Re: [PATCH] mm/vmalloc: avoid bogus -Wmaybe-uninitialized warning

[Andrew Morton]

On Tue, 18 Jun 2019 16:06:22 +0200 Uladzislau Rezki <urezki@xxxxxxxxx> wrote:

> On Tue, Jun 18, 2019 at 09:40:28AM -0400, Joel Fernandes wrote:
> > On Tue, Jun 18, 2019 at 5:27 AM Arnd Bergmann <arnd@xxxxxxxx> wrote:
> > >
> > > gcc gets confused in pcpu_get_vm_areas() because there are too many
> > > branches that affect whether 'lva' was initialized before it gets
> > > used:
> > >
> > > mm/vmalloc.c: In function 'pcpu_get_vm_areas':
> > > mm/vmalloc.c:991:4: error: 'lva' may be used uninitialized in this function [-Werror=maybe-uninitialized]
> > >     insert_vmap_area_augment(lva, &va->rb_node,
> > >     ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> > >      &free_vmap_area_root, &free_vmap_area_list);
> > >      ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> > > mm/vmalloc.c:916:20: note: 'lva' was declared here
> > >   struct vmap_area *lva;
> > >                     ^~~
> > >
> > > Add an intialization to NULL, and check whether this has changed
> > > before the first use.
> > >
> > > Fixes: 68ad4a330433 ("mm/vmalloc.c: keep track of free blocks for vmap allocation")
> > > Signed-off-by: Arnd Bergmann <arnd@xxxxxxxx>
> > > ---
> > >  mm/vmalloc.c | 9 +++++++--
> > >  1 file changed, 7 insertions(+), 2 deletions(-)
> > >
> > > diff --git a/mm/vmalloc.c b/mm/vmalloc.c
> > > index a9213fc3802d..42a6f795c3ee 100644
> > > --- a/mm/vmalloc.c
> > > +++ b/mm/vmalloc.c
> > > @@ -913,7 +913,12 @@ adjust_va_to_fit_type(struct vmap_area *va,
> > >         unsigned long nva_start_addr, unsigned long size,
> > >         enum fit_type type)
> > >  {
> > > -       struct vmap_area *lva;
> > > +       /*
> > > +        * GCC cannot always keep track of whether this variable
> > > +        * was initialized across many branches, therefore set
> > > +        * it NULL here to avoid a warning.
> > > +        */
> > > +       struct vmap_area *lva = NULL;
> > 
> > Fair enough, but is this 5-line comment really needed here?
> > 
> How it is rewritten now, probably not. I would just set it NULL and
> leave the comment, but that is IMHO. Anyway
> 

I agree - given that the patch does this:

@@ -972,7 +977,7 @@ adjust_va_to_fit_type(struct vmap_area *
 	if (type != FL_FIT_TYPE) {
 		augment_tree_propagate_from(va);
 
-		if (type == NE_FIT_TYPE)
+		if (lva)
 			insert_vmap_area_augment(lva, &va->rb_node,
 				&free_vmap_area_root, &free_vmap_area_list);
 	}

the comment simply isn't relevant any more.  Although I guess this
might be a bit helpful:

@@ -977,7 +972,7 @@ adjust_va_to_fit_type(struct vmap_area *
 	if (type != FL_FIT_TYPE) {
 		augment_tree_propagate_from(va);
 
-		if (lva)
+		if (lva)	/* type == NE_FIT_TYPE */
 			insert_vmap_area_augment(lva, &va->rb_node,
 				&free_vmap_area_root, &free_vmap_area_list);
 	}