> -----Original Message-----
> From: Zenm Chen <zenmchen@xxxxxxxxx>
> Sent: Tuesday, September 12, 2023 7:27 PM
> To: Jes.Sorensen@xxxxxxxxx
> Cc: kvalo@xxxxxxxxxx; linux-wireless@xxxxxxxxxxxxxxx; linux-kernel@xxxxxxxxxxxxxxx;
> rtl8821cerfe2@xxxxxxxxx; Ping-Ke Shih <pkshih@xxxxxxxxxxx>; Zenm Chen <zenmchen@xxxxxxxxx>
> Subject: [PATCH v2] wifi: rtl8xxxu: fix LED control code of RTL8192FU
>
> Some of the RTL8192FU-based wifi adapters use the register "REG_LEDCFG1"
> to control the LED, and some of them use the register "REG_LEDCFG0"
> instead. Currently rtl8xxxu controls the LED via writing the values
> to the register "REG_LEDCFG1" only. This caused a few RTL8192FU-based wifi
> adapters's LED don't blink according to the network activity. This patch
> will make rtl8xxxu write the correct values to the both register
> "REG_LEDCFG0" and "REG_LEDCFG1" to fix this issue.
>
> This was tested with these two wifi adapters:
> ASUS USB-N13 C1 (vid=0x0b05, pid=0x18f1, rfe_type=0x1)
> MERCURY MW310UH (vid=0x0bda, pid=0xf192, rfe_type=0x5)
>
> Signed-off-by: Zenm Chen <zenmchen@xxxxxxxxx>
> ---
> v2:
> - Explain why to fix the issue in this way in the commit message.
> - Split a long statement into short ones.
> - Use some of the definitions suggested by Ping-Ke Shih.
> ---
> .../realtek/rtl8xxxu/rtl8xxxu_8192f.c | 26 ++++++++++++-------
> .../wireless/realtek/rtl8xxxu/rtl8xxxu_regs.h | 3 +++
> 2 files changed, 20 insertions(+), 9 deletions(-)
>
> diff --git a/drivers/net/wireless/realtek/rtl8xxxu/rtl8xxxu_8192f.c
> b/drivers/net/wireless/realtek/rtl8xxxu/rtl8xxxu_8192f.c
> index 28e93835e05a..779f93afc22b 100644
> --- a/drivers/net/wireless/realtek/rtl8xxxu/rtl8xxxu_8192f.c
> +++ b/drivers/net/wireless/realtek/rtl8xxxu/rtl8xxxu_8192f.c
> @@ -2014,26 +2014,34 @@ static int rtl8192fu_led_brightness_set(struct led_classdev *led_cdev,
> struct rtl8xxxu_priv *priv = container_of(led_cdev,
> struct rtl8xxxu_priv,
> led_cdev);
> - u16 ledcfg;
> + u32 ledcfg;
>
> /* Values obtained by observing the USB traffic from the Windows driver. */
> rtl8xxxu_write32(priv, REG_SW_GPIO_SHARE_CTRL_0, 0x20080);
> rtl8xxxu_write32(priv, REG_SW_GPIO_SHARE_CTRL_1, 0x1b0000);
>
> - ledcfg = rtl8xxxu_read16(priv, REG_LEDCFG0);
> + ledcfg = rtl8xxxu_read32(priv, REG_LEDCFG0);
> +
> + /* Set LED0 GPIO enabled */
> + ledcfg |= LEDCFG0_LED0_GPIO_ENABLE;
>
> if (brightness == LED_OFF) {
> - /* Value obtained like above. */
> - ledcfg = BIT(1) | BIT(7);
> + /* Setting REG_LEDCFG0[15:0] to 0x0000 turns LED0/LED1 off. */
> + ledcfg &= ~GENMASK(15, 0);
> } else if (brightness == LED_ON) {
> - /* Value obtained like above. */
> - ledcfg = BIT(1) | BIT(7) | BIT(11);
> + /* Setting REG_LEDCFG0[15:0] to 0x0808 turns LED0/LED1 on. */
> + ledcfg &= ~GENMASK(15, 0);
> + ledcfg |= LEDCFG0_LED1SV | LEDCFG0_LED0SV;
> } else if (brightness == RTL8XXXU_HW_LED_CONTROL) {
> - /* Value obtained by brute force. */
> - ledcfg = BIT(8) | BIT(9);
> + /* Setting REG_LEDCFG0[15:0] to 0x0303 enables
> + * hardware-controlled blinking for LED0/LED1.
> + * The value 0x0303 is obtained by brute force.
I suppose this would not be a brute force if you can use register definitions
I provided by v1.
> + */
> + ledcfg &= ~GENMASK(15, 0);
> + ledcfg |= BIT(9) | BIT(8) | BIT(1) | BIT(0);
> }
>
> - rtl8xxxu_write16(priv, REG_LEDCFG0, ledcfg);
> + rtl8xxxu_write32(priv, REG_LEDCFG0, ledcfg);
>
> return 0;
> }
> diff --git a/drivers/net/wireless/realtek/rtl8xxxu/rtl8xxxu_regs.h
> b/drivers/net/wireless/realtek/rtl8xxxu/rtl8xxxu_regs.h
> index 920ee50e2115..5ce845f1d069 100644
> --- a/drivers/net/wireless/realtek/rtl8xxxu/rtl8xxxu_regs.h
> +++ b/drivers/net/wireless/realtek/rtl8xxxu/rtl8xxxu_regs.h
> @@ -146,6 +146,9 @@
> #define GPIO_INTM_EDGE_TRIG_IRQ BIT(9)
>
> #define REG_LEDCFG0 0x004c
> +#define LEDCFG0_LED0SV BIT(3)
> +#define LEDCFG0_LED1SV BIT(11)
> +#define LEDCFG0_LED0_GPIO_ENABLE BIT(21)
> #define LEDCFG0_DPDT_SELECT BIT(23)
> #define REG_LEDCFG1 0x004d
> #define LEDCFG1_HW_LED_CONTROL BIT(1)
> --
> 2.42.0
