Re: [PATCH v2 net 1/2] net: phylink: add lock for serializing concurrent pl->phydev writes with resolver

[Simon Horman <horms@kernel.org>]

On Wed, Sep 03, 2025 at 10:01:45PM +0300, Vladimir Oltean wrote:
> On Wed, Sep 03, 2025 at 07:48:58PM +0100, Simon Horman wrote:
> > On Wed, Sep 03, 2025 at 06:23:47PM +0300, Vladimir Oltean wrote:
> > > @@ -1582,8 +1584,11 @@ static void phylink_resolve(struct work_struct *w)
> > >  	struct phylink_link_state link_state;
> > >  	bool mac_config = false;
> > >  	bool retrigger = false;
> > > +	struct phy_device *phy;
> > >  	bool cur_link_state;
> > >  
> > > +	mutex_lock(&pl->phy_lock);
> > > +	phy = pl->phydev;
> > 
> > Hi Vladimir,
> > 
> > I guess this is an artifact of the development of this patchset.
> > Whatever the case, phy is set but otherwise unused in this function.
> > 
> > This makes CI lightup like a Christmas tree.
> > And it's a bit too early in the year for that.
> 
> Thanks for letting me know. It's an artifact of moving patch 1 in front
> of 2, and I'll address this for the next revision.
> 
> I downgraded to a slower computer for kernel compilation, and even
> though I did compile patch by patch this submission, I had to stop
> building with W=1 C=1 for some unrelated bisect and I forgot to turn
> them back on.
> 
> I don't have a great solution to this, except I'll try next time to set
> up a separate 'git worktree' for noisy stuff like bisection, and try to
> keep the net-next environment separate and always with build warnings
> and debug options enabled.

Understood. It's a tricky problem.

FWIIW, while it's not perfect - e.g. it doesn't exercise linking - building
individual objects does catch problems like this one with low CPU time
requirements.