Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Message-ID: <c79d5a16-2716-6cd8-8b6a-f0e5916a65cf@gmail.com>
Date:   Fri, 21 Jul 2023 11:23:15 +0800
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:102.0) Gecko/20100101
 Firefox/102.0 Thunderbird/102.13.0
Subject: Re: [PATCH v6] pwm: atmel: Enable clk when pwm already enabled in
 bootloader
To:     Thierry Reding <thierry.reding@gmail.com>
Cc:     claudiu.beznea@microchip.com, u.kleine-koenig@pengutronix.de,
        nicolas.ferre@microchip.com, alexandre.belloni@bootlin.com,
        linux-arm-kernel@lists.infradead.org, linux-pwm@vger.kernel.org,
        linux-kernel@vger.kernel.org
References: <20230716020652.18557-1-aarongt.shen@gmail.com>
 <ZLlO3vGr-ECdnmKA@orome>
Content-Language: en-US
From:   Guiting Shen <aarongt.shen@gmail.com>
In-Reply-To: <ZLlO3vGr-ECdnmKA@orome>
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 7bit
Precedence: bulk

On Thu, Jul 20, 2023 at 23:12:30PM GMT+8, Thierry Reding wrote:
> On Sun, Jul 16, 2023 at 10:06:52AM +0800, Guiting Shen wrote:
>> The driver would never call clk_enable() if the PWM channel was already
>> enabled in bootloader which lead to dump the warning message "the PWM
>> clock already disabled" when turning off the PWM channel.
>>
>> Add atmel_pwm_enable_clk_if_on() in probe function to enable clock if
>> the PWM channel was already enabled in bootloader.
>>
>> Signed-off-by: Guiting Shen <aarongt.shen@gmail.com>
>> ---
>>  drivers/pwm/pwm-atmel.c | 47 +++++++++++++++++++++++++++++++++++++++--
>>  1 file changed, 45 insertions(+), 2 deletions(-)
>>
>> diff --git a/drivers/pwm/pwm-atmel.c b/drivers/pwm/pwm-atmel.c
>> index cdbc23649032..fc89282db645 100644
>> --- a/drivers/pwm/pwm-atmel.c
>> +++ b/drivers/pwm/pwm-atmel.c
>> @@ -36,7 +36,7 @@
>>  #define PWM_SR			0x0C
>>  #define PWM_ISR			0x1C
>>  /* Bit field in SR */
>> -#define PWM_SR_ALL_CH_ON	0x0F
>> +#define PWM_SR_ALL_CH_MASK	0x0F
>>  
>>  /* The following register is PWM channel related registers */
>>  #define PWM_CH_REG_OFFSET	0x200
>> @@ -464,6 +464,42 @@ static const struct of_device_id atmel_pwm_dt_ids[] = {
>>  };
>>  MODULE_DEVICE_TABLE(of, atmel_pwm_dt_ids);
>>  
>> +static int atmel_pwm_enable_clk_if_on(struct atmel_pwm_chip *atmel_pwm, bool on)
>> +{
>> +	unsigned int i, cnt = 0;
>> +	int ret = 0;
>> +	u32 sr;
>> +
>> +	sr = atmel_pwm_readl(atmel_pwm, PWM_SR) & PWM_SR_ALL_CH_MASK;
>> +	if (!sr)
>> +		return 0;
>> +
>> +	cnt = bitmap_weight((unsigned long *)&sr, atmel_pwm->chip.npwm);
> 
> Tiny nit here: not sure if that cast is safe to do. You've got a 32-bit
> variable, but if you cast &sr to unsigned long * on a 64-bit machine it
> would cause hweight64() to get called and that would then read 64 bits
> from a 32-bit variable. This probably works most of the time because we
> don't read any of the upper bits, but it is strictly an illegal access
> and could be unaligned as well.

I found that the CONFIG_PWM_ATMEL was only in arch/arm/configs directory
which seems that the pwm-atmel driver only run in 32-bit soc machine.

> Should we just turn sr into an unsigned long to be safe here? No need to
> resend, I can make that change when I apply.

Yes, It is safer by doing this. Thank you.

-- 
Regards,
Guiting Shen