Re: [PATCH] irqdomain: do not shadow nr_irqs global definition

[Vincent MAILHOL]

On Fri. 29 Jul. 2022 at 20:12, Vincent MAILHOL
<mailhol.vincent@xxxxxxxxxx> wrote:
> On Fri. 29 juil. 2022 at 18:16, Marc Zyngier <maz@xxxxxxxxxx> wrote:
> > On 2022-07-29 10:03, Vincent Mailhol wrote:
> > > include/linux/irqnr.h declares a global variable named `nr_irqs'.
> > >
> > > The inline function irq_domain_alloc_irqs() from
> > > include/linux/irqdomain.h also uses the same name and thus shadow the
> > > global declaration. Rename the function argument from `nr_irqs' to
> > > `nbr_irqs'.
> > >
> > > This patch silences below -Wshadow warning:
> > >
> > > | In file included from ./arch/x86/include/asm/irqdomain.h:5,
> > > |                  from ./arch/x86/include/asm/msi.h:5,
> > > |                  from ./include/linux/msi.h:23,
> > > |                  from ./include/linux/kvm_host.h:19,
> > > |                  from arch/x86/kernel/../kvm/vmx/vmx.h:5,
> > > |                  from arch/x86/kernel/asm-offsets.c:22:
> > > | ./include/linux/irqdomain.h: In function 'irq_domain_alloc_irqs':
> > > | ./include/linux/irqdomain.h:514:38: warning: declaration of
> > > 'nr_irqs' shadows a global declaration [-Wshadow]
> > > |   514 |                         unsigned int nr_irqs, int node, void
> > > *arg)
> > > |       |                         ~~~~~~~~~~~~~^~~~~~~
> > > | In file included from ./include/linux/interrupt.h:10,
> > > |                  from ./include/linux/kernel_stat.h:9,
> > > |                  from ./include/linux/cgroup.h:26,
> > > |                  from ./include/linux/memcontrol.h:13,
> > > |                  from ./include/linux/swap.h:9,
> > > |                  from ./include/linux/suspend.h:5,
> > > |                  from arch/x86/kernel/asm-offsets.c:13:
> > > | ./include/linux/irqnr.h:8:12: note: shadowed declaration is here
> > > |     8 | extern int nr_irqs;
> > > |       |            ^~~~~~~
> > >
> > > Signed-off-by: Vincent Mailhol <mailhol.vincent@xxxxxxxxxx>
> > > ---
> > >  include/linux/irqdomain.h | 4 ++--
> > >  1 file changed, 2 insertions(+), 2 deletions(-)
> > >
> > > diff --git a/include/linux/irqdomain.h b/include/linux/irqdomain.h
> > > index 00d577f90883..76949bb029b0 100644
> > > --- a/include/linux/irqdomain.h
> > > +++ b/include/linux/irqdomain.h
> > > @@ -511,9 +511,9 @@ extern int irq_domain_activate_irq(struct irq_data
> > > *irq_data, bool early);
> > >  extern void irq_domain_deactivate_irq(struct irq_data *irq_data);
> > >
> > >  static inline int irq_domain_alloc_irqs(struct irq_domain *domain,
> > > -                     unsigned int nr_irqs, int node, void *arg)
> > > +                                     unsigned int nbr_irqs, int node, void *arg)
> > >  {
> > > -     return __irq_domain_alloc_irqs(domain, -1, nr_irqs, node, arg, false,
> > > +     return __irq_domain_alloc_irqs(domain, -1, nbr_irqs, node, arg,
> > > false,
> > >                                      NULL);
> > >  }
> >
> > I really don't think this is worth it. A function has its
> > own namespace, and this warning is on the long list of
> > "this is completely silly". Case in point:
> >
> > $ git grep 'unsigned int nr_irqs'| wc -l
> > 207
> >
> > Is anything broken? Not as far as I can tell.
>
> Nothing is broken, my only concern is that this occurs in a header
> file and thus the warning will pop up in random files which include
> this header.
>
> | git grep "unsigned int nr_irqs" include/ arch/*/include | wc -l
> | 17
> (and all static inline function which trigger the warning are from
> linux/irqdomain.h)
>
> Time to time, I check the W=2 output and spam from the headers makes
> it annoying to triage. I absolutely do not care of shadowing
> everywhere else outside of the includes directory.
>
> > If there was anything to fix, it is the top-level definition
> > that should be more indicative of its global status. But again,
> > there is nothing broken so far.
>
> This is also feasible, but it will become a tree-wide patch. And I am
> quite certain that it will be NACKed right away, so I will not try
> this alternate solution.

To add some figures, on a "make allyesconfig" (linux-next branch,
x86_64 build, gcc 12), there are in total 36005 -Wshadow and
irqdomain.h accounts for 7094 of all these (i.e. this patch suppresses
roughly one fifth of all -Wshadow warnings tree wide).


Yours sincerely,
Vincent Mailhol