Re: [PATCH v4 5/5] arm64: dts: imx8mq-librem5: Correct link frequency list

[sashiko-bot@kernel.org]

Thank you for your contribution! Sashiko AI review found 2 potential issue(s) to consider:
- [High] The exact link frequencies of 187500000 and 375000000 will fail the exact-match validation in the hi846 driver, causing the sensor probe to fail with -EINVAL.
- [Low] The arithmetic formula provided in the commit message is mathematically incorrect and misrepresents the driver's internal logic.
--

commit 1a95e1bc9a680037635b3deded93622acb02b6ed
Author: Pengyu Luo <mitltlatltl@gmail.com>

arm64: dts: imx8mq-librem5: Correct link frequency list

This commit updates the link frequencies for the hi846 camera sensor on the
imx8mq-librem5 board to match the expected values calculated from the mclk.

>     According to the hi846 datasheet and the register lists, they are
> 
>     mclk / prediv * multiplier / post_div1 / post_div2 =
>     25Mhz / 3 * 90 / 1 / {1,2} = 187.5Mhz, 375Mhz

This isn't a bug, but is the arithmetic in the commit message correct?

Mathematically, 25 / 3 * 90 is exactly 750. Dividing 750 by 1 and then by 1
or 2 results in 750 MHz and 375 MHz. To arrive at 187.5 MHz, the division
should be / 2 / 2, which corresponds to post_div1 = 2 and post_div2 = 2 in
the driver logic.

> diff --git a/arch/arm64/boot/dts/freescale/imx8mq-librem5.dtsi b/arch/arm64/boot/dts/freescale/imx8mq-librem5.dtsi
> index eee390c272105..959a5c8dcd0d3 100644
> --- a/arch/arm64/boot/dts/freescale/imx8mq-librem5.dtsi
> +++ b/arch/arm64/boot/dts/freescale/imx8mq-librem5.dtsi
> @@ -1122,7 +1122,7 @@ port {
>  			camera1_ep: endpoint {
>  				data-lanes = <1 2>;
>  				link-frequencies = /bits/ 64
> -					<80000000 200000000 300000000>;
> +					<187500000 375000000>;

Does this regression cause the sensor probe to fail with -EINVAL?

The hi846 driver dynamically calculates expected link frequencies from mclk
using the C expression: mclk / 3 * 90 / post_div1 / post_div2.

In C, operators with the same precedence are evaluated left-to-right. With
mclk configured to 25,000,000 Hz, the division 25000000 / 3 occurs first
and truncates to 8,333,333.

Subsequent multiplication by 90 and division yields 187,499,992 and
374,999,984 instead of exactly 187,500,000 and 375,000,000.

Because 187500000 != 187499992, the exact equality check in the driver:
    if (link_freq == ep->link_frequencies[j])
will fail.

This rejects these mathematically ideal frequencies and aborts the probe,
breaking camera initialization.

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260511103927.279550-1-mitltlatltl@gmail.com?part=5