linux-kernel - Re: [PATCH] leds: rgb: leds-qcom-lpg: Fix pwm resolution for Hi-Res PWMs

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <dfthocttum7dscotngi6l2hz6bpdwfgrdxpvkcv6bdux3lt66d@iqfvmntvzyut>
Date: Fri, 21 Feb 2025 00:35:08 +0100
From: Sebastian Reichel <sre@...nel.org>
To: Abel Vesa <abel.vesa@...aro.org>
Cc: Lee Jones <lee@...nel.org>, Pavel Machek <pavel@...nel.org>, 
	Anjelique Melendez <quic_amelende@...cinc.com>, Kamal Wadhwa <quic_kamalw@...cinc.com>, 
	Jishnu Prakash <jishnu.prakash@....qualcomm.com>, Bjorn Andersson <andersson@...nel.org>, 
	Konrad Dybcio <konradybcio@...nel.org>, Johan Hovold <johan@...nel.org>, Pavel Machek <pavel@....cz>, 
	linux-leds@...r.kernel.org, linux-kernel@...r.kernel.org, stable@...r.kernel.org
Subject: Re: [PATCH] leds: rgb: leds-qcom-lpg: Fix pwm resolution for Hi-Res
 PWMs

Hello Abel,

On Thu, Feb 20, 2025 at 12:31:00PM +0200, Abel Vesa wrote:
> Currently, for the high resolution PWMs, the resolution, clock,
> pre-divider and exponent are being selected based on period. Basically,
> the implementation loops over each one of these and tries to find the
> closest (higher) period based on the following formula:
> 
>                           period * refclk
> prediv_exp = log2 -------------------------------------
>                     NSEC_PER_SEC * pre_div * resolution
> 
> Since the resolution is power of 2, the actual period resulting is
> usually higher than what the resolution allows. That's why the duty
> cycle requested needs to be capped to the maximum value allowed by the
> resolution (known as PWM size).
> 
> Here is an example of how this can happen:
> 
> For a requested period of 5000000, the best clock is 19.2MHz, the best
> prediv is 5, the best exponent is 6 and the best resolution is 256.
> 
> Then, the pwm value is determined based on requested period and duty
> cycle, best prediv, best exponent and best clock, using the following
> formula:
> 
>                             duty * refclk
> pwm_value = ----------------------------------------------
>                 NSEC_PER_SEC * prediv * (1 << prediv_exp)
> 
> So in this specific scenario:
> 
> (5000000 * 19200000) / (1000000000 * 5 * (1 << 64)) = 300
> 
> With a resolution of 8 bits, this pwm value obviously goes over.
> 
> Therefore, the max pwm value allowed needs to be 255.
> 
> If not, the PMIC internal logic will only value that is under the set PWM
> size, resulting in a wrapped around PWM value.
> 
> This has been observed on Lenovo Thinkpad T14s Gen6 (LCD panel version)
> which uses one of the PMK8550 to control the LCD backlight.
> 
> Fix the value of the PWM by capping to a max based on the chosen
> resolution (PWM size).
> 
> Cc: stable@...r.kernel.org    # 6.4
> Fixes: b00d2ed37617 ("leds: rgb: leds-qcom-lpg: Add support for high resolution PWM")
> Signed-off-by: Abel Vesa <abel.vesa@...aro.org>
> ---
> Note: This fix is blocking backlight support on Lenovo Thinkpad T14s
> Gen6 (LCD version), for which I have patches ready to send once this
> patch is agreed on (review) and merged.
> ---

Do you know if the pwm duty cycle to pwm value calculation is
correct otherwise?

I'm asking because the max value is only used for capping, so with
this patch the maximum brightness will be reached at around 80% duty
cycle (i.e. when the wrap over happens without this patch).

Locally I'm currently remapping the duty cycle range to the PWM
value range, which means the display brightness increases
step-by-step until reaching 100% "duty cycle":

		val = (duty * 255) / chan->period;
		chan->pwm_value = min(val, 255);

But for the backlight control the absolute numbers do not really
matter and I have zero knowledge about the chip. So it might be
that the controller really can only go up to ~80% duty cycle at
these settings?

Greetings,

-- Sebastian

>  drivers/leds/rgb/leds-qcom-lpg.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/drivers/leds/rgb/leds-qcom-lpg.c b/drivers/leds/rgb/leds-qcom-lpg.c
> index f3c9ef2bfa572f9ee86c8b8aa37deb8231965490..146cd9b447787bf170310321e939022dfb176e9f 100644
> --- a/drivers/leds/rgb/leds-qcom-lpg.c
> +++ b/drivers/leds/rgb/leds-qcom-lpg.c
> @@ -529,7 +529,7 @@ static void lpg_calc_duty(struct lpg_channel *chan, uint64_t duty)
>  	unsigned int clk_rate;
>  
>  	if (chan->subtype == LPG_SUBTYPE_HI_RES_PWM) {
> -		max = LPG_RESOLUTION_15BIT - 1;
> +		max = BIT(lpg_pwm_resolution_hi_res[chan->pwm_resolution_sel]) - 1;
>  		clk_rate = lpg_clk_rates_hi_res[chan->clk_sel];
>  	} else {
>  		max = LPG_RESOLUTION_9BIT - 1;
> 
> ---
> base-commit: 50a0c754714aa3ea0b0e62f3765eb666a1579f24
> change-id: 20250220-leds-qcom-lpg-fix-max-pwm-on-hi-res-067e8782a79b
> 
> Best regards,
> -- 
> Abel Vesa <abel.vesa@...aro.org>
> 

Download attachment "signature.asc" of type "application/pgp-signature" (834 bytes)