Re: gcc-8: mm/swapfile.c:863:40: error: array subscript 1 is above array bounds of 'struct list_head[1]' [-Werror=array-bounds]

[Kairui Song]

On Tue, Aug 20, 2024 at 4:51 PM Chris Li <chrisl@xxxxxxxxxx> wrote:
>
> On Mon, Aug 19, 2024 at 10:05 PM Andrew Morton
> <akpm@xxxxxxxxxxxxxxxxxxxx> wrote:
> >
> > On Mon, 19 Aug 2024 19:44:25 +0800 Kairui Song <ryncsn@xxxxxxxxx> wrote:
> >
> > > --- a/mm/swapfile.c
> > > +++ b/mm/swapfile.c
> > > @@ -836,7 +836,7 @@ static unsigned long cluster_alloc_swap_entry(struct swap_info_struct *si, int o
> > >               goto done;
> > >
> > >       /* Order 0 stealing from higher order */
> > > -     for (int o = 1; o < PMD_ORDER; o++) {
> > > +     for (int o = 1; o < SWAP_NR_ORDERS; o++) {
> > >               /*
> > >                * Clusters here have at least one usable slots and can't fail order 0
> > >                * allocation, but reclaim may drop si->lock and race with another user.
> >
> > OK, I got that landed in the right place, but...
> >
> > The definition of `o' within the for statement isn't typical kernel
> > style - I'm surprised we didn't get a warning for this - maybe things
> > have changed when I wasn't looking.
>
> Noted.
>
> I did use the checkpatch.pl and fixed all the warnings before I sent
> the patch out.
> The checkpatch.pl script did not complain about this. Sure I can stay
> away from it.
> BTW, I did  a search on the kernel tree:
> $ rg 'for \(int' | wc -l
> 970
> $
> It seems pretty common in the kernel tree now.

Might be off topic from the issue...

I believe this issue it's not an upstream problem nowadays after
e8c07082a810 ("Kbuild: move to -std=gnu11"), I did notice a GCC error
after backporting these commits to an older kernel which still used
c89, but for upstream this should be OK?

>
> >
> > Also, this code makes no attempt to honor our "The preferred limit on
> > the length of a single line is 80 columns" objective.  There's just no
> > reason for comment blocks to violate this.
>
> I was wondering why the checkpatch.pl did not catch this, is there any
> config for checkpatch.pl I should apply?
>
> I typically invoke:
>
> ./scripts/checkpatch.pl -g HEAD

I found checkpatch.pl stopped checking for 80 columns limit after
commit bdc48fa11e46 ("checkpatch/coding-style: deprecate 80-column
warning") 4 years ago. But the 80 column limit seems still preferred?