Re: [RFC PATCH 1/5] mm: madvise: refactor madvise_populate()

[Lorenzo Stoakes <lorenzo.stoakes@oracle.com>]

On Tue, May 20, 2025 at 12:30:24PM +0200, David Hildenbrand wrote:
> On 19.05.25 22:52, Lorenzo Stoakes wrote:
> > Use a for-loop rather than a while with the update of the start argument at
> > the end of the while-loop.
> >
> > This is in preparation for a subsequent commit which modifies this
> > function, we therefore separate the refactoring from the actual change
> > cleanly by separating the two.
> >
> > Signed-off-by: Lorenzo Stoakes <lorenzo.stoakes@oracle.com>
> > ---
> >   mm/madvise.c | 39 ++++++++++++++++++++-------------------
> >   1 file changed, 20 insertions(+), 19 deletions(-)
> >
> > diff --git a/mm/madvise.c b/mm/madvise.c
> > index 8433ac9b27e0..63cc69daa4c7 100644
> > --- a/mm/madvise.c
> > +++ b/mm/madvise.c
> > @@ -967,32 +967,33 @@ static long madvise_populate(struct mm_struct *mm, unsigned long start,
> >   	int locked = 1;
> >   	long pages;
> > -	while (start < end) {
> > +	for (; start < end; start += pages * PAGE_SIZE) {
> >   		/* Populate (prefault) page tables readable/writable. */
> >   		pages = faultin_page_range(mm, start, end, write, &locked);
> >   		if (!locked) {
> >   			mmap_read_lock(mm);
> >   			locked = 1;
> >   		}
> > -		if (pages < 0) {
> > -			switch (pages) {
> > -			case -EINTR:
> > -				return -EINTR;
> > -			case -EINVAL: /* Incompatible mappings / permissions. */
> > -				return -EINVAL;
> > -			case -EHWPOISON:
> > -				return -EHWPOISON;
> > -			case -EFAULT: /* VM_FAULT_SIGBUS or VM_FAULT_SIGSEGV */
> > -				return -EFAULT;
> > -			default:
> > -				pr_warn_once("%s: unhandled return value: %ld\n",
> > -					     __func__, pages);
> > -				fallthrough;
> > -			case -ENOMEM: /* No VMA or out of memory. */
> > -				return -ENOMEM;
> > -			}
> > +
> > +		if (pages >= 0)
> > +			continue;
> > +
> > +		switch (pages) {
> > +		case -EINTR:
> > +			return -EINTR;
> > +		case -EINVAL: /* Incompatible mappings / permissions. */
> > +			return -EINVAL;
> > +		case -EHWPOISON:
> > +			return -EHWPOISON;
> > +		case -EFAULT: /* VM_FAULT_SIGBUS or VM_FAULT_SIGSEGV */
> > +			return -EFAULT;
> > +		default:
> > +			pr_warn_once("%s: unhandled return value: %ld\n",
> > +				     __func__, pages);
> > +			fallthrough;
> > +		case -ENOMEM: /* No VMA or out of memory. */
> > +			return -ENOMEM;
>
> Can we limit it to what the patch description says? "Use a for-loop rather
> than a while", or will that be a problem for the follow-up patch?

Well, kind of the point is that we can remove a level of indentation also, which
then makes life easier in subsequent patch.

Happy to change description or break into two (but that seems a bit over the top
maybe? :>)

Idea is that we clearly separate out the refactoring bit from the actual change
to the logic so it's not a pain to bisect/review.

>
> --
> Cheers,
>
> David / dhildenb
>