On 05/08/2025 03:40, Ping-Ke Shih wrote:
> Bitterblue Smith <rtl8821cerfe2@xxxxxxxxx> wrote:
>> On 01/08/2025 23:51, Bitterblue Smith wrote:
>>> It seems RTL8852CU can only use TX channels 0, 2, and 8 (for band 0),
>>> otherwise the chip stops working after downloading at maximum speed
>>> for a few seconds.
>>>
>>> Signed-off-by: Bitterblue Smith <rtl8821cerfe2@xxxxxxxxx>
>>> ---
>>> v2:
>>> - No change, messed up sending v1.
>>> ---
>>> drivers/net/wireless/realtek/rtw89/txrx.h | 19 +++++++++++++++++++
>>> 1 file changed, 19 insertions(+)
>>>
>>> diff --git a/drivers/net/wireless/realtek/rtw89/txrx.h b/drivers/net/wireless/realtek/rtw89/txrx.h
>>> index ec01bfc363da..318fd0ac8726 100644
>>> --- a/drivers/net/wireless/realtek/rtw89/txrx.h
>>> +++ b/drivers/net/wireless/realtek/rtw89/txrx.h
>>> @@ -734,6 +734,25 @@ rtw89_core_get_qsel_mgmt(struct rtw89_dev *rtwdev, struct rtw89_core_tx_request
>>>
>>> static inline u8 rtw89_core_get_ch_dma(struct rtw89_dev *rtwdev, u8 qsel)
>>> {
>>> + if (rtwdev->hci.type == RTW89_HCI_TYPE_USB &&
>>> + rtwdev->chip->chip_id == RTL8852C) {
>>> + switch (qsel) {
>>> + default:
>>> + rtw89_warn(rtwdev, "Cannot map qsel to dma: %d\n",
>>> + qsel);
>>> + fallthrough;
>>> + case RTW89_TX_QSEL_BE_0:
>>> + case RTW89_TX_QSEL_VO_0:
>>> + return RTW89_TXCH_ACH0;
>>> + case RTW89_TX_QSEL_BK_0:
>>> + case RTW89_TX_QSEL_VI_0:
>>> + return RTW89_TXCH_ACH2;
>>> + case RTW89_TX_QSEL_B0_MGMT:
>>> + case RTW89_TX_QSEL_B0_HI:
>>> + return RTW89_TXCH_CH8;
>>> + }
>>> + }
>>> +
>>> switch (qsel) {
>>> default:
>>> rtw89_warn(rtwdev, "Cannot map qsel to dma: %d\n", qsel);
>>
>> I see now that RTL8922AU needs the same change. And same in patch 2/11
>> as well.
>
> The coming chip RTL8922DE uses different mapping rule, so I add a new
> chip_ops:
>
> --- a/core.h
> +++ b/core.h
> @@ -4072,6 +4072,7 @@ struct rtw89_chip_ops {
> void (*fill_txdesc_fwcmd)(struct rtw89_dev *rtwdev,
> struct rtw89_tx_desc_info *desc_info,
> void *txdesc);
> + u8 (*get_ch_dma)(struct rtw89_dev *rtwdev, u8 qsel);
> int (*cfg_ctrl_path)(struct rtw89_dev *rtwdev, bool wl);
> int (*mac_cfg_gnt)(struct rtw89_dev *rtwdev,
> const struct rtw89_mac_ax_coex_gnt *gnt_cfg);
> @@ -7428,6 +7429,14 @@ void rtw89_chip_fill_txdesc_fwcmd(struct rtw89_dev *rtwdev,
> chip->ops->fill_txdesc_fwcmd(rtwdev, desc_info, txdesc);
> }
>
> +static inline
> +u8 rtw89_chip_get_ch_dma(struct rtw89_dev *rtwdev, u8 qsel)
> +{
> + const struct rtw89_chip_info *chip = rtwdev->chip;
> +
> + return chip->ops->get_ch_dma(rtwdev, qsel);
> +}
> +
> static inline
> void rtw89_chip_mac_cfg_gnt(struct rtw89_dev *rtwdev,
> const struct rtw89_mac_ax_coex_gnt *gnt_cfg)
>
> Also the rtw89_core_get_ch_dma() is moved to core.c, and add
> rtw89_core_get_ch_dma_v1() for RTL8922DE.
>
> However, USB can use different mapping rule, so we should extend the size
> to RTW89_HCI_TYPE_NUM. Then,
>
> return chip->ops->get_ch_dma[rtwdev->hci.type](rtwdev, qsel);
>
> If you want my patch as first patch of this patchset, I can share it as RFC.
>
Yes, that would be good.
>
> Since this is called in data path, I'd like to avoid if-condition to save
> a little execution time.
>
