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;
Date:   Sun, 28 May 2023 18:49:35 +0100
From:   Jonathan Cameron <jic23@kernel.org>
To:     Artur Rojek <contact@artur-rojek.eu>
Cc:     Paul Cercueil <paul@crapouillou.net>,
        Dmitry Torokhov <dmitry.torokhov@gmail.com>,
        Chris Morgan <macromorgan@hotmail.com>,
        Andy Shevchenko <andy.shevchenko@gmail.com>,
        linux-mips@vger.kernel.org, linux-iio@vger.kernel.org,
        linux-kernel@vger.kernel.org, linux-input@vger.kernel.org
Subject: Re: [PATCH v2 1/2] iio/adc: ingenic: Fix channel offsets in buffer
Message-ID: <20230528184935.15dd91fa@jic23-huawei>
In-Reply-To: <20230521225901.388455-2-contact@artur-rojek.eu>
References: <20230521225901.388455-1-contact@artur-rojek.eu>
        <20230521225901.388455-2-contact@artur-rojek.eu>
MIME-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit
Precedence: bulk

On Mon, 22 May 2023 00:59:00 +0200
Artur Rojek <contact@artur-rojek.eu> wrote:

> Consumers expect the buffer to only contain enabled channels. While
> preparing the buffer, the driver makes two mistakes:
> 1) It inserts empty data for disabled channels.
> 2) Each ADC readout contains samples for two 16-bit channels. If either
>    of them is active, the whole 32-bit sample is pushed into the buffer
>    as-is.
> 
> Both of those issues cause the active channels to appear at the wrong
> offsets in the buffer. Fix the above by demuxing samples for active
> channels only.
> 
> This has remained unnoticed, as all the consumers so far were only using
> channels 0 and 1, leaving them unaffected by changes introduced in this
> commit.
> 
> Signed-off-by: Artur Rojek <contact@artur-rojek.eu>
> Tested-by: Paul Cercueil <paul@crapouillou.net>

Lazy me suggestions that, as we didn't notice this before, clearly the
vast majority of times the channels are both enabled.
As such you 'could' just set available_scan_masks and burn the overhead
of reading channels you don't want, instead letting the IIO core demux
deal with the data movement if needed.

> ---
> 
> v2: - demux active channels from ADC readouts 
>     - clarify in the commit description that this patch doesn't impact
>       existing consumers of this driver
> 
>  drivers/iio/adc/ingenic-adc.c | 20 +++++++++++++-------
>  1 file changed, 13 insertions(+), 7 deletions(-)
> 
> diff --git a/drivers/iio/adc/ingenic-adc.c b/drivers/iio/adc/ingenic-adc.c
> index a7325dbbb99a..093710a7ad4c 100644
> --- a/drivers/iio/adc/ingenic-adc.c
> +++ b/drivers/iio/adc/ingenic-adc.c
> @@ -802,13 +802,19 @@ static irqreturn_t ingenic_adc_irq(int irq, void *data)
>  	struct ingenic_adc *adc = iio_priv(iio_dev);
>  	unsigned long mask = iio_dev->active_scan_mask[0];
>  	unsigned int i;
> -	u32 tdat[3];
> -
> -	for (i = 0; i < ARRAY_SIZE(tdat); mask >>= 2, i++) {
> -		if (mask & 0x3)
> -			tdat[i] = readl(adc->base + JZ_ADC_REG_ADTCH);
> -		else
> -			tdat[i] = 0;
> +	u16 tdat[6];
> +	u32 val;
> +
> +	memset(tdat, 0, ARRAY_SIZE(tdat));
> +	for (i = 0; mask && i < ARRAY_SIZE(tdat); mask >>= 2) {
> +		if (mask & 0x3) {
> +			val = readl(adc->base + JZ_ADC_REG_ADTCH);
> +			/* Two channels per sample. Demux active. */
> +			if (mask & BIT(0))
> +				tdat[i++] = val & 0xffff;
> +			if (mask & BIT(1))
> +				tdat[i++] = val >> 16;
> +		}
>  	}
>  
>  	iio_push_to_buffers(iio_dev, tdat);