8814 port-audit: fix 8812-on-8814 mis-ports + CCK/FW-boot bugs (no regressions)#96

Merged

josephnef merged 31 commits into

masterfrom

8814-port-audit

Jun 11, 2026

Commits on Jun 10, 2026

8814: clamp CCK TX rate to OFDM on 5GHz channels (fixes 5GHz TX)

josephnef
and
claude
committed
8814: env-gated deinit-before-init scaffold (DEVOURER_8814_DEINIT_BEFORE_INIT)
josephnef
committed
8814: add missing RFE GPIO pin-select at init (kernel PHY_SetRFEReg8814A bInit)
josephnef
committed
8814: port 8814 _InitBurstPktLen; 8812 body clobbered BCNQ1 boundary (kernel usb_halinit.c:_InitBurstPktLen)

josephnef
and
claude
committed
8814: keep USTIME_TSF/EDCA at the 100MHz MAC-table value (kernel usb_halinit.c:_InitEDCA_8814AUsb)

josephnef
and
claude
committed
8814: arm USB TX-agg block-descriptor config (kernel usb_halinit.c:usb_AggSettingTxUpdate_8814A)

josephnef
and
claude
committed
8814: program per-byte aggregation limits 0x4CA/0x4CB=0x36 (kernel usb_halinit.c:_InitMacConfigure_8814A)

josephnef
and
claude
committed
8814: restore NAV_UPPER=0xEB after the init zero-write (kernel usb_halinit.c HW_VAR_NAV_UPPER)

josephnef
and
claude
committed
8814: skip REG_USB_HRPWM init write (kernel usb_halinit.c:1354 has it commented out)

josephnef
and
claude
committed
docs: drop stale DEVOURER_FORCE_TXPWR; TX power runs by default

josephnef
and
claude
committed
8814: make deinit-before-init scaffold faithful to kernel card-disable (kernel usb_halinit.c:hal_carddisable_8814)

josephnef
and
claude
committed
pwrseq: poll retry delay is 10us upstream, not 10ms (kernel HalPwrSeqCmd.c:134)

josephnef
and
claude
committed
8814: FW-boot poll must require chip-set CPU_DL_READY (kernel rtl8814a_hal_init.c:649-656)

josephnef
and
claude
committed
8814: log the real fw version; header macro already offsets +4 (kernel GET_FIRMWARE_HDR_VERSION_3081)

josephnef
and
claude
committed
8814: correct register-name labels in fwdl comments (rtl8814a_spec.h)

josephnef
and
claude
committed
8814: 5GHz CCK clamp read an uninitialized channel - store it (kernel keys off maintained cur_channel)

josephnef
and
claude
committed
tx: VHT radiotap bandwidth never reached the descriptor (MHz literals vs CHANNEL_WIDTH enums)

josephnef
and
claude
committed
8814: fix crystal-cap trim bit positions + port RCK1 path sync (kernel phydm_cfotracking.c, rtl8814a_rf6052.c:143)

josephnef
and
claude
committed
8814: port secondary-CCA control write 0x577=0x03 (kernel usb_halinit.c:1250)

josephnef
and
claude
committed
8814: stop running 8812 TX-power-training writes (kernel PHY_SetTxPowerLevel8814 has none)

josephnef
and
claude
committed
8814: port the kernel RFE-type decision tree (kernel rtl8814a_hal_init.c:hal_ReadRFEType_8814A)

josephnef
and
claude
committed
8814: fix 5G EFUSE PG diff-block layout (kernel hal_com_phycfg.c:hal_load_pg_txpwr_info_path_5g)

josephnef
and
claude
committed
8814: per-Ntx TX-power diffs are cumulative over rate ranges (kernel hal_com_phycfg.c:2490-2601)

josephnef
and
claude
committed
8814: hygiene - label 0x670 as REG_CAMCMD, remove dead BIT0 LLT port

josephnef
and
claude
committed
rx: bulk-IN buffer must cover the 20K RX aggregate (kernel MAX_RECVBUF_SZ=32768)

josephnef
and
claude
committed
8814: clear USB_AGG_EN (0x283 bit7) like the kernel RX-agg setup (kernel usb_AggSettingRxUpdate_8814A)

josephnef
and
claude
committed
rx: parse hardening - physt gating, tail-fragment guard, 8814 DWORD4 caveat

josephnef
and
claude
committed
8814: RF reads must use the direct shadow blocks, not 8812 serial readback (kernel rtl8814a_phycfg.c:phy_RFRead_8814A)

josephnef
and
claude
committed
8814: rfe_type 2 5G pinmux nibble is 0x37173717 (kernel PHY_SetRFEReg8814A)

josephnef
and
claude
committed
8814: IGI floor must cover paths C/D too (kernel phydm_write_dig_reg, ODM_IC_AC_4SS)

josephnef
and
claude
committed
docs: 8814 port-audit report - ranked findings, coverage map, validation plan

josephnef
and
claude
committed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

8814 port-audit: fix 8812-on-8814 mis-ports + CCK/FW-boot bugs (no regressions)#96

8814 port-audit: fix 8812-on-8814 mis-ports + CCK/FW-boot bugs (no regressions)#96
josephnef merged 31 commits into
masterfrom
8814-port-audit

Commits on Jun 10, 2026

8814: clamp CCK TX rate to OFDM on 5GHz channels (fixes 5GHz TX)

8814: env-gated deinit-before-init scaffold (DEVOURER_8814_DEINIT_BEFORE_INIT)

8814: add missing RFE GPIO pin-select at init (kernel PHY_SetRFEReg8814A bInit)

8814: port 8814 _InitBurstPktLen; 8812 body clobbered BCNQ1 boundary (kernel usb_halinit.c:_InitBurstPktLen)

8814: keep USTIME_TSF/EDCA at the 100MHz MAC-table value (kernel usb_halinit.c:_InitEDCA_8814AUsb)

8814: arm USB TX-agg block-descriptor config (kernel usb_halinit.c:usb_AggSettingTxUpdate_8814A)

8814: program per-byte aggregation limits 0x4CA/0x4CB=0x36 (kernel usb_halinit.c:_InitMacConfigure_8814A)

8814: restore NAV_UPPER=0xEB after the init zero-write (kernel usb_halinit.c HW_VAR_NAV_UPPER)

8814: skip REG_USB_HRPWM init write (kernel usb_halinit.c:1354 has it commented out)

docs: drop stale DEVOURER_FORCE_TXPWR; TX power runs by default

8814: make deinit-before-init scaffold faithful to kernel card-disable (kernel usb_halinit.c:hal_carddisable_8814)

pwrseq: poll retry delay is 10us upstream, not 10ms (kernel HalPwrSeqCmd.c:134)

8814: FW-boot poll must require chip-set CPU_DL_READY (kernel rtl8814a_hal_init.c:649-656)

8814: log the real fw version; header macro already offsets +4 (kernel GET_FIRMWARE_HDR_VERSION_3081)

8814: correct register-name labels in fwdl comments (rtl8814a_spec.h)

8814: 5GHz CCK clamp read an uninitialized channel - store it (kernel keys off maintained cur_channel)

tx: VHT radiotap bandwidth never reached the descriptor (MHz literals vs CHANNEL_WIDTH enums)

8814: fix crystal-cap trim bit positions + port RCK1 path sync (kernel phydm_cfotracking.c, rtl8814a_rf6052.c:143)

8814: port secondary-CCA control write 0x577=0x03 (kernel usb_halinit.c:1250)

8814: stop running 8812 TX-power-training writes (kernel PHY_SetTxPowerLevel8814 has none)

8814: port the kernel RFE-type decision tree (kernel rtl8814a_hal_init.c:hal_ReadRFEType_8814A)

8814: fix 5G EFUSE PG diff-block layout (kernel hal_com_phycfg.c:hal_load_pg_txpwr_info_path_5g)

8814: per-Ntx TX-power diffs are cumulative over rate ranges (kernel hal_com_phycfg.c:2490-2601)

8814: hygiene - label 0x670 as REG_CAMCMD, remove dead BIT0 LLT port

rx: bulk-IN buffer must cover the 20K RX aggregate (kernel MAX_RECVBUF_SZ=32768)

8814: clear USB_AGG_EN (0x283 bit7) like the kernel RX-agg setup (kernel usb_AggSettingRxUpdate_8814A)

rx: parse hardening - physt gating, tail-fragment guard, 8814 DWORD4 caveat

8814: RF reads must use the direct shadow blocks, not 8812 serial readback (kernel rtl8814a_phycfg.c:phy_RFRead_8814A)

8814: rfe_type 2 5G pinmux nibble is 0x37173717 (kernel PHY_SetRFEReg8814A)

8814: IGI floor must cover paths C/D too (kernel phydm_write_dig_reg, ODM_IC_AC_4SS)

docs: 8814 port-audit report - ranked findings, coverage map, validation plan