Re: [PATCH rtw-next v3 2/5] wifi: rtw88: 88xxa: receive VHT SU beamformed transmissions

[Luka Gejak <luka.gejak@linux.dev>]

Hi Jie Feng,

On November 13, 2026 9:20:57 AM GMT+01:00, Jie Feng <hjf3108@gmail.com>
wrote:
> I can dig the vendor code for some confirmation that the csi param format
> is the same for the later chips as 88xxa.

I had a look, and I think all of these values can be named from Realtek's
own code, so nothing has to be guessed.

For csi_param, bf.c already documents the fields above
rtw_bf_enable_bfee_mu():

        /* nc index: 1 2T2R 0 1T1R
         * nr index: 1 use Nsts 0 use reg setting
         * codebookinfo: 1 802.11ac 3 802.11n
         */

and the 88xxa vendor values follow it. Per slot, hal_txbf_jaguar_enter()
uses 0x0109 and 0x0108 for VHT and 0x0309 and 0x0308 for HT, so only
codebookinfo and nc index change. csi_param could be built like in
rtw_bf_enable_bfee_su() and then copied into the high half for BFMER1.

For 0x9b4, rtw8812a_table.c already loads 0x00000000 and
rtw8821a_table.c loads 0x01081008, which fits the "nc =2" and "nc =1"
comments in the vendor code. If I read it right, the write only changes
something on an 8812au whose efuse says 1T1R. Two defines for the values
would make that clear.

For 0x0202, the halmac headers in Realtek's newer vendor drivers name the
bits of REG_SND_PTCL_CTRL. In halmac_bit_8822b.h:

        #define BIT_SHIFT_CSI_RPT_OFFSET_HT_V1_8822B 16
        #define BIT_SHIFT_R_CSI_RPT_OFFSET_VHT_V1_8822B 8

so 0x0202 looks like a CSI report offset of 2 for both HT and VHT. On
8821c and 8822c, bits 15:8 hold the VHT category instead, so maybe
Ping-Ke can confirm how it is on 8814a.

The low byte has the same layout on all of these chips, so the sounding
values could use names as well. 0xCB is USE_NSTS,
DISABLE_CHECK_VHTSIGB_CRC,
CSISEQ_SEL, HT_NDPA_EN and VHT_NDPA_EN, and 0xDB adds BFPARAM_SEL.

Best regards,
Luka Gejak