On Tue, Jul 01, 2025 at 12:19:47PM +0200, David Hildenbrand wrote:
> On 01.07.25 12:03, Lorenzo Stoakes wrote:
> > On Mon, Jun 30, 2025 at 02:59:54PM +0200, David Hildenbrand wrote:
> > > We can just look at the balloon device (stored in page->private), to see
> > > if the page is still part of the balloon.
> > >
> > > As isolated balloon pages cannot get released (they are taken off the
> > > balloon list while isolated), we don't have to worry about this case in
> > > the putback and migration callback. Add a WARN_ON_ONCE for now.
> > >
> > > Signed-off-by: David Hildenbrand <david@redhat.com>
> >
> > Seems reasonable, one comment below re: comment.
> >
> > Reviewed-by: Lorenzo Stoakes <lorenzo.stoakes@oracle.com>
> >
> > > ---
> > > include/linux/balloon_compaction.h | 4 +---
> > > mm/balloon_compaction.c | 11 +++++++++++
> > > 2 files changed, 12 insertions(+), 3 deletions(-)
> > >
> > > diff --git a/include/linux/balloon_compaction.h b/include/linux/balloon_compaction.h
> > > index bfc6e50bd004b..9bce8e9f5018c 100644
> > > --- a/include/linux/balloon_compaction.h
> > > +++ b/include/linux/balloon_compaction.h
> > > @@ -136,10 +136,8 @@ static inline gfp_t balloon_mapping_gfp_mask(void)
> > > */
> > > static inline void balloon_page_finalize(struct page *page)
> > > {
> > > - if (IS_ENABLED(CONFIG_BALLOON_COMPACTION)) {
> > > - __ClearPageMovable(page);
> > > + if (IS_ENABLED(CONFIG_BALLOON_COMPACTION))
> > > set_page_private(page, 0);
> > > - }
> > > /* PageOffline is sticky until the page is freed to the buddy. */
> > > }
> > >
> > > diff --git a/mm/balloon_compaction.c b/mm/balloon_compaction.c
> > > index ec176bdb8a78b..e4f1a122d786b 100644
> > > --- a/mm/balloon_compaction.c
> > > +++ b/mm/balloon_compaction.c
> > > @@ -206,6 +206,9 @@ static bool balloon_page_isolate(struct page *page, isolate_mode_t mode)
> > > struct balloon_dev_info *b_dev_info = balloon_page_device(page);
> > > unsigned long flags;
> > >
> > > + if (!b_dev_info)
> > > + return false;
> > > +
> > > spin_lock_irqsave(&b_dev_info->pages_lock, flags);
> > > list_del(&page->lru);
> > > b_dev_info->isolated_pages++;
> > > @@ -219,6 +222,10 @@ static void balloon_page_putback(struct page *page)
> > > struct balloon_dev_info *b_dev_info = balloon_page_device(page);
> > > unsigned long flags;
> > >
> > > + /* Isolated balloon pages cannot get deflated. */
> > > + if (WARN_ON_ONCE(!b_dev_info))
> > > + return;
> > > +
> > > spin_lock_irqsave(&b_dev_info->pages_lock, flags);
> > > list_add(&page->lru, &b_dev_info->pages);
> > > b_dev_info->isolated_pages--;
> > > @@ -234,6 +241,10 @@ static int balloon_page_migrate(struct page *newpage, struct page *page,
> > > VM_BUG_ON_PAGE(!PageLocked(page), page);
> > > VM_BUG_ON_PAGE(!PageLocked(newpage), newpage);
> > >
> > > + /* Isolated balloon pages cannot get deflated. */
> >
> > Hm do you mean migrated?
>
> Well, they can get migrated, obviously :)
Right yeah we isolate to migrate :P
I guess I was confused by the 'get' deflated, wrongly thinking putback was doing
this (I got confused about terminology), but I see in balloon_page_dequeue() we
balloon_page_finalize() which sets page->private = NULL which is what
balloon_page_device() returns.
OK I guess this is fine... :)
An aside, unrelated tot his series: it'd be nice to use 'deflate' consistently
in this code. We do __count_vm_event(BALLOON_DEFLATE) in
balloon_page_list_dequeue() but say 'deflate' nowhere else... well before this
patch :)
>
> Deflation would be the other code path where we would remove a balloon page
> from the balloon, and invalidate page->private, suddenly seeing !b_dev_info
> here.
ACtually it's 'balloon' not b_dev_info. Kind of out of scope for this patch but
would be good to rename.
>
> But that cannot happen, as isolation takes them off the balloon list. So
> deflating the balloon cannot find them until un-isolated.
Ack.
>
> --
> Cheers,
>
> David / dhildenb
>