal before other calibration) is also needed to avoid chip hang for chips before AR9485. This issue was originally observed with AR9382. During the issue kernel log was filled with following message ath: timeout (100000 us) on reg 0xa640: 0x00000001 & 0x00000001 != 0x00000000 ath: timeout (100000 us) on reg 0xa2c4: 0x00158dd9 & 0x00000001 != 0x00000000 ath: Unable to reset channel (2412 MHz), reset status -5 ath: Unable to set channel Signed-off-by: Rajkumar Manoharan <rmanoharan@atheros.com> Signed-off-by: John W. Linville <linville@tuxdriver.com>
2011-04-25 00:04:39 +08:00
return false;
}
return true;
}
static void ar9003_hw_tx_iq_cal_post_proc(struct ath_hw *ah, bool is_reusable)
{
struct ath_common *common = ath9k_hw_common(ah);
const u32 txiqcal_status[AR9300_MAX_CHAINS] = {
ath9k_hw: Fix Tx IQ Calibration hang issue in AR9003 chips On AR9003 chips, doing three IQ calibrations will possibly cause chip in stuck state. In noisy environment, chip could receive a packet during the middle of three calibrations and it causes the conflict of HW access and the eventual failure. It also causes IQ calibration outliers which results in poor Tx EVM. The IQ Cal procedure is after resetting the chip, run IQ cal 3 times per each cal cycle and find the two closest readings and average of two. The advantage of running Tx IQ cal more than once is that we can compare calibration results for the same gain setting over multiple iterations. Most of the cases the IQ failures were observed after first pass. For the AR9485 and later chips, Tx IQ Calibration is performed along with AGC cal. But for pre-AR9485 chips, Tx IQ cal HW has to be separated from the rest of calibration HW to avoid chip hang. After all calibrations are done in HW, we can start SW post-processing. By doing this way, we minimize the SW difference among all chips. The order of calibration (run IQ cal before other calibration) is also needed to avoid chip hang for chips before AR9485. This issue was originally observed with AR9382. During the issue kernel log was filled with following message ath: timeout (100000 us) on reg 0xa640: 0x00000001 & 0x00000001 != 0x00000000 ath: timeout (100000 us) on reg 0xa2c4: 0x00158dd9 & 0x00000001 != 0x00000000 ath: Unable to reset channel (2412 MHz), reset status -5 ath: Unable to set channel Signed-off-by: Rajkumar Manoharan <rmanoharan@atheros.com> Signed-off-by: John W. Linville <linville@tuxdriver.com>
2011-04-25 00:04:39 +08:00
AR_PHY_TX_IQCAL_STATUS_B0,
AR_PHY_TX_IQCAL_STATUS_B1,
AR_PHY_TX_IQCAL_STATUS_B2,
};
const u_int32_t chan_info_tab[] = {
AR_PHY_CHAN_INFO_TAB_0,
AR_PHY_CHAN_INFO_TAB_1,
AR_PHY_CHAN_INFO_TAB_2,
};
struct coeff coeff;
s32 iq_res[6];
u8 num_chains = 0;
ath9k_hw: Fix Tx IQ Calibration hang issue in AR9003 chips On AR9003 chips, doing three IQ calibrations will possibly cause chip in stuck state. In noisy environment, chip could receive a packet during the middle of three calibrations and it causes the conflict of HW access and the eventual failure. It also causes IQ calibration outliers which results in poor Tx EVM. The IQ Cal procedure is after resetting the chip, run IQ cal 3 times per each cal cycle and find the two closest readings and average of two. The advantage of running Tx IQ cal more than once is that we can compare calibration results for the same gain setting over multiple iterations. Most of the cases the IQ failures were observed after first pass. For the AR9485 and later chips, Tx IQ Calibration is performed along with AGC cal. But for pre-AR9485 chips, Tx IQ cal HW has to be separated from the rest of calibration HW to avoid chip hang. After all calibrations are done in HW, we can start SW post-processing. By doing this way, we minimize the SW difference among all chips. The order of calibration (run IQ cal before other calibration) is also needed to avoid chip hang for chips before AR9485. This issue was originally observed with AR9382. During the issue kernel log was filled with following message ath: timeout (100000 us) on reg 0xa640: 0x00000001 & 0x00000001 != 0x00000000 ath: timeout (100000 us) on reg 0xa2c4: 0x00158dd9 & 0x00000001 != 0x00000000 ath: Unable to reset channel (2412 MHz), reset status -5 ath: Unable to set channel Signed-off-by: Rajkumar Manoharan <rmanoharan@atheros.com> Signed-off-by: John W. Linville <linville@tuxdriver.com>
2011-04-25 00:04:39 +08:00
int i, im, j;
int nmeasurement;
for (i = 0; i < AR9300_MAX_CHAINS; i++) {
if (ah->txchainmask & (1 << i))
num_chains++;
}
ath9k_hw: Fix Tx IQ Calibration hang issue in AR9003 chips On AR9003 chips, doing three IQ calibrations will possibly cause chip in stuck state. In noisy environment, chip could receive a packet during the middle of three calibrations and it causes the conflict of HW access and the eventual failure. It also causes IQ calibration outliers which results in poor Tx EVM. The IQ Cal procedure is after resetting the chip, run IQ cal 3 times per each cal cycle and find the two closest readings and average of two. The advantage of running Tx IQ cal more than once is that we can compare calibration results for the same gain setting over multiple iterations. Most of the cases the IQ failures were observed after first pass. For the AR9485 and later chips, Tx IQ Calibration is performed along with AGC cal. But for pre-AR9485 chips, Tx IQ cal HW has to be separated from the rest of calibration HW to avoid chip hang. After all calibrations are done in HW, we can start SW post-processing. By doing this way, we minimize the SW difference among all chips. The order of calibration (run IQ cal before other calibration) is also needed to avoid chip hang for chips before AR9485. This issue was originally observed with AR9382. During the issue kernel log was filled with following message ath: timeout (100000 us) on reg 0xa640: 0x00000001 & 0x00000001 != 0x00000000 ath: timeout (100000 us) on reg 0xa2c4: 0x00158dd9 & 0x00000001 != 0x00000000 ath: Unable to reset channel (2412 MHz), reset status -5 ath: Unable to set channel Signed-off-by: Rajkumar Manoharan <rmanoharan@atheros.com> Signed-off-by: John W. Linville <linville@tuxdriver.com>
2011-04-25 00:04:39 +08:00
for (i = 0; i < num_chains; i++) {
nmeasurement = REG_READ_FIELD(ah,
AR_PHY_TX_IQCAL_STATUS_B0,
AR_PHY_CALIBRATED_GAINS_0);
if (nmeasurement > MAX_MEASUREMENT)
nmeasurement = MAX_MEASUREMENT;
ath9k_hw: Fix Tx IQ Calibration hang issue in AR9003 chips On AR9003 chips, doing three IQ calibrations will possibly cause chip in stuck state. In noisy environment, chip could receive a packet during the middle of three calibrations and it causes the conflict of HW access and the eventual failure. It also causes IQ calibration outliers which results in poor Tx EVM. The IQ Cal procedure is after resetting the chip, run IQ cal 3 times per each cal cycle and find the two closest readings and average of two. The advantage of running Tx IQ cal more than once is that we can compare calibration results for the same gain setting over multiple iterations. Most of the cases the IQ failures were observed after first pass. For the AR9485 and later chips, Tx IQ Calibration is performed along with AGC cal. But for pre-AR9485 chips, Tx IQ cal HW has to be separated from the rest of calibration HW to avoid chip hang. After all calibrations are done in HW, we can start SW post-processing. By doing this way, we minimize the SW difference among all chips. The order of calibration (run IQ cal before other calibration) is also needed to avoid chip hang for chips before AR9485. This issue was originally observed with AR9382. During the issue kernel log was filled with following message ath: timeout (100000 us) on reg 0xa640: 0x00000001 & 0x00000001 != 0x00000000 ath: timeout (100000 us) on reg 0xa2c4: 0x00158dd9 & 0x00000001 != 0x00000000 ath: Unable to reset channel (2412 MHz), reset status -5 ath: Unable to set channel Signed-off-by: Rajkumar Manoharan <rmanoharan@atheros.com> Signed-off-by: John W. Linville <linville@tuxdriver.com>
2011-04-25 00:04:39 +08:00
for (im = 0; im < nmeasurement; im++) {
ath_dbg(common, CALIBRATE,
"Doing Tx IQ Cal for chain %d\n", i);
ath9k_hw: Fix Tx IQ Calibration hang issue in AR9003 chips On AR9003 chips, doing three IQ calibrations will possibly cause chip in stuck state. In noisy environment, chip could receive a packet during the middle of three calibrations and it causes the conflict of HW access and the eventual failure. It also causes IQ calibration outliers which results in poor Tx EVM. The IQ Cal procedure is after resetting the chip, run IQ cal 3 times per each cal cycle and find the two closest readings and average of two. The advantage of running Tx IQ cal more than once is that we can compare calibration results for the same gain setting over multiple iterations. Most of the cases the IQ failures were observed after first pass. For the AR9485 and later chips, Tx IQ Calibration is performed along with AGC cal. But for pre-AR9485 chips, Tx IQ cal HW has to be separated from the rest of calibration HW to avoid chip hang. After all calibrations are done in HW, we can start SW post-processing. By doing this way, we minimize the SW difference among all chips. The order of calibration (run IQ cal before other calibration) is also needed to avoid chip hang for chips before AR9485. This issue was originally observed with AR9382. During the issue kernel log was filled with following message ath: timeout (100000 us) on reg 0xa640: 0x00000001 & 0x00000001 != 0x00000000 ath: timeout (100000 us) on reg 0xa2c4: 0x00158dd9 & 0x00000001 != 0x00000000 ath: Unable to reset channel (2412 MHz), reset status -5 ath: Unable to set channel Signed-off-by: Rajkumar Manoharan <rmanoharan@atheros.com> Signed-off-by: John W. Linville <linville@tuxdriver.com>
2011-04-25 00:04:39 +08:00
if (REG_READ(ah, txiqcal_status[i]) &
AR_PHY_TX_IQCAL_STATUS_FAILED) {
ath_dbg(common, CALIBRATE,
"Tx IQ Cal failed for chain %d\n", i);
ath9k_hw: Fix Tx IQ Calibration hang issue in AR9003 chips On AR9003 chips, doing three IQ calibrations will possibly cause chip in stuck state. In noisy environment, chip could receive a packet during the middle of three calibrations and it causes the conflict of HW access and the eventual failure. It also causes IQ calibration outliers which results in poor Tx EVM. The IQ Cal procedure is after resetting the chip, run IQ cal 3 times per each cal cycle and find the two closest readings and average of two. The advantage of running Tx IQ cal more than once is that we can compare calibration results for the same gain setting over multiple iterations. Most of the cases the IQ failures were observed after first pass. For the AR9485 and later chips, Tx IQ Calibration is performed along with AGC cal. But for pre-AR9485 chips, Tx IQ cal HW has to be separated from the rest of calibration HW to avoid chip hang. After all calibrations are done in HW, we can start SW post-processing. By doing this way, we minimize the SW difference among all chips. The order of calibration (run IQ cal before other calibration) is also needed to avoid chip hang for chips before AR9485. This issue was originally observed with AR9382. During the issue kernel log was filled with following message ath: timeout (100000 us) on reg 0xa640: 0x00000001 & 0x00000001 != 0x00000000 ath: timeout (100000 us) on reg 0xa2c4: 0x00158dd9 & 0x00000001 != 0x00000000 ath: Unable to reset channel (2412 MHz), reset status -5 ath: Unable to set channel Signed-off-by: Rajkumar Manoharan <rmanoharan@atheros.com> Signed-off-by: John W. Linville <linville@tuxdriver.com>
2011-04-25 00:04:39 +08:00
goto tx_iqcal_fail;
}
ath9k_hw: Fix Tx IQ Calibration hang issue in AR9003 chips On AR9003 chips, doing three IQ calibrations will possibly cause chip in stuck state. In noisy environment, chip could receive a packet during the middle of three calibrations and it causes the conflict of HW access and the eventual failure. It also causes IQ calibration outliers which results in poor Tx EVM. The IQ Cal procedure is after resetting the chip, run IQ cal 3 times per each cal cycle and find the two closest readings and average of two. The advantage of running Tx IQ cal more than once is that we can compare calibration results for the same gain setting over multiple iterations. Most of the cases the IQ failures were observed after first pass. For the AR9485 and later chips, Tx IQ Calibration is performed along with AGC cal. But for pre-AR9485 chips, Tx IQ cal HW has to be separated from the rest of calibration HW to avoid chip hang. After all calibrations are done in HW, we can start SW post-processing. By doing this way, we minimize the SW difference among all chips. The order of calibration (run IQ cal before other calibration) is also needed to avoid chip hang for chips before AR9485. This issue was originally observed with AR9382. During the issue kernel log was filled with following message ath: timeout (100000 us) on reg 0xa640: 0x00000001 & 0x00000001 != 0x00000000 ath: timeout (100000 us) on reg 0xa2c4: 0x00158dd9 & 0x00000001 != 0x00000000 ath: Unable to reset channel (2412 MHz), reset status -5 ath: Unable to set channel Signed-off-by: Rajkumar Manoharan <rmanoharan@atheros.com> Signed-off-by: John W. Linville <linville@tuxdriver.com>
2011-04-25 00:04:39 +08:00
for (j = 0; j < 3; j++) {
u32 idx = 2 * j, offset = 4 * (3 * im + j);
ath9k_hw: Fix Tx IQ Calibration hang issue in AR9003 chips On AR9003 chips, doing three IQ calibrations will possibly cause chip in stuck state. In noisy environment, chip could receive a packet during the middle of three calibrations and it causes the conflict of HW access and the eventual failure. It also causes IQ calibration outliers which results in poor Tx EVM. The IQ Cal procedure is after resetting the chip, run IQ cal 3 times per each cal cycle and find the two closest readings and average of two. The advantage of running Tx IQ cal more than once is that we can compare calibration results for the same gain setting over multiple iterations. Most of the cases the IQ failures were observed after first pass. For the AR9485 and later chips, Tx IQ Calibration is performed along with AGC cal. But for pre-AR9485 chips, Tx IQ cal HW has to be separated from the rest of calibration HW to avoid chip hang. After all calibrations are done in HW, we can start SW post-processing. By doing this way, we minimize the SW difference among all chips. The order of calibration (run IQ cal before other calibration) is also needed to avoid chip hang for chips before AR9485. This issue was originally observed with AR9382. During the issue kernel log was filled with following message ath: timeout (100000 us) on reg 0xa640: 0x00000001 & 0x00000001 != 0x00000000 ath: timeout (100000 us) on reg 0xa2c4: 0x00158dd9 & 0x00000001 != 0x00000000 ath: Unable to reset channel (2412 MHz), reset status -5 ath: Unable to set channel Signed-off-by: Rajkumar Manoharan <rmanoharan@atheros.com> Signed-off-by: John W. Linville <linville@tuxdriver.com>
2011-04-25 00:04:39 +08:00
REG_RMW_FIELD(ah,
AR_PHY_CHAN_INFO_MEMORY,
AR_PHY_CHAN_INFO_TAB_S2_READ,
0);
ath9k_hw: Fix Tx IQ Calibration hang issue in AR9003 chips On AR9003 chips, doing three IQ calibrations will possibly cause chip in stuck state. In noisy environment, chip could receive a packet during the middle of three calibrations and it causes the conflict of HW access and the eventual failure. It also causes IQ calibration outliers which results in poor Tx EVM. The IQ Cal procedure is after resetting the chip, run IQ cal 3 times per each cal cycle and find the two closest readings and average of two. The advantage of running Tx IQ cal more than once is that we can compare calibration results for the same gain setting over multiple iterations. Most of the cases the IQ failures were observed after first pass. For the AR9485 and later chips, Tx IQ Calibration is performed along with AGC cal. But for pre-AR9485 chips, Tx IQ cal HW has to be separated from the rest of calibration HW to avoid chip hang. After all calibrations are done in HW, we can start SW post-processing. By doing this way, we minimize the SW difference among all chips. The order of calibration (run IQ cal before other calibration) is also needed to avoid chip hang for chips before AR9485. This issue was originally observed with AR9382. During the issue kernel log was filled with following message ath: timeout (100000 us) on reg 0xa640: 0x00000001 & 0x00000001 != 0x00000000 ath: timeout (100000 us) on reg 0xa2c4: 0x00158dd9 & 0x00000001 != 0x00000000 ath: Unable to reset channel (2412 MHz), reset status -5 ath: Unable to set channel Signed-off-by: Rajkumar Manoharan <rmanoharan@atheros.com> Signed-off-by: John W. Linville <linville@tuxdriver.com>
2011-04-25 00:04:39 +08:00
/* 32 bits */
iq_res[idx] = REG_READ(ah,
chan_info_tab[i] +
offset);
ath9k_hw: Fix Tx IQ Calibration hang issue in AR9003 chips On AR9003 chips, doing three IQ calibrations will possibly cause chip in stuck state. In noisy environment, chip could receive a packet during the middle of three calibrations and it causes the conflict of HW access and the eventual failure. It also causes IQ calibration outliers which results in poor Tx EVM. The IQ Cal procedure is after resetting the chip, run IQ cal 3 times per each cal cycle and find the two closest readings and average of two. The advantage of running Tx IQ cal more than once is that we can compare calibration results for the same gain setting over multiple iterations. Most of the cases the IQ failures were observed after first pass. For the AR9485 and later chips, Tx IQ Calibration is performed along with AGC cal. But for pre-AR9485 chips, Tx IQ cal HW has to be separated from the rest of calibration HW to avoid chip hang. After all calibrations are done in HW, we can start SW post-processing. By doing this way, we minimize the SW difference among all chips. The order of calibration (run IQ cal before other calibration) is also needed to avoid chip hang for chips before AR9485. This issue was originally observed with AR9382. During the issue kernel log was filled with following message ath: timeout (100000 us) on reg 0xa640: 0x00000001 & 0x00000001 != 0x00000000 ath: timeout (100000 us) on reg 0xa2c4: 0x00158dd9 & 0x00000001 != 0x00000000 ath: Unable to reset channel (2412 MHz), reset status -5 ath: Unable to set channel Signed-off-by: Rajkumar Manoharan <rmanoharan@atheros.com> Signed-off-by: John W. Linville <linville@tuxdriver.com>
2011-04-25 00:04:39 +08:00
REG_RMW_FIELD(ah,
AR_PHY_CHAN_INFO_MEMORY,
AR_PHY_CHAN_INFO_TAB_S2_READ,
1);
ath9k_hw: Fix Tx IQ Calibration hang issue in AR9003 chips On AR9003 chips, doing three IQ calibrations will possibly cause chip in stuck state. In noisy environment, chip could receive a packet during the middle of three calibrations and it causes the conflict of HW access and the eventual failure. It also causes IQ calibration outliers which results in poor Tx EVM. The IQ Cal procedure is after resetting the chip, run IQ cal 3 times per each cal cycle and find the two closest readings and average of two. The advantage of running Tx IQ cal more than once is that we can compare calibration results for the same gain setting over multiple iterations. Most of the cases the IQ failures were observed after first pass. For the AR9485 and later chips, Tx IQ Calibration is performed along with AGC cal. But for pre-AR9485 chips, Tx IQ cal HW has to be separated from the rest of calibration HW to avoid chip hang. After all calibrations are done in HW, we can start SW post-processing. By doing this way, we minimize the SW difference among all chips. The order of calibration (run IQ cal before other calibration) is also needed to avoid chip hang for chips before AR9485. This issue was originally observed with AR9382. During the issue kernel log was filled with following message ath: timeout (100000 us) on reg 0xa640: 0x00000001 & 0x00000001 != 0x00000000 ath: timeout (100000 us) on reg 0xa2c4: 0x00158dd9 & 0x00000001 != 0x00000000 ath: Unable to reset channel (2412 MHz), reset status -5 ath: Unable to set channel Signed-off-by: Rajkumar Manoharan <rmanoharan@atheros.com> Signed-off-by: John W. Linville <linville@tuxdriver.com>
2011-04-25 00:04:39 +08:00
/* 16 bits */
iq_res[idx + 1] = 0xffff & REG_READ(ah,
chan_info_tab[i] + offset);
ath_dbg(common, CALIBRATE,
"IQ_RES[%d]=0x%x IQ_RES[%d]=0x%x\n",
ath9k_hw: Fix Tx IQ Calibration hang issue in AR9003 chips On AR9003 chips, doing three IQ calibrations will possibly cause chip in stuck state. In noisy environment, chip could receive a packet during the middle of three calibrations and it causes the conflict of HW access and the eventual failure. It also causes IQ calibration outliers which results in poor Tx EVM. The IQ Cal procedure is after resetting the chip, run IQ cal 3 times per each cal cycle and find the two closest readings and average of two. The advantage of running Tx IQ cal more than once is that we can compare calibration results for the same gain setting over multiple iterations. Most of the cases the IQ failures were observed after first pass. For the AR9485 and later chips, Tx IQ Calibration is performed along with AGC cal. But for pre-AR9485 chips, Tx IQ cal HW has to be separated from the rest of calibration HW to avoid chip hang. After all calibrations are done in HW, we can start SW post-processing. By doing this way, we minimize the SW difference among all chips. The order of calibration (run IQ cal before other calibration) is also needed to avoid chip hang for chips before AR9485. This issue was originally observed with AR9382. During the issue kernel log was filled with following message ath: timeout (100000 us) on reg 0xa640: 0x00000001 & 0x00000001 != 0x00000000 ath: timeout (100000 us) on reg 0xa2c4: 0x00158dd9 & 0x00000001 != 0x00000000 ath: Unable to reset channel (2412 MHz), reset status -5 ath: Unable to set channel Signed-off-by: Rajkumar Manoharan <rmanoharan@atheros.com> Signed-off-by: John W. Linville <linville@tuxdriver.com>
2011-04-25 00:04:39 +08:00
idx, iq_res[idx], idx + 1,
iq_res[idx + 1]);
}
ath9k_hw: Fix Tx IQ Calibration hang issue in AR9003 chips On AR9003 chips, doing three IQ calibrations will possibly cause chip in stuck state. In noisy environment, chip could receive a packet during the middle of three calibrations and it causes the conflict of HW access and the eventual failure. It also causes IQ calibration outliers which results in poor Tx EVM. The IQ Cal procedure is after resetting the chip, run IQ cal 3 times per each cal cycle and find the two closest readings and average of two. The advantage of running Tx IQ cal more than once is that we can compare calibration results for the same gain setting over multiple iterations. Most of the cases the IQ failures were observed after first pass. For the AR9485 and later chips, Tx IQ Calibration is performed along with AGC cal. But for pre-AR9485 chips, Tx IQ cal HW has to be separated from the rest of calibration HW to avoid chip hang. After all calibrations are done in HW, we can start SW post-processing. By doing this way, we minimize the SW difference among all chips. The order of calibration (run IQ cal before other calibration) is also needed to avoid chip hang for chips before AR9485. This issue was originally observed with AR9382. During the issue kernel log was filled with following message ath: timeout (100000 us) on reg 0xa640: 0x00000001 & 0x00000001 != 0x00000000 ath: timeout (100000 us) on reg 0xa2c4: 0x00158dd9 & 0x00000001 != 0x00000000 ath: Unable to reset channel (2412 MHz), reset status -5 ath: Unable to set channel Signed-off-by: Rajkumar Manoharan <rmanoharan@atheros.com> Signed-off-by: John W. Linville <linville@tuxdriver.com>
2011-04-25 00:04:39 +08:00
if (!ar9003_hw_calc_iq_corr(ah, i, iq_res,
coeff.iqc_coeff)) {
ath_dbg(common, CALIBRATE,
"Failed in calculation of IQ correction\n");
ath9k_hw: Fix Tx IQ Calibration hang issue in AR9003 chips On AR9003 chips, doing three IQ calibrations will possibly cause chip in stuck state. In noisy environment, chip could receive a packet during the middle of three calibrations and it causes the conflict of HW access and the eventual failure. It also causes IQ calibration outliers which results in poor Tx EVM. The IQ Cal procedure is after resetting the chip, run IQ cal 3 times per each cal cycle and find the two closest readings and average of two. The advantage of running Tx IQ cal more than once is that we can compare calibration results for the same gain setting over multiple iterations. Most of the cases the IQ failures were observed after first pass. For the AR9485 and later chips, Tx IQ Calibration is performed along with AGC cal. But for pre-AR9485 chips, Tx IQ cal HW has to be separated from the rest of calibration HW to avoid chip hang. After all calibrations are done in HW, we can start SW post-processing. By doing this way, we minimize the SW difference among all chips. The order of calibration (run IQ cal before other calibration) is also needed to avoid chip hang for chips before AR9485. This issue was originally observed with AR9382. During the issue kernel log was filled with following message ath: timeout (100000 us) on reg 0xa640: 0x00000001 & 0x00000001 != 0x00000000 ath: timeout (100000 us) on reg 0xa2c4: 0x00158dd9 & 0x00000001 != 0x00000000 ath: Unable to reset channel (2412 MHz), reset status -5 ath: Unable to set channel Signed-off-by: Rajkumar Manoharan <rmanoharan@atheros.com> Signed-off-by: John W. Linville <linville@tuxdriver.com>
2011-04-25 00:04:39 +08:00
goto tx_iqcal_fail;
}
ath9k_hw: Fix Tx IQ Calibration hang issue in AR9003 chips On AR9003 chips, doing three IQ calibrations will possibly cause chip in stuck state. In noisy environment, chip could receive a packet during the middle of three calibrations and it causes the conflict of HW access and the eventual failure. It also causes IQ calibration outliers which results in poor Tx EVM. The IQ Cal procedure is after resetting the chip, run IQ cal 3 times per each cal cycle and find the two closest readings and average of two. The advantage of running Tx IQ cal more than once is that we can compare calibration results for the same gain setting over multiple iterations. Most of the cases the IQ failures were observed after first pass. For the AR9485 and later chips, Tx IQ Calibration is performed along with AGC cal. But for pre-AR9485 chips, Tx IQ cal HW has to be separated from the rest of calibration HW to avoid chip hang. After all calibrations are done in HW, we can start SW post-processing. By doing this way, we minimize the SW difference among all chips. The order of calibration (run IQ cal before other calibration) is also needed to avoid chip hang for chips before AR9485. This issue was originally observed with AR9382. During the issue kernel log was filled with following message ath: timeout (100000 us) on reg 0xa640: 0x00000001 & 0x00000001 != 0x00000000 ath: timeout (100000 us) on reg 0xa2c4: 0x00158dd9 & 0x00000001 != 0x00000000 ath: Unable to reset channel (2412 MHz), reset status -5 ath: Unable to set channel Signed-off-by: Rajkumar Manoharan <rmanoharan@atheros.com> Signed-off-by: John W. Linville <linville@tuxdriver.com>
2011-04-25 00:04:39 +08:00
coeff.mag_coeff[i][im] = coeff.iqc_coeff[0] & 0x7f;
coeff.phs_coeff[i][im] =
(coeff.iqc_coeff[0] >> 7) & 0x7f;
ath9k_hw: Fix Tx IQ Calibration hang issue in AR9003 chips On AR9003 chips, doing three IQ calibrations will possibly cause chip in stuck state. In noisy environment, chip could receive a packet during the middle of three calibrations and it causes the conflict of HW access and the eventual failure. It also causes IQ calibration outliers which results in poor Tx EVM. The IQ Cal procedure is after resetting the chip, run IQ cal 3 times per each cal cycle and find the two closest readings and average of two. The advantage of running Tx IQ cal more than once is that we can compare calibration results for the same gain setting over multiple iterations. Most of the cases the IQ failures were observed after first pass. For the AR9485 and later chips, Tx IQ Calibration is performed along with AGC cal. But for pre-AR9485 chips, Tx IQ cal HW has to be separated from the rest of calibration HW to avoid chip hang. After all calibrations are done in HW, we can start SW post-processing. By doing this way, we minimize the SW difference among all chips. The order of calibration (run IQ cal before other calibration) is also needed to avoid chip hang for chips before AR9485. This issue was originally observed with AR9382. During the issue kernel log was filled with following message ath: timeout (100000 us) on reg 0xa640: 0x00000001 & 0x00000001 != 0x00000000 ath: timeout (100000 us) on reg 0xa2c4: 0x00158dd9 & 0x00000001 != 0x00000000 ath: Unable to reset channel (2412 MHz), reset status -5 ath: Unable to set channel Signed-off-by: Rajkumar Manoharan <rmanoharan@atheros.com> Signed-off-by: John W. Linville <linville@tuxdriver.com>
2011-04-25 00:04:39 +08:00
if (coeff.mag_coeff[i][im] > 63)
coeff.mag_coeff[i][im] -= 128;
if (coeff.phs_coeff[i][im] > 63)
coeff.phs_coeff[i][im] -= 128;
}
}
ar9003_hw_tx_iqcal_load_avg_2_passes(ah, num_chains,
&coeff, is_reusable);
return;
tx_iqcal_fail:
ath_dbg(common, CALIBRATE, "Tx IQ Cal failed\n");
return;
}
static void ar9003_hw_tx_iq_cal_reload(struct ath_hw *ah)
{
struct ath9k_hw_cal_data *caldata = ah->caldata;
u32 tx_corr_coeff[MAX_MEASUREMENT][AR9300_MAX_CHAINS];
int i, im;
memset(tx_corr_coeff, 0, sizeof(tx_corr_coeff));
for (i = 0; i < MAX_MEASUREMENT / 2; i++) {
tx_corr_coeff[i * 2][0] = tx_corr_coeff[(i * 2) + 1][0] =
AR_PHY_TX_IQCAL_CORR_COEFF_B0(i);
if (!AR_SREV_9485(ah)) {
tx_corr_coeff[i * 2][1] =
tx_corr_coeff[(i * 2) + 1][1] =
AR_PHY_TX_IQCAL_CORR_COEFF_B1(i);
tx_corr_coeff[i * 2][2] =
tx_corr_coeff[(i * 2) + 1][2] =
AR_PHY_TX_IQCAL_CORR_COEFF_B2(i);
}
}
for (i = 0; i < AR9300_MAX_CHAINS; i++) {
if (!(ah->txchainmask & (1 << i)))
continue;
for (im = 0; im < caldata->num_measures[i]; im++) {
if ((im % 2) == 0)
REG_RMW_FIELD(ah, tx_corr_coeff[im][i],
AR_PHY_TX_IQCAL_CORR_COEFF_00_COEFF_TABLE,
caldata->tx_corr_coeff[im][i]);
else
REG_RMW_FIELD(ah, tx_corr_coeff[im][i],
AR_PHY_TX_IQCAL_CORR_COEFF_01_COEFF_TABLE,
caldata->tx_corr_coeff[im][i]);
}
}
REG_RMW_FIELD(ah, AR_PHY_TX_IQCAL_CONTROL_3,
AR_PHY_TX_IQCAL_CONTROL_3_IQCORR_EN, 0x1);
REG_RMW_FIELD(ah, AR_PHY_RX_IQCAL_CORR_B0,
AR_PHY_RX_IQCAL_CORR_B0_LOOPBACK_IQCORR_EN, 0x1);
}
static bool ar9003_hw_rtt_restore(struct ath_hw *ah, struct ath9k_channel *chan)
{
struct ath9k_rtt_hist *hist;
u32 *table;
int i;
bool restore;
if (!ah->caldata)
return false;
hist = &ah->caldata->rtt_hist;
if (!hist->num_readings)
return false;
ar9003_hw_rtt_enable(ah);
ar9003_hw_rtt_set_mask(ah, 0x00);
for (i = 0; i < AR9300_MAX_CHAINS; i++) {
if (!(ah->rxchainmask & (1 << i)))
continue;
table = &hist->table[i][hist->num_readings][0];
ar9003_hw_rtt_load_hist(ah, i, table);
}
restore = ar9003_hw_rtt_force_restore(ah);
ar9003_hw_rtt_disable(ah);
return restore;
}
static bool ar9003_hw_init_cal(struct ath_hw *ah,
struct ath9k_channel *chan)
{
struct ath_common *common = ath9k_hw_common(ah);
struct ath9k_hw_cal_data *caldata = ah->caldata;
bool txiqcal_done = false, txclcal_done = false;
bool is_reusable = true, status = true;
bool run_rtt_cal = false, run_agc_cal;
bool rtt = !!(ah->caps.hw_caps & ATH9K_HW_CAP_RTT);
bool mci = !!(ah->caps.hw_caps & ATH9K_HW_CAP_MCI);
u32 agc_ctrl = 0, agc_supp_cals = AR_PHY_AGC_CONTROL_OFFSET_CAL |
AR_PHY_AGC_CONTROL_FLTR_CAL |
AR_PHY_AGC_CONTROL_PKDET_CAL;
int i, j;
u32 cl_idx[AR9300_MAX_CHAINS] = { AR_PHY_CL_TAB_0,
AR_PHY_CL_TAB_1,
AR_PHY_CL_TAB_2 };
if (rtt) {
if (!ar9003_hw_rtt_restore(ah, chan))
run_rtt_cal = true;
ath_dbg(common, CALIBRATE, "RTT restore %s\n",
run_rtt_cal ? "failed" : "succeed");
}
run_agc_cal = run_rtt_cal;
if (run_rtt_cal) {
ar9003_hw_rtt_enable(ah);
ar9003_hw_rtt_set_mask(ah, 0x00);
ar9003_hw_rtt_clear_hist(ah);
}
if (rtt && !run_rtt_cal) {
agc_ctrl = REG_READ(ah, AR_PHY_AGC_CONTROL);
agc_supp_cals &= agc_ctrl;
agc_ctrl &= ~(AR_PHY_AGC_CONTROL_OFFSET_CAL |
AR_PHY_AGC_CONTROL_FLTR_CAL |
AR_PHY_AGC_CONTROL_PKDET_CAL);
REG_WRITE(ah, AR_PHY_AGC_CONTROL, agc_ctrl);
}
if (ah->enabled_cals & TX_CL_CAL) {
if (caldata && caldata->done_txclcal_once)
REG_CLR_BIT(ah, AR_PHY_CL_CAL_CTL,
AR_PHY_CL_CAL_ENABLE);
else {
REG_SET_BIT(ah, AR_PHY_CL_CAL_CTL,
AR_PHY_CL_CAL_ENABLE);
run_agc_cal = true;
}
}
if (!(ah->enabled_cals & TX_IQ_CAL))
goto skip_tx_iqcal;
/* Do Tx IQ Calibration */
ath9k_hw: Fix Tx IQ Calibration hang issue in AR9003 chips On AR9003 chips, doing three IQ calibrations will possibly cause chip in stuck state. In noisy environment, chip could receive a packet during the middle of three calibrations and it causes the conflict of HW access and the eventual failure. It also causes IQ calibration outliers which results in poor Tx EVM. The IQ Cal procedure is after resetting the chip, run IQ cal 3 times per each cal cycle and find the two closest readings and average of two. The advantage of running Tx IQ cal more than once is that we can compare calibration results for the same gain setting over multiple iterations. Most of the cases the IQ failures were observed after first pass. For the AR9485 and later chips, Tx IQ Calibration is performed along with AGC cal. But for pre-AR9485 chips, Tx IQ cal HW has to be separated from the rest of calibration HW to avoid chip hang. After all calibrations are done in HW, we can start SW post-processing. By doing this way, we minimize the SW difference among all chips. The order of calibration (run IQ cal before other calibration) is also needed to avoid chip hang for chips before AR9485. This issue was originally observed with AR9382. During the issue kernel log was filled with following message ath: timeout (100000 us) on reg 0xa640: 0x00000001 & 0x00000001 != 0x00000000 ath: timeout (100000 us) on reg 0xa2c4: 0x00158dd9 & 0x00000001 != 0x00000000 ath: Unable to reset channel (2412 MHz), reset status -5 ath: Unable to set channel Signed-off-by: Rajkumar Manoharan <rmanoharan@atheros.com> Signed-off-by: John W. Linville <linville@tuxdriver.com>
2011-04-25 00:04:39 +08:00
REG_RMW_FIELD(ah, AR_PHY_TX_IQCAL_CONTROL_1,
AR_PHY_TX_IQCAL_CONTROL_1_IQCORR_I_Q_COFF_DELPT,
DELPT);
ath9k_hw: Fix Tx IQ Calibration hang issue in AR9003 chips On AR9003 chips, doing three IQ calibrations will possibly cause chip in stuck state. In noisy environment, chip could receive a packet during the middle of three calibrations and it causes the conflict of HW access and the eventual failure. It also causes IQ calibration outliers which results in poor Tx EVM. The IQ Cal procedure is after resetting the chip, run IQ cal 3 times per each cal cycle and find the two closest readings and average of two. The advantage of running Tx IQ cal more than once is that we can compare calibration results for the same gain setting over multiple iterations. Most of the cases the IQ failures were observed after first pass. For the AR9485 and later chips, Tx IQ Calibration is performed along with AGC cal. But for pre-AR9485 chips, Tx IQ cal HW has to be separated from the rest of calibration HW to avoid chip hang. After all calibrations are done in HW, we can start SW post-processing. By doing this way, we minimize the SW difference among all chips. The order of calibration (run IQ cal before other calibration) is also needed to avoid chip hang for chips before AR9485. This issue was originally observed with AR9382. During the issue kernel log was filled with following message ath: timeout (100000 us) on reg 0xa640: 0x00000001 & 0x00000001 != 0x00000000 ath: timeout (100000 us) on reg 0xa2c4: 0x00158dd9 & 0x00000001 != 0x00000000 ath: Unable to reset channel (2412 MHz), reset status -5 ath: Unable to set channel Signed-off-by: Rajkumar Manoharan <rmanoharan@atheros.com> Signed-off-by: John W. Linville <linville@tuxdriver.com>
2011-04-25 00:04:39 +08:00
/*
* For AR9485 or later chips, TxIQ cal runs as part of
* AGC calibration
*/
if (ah->enabled_cals & TX_IQ_ON_AGC_CAL) {
if (caldata && !caldata->done_txiqcal_once)
REG_SET_BIT(ah, AR_PHY_TX_IQCAL_CONTROL_0,
AR_PHY_TX_IQCAL_CONTROL_0_ENABLE_TXIQ_CAL);
else
REG_CLR_BIT(ah, AR_PHY_TX_IQCAL_CONTROL_0,
AR_PHY_TX_IQCAL_CONTROL_0_ENABLE_TXIQ_CAL);
txiqcal_done = run_agc_cal = true;
goto skip_tx_iqcal;
} else if (caldata && !caldata->done_txiqcal_once)
run_agc_cal = true;
if (mci && IS_CHAN_2GHZ(chan) && run_agc_cal)
ar9003_mci_init_cal_req(ah, &is_reusable);
txiqcal_done = ar9003_hw_tx_iq_cal_run(ah);
REG_WRITE(ah, AR_PHY_ACTIVE, AR_PHY_ACTIVE_DIS);
udelay(5);
REG_WRITE(ah, AR_PHY_ACTIVE, AR_PHY_ACTIVE_EN);
skip_tx_iqcal:
if (run_agc_cal || !(ah->ah_flags & AH_FASTCC)) {
/* Calibrate the AGC */
REG_WRITE(ah, AR_PHY_AGC_CONTROL,
REG_READ(ah, AR_PHY_AGC_CONTROL) |
AR_PHY_AGC_CONTROL_CAL);
/* Poll for offset calibration complete */
status = ath9k_hw_wait(ah, AR_PHY_AGC_CONTROL,
AR_PHY_AGC_CONTROL_CAL,
0, AH_WAIT_TIMEOUT);
}
if (mci && IS_CHAN_2GHZ(chan) && run_agc_cal)
ar9003_mci_init_cal_done(ah);
if (rtt && !run_rtt_cal) {
agc_ctrl |= agc_supp_cals;
REG_WRITE(ah, AR_PHY_AGC_CONTROL, agc_ctrl);
}
if (!status) {
if (run_rtt_cal)
ar9003_hw_rtt_disable(ah);
ath_dbg(common, CALIBRATE,
"offset calibration failed to complete in 1ms; noisy environment?\n");
return false;
}
ath9k_hw: Fix Tx IQ Calibration hang issue in AR9003 chips On AR9003 chips, doing three IQ calibrations will possibly cause chip in stuck state. In noisy environment, chip could receive a packet during the middle of three calibrations and it causes the conflict of HW access and the eventual failure. It also causes IQ calibration outliers which results in poor Tx EVM. The IQ Cal procedure is after resetting the chip, run IQ cal 3 times per each cal cycle and find the two closest readings and average of two. The advantage of running Tx IQ cal more than once is that we can compare calibration results for the same gain setting over multiple iterations. Most of the cases the IQ failures were observed after first pass. For the AR9485 and later chips, Tx IQ Calibration is performed along with AGC cal. But for pre-AR9485 chips, Tx IQ cal HW has to be separated from the rest of calibration HW to avoid chip hang. After all calibrations are done in HW, we can start SW post-processing. By doing this way, we minimize the SW difference among all chips. The order of calibration (run IQ cal before other calibration) is also needed to avoid chip hang for chips before AR9485. This issue was originally observed with AR9382. During the issue kernel log was filled with following message ath: timeout (100000 us) on reg 0xa640: 0x00000001 & 0x00000001 != 0x00000000 ath: timeout (100000 us) on reg 0xa2c4: 0x00158dd9 & 0x00000001 != 0x00000000 ath: Unable to reset channel (2412 MHz), reset status -5 ath: Unable to set channel Signed-off-by: Rajkumar Manoharan <rmanoharan@atheros.com> Signed-off-by: John W. Linville <linville@tuxdriver.com>
2011-04-25 00:04:39 +08:00
if (txiqcal_done)
ar9003_hw_tx_iq_cal_post_proc(ah, is_reusable);
else if (caldata && caldata->done_txiqcal_once)
ar9003_hw_tx_iq_cal_reload(ah);
#define CL_TAB_ENTRY(reg_base) (reg_base + (4 * j))
if (caldata && (ah->enabled_cals & TX_CL_CAL)) {
txclcal_done = !!(REG_READ(ah, AR_PHY_AGC_CONTROL) &
AR_PHY_AGC_CONTROL_CLC_SUCCESS);
if (caldata->done_txclcal_once) {
for (i = 0; i < AR9300_MAX_CHAINS; i++) {
if (!(ah->txchainmask & (1 << i)))
continue;
for (j = 0; j < MAX_CL_TAB_ENTRY; j++)
REG_WRITE(ah, CL_TAB_ENTRY(cl_idx[i]),
caldata->tx_clcal[i][j]);
}
} else if (is_reusable && txclcal_done) {
for (i = 0; i < AR9300_MAX_CHAINS; i++) {
if (!(ah->txchainmask & (1 << i)))
continue;
for (j = 0; j < MAX_CL_TAB_ENTRY; j++)
caldata->tx_clcal[i][j] =
REG_READ(ah,
CL_TAB_ENTRY(cl_idx[i]));
}
caldata->done_txclcal_once = true;
}
}
#undef CL_TAB_ENTRY
if (run_rtt_cal && caldata) {
struct ath9k_rtt_hist *hist = &caldata->rtt_hist;
if (is_reusable && (hist->num_readings < RTT_HIST_MAX)) {
u32 *table;
hist->num_readings++;
for (i = 0; i < AR9300_MAX_CHAINS; i++) {
if (!(ah->rxchainmask & (1 << i)))
continue;
table = &hist->table[i][hist->num_readings][0];
ar9003_hw_rtt_fill_hist(ah, i, table);
}
}
ar9003_hw_rtt_disable(ah);
}
/* Initialize list pointers */
ah->cal_list = ah->cal_list_last = ah->cal_list_curr = NULL;
ah->supp_cals = IQ_MISMATCH_CAL;
if (ah->supp_cals & IQ_MISMATCH_CAL) {
INIT_CAL(&ah->iq_caldata);
INSERT_CAL(ah, &ah->iq_caldata);
ath_dbg(common, CALIBRATE, "enabling IQ Calibration\n");
}
if (ah->supp_cals & TEMP_COMP_CAL) {
INIT_CAL(&ah->tempCompCalData);
INSERT_CAL(ah, &ah->tempCompCalData);
ath_dbg(common, CALIBRATE,
"enabling Temperature Compensation Calibration\n");
}
/* Initialize current pointer to first element in list */
ah->cal_list_curr = ah->cal_list;
if (ah->cal_list_curr)
ath9k_hw_reset_calibration(ah, ah->cal_list_curr);
if (caldata)
caldata->CalValid = 0;
return true;
}
void ar9003_hw_attach_calib_ops(struct ath_hw *ah)
{
struct ath_hw_private_ops *priv_ops = ath9k_hw_private_ops(ah);
struct ath_hw_ops *ops = ath9k_hw_ops(ah);
priv_ops->init_cal_settings = ar9003_hw_init_cal_settings;
priv_ops->init_cal = ar9003_hw_init_cal;
priv_ops->setup_calibration = ar9003_hw_setup_calibration;
ops->calibrate = ar9003_hw_calibrate;
}