openwrtv3/package/kernel/mac80211/patches/020-19-rt2x00-add-support-for-MT7620.patch

2078 lines
73 KiB
Diff
Raw Normal View History

From 41977e86c984fcdddb454a3d7887de5d47b5f530 Mon Sep 17 00:00:00 2001
From: Roman Yeryomin <roman@advem.lv>
Date: Tue, 21 Mar 2017 00:43:00 +0100
Subject: [PATCH 19/19] rt2x00: add support for MT7620
rt2x00: mt7620: lots of improvements This commit combines all the changes I've made on my staging tree into a single commit fixing many issues with our patch for MT7620. First of all, checkpatch.pl revealed numerous code style issues with the patch, so fix all the white-space and commets. Also use usleep_range instead of legacy timing and relax timing for VCO calibration just like the vendor driver does. Several line programming registers were commented out in the patch. Originally this came from the features present but disabled by default in the vendor's driver (RTMP_TEMPERATURE_CALIBRATION and ADJUST_POWER_CONSUMPTION_SUPPORT). Remove the dead code for now, it can easily be re-added if we actually intend to support those features. Move values from mt7620_freqconfig type into the existing rf_channel struct, this shouldn't be a new typedef and it is possible to use the existing struct because rf_channel got 4 32-bit fields, so two of the 8-bit values from mt7620_freqconfig can easily be stored in the same 32-bit field. Map values such that Rdiv -> rf1 N -> rf2 K -> rf3[0:7] D -> rf3[8:15] Ksd -> rf4 This makes the channel switching logic already look a bit more like what we are used to in rt2x00... Probably many of the read-modify-write calls could still be replaced by macros intended for that. iq calibration seems to be identical to RT5592, so just enable it. Test shows that this improves things quite a lot, datarates went up by a couple of megabits when running iperf, signal quality seems jumpy in the first few seconds once a station connencts, the stabelizes on a value significantly better than what it was before. Add description to the patch and reference the original OpenWrt commit by which it was added. The patch now passes checkpatch.pl and can thus be discussed with the upstream authors of the rt2x00 driver. Funded-by: https://www.kickstarter.com/projects/1327597961/better-support-for-mt7620a-n-in-openwrt-lede/ Signed-off-by: Daniel Golle <daniel@makrotopia.org>
2017-02-03 14:02:32 +00:00
Basic support for MT7620 built-in wireless radio was added to
OpenWrt in r41441. It has seen some heavy cleaning and refactoring
rt2x00: mt7620: lots of improvements This commit combines all the changes I've made on my staging tree into a single commit fixing many issues with our patch for MT7620. First of all, checkpatch.pl revealed numerous code style issues with the patch, so fix all the white-space and commets. Also use usleep_range instead of legacy timing and relax timing for VCO calibration just like the vendor driver does. Several line programming registers were commented out in the patch. Originally this came from the features present but disabled by default in the vendor's driver (RTMP_TEMPERATURE_CALIBRATION and ADJUST_POWER_CONSUMPTION_SUPPORT). Remove the dead code for now, it can easily be re-added if we actually intend to support those features. Move values from mt7620_freqconfig type into the existing rf_channel struct, this shouldn't be a new typedef and it is possible to use the existing struct because rf_channel got 4 32-bit fields, so two of the 8-bit values from mt7620_freqconfig can easily be stored in the same 32-bit field. Map values such that Rdiv -> rf1 N -> rf2 K -> rf3[0:7] D -> rf3[8:15] Ksd -> rf4 This makes the channel switching logic already look a bit more like what we are used to in rt2x00... Probably many of the read-modify-write calls could still be replaced by macros intended for that. iq calibration seems to be identical to RT5592, so just enable it. Test shows that this improves things quite a lot, datarates went up by a couple of megabits when running iperf, signal quality seems jumpy in the first few seconds once a station connencts, the stabelizes on a value significantly better than what it was before. Add description to the patch and reference the original OpenWrt commit by which it was added. The patch now passes checkpatch.pl and can thus be discussed with the upstream authors of the rt2x00 driver. Funded-by: https://www.kickstarter.com/projects/1327597961/better-support-for-mt7620a-n-in-openwrt-lede/ Signed-off-by: Daniel Golle <daniel@makrotopia.org>
2017-02-03 14:02:32 +00:00
since in order to match the Kernel's code quality standards.
Signed-off-by: Roman Yeryomin <roman@advem.lv>
Signed-off-by: Daniel Golle <daniel@makrotopia.org>
Acked-by: Stanislaw Gruszka <sgruszka@redhat.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
rt2x00: mt7620: lots of improvements This commit combines all the changes I've made on my staging tree into a single commit fixing many issues with our patch for MT7620. First of all, checkpatch.pl revealed numerous code style issues with the patch, so fix all the white-space and commets. Also use usleep_range instead of legacy timing and relax timing for VCO calibration just like the vendor driver does. Several line programming registers were commented out in the patch. Originally this came from the features present but disabled by default in the vendor's driver (RTMP_TEMPERATURE_CALIBRATION and ADJUST_POWER_CONSUMPTION_SUPPORT). Remove the dead code for now, it can easily be re-added if we actually intend to support those features. Move values from mt7620_freqconfig type into the existing rf_channel struct, this shouldn't be a new typedef and it is possible to use the existing struct because rf_channel got 4 32-bit fields, so two of the 8-bit values from mt7620_freqconfig can easily be stored in the same 32-bit field. Map values such that Rdiv -> rf1 N -> rf2 K -> rf3[0:7] D -> rf3[8:15] Ksd -> rf4 This makes the channel switching logic already look a bit more like what we are used to in rt2x00... Probably many of the read-modify-write calls could still be replaced by macros intended for that. iq calibration seems to be identical to RT5592, so just enable it. Test shows that this improves things quite a lot, datarates went up by a couple of megabits when running iperf, signal quality seems jumpy in the first few seconds once a station connencts, the stabelizes on a value significantly better than what it was before. Add description to the patch and reference the original OpenWrt commit by which it was added. The patch now passes checkpatch.pl and can thus be discussed with the upstream authors of the rt2x00 driver. Funded-by: https://www.kickstarter.com/projects/1327597961/better-support-for-mt7620a-n-in-openwrt-lede/ Signed-off-by: Daniel Golle <daniel@makrotopia.org>
2017-02-03 14:02:32 +00:00
---
drivers/net/wireless/ralink/rt2x00/Kconfig | 2 +-
drivers/net/wireless/ralink/rt2x00/rt2800.h | 177 +++
drivers/net/wireless/ralink/rt2x00/rt2800lib.c | 1421 +++++++++++++++++++++++-
drivers/net/wireless/ralink/rt2x00/rt2800lib.h | 4 +
drivers/net/wireless/ralink/rt2x00/rt2x00.h | 1 +
5 files changed, 1578 insertions(+), 27 deletions(-)
rt2x00: mt7620: lots of improvements This commit combines all the changes I've made on my staging tree into a single commit fixing many issues with our patch for MT7620. First of all, checkpatch.pl revealed numerous code style issues with the patch, so fix all the white-space and commets. Also use usleep_range instead of legacy timing and relax timing for VCO calibration just like the vendor driver does. Several line programming registers were commented out in the patch. Originally this came from the features present but disabled by default in the vendor's driver (RTMP_TEMPERATURE_CALIBRATION and ADJUST_POWER_CONSUMPTION_SUPPORT). Remove the dead code for now, it can easily be re-added if we actually intend to support those features. Move values from mt7620_freqconfig type into the existing rf_channel struct, this shouldn't be a new typedef and it is possible to use the existing struct because rf_channel got 4 32-bit fields, so two of the 8-bit values from mt7620_freqconfig can easily be stored in the same 32-bit field. Map values such that Rdiv -> rf1 N -> rf2 K -> rf3[0:7] D -> rf3[8:15] Ksd -> rf4 This makes the channel switching logic already look a bit more like what we are used to in rt2x00... Probably many of the read-modify-write calls could still be replaced by macros intended for that. iq calibration seems to be identical to RT5592, so just enable it. Test shows that this improves things quite a lot, datarates went up by a couple of megabits when running iperf, signal quality seems jumpy in the first few seconds once a station connencts, the stabelizes on a value significantly better than what it was before. Add description to the patch and reference the original OpenWrt commit by which it was added. The patch now passes checkpatch.pl and can thus be discussed with the upstream authors of the rt2x00 driver. Funded-by: https://www.kickstarter.com/projects/1327597961/better-support-for-mt7620a-n-in-openwrt-lede/ Signed-off-by: Daniel Golle <daniel@makrotopia.org>
2017-02-03 14:02:32 +00:00
diff --git a/drivers/net/wireless/ralink/rt2x00/Kconfig b/drivers/net/wireless/ralink/rt2x00/Kconfig
index de62f5dcb62f..a1d1cfe214d2 100644
--- a/drivers/net/wireless/ralink/rt2x00/Kconfig
+++ b/drivers/net/wireless/ralink/rt2x00/Kconfig
@@ -210,7 +210,7 @@ endif
config RT2800SOC
tristate "Ralink WiSoC support"
depends on m
- depends on SOC_RT288X || SOC_RT305X
+ depends on SOC_RT288X || SOC_RT305X || SOC_MT7620
select RT2X00_LIB_SOC
select RT2X00_LIB_MMIO
select RT2X00_LIB_CRYPTO
diff --git a/drivers/net/wireless/ralink/rt2x00/rt2800.h b/drivers/net/wireless/ralink/rt2x00/rt2800.h
index fd1dbd956bad..6a8c93fb6a43 100644
--- a/drivers/net/wireless/ralink/rt2x00/rt2800.h
+++ b/drivers/net/wireless/ralink/rt2x00/rt2800.h
@@ -79,6 +79,7 @@
#define RF5372 0x5372
#define RF5390 0x5390
#define RF5392 0x5392
+#define RF7620 0x7620
/*
* Chipset revisions.
@@ -639,6 +640,24 @@
#define RF_CSR_CFG_BUSY FIELD32(0x00020000)
/*
rt2x00: mt7620: yet another beauty session So here is another round of improvements for MT7620 WiFi. This commit fixes a few significant issues related to TX_PWR_CFG_x and TX_ALC and also makes the code more readable by adding register descriptions for things added for MT7620 and use the usual bit-field access macros and the now defined macros instead of plain bit-ops and magic numbers. Properly describe EEPROM_TARGET_POWER at word 0x68 (== byte 0xD0) and thereby fix internal TXALC which would otherwise just read out-of-bounds of the EEPROM map. Split-out tx-power/ALC related stuff into an additional function. Fix VCO calibration, it was carried out properly in the channel switching but incomplete in the actual VCO calibration function. Also there is no need to trigger VCO calibration in channel switching, the VCO calibration function is already being called at this point. Remove it from channel switching function to avoid redundant code. The TX power calibration differs significantly from all other Mediatek/Ralink chips: They finally allow 0.5dB steps stored as 8-bit values for (almost) each bitrate -- and promptly ran out of space and for some reason didn't want to change the EEPROM layout. The hence opted for a scheme of sharing values for some adjecent bitrates and a highly over-complicated (or obfuscated?) way to populate the TX_PWR_CFG_x registers with the values stored in the EEPROM. The code here now looks much less complicated than what you see in the vendor's driver, however, it does the exact same thing: bGpwrdeltaMinus is a constant and always TRUE, hence half of the code was dead. Gpwrdelta is always 0 (rather than using the value read from the EEPROM). What remains is some very grotesque effort to avoid 0x20, probably some hardware bug related to some misunderstanding of what a singed 8-bit value is (imagine: if it was a signed 6-bit value then someone could believe that 0x20 == 0x0). And then they didn't clean it up once they later on anandonned that whole story of having a constant offset for 40 MHz channels and just set the offset to be constant 0 -- there is no effort for avoiding 0x20 for the 20 MHz values stored in the EEPROM, hence that's probably just a forbidden value in the EEPROM specs and won't appear anyway... Anyway, the whole thing felt like solving some college math test where in the end everything cancels out and the result equals 0 ;) To make sure that channel bandwidth power compensation really doesn't need to be taken care of, output a warning when the corresponding value stored in the EEPROM is non-zero. Also there is no apparent reason to refrain from initializing RFCSR register 13, it doesn't fail what-so-ever. Signed-off-by: Daniel Golle <daniel@makrotopia.org>
2017-02-13 05:25:35 +00:00
+ * MT7620 RF registers (reversed order)
+ */
+#define RF_CSR_CFG_DATA_MT7620 FIELD32(0x0000ff00)
+#define RF_CSR_CFG_REGNUM_MT7620 FIELD32(0x03ff0000)
+#define RF_CSR_CFG_WRITE_MT7620 FIELD32(0x00000010)
+#define RF_CSR_CFG_BUSY_MT7620 FIELD32(0x00000001)
+
+/* undocumented registers for calibration of new MAC */
+#define RF_CONTROL0 0x0518
+#define RF_BYPASS0 0x051c
+#define RF_CONTROL1 0x0520
+#define RF_BYPASS1 0x0524
+#define RF_CONTROL2 0x0528
+#define RF_BYPASS2 0x052c
+#define RF_CONTROL3 0x0530
+#define RF_BYPASS3 0x0534
+
+/*
* EFUSE_CSR: RT30x0 EEPROM
*/
#define EFUSE_CTRL 0x0580
@@ -1022,6 +1041,16 @@
#define AUTOWAKEUP_CFG_AUTOWAKE FIELD32(0x00008000)
/*
rt2x00: mt7620: yet another beauty session So here is another round of improvements for MT7620 WiFi. This commit fixes a few significant issues related to TX_PWR_CFG_x and TX_ALC and also makes the code more readable by adding register descriptions for things added for MT7620 and use the usual bit-field access macros and the now defined macros instead of plain bit-ops and magic numbers. Properly describe EEPROM_TARGET_POWER at word 0x68 (== byte 0xD0) and thereby fix internal TXALC which would otherwise just read out-of-bounds of the EEPROM map. Split-out tx-power/ALC related stuff into an additional function. Fix VCO calibration, it was carried out properly in the channel switching but incomplete in the actual VCO calibration function. Also there is no need to trigger VCO calibration in channel switching, the VCO calibration function is already being called at this point. Remove it from channel switching function to avoid redundant code. The TX power calibration differs significantly from all other Mediatek/Ralink chips: They finally allow 0.5dB steps stored as 8-bit values for (almost) each bitrate -- and promptly ran out of space and for some reason didn't want to change the EEPROM layout. The hence opted for a scheme of sharing values for some adjecent bitrates and a highly over-complicated (or obfuscated?) way to populate the TX_PWR_CFG_x registers with the values stored in the EEPROM. The code here now looks much less complicated than what you see in the vendor's driver, however, it does the exact same thing: bGpwrdeltaMinus is a constant and always TRUE, hence half of the code was dead. Gpwrdelta is always 0 (rather than using the value read from the EEPROM). What remains is some very grotesque effort to avoid 0x20, probably some hardware bug related to some misunderstanding of what a singed 8-bit value is (imagine: if it was a signed 6-bit value then someone could believe that 0x20 == 0x0). And then they didn't clean it up once they later on anandonned that whole story of having a constant offset for 40 MHz channels and just set the offset to be constant 0 -- there is no effort for avoiding 0x20 for the 20 MHz values stored in the EEPROM, hence that's probably just a forbidden value in the EEPROM specs and won't appear anyway... Anyway, the whole thing felt like solving some college math test where in the end everything cancels out and the result equals 0 ;) To make sure that channel bandwidth power compensation really doesn't need to be taken care of, output a warning when the corresponding value stored in the EEPROM is non-zero. Also there is no apparent reason to refrain from initializing RFCSR register 13, it doesn't fail what-so-ever. Signed-off-by: Daniel Golle <daniel@makrotopia.org>
2017-02-13 05:25:35 +00:00
+ * MIMO_PS_CFG: MIMO Power-save Configuration
+ */
+#define MIMO_PS_CFG 0x1210
rt2x00: mt7620: yet another beauty session So here is another round of improvements for MT7620 WiFi. This commit fixes a few significant issues related to TX_PWR_CFG_x and TX_ALC and also makes the code more readable by adding register descriptions for things added for MT7620 and use the usual bit-field access macros and the now defined macros instead of plain bit-ops and magic numbers. Properly describe EEPROM_TARGET_POWER at word 0x68 (== byte 0xD0) and thereby fix internal TXALC which would otherwise just read out-of-bounds of the EEPROM map. Split-out tx-power/ALC related stuff into an additional function. Fix VCO calibration, it was carried out properly in the channel switching but incomplete in the actual VCO calibration function. Also there is no need to trigger VCO calibration in channel switching, the VCO calibration function is already being called at this point. Remove it from channel switching function to avoid redundant code. The TX power calibration differs significantly from all other Mediatek/Ralink chips: They finally allow 0.5dB steps stored as 8-bit values for (almost) each bitrate -- and promptly ran out of space and for some reason didn't want to change the EEPROM layout. The hence opted for a scheme of sharing values for some adjecent bitrates and a highly over-complicated (or obfuscated?) way to populate the TX_PWR_CFG_x registers with the values stored in the EEPROM. The code here now looks much less complicated than what you see in the vendor's driver, however, it does the exact same thing: bGpwrdeltaMinus is a constant and always TRUE, hence half of the code was dead. Gpwrdelta is always 0 (rather than using the value read from the EEPROM). What remains is some very grotesque effort to avoid 0x20, probably some hardware bug related to some misunderstanding of what a singed 8-bit value is (imagine: if it was a signed 6-bit value then someone could believe that 0x20 == 0x0). And then they didn't clean it up once they later on anandonned that whole story of having a constant offset for 40 MHz channels and just set the offset to be constant 0 -- there is no effort for avoiding 0x20 for the 20 MHz values stored in the EEPROM, hence that's probably just a forbidden value in the EEPROM specs and won't appear anyway... Anyway, the whole thing felt like solving some college math test where in the end everything cancels out and the result equals 0 ;) To make sure that channel bandwidth power compensation really doesn't need to be taken care of, output a warning when the corresponding value stored in the EEPROM is non-zero. Also there is no apparent reason to refrain from initializing RFCSR register 13, it doesn't fail what-so-ever. Signed-off-by: Daniel Golle <daniel@makrotopia.org>
2017-02-13 05:25:35 +00:00
+#define MIMO_PS_CFG_MMPS_BB_EN FIELD32(0x00000001)
+#define MIMO_PS_CFG_MMPS_RX_ANT_NUM FIELD32(0x00000006)
+#define MIMO_PS_CFG_MMPS_RF_EN FIELD32(0x00000008)
+#define MIMO_PS_CFG_RX_STBY_POL FIELD32(0x00000010)
+#define MIMO_PS_CFG_RX_RX_STBY0 FIELD32(0x00000020)
+
+/*
* EDCA_AC0_CFG:
*/
#define EDCA_AC0_CFG 0x1300
@@ -1095,6 +1124,12 @@
rt2x00: mt7620: yet another beauty session So here is another round of improvements for MT7620 WiFi. This commit fixes a few significant issues related to TX_PWR_CFG_x and TX_ALC and also makes the code more readable by adding register descriptions for things added for MT7620 and use the usual bit-field access macros and the now defined macros instead of plain bit-ops and magic numbers. Properly describe EEPROM_TARGET_POWER at word 0x68 (== byte 0xD0) and thereby fix internal TXALC which would otherwise just read out-of-bounds of the EEPROM map. Split-out tx-power/ALC related stuff into an additional function. Fix VCO calibration, it was carried out properly in the channel switching but incomplete in the actual VCO calibration function. Also there is no need to trigger VCO calibration in channel switching, the VCO calibration function is already being called at this point. Remove it from channel switching function to avoid redundant code. The TX power calibration differs significantly from all other Mediatek/Ralink chips: They finally allow 0.5dB steps stored as 8-bit values for (almost) each bitrate -- and promptly ran out of space and for some reason didn't want to change the EEPROM layout. The hence opted for a scheme of sharing values for some adjecent bitrates and a highly over-complicated (or obfuscated?) way to populate the TX_PWR_CFG_x registers with the values stored in the EEPROM. The code here now looks much less complicated than what you see in the vendor's driver, however, it does the exact same thing: bGpwrdeltaMinus is a constant and always TRUE, hence half of the code was dead. Gpwrdelta is always 0 (rather than using the value read from the EEPROM). What remains is some very grotesque effort to avoid 0x20, probably some hardware bug related to some misunderstanding of what a singed 8-bit value is (imagine: if it was a signed 6-bit value then someone could believe that 0x20 == 0x0). And then they didn't clean it up once they later on anandonned that whole story of having a constant offset for 40 MHz channels and just set the offset to be constant 0 -- there is no effort for avoiding 0x20 for the 20 MHz values stored in the EEPROM, hence that's probably just a forbidden value in the EEPROM specs and won't appear anyway... Anyway, the whole thing felt like solving some college math test where in the end everything cancels out and the result equals 0 ;) To make sure that channel bandwidth power compensation really doesn't need to be taken care of, output a warning when the corresponding value stored in the EEPROM is non-zero. Also there is no apparent reason to refrain from initializing RFCSR register 13, it doesn't fail what-so-ever. Signed-off-by: Daniel Golle <daniel@makrotopia.org>
2017-02-13 05:25:35 +00:00
#define TX_PWR_CFG_0_OFDM6_CH1 FIELD32(0x00f00000)
#define TX_PWR_CFG_0_OFDM12_CH0 FIELD32(0x0f000000)
#define TX_PWR_CFG_0_OFDM12_CH1 FIELD32(0xf0000000)
+/* bits for new 2T devices */
+#define TX_PWR_CFG_0B_1MBS_2MBS FIELD32(0x000000ff)
+#define TX_PWR_CFG_0B_5MBS_11MBS FIELD32(0x0000ff00)
+#define TX_PWR_CFG_0B_6MBS_9MBS FIELD32(0x00ff0000)
+#define TX_PWR_CFG_0B_12MBS_18MBS FIELD32(0xff000000)
+
/*
* TX_PWR_CFG_1:
@@ -1117,6 +1152,11 @@
rt2x00: mt7620: yet another beauty session So here is another round of improvements for MT7620 WiFi. This commit fixes a few significant issues related to TX_PWR_CFG_x and TX_ALC and also makes the code more readable by adding register descriptions for things added for MT7620 and use the usual bit-field access macros and the now defined macros instead of plain bit-ops and magic numbers. Properly describe EEPROM_TARGET_POWER at word 0x68 (== byte 0xD0) and thereby fix internal TXALC which would otherwise just read out-of-bounds of the EEPROM map. Split-out tx-power/ALC related stuff into an additional function. Fix VCO calibration, it was carried out properly in the channel switching but incomplete in the actual VCO calibration function. Also there is no need to trigger VCO calibration in channel switching, the VCO calibration function is already being called at this point. Remove it from channel switching function to avoid redundant code. The TX power calibration differs significantly from all other Mediatek/Ralink chips: They finally allow 0.5dB steps stored as 8-bit values for (almost) each bitrate -- and promptly ran out of space and for some reason didn't want to change the EEPROM layout. The hence opted for a scheme of sharing values for some adjecent bitrates and a highly over-complicated (or obfuscated?) way to populate the TX_PWR_CFG_x registers with the values stored in the EEPROM. The code here now looks much less complicated than what you see in the vendor's driver, however, it does the exact same thing: bGpwrdeltaMinus is a constant and always TRUE, hence half of the code was dead. Gpwrdelta is always 0 (rather than using the value read from the EEPROM). What remains is some very grotesque effort to avoid 0x20, probably some hardware bug related to some misunderstanding of what a singed 8-bit value is (imagine: if it was a signed 6-bit value then someone could believe that 0x20 == 0x0). And then they didn't clean it up once they later on anandonned that whole story of having a constant offset for 40 MHz channels and just set the offset to be constant 0 -- there is no effort for avoiding 0x20 for the 20 MHz values stored in the EEPROM, hence that's probably just a forbidden value in the EEPROM specs and won't appear anyway... Anyway, the whole thing felt like solving some college math test where in the end everything cancels out and the result equals 0 ;) To make sure that channel bandwidth power compensation really doesn't need to be taken care of, output a warning when the corresponding value stored in the EEPROM is non-zero. Also there is no apparent reason to refrain from initializing RFCSR register 13, it doesn't fail what-so-ever. Signed-off-by: Daniel Golle <daniel@makrotopia.org>
2017-02-13 05:25:35 +00:00
#define TX_PWR_CFG_1_MCS0_CH1 FIELD32(0x00f00000)
#define TX_PWR_CFG_1_MCS2_CH0 FIELD32(0x0f000000)
#define TX_PWR_CFG_1_MCS2_CH1 FIELD32(0xf0000000)
+/* bits for new 2T devices */
+#define TX_PWR_CFG_1B_24MBS_36MBS FIELD32(0x000000ff)
+#define TX_PWR_CFG_1B_48MBS FIELD32(0x0000ff00)
+#define TX_PWR_CFG_1B_MCS0_MCS1 FIELD32(0x00ff0000)
+#define TX_PWR_CFG_1B_MCS2_MCS3 FIELD32(0xff000000)
/*
* TX_PWR_CFG_2:
@@ -1139,6 +1179,11 @@
rt2x00: mt7620: yet another beauty session So here is another round of improvements for MT7620 WiFi. This commit fixes a few significant issues related to TX_PWR_CFG_x and TX_ALC and also makes the code more readable by adding register descriptions for things added for MT7620 and use the usual bit-field access macros and the now defined macros instead of plain bit-ops and magic numbers. Properly describe EEPROM_TARGET_POWER at word 0x68 (== byte 0xD0) and thereby fix internal TXALC which would otherwise just read out-of-bounds of the EEPROM map. Split-out tx-power/ALC related stuff into an additional function. Fix VCO calibration, it was carried out properly in the channel switching but incomplete in the actual VCO calibration function. Also there is no need to trigger VCO calibration in channel switching, the VCO calibration function is already being called at this point. Remove it from channel switching function to avoid redundant code. The TX power calibration differs significantly from all other Mediatek/Ralink chips: They finally allow 0.5dB steps stored as 8-bit values for (almost) each bitrate -- and promptly ran out of space and for some reason didn't want to change the EEPROM layout. The hence opted for a scheme of sharing values for some adjecent bitrates and a highly over-complicated (or obfuscated?) way to populate the TX_PWR_CFG_x registers with the values stored in the EEPROM. The code here now looks much less complicated than what you see in the vendor's driver, however, it does the exact same thing: bGpwrdeltaMinus is a constant and always TRUE, hence half of the code was dead. Gpwrdelta is always 0 (rather than using the value read from the EEPROM). What remains is some very grotesque effort to avoid 0x20, probably some hardware bug related to some misunderstanding of what a singed 8-bit value is (imagine: if it was a signed 6-bit value then someone could believe that 0x20 == 0x0). And then they didn't clean it up once they later on anandonned that whole story of having a constant offset for 40 MHz channels and just set the offset to be constant 0 -- there is no effort for avoiding 0x20 for the 20 MHz values stored in the EEPROM, hence that's probably just a forbidden value in the EEPROM specs and won't appear anyway... Anyway, the whole thing felt like solving some college math test where in the end everything cancels out and the result equals 0 ;) To make sure that channel bandwidth power compensation really doesn't need to be taken care of, output a warning when the corresponding value stored in the EEPROM is non-zero. Also there is no apparent reason to refrain from initializing RFCSR register 13, it doesn't fail what-so-ever. Signed-off-by: Daniel Golle <daniel@makrotopia.org>
2017-02-13 05:25:35 +00:00
#define TX_PWR_CFG_2_MCS8_CH1 FIELD32(0x00f00000)
#define TX_PWR_CFG_2_MCS10_CH0 FIELD32(0x0f000000)
#define TX_PWR_CFG_2_MCS10_CH1 FIELD32(0xf0000000)
+/* bits for new 2T devices */
+#define TX_PWR_CFG_2B_MCS4_MCS5 FIELD32(0x000000ff)
+#define TX_PWR_CFG_2B_MCS6_MCS7 FIELD32(0x0000ff00)
+#define TX_PWR_CFG_2B_MCS8_MCS9 FIELD32(0x00ff0000)
+#define TX_PWR_CFG_2B_MCS10_MCS11 FIELD32(0xff000000)
/*
* TX_PWR_CFG_3:
@@ -1161,6 +1206,11 @@
rt2x00: mt7620: yet another beauty session So here is another round of improvements for MT7620 WiFi. This commit fixes a few significant issues related to TX_PWR_CFG_x and TX_ALC and also makes the code more readable by adding register descriptions for things added for MT7620 and use the usual bit-field access macros and the now defined macros instead of plain bit-ops and magic numbers. Properly describe EEPROM_TARGET_POWER at word 0x68 (== byte 0xD0) and thereby fix internal TXALC which would otherwise just read out-of-bounds of the EEPROM map. Split-out tx-power/ALC related stuff into an additional function. Fix VCO calibration, it was carried out properly in the channel switching but incomplete in the actual VCO calibration function. Also there is no need to trigger VCO calibration in channel switching, the VCO calibration function is already being called at this point. Remove it from channel switching function to avoid redundant code. The TX power calibration differs significantly from all other Mediatek/Ralink chips: They finally allow 0.5dB steps stored as 8-bit values for (almost) each bitrate -- and promptly ran out of space and for some reason didn't want to change the EEPROM layout. The hence opted for a scheme of sharing values for some adjecent bitrates and a highly over-complicated (or obfuscated?) way to populate the TX_PWR_CFG_x registers with the values stored in the EEPROM. The code here now looks much less complicated than what you see in the vendor's driver, however, it does the exact same thing: bGpwrdeltaMinus is a constant and always TRUE, hence half of the code was dead. Gpwrdelta is always 0 (rather than using the value read from the EEPROM). What remains is some very grotesque effort to avoid 0x20, probably some hardware bug related to some misunderstanding of what a singed 8-bit value is (imagine: if it was a signed 6-bit value then someone could believe that 0x20 == 0x0). And then they didn't clean it up once they later on anandonned that whole story of having a constant offset for 40 MHz channels and just set the offset to be constant 0 -- there is no effort for avoiding 0x20 for the 20 MHz values stored in the EEPROM, hence that's probably just a forbidden value in the EEPROM specs and won't appear anyway... Anyway, the whole thing felt like solving some college math test where in the end everything cancels out and the result equals 0 ;) To make sure that channel bandwidth power compensation really doesn't need to be taken care of, output a warning when the corresponding value stored in the EEPROM is non-zero. Also there is no apparent reason to refrain from initializing RFCSR register 13, it doesn't fail what-so-ever. Signed-off-by: Daniel Golle <daniel@makrotopia.org>
2017-02-13 05:25:35 +00:00
#define TX_PWR_CFG_3_STBC0_CH1 FIELD32(0x00f00000)
#define TX_PWR_CFG_3_STBC2_CH0 FIELD32(0x0f000000)
#define TX_PWR_CFG_3_STBC2_CH1 FIELD32(0xf0000000)
+/* bits for new 2T devices */
+#define TX_PWR_CFG_3B_MCS12_MCS13 FIELD32(0x000000ff)
+#define TX_PWR_CFG_3B_MCS14 FIELD32(0x0000ff00)
+#define TX_PWR_CFG_3B_STBC_MCS0_MCS1 FIELD32(0x00ff0000)
+#define TX_PWR_CFG_3B_STBC_MCS2_MSC3 FIELD32(0xff000000)
/*
* TX_PWR_CFG_4:
@@ -1175,6 +1225,9 @@
#define TX_PWR_CFG_4_STBC4_CH1 FIELD32(0x000000f0)
#define TX_PWR_CFG_4_STBC6_CH0 FIELD32(0x00000f00)
#define TX_PWR_CFG_4_STBC6_CH1 FIELD32(0x0000f000)
rt2x00: mt7620: yet another beauty session So here is another round of improvements for MT7620 WiFi. This commit fixes a few significant issues related to TX_PWR_CFG_x and TX_ALC and also makes the code more readable by adding register descriptions for things added for MT7620 and use the usual bit-field access macros and the now defined macros instead of plain bit-ops and magic numbers. Properly describe EEPROM_TARGET_POWER at word 0x68 (== byte 0xD0) and thereby fix internal TXALC which would otherwise just read out-of-bounds of the EEPROM map. Split-out tx-power/ALC related stuff into an additional function. Fix VCO calibration, it was carried out properly in the channel switching but incomplete in the actual VCO calibration function. Also there is no need to trigger VCO calibration in channel switching, the VCO calibration function is already being called at this point. Remove it from channel switching function to avoid redundant code. The TX power calibration differs significantly from all other Mediatek/Ralink chips: They finally allow 0.5dB steps stored as 8-bit values for (almost) each bitrate -- and promptly ran out of space and for some reason didn't want to change the EEPROM layout. The hence opted for a scheme of sharing values for some adjecent bitrates and a highly over-complicated (or obfuscated?) way to populate the TX_PWR_CFG_x registers with the values stored in the EEPROM. The code here now looks much less complicated than what you see in the vendor's driver, however, it does the exact same thing: bGpwrdeltaMinus is a constant and always TRUE, hence half of the code was dead. Gpwrdelta is always 0 (rather than using the value read from the EEPROM). What remains is some very grotesque effort to avoid 0x20, probably some hardware bug related to some misunderstanding of what a singed 8-bit value is (imagine: if it was a signed 6-bit value then someone could believe that 0x20 == 0x0). And then they didn't clean it up once they later on anandonned that whole story of having a constant offset for 40 MHz channels and just set the offset to be constant 0 -- there is no effort for avoiding 0x20 for the 20 MHz values stored in the EEPROM, hence that's probably just a forbidden value in the EEPROM specs and won't appear anyway... Anyway, the whole thing felt like solving some college math test where in the end everything cancels out and the result equals 0 ;) To make sure that channel bandwidth power compensation really doesn't need to be taken care of, output a warning when the corresponding value stored in the EEPROM is non-zero. Also there is no apparent reason to refrain from initializing RFCSR register 13, it doesn't fail what-so-ever. Signed-off-by: Daniel Golle <daniel@makrotopia.org>
2017-02-13 05:25:35 +00:00
+/* bits for new 2T devices */
+#define TX_PWR_CFG_4B_STBC_MCS4_MCS5 FIELD32(0x000000ff)
+#define TX_PWR_CFG_4B_STBC_MCS6 FIELD32(0x0000ff00)
/*
* TX_PIN_CFG:
@@ -1201,6 +1254,8 @@
#define TX_PIN_CFG_RFTR_POL FIELD32(0x00020000)
#define TX_PIN_CFG_TRSW_EN FIELD32(0x00040000)
#define TX_PIN_CFG_TRSW_POL FIELD32(0x00080000)
rt2x00: mt7620: yet another beauty session So here is another round of improvements for MT7620 WiFi. This commit fixes a few significant issues related to TX_PWR_CFG_x and TX_ALC and also makes the code more readable by adding register descriptions for things added for MT7620 and use the usual bit-field access macros and the now defined macros instead of plain bit-ops and magic numbers. Properly describe EEPROM_TARGET_POWER at word 0x68 (== byte 0xD0) and thereby fix internal TXALC which would otherwise just read out-of-bounds of the EEPROM map. Split-out tx-power/ALC related stuff into an additional function. Fix VCO calibration, it was carried out properly in the channel switching but incomplete in the actual VCO calibration function. Also there is no need to trigger VCO calibration in channel switching, the VCO calibration function is already being called at this point. Remove it from channel switching function to avoid redundant code. The TX power calibration differs significantly from all other Mediatek/Ralink chips: They finally allow 0.5dB steps stored as 8-bit values for (almost) each bitrate -- and promptly ran out of space and for some reason didn't want to change the EEPROM layout. The hence opted for a scheme of sharing values for some adjecent bitrates and a highly over-complicated (or obfuscated?) way to populate the TX_PWR_CFG_x registers with the values stored in the EEPROM. The code here now looks much less complicated than what you see in the vendor's driver, however, it does the exact same thing: bGpwrdeltaMinus is a constant and always TRUE, hence half of the code was dead. Gpwrdelta is always 0 (rather than using the value read from the EEPROM). What remains is some very grotesque effort to avoid 0x20, probably some hardware bug related to some misunderstanding of what a singed 8-bit value is (imagine: if it was a signed 6-bit value then someone could believe that 0x20 == 0x0). And then they didn't clean it up once they later on anandonned that whole story of having a constant offset for 40 MHz channels and just set the offset to be constant 0 -- there is no effort for avoiding 0x20 for the 20 MHz values stored in the EEPROM, hence that's probably just a forbidden value in the EEPROM specs and won't appear anyway... Anyway, the whole thing felt like solving some college math test where in the end everything cancels out and the result equals 0 ;) To make sure that channel bandwidth power compensation really doesn't need to be taken care of, output a warning when the corresponding value stored in the EEPROM is non-zero. Also there is no apparent reason to refrain from initializing RFCSR register 13, it doesn't fail what-so-ever. Signed-off-by: Daniel Golle <daniel@makrotopia.org>
2017-02-13 05:25:35 +00:00
+#define TX_PIN_CFG_RFRX_EN FIELD32(0x00100000)
+#define TX_PIN_CFG_RFRX_POL FIELD32(0x00200000)
#define TX_PIN_CFG_PA_PE_A2_EN FIELD32(0x01000000)
#define TX_PIN_CFG_PA_PE_G2_EN FIELD32(0x02000000)
#define TX_PIN_CFG_PA_PE_A2_POL FIELD32(0x04000000)
@@ -1547,6 +1602,95 @@
#define TX_PWR_CFG_4_EXT_STBC4_CH2 FIELD32(0x0000000f)
#define TX_PWR_CFG_4_EXT_STBC6_CH2 FIELD32(0x00000f00)
rt2x00: mt7620: yet another beauty session So here is another round of improvements for MT7620 WiFi. This commit fixes a few significant issues related to TX_PWR_CFG_x and TX_ALC and also makes the code more readable by adding register descriptions for things added for MT7620 and use the usual bit-field access macros and the now defined macros instead of plain bit-ops and magic numbers. Properly describe EEPROM_TARGET_POWER at word 0x68 (== byte 0xD0) and thereby fix internal TXALC which would otherwise just read out-of-bounds of the EEPROM map. Split-out tx-power/ALC related stuff into an additional function. Fix VCO calibration, it was carried out properly in the channel switching but incomplete in the actual VCO calibration function. Also there is no need to trigger VCO calibration in channel switching, the VCO calibration function is already being called at this point. Remove it from channel switching function to avoid redundant code. The TX power calibration differs significantly from all other Mediatek/Ralink chips: They finally allow 0.5dB steps stored as 8-bit values for (almost) each bitrate -- and promptly ran out of space and for some reason didn't want to change the EEPROM layout. The hence opted for a scheme of sharing values for some adjecent bitrates and a highly over-complicated (or obfuscated?) way to populate the TX_PWR_CFG_x registers with the values stored in the EEPROM. The code here now looks much less complicated than what you see in the vendor's driver, however, it does the exact same thing: bGpwrdeltaMinus is a constant and always TRUE, hence half of the code was dead. Gpwrdelta is always 0 (rather than using the value read from the EEPROM). What remains is some very grotesque effort to avoid 0x20, probably some hardware bug related to some misunderstanding of what a singed 8-bit value is (imagine: if it was a signed 6-bit value then someone could believe that 0x20 == 0x0). And then they didn't clean it up once they later on anandonned that whole story of having a constant offset for 40 MHz channels and just set the offset to be constant 0 -- there is no effort for avoiding 0x20 for the 20 MHz values stored in the EEPROM, hence that's probably just a forbidden value in the EEPROM specs and won't appear anyway... Anyway, the whole thing felt like solving some college math test where in the end everything cancels out and the result equals 0 ;) To make sure that channel bandwidth power compensation really doesn't need to be taken care of, output a warning when the corresponding value stored in the EEPROM is non-zero. Also there is no apparent reason to refrain from initializing RFCSR register 13, it doesn't fail what-so-ever. Signed-off-by: Daniel Golle <daniel@makrotopia.org>
2017-02-13 05:25:35 +00:00
+/* TXn_RF_GAIN_CORRECT: RF Gain Correction for each RF_ALC[3:2]
+ * Unit: 0.1 dB, Range: -3.2 dB to 3.1 dB
+ */
+#define TX0_RF_GAIN_CORRECT 0x13a0
rt2x00: mt7620: yet another beauty session So here is another round of improvements for MT7620 WiFi. This commit fixes a few significant issues related to TX_PWR_CFG_x and TX_ALC and also makes the code more readable by adding register descriptions for things added for MT7620 and use the usual bit-field access macros and the now defined macros instead of plain bit-ops and magic numbers. Properly describe EEPROM_TARGET_POWER at word 0x68 (== byte 0xD0) and thereby fix internal TXALC which would otherwise just read out-of-bounds of the EEPROM map. Split-out tx-power/ALC related stuff into an additional function. Fix VCO calibration, it was carried out properly in the channel switching but incomplete in the actual VCO calibration function. Also there is no need to trigger VCO calibration in channel switching, the VCO calibration function is already being called at this point. Remove it from channel switching function to avoid redundant code. The TX power calibration differs significantly from all other Mediatek/Ralink chips: They finally allow 0.5dB steps stored as 8-bit values for (almost) each bitrate -- and promptly ran out of space and for some reason didn't want to change the EEPROM layout. The hence opted for a scheme of sharing values for some adjecent bitrates and a highly over-complicated (or obfuscated?) way to populate the TX_PWR_CFG_x registers with the values stored in the EEPROM. The code here now looks much less complicated than what you see in the vendor's driver, however, it does the exact same thing: bGpwrdeltaMinus is a constant and always TRUE, hence half of the code was dead. Gpwrdelta is always 0 (rather than using the value read from the EEPROM). What remains is some very grotesque effort to avoid 0x20, probably some hardware bug related to some misunderstanding of what a singed 8-bit value is (imagine: if it was a signed 6-bit value then someone could believe that 0x20 == 0x0). And then they didn't clean it up once they later on anandonned that whole story of having a constant offset for 40 MHz channels and just set the offset to be constant 0 -- there is no effort for avoiding 0x20 for the 20 MHz values stored in the EEPROM, hence that's probably just a forbidden value in the EEPROM specs and won't appear anyway... Anyway, the whole thing felt like solving some college math test where in the end everything cancels out and the result equals 0 ;) To make sure that channel bandwidth power compensation really doesn't need to be taken care of, output a warning when the corresponding value stored in the EEPROM is non-zero. Also there is no apparent reason to refrain from initializing RFCSR register 13, it doesn't fail what-so-ever. Signed-off-by: Daniel Golle <daniel@makrotopia.org>
2017-02-13 05:25:35 +00:00
+#define TX0_RF_GAIN_CORRECT_GAIN_CORR_0 FIELD32(0x0000003f)
+#define TX0_RF_GAIN_CORRECT_GAIN_CORR_1 FIELD32(0x00003f00)
+#define TX0_RF_GAIN_CORRECT_GAIN_CORR_2 FIELD32(0x003f0000)
+#define TX0_RF_GAIN_CORRECT_GAIN_CORR_3 FIELD32(0x3f000000)
+
+#define TX1_RF_GAIN_CORRECT 0x13a4
rt2x00: mt7620: yet another beauty session So here is another round of improvements for MT7620 WiFi. This commit fixes a few significant issues related to TX_PWR_CFG_x and TX_ALC and also makes the code more readable by adding register descriptions for things added for MT7620 and use the usual bit-field access macros and the now defined macros instead of plain bit-ops and magic numbers. Properly describe EEPROM_TARGET_POWER at word 0x68 (== byte 0xD0) and thereby fix internal TXALC which would otherwise just read out-of-bounds of the EEPROM map. Split-out tx-power/ALC related stuff into an additional function. Fix VCO calibration, it was carried out properly in the channel switching but incomplete in the actual VCO calibration function. Also there is no need to trigger VCO calibration in channel switching, the VCO calibration function is already being called at this point. Remove it from channel switching function to avoid redundant code. The TX power calibration differs significantly from all other Mediatek/Ralink chips: They finally allow 0.5dB steps stored as 8-bit values for (almost) each bitrate -- and promptly ran out of space and for some reason didn't want to change the EEPROM layout. The hence opted for a scheme of sharing values for some adjecent bitrates and a highly over-complicated (or obfuscated?) way to populate the TX_PWR_CFG_x registers with the values stored in the EEPROM. The code here now looks much less complicated than what you see in the vendor's driver, however, it does the exact same thing: bGpwrdeltaMinus is a constant and always TRUE, hence half of the code was dead. Gpwrdelta is always 0 (rather than using the value read from the EEPROM). What remains is some very grotesque effort to avoid 0x20, probably some hardware bug related to some misunderstanding of what a singed 8-bit value is (imagine: if it was a signed 6-bit value then someone could believe that 0x20 == 0x0). And then they didn't clean it up once they later on anandonned that whole story of having a constant offset for 40 MHz channels and just set the offset to be constant 0 -- there is no effort for avoiding 0x20 for the 20 MHz values stored in the EEPROM, hence that's probably just a forbidden value in the EEPROM specs and won't appear anyway... Anyway, the whole thing felt like solving some college math test where in the end everything cancels out and the result equals 0 ;) To make sure that channel bandwidth power compensation really doesn't need to be taken care of, output a warning when the corresponding value stored in the EEPROM is non-zero. Also there is no apparent reason to refrain from initializing RFCSR register 13, it doesn't fail what-so-ever. Signed-off-by: Daniel Golle <daniel@makrotopia.org>
2017-02-13 05:25:35 +00:00
+#define TX1_RF_GAIN_CORRECT_GAIN_CORR_0 FIELD32(0x0000003f)
+#define TX1_RF_GAIN_CORRECT_GAIN_CORR_1 FIELD32(0x00003f00)
+#define TX1_RF_GAIN_CORRECT_GAIN_CORR_2 FIELD32(0x003f0000)
+#define TX1_RF_GAIN_CORRECT_GAIN_CORR_3 FIELD32(0x3f000000)
+
+/* TXn_RF_GAIN_ATTEN: TXn RF Gain Attenuation Level
+ * Format: 7-bit, signed value
+ * Unit: 0.5 dB, Range: -20 dB to -5 dB
+ */
+#define TX0_RF_GAIN_ATTEN 0x13a8
rt2x00: mt7620: yet another beauty session So here is another round of improvements for MT7620 WiFi. This commit fixes a few significant issues related to TX_PWR_CFG_x and TX_ALC and also makes the code more readable by adding register descriptions for things added for MT7620 and use the usual bit-field access macros and the now defined macros instead of plain bit-ops and magic numbers. Properly describe EEPROM_TARGET_POWER at word 0x68 (== byte 0xD0) and thereby fix internal TXALC which would otherwise just read out-of-bounds of the EEPROM map. Split-out tx-power/ALC related stuff into an additional function. Fix VCO calibration, it was carried out properly in the channel switching but incomplete in the actual VCO calibration function. Also there is no need to trigger VCO calibration in channel switching, the VCO calibration function is already being called at this point. Remove it from channel switching function to avoid redundant code. The TX power calibration differs significantly from all other Mediatek/Ralink chips: They finally allow 0.5dB steps stored as 8-bit values for (almost) each bitrate -- and promptly ran out of space and for some reason didn't want to change the EEPROM layout. The hence opted for a scheme of sharing values for some adjecent bitrates and a highly over-complicated (or obfuscated?) way to populate the TX_PWR_CFG_x registers with the values stored in the EEPROM. The code here now looks much less complicated than what you see in the vendor's driver, however, it does the exact same thing: bGpwrdeltaMinus is a constant and always TRUE, hence half of the code was dead. Gpwrdelta is always 0 (rather than using the value read from the EEPROM). What remains is some very grotesque effort to avoid 0x20, probably some hardware bug related to some misunderstanding of what a singed 8-bit value is (imagine: if it was a signed 6-bit value then someone could believe that 0x20 == 0x0). And then they didn't clean it up once they later on anandonned that whole story of having a constant offset for 40 MHz channels and just set the offset to be constant 0 -- there is no effort for avoiding 0x20 for the 20 MHz values stored in the EEPROM, hence that's probably just a forbidden value in the EEPROM specs and won't appear anyway... Anyway, the whole thing felt like solving some college math test where in the end everything cancels out and the result equals 0 ;) To make sure that channel bandwidth power compensation really doesn't need to be taken care of, output a warning when the corresponding value stored in the EEPROM is non-zero. Also there is no apparent reason to refrain from initializing RFCSR register 13, it doesn't fail what-so-ever. Signed-off-by: Daniel Golle <daniel@makrotopia.org>
2017-02-13 05:25:35 +00:00
+#define TX0_RF_GAIN_ATTEN_LEVEL_0 FIELD32(0x0000007f)
+#define TX0_RF_GAIN_ATTEN_LEVEL_1 FIELD32(0x00007f00)
+#define TX0_RF_GAIN_ATTEN_LEVEL_2 FIELD32(0x007f0000)
+#define TX0_RF_GAIN_ATTEN_LEVEL_3 FIELD32(0x7f000000)
+#define TX1_RF_GAIN_ATTEN 0x13ac
rt2x00: mt7620: yet another beauty session So here is another round of improvements for MT7620 WiFi. This commit fixes a few significant issues related to TX_PWR_CFG_x and TX_ALC and also makes the code more readable by adding register descriptions for things added for MT7620 and use the usual bit-field access macros and the now defined macros instead of plain bit-ops and magic numbers. Properly describe EEPROM_TARGET_POWER at word 0x68 (== byte 0xD0) and thereby fix internal TXALC which would otherwise just read out-of-bounds of the EEPROM map. Split-out tx-power/ALC related stuff into an additional function. Fix VCO calibration, it was carried out properly in the channel switching but incomplete in the actual VCO calibration function. Also there is no need to trigger VCO calibration in channel switching, the VCO calibration function is already being called at this point. Remove it from channel switching function to avoid redundant code. The TX power calibration differs significantly from all other Mediatek/Ralink chips: They finally allow 0.5dB steps stored as 8-bit values for (almost) each bitrate -- and promptly ran out of space and for some reason didn't want to change the EEPROM layout. The hence opted for a scheme of sharing values for some adjecent bitrates and a highly over-complicated (or obfuscated?) way to populate the TX_PWR_CFG_x registers with the values stored in the EEPROM. The code here now looks much less complicated than what you see in the vendor's driver, however, it does the exact same thing: bGpwrdeltaMinus is a constant and always TRUE, hence half of the code was dead. Gpwrdelta is always 0 (rather than using the value read from the EEPROM). What remains is some very grotesque effort to avoid 0x20, probably some hardware bug related to some misunderstanding of what a singed 8-bit value is (imagine: if it was a signed 6-bit value then someone could believe that 0x20 == 0x0). And then they didn't clean it up once they later on anandonned that whole story of having a constant offset for 40 MHz channels and just set the offset to be constant 0 -- there is no effort for avoiding 0x20 for the 20 MHz values stored in the EEPROM, hence that's probably just a forbidden value in the EEPROM specs and won't appear anyway... Anyway, the whole thing felt like solving some college math test where in the end everything cancels out and the result equals 0 ;) To make sure that channel bandwidth power compensation really doesn't need to be taken care of, output a warning when the corresponding value stored in the EEPROM is non-zero. Also there is no apparent reason to refrain from initializing RFCSR register 13, it doesn't fail what-so-ever. Signed-off-by: Daniel Golle <daniel@makrotopia.org>
2017-02-13 05:25:35 +00:00
+#define TX1_RF_GAIN_ATTEN_LEVEL_0 FIELD32(0x0000007f)
+#define TX1_RF_GAIN_ATTEN_LEVEL_1 FIELD32(0x00007f00)
+#define TX1_RF_GAIN_ATTEN_LEVEL_2 FIELD32(0x007f0000)
+#define TX1_RF_GAIN_ATTEN_LEVEL_3 FIELD32(0x7f000000)
+
+/* TX_ALC_CFG_0: TX Automatic Level Control Configuration 0
+ * TX_ALC_LIMIT_n: TXn upper limit
+ * TX_ALC_CH_INIT_n: TXn channel initial transmission gain
+ * Unit: 0.5 dB, Range: 0 to 23.5 dB
+ */
+#define TX_ALC_CFG_0 0x13b0
+#define TX_ALC_CFG_0_CH_INIT_0 FIELD32(0x0000003f)
+#define TX_ALC_CFG_0_CH_INIT_1 FIELD32(0x00003f00)
+#define TX_ALC_CFG_0_LIMIT_0 FIELD32(0x003f0000)
+#define TX_ALC_CFG_0_LIMIT_1 FIELD32(0x3f000000)
+
+/* TX_ALC_CFG_1: TX Automatic Level Control Configuration 1
+ * TX_TEMP_COMP: TX Power Temperature Compensation
+ * Unit: 0.5 dB, Range: -10 dB to 10 dB
+ * TXn_GAIN_FINE: TXn Gain Fine Adjustment
+ * Unit: 0.1 dB, Range: -0.8 dB to 0.7 dB
+ * RF_TOS_DLY: Sets the RF_TOS_EN assertion delay after
+ * deassertion of PA_PE.
+ * Unit: 0.25 usec
+ * TXn_RF_GAIN_ATTEN: TXn RF gain attentuation selector
+ * RF_TOS_TIMEOUT: time-out value for RF_TOS_ENABLE
+ * deassertion if RF_TOS_DONE is missing.
+ * Unit: 0.25 usec
+ * RF_TOS_ENABLE: TX offset calibration enable
+ * ROS_BUSY_EN: RX offset calibration busy enable
+ */
+#define TX_ALC_CFG_1 0x13b4
+#define TX_ALC_CFG_1_TX_TEMP_COMP FIELD32(0x0000003f)
+#define TX_ALC_CFG_1_TX0_GAIN_FINE FIELD32(0x00000f00)
+#define TX_ALC_CFG_1_TX1_GAIN_FINE FIELD32(0x0000f000)
+#define TX_ALC_CFG_1_RF_TOS_DLY FIELD32(0x00070000)
+#define TX_ALC_CFG_1_TX0_RF_GAIN_ATTEN FIELD32(0x00300000)
+#define TX_ALC_CFG_1_TX1_RF_GAIN_ATTEN FIELD32(0x00c00000)
+#define TX_ALC_CFG_1_RF_TOS_TIMEOUT FIELD32(0x3f000000)
+#define TX_ALC_CFG_1_RF_TOS_ENABLE FIELD32(0x40000000)
+#define TX_ALC_CFG_1_ROS_BUSY_EN FIELD32(0x80000000)
+
+/* TXn_BB_GAIN_ATTEN: TXn RF Gain Attenuation Level
+ * Format: 5-bit signed values
+ * Unit: 0.5 dB, Range: -8 dB to 7 dB
+ */
+#define TX0_BB_GAIN_ATTEN 0x13c0
rt2x00: mt7620: yet another beauty session So here is another round of improvements for MT7620 WiFi. This commit fixes a few significant issues related to TX_PWR_CFG_x and TX_ALC and also makes the code more readable by adding register descriptions for things added for MT7620 and use the usual bit-field access macros and the now defined macros instead of plain bit-ops and magic numbers. Properly describe EEPROM_TARGET_POWER at word 0x68 (== byte 0xD0) and thereby fix internal TXALC which would otherwise just read out-of-bounds of the EEPROM map. Split-out tx-power/ALC related stuff into an additional function. Fix VCO calibration, it was carried out properly in the channel switching but incomplete in the actual VCO calibration function. Also there is no need to trigger VCO calibration in channel switching, the VCO calibration function is already being called at this point. Remove it from channel switching function to avoid redundant code. The TX power calibration differs significantly from all other Mediatek/Ralink chips: They finally allow 0.5dB steps stored as 8-bit values for (almost) each bitrate -- and promptly ran out of space and for some reason didn't want to change the EEPROM layout. The hence opted for a scheme of sharing values for some adjecent bitrates and a highly over-complicated (or obfuscated?) way to populate the TX_PWR_CFG_x registers with the values stored in the EEPROM. The code here now looks much less complicated than what you see in the vendor's driver, however, it does the exact same thing: bGpwrdeltaMinus is a constant and always TRUE, hence half of the code was dead. Gpwrdelta is always 0 (rather than using the value read from the EEPROM). What remains is some very grotesque effort to avoid 0x20, probably some hardware bug related to some misunderstanding of what a singed 8-bit value is (imagine: if it was a signed 6-bit value then someone could believe that 0x20 == 0x0). And then they didn't clean it up once they later on anandonned that whole story of having a constant offset for 40 MHz channels and just set the offset to be constant 0 -- there is no effort for avoiding 0x20 for the 20 MHz values stored in the EEPROM, hence that's probably just a forbidden value in the EEPROM specs and won't appear anyway... Anyway, the whole thing felt like solving some college math test where in the end everything cancels out and the result equals 0 ;) To make sure that channel bandwidth power compensation really doesn't need to be taken care of, output a warning when the corresponding value stored in the EEPROM is non-zero. Also there is no apparent reason to refrain from initializing RFCSR register 13, it doesn't fail what-so-ever. Signed-off-by: Daniel Golle <daniel@makrotopia.org>
2017-02-13 05:25:35 +00:00
+#define TX0_BB_GAIN_ATTEN_LEVEL_0 FIELD32(0x0000001f)
+#define TX0_BB_GAIN_ATTEN_LEVEL_1 FIELD32(0x00001f00)
+#define TX0_BB_GAIN_ATTEN_LEVEL_2 FIELD32(0x001f0000)
+#define TX0_BB_GAIN_ATTEN_LEVEL_3 FIELD32(0x1f000000)
+#define TX1_BB_GAIN_ATTEN 0x13c4
rt2x00: mt7620: yet another beauty session So here is another round of improvements for MT7620 WiFi. This commit fixes a few significant issues related to TX_PWR_CFG_x and TX_ALC and also makes the code more readable by adding register descriptions for things added for MT7620 and use the usual bit-field access macros and the now defined macros instead of plain bit-ops and magic numbers. Properly describe EEPROM_TARGET_POWER at word 0x68 (== byte 0xD0) and thereby fix internal TXALC which would otherwise just read out-of-bounds of the EEPROM map. Split-out tx-power/ALC related stuff into an additional function. Fix VCO calibration, it was carried out properly in the channel switching but incomplete in the actual VCO calibration function. Also there is no need to trigger VCO calibration in channel switching, the VCO calibration function is already being called at this point. Remove it from channel switching function to avoid redundant code. The TX power calibration differs significantly from all other Mediatek/Ralink chips: They finally allow 0.5dB steps stored as 8-bit values for (almost) each bitrate -- and promptly ran out of space and for some reason didn't want to change the EEPROM layout. The hence opted for a scheme of sharing values for some adjecent bitrates and a highly over-complicated (or obfuscated?) way to populate the TX_PWR_CFG_x registers with the values stored in the EEPROM. The code here now looks much less complicated than what you see in the vendor's driver, however, it does the exact same thing: bGpwrdeltaMinus is a constant and always TRUE, hence half of the code was dead. Gpwrdelta is always 0 (rather than using the value read from the EEPROM). What remains is some very grotesque effort to avoid 0x20, probably some hardware bug related to some misunderstanding of what a singed 8-bit value is (imagine: if it was a signed 6-bit value then someone could believe that 0x20 == 0x0). And then they didn't clean it up once they later on anandonned that whole story of having a constant offset for 40 MHz channels and just set the offset to be constant 0 -- there is no effort for avoiding 0x20 for the 20 MHz values stored in the EEPROM, hence that's probably just a forbidden value in the EEPROM specs and won't appear anyway... Anyway, the whole thing felt like solving some college math test where in the end everything cancels out and the result equals 0 ;) To make sure that channel bandwidth power compensation really doesn't need to be taken care of, output a warning when the corresponding value stored in the EEPROM is non-zero. Also there is no apparent reason to refrain from initializing RFCSR register 13, it doesn't fail what-so-ever. Signed-off-by: Daniel Golle <daniel@makrotopia.org>
2017-02-13 05:25:35 +00:00
+#define TX1_BB_GAIN_ATTEN_LEVEL_0 FIELD32(0x0000001f)
+#define TX1_BB_GAIN_ATTEN_LEVEL_1 FIELD32(0x00001f00)
+#define TX1_BB_GAIN_ATTEN_LEVEL_2 FIELD32(0x001f0000)
+#define TX1_BB_GAIN_ATTEN_LEVEL_3 FIELD32(0x1f000000)
+
+/* TX_ALC_VGA3: TX Automatic Level Correction Variable Gain Amplifier 3 */
+#define TX_ALC_VGA3 0x13c8
rt2x00: mt7620: yet another beauty session So here is another round of improvements for MT7620 WiFi. This commit fixes a few significant issues related to TX_PWR_CFG_x and TX_ALC and also makes the code more readable by adding register descriptions for things added for MT7620 and use the usual bit-field access macros and the now defined macros instead of plain bit-ops and magic numbers. Properly describe EEPROM_TARGET_POWER at word 0x68 (== byte 0xD0) and thereby fix internal TXALC which would otherwise just read out-of-bounds of the EEPROM map. Split-out tx-power/ALC related stuff into an additional function. Fix VCO calibration, it was carried out properly in the channel switching but incomplete in the actual VCO calibration function. Also there is no need to trigger VCO calibration in channel switching, the VCO calibration function is already being called at this point. Remove it from channel switching function to avoid redundant code. The TX power calibration differs significantly from all other Mediatek/Ralink chips: They finally allow 0.5dB steps stored as 8-bit values for (almost) each bitrate -- and promptly ran out of space and for some reason didn't want to change the EEPROM layout. The hence opted for a scheme of sharing values for some adjecent bitrates and a highly over-complicated (or obfuscated?) way to populate the TX_PWR_CFG_x registers with the values stored in the EEPROM. The code here now looks much less complicated than what you see in the vendor's driver, however, it does the exact same thing: bGpwrdeltaMinus is a constant and always TRUE, hence half of the code was dead. Gpwrdelta is always 0 (rather than using the value read from the EEPROM). What remains is some very grotesque effort to avoid 0x20, probably some hardware bug related to some misunderstanding of what a singed 8-bit value is (imagine: if it was a signed 6-bit value then someone could believe that 0x20 == 0x0). And then they didn't clean it up once they later on anandonned that whole story of having a constant offset for 40 MHz channels and just set the offset to be constant 0 -- there is no effort for avoiding 0x20 for the 20 MHz values stored in the EEPROM, hence that's probably just a forbidden value in the EEPROM specs and won't appear anyway... Anyway, the whole thing felt like solving some college math test where in the end everything cancels out and the result equals 0 ;) To make sure that channel bandwidth power compensation really doesn't need to be taken care of, output a warning when the corresponding value stored in the EEPROM is non-zero. Also there is no apparent reason to refrain from initializing RFCSR register 13, it doesn't fail what-so-ever. Signed-off-by: Daniel Golle <daniel@makrotopia.org>
2017-02-13 05:25:35 +00:00
+#define TX_ALC_VGA3_TX0_ALC_VGA3 FIELD32(0x0000001f)
+#define TX_ALC_VGA3_TX1_ALC_VGA3 FIELD32(0x00001f00)
+#define TX_ALC_VGA3_TX0_ALC_VGA2 FIELD32(0x001f0000)
+#define TX_ALC_VGA3_TX1_ALC_VGA2 FIELD32(0x1f000000)
+
/* TX_PWR_CFG_7 */
#define TX_PWR_CFG_7 0x13d4
#define TX_PWR_CFG_7_OFDM54_CH0 FIELD32(0x0000000f)
@@ -1555,6 +1699,10 @@
rt2x00: mt7620: yet another beauty session So here is another round of improvements for MT7620 WiFi. This commit fixes a few significant issues related to TX_PWR_CFG_x and TX_ALC and also makes the code more readable by adding register descriptions for things added for MT7620 and use the usual bit-field access macros and the now defined macros instead of plain bit-ops and magic numbers. Properly describe EEPROM_TARGET_POWER at word 0x68 (== byte 0xD0) and thereby fix internal TXALC which would otherwise just read out-of-bounds of the EEPROM map. Split-out tx-power/ALC related stuff into an additional function. Fix VCO calibration, it was carried out properly in the channel switching but incomplete in the actual VCO calibration function. Also there is no need to trigger VCO calibration in channel switching, the VCO calibration function is already being called at this point. Remove it from channel switching function to avoid redundant code. The TX power calibration differs significantly from all other Mediatek/Ralink chips: They finally allow 0.5dB steps stored as 8-bit values for (almost) each bitrate -- and promptly ran out of space and for some reason didn't want to change the EEPROM layout. The hence opted for a scheme of sharing values for some adjecent bitrates and a highly over-complicated (or obfuscated?) way to populate the TX_PWR_CFG_x registers with the values stored in the EEPROM. The code here now looks much less complicated than what you see in the vendor's driver, however, it does the exact same thing: bGpwrdeltaMinus is a constant and always TRUE, hence half of the code was dead. Gpwrdelta is always 0 (rather than using the value read from the EEPROM). What remains is some very grotesque effort to avoid 0x20, probably some hardware bug related to some misunderstanding of what a singed 8-bit value is (imagine: if it was a signed 6-bit value then someone could believe that 0x20 == 0x0). And then they didn't clean it up once they later on anandonned that whole story of having a constant offset for 40 MHz channels and just set the offset to be constant 0 -- there is no effort for avoiding 0x20 for the 20 MHz values stored in the EEPROM, hence that's probably just a forbidden value in the EEPROM specs and won't appear anyway... Anyway, the whole thing felt like solving some college math test where in the end everything cancels out and the result equals 0 ;) To make sure that channel bandwidth power compensation really doesn't need to be taken care of, output a warning when the corresponding value stored in the EEPROM is non-zero. Also there is no apparent reason to refrain from initializing RFCSR register 13, it doesn't fail what-so-ever. Signed-off-by: Daniel Golle <daniel@makrotopia.org>
2017-02-13 05:25:35 +00:00
#define TX_PWR_CFG_7_MCS7_CH0 FIELD32(0x000f0000)
#define TX_PWR_CFG_7_MCS7_CH1 FIELD32(0x00f00000)
#define TX_PWR_CFG_7_MCS7_CH2 FIELD32(0x0f000000)
+/* bits for new 2T devices */
+#define TX_PWR_CFG_7B_54MBS FIELD32(0x000000ff)
+#define TX_PWR_CFG_7B_MCS7 FIELD32(0x00ff0000)
+
/* TX_PWR_CFG_8 */
#define TX_PWR_CFG_8 0x13d8
@@ -1564,12 +1712,17 @@
rt2x00: mt7620: yet another beauty session So here is another round of improvements for MT7620 WiFi. This commit fixes a few significant issues related to TX_PWR_CFG_x and TX_ALC and also makes the code more readable by adding register descriptions for things added for MT7620 and use the usual bit-field access macros and the now defined macros instead of plain bit-ops and magic numbers. Properly describe EEPROM_TARGET_POWER at word 0x68 (== byte 0xD0) and thereby fix internal TXALC which would otherwise just read out-of-bounds of the EEPROM map. Split-out tx-power/ALC related stuff into an additional function. Fix VCO calibration, it was carried out properly in the channel switching but incomplete in the actual VCO calibration function. Also there is no need to trigger VCO calibration in channel switching, the VCO calibration function is already being called at this point. Remove it from channel switching function to avoid redundant code. The TX power calibration differs significantly from all other Mediatek/Ralink chips: They finally allow 0.5dB steps stored as 8-bit values for (almost) each bitrate -- and promptly ran out of space and for some reason didn't want to change the EEPROM layout. The hence opted for a scheme of sharing values for some adjecent bitrates and a highly over-complicated (or obfuscated?) way to populate the TX_PWR_CFG_x registers with the values stored in the EEPROM. The code here now looks much less complicated than what you see in the vendor's driver, however, it does the exact same thing: bGpwrdeltaMinus is a constant and always TRUE, hence half of the code was dead. Gpwrdelta is always 0 (rather than using the value read from the EEPROM). What remains is some very grotesque effort to avoid 0x20, probably some hardware bug related to some misunderstanding of what a singed 8-bit value is (imagine: if it was a signed 6-bit value then someone could believe that 0x20 == 0x0). And then they didn't clean it up once they later on anandonned that whole story of having a constant offset for 40 MHz channels and just set the offset to be constant 0 -- there is no effort for avoiding 0x20 for the 20 MHz values stored in the EEPROM, hence that's probably just a forbidden value in the EEPROM specs and won't appear anyway... Anyway, the whole thing felt like solving some college math test where in the end everything cancels out and the result equals 0 ;) To make sure that channel bandwidth power compensation really doesn't need to be taken care of, output a warning when the corresponding value stored in the EEPROM is non-zero. Also there is no apparent reason to refrain from initializing RFCSR register 13, it doesn't fail what-so-ever. Signed-off-by: Daniel Golle <daniel@makrotopia.org>
2017-02-13 05:25:35 +00:00
#define TX_PWR_CFG_8_MCS23_CH0 FIELD32(0x000f0000)
#define TX_PWR_CFG_8_MCS23_CH1 FIELD32(0x00f00000)
#define TX_PWR_CFG_8_MCS23_CH2 FIELD32(0x0f000000)
+/* bits for new 2T devices */
+#define TX_PWR_CFG_8B_MCS15 FIELD32(0x000000ff)
+
/* TX_PWR_CFG_9 */
#define TX_PWR_CFG_9 0x13dc
#define TX_PWR_CFG_9_STBC7_CH0 FIELD32(0x0000000f)
#define TX_PWR_CFG_9_STBC7_CH1 FIELD32(0x000000f0)
#define TX_PWR_CFG_9_STBC7_CH2 FIELD32(0x00000f00)
+/* bits for new 2T devices */
+#define TX_PWR_CFG_9B_STBC_MCS7 FIELD32(0x000000ff)
/*
* RX_FILTER_CFG: RX configuration register.
@@ -2137,11 +2290,14 @@ struct mac_iveiv_entry {
#define RFCSR1_TX1_PD FIELD8(0x20)
#define RFCSR1_RX2_PD FIELD8(0x40)
#define RFCSR1_TX2_PD FIELD8(0x80)
+#define RFCSR1_TX2_EN_MT7620 FIELD8(0x02)
/*
* RFCSR 2:
*/
#define RFCSR2_RESCAL_EN FIELD8(0x80)
+#define RFCSR2_RX2_EN_MT7620 FIELD8(0x02)
+#define RFCSR2_TX2_EN_MT7620 FIELD8(0x20)
/*
* RFCSR 3:
@@ -2160,6 +2316,12 @@ struct mac_iveiv_entry {
rt2x00: mt7620: yet another beauty session So here is another round of improvements for MT7620 WiFi. This commit fixes a few significant issues related to TX_PWR_CFG_x and TX_ALC and also makes the code more readable by adding register descriptions for things added for MT7620 and use the usual bit-field access macros and the now defined macros instead of plain bit-ops and magic numbers. Properly describe EEPROM_TARGET_POWER at word 0x68 (== byte 0xD0) and thereby fix internal TXALC which would otherwise just read out-of-bounds of the EEPROM map. Split-out tx-power/ALC related stuff into an additional function. Fix VCO calibration, it was carried out properly in the channel switching but incomplete in the actual VCO calibration function. Also there is no need to trigger VCO calibration in channel switching, the VCO calibration function is already being called at this point. Remove it from channel switching function to avoid redundant code. The TX power calibration differs significantly from all other Mediatek/Ralink chips: They finally allow 0.5dB steps stored as 8-bit values for (almost) each bitrate -- and promptly ran out of space and for some reason didn't want to change the EEPROM layout. The hence opted for a scheme of sharing values for some adjecent bitrates and a highly over-complicated (or obfuscated?) way to populate the TX_PWR_CFG_x registers with the values stored in the EEPROM. The code here now looks much less complicated than what you see in the vendor's driver, however, it does the exact same thing: bGpwrdeltaMinus is a constant and always TRUE, hence half of the code was dead. Gpwrdelta is always 0 (rather than using the value read from the EEPROM). What remains is some very grotesque effort to avoid 0x20, probably some hardware bug related to some misunderstanding of what a singed 8-bit value is (imagine: if it was a signed 6-bit value then someone could believe that 0x20 == 0x0). And then they didn't clean it up once they later on anandonned that whole story of having a constant offset for 40 MHz channels and just set the offset to be constant 0 -- there is no effort for avoiding 0x20 for the 20 MHz values stored in the EEPROM, hence that's probably just a forbidden value in the EEPROM specs and won't appear anyway... Anyway, the whole thing felt like solving some college math test where in the end everything cancels out and the result equals 0 ;) To make sure that channel bandwidth power compensation really doesn't need to be taken care of, output a warning when the corresponding value stored in the EEPROM is non-zero. Also there is no apparent reason to refrain from initializing RFCSR register 13, it doesn't fail what-so-ever. Signed-off-by: Daniel Golle <daniel@makrotopia.org>
2017-02-13 05:25:35 +00:00
#define RFCSR3_BIT5 FIELD8(0x20)
/*
+ * RFCSR 4:
+ * VCOCAL_EN used by MT7620
+ */
+#define RFCSR4_VCOCAL_EN FIELD8(0x80)
+
+/*
* FRCSR 5:
*/
#define RFCSR5_R1 FIELD8(0x0c)
@@ -2214,6 +2376,7 @@ struct mac_iveiv_entry {
*/
#define RFCSR13_TX_POWER FIELD8(0x1f)
#define RFCSR13_DR0 FIELD8(0xe0)
+#define RFCSR13_RDIV_MT7620 FIELD8(0x03)
/*
* RFCSR 15:
@@ -2224,6 +2387,8 @@ struct mac_iveiv_entry {
* RFCSR 16:
*/
#define RFCSR16_TXMIXER_GAIN FIELD8(0x07)
+#define RFCSR16_RF_PLL_FREQ_SEL_MT7620 FIELD8(0x0F)
+#define RFCSR16_SDM_MODE_MT7620 FIELD8(0xE0)
/*
* RFCSR 17:
@@ -2236,6 +2401,8 @@ struct mac_iveiv_entry {
/* RFCSR 18 */
#define RFCSR18_XO_TUNE_BYPASS FIELD8(0x40)
+/* RFCSR 19 */
+#define RFCSR19_K FIELD8(0x03)
/*
* RFCSR 20:
@@ -2246,11 +2413,14 @@ struct mac_iveiv_entry {
* RFCSR 21:
*/
#define RFCSR21_RX_LO2_EN FIELD8(0x08)
+#define RFCSR21_BIT1 FIELD8(0x01)
+#define RFCSR21_BIT8 FIELD8(0x80)
/*
* RFCSR 22:
*/
#define RFCSR22_BASEBAND_LOOPBACK FIELD8(0x01)
+#define RFCSR22_FREQPLAN_D_MT7620 FIELD8(0x07)
/*
* RFCSR 23:
@@ -2273,6 +2443,11 @@ struct mac_iveiv_entry {
#define RFCSR27_R4 FIELD8(0x40)
/*
+ * RFCSR 28:
+ */
+#define RFCSR28_CH11_HT40 FIELD8(0x04)
+
+/*
* RFCSR 29:
*/
#define RFCSR29_ADC6_TEST FIELD8(0x01)
@@ -2333,6 +2508,7 @@ struct mac_iveiv_entry {
*/
#define RFCSR42_BIT1 FIELD8(0x01)
#define RFCSR42_BIT4 FIELD8(0x08)
+#define RFCSR42_TX2_EN_MT7620 FIELD8(0x40)
/*
* RFCSR 49:
@@ -2435,6 +2611,7 @@ enum rt2800_eeprom_word {
rt2x00: mt7620: yet another beauty session So here is another round of improvements for MT7620 WiFi. This commit fixes a few significant issues related to TX_PWR_CFG_x and TX_ALC and also makes the code more readable by adding register descriptions for things added for MT7620 and use the usual bit-field access macros and the now defined macros instead of plain bit-ops and magic numbers. Properly describe EEPROM_TARGET_POWER at word 0x68 (== byte 0xD0) and thereby fix internal TXALC which would otherwise just read out-of-bounds of the EEPROM map. Split-out tx-power/ALC related stuff into an additional function. Fix VCO calibration, it was carried out properly in the channel switching but incomplete in the actual VCO calibration function. Also there is no need to trigger VCO calibration in channel switching, the VCO calibration function is already being called at this point. Remove it from channel switching function to avoid redundant code. The TX power calibration differs significantly from all other Mediatek/Ralink chips: They finally allow 0.5dB steps stored as 8-bit values for (almost) each bitrate -- and promptly ran out of space and for some reason didn't want to change the EEPROM layout. The hence opted for a scheme of sharing values for some adjecent bitrates and a highly over-complicated (or obfuscated?) way to populate the TX_PWR_CFG_x registers with the values stored in the EEPROM. The code here now looks much less complicated than what you see in the vendor's driver, however, it does the exact same thing: bGpwrdeltaMinus is a constant and always TRUE, hence half of the code was dead. Gpwrdelta is always 0 (rather than using the value read from the EEPROM). What remains is some very grotesque effort to avoid 0x20, probably some hardware bug related to some misunderstanding of what a singed 8-bit value is (imagine: if it was a signed 6-bit value then someone could believe that 0x20 == 0x0). And then they didn't clean it up once they later on anandonned that whole story of having a constant offset for 40 MHz channels and just set the offset to be constant 0 -- there is no effort for avoiding 0x20 for the 20 MHz values stored in the EEPROM, hence that's probably just a forbidden value in the EEPROM specs and won't appear anyway... Anyway, the whole thing felt like solving some college math test where in the end everything cancels out and the result equals 0 ;) To make sure that channel bandwidth power compensation really doesn't need to be taken care of, output a warning when the corresponding value stored in the EEPROM is non-zero. Also there is no apparent reason to refrain from initializing RFCSR register 13, it doesn't fail what-so-ever. Signed-off-by: Daniel Golle <daniel@makrotopia.org>
2017-02-13 05:25:35 +00:00
EEPROM_TSSI_BOUND_BG5,
EEPROM_TXPOWER_A1,
EEPROM_TXPOWER_A2,
+ EEPROM_TXPOWER_INIT,
EEPROM_TSSI_BOUND_A1,
EEPROM_TSSI_BOUND_A2,
EEPROM_TSSI_BOUND_A3,
diff --git a/drivers/net/wireless/ralink/rt2x00/rt2800lib.c b/drivers/net/wireless/ralink/rt2x00/rt2800lib.c
index 8d00c599e47a..201b12ed90c6 100644
--- a/drivers/net/wireless/ralink/rt2x00/rt2800lib.c
+++ b/drivers/net/wireless/ralink/rt2x00/rt2800lib.c
@@ -59,6 +59,9 @@
rt2800_regbusy_read((__dev), BBP_CSR_CFG, BBP_CSR_CFG_BUSY, (__reg))
#define WAIT_FOR_RFCSR(__dev, __reg) \
rt2800_regbusy_read((__dev), RF_CSR_CFG, RF_CSR_CFG_BUSY, (__reg))
+#define WAIT_FOR_RFCSR_MT7620(__dev, __reg) \
rt2x00: mt7620: lots of improvements This commit combines all the changes I've made on my staging tree into a single commit fixing many issues with our patch for MT7620. First of all, checkpatch.pl revealed numerous code style issues with the patch, so fix all the white-space and commets. Also use usleep_range instead of legacy timing and relax timing for VCO calibration just like the vendor driver does. Several line programming registers were commented out in the patch. Originally this came from the features present but disabled by default in the vendor's driver (RTMP_TEMPERATURE_CALIBRATION and ADJUST_POWER_CONSUMPTION_SUPPORT). Remove the dead code for now, it can easily be re-added if we actually intend to support those features. Move values from mt7620_freqconfig type into the existing rf_channel struct, this shouldn't be a new typedef and it is possible to use the existing struct because rf_channel got 4 32-bit fields, so two of the 8-bit values from mt7620_freqconfig can easily be stored in the same 32-bit field. Map values such that Rdiv -> rf1 N -> rf2 K -> rf3[0:7] D -> rf3[8:15] Ksd -> rf4 This makes the channel switching logic already look a bit more like what we are used to in rt2x00... Probably many of the read-modify-write calls could still be replaced by macros intended for that. iq calibration seems to be identical to RT5592, so just enable it. Test shows that this improves things quite a lot, datarates went up by a couple of megabits when running iperf, signal quality seems jumpy in the first few seconds once a station connencts, the stabelizes on a value significantly better than what it was before. Add description to the patch and reference the original OpenWrt commit by which it was added. The patch now passes checkpatch.pl and can thus be discussed with the upstream authors of the rt2x00 driver. Funded-by: https://www.kickstarter.com/projects/1327597961/better-support-for-mt7620a-n-in-openwrt-lede/ Signed-off-by: Daniel Golle <daniel@makrotopia.org>
2017-02-03 14:02:32 +00:00
+ rt2800_regbusy_read((__dev), RF_CSR_CFG, RF_CSR_CFG_BUSY_MT7620, \
+ (__reg))
#define WAIT_FOR_RF(__dev, __reg) \
rt2800_regbusy_read((__dev), RF_CSR_CFG0, RF_CSR_CFG0_BUSY, (__reg))
#define WAIT_FOR_MCU(__dev, __reg) \
@@ -150,19 +153,56 @@ static void rt2800_rfcsr_write(struct rt2x00_dev *rt2x00dev,
* Wait until the RFCSR becomes available, afterwards we
* can safely write the new data into the register.
*/
- if (WAIT_FOR_RFCSR(rt2x00dev, &reg)) {
- reg = 0;
- rt2x00_set_field32(&reg, RF_CSR_CFG_DATA, value);
- rt2x00_set_field32(&reg, RF_CSR_CFG_REGNUM, word);
- rt2x00_set_field32(&reg, RF_CSR_CFG_WRITE, 1);
- rt2x00_set_field32(&reg, RF_CSR_CFG_BUSY, 1);
+ switch (rt2x00dev->chip.rt) {
+ case RT6352:
+ if (WAIT_FOR_RFCSR_MT7620(rt2x00dev, &reg)) {
+ reg = 0;
+ rt2x00_set_field32(&reg, RF_CSR_CFG_DATA_MT7620, value);
rt2x00: mt7620: lots of improvements This commit combines all the changes I've made on my staging tree into a single commit fixing many issues with our patch for MT7620. First of all, checkpatch.pl revealed numerous code style issues with the patch, so fix all the white-space and commets. Also use usleep_range instead of legacy timing and relax timing for VCO calibration just like the vendor driver does. Several line programming registers were commented out in the patch. Originally this came from the features present but disabled by default in the vendor's driver (RTMP_TEMPERATURE_CALIBRATION and ADJUST_POWER_CONSUMPTION_SUPPORT). Remove the dead code for now, it can easily be re-added if we actually intend to support those features. Move values from mt7620_freqconfig type into the existing rf_channel struct, this shouldn't be a new typedef and it is possible to use the existing struct because rf_channel got 4 32-bit fields, so two of the 8-bit values from mt7620_freqconfig can easily be stored in the same 32-bit field. Map values such that Rdiv -> rf1 N -> rf2 K -> rf3[0:7] D -> rf3[8:15] Ksd -> rf4 This makes the channel switching logic already look a bit more like what we are used to in rt2x00... Probably many of the read-modify-write calls could still be replaced by macros intended for that. iq calibration seems to be identical to RT5592, so just enable it. Test shows that this improves things quite a lot, datarates went up by a couple of megabits when running iperf, signal quality seems jumpy in the first few seconds once a station connencts, the stabelizes on a value significantly better than what it was before. Add description to the patch and reference the original OpenWrt commit by which it was added. The patch now passes checkpatch.pl and can thus be discussed with the upstream authors of the rt2x00 driver. Funded-by: https://www.kickstarter.com/projects/1327597961/better-support-for-mt7620a-n-in-openwrt-lede/ Signed-off-by: Daniel Golle <daniel@makrotopia.org>
2017-02-03 14:02:32 +00:00
+ rt2x00_set_field32(&reg, RF_CSR_CFG_REGNUM_MT7620,
+ word);
+ rt2x00_set_field32(&reg, RF_CSR_CFG_WRITE_MT7620, 1);
+ rt2x00_set_field32(&reg, RF_CSR_CFG_BUSY_MT7620, 1);
rt2x00: mt7620: lots of improvements This commit combines all the changes I've made on my staging tree into a single commit fixing many issues with our patch for MT7620. First of all, checkpatch.pl revealed numerous code style issues with the patch, so fix all the white-space and commets. Also use usleep_range instead of legacy timing and relax timing for VCO calibration just like the vendor driver does. Several line programming registers were commented out in the patch. Originally this came from the features present but disabled by default in the vendor's driver (RTMP_TEMPERATURE_CALIBRATION and ADJUST_POWER_CONSUMPTION_SUPPORT). Remove the dead code for now, it can easily be re-added if we actually intend to support those features. Move values from mt7620_freqconfig type into the existing rf_channel struct, this shouldn't be a new typedef and it is possible to use the existing struct because rf_channel got 4 32-bit fields, so two of the 8-bit values from mt7620_freqconfig can easily be stored in the same 32-bit field. Map values such that Rdiv -> rf1 N -> rf2 K -> rf3[0:7] D -> rf3[8:15] Ksd -> rf4 This makes the channel switching logic already look a bit more like what we are used to in rt2x00... Probably many of the read-modify-write calls could still be replaced by macros intended for that. iq calibration seems to be identical to RT5592, so just enable it. Test shows that this improves things quite a lot, datarates went up by a couple of megabits when running iperf, signal quality seems jumpy in the first few seconds once a station connencts, the stabelizes on a value significantly better than what it was before. Add description to the patch and reference the original OpenWrt commit by which it was added. The patch now passes checkpatch.pl and can thus be discussed with the upstream authors of the rt2x00 driver. Funded-by: https://www.kickstarter.com/projects/1327597961/better-support-for-mt7620a-n-in-openwrt-lede/ Signed-off-by: Daniel Golle <daniel@makrotopia.org>
2017-02-03 14:02:32 +00:00
+
+ rt2800_register_write_lock(rt2x00dev, RF_CSR_CFG, reg);
+ }
+ break;
- rt2800_register_write_lock(rt2x00dev, RF_CSR_CFG, reg);
+ default:
+ if (WAIT_FOR_RFCSR(rt2x00dev, &reg)) {
+ reg = 0;
+ rt2x00_set_field32(&reg, RF_CSR_CFG_DATA, value);
+ rt2x00_set_field32(&reg, RF_CSR_CFG_REGNUM, word);
+ rt2x00_set_field32(&reg, RF_CSR_CFG_WRITE, 1);
+ rt2x00_set_field32(&reg, RF_CSR_CFG_BUSY, 1);
+
+ rt2800_register_write_lock(rt2x00dev, RF_CSR_CFG, reg);
+ }
+ break;
}
mutex_unlock(&rt2x00dev->csr_mutex);
}
+static void rt2800_rfcsr_write_bank(struct rt2x00_dev *rt2x00dev, const u8 bank,
+ const unsigned int reg, const u8 value)
+{
+ rt2800_rfcsr_write(rt2x00dev, (reg | (bank << 6)), value);
+}
+
+static void rt2800_rfcsr_write_chanreg(struct rt2x00_dev *rt2x00dev,
+ const unsigned int reg, const u8 value)
+{
+ rt2800_rfcsr_write_bank(rt2x00dev, 4, reg, value);
+ rt2800_rfcsr_write_bank(rt2x00dev, 6, reg, value);
+}
+
+static void rt2800_rfcsr_write_dccal(struct rt2x00_dev *rt2x00dev,
+ const unsigned int reg, const u8 value)
+{
+ rt2800_rfcsr_write_bank(rt2x00dev, 5, reg, value);
+ rt2800_rfcsr_write_bank(rt2x00dev, 7, reg, value);
+}
+
static void rt2800_rfcsr_read(struct rt2x00_dev *rt2x00dev,
const unsigned int word, u8 *value)
{
@@ -178,22 +218,48 @@ static void rt2800_rfcsr_read(struct rt2x00_dev *rt2x00dev,
* doesn't become available in time, reg will be 0xffffffff
* which means we return 0xff to the caller.
*/
- if (WAIT_FOR_RFCSR(rt2x00dev, &reg)) {
- reg = 0;
- rt2x00_set_field32(&reg, RF_CSR_CFG_REGNUM, word);
- rt2x00_set_field32(&reg, RF_CSR_CFG_WRITE, 0);
- rt2x00_set_field32(&reg, RF_CSR_CFG_BUSY, 1);
+ switch (rt2x00dev->chip.rt) {
+ case RT6352:
+ if (WAIT_FOR_RFCSR_MT7620(rt2x00dev, &reg)) {
+ reg = 0;
rt2x00: mt7620: lots of improvements This commit combines all the changes I've made on my staging tree into a single commit fixing many issues with our patch for MT7620. First of all, checkpatch.pl revealed numerous code style issues with the patch, so fix all the white-space and commets. Also use usleep_range instead of legacy timing and relax timing for VCO calibration just like the vendor driver does. Several line programming registers were commented out in the patch. Originally this came from the features present but disabled by default in the vendor's driver (RTMP_TEMPERATURE_CALIBRATION and ADJUST_POWER_CONSUMPTION_SUPPORT). Remove the dead code for now, it can easily be re-added if we actually intend to support those features. Move values from mt7620_freqconfig type into the existing rf_channel struct, this shouldn't be a new typedef and it is possible to use the existing struct because rf_channel got 4 32-bit fields, so two of the 8-bit values from mt7620_freqconfig can easily be stored in the same 32-bit field. Map values such that Rdiv -> rf1 N -> rf2 K -> rf3[0:7] D -> rf3[8:15] Ksd -> rf4 This makes the channel switching logic already look a bit more like what we are used to in rt2x00... Probably many of the read-modify-write calls could still be replaced by macros intended for that. iq calibration seems to be identical to RT5592, so just enable it. Test shows that this improves things quite a lot, datarates went up by a couple of megabits when running iperf, signal quality seems jumpy in the first few seconds once a station connencts, the stabelizes on a value significantly better than what it was before. Add description to the patch and reference the original OpenWrt commit by which it was added. The patch now passes checkpatch.pl and can thus be discussed with the upstream authors of the rt2x00 driver. Funded-by: https://www.kickstarter.com/projects/1327597961/better-support-for-mt7620a-n-in-openwrt-lede/ Signed-off-by: Daniel Golle <daniel@makrotopia.org>
2017-02-03 14:02:32 +00:00
+ rt2x00_set_field32(&reg, RF_CSR_CFG_REGNUM_MT7620,
+ word);
+ rt2x00_set_field32(&reg, RF_CSR_CFG_WRITE_MT7620, 0);
+ rt2x00_set_field32(&reg, RF_CSR_CFG_BUSY_MT7620, 1);
- rt2800_register_write_lock(rt2x00dev, RF_CSR_CFG, reg);
+ rt2800_register_write_lock(rt2x00dev, RF_CSR_CFG, reg);
- WAIT_FOR_RFCSR(rt2x00dev, &reg);
- }
+ WAIT_FOR_RFCSR_MT7620(rt2x00dev, &reg);
+ }
- *value = rt2x00_get_field32(reg, RF_CSR_CFG_DATA);
+ *value = rt2x00_get_field32(reg, RF_CSR_CFG_DATA_MT7620);
+ break;
+
+ default:
+ if (WAIT_FOR_RFCSR(rt2x00dev, &reg)) {
+ reg = 0;
+ rt2x00_set_field32(&reg, RF_CSR_CFG_REGNUM, word);
+ rt2x00_set_field32(&reg, RF_CSR_CFG_WRITE, 0);
+ rt2x00_set_field32(&reg, RF_CSR_CFG_BUSY, 1);
rt2x00: mt7620: yet another beauty session So here is another round of improvements for MT7620 WiFi. This commit fixes a few significant issues related to TX_PWR_CFG_x and TX_ALC and also makes the code more readable by adding register descriptions for things added for MT7620 and use the usual bit-field access macros and the now defined macros instead of plain bit-ops and magic numbers. Properly describe EEPROM_TARGET_POWER at word 0x68 (== byte 0xD0) and thereby fix internal TXALC which would otherwise just read out-of-bounds of the EEPROM map. Split-out tx-power/ALC related stuff into an additional function. Fix VCO calibration, it was carried out properly in the channel switching but incomplete in the actual VCO calibration function. Also there is no need to trigger VCO calibration in channel switching, the VCO calibration function is already being called at this point. Remove it from channel switching function to avoid redundant code. The TX power calibration differs significantly from all other Mediatek/Ralink chips: They finally allow 0.5dB steps stored as 8-bit values for (almost) each bitrate -- and promptly ran out of space and for some reason didn't want to change the EEPROM layout. The hence opted for a scheme of sharing values for some adjecent bitrates and a highly over-complicated (or obfuscated?) way to populate the TX_PWR_CFG_x registers with the values stored in the EEPROM. The code here now looks much less complicated than what you see in the vendor's driver, however, it does the exact same thing: bGpwrdeltaMinus is a constant and always TRUE, hence half of the code was dead. Gpwrdelta is always 0 (rather than using the value read from the EEPROM). What remains is some very grotesque effort to avoid 0x20, probably some hardware bug related to some misunderstanding of what a singed 8-bit value is (imagine: if it was a signed 6-bit value then someone could believe that 0x20 == 0x0). And then they didn't clean it up once they later on anandonned that whole story of having a constant offset for 40 MHz channels and just set the offset to be constant 0 -- there is no effort for avoiding 0x20 for the 20 MHz values stored in the EEPROM, hence that's probably just a forbidden value in the EEPROM specs and won't appear anyway... Anyway, the whole thing felt like solving some college math test where in the end everything cancels out and the result equals 0 ;) To make sure that channel bandwidth power compensation really doesn't need to be taken care of, output a warning when the corresponding value stored in the EEPROM is non-zero. Also there is no apparent reason to refrain from initializing RFCSR register 13, it doesn't fail what-so-ever. Signed-off-by: Daniel Golle <daniel@makrotopia.org>
2017-02-13 05:25:35 +00:00
+
+ rt2800_register_write_lock(rt2x00dev, RF_CSR_CFG, reg);
+
+ WAIT_FOR_RFCSR(rt2x00dev, &reg);
+ }
+
+ *value = rt2x00_get_field32(reg, RF_CSR_CFG_DATA);
+ break;
+ }
mutex_unlock(&rt2x00dev->csr_mutex);
}
+static void rt2800_rfcsr_read_bank(struct rt2x00_dev *rt2x00dev, const u8 bank,
+ const unsigned int reg, u8 *value)
+{
+ rt2800_rfcsr_read(rt2x00dev, (reg | (bank << 6)), value);
+}
+
static void rt2800_rf_write(struct rt2x00_dev *rt2x00dev,
const unsigned int word, const u32 value)
{
@@ -250,6 +316,7 @@ static const unsigned int rt2800_eeprom_map[EEPROM_WORD_COUNT] = {
rt2x00: mt7620: yet another beauty session So here is another round of improvements for MT7620 WiFi. This commit fixes a few significant issues related to TX_PWR_CFG_x and TX_ALC and also makes the code more readable by adding register descriptions for things added for MT7620 and use the usual bit-field access macros and the now defined macros instead of plain bit-ops and magic numbers. Properly describe EEPROM_TARGET_POWER at word 0x68 (== byte 0xD0) and thereby fix internal TXALC which would otherwise just read out-of-bounds of the EEPROM map. Split-out tx-power/ALC related stuff into an additional function. Fix VCO calibration, it was carried out properly in the channel switching but incomplete in the actual VCO calibration function. Also there is no need to trigger VCO calibration in channel switching, the VCO calibration function is already being called at this point. Remove it from channel switching function to avoid redundant code. The TX power calibration differs significantly from all other Mediatek/Ralink chips: They finally allow 0.5dB steps stored as 8-bit values for (almost) each bitrate -- and promptly ran out of space and for some reason didn't want to change the EEPROM layout. The hence opted for a scheme of sharing values for some adjecent bitrates and a highly over-complicated (or obfuscated?) way to populate the TX_PWR_CFG_x registers with the values stored in the EEPROM. The code here now looks much less complicated than what you see in the vendor's driver, however, it does the exact same thing: bGpwrdeltaMinus is a constant and always TRUE, hence half of the code was dead. Gpwrdelta is always 0 (rather than using the value read from the EEPROM). What remains is some very grotesque effort to avoid 0x20, probably some hardware bug related to some misunderstanding of what a singed 8-bit value is (imagine: if it was a signed 6-bit value then someone could believe that 0x20 == 0x0). And then they didn't clean it up once they later on anandonned that whole story of having a constant offset for 40 MHz channels and just set the offset to be constant 0 -- there is no effort for avoiding 0x20 for the 20 MHz values stored in the EEPROM, hence that's probably just a forbidden value in the EEPROM specs and won't appear anyway... Anyway, the whole thing felt like solving some college math test where in the end everything cancels out and the result equals 0 ;) To make sure that channel bandwidth power compensation really doesn't need to be taken care of, output a warning when the corresponding value stored in the EEPROM is non-zero. Also there is no apparent reason to refrain from initializing RFCSR register 13, it doesn't fail what-so-ever. Signed-off-by: Daniel Golle <daniel@makrotopia.org>
2017-02-13 05:25:35 +00:00
[EEPROM_TSSI_BOUND_BG5] = 0x003b,
[EEPROM_TXPOWER_A1] = 0x003c,
[EEPROM_TXPOWER_A2] = 0x0053,
+ [EEPROM_TXPOWER_INIT] = 0x0068,
[EEPROM_TSSI_BOUND_A1] = 0x006a,
[EEPROM_TSSI_BOUND_A2] = 0x006b,
[EEPROM_TSSI_BOUND_A3] = 0x006c,
@@ -524,6 +591,7 @@ void rt2800_get_txwi_rxwi_size(struct rt2x00_dev *rt2x00dev,
break;
case RT5592:
+ case RT6352:
*txwi_size = TXWI_DESC_SIZE_5WORDS;
*rxwi_size = RXWI_DESC_SIZE_6WORDS;
break;
@@ -2810,7 +2878,8 @@ static void rt2800_config_channel_rf53xx(struct rt2x00_dev *rt2x00dev,
rt2800_rfcsr_write(rt2x00dev, 59,
r59_nonbt_rev[idx]);
} else if (rt2x00_rt(rt2x00dev, RT5390) ||
- rt2x00_rt(rt2x00dev, RT5392)) {
+ rt2x00_rt(rt2x00dev, RT5392) ||
+ rt2x00_rt(rt2x00dev, RT6352)) {
static const char r59_non_bt[] = {0x8f, 0x8f,
0x8f, 0x8f, 0x8f, 0x8f, 0x8f, 0x8d,
0x8a, 0x88, 0x88, 0x87, 0x87, 0x86};
@@ -3104,6 +3173,242 @@ static void rt2800_config_channel_rf55xx(struct rt2x00_dev *rt2x00dev,
rt2800_bbp_write(rt2x00dev, 196, (rf->channel <= 14) ? 0x19 : 0x7F);
}
+static void rt2800_config_channel_rf7620(struct rt2x00_dev *rt2x00dev,
+ struct ieee80211_conf *conf,
+ struct rf_channel *rf,
+ struct channel_info *info)
+{
+ struct rt2800_drv_data *drv_data = rt2x00dev->drv_data;
+ u8 rx_agc_fc, tx_agc_fc;
+ u8 rfcsr;
+
+ /* Frequeny plan setting */
+ /* Rdiv setting (set 0x03 if Xtal==20)
rt2x00: mt7620: lots of improvements This commit combines all the changes I've made on my staging tree into a single commit fixing many issues with our patch for MT7620. First of all, checkpatch.pl revealed numerous code style issues with the patch, so fix all the white-space and commets. Also use usleep_range instead of legacy timing and relax timing for VCO calibration just like the vendor driver does. Several line programming registers were commented out in the patch. Originally this came from the features present but disabled by default in the vendor's driver (RTMP_TEMPERATURE_CALIBRATION and ADJUST_POWER_CONSUMPTION_SUPPORT). Remove the dead code for now, it can easily be re-added if we actually intend to support those features. Move values from mt7620_freqconfig type into the existing rf_channel struct, this shouldn't be a new typedef and it is possible to use the existing struct because rf_channel got 4 32-bit fields, so two of the 8-bit values from mt7620_freqconfig can easily be stored in the same 32-bit field. Map values such that Rdiv -> rf1 N -> rf2 K -> rf3[0:7] D -> rf3[8:15] Ksd -> rf4 This makes the channel switching logic already look a bit more like what we are used to in rt2x00... Probably many of the read-modify-write calls could still be replaced by macros intended for that. iq calibration seems to be identical to RT5592, so just enable it. Test shows that this improves things quite a lot, datarates went up by a couple of megabits when running iperf, signal quality seems jumpy in the first few seconds once a station connencts, the stabelizes on a value significantly better than what it was before. Add description to the patch and reference the original OpenWrt commit by which it was added. The patch now passes checkpatch.pl and can thus be discussed with the upstream authors of the rt2x00 driver. Funded-by: https://www.kickstarter.com/projects/1327597961/better-support-for-mt7620a-n-in-openwrt-lede/ Signed-off-by: Daniel Golle <daniel@makrotopia.org>
2017-02-03 14:02:32 +00:00
+ * R13[1:0]
+ */
+ rt2800_rfcsr_read(rt2x00dev, 13, &rfcsr);
+ rt2x00_set_field8(&rfcsr, RFCSR13_RDIV_MT7620,
+ rt2800_clk_is_20mhz(rt2x00dev) ? 3 : 0);
+ rt2800_rfcsr_write(rt2x00dev, 13, rfcsr);
+
+ /* N setting
+ * R20[7:0] in rf->rf1
+ * R21[0] always 0
+ */
+ rt2800_rfcsr_read(rt2x00dev, 20, &rfcsr);
+ rfcsr = (rf->rf1 & 0x00ff);
+ rt2800_rfcsr_write(rt2x00dev, 20, rfcsr);
+
+ rt2800_rfcsr_read(rt2x00dev, 21, &rfcsr);
+ rt2x00_set_field8(&rfcsr, RFCSR21_BIT1, 0);
+ rt2800_rfcsr_write(rt2x00dev, 21, rfcsr);
+
+ /* K setting (always 0)
+ * R16[3:0] (RF PLL freq selection)
+ */
+ rt2800_rfcsr_read(rt2x00dev, 16, &rfcsr);
+ rt2x00_set_field8(&rfcsr, RFCSR16_RF_PLL_FREQ_SEL_MT7620, 0);
+ rt2800_rfcsr_write(rt2x00dev, 16, rfcsr);
+
+ /* D setting (always 0)
+ * R22[2:0] (D=15, R22[2:0]=<111>)
+ */
+ rt2800_rfcsr_read(rt2x00dev, 22, &rfcsr);
+ rt2x00_set_field8(&rfcsr, RFCSR22_FREQPLAN_D_MT7620, 0);
+ rt2800_rfcsr_write(rt2x00dev, 22, rfcsr);
+
+ /* Ksd setting
+ * Ksd: R17<7:0> in rf->rf2
+ * R18<7:0> in rf->rf3
+ * R19<1:0> in rf->rf4
+ */
+ rt2800_rfcsr_read(rt2x00dev, 17, &rfcsr);
+ rfcsr = rf->rf2;
+ rt2800_rfcsr_write(rt2x00dev, 17, rfcsr);
+
+ rt2800_rfcsr_read(rt2x00dev, 18, &rfcsr);
+ rfcsr = rf->rf3;
+ rt2800_rfcsr_write(rt2x00dev, 18, rfcsr);
+
+ rt2800_rfcsr_read(rt2x00dev, 19, &rfcsr);
+ rt2x00_set_field8(&rfcsr, RFCSR19_K, rf->rf4);
+ rt2800_rfcsr_write(rt2x00dev, 19, rfcsr);
+
+ /* Default: XO=20MHz , SDM mode */
+ rt2800_rfcsr_read(rt2x00dev, 16, &rfcsr);
+ rt2x00_set_field8(&rfcsr, RFCSR16_SDM_MODE_MT7620, 0x80);
+ rt2800_rfcsr_write(rt2x00dev, 16, rfcsr);
+
+ rt2800_rfcsr_read(rt2x00dev, 21, &rfcsr);
+ rt2x00_set_field8(&rfcsr, RFCSR21_BIT8, 1);
+ rt2800_rfcsr_write(rt2x00dev, 21, rfcsr);
+
+ rt2800_rfcsr_read(rt2x00dev, 1, &rfcsr);
+ rt2x00_set_field8(&rfcsr, RFCSR1_TX2_EN_MT7620,
+ rt2x00dev->default_ant.tx_chain_num != 1);
+ rt2800_rfcsr_write(rt2x00dev, 1, rfcsr);
+
+ rt2800_rfcsr_read(rt2x00dev, 2, &rfcsr);
+ rt2x00_set_field8(&rfcsr, RFCSR2_TX2_EN_MT7620,
+ rt2x00dev->default_ant.tx_chain_num != 1);
+ rt2x00_set_field8(&rfcsr, RFCSR2_RX2_EN_MT7620,
+ rt2x00dev->default_ant.rx_chain_num != 1);
+ rt2800_rfcsr_write(rt2x00dev, 2, rfcsr);
+
+ rt2800_rfcsr_read(rt2x00dev, 42, &rfcsr);
+ rt2x00_set_field8(&rfcsr, RFCSR42_TX2_EN_MT7620,
+ rt2x00dev->default_ant.tx_chain_num != 1);
+ rt2800_rfcsr_write(rt2x00dev, 42, rfcsr);
+
+ /* RF for DC Cal BW */
+ if (conf_is_ht40(conf)) {
+ rt2800_rfcsr_write_dccal(rt2x00dev, 6, 0x10);
+ rt2800_rfcsr_write_dccal(rt2x00dev, 7, 0x10);
+ rt2800_rfcsr_write_dccal(rt2x00dev, 8, 0x04);
+ rt2800_rfcsr_write_dccal(rt2x00dev, 58, 0x10);
+ rt2800_rfcsr_write_dccal(rt2x00dev, 59, 0x10);
+ } else {
+ rt2800_rfcsr_write_dccal(rt2x00dev, 6, 0x20);
+ rt2800_rfcsr_write_dccal(rt2x00dev, 7, 0x20);
+ rt2800_rfcsr_write_dccal(rt2x00dev, 8, 0x00);
+ rt2800_rfcsr_write_dccal(rt2x00dev, 58, 0x20);
+ rt2800_rfcsr_write_dccal(rt2x00dev, 59, 0x20);
+ }
+
+ if (conf_is_ht40(conf)) {
+ rt2800_rfcsr_write_dccal(rt2x00dev, 58, 0x08);
+ rt2800_rfcsr_write_dccal(rt2x00dev, 59, 0x08);
+ } else {
+ rt2800_rfcsr_write_dccal(rt2x00dev, 58, 0x28);
+ rt2800_rfcsr_write_dccal(rt2x00dev, 59, 0x28);
+ }
+
+ rt2800_rfcsr_read(rt2x00dev, 28, &rfcsr);
+ rt2x00_set_field8(&rfcsr, RFCSR28_CH11_HT40,
+ conf_is_ht40(conf) && (rf->channel == 11));
+ rt2800_rfcsr_write(rt2x00dev, 28, rfcsr);
+
rt2x00: mt7620: yet another beauty session So here is another round of improvements for MT7620 WiFi. This commit fixes a few significant issues related to TX_PWR_CFG_x and TX_ALC and also makes the code more readable by adding register descriptions for things added for MT7620 and use the usual bit-field access macros and the now defined macros instead of plain bit-ops and magic numbers. Properly describe EEPROM_TARGET_POWER at word 0x68 (== byte 0xD0) and thereby fix internal TXALC which would otherwise just read out-of-bounds of the EEPROM map. Split-out tx-power/ALC related stuff into an additional function. Fix VCO calibration, it was carried out properly in the channel switching but incomplete in the actual VCO calibration function. Also there is no need to trigger VCO calibration in channel switching, the VCO calibration function is already being called at this point. Remove it from channel switching function to avoid redundant code. The TX power calibration differs significantly from all other Mediatek/Ralink chips: They finally allow 0.5dB steps stored as 8-bit values for (almost) each bitrate -- and promptly ran out of space and for some reason didn't want to change the EEPROM layout. The hence opted for a scheme of sharing values for some adjecent bitrates and a highly over-complicated (or obfuscated?) way to populate the TX_PWR_CFG_x registers with the values stored in the EEPROM. The code here now looks much less complicated than what you see in the vendor's driver, however, it does the exact same thing: bGpwrdeltaMinus is a constant and always TRUE, hence half of the code was dead. Gpwrdelta is always 0 (rather than using the value read from the EEPROM). What remains is some very grotesque effort to avoid 0x20, probably some hardware bug related to some misunderstanding of what a singed 8-bit value is (imagine: if it was a signed 6-bit value then someone could believe that 0x20 == 0x0). And then they didn't clean it up once they later on anandonned that whole story of having a constant offset for 40 MHz channels and just set the offset to be constant 0 -- there is no effort for avoiding 0x20 for the 20 MHz values stored in the EEPROM, hence that's probably just a forbidden value in the EEPROM specs and won't appear anyway... Anyway, the whole thing felt like solving some college math test where in the end everything cancels out and the result equals 0 ;) To make sure that channel bandwidth power compensation really doesn't need to be taken care of, output a warning when the corresponding value stored in the EEPROM is non-zero. Also there is no apparent reason to refrain from initializing RFCSR register 13, it doesn't fail what-so-ever. Signed-off-by: Daniel Golle <daniel@makrotopia.org>
2017-02-13 05:25:35 +00:00
+ if (!test_bit(DEVICE_STATE_SCANNING, &rt2x00dev->flags)) {
+ if (conf_is_ht40(conf)) {
+ rx_agc_fc = drv_data->rx_calibration_bw40;
+ tx_agc_fc = drv_data->tx_calibration_bw40;
rt2x00: mt7620: yet another beauty session So here is another round of improvements for MT7620 WiFi. This commit fixes a few significant issues related to TX_PWR_CFG_x and TX_ALC and also makes the code more readable by adding register descriptions for things added for MT7620 and use the usual bit-field access macros and the now defined macros instead of plain bit-ops and magic numbers. Properly describe EEPROM_TARGET_POWER at word 0x68 (== byte 0xD0) and thereby fix internal TXALC which would otherwise just read out-of-bounds of the EEPROM map. Split-out tx-power/ALC related stuff into an additional function. Fix VCO calibration, it was carried out properly in the channel switching but incomplete in the actual VCO calibration function. Also there is no need to trigger VCO calibration in channel switching, the VCO calibration function is already being called at this point. Remove it from channel switching function to avoid redundant code. The TX power calibration differs significantly from all other Mediatek/Ralink chips: They finally allow 0.5dB steps stored as 8-bit values for (almost) each bitrate -- and promptly ran out of space and for some reason didn't want to change the EEPROM layout. The hence opted for a scheme of sharing values for some adjecent bitrates and a highly over-complicated (or obfuscated?) way to populate the TX_PWR_CFG_x registers with the values stored in the EEPROM. The code here now looks much less complicated than what you see in the vendor's driver, however, it does the exact same thing: bGpwrdeltaMinus is a constant and always TRUE, hence half of the code was dead. Gpwrdelta is always 0 (rather than using the value read from the EEPROM). What remains is some very grotesque effort to avoid 0x20, probably some hardware bug related to some misunderstanding of what a singed 8-bit value is (imagine: if it was a signed 6-bit value then someone could believe that 0x20 == 0x0). And then they didn't clean it up once they later on anandonned that whole story of having a constant offset for 40 MHz channels and just set the offset to be constant 0 -- there is no effort for avoiding 0x20 for the 20 MHz values stored in the EEPROM, hence that's probably just a forbidden value in the EEPROM specs and won't appear anyway... Anyway, the whole thing felt like solving some college math test where in the end everything cancels out and the result equals 0 ;) To make sure that channel bandwidth power compensation really doesn't need to be taken care of, output a warning when the corresponding value stored in the EEPROM is non-zero. Also there is no apparent reason to refrain from initializing RFCSR register 13, it doesn't fail what-so-ever. Signed-off-by: Daniel Golle <daniel@makrotopia.org>
2017-02-13 05:25:35 +00:00
+ } else {
+ rx_agc_fc = drv_data->rx_calibration_bw20;
+ tx_agc_fc = drv_data->tx_calibration_bw20;
rt2x00: mt7620: yet another beauty session So here is another round of improvements for MT7620 WiFi. This commit fixes a few significant issues related to TX_PWR_CFG_x and TX_ALC and also makes the code more readable by adding register descriptions for things added for MT7620 and use the usual bit-field access macros and the now defined macros instead of plain bit-ops and magic numbers. Properly describe EEPROM_TARGET_POWER at word 0x68 (== byte 0xD0) and thereby fix internal TXALC which would otherwise just read out-of-bounds of the EEPROM map. Split-out tx-power/ALC related stuff into an additional function. Fix VCO calibration, it was carried out properly in the channel switching but incomplete in the actual VCO calibration function. Also there is no need to trigger VCO calibration in channel switching, the VCO calibration function is already being called at this point. Remove it from channel switching function to avoid redundant code. The TX power calibration differs significantly from all other Mediatek/Ralink chips: They finally allow 0.5dB steps stored as 8-bit values for (almost) each bitrate -- and promptly ran out of space and for some reason didn't want to change the EEPROM layout. The hence opted for a scheme of sharing values for some adjecent bitrates and a highly over-complicated (or obfuscated?) way to populate the TX_PWR_CFG_x registers with the values stored in the EEPROM. The code here now looks much less complicated than what you see in the vendor's driver, however, it does the exact same thing: bGpwrdeltaMinus is a constant and always TRUE, hence half of the code was dead. Gpwrdelta is always 0 (rather than using the value read from the EEPROM). What remains is some very grotesque effort to avoid 0x20, probably some hardware bug related to some misunderstanding of what a singed 8-bit value is (imagine: if it was a signed 6-bit value then someone could believe that 0x20 == 0x0). And then they didn't clean it up once they later on anandonned that whole story of having a constant offset for 40 MHz channels and just set the offset to be constant 0 -- there is no effort for avoiding 0x20 for the 20 MHz values stored in the EEPROM, hence that's probably just a forbidden value in the EEPROM specs and won't appear anyway... Anyway, the whole thing felt like solving some college math test where in the end everything cancels out and the result equals 0 ;) To make sure that channel bandwidth power compensation really doesn't need to be taken care of, output a warning when the corresponding value stored in the EEPROM is non-zero. Also there is no apparent reason to refrain from initializing RFCSR register 13, it doesn't fail what-so-ever. Signed-off-by: Daniel Golle <daniel@makrotopia.org>
2017-02-13 05:25:35 +00:00
+ }
+ rt2800_rfcsr_read_bank(rt2x00dev, 5, 6, &rfcsr);
+ rfcsr &= (~0x3F);
+ rfcsr |= rx_agc_fc;
rt2x00: mt7620: yet another beauty session So here is another round of improvements for MT7620 WiFi. This commit fixes a few significant issues related to TX_PWR_CFG_x and TX_ALC and also makes the code more readable by adding register descriptions for things added for MT7620 and use the usual bit-field access macros and the now defined macros instead of plain bit-ops and magic numbers. Properly describe EEPROM_TARGET_POWER at word 0x68 (== byte 0xD0) and thereby fix internal TXALC which would otherwise just read out-of-bounds of the EEPROM map. Split-out tx-power/ALC related stuff into an additional function. Fix VCO calibration, it was carried out properly in the channel switching but incomplete in the actual VCO calibration function. Also there is no need to trigger VCO calibration in channel switching, the VCO calibration function is already being called at this point. Remove it from channel switching function to avoid redundant code. The TX power calibration differs significantly from all other Mediatek/Ralink chips: They finally allow 0.5dB steps stored as 8-bit values for (almost) each bitrate -- and promptly ran out of space and for some reason didn't want to change the EEPROM layout. The hence opted for a scheme of sharing values for some adjecent bitrates and a highly over-complicated (or obfuscated?) way to populate the TX_PWR_CFG_x registers with the values stored in the EEPROM. The code here now looks much less complicated than what you see in the vendor's driver, however, it does the exact same thing: bGpwrdeltaMinus is a constant and always TRUE, hence half of the code was dead. Gpwrdelta is always 0 (rather than using the value read from the EEPROM). What remains is some very grotesque effort to avoid 0x20, probably some hardware bug related to some misunderstanding of what a singed 8-bit value is (imagine: if it was a signed 6-bit value then someone could believe that 0x20 == 0x0). And then they didn't clean it up once they later on anandonned that whole story of having a constant offset for 40 MHz channels and just set the offset to be constant 0 -- there is no effort for avoiding 0x20 for the 20 MHz values stored in the EEPROM, hence that's probably just a forbidden value in the EEPROM specs and won't appear anyway... Anyway, the whole thing felt like solving some college math test where in the end everything cancels out and the result equals 0 ;) To make sure that channel bandwidth power compensation really doesn't need to be taken care of, output a warning when the corresponding value stored in the EEPROM is non-zero. Also there is no apparent reason to refrain from initializing RFCSR register 13, it doesn't fail what-so-ever. Signed-off-by: Daniel Golle <daniel@makrotopia.org>
2017-02-13 05:25:35 +00:00
+ rt2800_rfcsr_write_bank(rt2x00dev, 5, 6, rfcsr);
+ rt2800_rfcsr_read_bank(rt2x00dev, 5, 7, &rfcsr);
+ rfcsr &= (~0x3F);
+ rfcsr |= rx_agc_fc;
rt2x00: mt7620: yet another beauty session So here is another round of improvements for MT7620 WiFi. This commit fixes a few significant issues related to TX_PWR_CFG_x and TX_ALC and also makes the code more readable by adding register descriptions for things added for MT7620 and use the usual bit-field access macros and the now defined macros instead of plain bit-ops and magic numbers. Properly describe EEPROM_TARGET_POWER at word 0x68 (== byte 0xD0) and thereby fix internal TXALC which would otherwise just read out-of-bounds of the EEPROM map. Split-out tx-power/ALC related stuff into an additional function. Fix VCO calibration, it was carried out properly in the channel switching but incomplete in the actual VCO calibration function. Also there is no need to trigger VCO calibration in channel switching, the VCO calibration function is already being called at this point. Remove it from channel switching function to avoid redundant code. The TX power calibration differs significantly from all other Mediatek/Ralink chips: They finally allow 0.5dB steps stored as 8-bit values for (almost) each bitrate -- and promptly ran out of space and for some reason didn't want to change the EEPROM layout. The hence opted for a scheme of sharing values for some adjecent bitrates and a highly over-complicated (or obfuscated?) way to populate the TX_PWR_CFG_x registers with the values stored in the EEPROM. The code here now looks much less complicated than what you see in the vendor's driver, however, it does the exact same thing: bGpwrdeltaMinus is a constant and always TRUE, hence half of the code was dead. Gpwrdelta is always 0 (rather than using the value read from the EEPROM). What remains is some very grotesque effort to avoid 0x20, probably some hardware bug related to some misunderstanding of what a singed 8-bit value is (imagine: if it was a signed 6-bit value then someone could believe that 0x20 == 0x0). And then they didn't clean it up once they later on anandonned that whole story of having a constant offset for 40 MHz channels and just set the offset to be constant 0 -- there is no effort for avoiding 0x20 for the 20 MHz values stored in the EEPROM, hence that's probably just a forbidden value in the EEPROM specs and won't appear anyway... Anyway, the whole thing felt like solving some college math test where in the end everything cancels out and the result equals 0 ;) To make sure that channel bandwidth power compensation really doesn't need to be taken care of, output a warning when the corresponding value stored in the EEPROM is non-zero. Also there is no apparent reason to refrain from initializing RFCSR register 13, it doesn't fail what-so-ever. Signed-off-by: Daniel Golle <daniel@makrotopia.org>
2017-02-13 05:25:35 +00:00
+ rt2800_rfcsr_write_bank(rt2x00dev, 5, 7, rfcsr);
+ rt2800_rfcsr_read_bank(rt2x00dev, 7, 6, &rfcsr);
+ rfcsr &= (~0x3F);
+ rfcsr |= rx_agc_fc;
rt2x00: mt7620: yet another beauty session So here is another round of improvements for MT7620 WiFi. This commit fixes a few significant issues related to TX_PWR_CFG_x and TX_ALC and also makes the code more readable by adding register descriptions for things added for MT7620 and use the usual bit-field access macros and the now defined macros instead of plain bit-ops and magic numbers. Properly describe EEPROM_TARGET_POWER at word 0x68 (== byte 0xD0) and thereby fix internal TXALC which would otherwise just read out-of-bounds of the EEPROM map. Split-out tx-power/ALC related stuff into an additional function. Fix VCO calibration, it was carried out properly in the channel switching but incomplete in the actual VCO calibration function. Also there is no need to trigger VCO calibration in channel switching, the VCO calibration function is already being called at this point. Remove it from channel switching function to avoid redundant code. The TX power calibration differs significantly from all other Mediatek/Ralink chips: They finally allow 0.5dB steps stored as 8-bit values for (almost) each bitrate -- and promptly ran out of space and for some reason didn't want to change the EEPROM layout. The hence opted for a scheme of sharing values for some adjecent bitrates and a highly over-complicated (or obfuscated?) way to populate the TX_PWR_CFG_x registers with the values stored in the EEPROM. The code here now looks much less complicated than what you see in the vendor's driver, however, it does the exact same thing: bGpwrdeltaMinus is a constant and always TRUE, hence half of the code was dead. Gpwrdelta is always 0 (rather than using the value read from the EEPROM). What remains is some very grotesque effort to avoid 0x20, probably some hardware bug related to some misunderstanding of what a singed 8-bit value is (imagine: if it was a signed 6-bit value then someone could believe that 0x20 == 0x0). And then they didn't clean it up once they later on anandonned that whole story of having a constant offset for 40 MHz channels and just set the offset to be constant 0 -- there is no effort for avoiding 0x20 for the 20 MHz values stored in the EEPROM, hence that's probably just a forbidden value in the EEPROM specs and won't appear anyway... Anyway, the whole thing felt like solving some college math test where in the end everything cancels out and the result equals 0 ;) To make sure that channel bandwidth power compensation really doesn't need to be taken care of, output a warning when the corresponding value stored in the EEPROM is non-zero. Also there is no apparent reason to refrain from initializing RFCSR register 13, it doesn't fail what-so-ever. Signed-off-by: Daniel Golle <daniel@makrotopia.org>
2017-02-13 05:25:35 +00:00
+ rt2800_rfcsr_write_bank(rt2x00dev, 7, 6, rfcsr);
+ rt2800_rfcsr_read_bank(rt2x00dev, 7, 7, &rfcsr);
+ rfcsr &= (~0x3F);
+ rfcsr |= rx_agc_fc;
rt2x00: mt7620: yet another beauty session So here is another round of improvements for MT7620 WiFi. This commit fixes a few significant issues related to TX_PWR_CFG_x and TX_ALC and also makes the code more readable by adding register descriptions for things added for MT7620 and use the usual bit-field access macros and the now defined macros instead of plain bit-ops and magic numbers. Properly describe EEPROM_TARGET_POWER at word 0x68 (== byte 0xD0) and thereby fix internal TXALC which would otherwise just read out-of-bounds of the EEPROM map. Split-out tx-power/ALC related stuff into an additional function. Fix VCO calibration, it was carried out properly in the channel switching but incomplete in the actual VCO calibration function. Also there is no need to trigger VCO calibration in channel switching, the VCO calibration function is already being called at this point. Remove it from channel switching function to avoid redundant code. The TX power calibration differs significantly from all other Mediatek/Ralink chips: They finally allow 0.5dB steps stored as 8-bit values for (almost) each bitrate -- and promptly ran out of space and for some reason didn't want to change the EEPROM layout. The hence opted for a scheme of sharing values for some adjecent bitrates and a highly over-complicated (or obfuscated?) way to populate the TX_PWR_CFG_x registers with the values stored in the EEPROM. The code here now looks much less complicated than what you see in the vendor's driver, however, it does the exact same thing: bGpwrdeltaMinus is a constant and always TRUE, hence half of the code was dead. Gpwrdelta is always 0 (rather than using the value read from the EEPROM). What remains is some very grotesque effort to avoid 0x20, probably some hardware bug related to some misunderstanding of what a singed 8-bit value is (imagine: if it was a signed 6-bit value then someone could believe that 0x20 == 0x0). And then they didn't clean it up once they later on anandonned that whole story of having a constant offset for 40 MHz channels and just set the offset to be constant 0 -- there is no effort for avoiding 0x20 for the 20 MHz values stored in the EEPROM, hence that's probably just a forbidden value in the EEPROM specs and won't appear anyway... Anyway, the whole thing felt like solving some college math test where in the end everything cancels out and the result equals 0 ;) To make sure that channel bandwidth power compensation really doesn't need to be taken care of, output a warning when the corresponding value stored in the EEPROM is non-zero. Also there is no apparent reason to refrain from initializing RFCSR register 13, it doesn't fail what-so-ever. Signed-off-by: Daniel Golle <daniel@makrotopia.org>
2017-02-13 05:25:35 +00:00
+ rt2800_rfcsr_write_bank(rt2x00dev, 7, 7, rfcsr);
+
+ rt2800_rfcsr_read_bank(rt2x00dev, 5, 58, &rfcsr);
+ rfcsr &= (~0x3F);
+ rfcsr |= tx_agc_fc;
rt2x00: mt7620: yet another beauty session So here is another round of improvements for MT7620 WiFi. This commit fixes a few significant issues related to TX_PWR_CFG_x and TX_ALC and also makes the code more readable by adding register descriptions for things added for MT7620 and use the usual bit-field access macros and the now defined macros instead of plain bit-ops and magic numbers. Properly describe EEPROM_TARGET_POWER at word 0x68 (== byte 0xD0) and thereby fix internal TXALC which would otherwise just read out-of-bounds of the EEPROM map. Split-out tx-power/ALC related stuff into an additional function. Fix VCO calibration, it was carried out properly in the channel switching but incomplete in the actual VCO calibration function. Also there is no need to trigger VCO calibration in channel switching, the VCO calibration function is already being called at this point. Remove it from channel switching function to avoid redundant code. The TX power calibration differs significantly from all other Mediatek/Ralink chips: They finally allow 0.5dB steps stored as 8-bit values for (almost) each bitrate -- and promptly ran out of space and for some reason didn't want to change the EEPROM layout. The hence opted for a scheme of sharing values for some adjecent bitrates and a highly over-complicated (or obfuscated?) way to populate the TX_PWR_CFG_x registers with the values stored in the EEPROM. The code here now looks much less complicated than what you see in the vendor's driver, however, it does the exact same thing: bGpwrdeltaMinus is a constant and always TRUE, hence half of the code was dead. Gpwrdelta is always 0 (rather than using the value read from the EEPROM). What remains is some very grotesque effort to avoid 0x20, probably some hardware bug related to some misunderstanding of what a singed 8-bit value is (imagine: if it was a signed 6-bit value then someone could believe that 0x20 == 0x0). And then they didn't clean it up once they later on anandonned that whole story of having a constant offset for 40 MHz channels and just set the offset to be constant 0 -- there is no effort for avoiding 0x20 for the 20 MHz values stored in the EEPROM, hence that's probably just a forbidden value in the EEPROM specs and won't appear anyway... Anyway, the whole thing felt like solving some college math test where in the end everything cancels out and the result equals 0 ;) To make sure that channel bandwidth power compensation really doesn't need to be taken care of, output a warning when the corresponding value stored in the EEPROM is non-zero. Also there is no apparent reason to refrain from initializing RFCSR register 13, it doesn't fail what-so-ever. Signed-off-by: Daniel Golle <daniel@makrotopia.org>
2017-02-13 05:25:35 +00:00
+ rt2800_rfcsr_write_bank(rt2x00dev, 5, 58, rfcsr);
+ rt2800_rfcsr_read_bank(rt2x00dev, 5, 59, &rfcsr);
+ rfcsr &= (~0x3F);
+ rfcsr |= tx_agc_fc;
rt2x00: mt7620: yet another beauty session So here is another round of improvements for MT7620 WiFi. This commit fixes a few significant issues related to TX_PWR_CFG_x and TX_ALC and also makes the code more readable by adding register descriptions for things added for MT7620 and use the usual bit-field access macros and the now defined macros instead of plain bit-ops and magic numbers. Properly describe EEPROM_TARGET_POWER at word 0x68 (== byte 0xD0) and thereby fix internal TXALC which would otherwise just read out-of-bounds of the EEPROM map. Split-out tx-power/ALC related stuff into an additional function. Fix VCO calibration, it was carried out properly in the channel switching but incomplete in the actual VCO calibration function. Also there is no need to trigger VCO calibration in channel switching, the VCO calibration function is already being called at this point. Remove it from channel switching function to avoid redundant code. The TX power calibration differs significantly from all other Mediatek/Ralink chips: They finally allow 0.5dB steps stored as 8-bit values for (almost) each bitrate -- and promptly ran out of space and for some reason didn't want to change the EEPROM layout. The hence opted for a scheme of sharing values for some adjecent bitrates and a highly over-complicated (or obfuscated?) way to populate the TX_PWR_CFG_x registers with the values stored in the EEPROM. The code here now looks much less complicated than what you see in the vendor's driver, however, it does the exact same thing: bGpwrdeltaMinus is a constant and always TRUE, hence half of the code was dead. Gpwrdelta is always 0 (rather than using the value read from the EEPROM). What remains is some very grotesque effort to avoid 0x20, probably some hardware bug related to some misunderstanding of what a singed 8-bit value is (imagine: if it was a signed 6-bit value then someone could believe that 0x20 == 0x0). And then they didn't clean it up once they later on anandonned that whole story of having a constant offset for 40 MHz channels and just set the offset to be constant 0 -- there is no effort for avoiding 0x20 for the 20 MHz values stored in the EEPROM, hence that's probably just a forbidden value in the EEPROM specs and won't appear anyway... Anyway, the whole thing felt like solving some college math test where in the end everything cancels out and the result equals 0 ;) To make sure that channel bandwidth power compensation really doesn't need to be taken care of, output a warning when the corresponding value stored in the EEPROM is non-zero. Also there is no apparent reason to refrain from initializing RFCSR register 13, it doesn't fail what-so-ever. Signed-off-by: Daniel Golle <daniel@makrotopia.org>
2017-02-13 05:25:35 +00:00
+ rt2800_rfcsr_write_bank(rt2x00dev, 5, 59, rfcsr);
+ rt2800_rfcsr_read_bank(rt2x00dev, 7, 58, &rfcsr);
+ rfcsr &= (~0x3F);
+ rfcsr |= tx_agc_fc;
rt2x00: mt7620: yet another beauty session So here is another round of improvements for MT7620 WiFi. This commit fixes a few significant issues related to TX_PWR_CFG_x and TX_ALC and also makes the code more readable by adding register descriptions for things added for MT7620 and use the usual bit-field access macros and the now defined macros instead of plain bit-ops and magic numbers. Properly describe EEPROM_TARGET_POWER at word 0x68 (== byte 0xD0) and thereby fix internal TXALC which would otherwise just read out-of-bounds of the EEPROM map. Split-out tx-power/ALC related stuff into an additional function. Fix VCO calibration, it was carried out properly in the channel switching but incomplete in the actual VCO calibration function. Also there is no need to trigger VCO calibration in channel switching, the VCO calibration function is already being called at this point. Remove it from channel switching function to avoid redundant code. The TX power calibration differs significantly from all other Mediatek/Ralink chips: They finally allow 0.5dB steps stored as 8-bit values for (almost) each bitrate -- and promptly ran out of space and for some reason didn't want to change the EEPROM layout. The hence opted for a scheme of sharing values for some adjecent bitrates and a highly over-complicated (or obfuscated?) way to populate the TX_PWR_CFG_x registers with the values stored in the EEPROM. The code here now looks much less complicated than what you see in the vendor's driver, however, it does the exact same thing: bGpwrdeltaMinus is a constant and always TRUE, hence half of the code was dead. Gpwrdelta is always 0 (rather than using the value read from the EEPROM). What remains is some very grotesque effort to avoid 0x20, probably some hardware bug related to some misunderstanding of what a singed 8-bit value is (imagine: if it was a signed 6-bit value then someone could believe that 0x20 == 0x0). And then they didn't clean it up once they later on anandonned that whole story of having a constant offset for 40 MHz channels and just set the offset to be constant 0 -- there is no effort for avoiding 0x20 for the 20 MHz values stored in the EEPROM, hence that's probably just a forbidden value in the EEPROM specs and won't appear anyway... Anyway, the whole thing felt like solving some college math test where in the end everything cancels out and the result equals 0 ;) To make sure that channel bandwidth power compensation really doesn't need to be taken care of, output a warning when the corresponding value stored in the EEPROM is non-zero. Also there is no apparent reason to refrain from initializing RFCSR register 13, it doesn't fail what-so-ever. Signed-off-by: Daniel Golle <daniel@makrotopia.org>
2017-02-13 05:25:35 +00:00
+ rt2800_rfcsr_write_bank(rt2x00dev, 7, 58, rfcsr);
+ rt2800_rfcsr_read_bank(rt2x00dev, 7, 59, &rfcsr);
+ rfcsr &= (~0x3F);
+ rfcsr |= tx_agc_fc;
rt2x00: mt7620: yet another beauty session So here is another round of improvements for MT7620 WiFi. This commit fixes a few significant issues related to TX_PWR_CFG_x and TX_ALC and also makes the code more readable by adding register descriptions for things added for MT7620 and use the usual bit-field access macros and the now defined macros instead of plain bit-ops and magic numbers. Properly describe EEPROM_TARGET_POWER at word 0x68 (== byte 0xD0) and thereby fix internal TXALC which would otherwise just read out-of-bounds of the EEPROM map. Split-out tx-power/ALC related stuff into an additional function. Fix VCO calibration, it was carried out properly in the channel switching but incomplete in the actual VCO calibration function. Also there is no need to trigger VCO calibration in channel switching, the VCO calibration function is already being called at this point. Remove it from channel switching function to avoid redundant code. The TX power calibration differs significantly from all other Mediatek/Ralink chips: They finally allow 0.5dB steps stored as 8-bit values for (almost) each bitrate -- and promptly ran out of space and for some reason didn't want to change the EEPROM layout. The hence opted for a scheme of sharing values for some adjecent bitrates and a highly over-complicated (or obfuscated?) way to populate the TX_PWR_CFG_x registers with the values stored in the EEPROM. The code here now looks much less complicated than what you see in the vendor's driver, however, it does the exact same thing: bGpwrdeltaMinus is a constant and always TRUE, hence half of the code was dead. Gpwrdelta is always 0 (rather than using the value read from the EEPROM). What remains is some very grotesque effort to avoid 0x20, probably some hardware bug related to some misunderstanding of what a singed 8-bit value is (imagine: if it was a signed 6-bit value then someone could believe that 0x20 == 0x0). And then they didn't clean it up once they later on anandonned that whole story of having a constant offset for 40 MHz channels and just set the offset to be constant 0 -- there is no effort for avoiding 0x20 for the 20 MHz values stored in the EEPROM, hence that's probably just a forbidden value in the EEPROM specs and won't appear anyway... Anyway, the whole thing felt like solving some college math test where in the end everything cancels out and the result equals 0 ;) To make sure that channel bandwidth power compensation really doesn't need to be taken care of, output a warning when the corresponding value stored in the EEPROM is non-zero. Also there is no apparent reason to refrain from initializing RFCSR register 13, it doesn't fail what-so-ever. Signed-off-by: Daniel Golle <daniel@makrotopia.org>
2017-02-13 05:25:35 +00:00
+ rt2800_rfcsr_write_bank(rt2x00dev, 7, 59, rfcsr);
+ }
rt2x00: mt7620: yet another beauty session So here is another round of improvements for MT7620 WiFi. This commit fixes a few significant issues related to TX_PWR_CFG_x and TX_ALC and also makes the code more readable by adding register descriptions for things added for MT7620 and use the usual bit-field access macros and the now defined macros instead of plain bit-ops and magic numbers. Properly describe EEPROM_TARGET_POWER at word 0x68 (== byte 0xD0) and thereby fix internal TXALC which would otherwise just read out-of-bounds of the EEPROM map. Split-out tx-power/ALC related stuff into an additional function. Fix VCO calibration, it was carried out properly in the channel switching but incomplete in the actual VCO calibration function. Also there is no need to trigger VCO calibration in channel switching, the VCO calibration function is already being called at this point. Remove it from channel switching function to avoid redundant code. The TX power calibration differs significantly from all other Mediatek/Ralink chips: They finally allow 0.5dB steps stored as 8-bit values for (almost) each bitrate -- and promptly ran out of space and for some reason didn't want to change the EEPROM layout. The hence opted for a scheme of sharing values for some adjecent bitrates and a highly over-complicated (or obfuscated?) way to populate the TX_PWR_CFG_x registers with the values stored in the EEPROM. The code here now looks much less complicated than what you see in the vendor's driver, however, it does the exact same thing: bGpwrdeltaMinus is a constant and always TRUE, hence half of the code was dead. Gpwrdelta is always 0 (rather than using the value read from the EEPROM). What remains is some very grotesque effort to avoid 0x20, probably some hardware bug related to some misunderstanding of what a singed 8-bit value is (imagine: if it was a signed 6-bit value then someone could believe that 0x20 == 0x0). And then they didn't clean it up once they later on anandonned that whole story of having a constant offset for 40 MHz channels and just set the offset to be constant 0 -- there is no effort for avoiding 0x20 for the 20 MHz values stored in the EEPROM, hence that's probably just a forbidden value in the EEPROM specs and won't appear anyway... Anyway, the whole thing felt like solving some college math test where in the end everything cancels out and the result equals 0 ;) To make sure that channel bandwidth power compensation really doesn't need to be taken care of, output a warning when the corresponding value stored in the EEPROM is non-zero. Also there is no apparent reason to refrain from initializing RFCSR register 13, it doesn't fail what-so-ever. Signed-off-by: Daniel Golle <daniel@makrotopia.org>
2017-02-13 05:25:35 +00:00
+}
+
+static void rt2800_config_alc(struct rt2x00_dev *rt2x00dev,
+ struct ieee80211_channel *chan,
+ int power_level) {
+ u16 eeprom, target_power, max_power;
+ u32 mac_sys_ctrl, mac_status;
+ u32 reg;
+ u8 bbp;
+ int i;
+
+ /* hardware unit is 0.5dBm, limited to 23.5dBm */
+ power_level *= 2;
+ if (power_level > 0x2f)
+ power_level = 0x2f;
+
+ max_power = chan->max_power * 2;
+ if (max_power > 0x2f)
+ max_power = 0x2f;
+
+ rt2800_register_read(rt2x00dev, TX_ALC_CFG_0, &reg);
+ rt2x00_set_field32(&reg, TX_ALC_CFG_0_CH_INIT_0, power_level);
+ rt2x00_set_field32(&reg, TX_ALC_CFG_0_CH_INIT_1, power_level);
+ rt2x00_set_field32(&reg, TX_ALC_CFG_0_LIMIT_0, max_power);
+ rt2x00_set_field32(&reg, TX_ALC_CFG_0_LIMIT_1, max_power);
+
+ rt2800_eeprom_read(rt2x00dev, EEPROM_NIC_CONF1, &eeprom);
+ if (rt2x00_get_field16(eeprom, EEPROM_NIC_CONF1_INTERNAL_TX_ALC)) {
rt2x00: mt7620: yet another beauty session So here is another round of improvements for MT7620 WiFi. This commit fixes a few significant issues related to TX_PWR_CFG_x and TX_ALC and also makes the code more readable by adding register descriptions for things added for MT7620 and use the usual bit-field access macros and the now defined macros instead of plain bit-ops and magic numbers. Properly describe EEPROM_TARGET_POWER at word 0x68 (== byte 0xD0) and thereby fix internal TXALC which would otherwise just read out-of-bounds of the EEPROM map. Split-out tx-power/ALC related stuff into an additional function. Fix VCO calibration, it was carried out properly in the channel switching but incomplete in the actual VCO calibration function. Also there is no need to trigger VCO calibration in channel switching, the VCO calibration function is already being called at this point. Remove it from channel switching function to avoid redundant code. The TX power calibration differs significantly from all other Mediatek/Ralink chips: They finally allow 0.5dB steps stored as 8-bit values for (almost) each bitrate -- and promptly ran out of space and for some reason didn't want to change the EEPROM layout. The hence opted for a scheme of sharing values for some adjecent bitrates and a highly over-complicated (or obfuscated?) way to populate the TX_PWR_CFG_x registers with the values stored in the EEPROM. The code here now looks much less complicated than what you see in the vendor's driver, however, it does the exact same thing: bGpwrdeltaMinus is a constant and always TRUE, hence half of the code was dead. Gpwrdelta is always 0 (rather than using the value read from the EEPROM). What remains is some very grotesque effort to avoid 0x20, probably some hardware bug related to some misunderstanding of what a singed 8-bit value is (imagine: if it was a signed 6-bit value then someone could believe that 0x20 == 0x0). And then they didn't clean it up once they later on anandonned that whole story of having a constant offset for 40 MHz channels and just set the offset to be constant 0 -- there is no effort for avoiding 0x20 for the 20 MHz values stored in the EEPROM, hence that's probably just a forbidden value in the EEPROM specs and won't appear anyway... Anyway, the whole thing felt like solving some college math test where in the end everything cancels out and the result equals 0 ;) To make sure that channel bandwidth power compensation really doesn't need to be taken care of, output a warning when the corresponding value stored in the EEPROM is non-zero. Also there is no apparent reason to refrain from initializing RFCSR register 13, it doesn't fail what-so-ever. Signed-off-by: Daniel Golle <daniel@makrotopia.org>
2017-02-13 05:25:35 +00:00
+ /* init base power by eeprom target power */
+ rt2800_eeprom_read(rt2x00dev, EEPROM_TXPOWER_INIT,
+ &target_power);
+ rt2x00_set_field32(&reg, TX_ALC_CFG_0_CH_INIT_0, target_power);
+ rt2x00_set_field32(&reg, TX_ALC_CFG_0_CH_INIT_1, target_power);
+ }
rt2x00: mt7620: yet another beauty session So here is another round of improvements for MT7620 WiFi. This commit fixes a few significant issues related to TX_PWR_CFG_x and TX_ALC and also makes the code more readable by adding register descriptions for things added for MT7620 and use the usual bit-field access macros and the now defined macros instead of plain bit-ops and magic numbers. Properly describe EEPROM_TARGET_POWER at word 0x68 (== byte 0xD0) and thereby fix internal TXALC which would otherwise just read out-of-bounds of the EEPROM map. Split-out tx-power/ALC related stuff into an additional function. Fix VCO calibration, it was carried out properly in the channel switching but incomplete in the actual VCO calibration function. Also there is no need to trigger VCO calibration in channel switching, the VCO calibration function is already being called at this point. Remove it from channel switching function to avoid redundant code. The TX power calibration differs significantly from all other Mediatek/Ralink chips: They finally allow 0.5dB steps stored as 8-bit values for (almost) each bitrate -- and promptly ran out of space and for some reason didn't want to change the EEPROM layout. The hence opted for a scheme of sharing values for some adjecent bitrates and a highly over-complicated (or obfuscated?) way to populate the TX_PWR_CFG_x registers with the values stored in the EEPROM. The code here now looks much less complicated than what you see in the vendor's driver, however, it does the exact same thing: bGpwrdeltaMinus is a constant and always TRUE, hence half of the code was dead. Gpwrdelta is always 0 (rather than using the value read from the EEPROM). What remains is some very grotesque effort to avoid 0x20, probably some hardware bug related to some misunderstanding of what a singed 8-bit value is (imagine: if it was a signed 6-bit value then someone could believe that 0x20 == 0x0). And then they didn't clean it up once they later on anandonned that whole story of having a constant offset for 40 MHz channels and just set the offset to be constant 0 -- there is no effort for avoiding 0x20 for the 20 MHz values stored in the EEPROM, hence that's probably just a forbidden value in the EEPROM specs and won't appear anyway... Anyway, the whole thing felt like solving some college math test where in the end everything cancels out and the result equals 0 ;) To make sure that channel bandwidth power compensation really doesn't need to be taken care of, output a warning when the corresponding value stored in the EEPROM is non-zero. Also there is no apparent reason to refrain from initializing RFCSR register 13, it doesn't fail what-so-ever. Signed-off-by: Daniel Golle <daniel@makrotopia.org>
2017-02-13 05:25:35 +00:00
+ rt2800_register_write(rt2x00dev, TX_ALC_CFG_0, reg);
+
rt2x00: mt7620: yet another beauty session So here is another round of improvements for MT7620 WiFi. This commit fixes a few significant issues related to TX_PWR_CFG_x and TX_ALC and also makes the code more readable by adding register descriptions for things added for MT7620 and use the usual bit-field access macros and the now defined macros instead of plain bit-ops and magic numbers. Properly describe EEPROM_TARGET_POWER at word 0x68 (== byte 0xD0) and thereby fix internal TXALC which would otherwise just read out-of-bounds of the EEPROM map. Split-out tx-power/ALC related stuff into an additional function. Fix VCO calibration, it was carried out properly in the channel switching but incomplete in the actual VCO calibration function. Also there is no need to trigger VCO calibration in channel switching, the VCO calibration function is already being called at this point. Remove it from channel switching function to avoid redundant code. The TX power calibration differs significantly from all other Mediatek/Ralink chips: They finally allow 0.5dB steps stored as 8-bit values for (almost) each bitrate -- and promptly ran out of space and for some reason didn't want to change the EEPROM layout. The hence opted for a scheme of sharing values for some adjecent bitrates and a highly over-complicated (or obfuscated?) way to populate the TX_PWR_CFG_x registers with the values stored in the EEPROM. The code here now looks much less complicated than what you see in the vendor's driver, however, it does the exact same thing: bGpwrdeltaMinus is a constant and always TRUE, hence half of the code was dead. Gpwrdelta is always 0 (rather than using the value read from the EEPROM). What remains is some very grotesque effort to avoid 0x20, probably some hardware bug related to some misunderstanding of what a singed 8-bit value is (imagine: if it was a signed 6-bit value then someone could believe that 0x20 == 0x0). And then they didn't clean it up once they later on anandonned that whole story of having a constant offset for 40 MHz channels and just set the offset to be constant 0 -- there is no effort for avoiding 0x20 for the 20 MHz values stored in the EEPROM, hence that's probably just a forbidden value in the EEPROM specs and won't appear anyway... Anyway, the whole thing felt like solving some college math test where in the end everything cancels out and the result equals 0 ;) To make sure that channel bandwidth power compensation really doesn't need to be taken care of, output a warning when the corresponding value stored in the EEPROM is non-zero. Also there is no apparent reason to refrain from initializing RFCSR register 13, it doesn't fail what-so-ever. Signed-off-by: Daniel Golle <daniel@makrotopia.org>
2017-02-13 05:25:35 +00:00
+ rt2800_register_read(rt2x00dev, TX_ALC_CFG_1, &reg);
+ rt2x00_set_field32(&reg, TX_ALC_CFG_1_TX_TEMP_COMP, 0);
+ rt2800_register_write(rt2x00dev, TX_ALC_CFG_1, reg);
+
+ /* Save MAC SYS CTRL registers */
+ rt2800_register_read(rt2x00dev, MAC_SYS_CTRL, &mac_sys_ctrl);
+ /* Disable Tx/Rx */
+ rt2800_register_write(rt2x00dev, MAC_SYS_CTRL, 0);
+ /* Check MAC Tx/Rx idle */
+ for (i = 0; i < 10000; i++) {
rt2x00: mt7620: yet another beauty session So here is another round of improvements for MT7620 WiFi. This commit fixes a few significant issues related to TX_PWR_CFG_x and TX_ALC and also makes the code more readable by adding register descriptions for things added for MT7620 and use the usual bit-field access macros and the now defined macros instead of plain bit-ops and magic numbers. Properly describe EEPROM_TARGET_POWER at word 0x68 (== byte 0xD0) and thereby fix internal TXALC which would otherwise just read out-of-bounds of the EEPROM map. Split-out tx-power/ALC related stuff into an additional function. Fix VCO calibration, it was carried out properly in the channel switching but incomplete in the actual VCO calibration function. Also there is no need to trigger VCO calibration in channel switching, the VCO calibration function is already being called at this point. Remove it from channel switching function to avoid redundant code. The TX power calibration differs significantly from all other Mediatek/Ralink chips: They finally allow 0.5dB steps stored as 8-bit values for (almost) each bitrate -- and promptly ran out of space and for some reason didn't want to change the EEPROM layout. The hence opted for a scheme of sharing values for some adjecent bitrates and a highly over-complicated (or obfuscated?) way to populate the TX_PWR_CFG_x registers with the values stored in the EEPROM. The code here now looks much less complicated than what you see in the vendor's driver, however, it does the exact same thing: bGpwrdeltaMinus is a constant and always TRUE, hence half of the code was dead. Gpwrdelta is always 0 (rather than using the value read from the EEPROM). What remains is some very grotesque effort to avoid 0x20, probably some hardware bug related to some misunderstanding of what a singed 8-bit value is (imagine: if it was a signed 6-bit value then someone could believe that 0x20 == 0x0). And then they didn't clean it up once they later on anandonned that whole story of having a constant offset for 40 MHz channels and just set the offset to be constant 0 -- there is no effort for avoiding 0x20 for the 20 MHz values stored in the EEPROM, hence that's probably just a forbidden value in the EEPROM specs and won't appear anyway... Anyway, the whole thing felt like solving some college math test where in the end everything cancels out and the result equals 0 ;) To make sure that channel bandwidth power compensation really doesn't need to be taken care of, output a warning when the corresponding value stored in the EEPROM is non-zero. Also there is no apparent reason to refrain from initializing RFCSR register 13, it doesn't fail what-so-ever. Signed-off-by: Daniel Golle <daniel@makrotopia.org>
2017-02-13 05:25:35 +00:00
+ rt2800_register_read(rt2x00dev, MAC_STATUS_CFG,
+ &mac_status);
+ if (mac_status & 0x3)
rt2x00: mt7620: lots of improvements This commit combines all the changes I've made on my staging tree into a single commit fixing many issues with our patch for MT7620. First of all, checkpatch.pl revealed numerous code style issues with the patch, so fix all the white-space and commets. Also use usleep_range instead of legacy timing and relax timing for VCO calibration just like the vendor driver does. Several line programming registers were commented out in the patch. Originally this came from the features present but disabled by default in the vendor's driver (RTMP_TEMPERATURE_CALIBRATION and ADJUST_POWER_CONSUMPTION_SUPPORT). Remove the dead code for now, it can easily be re-added if we actually intend to support those features. Move values from mt7620_freqconfig type into the existing rf_channel struct, this shouldn't be a new typedef and it is possible to use the existing struct because rf_channel got 4 32-bit fields, so two of the 8-bit values from mt7620_freqconfig can easily be stored in the same 32-bit field. Map values such that Rdiv -> rf1 N -> rf2 K -> rf3[0:7] D -> rf3[8:15] Ksd -> rf4 This makes the channel switching logic already look a bit more like what we are used to in rt2x00... Probably many of the read-modify-write calls could still be replaced by macros intended for that. iq calibration seems to be identical to RT5592, so just enable it. Test shows that this improves things quite a lot, datarates went up by a couple of megabits when running iperf, signal quality seems jumpy in the first few seconds once a station connencts, the stabelizes on a value significantly better than what it was before. Add description to the patch and reference the original OpenWrt commit by which it was added. The patch now passes checkpatch.pl and can thus be discussed with the upstream authors of the rt2x00 driver. Funded-by: https://www.kickstarter.com/projects/1327597961/better-support-for-mt7620a-n-in-openwrt-lede/ Signed-off-by: Daniel Golle <daniel@makrotopia.org>
2017-02-03 14:02:32 +00:00
+ usleep_range(50, 200);
+ else
+ break;
+ }
+
+ if (i == 10000)
+ rt2x00_warn(rt2x00dev, "Wait MAC Status to MAX !!!\n");
+
rt2x00: mt7620: yet another beauty session So here is another round of improvements for MT7620 WiFi. This commit fixes a few significant issues related to TX_PWR_CFG_x and TX_ALC and also makes the code more readable by adding register descriptions for things added for MT7620 and use the usual bit-field access macros and the now defined macros instead of plain bit-ops and magic numbers. Properly describe EEPROM_TARGET_POWER at word 0x68 (== byte 0xD0) and thereby fix internal TXALC which would otherwise just read out-of-bounds of the EEPROM map. Split-out tx-power/ALC related stuff into an additional function. Fix VCO calibration, it was carried out properly in the channel switching but incomplete in the actual VCO calibration function. Also there is no need to trigger VCO calibration in channel switching, the VCO calibration function is already being called at this point. Remove it from channel switching function to avoid redundant code. The TX power calibration differs significantly from all other Mediatek/Ralink chips: They finally allow 0.5dB steps stored as 8-bit values for (almost) each bitrate -- and promptly ran out of space and for some reason didn't want to change the EEPROM layout. The hence opted for a scheme of sharing values for some adjecent bitrates and a highly over-complicated (or obfuscated?) way to populate the TX_PWR_CFG_x registers with the values stored in the EEPROM. The code here now looks much less complicated than what you see in the vendor's driver, however, it does the exact same thing: bGpwrdeltaMinus is a constant and always TRUE, hence half of the code was dead. Gpwrdelta is always 0 (rather than using the value read from the EEPROM). What remains is some very grotesque effort to avoid 0x20, probably some hardware bug related to some misunderstanding of what a singed 8-bit value is (imagine: if it was a signed 6-bit value then someone could believe that 0x20 == 0x0). And then they didn't clean it up once they later on anandonned that whole story of having a constant offset for 40 MHz channels and just set the offset to be constant 0 -- there is no effort for avoiding 0x20 for the 20 MHz values stored in the EEPROM, hence that's probably just a forbidden value in the EEPROM specs and won't appear anyway... Anyway, the whole thing felt like solving some college math test where in the end everything cancels out and the result equals 0 ;) To make sure that channel bandwidth power compensation really doesn't need to be taken care of, output a warning when the corresponding value stored in the EEPROM is non-zero. Also there is no apparent reason to refrain from initializing RFCSR register 13, it doesn't fail what-so-ever. Signed-off-by: Daniel Golle <daniel@makrotopia.org>
2017-02-13 05:25:35 +00:00
+ if (chan->center_freq > 2457) {
+ rt2800_bbp_read(rt2x00dev, 30, &bbp);
+ bbp = 0x40;
+ rt2800_bbp_write(rt2x00dev, 30, bbp);
+ rt2800_rfcsr_write(rt2x00dev, 39, 0);
+ if (rt2x00_has_cap_external_lna_bg(rt2x00dev))
+ rt2800_rfcsr_write(rt2x00dev, 42, 0xfb);
+ else
+ rt2800_rfcsr_write(rt2x00dev, 42, 0x7b);
+ } else {
+ rt2800_bbp_read(rt2x00dev, 30, &bbp);
+ bbp = 0x1f;
+ rt2800_bbp_write(rt2x00dev, 30, bbp);
+ rt2800_rfcsr_write(rt2x00dev, 39, 0x80);
+ if (rt2x00_has_cap_external_lna_bg(rt2x00dev))
+ rt2800_rfcsr_write(rt2x00dev, 42, 0xdb);
+ else
+ rt2800_rfcsr_write(rt2x00dev, 42, 0x5b);
+ }
+ rt2800_register_write(rt2x00dev, MAC_SYS_CTRL, mac_sys_ctrl);
rt2x00: mt7620: lots of improvements This commit combines all the changes I've made on my staging tree into a single commit fixing many issues with our patch for MT7620. First of all, checkpatch.pl revealed numerous code style issues with the patch, so fix all the white-space and commets. Also use usleep_range instead of legacy timing and relax timing for VCO calibration just like the vendor driver does. Several line programming registers were commented out in the patch. Originally this came from the features present but disabled by default in the vendor's driver (RTMP_TEMPERATURE_CALIBRATION and ADJUST_POWER_CONSUMPTION_SUPPORT). Remove the dead code for now, it can easily be re-added if we actually intend to support those features. Move values from mt7620_freqconfig type into the existing rf_channel struct, this shouldn't be a new typedef and it is possible to use the existing struct because rf_channel got 4 32-bit fields, so two of the 8-bit values from mt7620_freqconfig can easily be stored in the same 32-bit field. Map values such that Rdiv -> rf1 N -> rf2 K -> rf3[0:7] D -> rf3[8:15] Ksd -> rf4 This makes the channel switching logic already look a bit more like what we are used to in rt2x00... Probably many of the read-modify-write calls could still be replaced by macros intended for that. iq calibration seems to be identical to RT5592, so just enable it. Test shows that this improves things quite a lot, datarates went up by a couple of megabits when running iperf, signal quality seems jumpy in the first few seconds once a station connencts, the stabelizes on a value significantly better than what it was before. Add description to the patch and reference the original OpenWrt commit by which it was added. The patch now passes checkpatch.pl and can thus be discussed with the upstream authors of the rt2x00 driver. Funded-by: https://www.kickstarter.com/projects/1327597961/better-support-for-mt7620a-n-in-openwrt-lede/ Signed-off-by: Daniel Golle <daniel@makrotopia.org>
2017-02-03 14:02:32 +00:00
+}
+
static void rt2800_bbp_write_with_rx_chain(struct rt2x00_dev *rt2x00dev,
const unsigned int word,
const u8 value)
@@ -3228,7 +3533,7 @@ static void rt2800_config_channel(struct rt2x00_dev *rt2x00dev,
struct channel_info *info)
{
u32 reg;
- unsigned int tx_pin;
+ u32 tx_pin;
u8 bbp, rfcsr;
info->default_power1 = rt2800_txpower_to_dev(rt2x00dev, rf->channel,
@@ -3273,6 +3578,9 @@ static void rt2800_config_channel(struct rt2x00_dev *rt2x00dev,
case RF5592:
rt2800_config_channel_rf55xx(rt2x00dev, conf, rf, info);
break;
+ case RF7620:
+ rt2800_config_channel_rf7620(rt2x00dev, conf, rf, info);
+ break;
default:
rt2800_config_channel_rf2xxx(rt2x00dev, conf, rf, info);
}
@@ -3347,7 +3655,8 @@ static void rt2800_config_channel(struct rt2x00_dev *rt2x00dev,
if (rf->channel <= 14) {
if (!rt2x00_rt(rt2x00dev, RT5390) &&
- !rt2x00_rt(rt2x00dev, RT5392)) {
+ !rt2x00_rt(rt2x00dev, RT5392) &&
+ !rt2x00_rt(rt2x00dev, RT6352)) {
if (rt2x00_has_cap_external_lna_bg(rt2x00dev)) {
rt2800_bbp_write(rt2x00dev, 82, 0x62);
rt2800_bbp_write(rt2x00dev, 75, 0x46);
@@ -3367,7 +3676,7 @@ static void rt2800_config_channel(struct rt2x00_dev *rt2x00dev,
rt2800_bbp_write(rt2x00dev, 82, 0x94);
else if (rt2x00_rt(rt2x00dev, RT3593))
rt2800_bbp_write(rt2x00dev, 82, 0x82);
- else
+ else if (!rt2x00_rt(rt2x00dev, RT6352))
rt2800_bbp_write(rt2x00dev, 82, 0xf2);
if (rt2x00_rt(rt2x00dev, RT3593))
@@ -3388,7 +3697,7 @@ static void rt2800_config_channel(struct rt2x00_dev *rt2x00dev,
if (rt2x00_rt(rt2x00dev, RT3572))
rt2800_rfcsr_write(rt2x00dev, 8, 0);
- tx_pin = 0;
+ rt2800_register_read(rt2x00dev, TX_PIN_CFG, &tx_pin);
switch (rt2x00dev->default_ant.tx_chain_num) {
case 3:
@@ -3437,6 +3746,7 @@ static void rt2800_config_channel(struct rt2x00_dev *rt2x00dev,
rt2x00_set_field32(&tx_pin, TX_PIN_CFG_RFTR_EN, 1);
rt2x00_set_field32(&tx_pin, TX_PIN_CFG_TRSW_EN, 1);
+ rt2x00_set_field32(&tx_pin, TX_PIN_CFG_RFRX_EN, 1); /* mt7620 */
rt2800_register_write(rt2x00dev, TX_PIN_CFG, tx_pin);
@@ -3495,7 +3805,7 @@ static void rt2800_config_channel(struct rt2x00_dev *rt2x00dev,
rt2x00: mt7620: lots of improvements This commit combines all the changes I've made on my staging tree into a single commit fixing many issues with our patch for MT7620. First of all, checkpatch.pl revealed numerous code style issues with the patch, so fix all the white-space and commets. Also use usleep_range instead of legacy timing and relax timing for VCO calibration just like the vendor driver does. Several line programming registers were commented out in the patch. Originally this came from the features present but disabled by default in the vendor's driver (RTMP_TEMPERATURE_CALIBRATION and ADJUST_POWER_CONSUMPTION_SUPPORT). Remove the dead code for now, it can easily be re-added if we actually intend to support those features. Move values from mt7620_freqconfig type into the existing rf_channel struct, this shouldn't be a new typedef and it is possible to use the existing struct because rf_channel got 4 32-bit fields, so two of the 8-bit values from mt7620_freqconfig can easily be stored in the same 32-bit field. Map values such that Rdiv -> rf1 N -> rf2 K -> rf3[0:7] D -> rf3[8:15] Ksd -> rf4 This makes the channel switching logic already look a bit more like what we are used to in rt2x00... Probably many of the read-modify-write calls could still be replaced by macros intended for that. iq calibration seems to be identical to RT5592, so just enable it. Test shows that this improves things quite a lot, datarates went up by a couple of megabits when running iperf, signal quality seems jumpy in the first few seconds once a station connencts, the stabelizes on a value significantly better than what it was before. Add description to the patch and reference the original OpenWrt commit by which it was added. The patch now passes checkpatch.pl and can thus be discussed with the upstream authors of the rt2x00 driver. Funded-by: https://www.kickstarter.com/projects/1327597961/better-support-for-mt7620a-n-in-openwrt-lede/ Signed-off-by: Daniel Golle <daniel@makrotopia.org>
2017-02-03 14:02:32 +00:00
usleep_range(1000, 1500);
}
- if (rt2x00_rt(rt2x00dev, RT5592)) {
+ if (rt2x00_rt(rt2x00dev, RT5592) || rt2x00_rt(rt2x00dev, RT6352)) {
rt2x00: mt7620: lots of improvements This commit combines all the changes I've made on my staging tree into a single commit fixing many issues with our patch for MT7620. First of all, checkpatch.pl revealed numerous code style issues with the patch, so fix all the white-space and commets. Also use usleep_range instead of legacy timing and relax timing for VCO calibration just like the vendor driver does. Several line programming registers were commented out in the patch. Originally this came from the features present but disabled by default in the vendor's driver (RTMP_TEMPERATURE_CALIBRATION and ADJUST_POWER_CONSUMPTION_SUPPORT). Remove the dead code for now, it can easily be re-added if we actually intend to support those features. Move values from mt7620_freqconfig type into the existing rf_channel struct, this shouldn't be a new typedef and it is possible to use the existing struct because rf_channel got 4 32-bit fields, so two of the 8-bit values from mt7620_freqconfig can easily be stored in the same 32-bit field. Map values such that Rdiv -> rf1 N -> rf2 K -> rf3[0:7] D -> rf3[8:15] Ksd -> rf4 This makes the channel switching logic already look a bit more like what we are used to in rt2x00... Probably many of the read-modify-write calls could still be replaced by macros intended for that. iq calibration seems to be identical to RT5592, so just enable it. Test shows that this improves things quite a lot, datarates went up by a couple of megabits when running iperf, signal quality seems jumpy in the first few seconds once a station connencts, the stabelizes on a value significantly better than what it was before. Add description to the patch and reference the original OpenWrt commit by which it was added. The patch now passes checkpatch.pl and can thus be discussed with the upstream authors of the rt2x00 driver. Funded-by: https://www.kickstarter.com/projects/1327597961/better-support-for-mt7620a-n-in-openwrt-lede/ Signed-off-by: Daniel Golle <daniel@makrotopia.org>
2017-02-03 14:02:32 +00:00
rt2800_bbp_write(rt2x00dev, 195, 141);
rt2800_bbp_write(rt2x00dev, 196, conf_is_ht40(conf) ? 0x10 : 0x1a);
@@ -4182,6 +4492,128 @@ static void rt2800_config_txpower_rt3593(struct rt2x00_dev *rt2x00dev,
rt2x00: mt7620: yet another beauty session So here is another round of improvements for MT7620 WiFi. This commit fixes a few significant issues related to TX_PWR_CFG_x and TX_ALC and also makes the code more readable by adding register descriptions for things added for MT7620 and use the usual bit-field access macros and the now defined macros instead of plain bit-ops and magic numbers. Properly describe EEPROM_TARGET_POWER at word 0x68 (== byte 0xD0) and thereby fix internal TXALC which would otherwise just read out-of-bounds of the EEPROM map. Split-out tx-power/ALC related stuff into an additional function. Fix VCO calibration, it was carried out properly in the channel switching but incomplete in the actual VCO calibration function. Also there is no need to trigger VCO calibration in channel switching, the VCO calibration function is already being called at this point. Remove it from channel switching function to avoid redundant code. The TX power calibration differs significantly from all other Mediatek/Ralink chips: They finally allow 0.5dB steps stored as 8-bit values for (almost) each bitrate -- and promptly ran out of space and for some reason didn't want to change the EEPROM layout. The hence opted for a scheme of sharing values for some adjecent bitrates and a highly over-complicated (or obfuscated?) way to populate the TX_PWR_CFG_x registers with the values stored in the EEPROM. The code here now looks much less complicated than what you see in the vendor's driver, however, it does the exact same thing: bGpwrdeltaMinus is a constant and always TRUE, hence half of the code was dead. Gpwrdelta is always 0 (rather than using the value read from the EEPROM). What remains is some very grotesque effort to avoid 0x20, probably some hardware bug related to some misunderstanding of what a singed 8-bit value is (imagine: if it was a signed 6-bit value then someone could believe that 0x20 == 0x0). And then they didn't clean it up once they later on anandonned that whole story of having a constant offset for 40 MHz channels and just set the offset to be constant 0 -- there is no effort for avoiding 0x20 for the 20 MHz values stored in the EEPROM, hence that's probably just a forbidden value in the EEPROM specs and won't appear anyway... Anyway, the whole thing felt like solving some college math test where in the end everything cancels out and the result equals 0 ;) To make sure that channel bandwidth power compensation really doesn't need to be taken care of, output a warning when the corresponding value stored in the EEPROM is non-zero. Also there is no apparent reason to refrain from initializing RFCSR register 13, it doesn't fail what-so-ever. Signed-off-by: Daniel Golle <daniel@makrotopia.org>
2017-02-13 05:25:35 +00:00
(unsigned long) regs[i]);
}
+static void rt2800_config_txpower_rt6352(struct rt2x00_dev *rt2x00dev,
rt2x00: mt7620: yet another beauty session So here is another round of improvements for MT7620 WiFi. This commit fixes a few significant issues related to TX_PWR_CFG_x and TX_ALC and also makes the code more readable by adding register descriptions for things added for MT7620 and use the usual bit-field access macros and the now defined macros instead of plain bit-ops and magic numbers. Properly describe EEPROM_TARGET_POWER at word 0x68 (== byte 0xD0) and thereby fix internal TXALC which would otherwise just read out-of-bounds of the EEPROM map. Split-out tx-power/ALC related stuff into an additional function. Fix VCO calibration, it was carried out properly in the channel switching but incomplete in the actual VCO calibration function. Also there is no need to trigger VCO calibration in channel switching, the VCO calibration function is already being called at this point. Remove it from channel switching function to avoid redundant code. The TX power calibration differs significantly from all other Mediatek/Ralink chips: They finally allow 0.5dB steps stored as 8-bit values for (almost) each bitrate -- and promptly ran out of space and for some reason didn't want to change the EEPROM layout. The hence opted for a scheme of sharing values for some adjecent bitrates and a highly over-complicated (or obfuscated?) way to populate the TX_PWR_CFG_x registers with the values stored in the EEPROM. The code here now looks much less complicated than what you see in the vendor's driver, however, it does the exact same thing: bGpwrdeltaMinus is a constant and always TRUE, hence half of the code was dead. Gpwrdelta is always 0 (rather than using the value read from the EEPROM). What remains is some very grotesque effort to avoid 0x20, probably some hardware bug related to some misunderstanding of what a singed 8-bit value is (imagine: if it was a signed 6-bit value then someone could believe that 0x20 == 0x0). And then they didn't clean it up once they later on anandonned that whole story of having a constant offset for 40 MHz channels and just set the offset to be constant 0 -- there is no effort for avoiding 0x20 for the 20 MHz values stored in the EEPROM, hence that's probably just a forbidden value in the EEPROM specs and won't appear anyway... Anyway, the whole thing felt like solving some college math test where in the end everything cancels out and the result equals 0 ;) To make sure that channel bandwidth power compensation really doesn't need to be taken care of, output a warning when the corresponding value stored in the EEPROM is non-zero. Also there is no apparent reason to refrain from initializing RFCSR register 13, it doesn't fail what-so-ever. Signed-off-by: Daniel Golle <daniel@makrotopia.org>
2017-02-13 05:25:35 +00:00
+ struct ieee80211_channel *chan,
+ int power_level)
+{
+ u32 reg, pwreg;
+ u16 eeprom;
+ u32 data, gdata;
+ u8 t, i;
+ enum nl80211_band band = chan->band;
+ int delta;
+
+ /* Warn user if bw_comp is set in EEPROM */
+ delta = rt2800_get_txpower_bw_comp(rt2x00dev, band);
+
+ if (delta)
+ rt2x00_warn(rt2x00dev, "ignoring EEPROM HT40 power delta: %d\n",
+ delta);
+
+ /* populate TX_PWR_CFG_0 up to TX_PWR_CFG_4 from EEPROM for HT20, limit
rt2x00: mt7620: yet another beauty session So here is another round of improvements for MT7620 WiFi. This commit fixes a few significant issues related to TX_PWR_CFG_x and TX_ALC and also makes the code more readable by adding register descriptions for things added for MT7620 and use the usual bit-field access macros and the now defined macros instead of plain bit-ops and magic numbers. Properly describe EEPROM_TARGET_POWER at word 0x68 (== byte 0xD0) and thereby fix internal TXALC which would otherwise just read out-of-bounds of the EEPROM map. Split-out tx-power/ALC related stuff into an additional function. Fix VCO calibration, it was carried out properly in the channel switching but incomplete in the actual VCO calibration function. Also there is no need to trigger VCO calibration in channel switching, the VCO calibration function is already being called at this point. Remove it from channel switching function to avoid redundant code. The TX power calibration differs significantly from all other Mediatek/Ralink chips: They finally allow 0.5dB steps stored as 8-bit values for (almost) each bitrate -- and promptly ran out of space and for some reason didn't want to change the EEPROM layout. The hence opted for a scheme of sharing values for some adjecent bitrates and a highly over-complicated (or obfuscated?) way to populate the TX_PWR_CFG_x registers with the values stored in the EEPROM. The code here now looks much less complicated than what you see in the vendor's driver, however, it does the exact same thing: bGpwrdeltaMinus is a constant and always TRUE, hence half of the code was dead. Gpwrdelta is always 0 (rather than using the value read from the EEPROM). What remains is some very grotesque effort to avoid 0x20, probably some hardware bug related to some misunderstanding of what a singed 8-bit value is (imagine: if it was a signed 6-bit value then someone could believe that 0x20 == 0x0). And then they didn't clean it up once they later on anandonned that whole story of having a constant offset for 40 MHz channels and just set the offset to be constant 0 -- there is no effort for avoiding 0x20 for the 20 MHz values stored in the EEPROM, hence that's probably just a forbidden value in the EEPROM specs and won't appear anyway... Anyway, the whole thing felt like solving some college math test where in the end everything cancels out and the result equals 0 ;) To make sure that channel bandwidth power compensation really doesn't need to be taken care of, output a warning when the corresponding value stored in the EEPROM is non-zero. Also there is no apparent reason to refrain from initializing RFCSR register 13, it doesn't fail what-so-ever. Signed-off-by: Daniel Golle <daniel@makrotopia.org>
2017-02-13 05:25:35 +00:00
+ * value to 0x3f and replace 0x20 by 0x21 as this is what the vendor
+ * driver does as well, though it looks kinda wrong.
+ * Maybe some misunderstanding of what a signed 8-bit value is? Maybe
+ * the hardware has a problem handling 0x20, and as the code initially
+ * used a fixed offset between HT20 and HT40 rates they had to work-
+ * around that issue and most likely just forgot about it later on.
+ * Maybe we should use rt2800_get_txpower_bw_comp() here as well,
+ * however, the corresponding EEPROM value is not respected by the
+ * vendor driver, so maybe this is rather being taken care of the
+ * TXALC and the driver doesn't need to handle it...?
+ * Though this is all very awkward, just do as they did, as that's what
+ * board vendors expected when they populated the EEPROM...
+ */
+ for (i = 0; i < 5; i++) {
+ rt2800_eeprom_read_from_array(rt2x00dev, EEPROM_TXPOWER_BYRATE,
+ i * 2, &eeprom);
+
+ data = eeprom;
+
+ t = eeprom & 0x3f;
+ if (t == 32)
+ t++;
+
+ gdata = t;
+
+ t = (eeprom & 0x3f00) >> 8;
+ if (t == 32)
+ t++;
+
+ gdata |= (t << 8);
+
+ rt2800_eeprom_read_from_array(rt2x00dev, EEPROM_TXPOWER_BYRATE,
+ (i * 2) + 1, &eeprom);
+
+ t = eeprom & 0x3f;
+ if (t == 32)
+ t++;
+
+ gdata |= (t << 16);
+
+ t = (eeprom & 0x3f00) >> 8;
+ if (t == 32)
+ t++;
+
+ gdata |= (t << 24);
+ data |= (eeprom << 16);
+
+ if (!test_bit(CONFIG_CHANNEL_HT40, &rt2x00dev->flags)) {
+ /* HT20 */
+ if (data != 0xffffffff)
+ rt2800_register_write(rt2x00dev,
+ TX_PWR_CFG_0 + (i * 4),
+ data);
+ } else {
+ /* HT40 */
+ if (gdata != 0xffffffff)
+ rt2800_register_write(rt2x00dev,
+ TX_PWR_CFG_0 + (i * 4),
+ gdata);
+ }
+ }
+
+ /* Aparently Ralink ran out of space in the BYRATE calibration section
+ * of the EERPOM which is copied to the corresponding TX_PWR_CFG_x
+ * registers. As recent 2T chips use 8-bit instead of 4-bit values for
+ * power-offsets more space would be needed. Ralink decided to keep the
+ * EEPROM layout untouched and rather have some shared values covering
+ * multiple bitrates.
rt2x00: mt7620: yet another beauty session So here is another round of improvements for MT7620 WiFi. This commit fixes a few significant issues related to TX_PWR_CFG_x and TX_ALC and also makes the code more readable by adding register descriptions for things added for MT7620 and use the usual bit-field access macros and the now defined macros instead of plain bit-ops and magic numbers. Properly describe EEPROM_TARGET_POWER at word 0x68 (== byte 0xD0) and thereby fix internal TXALC which would otherwise just read out-of-bounds of the EEPROM map. Split-out tx-power/ALC related stuff into an additional function. Fix VCO calibration, it was carried out properly in the channel switching but incomplete in the actual VCO calibration function. Also there is no need to trigger VCO calibration in channel switching, the VCO calibration function is already being called at this point. Remove it from channel switching function to avoid redundant code. The TX power calibration differs significantly from all other Mediatek/Ralink chips: They finally allow 0.5dB steps stored as 8-bit values for (almost) each bitrate -- and promptly ran out of space and for some reason didn't want to change the EEPROM layout. The hence opted for a scheme of sharing values for some adjecent bitrates and a highly over-complicated (or obfuscated?) way to populate the TX_PWR_CFG_x registers with the values stored in the EEPROM. The code here now looks much less complicated than what you see in the vendor's driver, however, it does the exact same thing: bGpwrdeltaMinus is a constant and always TRUE, hence half of the code was dead. Gpwrdelta is always 0 (rather than using the value read from the EEPROM). What remains is some very grotesque effort to avoid 0x20, probably some hardware bug related to some misunderstanding of what a singed 8-bit value is (imagine: if it was a signed 6-bit value then someone could believe that 0x20 == 0x0). And then they didn't clean it up once they later on anandonned that whole story of having a constant offset for 40 MHz channels and just set the offset to be constant 0 -- there is no effort for avoiding 0x20 for the 20 MHz values stored in the EEPROM, hence that's probably just a forbidden value in the EEPROM specs and won't appear anyway... Anyway, the whole thing felt like solving some college math test where in the end everything cancels out and the result equals 0 ;) To make sure that channel bandwidth power compensation really doesn't need to be taken care of, output a warning when the corresponding value stored in the EEPROM is non-zero. Also there is no apparent reason to refrain from initializing RFCSR register 13, it doesn't fail what-so-ever. Signed-off-by: Daniel Golle <daniel@makrotopia.org>
2017-02-13 05:25:35 +00:00
+ * Populate the registers not covered by the EEPROM in the same way the
+ * vendor driver does.
+ */
+
+ /* For OFDM 54MBS use value from OFDM 48MBS */
+ pwreg = 0;
+ rt2800_register_read(rt2x00dev, TX_PWR_CFG_1, &reg);
+ t = rt2x00_get_field32(reg, TX_PWR_CFG_1B_48MBS);
+ rt2x00_set_field32(&pwreg, TX_PWR_CFG_7B_54MBS, t);
+
+ /* For MCS 7 use value from MCS 6 */
+ rt2800_register_read(rt2x00dev, TX_PWR_CFG_2, &reg);
+ t = rt2x00_get_field32(reg, TX_PWR_CFG_2B_MCS6_MCS7);
+ rt2x00_set_field32(&pwreg, TX_PWR_CFG_7B_MCS7, t);
+ rt2800_register_write(rt2x00dev, TX_PWR_CFG_7, pwreg);
+
+ /* For MCS 15 use value from MCS 14 */
+ pwreg = 0;
+ rt2800_register_read(rt2x00dev, TX_PWR_CFG_3, &reg);
+ t = rt2x00_get_field32(reg, TX_PWR_CFG_3B_MCS14);
+ rt2x00_set_field32(&pwreg, TX_PWR_CFG_8B_MCS15, t);
+ rt2800_register_write(rt2x00dev, TX_PWR_CFG_8, pwreg);
+
+ /* For STBC MCS 7 use value from STBC MCS 6 */
+ pwreg = 0;
+ rt2800_register_read(rt2x00dev, TX_PWR_CFG_4, &reg);
+ t = rt2x00_get_field32(reg, TX_PWR_CFG_4B_STBC_MCS6);
+ rt2x00_set_field32(&pwreg, TX_PWR_CFG_9B_STBC_MCS7, t);
+ rt2800_register_write(rt2x00dev, TX_PWR_CFG_9, pwreg);
+
+ rt2800_config_alc(rt2x00dev, chan, power_level);
+
+ /* TODO: temperature compensation code! */
+}
+
/*
* We configure transmit power using MAC TX_PWR_CFG_{0,...,N} registers and
* BBP R1 register. TX_PWR_CFG_X allow to configure per rate TX power values,
@@ -4378,6 +4810,8 @@ static void rt2800_config_txpower(struct rt2x00_dev *rt2x00dev,
{
if (rt2x00_rt(rt2x00dev, RT3593))
rt2x00: mt7620: yet another beauty session So here is another round of improvements for MT7620 WiFi. This commit fixes a few significant issues related to TX_PWR_CFG_x and TX_ALC and also makes the code more readable by adding register descriptions for things added for MT7620 and use the usual bit-field access macros and the now defined macros instead of plain bit-ops and magic numbers. Properly describe EEPROM_TARGET_POWER at word 0x68 (== byte 0xD0) and thereby fix internal TXALC which would otherwise just read out-of-bounds of the EEPROM map. Split-out tx-power/ALC related stuff into an additional function. Fix VCO calibration, it was carried out properly in the channel switching but incomplete in the actual VCO calibration function. Also there is no need to trigger VCO calibration in channel switching, the VCO calibration function is already being called at this point. Remove it from channel switching function to avoid redundant code. The TX power calibration differs significantly from all other Mediatek/Ralink chips: They finally allow 0.5dB steps stored as 8-bit values for (almost) each bitrate -- and promptly ran out of space and for some reason didn't want to change the EEPROM layout. The hence opted for a scheme of sharing values for some adjecent bitrates and a highly over-complicated (or obfuscated?) way to populate the TX_PWR_CFG_x registers with the values stored in the EEPROM. The code here now looks much less complicated than what you see in the vendor's driver, however, it does the exact same thing: bGpwrdeltaMinus is a constant and always TRUE, hence half of the code was dead. Gpwrdelta is always 0 (rather than using the value read from the EEPROM). What remains is some very grotesque effort to avoid 0x20, probably some hardware bug related to some misunderstanding of what a singed 8-bit value is (imagine: if it was a signed 6-bit value then someone could believe that 0x20 == 0x0). And then they didn't clean it up once they later on anandonned that whole story of having a constant offset for 40 MHz channels and just set the offset to be constant 0 -- there is no effort for avoiding 0x20 for the 20 MHz values stored in the EEPROM, hence that's probably just a forbidden value in the EEPROM specs and won't appear anyway... Anyway, the whole thing felt like solving some college math test where in the end everything cancels out and the result equals 0 ;) To make sure that channel bandwidth power compensation really doesn't need to be taken care of, output a warning when the corresponding value stored in the EEPROM is non-zero. Also there is no apparent reason to refrain from initializing RFCSR register 13, it doesn't fail what-so-ever. Signed-off-by: Daniel Golle <daniel@makrotopia.org>
2017-02-13 05:25:35 +00:00
rt2800_config_txpower_rt3593(rt2x00dev, chan, power_level);
+ else if (rt2x00_rt(rt2x00dev, RT6352))
+ rt2800_config_txpower_rt6352(rt2x00dev, chan, power_level);
rt2x00: mt7620: yet another beauty session So here is another round of improvements for MT7620 WiFi. This commit fixes a few significant issues related to TX_PWR_CFG_x and TX_ALC and also makes the code more readable by adding register descriptions for things added for MT7620 and use the usual bit-field access macros and the now defined macros instead of plain bit-ops and magic numbers. Properly describe EEPROM_TARGET_POWER at word 0x68 (== byte 0xD0) and thereby fix internal TXALC which would otherwise just read out-of-bounds of the EEPROM map. Split-out tx-power/ALC related stuff into an additional function. Fix VCO calibration, it was carried out properly in the channel switching but incomplete in the actual VCO calibration function. Also there is no need to trigger VCO calibration in channel switching, the VCO calibration function is already being called at this point. Remove it from channel switching function to avoid redundant code. The TX power calibration differs significantly from all other Mediatek/Ralink chips: They finally allow 0.5dB steps stored as 8-bit values for (almost) each bitrate -- and promptly ran out of space and for some reason didn't want to change the EEPROM layout. The hence opted for a scheme of sharing values for some adjecent bitrates and a highly over-complicated (or obfuscated?) way to populate the TX_PWR_CFG_x registers with the values stored in the EEPROM. The code here now looks much less complicated than what you see in the vendor's driver, however, it does the exact same thing: bGpwrdeltaMinus is a constant and always TRUE, hence half of the code was dead. Gpwrdelta is always 0 (rather than using the value read from the EEPROM). What remains is some very grotesque effort to avoid 0x20, probably some hardware bug related to some misunderstanding of what a singed 8-bit value is (imagine: if it was a signed 6-bit value then someone could believe that 0x20 == 0x0). And then they didn't clean it up once they later on anandonned that whole story of having a constant offset for 40 MHz channels and just set the offset to be constant 0 -- there is no effort for avoiding 0x20 for the 20 MHz values stored in the EEPROM, hence that's probably just a forbidden value in the EEPROM specs and won't appear anyway... Anyway, the whole thing felt like solving some college math test where in the end everything cancels out and the result equals 0 ;) To make sure that channel bandwidth power compensation really doesn't need to be taken care of, output a warning when the corresponding value stored in the EEPROM is non-zero. Also there is no apparent reason to refrain from initializing RFCSR register 13, it doesn't fail what-so-ever. Signed-off-by: Daniel Golle <daniel@makrotopia.org>
2017-02-13 05:25:35 +00:00
else
rt2800_config_txpower_rt28xx(rt2x00dev, chan, power_level);
}
@@ -4393,6 +4827,7 @@ void rt2800_vco_calibration(struct rt2x00_dev *rt2x00dev)
rt2x00: mt7620: yet another beauty session So here is another round of improvements for MT7620 WiFi. This commit fixes a few significant issues related to TX_PWR_CFG_x and TX_ALC and also makes the code more readable by adding register descriptions for things added for MT7620 and use the usual bit-field access macros and the now defined macros instead of plain bit-ops and magic numbers. Properly describe EEPROM_TARGET_POWER at word 0x68 (== byte 0xD0) and thereby fix internal TXALC which would otherwise just read out-of-bounds of the EEPROM map. Split-out tx-power/ALC related stuff into an additional function. Fix VCO calibration, it was carried out properly in the channel switching but incomplete in the actual VCO calibration function. Also there is no need to trigger VCO calibration in channel switching, the VCO calibration function is already being called at this point. Remove it from channel switching function to avoid redundant code. The TX power calibration differs significantly from all other Mediatek/Ralink chips: They finally allow 0.5dB steps stored as 8-bit values for (almost) each bitrate -- and promptly ran out of space and for some reason didn't want to change the EEPROM layout. The hence opted for a scheme of sharing values for some adjecent bitrates and a highly over-complicated (or obfuscated?) way to populate the TX_PWR_CFG_x registers with the values stored in the EEPROM. The code here now looks much less complicated than what you see in the vendor's driver, however, it does the exact same thing: bGpwrdeltaMinus is a constant and always TRUE, hence half of the code was dead. Gpwrdelta is always 0 (rather than using the value read from the EEPROM). What remains is some very grotesque effort to avoid 0x20, probably some hardware bug related to some misunderstanding of what a singed 8-bit value is (imagine: if it was a signed 6-bit value then someone could believe that 0x20 == 0x0). And then they didn't clean it up once they later on anandonned that whole story of having a constant offset for 40 MHz channels and just set the offset to be constant 0 -- there is no effort for avoiding 0x20 for the 20 MHz values stored in the EEPROM, hence that's probably just a forbidden value in the EEPROM specs and won't appear anyway... Anyway, the whole thing felt like solving some college math test where in the end everything cancels out and the result equals 0 ;) To make sure that channel bandwidth power compensation really doesn't need to be taken care of, output a warning when the corresponding value stored in the EEPROM is non-zero. Also there is no apparent reason to refrain from initializing RFCSR register 13, it doesn't fail what-so-ever. Signed-off-by: Daniel Golle <daniel@makrotopia.org>
2017-02-13 05:25:35 +00:00
{
u32 tx_pin;
u8 rfcsr;
+ unsigned long min_sleep = 0;
/*
* A voltage-controlled oscillator(VCO) is an electronic oscillator
@@ -4431,6 +4866,15 @@ void rt2800_vco_calibration(struct rt2x00_dev *rt2x00dev)
rt2x00: mt7620: yet another beauty session So here is another round of improvements for MT7620 WiFi. This commit fixes a few significant issues related to TX_PWR_CFG_x and TX_ALC and also makes the code more readable by adding register descriptions for things added for MT7620 and use the usual bit-field access macros and the now defined macros instead of plain bit-ops and magic numbers. Properly describe EEPROM_TARGET_POWER at word 0x68 (== byte 0xD0) and thereby fix internal TXALC which would otherwise just read out-of-bounds of the EEPROM map. Split-out tx-power/ALC related stuff into an additional function. Fix VCO calibration, it was carried out properly in the channel switching but incomplete in the actual VCO calibration function. Also there is no need to trigger VCO calibration in channel switching, the VCO calibration function is already being called at this point. Remove it from channel switching function to avoid redundant code. The TX power calibration differs significantly from all other Mediatek/Ralink chips: They finally allow 0.5dB steps stored as 8-bit values for (almost) each bitrate -- and promptly ran out of space and for some reason didn't want to change the EEPROM layout. The hence opted for a scheme of sharing values for some adjecent bitrates and a highly over-complicated (or obfuscated?) way to populate the TX_PWR_CFG_x registers with the values stored in the EEPROM. The code here now looks much less complicated than what you see in the vendor's driver, however, it does the exact same thing: bGpwrdeltaMinus is a constant and always TRUE, hence half of the code was dead. Gpwrdelta is always 0 (rather than using the value read from the EEPROM). What remains is some very grotesque effort to avoid 0x20, probably some hardware bug related to some misunderstanding of what a singed 8-bit value is (imagine: if it was a signed 6-bit value then someone could believe that 0x20 == 0x0). And then they didn't clean it up once they later on anandonned that whole story of having a constant offset for 40 MHz channels and just set the offset to be constant 0 -- there is no effort for avoiding 0x20 for the 20 MHz values stored in the EEPROM, hence that's probably just a forbidden value in the EEPROM specs and won't appear anyway... Anyway, the whole thing felt like solving some college math test where in the end everything cancels out and the result equals 0 ;) To make sure that channel bandwidth power compensation really doesn't need to be taken care of, output a warning when the corresponding value stored in the EEPROM is non-zero. Also there is no apparent reason to refrain from initializing RFCSR register 13, it doesn't fail what-so-ever. Signed-off-by: Daniel Golle <daniel@makrotopia.org>
2017-02-13 05:25:35 +00:00
rt2800_rfcsr_read(rt2x00dev, 3, &rfcsr);
rt2x00_set_field8(&rfcsr, RFCSR3_VCOCAL_EN, 1);
rt2800_rfcsr_write(rt2x00dev, 3, rfcsr);
rt2x00: mt7620: yet another beauty session So here is another round of improvements for MT7620 WiFi. This commit fixes a few significant issues related to TX_PWR_CFG_x and TX_ALC and also makes the code more readable by adding register descriptions for things added for MT7620 and use the usual bit-field access macros and the now defined macros instead of plain bit-ops and magic numbers. Properly describe EEPROM_TARGET_POWER at word 0x68 (== byte 0xD0) and thereby fix internal TXALC which would otherwise just read out-of-bounds of the EEPROM map. Split-out tx-power/ALC related stuff into an additional function. Fix VCO calibration, it was carried out properly in the channel switching but incomplete in the actual VCO calibration function. Also there is no need to trigger VCO calibration in channel switching, the VCO calibration function is already being called at this point. Remove it from channel switching function to avoid redundant code. The TX power calibration differs significantly from all other Mediatek/Ralink chips: They finally allow 0.5dB steps stored as 8-bit values for (almost) each bitrate -- and promptly ran out of space and for some reason didn't want to change the EEPROM layout. The hence opted for a scheme of sharing values for some adjecent bitrates and a highly over-complicated (or obfuscated?) way to populate the TX_PWR_CFG_x registers with the values stored in the EEPROM. The code here now looks much less complicated than what you see in the vendor's driver, however, it does the exact same thing: bGpwrdeltaMinus is a constant and always TRUE, hence half of the code was dead. Gpwrdelta is always 0 (rather than using the value read from the EEPROM). What remains is some very grotesque effort to avoid 0x20, probably some hardware bug related to some misunderstanding of what a singed 8-bit value is (imagine: if it was a signed 6-bit value then someone could believe that 0x20 == 0x0). And then they didn't clean it up once they later on anandonned that whole story of having a constant offset for 40 MHz channels and just set the offset to be constant 0 -- there is no effort for avoiding 0x20 for the 20 MHz values stored in the EEPROM, hence that's probably just a forbidden value in the EEPROM specs and won't appear anyway... Anyway, the whole thing felt like solving some college math test where in the end everything cancels out and the result equals 0 ;) To make sure that channel bandwidth power compensation really doesn't need to be taken care of, output a warning when the corresponding value stored in the EEPROM is non-zero. Also there is no apparent reason to refrain from initializing RFCSR register 13, it doesn't fail what-so-ever. Signed-off-by: Daniel Golle <daniel@makrotopia.org>
2017-02-13 05:25:35 +00:00
+ min_sleep = 1000;
+ break;
+ case RF7620:
rt2x00: mt7620: yet another beauty session So here is another round of improvements for MT7620 WiFi. This commit fixes a few significant issues related to TX_PWR_CFG_x and TX_ALC and also makes the code more readable by adding register descriptions for things added for MT7620 and use the usual bit-field access macros and the now defined macros instead of plain bit-ops and magic numbers. Properly describe EEPROM_TARGET_POWER at word 0x68 (== byte 0xD0) and thereby fix internal TXALC which would otherwise just read out-of-bounds of the EEPROM map. Split-out tx-power/ALC related stuff into an additional function. Fix VCO calibration, it was carried out properly in the channel switching but incomplete in the actual VCO calibration function. Also there is no need to trigger VCO calibration in channel switching, the VCO calibration function is already being called at this point. Remove it from channel switching function to avoid redundant code. The TX power calibration differs significantly from all other Mediatek/Ralink chips: They finally allow 0.5dB steps stored as 8-bit values for (almost) each bitrate -- and promptly ran out of space and for some reason didn't want to change the EEPROM layout. The hence opted for a scheme of sharing values for some adjecent bitrates and a highly over-complicated (or obfuscated?) way to populate the TX_PWR_CFG_x registers with the values stored in the EEPROM. The code here now looks much less complicated than what you see in the vendor's driver, however, it does the exact same thing: bGpwrdeltaMinus is a constant and always TRUE, hence half of the code was dead. Gpwrdelta is always 0 (rather than using the value read from the EEPROM). What remains is some very grotesque effort to avoid 0x20, probably some hardware bug related to some misunderstanding of what a singed 8-bit value is (imagine: if it was a signed 6-bit value then someone could believe that 0x20 == 0x0). And then they didn't clean it up once they later on anandonned that whole story of having a constant offset for 40 MHz channels and just set the offset to be constant 0 -- there is no effort for avoiding 0x20 for the 20 MHz values stored in the EEPROM, hence that's probably just a forbidden value in the EEPROM specs and won't appear anyway... Anyway, the whole thing felt like solving some college math test where in the end everything cancels out and the result equals 0 ;) To make sure that channel bandwidth power compensation really doesn't need to be taken care of, output a warning when the corresponding value stored in the EEPROM is non-zero. Also there is no apparent reason to refrain from initializing RFCSR register 13, it doesn't fail what-so-ever. Signed-off-by: Daniel Golle <daniel@makrotopia.org>
2017-02-13 05:25:35 +00:00
+ rt2800_rfcsr_write(rt2x00dev, 5, 0x40);
+ rt2800_rfcsr_write(rt2x00dev, 4, 0x0C);
+ rt2800_rfcsr_read(rt2x00dev, 4, &rfcsr);
rt2x00: mt7620: yet another beauty session So here is another round of improvements for MT7620 WiFi. This commit fixes a few significant issues related to TX_PWR_CFG_x and TX_ALC and also makes the code more readable by adding register descriptions for things added for MT7620 and use the usual bit-field access macros and the now defined macros instead of plain bit-ops and magic numbers. Properly describe EEPROM_TARGET_POWER at word 0x68 (== byte 0xD0) and thereby fix internal TXALC which would otherwise just read out-of-bounds of the EEPROM map. Split-out tx-power/ALC related stuff into an additional function. Fix VCO calibration, it was carried out properly in the channel switching but incomplete in the actual VCO calibration function. Also there is no need to trigger VCO calibration in channel switching, the VCO calibration function is already being called at this point. Remove it from channel switching function to avoid redundant code. The TX power calibration differs significantly from all other Mediatek/Ralink chips: They finally allow 0.5dB steps stored as 8-bit values for (almost) each bitrate -- and promptly ran out of space and for some reason didn't want to change the EEPROM layout. The hence opted for a scheme of sharing values for some adjecent bitrates and a highly over-complicated (or obfuscated?) way to populate the TX_PWR_CFG_x registers with the values stored in the EEPROM. The code here now looks much less complicated than what you see in the vendor's driver, however, it does the exact same thing: bGpwrdeltaMinus is a constant and always TRUE, hence half of the code was dead. Gpwrdelta is always 0 (rather than using the value read from the EEPROM). What remains is some very grotesque effort to avoid 0x20, probably some hardware bug related to some misunderstanding of what a singed 8-bit value is (imagine: if it was a signed 6-bit value then someone could believe that 0x20 == 0x0). And then they didn't clean it up once they later on anandonned that whole story of having a constant offset for 40 MHz channels and just set the offset to be constant 0 -- there is no effort for avoiding 0x20 for the 20 MHz values stored in the EEPROM, hence that's probably just a forbidden value in the EEPROM specs and won't appear anyway... Anyway, the whole thing felt like solving some college math test where in the end everything cancels out and the result equals 0 ;) To make sure that channel bandwidth power compensation really doesn't need to be taken care of, output a warning when the corresponding value stored in the EEPROM is non-zero. Also there is no apparent reason to refrain from initializing RFCSR register 13, it doesn't fail what-so-ever. Signed-off-by: Daniel Golle <daniel@makrotopia.org>
2017-02-13 05:25:35 +00:00
+ rt2x00_set_field8(&rfcsr, RFCSR4_VCOCAL_EN, 1);
+ rt2800_rfcsr_write(rt2x00dev, 4, rfcsr);
rt2x00: mt7620: yet another beauty session So here is another round of improvements for MT7620 WiFi. This commit fixes a few significant issues related to TX_PWR_CFG_x and TX_ALC and also makes the code more readable by adding register descriptions for things added for MT7620 and use the usual bit-field access macros and the now defined macros instead of plain bit-ops and magic numbers. Properly describe EEPROM_TARGET_POWER at word 0x68 (== byte 0xD0) and thereby fix internal TXALC which would otherwise just read out-of-bounds of the EEPROM map. Split-out tx-power/ALC related stuff into an additional function. Fix VCO calibration, it was carried out properly in the channel switching but incomplete in the actual VCO calibration function. Also there is no need to trigger VCO calibration in channel switching, the VCO calibration function is already being called at this point. Remove it from channel switching function to avoid redundant code. The TX power calibration differs significantly from all other Mediatek/Ralink chips: They finally allow 0.5dB steps stored as 8-bit values for (almost) each bitrate -- and promptly ran out of space and for some reason didn't want to change the EEPROM layout. The hence opted for a scheme of sharing values for some adjecent bitrates and a highly over-complicated (or obfuscated?) way to populate the TX_PWR_CFG_x registers with the values stored in the EEPROM. The code here now looks much less complicated than what you see in the vendor's driver, however, it does the exact same thing: bGpwrdeltaMinus is a constant and always TRUE, hence half of the code was dead. Gpwrdelta is always 0 (rather than using the value read from the EEPROM). What remains is some very grotesque effort to avoid 0x20, probably some hardware bug related to some misunderstanding of what a singed 8-bit value is (imagine: if it was a signed 6-bit value then someone could believe that 0x20 == 0x0). And then they didn't clean it up once they later on anandonned that whole story of having a constant offset for 40 MHz channels and just set the offset to be constant 0 -- there is no effort for avoiding 0x20 for the 20 MHz values stored in the EEPROM, hence that's probably just a forbidden value in the EEPROM specs and won't appear anyway... Anyway, the whole thing felt like solving some college math test where in the end everything cancels out and the result equals 0 ;) To make sure that channel bandwidth power compensation really doesn't need to be taken care of, output a warning when the corresponding value stored in the EEPROM is non-zero. Also there is no apparent reason to refrain from initializing RFCSR register 13, it doesn't fail what-so-ever. Signed-off-by: Daniel Golle <daniel@makrotopia.org>
2017-02-13 05:25:35 +00:00
+ min_sleep = 2000;
break;
default:
WARN_ONCE(1, "Not supported RF chipet %x for VCO recalibration",
@@ -4438,7 +4882,8 @@ void rt2800_vco_calibration(struct rt2x00_dev *rt2x00dev)
rt2x00: mt7620: yet another beauty session So here is another round of improvements for MT7620 WiFi. This commit fixes a few significant issues related to TX_PWR_CFG_x and TX_ALC and also makes the code more readable by adding register descriptions for things added for MT7620 and use the usual bit-field access macros and the now defined macros instead of plain bit-ops and magic numbers. Properly describe EEPROM_TARGET_POWER at word 0x68 (== byte 0xD0) and thereby fix internal TXALC which would otherwise just read out-of-bounds of the EEPROM map. Split-out tx-power/ALC related stuff into an additional function. Fix VCO calibration, it was carried out properly in the channel switching but incomplete in the actual VCO calibration function. Also there is no need to trigger VCO calibration in channel switching, the VCO calibration function is already being called at this point. Remove it from channel switching function to avoid redundant code. The TX power calibration differs significantly from all other Mediatek/Ralink chips: They finally allow 0.5dB steps stored as 8-bit values for (almost) each bitrate -- and promptly ran out of space and for some reason didn't want to change the EEPROM layout. The hence opted for a scheme of sharing values for some adjecent bitrates and a highly over-complicated (or obfuscated?) way to populate the TX_PWR_CFG_x registers with the values stored in the EEPROM. The code here now looks much less complicated than what you see in the vendor's driver, however, it does the exact same thing: bGpwrdeltaMinus is a constant and always TRUE, hence half of the code was dead. Gpwrdelta is always 0 (rather than using the value read from the EEPROM). What remains is some very grotesque effort to avoid 0x20, probably some hardware bug related to some misunderstanding of what a singed 8-bit value is (imagine: if it was a signed 6-bit value then someone could believe that 0x20 == 0x0). And then they didn't clean it up once they later on anandonned that whole story of having a constant offset for 40 MHz channels and just set the offset to be constant 0 -- there is no effort for avoiding 0x20 for the 20 MHz values stored in the EEPROM, hence that's probably just a forbidden value in the EEPROM specs and won't appear anyway... Anyway, the whole thing felt like solving some college math test where in the end everything cancels out and the result equals 0 ;) To make sure that channel bandwidth power compensation really doesn't need to be taken care of, output a warning when the corresponding value stored in the EEPROM is non-zero. Also there is no apparent reason to refrain from initializing RFCSR register 13, it doesn't fail what-so-ever. Signed-off-by: Daniel Golle <daniel@makrotopia.org>
2017-02-13 05:25:35 +00:00
return;
}
- usleep_range(1000, 1500);
+ if (min_sleep > 0)
+ usleep_range(min_sleep, min_sleep * 2);
rt2800_register_read(rt2x00dev, TX_PIN_CFG, &tx_pin);
if (rt2x00dev->rf_channel <= 14) {
@@ -4470,6 +4915,42 @@ void rt2800_vco_calibration(struct rt2x00_dev *rt2x00dev)
rt2x00: mt7620: yet another beauty session So here is another round of improvements for MT7620 WiFi. This commit fixes a few significant issues related to TX_PWR_CFG_x and TX_ALC and also makes the code more readable by adding register descriptions for things added for MT7620 and use the usual bit-field access macros and the now defined macros instead of plain bit-ops and magic numbers. Properly describe EEPROM_TARGET_POWER at word 0x68 (== byte 0xD0) and thereby fix internal TXALC which would otherwise just read out-of-bounds of the EEPROM map. Split-out tx-power/ALC related stuff into an additional function. Fix VCO calibration, it was carried out properly in the channel switching but incomplete in the actual VCO calibration function. Also there is no need to trigger VCO calibration in channel switching, the VCO calibration function is already being called at this point. Remove it from channel switching function to avoid redundant code. The TX power calibration differs significantly from all other Mediatek/Ralink chips: They finally allow 0.5dB steps stored as 8-bit values for (almost) each bitrate -- and promptly ran out of space and for some reason didn't want to change the EEPROM layout. The hence opted for a scheme of sharing values for some adjecent bitrates and a highly over-complicated (or obfuscated?) way to populate the TX_PWR_CFG_x registers with the values stored in the EEPROM. The code here now looks much less complicated than what you see in the vendor's driver, however, it does the exact same thing: bGpwrdeltaMinus is a constant and always TRUE, hence half of the code was dead. Gpwrdelta is always 0 (rather than using the value read from the EEPROM). What remains is some very grotesque effort to avoid 0x20, probably some hardware bug related to some misunderstanding of what a singed 8-bit value is (imagine: if it was a signed 6-bit value then someone could believe that 0x20 == 0x0). And then they didn't clean it up once they later on anandonned that whole story of having a constant offset for 40 MHz channels and just set the offset to be constant 0 -- there is no effort for avoiding 0x20 for the 20 MHz values stored in the EEPROM, hence that's probably just a forbidden value in the EEPROM specs and won't appear anyway... Anyway, the whole thing felt like solving some college math test where in the end everything cancels out and the result equals 0 ;) To make sure that channel bandwidth power compensation really doesn't need to be taken care of, output a warning when the corresponding value stored in the EEPROM is non-zero. Also there is no apparent reason to refrain from initializing RFCSR register 13, it doesn't fail what-so-ever. Signed-off-by: Daniel Golle <daniel@makrotopia.org>
2017-02-13 05:25:35 +00:00
}
rt2800_register_write(rt2x00dev, TX_PIN_CFG, tx_pin);
+ if (rt2x00_rt(rt2x00dev, RT6352)) {
rt2x00: mt7620: yet another beauty session So here is another round of improvements for MT7620 WiFi. This commit fixes a few significant issues related to TX_PWR_CFG_x and TX_ALC and also makes the code more readable by adding register descriptions for things added for MT7620 and use the usual bit-field access macros and the now defined macros instead of plain bit-ops and magic numbers. Properly describe EEPROM_TARGET_POWER at word 0x68 (== byte 0xD0) and thereby fix internal TXALC which would otherwise just read out-of-bounds of the EEPROM map. Split-out tx-power/ALC related stuff into an additional function. Fix VCO calibration, it was carried out properly in the channel switching but incomplete in the actual VCO calibration function. Also there is no need to trigger VCO calibration in channel switching, the VCO calibration function is already being called at this point. Remove it from channel switching function to avoid redundant code. The TX power calibration differs significantly from all other Mediatek/Ralink chips: They finally allow 0.5dB steps stored as 8-bit values for (almost) each bitrate -- and promptly ran out of space and for some reason didn't want to change the EEPROM layout. The hence opted for a scheme of sharing values for some adjecent bitrates and a highly over-complicated (or obfuscated?) way to populate the TX_PWR_CFG_x registers with the values stored in the EEPROM. The code here now looks much less complicated than what you see in the vendor's driver, however, it does the exact same thing: bGpwrdeltaMinus is a constant and always TRUE, hence half of the code was dead. Gpwrdelta is always 0 (rather than using the value read from the EEPROM). What remains is some very grotesque effort to avoid 0x20, probably some hardware bug related to some misunderstanding of what a singed 8-bit value is (imagine: if it was a signed 6-bit value then someone could believe that 0x20 == 0x0). And then they didn't clean it up once they later on anandonned that whole story of having a constant offset for 40 MHz channels and just set the offset to be constant 0 -- there is no effort for avoiding 0x20 for the 20 MHz values stored in the EEPROM, hence that's probably just a forbidden value in the EEPROM specs and won't appear anyway... Anyway, the whole thing felt like solving some college math test where in the end everything cancels out and the result equals 0 ;) To make sure that channel bandwidth power compensation really doesn't need to be taken care of, output a warning when the corresponding value stored in the EEPROM is non-zero. Also there is no apparent reason to refrain from initializing RFCSR register 13, it doesn't fail what-so-ever. Signed-off-by: Daniel Golle <daniel@makrotopia.org>
2017-02-13 05:25:35 +00:00
+ if (rt2x00dev->default_ant.tx_chain_num == 1) {
+ rt2800_bbp_write(rt2x00dev, 91, 0x07);
+ rt2800_bbp_write(rt2x00dev, 95, 0x1A);
+ rt2800_bbp_write(rt2x00dev, 195, 128);
+ rt2800_bbp_write(rt2x00dev, 196, 0xA0);
+ rt2800_bbp_write(rt2x00dev, 195, 170);
+ rt2800_bbp_write(rt2x00dev, 196, 0x12);
+ rt2800_bbp_write(rt2x00dev, 195, 171);
+ rt2800_bbp_write(rt2x00dev, 196, 0x10);
+ } else {
+ rt2800_bbp_write(rt2x00dev, 91, 0x06);
+ rt2800_bbp_write(rt2x00dev, 95, 0x9A);
+ rt2800_bbp_write(rt2x00dev, 195, 128);
+ rt2800_bbp_write(rt2x00dev, 196, 0xE0);
+ rt2800_bbp_write(rt2x00dev, 195, 170);
+ rt2800_bbp_write(rt2x00dev, 196, 0x30);
+ rt2800_bbp_write(rt2x00dev, 195, 171);
+ rt2800_bbp_write(rt2x00dev, 196, 0x30);
+ }
+
+ if (rt2x00_has_cap_external_lna_bg(rt2x00dev)) {
+ rt2800_bbp_write(rt2x00dev, 75, 0x60);
+ rt2800_bbp_write(rt2x00dev, 76, 0x44);
+ rt2800_bbp_write(rt2x00dev, 79, 0x1C);
+ rt2800_bbp_write(rt2x00dev, 80, 0x0C);
+ rt2800_bbp_write(rt2x00dev, 82, 0xB6);
+ }
+
+ /* On 11A, We should delay and wait RF/BBP to be stable
+ * and the appropriate time should be 1000 micro seconds
+ * 2005/06/05 - On 11G, we also need this delay time.
+ * Otherwise it's difficult to pass the WHQL.
+ */
+ usleep_range(1000, 1500);
+ }
}
EXPORT_SYMBOL_GPL(rt2800_vco_calibration);
@@ -4568,7 +5049,8 @@ static u8 rt2800_get_default_vgc(struct rt2x00_dev *rt2x00dev)
rt2x00_rt(rt2x00dev, RT3593) ||
rt2x00_rt(rt2x00dev, RT5390) ||
rt2x00_rt(rt2x00dev, RT5392) ||
- rt2x00_rt(rt2x00dev, RT5592))
+ rt2x00_rt(rt2x00dev, RT5592) ||
+ rt2x00_rt(rt2x00dev, RT6352))
vgc = 0x1c + (2 * rt2x00dev->lna_gain);
else
vgc = 0x2e + rt2x00dev->lna_gain;
@@ -4795,7 +5277,8 @@ static int rt2800_init_registers(struct rt2x00_dev *rt2x00dev)
0x00000000);
}
} else if (rt2x00_rt(rt2x00dev, RT5390) ||
- rt2x00_rt(rt2x00dev, RT5392)) {
+ rt2x00_rt(rt2x00dev, RT5392) ||
+ rt2x00_rt(rt2x00dev, RT6352)) {
rt2800_register_write(rt2x00dev, TX_SW_CFG0, 0x00000404);
rt2800_register_write(rt2x00dev, TX_SW_CFG1, 0x00080606);
rt2800_register_write(rt2x00dev, TX_SW_CFG2, 0x00000000);
@@ -4805,6 +5288,24 @@ static int rt2800_init_registers(struct rt2x00_dev *rt2x00dev)
rt2800_register_write(rt2x00dev, TX_SW_CFG2, 0x00000000);
} else if (rt2x00_rt(rt2x00dev, RT5350)) {
rt2800_register_write(rt2x00dev, TX_SW_CFG0, 0x00000404);
+ } else if (rt2x00_rt(rt2x00dev, RT6352)) {
+ rt2800_register_write(rt2x00dev, TX_SW_CFG0, 0x00000401);
+ rt2800_register_write(rt2x00dev, TX_SW_CFG1, 0x000C0000);
+ rt2800_register_write(rt2x00dev, TX_SW_CFG2, 0x00000000);
+ rt2800_register_write(rt2x00dev, MIMO_PS_CFG, 0x00000002);
+ rt2800_register_write(rt2x00dev, TX_PIN_CFG, 0x00150F0F);
+ rt2800_register_write(rt2x00dev, TX_ALC_VGA3, 0x06060606);
+ rt2800_register_write(rt2x00dev, TX0_BB_GAIN_ATTEN, 0x0);
+ rt2800_register_write(rt2x00dev, TX1_BB_GAIN_ATTEN, 0x0);
+ rt2800_register_write(rt2x00dev, TX0_RF_GAIN_ATTEN, 0x6C6C666C);
+ rt2800_register_write(rt2x00dev, TX1_RF_GAIN_ATTEN, 0x6C6C666C);
+ rt2800_register_write(rt2x00dev, TX0_RF_GAIN_CORRECT,
+ 0x3630363A);
+ rt2800_register_write(rt2x00dev, TX1_RF_GAIN_CORRECT,
+ 0x3630363A);
rt2x00: mt7620: yet another beauty session So here is another round of improvements for MT7620 WiFi. This commit fixes a few significant issues related to TX_PWR_CFG_x and TX_ALC and also makes the code more readable by adding register descriptions for things added for MT7620 and use the usual bit-field access macros and the now defined macros instead of plain bit-ops and magic numbers. Properly describe EEPROM_TARGET_POWER at word 0x68 (== byte 0xD0) and thereby fix internal TXALC which would otherwise just read out-of-bounds of the EEPROM map. Split-out tx-power/ALC related stuff into an additional function. Fix VCO calibration, it was carried out properly in the channel switching but incomplete in the actual VCO calibration function. Also there is no need to trigger VCO calibration in channel switching, the VCO calibration function is already being called at this point. Remove it from channel switching function to avoid redundant code. The TX power calibration differs significantly from all other Mediatek/Ralink chips: They finally allow 0.5dB steps stored as 8-bit values for (almost) each bitrate -- and promptly ran out of space and for some reason didn't want to change the EEPROM layout. The hence opted for a scheme of sharing values for some adjecent bitrates and a highly over-complicated (or obfuscated?) way to populate the TX_PWR_CFG_x registers with the values stored in the EEPROM. The code here now looks much less complicated than what you see in the vendor's driver, however, it does the exact same thing: bGpwrdeltaMinus is a constant and always TRUE, hence half of the code was dead. Gpwrdelta is always 0 (rather than using the value read from the EEPROM). What remains is some very grotesque effort to avoid 0x20, probably some hardware bug related to some misunderstanding of what a singed 8-bit value is (imagine: if it was a signed 6-bit value then someone could believe that 0x20 == 0x0). And then they didn't clean it up once they later on anandonned that whole story of having a constant offset for 40 MHz channels and just set the offset to be constant 0 -- there is no effort for avoiding 0x20 for the 20 MHz values stored in the EEPROM, hence that's probably just a forbidden value in the EEPROM specs and won't appear anyway... Anyway, the whole thing felt like solving some college math test where in the end everything cancels out and the result equals 0 ;) To make sure that channel bandwidth power compensation really doesn't need to be taken care of, output a warning when the corresponding value stored in the EEPROM is non-zero. Also there is no apparent reason to refrain from initializing RFCSR register 13, it doesn't fail what-so-ever. Signed-off-by: Daniel Golle <daniel@makrotopia.org>
2017-02-13 05:25:35 +00:00
+ rt2800_register_read(rt2x00dev, TX_ALC_CFG_1, &reg);
+ rt2x00_set_field32(&reg, TX_ALC_CFG_1_ROS_BUSY_EN, 0);
+ rt2800_register_write(rt2x00dev, TX_ALC_CFG_1, reg);
} else {
rt2800_register_write(rt2x00dev, TX_SW_CFG0, 0x00000000);
rt2800_register_write(rt2x00dev, TX_SW_CFG1, 0x00080606);
@@ -5786,6 +6287,231 @@ static void rt2800_init_bbp_5592(struct rt2x00_dev *rt2x00dev)
rt2800_bbp_write(rt2x00dev, 103, 0xc0);
}
+static void rt2800_bbp_glrt_write(struct rt2x00_dev *rt2x00dev,
+ const u8 reg, const u8 value)
+{
+ rt2800_bbp_write(rt2x00dev, 195, reg);
+ rt2800_bbp_write(rt2x00dev, 196, value);
+}
+
+static void rt2800_bbp_dcoc_write(struct rt2x00_dev *rt2x00dev,
+ const u8 reg, const u8 value)
+{
+ rt2800_bbp_write(rt2x00dev, 158, reg);
+ rt2800_bbp_write(rt2x00dev, 159, value);
+}
+
+static void rt2800_bbp_dcoc_read(struct rt2x00_dev *rt2x00dev,
+ const u8 reg, u8 *value)
+{
+ rt2800_bbp_write(rt2x00dev, 158, reg);
+ rt2800_bbp_read(rt2x00dev, 159, value);
+}
+
+static void rt2800_init_bbp_6352(struct rt2x00_dev *rt2x00dev)
+{
+ u8 bbp;
+
+ /* Apply Maximum Likelihood Detection (MLD) for 2 stream case */
+ rt2800_bbp_read(rt2x00dev, 105, &bbp);
+ rt2x00_set_field8(&bbp, BBP105_MLD,
+ rt2x00dev->default_ant.rx_chain_num == 2);
+ rt2800_bbp_write(rt2x00dev, 105, bbp);
+
+ /* Avoid data loss and CRC errors */
+ rt2800_bbp4_mac_if_ctrl(rt2x00dev);
+
+ /* Fix I/Q swap issue */
+ rt2800_bbp_read(rt2x00dev, 1, &bbp);
+ bbp |= 0x04;
+ rt2800_bbp_write(rt2x00dev, 1, bbp);
+
+ /* BBP for G band */
+ rt2800_bbp_write(rt2x00dev, 3, 0x08);
+ rt2800_bbp_write(rt2x00dev, 4, 0x00); /* rt2800_bbp4_mac_if_ctrl? */
+ rt2800_bbp_write(rt2x00dev, 6, 0x08);
+ rt2800_bbp_write(rt2x00dev, 14, 0x09);
+ rt2800_bbp_write(rt2x00dev, 15, 0xFF);
+ rt2800_bbp_write(rt2x00dev, 16, 0x01);
+ rt2800_bbp_write(rt2x00dev, 20, 0x06);
+ rt2800_bbp_write(rt2x00dev, 21, 0x00);
+ rt2800_bbp_write(rt2x00dev, 22, 0x00);
+ rt2800_bbp_write(rt2x00dev, 27, 0x00);
+ rt2800_bbp_write(rt2x00dev, 28, 0x00);
+ rt2800_bbp_write(rt2x00dev, 30, 0x00);
+ rt2800_bbp_write(rt2x00dev, 31, 0x48);
+ rt2800_bbp_write(rt2x00dev, 47, 0x40);
+ rt2800_bbp_write(rt2x00dev, 62, 0x00);
+ rt2800_bbp_write(rt2x00dev, 63, 0x00);
+ rt2800_bbp_write(rt2x00dev, 64, 0x00);
+ rt2800_bbp_write(rt2x00dev, 65, 0x2C);
+ rt2800_bbp_write(rt2x00dev, 66, 0x1C);
+ rt2800_bbp_write(rt2x00dev, 67, 0x20);
+ rt2800_bbp_write(rt2x00dev, 68, 0xDD);
+ rt2800_bbp_write(rt2x00dev, 69, 0x10);
+ rt2800_bbp_write(rt2x00dev, 70, 0x05);
+ rt2800_bbp_write(rt2x00dev, 73, 0x18);
+ rt2800_bbp_write(rt2x00dev, 74, 0x0F);
+ rt2800_bbp_write(rt2x00dev, 75, 0x60);
+ rt2800_bbp_write(rt2x00dev, 76, 0x44);
+ rt2800_bbp_write(rt2x00dev, 77, 0x59);
+ rt2800_bbp_write(rt2x00dev, 78, 0x1E);
+ rt2800_bbp_write(rt2x00dev, 79, 0x1C);
+ rt2800_bbp_write(rt2x00dev, 80, 0x0C);
+ rt2800_bbp_write(rt2x00dev, 81, 0x3A);
+ rt2800_bbp_write(rt2x00dev, 82, 0xB6);
+ rt2800_bbp_write(rt2x00dev, 83, 0x9A);
+ rt2800_bbp_write(rt2x00dev, 84, 0x9A);
+ rt2800_bbp_write(rt2x00dev, 86, 0x38);
+ rt2800_bbp_write(rt2x00dev, 88, 0x90);
+ rt2800_bbp_write(rt2x00dev, 91, 0x04);
+ rt2800_bbp_write(rt2x00dev, 92, 0x02);
+ rt2800_bbp_write(rt2x00dev, 95, 0x9A);
+ rt2800_bbp_write(rt2x00dev, 96, 0x00);
+ rt2800_bbp_write(rt2x00dev, 103, 0xC0);
+ rt2800_bbp_write(rt2x00dev, 104, 0x92);
+ /* FIXME BBP105 owerwrite */
+ rt2800_bbp_write(rt2x00dev, 105, 0x3C);
+ rt2800_bbp_write(rt2x00dev, 106, 0x12);
+ rt2800_bbp_write(rt2x00dev, 109, 0x00);
+ rt2800_bbp_write(rt2x00dev, 134, 0x10);
+ rt2800_bbp_write(rt2x00dev, 135, 0xA6);
+ rt2800_bbp_write(rt2x00dev, 137, 0x04);
+ rt2800_bbp_write(rt2x00dev, 142, 0x30);
+ rt2800_bbp_write(rt2x00dev, 143, 0xF7);
+ rt2800_bbp_write(rt2x00dev, 160, 0xEC);
+ rt2800_bbp_write(rt2x00dev, 161, 0xC4);
+ rt2800_bbp_write(rt2x00dev, 162, 0x77);
+ rt2800_bbp_write(rt2x00dev, 163, 0xF9);
+ rt2800_bbp_write(rt2x00dev, 164, 0x00);
+ rt2800_bbp_write(rt2x00dev, 165, 0x00);
+ rt2800_bbp_write(rt2x00dev, 186, 0x00);
+ rt2800_bbp_write(rt2x00dev, 187, 0x00);
+ rt2800_bbp_write(rt2x00dev, 188, 0x00);
+ rt2800_bbp_write(rt2x00dev, 186, 0x00);
+ rt2800_bbp_write(rt2x00dev, 187, 0x01);
+ rt2800_bbp_write(rt2x00dev, 188, 0x00);
+ rt2800_bbp_write(rt2x00dev, 189, 0x00);
+
+ rt2800_bbp_write(rt2x00dev, 91, 0x06);
+ rt2800_bbp_write(rt2x00dev, 92, 0x04);
+ rt2800_bbp_write(rt2x00dev, 93, 0x54);
+ rt2800_bbp_write(rt2x00dev, 99, 0x50);
+ rt2800_bbp_write(rt2x00dev, 148, 0x84);
+ rt2800_bbp_write(rt2x00dev, 167, 0x80);
+ rt2800_bbp_write(rt2x00dev, 178, 0xFF);
+ rt2800_bbp_write(rt2x00dev, 106, 0x13);
+
+ /* BBP for G band GLRT function (BBP_128 ~ BBP_221) */
+ rt2800_bbp_glrt_write(rt2x00dev, 0, 0x00);
rt2x00: mt7620: yet another beauty session So here is another round of improvements for MT7620 WiFi. This commit fixes a few significant issues related to TX_PWR_CFG_x and TX_ALC and also makes the code more readable by adding register descriptions for things added for MT7620 and use the usual bit-field access macros and the now defined macros instead of plain bit-ops and magic numbers. Properly describe EEPROM_TARGET_POWER at word 0x68 (== byte 0xD0) and thereby fix internal TXALC which would otherwise just read out-of-bounds of the EEPROM map. Split-out tx-power/ALC related stuff into an additional function. Fix VCO calibration, it was carried out properly in the channel switching but incomplete in the actual VCO calibration function. Also there is no need to trigger VCO calibration in channel switching, the VCO calibration function is already being called at this point. Remove it from channel switching function to avoid redundant code. The TX power calibration differs significantly from all other Mediatek/Ralink chips: They finally allow 0.5dB steps stored as 8-bit values for (almost) each bitrate -- and promptly ran out of space and for some reason didn't want to change the EEPROM layout. The hence opted for a scheme of sharing values for some adjecent bitrates and a highly over-complicated (or obfuscated?) way to populate the TX_PWR_CFG_x registers with the values stored in the EEPROM. The code here now looks much less complicated than what you see in the vendor's driver, however, it does the exact same thing: bGpwrdeltaMinus is a constant and always TRUE, hence half of the code was dead. Gpwrdelta is always 0 (rather than using the value read from the EEPROM). What remains is some very grotesque effort to avoid 0x20, probably some hardware bug related to some misunderstanding of what a singed 8-bit value is (imagine: if it was a signed 6-bit value then someone could believe that 0x20 == 0x0). And then they didn't clean it up once they later on anandonned that whole story of having a constant offset for 40 MHz channels and just set the offset to be constant 0 -- there is no effort for avoiding 0x20 for the 20 MHz values stored in the EEPROM, hence that's probably just a forbidden value in the EEPROM specs and won't appear anyway... Anyway, the whole thing felt like solving some college math test where in the end everything cancels out and the result equals 0 ;) To make sure that channel bandwidth power compensation really doesn't need to be taken care of, output a warning when the corresponding value stored in the EEPROM is non-zero. Also there is no apparent reason to refrain from initializing RFCSR register 13, it doesn't fail what-so-ever. Signed-off-by: Daniel Golle <daniel@makrotopia.org>
2017-02-13 05:25:35 +00:00
+ rt2800_bbp_glrt_write(rt2x00dev, 1, 0x14);
+ rt2800_bbp_glrt_write(rt2x00dev, 2, 0x20);
+ rt2800_bbp_glrt_write(rt2x00dev, 3, 0x0A);
+ rt2800_bbp_glrt_write(rt2x00dev, 10, 0x16);
+ rt2800_bbp_glrt_write(rt2x00dev, 11, 0x06);
+ rt2800_bbp_glrt_write(rt2x00dev, 12, 0x02);
+ rt2800_bbp_glrt_write(rt2x00dev, 13, 0x07);
+ rt2800_bbp_glrt_write(rt2x00dev, 14, 0x05);
+ rt2800_bbp_glrt_write(rt2x00dev, 15, 0x09);
+ rt2800_bbp_glrt_write(rt2x00dev, 16, 0x20);
+ rt2800_bbp_glrt_write(rt2x00dev, 17, 0x08);
+ rt2800_bbp_glrt_write(rt2x00dev, 18, 0x4A);
+ rt2800_bbp_glrt_write(rt2x00dev, 19, 0x00);
+ rt2800_bbp_glrt_write(rt2x00dev, 20, 0x00);
+ rt2800_bbp_glrt_write(rt2x00dev, 128, 0xE0);
+ rt2800_bbp_glrt_write(rt2x00dev, 129, 0x1F);
+ rt2800_bbp_glrt_write(rt2x00dev, 130, 0x4F);
+ rt2800_bbp_glrt_write(rt2x00dev, 131, 0x32);
+ rt2800_bbp_glrt_write(rt2x00dev, 132, 0x08);
+ rt2800_bbp_glrt_write(rt2x00dev, 133, 0x28);
+ rt2800_bbp_glrt_write(rt2x00dev, 134, 0x19);
+ rt2800_bbp_glrt_write(rt2x00dev, 135, 0x0A);
+ rt2800_bbp_glrt_write(rt2x00dev, 138, 0x16);
+ rt2800_bbp_glrt_write(rt2x00dev, 139, 0x10);
+ rt2800_bbp_glrt_write(rt2x00dev, 140, 0x10);
+ rt2800_bbp_glrt_write(rt2x00dev, 141, 0x1A);
+ rt2800_bbp_glrt_write(rt2x00dev, 142, 0x36);
+ rt2800_bbp_glrt_write(rt2x00dev, 143, 0x2C);
+ rt2800_bbp_glrt_write(rt2x00dev, 144, 0x26);
+ rt2800_bbp_glrt_write(rt2x00dev, 145, 0x24);
+ rt2800_bbp_glrt_write(rt2x00dev, 146, 0x42);
+ rt2800_bbp_glrt_write(rt2x00dev, 147, 0x40);
+ rt2800_bbp_glrt_write(rt2x00dev, 148, 0x30);
+ rt2800_bbp_glrt_write(rt2x00dev, 149, 0x29);
+ rt2800_bbp_glrt_write(rt2x00dev, 150, 0x4C);
+ rt2800_bbp_glrt_write(rt2x00dev, 151, 0x46);
+ rt2800_bbp_glrt_write(rt2x00dev, 152, 0x3D);
+ rt2800_bbp_glrt_write(rt2x00dev, 153, 0x40);
+ rt2800_bbp_glrt_write(rt2x00dev, 154, 0x3E);
+ rt2800_bbp_glrt_write(rt2x00dev, 155, 0x38);
+ rt2800_bbp_glrt_write(rt2x00dev, 156, 0x3D);
+ rt2800_bbp_glrt_write(rt2x00dev, 157, 0x2F);
+ rt2800_bbp_glrt_write(rt2x00dev, 158, 0x3C);
+ rt2800_bbp_glrt_write(rt2x00dev, 159, 0x34);
+ rt2800_bbp_glrt_write(rt2x00dev, 160, 0x2C);
+ rt2800_bbp_glrt_write(rt2x00dev, 161, 0x2F);
+ rt2800_bbp_glrt_write(rt2x00dev, 162, 0x3C);
+ rt2800_bbp_glrt_write(rt2x00dev, 163, 0x35);
+ rt2800_bbp_glrt_write(rt2x00dev, 164, 0x2E);
+ rt2800_bbp_glrt_write(rt2x00dev, 165, 0x2F);
+ rt2800_bbp_glrt_write(rt2x00dev, 166, 0x49);
+ rt2800_bbp_glrt_write(rt2x00dev, 167, 0x41);
+ rt2800_bbp_glrt_write(rt2x00dev, 168, 0x36);
+ rt2800_bbp_glrt_write(rt2x00dev, 169, 0x39);
+ rt2800_bbp_glrt_write(rt2x00dev, 170, 0x30);
+ rt2800_bbp_glrt_write(rt2x00dev, 171, 0x30);
+ rt2800_bbp_glrt_write(rt2x00dev, 172, 0x0E);
+ rt2800_bbp_glrt_write(rt2x00dev, 173, 0x0D);
+ rt2800_bbp_glrt_write(rt2x00dev, 174, 0x28);
+ rt2800_bbp_glrt_write(rt2x00dev, 175, 0x21);
+ rt2800_bbp_glrt_write(rt2x00dev, 176, 0x1C);
+ rt2800_bbp_glrt_write(rt2x00dev, 177, 0x16);
+ rt2800_bbp_glrt_write(rt2x00dev, 178, 0x50);
+ rt2800_bbp_glrt_write(rt2x00dev, 179, 0x4A);
+ rt2800_bbp_glrt_write(rt2x00dev, 180, 0x43);
+ rt2800_bbp_glrt_write(rt2x00dev, 181, 0x50);
+ rt2800_bbp_glrt_write(rt2x00dev, 182, 0x10);
+ rt2800_bbp_glrt_write(rt2x00dev, 183, 0x10);
+ rt2800_bbp_glrt_write(rt2x00dev, 184, 0x10);
+ rt2800_bbp_glrt_write(rt2x00dev, 185, 0x10);
+ rt2800_bbp_glrt_write(rt2x00dev, 200, 0x7D);
+ rt2800_bbp_glrt_write(rt2x00dev, 201, 0x14);
+ rt2800_bbp_glrt_write(rt2x00dev, 202, 0x32);
+ rt2800_bbp_glrt_write(rt2x00dev, 203, 0x2C);
+ rt2800_bbp_glrt_write(rt2x00dev, 204, 0x36);
+ rt2800_bbp_glrt_write(rt2x00dev, 205, 0x4C);
+ rt2800_bbp_glrt_write(rt2x00dev, 206, 0x43);
+ rt2800_bbp_glrt_write(rt2x00dev, 207, 0x2C);
+ rt2800_bbp_glrt_write(rt2x00dev, 208, 0x2E);
+ rt2800_bbp_glrt_write(rt2x00dev, 209, 0x36);
+ rt2800_bbp_glrt_write(rt2x00dev, 210, 0x30);
+ rt2800_bbp_glrt_write(rt2x00dev, 211, 0x6E);
+
+ /* BBP for G band DCOC function */
+ rt2800_bbp_dcoc_write(rt2x00dev, 140, 0x0C);
+ rt2800_bbp_dcoc_write(rt2x00dev, 141, 0x00);
+ rt2800_bbp_dcoc_write(rt2x00dev, 142, 0x10);
+ rt2800_bbp_dcoc_write(rt2x00dev, 143, 0x10);
+ rt2800_bbp_dcoc_write(rt2x00dev, 144, 0x10);
+ rt2800_bbp_dcoc_write(rt2x00dev, 145, 0x10);
+ rt2800_bbp_dcoc_write(rt2x00dev, 146, 0x08);
+ rt2800_bbp_dcoc_write(rt2x00dev, 147, 0x40);
+ rt2800_bbp_dcoc_write(rt2x00dev, 148, 0x04);
+ rt2800_bbp_dcoc_write(rt2x00dev, 149, 0x04);
+ rt2800_bbp_dcoc_write(rt2x00dev, 150, 0x08);
+ rt2800_bbp_dcoc_write(rt2x00dev, 151, 0x08);
+ rt2800_bbp_dcoc_write(rt2x00dev, 152, 0x03);
+ rt2800_bbp_dcoc_write(rt2x00dev, 153, 0x03);
+ rt2800_bbp_dcoc_write(rt2x00dev, 154, 0x03);
+ rt2800_bbp_dcoc_write(rt2x00dev, 155, 0x02);
+ rt2800_bbp_dcoc_write(rt2x00dev, 156, 0x40);
+ rt2800_bbp_dcoc_write(rt2x00dev, 157, 0x40);
+ rt2800_bbp_dcoc_write(rt2x00dev, 158, 0x64);
+ rt2800_bbp_dcoc_write(rt2x00dev, 159, 0x64);
+
+ rt2800_bbp4_mac_if_ctrl(rt2x00dev);
+}
+
static void rt2800_init_bbp(struct rt2x00_dev *rt2x00dev)
{
unsigned int i;
@@ -5830,6 +6556,9 @@ static void rt2800_init_bbp(struct rt2x00_dev *rt2x00dev)
case RT5592:
rt2800_init_bbp_5592(rt2x00dev);
return;
+ case RT6352:
+ rt2800_init_bbp_6352(rt2x00dev);
+ break;
}
for (i = 0; i < EEPROM_BBP_SIZE; i++) {
@@ -6901,6 +7630,615 @@ static void rt2800_init_rfcsr_5592(struct rt2x00_dev *rt2x00dev)
rt2800_led_open_drain_enable(rt2x00dev);
}
+static void rt2800_bbp_core_soft_reset(struct rt2x00_dev *rt2x00dev,
+ bool set_bw, bool is_ht40)
+{
+ u8 bbp_val;
+
+ rt2800_bbp_read(rt2x00dev, 21, &bbp_val);
+ bbp_val |= 0x1;
+ rt2800_bbp_write(rt2x00dev, 21, bbp_val);
+ usleep_range(100, 200);
+
+ if (set_bw) {
+ rt2800_bbp_read(rt2x00dev, 4, &bbp_val);
+ rt2x00_set_field8(&bbp_val, BBP4_BANDWIDTH, 2 * is_ht40);
+ rt2800_bbp_write(rt2x00dev, 4, bbp_val);
+ usleep_range(100, 200);
+ }
+
+ rt2800_bbp_read(rt2x00dev, 21, &bbp_val);
+ bbp_val &= (~0x1);
+ rt2800_bbp_write(rt2x00dev, 21, bbp_val);
+ usleep_range(100, 200);
+}
+
+static int rt2800_rf_lp_config(struct rt2x00_dev *rt2x00dev, bool btxcal)
+{
+ u8 rf_val;
+
+ if (btxcal)
+ rt2800_register_write(rt2x00dev, RF_CONTROL0, 0x04);
+ else
+ rt2800_register_write(rt2x00dev, RF_CONTROL0, 0x02);
+
+ rt2800_register_write(rt2x00dev, RF_BYPASS0, 0x06);
+
+ rt2800_rfcsr_read_bank(rt2x00dev, 5, 17, &rf_val);
+ rf_val |= 0x80;
+ rt2800_rfcsr_write_bank(rt2x00dev, 5, 17, rf_val);
+
+ if (btxcal) {
+ rt2800_rfcsr_write_bank(rt2x00dev, 5, 18, 0xC1);
+ rt2800_rfcsr_write_bank(rt2x00dev, 5, 19, 0x20);
+ rt2800_rfcsr_write_bank(rt2x00dev, 5, 20, 0x02);
+ rt2800_rfcsr_read_bank(rt2x00dev, 5, 3, &rf_val);
+ rf_val &= (~0x3F);
+ rf_val |= 0x3F;
+ rt2800_rfcsr_write_bank(rt2x00dev, 5, 3, rf_val);
+ rt2800_rfcsr_read_bank(rt2x00dev, 5, 4, &rf_val);
+ rf_val &= (~0x3F);
+ rf_val |= 0x3F;
+ rt2800_rfcsr_write_bank(rt2x00dev, 5, 4, rf_val);
+ rt2800_rfcsr_write_bank(rt2x00dev, 5, 5, 0x31);
+ } else {
+ rt2800_rfcsr_write_bank(rt2x00dev, 5, 18, 0xF1);
+ rt2800_rfcsr_write_bank(rt2x00dev, 5, 19, 0x18);
+ rt2800_rfcsr_write_bank(rt2x00dev, 5, 20, 0x02);
+ rt2800_rfcsr_read_bank(rt2x00dev, 5, 3, &rf_val);
+ rf_val &= (~0x3F);
+ rf_val |= 0x34;
+ rt2800_rfcsr_write_bank(rt2x00dev, 5, 3, rf_val);
+ rt2800_rfcsr_read_bank(rt2x00dev, 5, 4, &rf_val);
+ rf_val &= (~0x3F);
+ rf_val |= 0x34;
+ rt2800_rfcsr_write_bank(rt2x00dev, 5, 4, rf_val);
+ }
+
+ return 0;
+}
+
+static char rt2800_lp_tx_filter_bw_cal(struct rt2x00_dev *rt2x00dev)
+{
+ unsigned int cnt;
+ u8 bbp_val;
+ char cal_val;
+
+ rt2800_bbp_dcoc_write(rt2x00dev, 0, 0x82);
+
+ cnt = 0;
+ do {
+ usleep_range(500, 2000);
+ rt2800_bbp_read(rt2x00dev, 159, &bbp_val);
+ if (bbp_val == 0x02 || cnt == 20)
+ break;
+
+ cnt++;
+ } while (cnt < 20);
+
+ rt2800_bbp_dcoc_read(rt2x00dev, 0x39, &bbp_val);
+ cal_val = bbp_val & 0x7F;
+ if (cal_val >= 0x40)
+ cal_val -= 128;
+
+ return cal_val;
+}
+
+static void rt2800_bw_filter_calibration(struct rt2x00_dev *rt2x00dev,
+ bool btxcal)
+{
+ struct rt2800_drv_data *drv_data = rt2x00dev->drv_data;
+ u8 tx_agc_fc = 0, rx_agc_fc = 0, cmm_agc_fc;
+ u8 filter_target;
+ u8 tx_filter_target_20m = 0x09, tx_filter_target_40m = 0x02;
+ u8 rx_filter_target_20m = 0x27, rx_filter_target_40m = 0x31;
+ int loop = 0, is_ht40, cnt;
+ u8 bbp_val, rf_val;
+ char cal_r32_init, cal_r32_val, cal_diff;
+ u8 saverfb5r00, saverfb5r01, saverfb5r03, saverfb5r04, saverfb5r05;
+ u8 saverfb5r06, saverfb5r07;
+ u8 saverfb5r08, saverfb5r17, saverfb5r18, saverfb5r19, saverfb5r20;
+ u8 saverfb5r37, saverfb5r38, saverfb5r39, saverfb5r40, saverfb5r41;
+ u8 saverfb5r42, saverfb5r43, saverfb5r44, saverfb5r45, saverfb5r46;
+ u8 saverfb5r58, saverfb5r59;
+ u8 savebbp159r0, savebbp159r2, savebbpr23;
+ u32 MAC_RF_CONTROL0, MAC_RF_BYPASS0;
+
+ /* Save MAC registers */
+ rt2800_register_read(rt2x00dev, RF_CONTROL0, &MAC_RF_CONTROL0);
+ rt2800_register_read(rt2x00dev, RF_BYPASS0, &MAC_RF_BYPASS0);
+
+ /* save BBP registers */
+ rt2800_bbp_read(rt2x00dev, 23, &savebbpr23);
+
+ rt2800_bbp_dcoc_read(rt2x00dev, 0, &savebbp159r0);
+ rt2800_bbp_dcoc_read(rt2x00dev, 2, &savebbp159r2);
+
+ /* Save RF registers */
+ rt2800_rfcsr_read_bank(rt2x00dev, 5, 0, &saverfb5r00);
+ rt2800_rfcsr_read_bank(rt2x00dev, 5, 1, &saverfb5r01);
+ rt2800_rfcsr_read_bank(rt2x00dev, 5, 3, &saverfb5r03);
+ rt2800_rfcsr_read_bank(rt2x00dev, 5, 4, &saverfb5r04);
+ rt2800_rfcsr_read_bank(rt2x00dev, 5, 5, &saverfb5r05);
+ rt2800_rfcsr_read_bank(rt2x00dev, 5, 6, &saverfb5r06);
+ rt2800_rfcsr_read_bank(rt2x00dev, 5, 7, &saverfb5r07);
+ rt2800_rfcsr_read_bank(rt2x00dev, 5, 8, &saverfb5r08);
+ rt2800_rfcsr_read_bank(rt2x00dev, 5, 17, &saverfb5r17);
+ rt2800_rfcsr_read_bank(rt2x00dev, 5, 18, &saverfb5r18);
+ rt2800_rfcsr_read_bank(rt2x00dev, 5, 19, &saverfb5r19);
+ rt2800_rfcsr_read_bank(rt2x00dev, 5, 20, &saverfb5r20);
+
+ rt2800_rfcsr_read_bank(rt2x00dev, 5, 37, &saverfb5r37);
+ rt2800_rfcsr_read_bank(rt2x00dev, 5, 38, &saverfb5r38);
+ rt2800_rfcsr_read_bank(rt2x00dev, 5, 39, &saverfb5r39);
+ rt2800_rfcsr_read_bank(rt2x00dev, 5, 40, &saverfb5r40);
+ rt2800_rfcsr_read_bank(rt2x00dev, 5, 41, &saverfb5r41);
+ rt2800_rfcsr_read_bank(rt2x00dev, 5, 42, &saverfb5r42);
+ rt2800_rfcsr_read_bank(rt2x00dev, 5, 43, &saverfb5r43);
+ rt2800_rfcsr_read_bank(rt2x00dev, 5, 44, &saverfb5r44);
+ rt2800_rfcsr_read_bank(rt2x00dev, 5, 45, &saverfb5r45);
+ rt2800_rfcsr_read_bank(rt2x00dev, 5, 46, &saverfb5r46);
+
+ rt2800_rfcsr_read_bank(rt2x00dev, 5, 58, &saverfb5r58);
+ rt2800_rfcsr_read_bank(rt2x00dev, 5, 59, &saverfb5r59);
+
+ rt2800_rfcsr_read_bank(rt2x00dev, 5, 0, &rf_val);
+ rf_val |= 0x3;
+ rt2800_rfcsr_write_bank(rt2x00dev, 5, 0, rf_val);
+
+ rt2800_rfcsr_read_bank(rt2x00dev, 5, 1, &rf_val);
+ rf_val |= 0x1;
+ rt2800_rfcsr_write_bank(rt2x00dev, 5, 1, rf_val);
+
+ cnt = 0;
+ do {
+ usleep_range(500, 2000);
+ rt2800_rfcsr_read_bank(rt2x00dev, 5, 1, &rf_val);
+ if (((rf_val & 0x1) == 0x00) || (cnt == 40))
+ break;
+ cnt++;
+ } while (cnt < 40);
+
+ rt2800_rfcsr_read_bank(rt2x00dev, 5, 0, &rf_val);
+ rf_val &= (~0x3);
+ rf_val |= 0x1;
+ rt2800_rfcsr_write_bank(rt2x00dev, 5, 0, rf_val);
+
+ /* I-3 */
+ rt2800_bbp_read(rt2x00dev, 23, &bbp_val);
+ bbp_val &= (~0x1F);
+ bbp_val |= 0x10;
+ rt2800_bbp_write(rt2x00dev, 23, bbp_val);
+
+ do {
+ /* I-4,5,6,7,8,9 */
+ if (loop == 0) {
+ is_ht40 = false;
+
+ if (btxcal)
+ filter_target = tx_filter_target_20m;
+ else
+ filter_target = rx_filter_target_20m;
+ } else {
+ is_ht40 = true;
+
+ if (btxcal)
+ filter_target = tx_filter_target_40m;
+ else
+ filter_target = rx_filter_target_40m;
+ }
+
+ rt2800_rfcsr_read_bank(rt2x00dev, 5, 8, &rf_val);
+ rf_val &= (~0x04);
+ if (loop == 1)
+ rf_val |= 0x4;
+
+ rt2800_rfcsr_write_bank(rt2x00dev, 5, 8, rf_val);
+
+ rt2800_bbp_core_soft_reset(rt2x00dev, true, is_ht40);
+
+ rt2800_rf_lp_config(rt2x00dev, btxcal);
+ if (btxcal) {
+ tx_agc_fc = 0;
+ rt2800_rfcsr_read_bank(rt2x00dev, 5, 58, &rf_val);
+ rf_val &= (~0x7F);
+ rt2800_rfcsr_write_bank(rt2x00dev, 5, 58, rf_val);
+ rt2800_rfcsr_read_bank(rt2x00dev, 5, 59, &rf_val);
+ rf_val &= (~0x7F);
+ rt2800_rfcsr_write_bank(rt2x00dev, 5, 59, rf_val);
+ } else {
+ rx_agc_fc = 0;
+ rt2800_rfcsr_read_bank(rt2x00dev, 5, 6, &rf_val);
+ rf_val &= (~0x7F);
+ rt2800_rfcsr_write_bank(rt2x00dev, 5, 6, rf_val);
+ rt2800_rfcsr_read_bank(rt2x00dev, 5, 7, &rf_val);
+ rf_val &= (~0x7F);
+ rt2800_rfcsr_write_bank(rt2x00dev, 5, 7, rf_val);
+ }
+
+ usleep_range(1000, 2000);
+
+ rt2800_bbp_dcoc_read(rt2x00dev, 2, &bbp_val);
+ bbp_val &= (~0x6);
+ rt2800_bbp_dcoc_write(rt2x00dev, 2, bbp_val);
+
+ rt2800_bbp_core_soft_reset(rt2x00dev, false, is_ht40);
+
+ cal_r32_init = rt2800_lp_tx_filter_bw_cal(rt2x00dev);
+
+ rt2800_bbp_dcoc_read(rt2x00dev, 2, &bbp_val);
+ bbp_val |= 0x6;
+ rt2800_bbp_dcoc_write(rt2x00dev, 2, bbp_val);
+do_cal:
+ if (btxcal) {
+ rt2800_rfcsr_read_bank(rt2x00dev, 5, 58, &rf_val);
+ rf_val &= (~0x7F);
+ rf_val |= tx_agc_fc;
+ rt2800_rfcsr_write_bank(rt2x00dev, 5, 58, rf_val);
+ rt2800_rfcsr_read_bank(rt2x00dev, 5, 59, &rf_val);
+ rf_val &= (~0x7F);
+ rf_val |= tx_agc_fc;
+ rt2800_rfcsr_write_bank(rt2x00dev, 5, 59, rf_val);
+ } else {
+ rt2800_rfcsr_read_bank(rt2x00dev, 5, 6, &rf_val);
+ rf_val &= (~0x7F);
+ rf_val |= rx_agc_fc;
+ rt2800_rfcsr_write_bank(rt2x00dev, 5, 6, rf_val);
+ rt2800_rfcsr_read_bank(rt2x00dev, 5, 7, &rf_val);
+ rf_val &= (~0x7F);
+ rf_val |= rx_agc_fc;
+ rt2800_rfcsr_write_bank(rt2x00dev, 5, 7, rf_val);
+ }
+
+ usleep_range(500, 1000);
+
+ rt2800_bbp_core_soft_reset(rt2x00dev, false, is_ht40);
+
+ cal_r32_val = rt2800_lp_tx_filter_bw_cal(rt2x00dev);
+
+ cal_diff = cal_r32_init - cal_r32_val;
+
+ if (btxcal)
+ cmm_agc_fc = tx_agc_fc;
+ else
+ cmm_agc_fc = rx_agc_fc;
+
+ if (((cal_diff > filter_target) && (cmm_agc_fc == 0)) ||
+ ((cal_diff < filter_target) && (cmm_agc_fc == 0x3f))) {
+ if (btxcal)
+ tx_agc_fc = 0;
+ else
+ rx_agc_fc = 0;
+ } else if ((cal_diff <= filter_target) && (cmm_agc_fc < 0x3f)) {
+ if (btxcal)
+ tx_agc_fc++;
+ else
+ rx_agc_fc++;
+ goto do_cal;
+ }
+
+ if (btxcal) {
+ if (loop == 0)
+ drv_data->tx_calibration_bw20 = tx_agc_fc;
+ else
+ drv_data->tx_calibration_bw40 = tx_agc_fc;
+ } else {
+ if (loop == 0)
+ drv_data->rx_calibration_bw20 = rx_agc_fc;
+ else
+ drv_data->rx_calibration_bw40 = rx_agc_fc;
+ }
+
+ loop++;
+ } while (loop <= 1);
+
+ rt2800_rfcsr_write_bank(rt2x00dev, 5, 0, saverfb5r00);
+ rt2800_rfcsr_write_bank(rt2x00dev, 5, 1, saverfb5r01);
+ rt2800_rfcsr_write_bank(rt2x00dev, 5, 3, saverfb5r03);
+ rt2800_rfcsr_write_bank(rt2x00dev, 5, 4, saverfb5r04);
+ rt2800_rfcsr_write_bank(rt2x00dev, 5, 5, saverfb5r05);
+ rt2800_rfcsr_write_bank(rt2x00dev, 5, 6, saverfb5r06);
+ rt2800_rfcsr_write_bank(rt2x00dev, 5, 7, saverfb5r07);
+ rt2800_rfcsr_write_bank(rt2x00dev, 5, 8, saverfb5r08);
+ rt2800_rfcsr_write_bank(rt2x00dev, 5, 17, saverfb5r17);
+ rt2800_rfcsr_write_bank(rt2x00dev, 5, 18, saverfb5r18);
+ rt2800_rfcsr_write_bank(rt2x00dev, 5, 19, saverfb5r19);
+ rt2800_rfcsr_write_bank(rt2x00dev, 5, 20, saverfb5r20);
+
+ rt2800_rfcsr_write_bank(rt2x00dev, 5, 37, saverfb5r37);
+ rt2800_rfcsr_write_bank(rt2x00dev, 5, 38, saverfb5r38);
+ rt2800_rfcsr_write_bank(rt2x00dev, 5, 39, saverfb5r39);
+ rt2800_rfcsr_write_bank(rt2x00dev, 5, 40, saverfb5r40);
+ rt2800_rfcsr_write_bank(rt2x00dev, 5, 41, saverfb5r41);
+ rt2800_rfcsr_write_bank(rt2x00dev, 5, 42, saverfb5r42);
+ rt2800_rfcsr_write_bank(rt2x00dev, 5, 43, saverfb5r43);
+ rt2800_rfcsr_write_bank(rt2x00dev, 5, 44, saverfb5r44);
+ rt2800_rfcsr_write_bank(rt2x00dev, 5, 45, saverfb5r45);
+ rt2800_rfcsr_write_bank(rt2x00dev, 5, 46, saverfb5r46);
+
+ rt2800_rfcsr_write_bank(rt2x00dev, 5, 58, saverfb5r58);
+ rt2800_rfcsr_write_bank(rt2x00dev, 5, 59, saverfb5r59);
+
+ rt2800_bbp_write(rt2x00dev, 23, savebbpr23);
+
+ rt2800_bbp_dcoc_write(rt2x00dev, 0, savebbp159r0);
+ rt2800_bbp_dcoc_write(rt2x00dev, 2, savebbp159r2);
+
+ rt2800_bbp_read(rt2x00dev, 4, &bbp_val);
+ rt2x00_set_field8(&bbp_val, BBP4_BANDWIDTH,
+ 2 * test_bit(CONFIG_CHANNEL_HT40, &rt2x00dev->flags));
+ rt2800_bbp_write(rt2x00dev, 4, bbp_val);
+
+ rt2800_register_write(rt2x00dev, RF_CONTROL0, MAC_RF_CONTROL0);
+ rt2800_register_write(rt2x00dev, RF_BYPASS0, MAC_RF_BYPASS0);
+}
+
+static void rt2800_init_rfcsr_6352(struct rt2x00_dev *rt2x00dev)
+{
+ /* Initialize RF central register to default value */
+ rt2800_rfcsr_write(rt2x00dev, 0, 0x02);
+ rt2800_rfcsr_write(rt2x00dev, 1, 0x03);
+ rt2800_rfcsr_write(rt2x00dev, 2, 0x33);
+ rt2800_rfcsr_write(rt2x00dev, 3, 0xFF);
+ rt2800_rfcsr_write(rt2x00dev, 4, 0x0C);
rt2x00: mt7620: yet another beauty session So here is another round of improvements for MT7620 WiFi. This commit fixes a few significant issues related to TX_PWR_CFG_x and TX_ALC and also makes the code more readable by adding register descriptions for things added for MT7620 and use the usual bit-field access macros and the now defined macros instead of plain bit-ops and magic numbers. Properly describe EEPROM_TARGET_POWER at word 0x68 (== byte 0xD0) and thereby fix internal TXALC which would otherwise just read out-of-bounds of the EEPROM map. Split-out tx-power/ALC related stuff into an additional function. Fix VCO calibration, it was carried out properly in the channel switching but incomplete in the actual VCO calibration function. Also there is no need to trigger VCO calibration in channel switching, the VCO calibration function is already being called at this point. Remove it from channel switching function to avoid redundant code. The TX power calibration differs significantly from all other Mediatek/Ralink chips: They finally allow 0.5dB steps stored as 8-bit values for (almost) each bitrate -- and promptly ran out of space and for some reason didn't want to change the EEPROM layout. The hence opted for a scheme of sharing values for some adjecent bitrates and a highly over-complicated (or obfuscated?) way to populate the TX_PWR_CFG_x registers with the values stored in the EEPROM. The code here now looks much less complicated than what you see in the vendor's driver, however, it does the exact same thing: bGpwrdeltaMinus is a constant and always TRUE, hence half of the code was dead. Gpwrdelta is always 0 (rather than using the value read from the EEPROM). What remains is some very grotesque effort to avoid 0x20, probably some hardware bug related to some misunderstanding of what a singed 8-bit value is (imagine: if it was a signed 6-bit value then someone could believe that 0x20 == 0x0). And then they didn't clean it up once they later on anandonned that whole story of having a constant offset for 40 MHz channels and just set the offset to be constant 0 -- there is no effort for avoiding 0x20 for the 20 MHz values stored in the EEPROM, hence that's probably just a forbidden value in the EEPROM specs and won't appear anyway... Anyway, the whole thing felt like solving some college math test where in the end everything cancels out and the result equals 0 ;) To make sure that channel bandwidth power compensation really doesn't need to be taken care of, output a warning when the corresponding value stored in the EEPROM is non-zero. Also there is no apparent reason to refrain from initializing RFCSR register 13, it doesn't fail what-so-ever. Signed-off-by: Daniel Golle <daniel@makrotopia.org>
2017-02-13 05:25:35 +00:00
+ rt2800_rfcsr_write(rt2x00dev, 5, 0x40);
+ rt2800_rfcsr_write(rt2x00dev, 6, 0x00);
+ rt2800_rfcsr_write(rt2x00dev, 7, 0x00);
+ rt2800_rfcsr_write(rt2x00dev, 8, 0x00);
+ rt2800_rfcsr_write(rt2x00dev, 9, 0x00);
+ rt2800_rfcsr_write(rt2x00dev, 10, 0x00);
+ rt2800_rfcsr_write(rt2x00dev, 11, 0x00);
rt2x00: mt7620: yet another beauty session So here is another round of improvements for MT7620 WiFi. This commit fixes a few significant issues related to TX_PWR_CFG_x and TX_ALC and also makes the code more readable by adding register descriptions for things added for MT7620 and use the usual bit-field access macros and the now defined macros instead of plain bit-ops and magic numbers. Properly describe EEPROM_TARGET_POWER at word 0x68 (== byte 0xD0) and thereby fix internal TXALC which would otherwise just read out-of-bounds of the EEPROM map. Split-out tx-power/ALC related stuff into an additional function. Fix VCO calibration, it was carried out properly in the channel switching but incomplete in the actual VCO calibration function. Also there is no need to trigger VCO calibration in channel switching, the VCO calibration function is already being called at this point. Remove it from channel switching function to avoid redundant code. The TX power calibration differs significantly from all other Mediatek/Ralink chips: They finally allow 0.5dB steps stored as 8-bit values for (almost) each bitrate -- and promptly ran out of space and for some reason didn't want to change the EEPROM layout. The hence opted for a scheme of sharing values for some adjecent bitrates and a highly over-complicated (or obfuscated?) way to populate the TX_PWR_CFG_x registers with the values stored in the EEPROM. The code here now looks much less complicated than what you see in the vendor's driver, however, it does the exact same thing: bGpwrdeltaMinus is a constant and always TRUE, hence half of the code was dead. Gpwrdelta is always 0 (rather than using the value read from the EEPROM). What remains is some very grotesque effort to avoid 0x20, probably some hardware bug related to some misunderstanding of what a singed 8-bit value is (imagine: if it was a signed 6-bit value then someone could believe that 0x20 == 0x0). And then they didn't clean it up once they later on anandonned that whole story of having a constant offset for 40 MHz channels and just set the offset to be constant 0 -- there is no effort for avoiding 0x20 for the 20 MHz values stored in the EEPROM, hence that's probably just a forbidden value in the EEPROM specs and won't appear anyway... Anyway, the whole thing felt like solving some college math test where in the end everything cancels out and the result equals 0 ;) To make sure that channel bandwidth power compensation really doesn't need to be taken care of, output a warning when the corresponding value stored in the EEPROM is non-zero. Also there is no apparent reason to refrain from initializing RFCSR register 13, it doesn't fail what-so-ever. Signed-off-by: Daniel Golle <daniel@makrotopia.org>
2017-02-13 05:25:35 +00:00
+ rt2800_rfcsr_write(rt2x00dev, 12, rt2x00dev->freq_offset);
+ rt2800_rfcsr_write(rt2x00dev, 13, 0x00);
+ rt2800_rfcsr_write(rt2x00dev, 14, 0x40);
+ rt2800_rfcsr_write(rt2x00dev, 15, 0x22);
+ rt2800_rfcsr_write(rt2x00dev, 16, 0x4C);
+ rt2800_rfcsr_write(rt2x00dev, 17, 0x00);
+ rt2800_rfcsr_write(rt2x00dev, 18, 0x00);
+ rt2800_rfcsr_write(rt2x00dev, 19, 0x00);
+ rt2800_rfcsr_write(rt2x00dev, 20, 0xA0);
+ rt2800_rfcsr_write(rt2x00dev, 21, 0x12);
+ rt2800_rfcsr_write(rt2x00dev, 22, 0x07);
+ rt2800_rfcsr_write(rt2x00dev, 23, 0x13);
+ rt2800_rfcsr_write(rt2x00dev, 24, 0xFE);
+ rt2800_rfcsr_write(rt2x00dev, 25, 0x24);
+ rt2800_rfcsr_write(rt2x00dev, 26, 0x7A);
+ rt2800_rfcsr_write(rt2x00dev, 27, 0x00);
+ rt2800_rfcsr_write(rt2x00dev, 28, 0x00);
+ rt2800_rfcsr_write(rt2x00dev, 29, 0x05);
+ rt2800_rfcsr_write(rt2x00dev, 30, 0x00);
+ rt2800_rfcsr_write(rt2x00dev, 31, 0x00);
+ rt2800_rfcsr_write(rt2x00dev, 32, 0x00);
+ rt2800_rfcsr_write(rt2x00dev, 33, 0x00);
+ rt2800_rfcsr_write(rt2x00dev, 34, 0x00);
+ rt2800_rfcsr_write(rt2x00dev, 35, 0x00);
+ rt2800_rfcsr_write(rt2x00dev, 36, 0x00);
+ rt2800_rfcsr_write(rt2x00dev, 37, 0x00);
+ rt2800_rfcsr_write(rt2x00dev, 38, 0x00);
+ rt2800_rfcsr_write(rt2x00dev, 39, 0x00);
+ rt2800_rfcsr_write(rt2x00dev, 40, 0x00);
+ rt2800_rfcsr_write(rt2x00dev, 41, 0xD0);
+ rt2800_rfcsr_write(rt2x00dev, 42, 0x5B);
+ rt2800_rfcsr_write(rt2x00dev, 43, 0x00);
+
+ rt2800_rfcsr_write(rt2x00dev, 11, 0x21);
+ if (rt2800_clk_is_20mhz(rt2x00dev))
+ rt2800_rfcsr_write(rt2x00dev, 13, 0x03);
+ else
+ rt2800_rfcsr_write(rt2x00dev, 13, 0x00);
+ rt2800_rfcsr_write(rt2x00dev, 14, 0x7C);
+ rt2800_rfcsr_write(rt2x00dev, 16, 0x80);
+ rt2800_rfcsr_write(rt2x00dev, 17, 0x99);
+ rt2800_rfcsr_write(rt2x00dev, 18, 0x99);
+ rt2800_rfcsr_write(rt2x00dev, 19, 0x09);
+ rt2800_rfcsr_write(rt2x00dev, 20, 0x50);
+ rt2800_rfcsr_write(rt2x00dev, 21, 0xB0);
+ rt2800_rfcsr_write(rt2x00dev, 22, 0x00);
+ rt2800_rfcsr_write(rt2x00dev, 23, 0x06);
+ rt2800_rfcsr_write(rt2x00dev, 24, 0x00);
+ rt2800_rfcsr_write(rt2x00dev, 25, 0x00);
+ rt2800_rfcsr_write(rt2x00dev, 26, 0x5D);
+ rt2800_rfcsr_write(rt2x00dev, 27, 0x00);
+ rt2800_rfcsr_write(rt2x00dev, 28, 0x61);
+ rt2800_rfcsr_write(rt2x00dev, 29, 0xB5);
+ rt2800_rfcsr_write(rt2x00dev, 43, 0x02);
+
+ rt2800_rfcsr_write(rt2x00dev, 28, 0x62);
+ rt2800_rfcsr_write(rt2x00dev, 29, 0xAD);
+ rt2800_rfcsr_write(rt2x00dev, 39, 0x80);
+
+ /* Initialize RF channel register to default value */
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 0, 0x03);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 1, 0x00);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 2, 0x00);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 3, 0x00);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 4, 0x00);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 5, 0x08);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 6, 0x00);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 7, 0x51);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 8, 0x53);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 9, 0x16);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 10, 0x61);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 11, 0x53);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 12, 0x22);
rt2x00: mt7620: yet another beauty session So here is another round of improvements for MT7620 WiFi. This commit fixes a few significant issues related to TX_PWR_CFG_x and TX_ALC and also makes the code more readable by adding register descriptions for things added for MT7620 and use the usual bit-field access macros and the now defined macros instead of plain bit-ops and magic numbers. Properly describe EEPROM_TARGET_POWER at word 0x68 (== byte 0xD0) and thereby fix internal TXALC which would otherwise just read out-of-bounds of the EEPROM map. Split-out tx-power/ALC related stuff into an additional function. Fix VCO calibration, it was carried out properly in the channel switching but incomplete in the actual VCO calibration function. Also there is no need to trigger VCO calibration in channel switching, the VCO calibration function is already being called at this point. Remove it from channel switching function to avoid redundant code. The TX power calibration differs significantly from all other Mediatek/Ralink chips: They finally allow 0.5dB steps stored as 8-bit values for (almost) each bitrate -- and promptly ran out of space and for some reason didn't want to change the EEPROM layout. The hence opted for a scheme of sharing values for some adjecent bitrates and a highly over-complicated (or obfuscated?) way to populate the TX_PWR_CFG_x registers with the values stored in the EEPROM. The code here now looks much less complicated than what you see in the vendor's driver, however, it does the exact same thing: bGpwrdeltaMinus is a constant and always TRUE, hence half of the code was dead. Gpwrdelta is always 0 (rather than using the value read from the EEPROM). What remains is some very grotesque effort to avoid 0x20, probably some hardware bug related to some misunderstanding of what a singed 8-bit value is (imagine: if it was a signed 6-bit value then someone could believe that 0x20 == 0x0). And then they didn't clean it up once they later on anandonned that whole story of having a constant offset for 40 MHz channels and just set the offset to be constant 0 -- there is no effort for avoiding 0x20 for the 20 MHz values stored in the EEPROM, hence that's probably just a forbidden value in the EEPROM specs and won't appear anyway... Anyway, the whole thing felt like solving some college math test where in the end everything cancels out and the result equals 0 ;) To make sure that channel bandwidth power compensation really doesn't need to be taken care of, output a warning when the corresponding value stored in the EEPROM is non-zero. Also there is no apparent reason to refrain from initializing RFCSR register 13, it doesn't fail what-so-ever. Signed-off-by: Daniel Golle <daniel@makrotopia.org>
2017-02-13 05:25:35 +00:00
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 13, 0x3D);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 14, 0x06);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 15, 0x13);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 16, 0x22);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 17, 0x27);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 18, 0x02);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 19, 0xA7);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 20, 0x01);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 21, 0x52);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 22, 0x80);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 23, 0xB3);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 24, 0x00);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 25, 0x00);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 26, 0x00);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 27, 0x00);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 28, 0x5C);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 29, 0x6B);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 30, 0x6B);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 31, 0x31);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 32, 0x5D);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 33, 0x00);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 34, 0xE6);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 35, 0x55);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 36, 0x00);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 37, 0xBB);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 38, 0xB3);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 39, 0xB3);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 40, 0x03);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 41, 0x00);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 42, 0x00);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 43, 0xB3);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 44, 0xD3);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 45, 0xD5);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 46, 0x07);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 47, 0x68);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 48, 0xEF);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 49, 0x1C);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 54, 0x07);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 55, 0xA8);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 56, 0x85);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 57, 0x10);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 58, 0x07);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 59, 0x6A);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 60, 0x85);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 61, 0x10);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 62, 0x1C);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 63, 0x00);
+
rt2x00: mt7620: lots of improvements This commit combines all the changes I've made on my staging tree into a single commit fixing many issues with our patch for MT7620. First of all, checkpatch.pl revealed numerous code style issues with the patch, so fix all the white-space and commets. Also use usleep_range instead of legacy timing and relax timing for VCO calibration just like the vendor driver does. Several line programming registers were commented out in the patch. Originally this came from the features present but disabled by default in the vendor's driver (RTMP_TEMPERATURE_CALIBRATION and ADJUST_POWER_CONSUMPTION_SUPPORT). Remove the dead code for now, it can easily be re-added if we actually intend to support those features. Move values from mt7620_freqconfig type into the existing rf_channel struct, this shouldn't be a new typedef and it is possible to use the existing struct because rf_channel got 4 32-bit fields, so two of the 8-bit values from mt7620_freqconfig can easily be stored in the same 32-bit field. Map values such that Rdiv -> rf1 N -> rf2 K -> rf3[0:7] D -> rf3[8:15] Ksd -> rf4 This makes the channel switching logic already look a bit more like what we are used to in rt2x00... Probably many of the read-modify-write calls could still be replaced by macros intended for that. iq calibration seems to be identical to RT5592, so just enable it. Test shows that this improves things quite a lot, datarates went up by a couple of megabits when running iperf, signal quality seems jumpy in the first few seconds once a station connencts, the stabelizes on a value significantly better than what it was before. Add description to the patch and reference the original OpenWrt commit by which it was added. The patch now passes checkpatch.pl and can thus be discussed with the upstream authors of the rt2x00 driver. Funded-by: https://www.kickstarter.com/projects/1327597961/better-support-for-mt7620a-n-in-openwrt-lede/ Signed-off-by: Daniel Golle <daniel@makrotopia.org>
2017-02-03 14:02:32 +00:00
+ rt2800_rfcsr_write_bank(rt2x00dev, 6, 45, 0xC5);
+
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 9, 0x47);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 10, 0x71);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 11, 0x33);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 14, 0x0E);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 17, 0x23);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 19, 0xA4);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 20, 0x02);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 21, 0x12);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 28, 0x1C);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 29, 0xEB);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 32, 0x7D);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 34, 0xD6);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 36, 0x08);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 38, 0xB4);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 43, 0xD3);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 44, 0xB3);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 45, 0xD5);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 46, 0x27);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 47, 0x69);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 48, 0xFF);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 54, 0x20);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 55, 0x66);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 56, 0xFF);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 57, 0x1C);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 58, 0x20);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 59, 0x6B);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 60, 0xF7);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 61, 0x09);
+
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 10, 0x51);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 14, 0x06);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 19, 0xA7);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 28, 0x2C);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 55, 0x64);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 8, 0x51);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 9, 0x36);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 11, 0x53);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 14, 0x16);
+
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 47, 0x6C);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 48, 0xFC);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 49, 0x1F);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 54, 0x27);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 55, 0x66);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 59, 0x6B);
+
+ /* Initialize RF channel register for DRQFN */
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 43, 0xD3);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 44, 0xE3);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 45, 0xE5);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 47, 0x28);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 55, 0x68);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 56, 0xF7);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 58, 0x02);
+ rt2800_rfcsr_write_chanreg(rt2x00dev, 60, 0xC7);
+
+ /* Initialize RF DC calibration register to default value */
+ rt2800_rfcsr_write_dccal(rt2x00dev, 0, 0x47);
+ rt2800_rfcsr_write_dccal(rt2x00dev, 1, 0x00);
+ rt2800_rfcsr_write_dccal(rt2x00dev, 2, 0x00);
+ rt2800_rfcsr_write_dccal(rt2x00dev, 3, 0x00);
+ rt2800_rfcsr_write_dccal(rt2x00dev, 4, 0x00);
+ rt2800_rfcsr_write_dccal(rt2x00dev, 5, 0x00);
+ rt2800_rfcsr_write_dccal(rt2x00dev, 6, 0x10);
+ rt2800_rfcsr_write_dccal(rt2x00dev, 7, 0x10);
+ rt2800_rfcsr_write_dccal(rt2x00dev, 8, 0x04);
+ rt2800_rfcsr_write_dccal(rt2x00dev, 9, 0x00);
+ rt2800_rfcsr_write_dccal(rt2x00dev, 10, 0x07);
+ rt2800_rfcsr_write_dccal(rt2x00dev, 11, 0x01);
+ rt2800_rfcsr_write_dccal(rt2x00dev, 12, 0x07);
+ rt2800_rfcsr_write_dccal(rt2x00dev, 13, 0x07);
+ rt2800_rfcsr_write_dccal(rt2x00dev, 14, 0x07);
+ rt2800_rfcsr_write_dccal(rt2x00dev, 15, 0x20);
+ rt2800_rfcsr_write_dccal(rt2x00dev, 16, 0x22);
+ rt2800_rfcsr_write_dccal(rt2x00dev, 17, 0x00);
+ rt2800_rfcsr_write_dccal(rt2x00dev, 18, 0x00);
+ rt2800_rfcsr_write_dccal(rt2x00dev, 19, 0x00);
+ rt2800_rfcsr_write_dccal(rt2x00dev, 20, 0x00);
+ rt2800_rfcsr_write_dccal(rt2x00dev, 21, 0xF1);
+ rt2800_rfcsr_write_dccal(rt2x00dev, 22, 0x11);
+ rt2800_rfcsr_write_dccal(rt2x00dev, 23, 0x02);
+ rt2800_rfcsr_write_dccal(rt2x00dev, 24, 0x41);
+ rt2800_rfcsr_write_dccal(rt2x00dev, 25, 0x20);
+ rt2800_rfcsr_write_dccal(rt2x00dev, 26, 0x00);
+ rt2800_rfcsr_write_dccal(rt2x00dev, 27, 0xD7);
+ rt2800_rfcsr_write_dccal(rt2x00dev, 28, 0xA2);
+ rt2800_rfcsr_write_dccal(rt2x00dev, 29, 0x20);
+ rt2800_rfcsr_write_dccal(rt2x00dev, 30, 0x49);
+ rt2800_rfcsr_write_dccal(rt2x00dev, 31, 0x20);
+ rt2800_rfcsr_write_dccal(rt2x00dev, 32, 0x04);
+ rt2800_rfcsr_write_dccal(rt2x00dev, 33, 0xF1);
+ rt2800_rfcsr_write_dccal(rt2x00dev, 34, 0xA1);
+ rt2800_rfcsr_write_dccal(rt2x00dev, 35, 0x01);
+ rt2800_rfcsr_write_dccal(rt2x00dev, 41, 0x00);
+ rt2800_rfcsr_write_dccal(rt2x00dev, 42, 0x00);
+ rt2800_rfcsr_write_dccal(rt2x00dev, 43, 0x00);
+ rt2800_rfcsr_write_dccal(rt2x00dev, 44, 0x00);
+ rt2800_rfcsr_write_dccal(rt2x00dev, 45, 0x00);
+ rt2800_rfcsr_write_dccal(rt2x00dev, 46, 0x00);
+ rt2800_rfcsr_write_dccal(rt2x00dev, 47, 0x3E);
+ rt2800_rfcsr_write_dccal(rt2x00dev, 48, 0x3D);
+ rt2800_rfcsr_write_dccal(rt2x00dev, 49, 0x3E);
+ rt2800_rfcsr_write_dccal(rt2x00dev, 50, 0x3D);
+ rt2800_rfcsr_write_dccal(rt2x00dev, 51, 0x3E);
+ rt2800_rfcsr_write_dccal(rt2x00dev, 52, 0x3D);
+ rt2800_rfcsr_write_dccal(rt2x00dev, 53, 0x00);
+ rt2800_rfcsr_write_dccal(rt2x00dev, 54, 0x00);
+ rt2800_rfcsr_write_dccal(rt2x00dev, 55, 0x00);
+ rt2800_rfcsr_write_dccal(rt2x00dev, 56, 0x00);
+ rt2800_rfcsr_write_dccal(rt2x00dev, 57, 0x00);
+ rt2800_rfcsr_write_dccal(rt2x00dev, 58, 0x10);
+ rt2800_rfcsr_write_dccal(rt2x00dev, 59, 0x10);
+ rt2800_rfcsr_write_dccal(rt2x00dev, 60, 0x0A);
+ rt2800_rfcsr_write_dccal(rt2x00dev, 61, 0x00);
+ rt2800_rfcsr_write_dccal(rt2x00dev, 62, 0x00);
+ rt2800_rfcsr_write_dccal(rt2x00dev, 63, 0x00);
+
+ rt2800_rfcsr_write_dccal(rt2x00dev, 3, 0x08);
+ rt2800_rfcsr_write_dccal(rt2x00dev, 4, 0x04);
+ rt2800_rfcsr_write_dccal(rt2x00dev, 5, 0x20);
+
+ rt2800_rfcsr_write_dccal(rt2x00dev, 5, 0x00);
+ rt2800_rfcsr_write_dccal(rt2x00dev, 17, 0x7C);
+
+ rt2800_bw_filter_calibration(rt2x00dev, true);
+ rt2800_bw_filter_calibration(rt2x00dev, false);
+}
+
static void rt2800_init_rfcsr(struct rt2x00_dev *rt2x00dev)
{
if (rt2800_is_305x_soc(rt2x00dev)) {
@@ -6941,6 +8279,9 @@ static void rt2800_init_rfcsr(struct rt2x00_dev *rt2x00dev)
case RT5592:
rt2800_init_rfcsr_5592(rt2x00dev);
break;
+ case RT6352:
+ rt2800_init_rfcsr_6352(rt2x00dev);
+ break;
}
}
@@ -7307,7 +8648,8 @@ static int rt2800_init_eeprom(struct rt2x00_dev *rt2x00dev)
*/
if (rt2x00_rt(rt2x00dev, RT3290) ||
rt2x00_rt(rt2x00dev, RT5390) ||
- rt2x00_rt(rt2x00dev, RT5392))
+ rt2x00_rt(rt2x00dev, RT5392) ||
+ rt2x00_rt(rt2x00dev, RT6352))
rt2800_eeprom_read(rt2x00dev, EEPROM_CHIP_ID, &rf);
else if (rt2x00_rt(rt2x00dev, RT3352))
rf = RF3322;
@@ -7339,6 +8681,7 @@ static int rt2800_init_eeprom(struct rt2x00_dev *rt2x00dev)
case RF5390:
case RF5392:
case RF5592:
+ case RF7620:
break;
default:
rt2x00_err(rt2x00dev, "Invalid RF chipset 0x%04x detected\n",
@@ -7746,6 +9089,23 @@ static const struct rf_channel rf_vals_5592_xtal40[] = {
rt2x00: mt7620: lots of improvements This commit combines all the changes I've made on my staging tree into a single commit fixing many issues with our patch for MT7620. First of all, checkpatch.pl revealed numerous code style issues with the patch, so fix all the white-space and commets. Also use usleep_range instead of legacy timing and relax timing for VCO calibration just like the vendor driver does. Several line programming registers were commented out in the patch. Originally this came from the features present but disabled by default in the vendor's driver (RTMP_TEMPERATURE_CALIBRATION and ADJUST_POWER_CONSUMPTION_SUPPORT). Remove the dead code for now, it can easily be re-added if we actually intend to support those features. Move values from mt7620_freqconfig type into the existing rf_channel struct, this shouldn't be a new typedef and it is possible to use the existing struct because rf_channel got 4 32-bit fields, so two of the 8-bit values from mt7620_freqconfig can easily be stored in the same 32-bit field. Map values such that Rdiv -> rf1 N -> rf2 K -> rf3[0:7] D -> rf3[8:15] Ksd -> rf4 This makes the channel switching logic already look a bit more like what we are used to in rt2x00... Probably many of the read-modify-write calls could still be replaced by macros intended for that. iq calibration seems to be identical to RT5592, so just enable it. Test shows that this improves things quite a lot, datarates went up by a couple of megabits when running iperf, signal quality seems jumpy in the first few seconds once a station connencts, the stabelizes on a value significantly better than what it was before. Add description to the patch and reference the original OpenWrt commit by which it was added. The patch now passes checkpatch.pl and can thus be discussed with the upstream authors of the rt2x00 driver. Funded-by: https://www.kickstarter.com/projects/1327597961/better-support-for-mt7620a-n-in-openwrt-lede/ Signed-off-by: Daniel Golle <daniel@makrotopia.org>
2017-02-03 14:02:32 +00:00
{196, 83, 0, 12, 1},
};
+static const struct rf_channel rf_vals_7620[] = {
+ {1, 0x50, 0x99, 0x99, 1},
+ {2, 0x50, 0x44, 0x44, 2},
+ {3, 0x50, 0xEE, 0xEE, 2},
+ {4, 0x50, 0x99, 0x99, 3},
+ {5, 0x51, 0x44, 0x44, 0},
+ {6, 0x51, 0xEE, 0xEE, 0},
+ {7, 0x51, 0x99, 0x99, 1},
+ {8, 0x51, 0x44, 0x44, 2},
+ {9, 0x51, 0xEE, 0xEE, 2},
+ {10, 0x51, 0x99, 0x99, 3},
+ {11, 0x52, 0x44, 0x44, 0},
+ {12, 0x52, 0xEE, 0xEE, 0},
+ {13, 0x52, 0x99, 0x99, 1},
+ {14, 0x52, 0x33, 0x33, 3},
rt2x00: mt7620: lots of improvements This commit combines all the changes I've made on my staging tree into a single commit fixing many issues with our patch for MT7620. First of all, checkpatch.pl revealed numerous code style issues with the patch, so fix all the white-space and commets. Also use usleep_range instead of legacy timing and relax timing for VCO calibration just like the vendor driver does. Several line programming registers were commented out in the patch. Originally this came from the features present but disabled by default in the vendor's driver (RTMP_TEMPERATURE_CALIBRATION and ADJUST_POWER_CONSUMPTION_SUPPORT). Remove the dead code for now, it can easily be re-added if we actually intend to support those features. Move values from mt7620_freqconfig type into the existing rf_channel struct, this shouldn't be a new typedef and it is possible to use the existing struct because rf_channel got 4 32-bit fields, so two of the 8-bit values from mt7620_freqconfig can easily be stored in the same 32-bit field. Map values such that Rdiv -> rf1 N -> rf2 K -> rf3[0:7] D -> rf3[8:15] Ksd -> rf4 This makes the channel switching logic already look a bit more like what we are used to in rt2x00... Probably many of the read-modify-write calls could still be replaced by macros intended for that. iq calibration seems to be identical to RT5592, so just enable it. Test shows that this improves things quite a lot, datarates went up by a couple of megabits when running iperf, signal quality seems jumpy in the first few seconds once a station connencts, the stabelizes on a value significantly better than what it was before. Add description to the patch and reference the original OpenWrt commit by which it was added. The patch now passes checkpatch.pl and can thus be discussed with the upstream authors of the rt2x00 driver. Funded-by: https://www.kickstarter.com/projects/1327597961/better-support-for-mt7620a-n-in-openwrt-lede/ Signed-off-by: Daniel Golle <daniel@makrotopia.org>
2017-02-03 14:02:32 +00:00
+};
+
static int rt2800_probe_hw_mode(struct rt2x00_dev *rt2x00dev)
{
struct hw_mode_spec *spec = &rt2x00dev->spec;
@@ -7849,6 +9209,11 @@ static int rt2800_probe_hw_mode(struct rt2x00_dev *rt2x00dev)
rt2x00: mt7620: lots of improvements This commit combines all the changes I've made on my staging tree into a single commit fixing many issues with our patch for MT7620. First of all, checkpatch.pl revealed numerous code style issues with the patch, so fix all the white-space and commets. Also use usleep_range instead of legacy timing and relax timing for VCO calibration just like the vendor driver does. Several line programming registers were commented out in the patch. Originally this came from the features present but disabled by default in the vendor's driver (RTMP_TEMPERATURE_CALIBRATION and ADJUST_POWER_CONSUMPTION_SUPPORT). Remove the dead code for now, it can easily be re-added if we actually intend to support those features. Move values from mt7620_freqconfig type into the existing rf_channel struct, this shouldn't be a new typedef and it is possible to use the existing struct because rf_channel got 4 32-bit fields, so two of the 8-bit values from mt7620_freqconfig can easily be stored in the same 32-bit field. Map values such that Rdiv -> rf1 N -> rf2 K -> rf3[0:7] D -> rf3[8:15] Ksd -> rf4 This makes the channel switching logic already look a bit more like what we are used to in rt2x00... Probably many of the read-modify-write calls could still be replaced by macros intended for that. iq calibration seems to be identical to RT5592, so just enable it. Test shows that this improves things quite a lot, datarates went up by a couple of megabits when running iperf, signal quality seems jumpy in the first few seconds once a station connencts, the stabelizes on a value significantly better than what it was before. Add description to the patch and reference the original OpenWrt commit by which it was added. The patch now passes checkpatch.pl and can thus be discussed with the upstream authors of the rt2x00 driver. Funded-by: https://www.kickstarter.com/projects/1327597961/better-support-for-mt7620a-n-in-openwrt-lede/ Signed-off-by: Daniel Golle <daniel@makrotopia.org>
2017-02-03 14:02:32 +00:00
spec->channels = rf_vals_3x;
break;
+ case RF7620:
rt2x00: mt7620: lots of improvements This commit combines all the changes I've made on my staging tree into a single commit fixing many issues with our patch for MT7620. First of all, checkpatch.pl revealed numerous code style issues with the patch, so fix all the white-space and commets. Also use usleep_range instead of legacy timing and relax timing for VCO calibration just like the vendor driver does. Several line programming registers were commented out in the patch. Originally this came from the features present but disabled by default in the vendor's driver (RTMP_TEMPERATURE_CALIBRATION and ADJUST_POWER_CONSUMPTION_SUPPORT). Remove the dead code for now, it can easily be re-added if we actually intend to support those features. Move values from mt7620_freqconfig type into the existing rf_channel struct, this shouldn't be a new typedef and it is possible to use the existing struct because rf_channel got 4 32-bit fields, so two of the 8-bit values from mt7620_freqconfig can easily be stored in the same 32-bit field. Map values such that Rdiv -> rf1 N -> rf2 K -> rf3[0:7] D -> rf3[8:15] Ksd -> rf4 This makes the channel switching logic already look a bit more like what we are used to in rt2x00... Probably many of the read-modify-write calls could still be replaced by macros intended for that. iq calibration seems to be identical to RT5592, so just enable it. Test shows that this improves things quite a lot, datarates went up by a couple of megabits when running iperf, signal quality seems jumpy in the first few seconds once a station connencts, the stabelizes on a value significantly better than what it was before. Add description to the patch and reference the original OpenWrt commit by which it was added. The patch now passes checkpatch.pl and can thus be discussed with the upstream authors of the rt2x00 driver. Funded-by: https://www.kickstarter.com/projects/1327597961/better-support-for-mt7620a-n-in-openwrt-lede/ Signed-off-by: Daniel Golle <daniel@makrotopia.org>
2017-02-03 14:02:32 +00:00
+ spec->num_channels = ARRAY_SIZE(rf_vals_7620);
+ spec->channels = rf_vals_7620;
+ break;
+
case RF3052:
case RF3053:
spec->num_channels = ARRAY_SIZE(rf_vals_3x);
@@ -7980,6 +9345,7 @@ static int rt2800_probe_hw_mode(struct rt2x00_dev *rt2x00dev)
case RF5390:
case RF5392:
case RF5592:
+ case RF7620:
__set_bit(CAPABILITY_VCO_RECALIBRATION, &rt2x00dev->cap_flags);
break;
}
@@ -8024,6 +9390,9 @@ static int rt2800_probe_rt(struct rt2x00_dev *rt2x00dev)
return -ENODEV;
}
+ if (rt == RT5390 && rt2x00_is_soc(rt2x00dev))
+ rt = RT6352;
+
rt2x00_set_rt(rt2x00dev, rt, rev);
return 0;
diff --git a/drivers/net/wireless/ralink/rt2x00/rt2800lib.h b/drivers/net/wireless/ralink/rt2x00/rt2800lib.h
index d9ef260d542a..f357531d9488 100644
--- a/drivers/net/wireless/ralink/rt2x00/rt2800lib.h
+++ b/drivers/net/wireless/ralink/rt2x00/rt2800lib.h
@@ -33,6 +33,10 @@
struct rt2800_drv_data {
u8 calibration_bw20;
u8 calibration_bw40;
+ char rx_calibration_bw20;
+ char rx_calibration_bw40;
+ char tx_calibration_bw20;
+ char tx_calibration_bw40;
u8 bbp25;
u8 bbp26;
u8 txmixer_gain_24g;
diff --git a/drivers/net/wireless/ralink/rt2x00/rt2x00.h b/drivers/net/wireless/ralink/rt2x00/rt2x00.h
index ce340bfd71a0..8fdd2f9726ee 100644
--- a/drivers/net/wireless/ralink/rt2x00/rt2x00.h
+++ b/drivers/net/wireless/ralink/rt2x00/rt2x00.h
@@ -174,6 +174,7 @@ struct rt2x00_chip {
#define RT5390 0x5390 /* 2.4GHz */
#define RT5392 0x5392 /* 2.4GHz */
#define RT5592 0x5592
+#define RT6352 0x6352 /* WSOC 2.4GHz */
u16 rf;
u16 rev;
--
2.12.1