Re: [PATCH net v1] net: phy: dp83640: fix per-bus clock lifetime

[luoxuanqiang <xuanqiang.luo@linux.dev>]

在 2026/7/29 21:39, Andrew Lunn 写道:
> On Tue, Jul 28, 2026 at 07:03:40PM -0700, Jakub Kicinski wrote:
>> On Fri, 24 Jul 2026 10:35:32 +0800 xuanqiang.luo@linux.dev wrote:
>>> Commit 42e2a9e11a1d ("net: phy: dp83640: improve phydev and driver
>>> removal handling") moved per-bus clock cleanup from module exit to the
>>> remove path. This leaves two lifetime problems.
>> Isn't this sort of shared PHY resource problem supposed to be solved
>> using the PHY package API?
> Ah! Sorry, my reply was completely wrong, now that i have looked at
> the patch.
>
> We have clocks and we have clocks. To me, a clock is something which
> uses the common clock framework, because it is a clock. clk_get,
> clk_enable, clk_disable.
>
> And then we have these PTP things, which some people call clocks, but
> they are something else. And in this patch, we are talking about this
> something else.
>
> And you are correct, it does appear to be one of these PTP things
> which is shared by a number of PHYs. So yes, the PHY package API
> should be used here.
>
> Xuanqiang please take a look at drivers/net/phy/phy_package.c, and the
> PHYs which use it. It should help solve your lifetime problems.
>
>      Andrew
>
> ---
> pw-bot: cr
>
Thanks Jakub and Andrew for pointing this out.

That does look like a much cleaner way to handle the lifetime of the
shared PTP clock.

I'll rework the patch and send a v2.