RE: [PATCH net-next v3 5/5] net: phy: Move phy_init_hw() from phy_resume() to __phy_resume()

[Biju Das]

Hi Andrew,

> -----Original Message-----
> From: Andrew Lunn <andrew@xxxxxxx>
> Sent: 14 April 2026 17:03
> Subject: Re: [PATCH net-next v3 5/5] net: phy: Move phy_init_hw() from phy_resume() to __phy_resume()
> 
> On Sun, Apr 12, 2026 at 03:00:27PM +0100, Biju wrote:
> > From: Biju Das <biju.das.jz@xxxxxxxxxxxxxx>
> >
> > Now that redundant locking has been removed from PHY driver callbacks,
> > phy_init_hw() can be called with phydev->lock held.
> >
> > Many MAC drivers and the phylink framework resume the PHY via
> > phy_start(), which invokes __phy_resume() directly without going
> > through phy_resume(). Keeping phy_init_hw() in phy_resume() means it
> > is not called in this path.
> >
> > Move phy_init_hw() into __phy_resume() so that PHY soft reset and
> > re-initialisation happen unconditionally on every resume, regardless
> > of which code path triggers it.
> 
> I would change the order of these patches. First remove the redundant locks. You can then put
> phy_init_hw() into __phy_resume(), rather than first moving it into phy_resume() and then
> __phy_resume().

Agreed.

Cheers,
Biju