Re: [PATCH net v2] net: macb: fix ordering around PTP timestamp read

[Jakub Kicinski]

On Thu, 10 Sep 2026 11:06:52 +0700 James Clark wrote:
> +	/* ptp_read_system_prets() uses smp_rmb() internally,
> +	 * which does not guarantee ordering against MMIO reads.
> +	 */

IMHO the fact that the gem_readl() is a relaxed read is more relevant /
unusual in this case?

Also clashiko points out that this function may be called per packet
with sts as NULL so it's worth wrapping these barriers in if, I think
that's a good suggestion?

https://netdev-ai.bots.linux.dev/sashiko/#/patchset/20260910040658.18359-1-jjc@xxxxxxxxxx

> +	rmb();
>  	first = gem_readl(bp, TN);
> +	/* Ensure the PHC read completes before taking
> +	 * the post timestamp.
> +	 */
> +	rmb();
-- 
pw-bot: cr