Re: [PATCH v6 02/27] of: dynamic: Fix overlayed devices not probing because of fw_devlink

[Herve Codina]

Hi Geert, Doug,

On Thu, 23 Apr 2026 08:53:33 +0200
Geert Uytterhoeven <geert@xxxxxxxxxxxxxx> wrote:

> Hi Doug,
> 
> On Wed, 22 Apr 2026 at 21:40, Doug Anderson <dianders@xxxxxxxxxxxx> wrote:
> > On Wed, Apr 22, 2026 at 12:27 PM Andy Shevchenko
> > <andriy.shevchenko@xxxxxxxxxxxxxxx> wrote:  
> > > On Wed, Apr 22, 2026 at 11:51:36AM -0700, Doug Anderson wrote:  
> > > > On Wed, Apr 22, 2026 at 10:44 AM Geert Uytterhoeven
> > > > <geert@xxxxxxxxxxxxxx> wrote:  
> > > > > On Wed, 25 Mar 2026 at 15:36, Herve Codina <herve.codina@xxxxxxxxxxx> wrote:  
> > >
> > > ...
> > >  
> > > > > > +               if (fwnode->flags & FWNODE_FLAG_NOT_DEVICE)  
> > > > >
> > > > > After commit f72e77c33e4b5657 ("device property: Make modifications
> > > > > of fwnode "flags" thread safe"), this must be changed to:
> > > > >
> > > > >     if (fwnode_test_flag(fwnode, FWNODE_FLAG_NOT_DEVICE))
> > > > >
> > > > > With this change my DT overlays are working again.
> > > > >
> > > > > I guess keeping the FWNODE_FLAG_* names, but changing their values
> > > > > from bit masks to bit numbers was probably not such a good idea,
> > > > > as it fails to catch missing conversions...  
> > > >
> > > > Crud, I should have thought about that. :( Do you think it's worth it
> > > > to do a rename at this point to catch future problems?  
> 
> I am afraid it is too late (it is already in Linus' tree), and would
> add only more to the confusion.  We just have to take care when this
> patch (and 01/27) is backported to stable.  Fortunately the reverse
> is caught, as the fwnode_*_flag() helpers did not exist before.
> 
> > > You can try locally with allyesconfig / allmodconfig and see the outcome.
> > > If there are missed cases, and their amount reasonably low, I won't bother
> > > in this case.  
> >
> > I believe I caught all of them at the time of my CL, but the problem
> > is in-flight stuff, right? I think the problem Geert is pointing out
> > isn't something I missed in my original patch but rather an
> > interaction between my patch and this in-flight patch.  
> 
> There are no unconverted cases in upstream or linux-next
> (I checked yesterday, and rechecked this morning).
> 

Thanks for having pointed out this modification.
It will save some debug hours on my side when I test the next iteration. 

For this in-flight series, I will use new accessors introduced in
f72e77c33e4b5657 ("device property: Make modifications of fwnode "flags"
thread safe")

Best regards,
Hervé