Re: [PATCH] staging: rtl8723bs: remove dead antenna diversity infrastructure (FAT/SWAT)

[Greg Kroah-Hartman <gregkh@linuxfoundation.org>]

On Fri, Oct 02, 2026 at 07:33:41PM -0500, Marinela Tatiana Selseth wrote:
> Deep analysis reveals that the driver contains major vestigial out-of-tree
> architectural frameworks for Fast Antenna Training (FAT) and Software
> Antenna Switching (SWAS) that are completely uninitialized, unread, or
> write-only.

what tool did this "deep analysis"?

> 
> Remove the 'struct fat_t' definition from hal/odm.h and refactor the
> fallback path in hal/rtl8723b_phycfg.c. Because the unassigned structural
> member 'RxIdleAnt' naturally defaults to 0, the conditional path check
> statically resolves to false, permanently routing RFPath to RF_PATH_B
> whenever antenna diversity is active. Hardcode this layout behavior to
> match factual reality and remove the now unused macro 'MAIN_ANT'.
> 
> Additionally, purge the write-only 'struct swat_t' infrastructure from
> hal/odm.h and include/hal_data.h. Inside hal/odm.c, eliminate the dead
> variable tracking assignments while explicitly preserving the hardware
> register read to rDPDT_control as a dummy read to ensure initialization
> sequencing and hardware timing remain completely unaltered.
> 
> Lastly, clean up legacy comment baggage across hal/odm.c and
> hal/odm_DIG.c. Remove arbitrary spreadsheet digits and references to
> unsupported legacy chipsets (92E, 8723A, 8821, 8812) to ensure inline
> notes are highly focused on the active RTL8723B Bluetooth coexistence
> parameters.
> 
> Signed-off-by: Marinela Tatiana Selseth <marinela.selseth@firmwaredesign.org>
> ---
>  drivers/staging/rtl8723bs/hal/odm.c           |  12 +--
>  drivers/staging/rtl8723bs/hal/odm.h           | 100 ------------------
>  drivers/staging/rtl8723bs/hal/odm_DIG.c       |   5 +-
>  .../staging/rtl8723bs/hal/rtl8723b_phycfg.c   |  11 +-
>  drivers/staging/rtl8723bs/include/hal_data.h  |   1 -
>  5 files changed, 11 insertions(+), 118 deletions(-)
> 
> diff --git a/drivers/staging/rtl8723bs/hal/odm.c b/drivers/staging/rtl8723bs/hal/odm.c
> index f3989fa33d9f..00b938071dee 100644
> --- a/drivers/staging/rtl8723bs/hal/odm.c
> +++ b/drivers/staging/rtl8723bs/hal/odm.c
> @@ -461,12 +461,8 @@ static void odm_RSSIMonitorCheck(struct dm_odm_t *pDM_Odm)
>  /* 3 ============================================================ */
>  static void odm_SwAntDetectInit(struct dm_odm_t *pDM_Odm)
>  {
> -	struct swat_t *pDM_SWAT_Table = &pDM_Odm->DM_SWAT_Table;
> -
> -	pDM_SWAT_Table->SWAS_NoLink_BK_Reg92c = rtw_read32(pDM_Odm->Adapter, rDPDT_control);
> -	pDM_SWAT_Table->PreAntenna = MAIN_ANT;
> -	pDM_SWAT_Table->CurAntenna = MAIN_ANT;
> -	pDM_SWAT_Table->SWAS_NoLink_State = 0;
> +	/* Dummy read rDPDT_control to preserve initialization hardware sequencing */
> +	rtw_read32(pDM_Odm->Adapter, rDPDT_control);

While it's nice to preserve this, how about testing on real hardware to
see if it too can be dropped?

thanks,

greg k-h