Re: [PATCH v1 02/29] mm/balloon_compaction: convert balloon_page_delete() to balloon_page_finalize()

[Lorenzo Stoakes]

On Tue, Jul 01, 2025 at 09:58:09AM +0200, David Hildenbrand wrote:
> On 30.06.25 17:15, Lorenzo Stoakes wrote:
> > On Mon, Jun 30, 2025 at 02:59:43PM +0200, David Hildenbrand wrote:
> > > Let's move the removal of the page from the balloon list into the single
> > > caller, to remove the dependency on the PG_isolated flag and clarify
> > > locking requirements.
> > >
> > > We'll shuffle the operations a bit such that they logically make more sense
> > > (e.g., remove from the list before clearing flags).
> > >
> > > In balloon migration functions we can now move the balloon_page_finalize()
> > > out of the balloon lock and perform the finalization just before dropping
> > > the balloon reference.
> > >
> > > Document that the page lock is currently required when modifying the
> > > movability aspects of a page; hopefully we can soon decouple this from the
> > > page lock.
> > >
> > > Signed-off-by: David Hildenbrand <david@xxxxxxxxxx>

Based on below this LGTM, so:

Reviewed-by: Lorenzo Stoakes <lorenzo.stoakes@xxxxxxxxxx>

> > > ---
> > >   arch/powerpc/platforms/pseries/cmm.c |  2 +-
> > >   drivers/misc/vmw_balloon.c           |  3 +-
> > >   drivers/virtio/virtio_balloon.c      |  4 +--
> > >   include/linux/balloon_compaction.h   | 43 +++++++++++-----------------
> > >   mm/balloon_compaction.c              |  3 +-
> > >   5 files changed, 21 insertions(+), 34 deletions(-)
> > >
> > > diff --git a/arch/powerpc/platforms/pseries/cmm.c b/arch/powerpc/platforms/pseries/cmm.c
> > > index 5f4037c1d7fe8..5e0a718d1be7b 100644
> > > --- a/arch/powerpc/platforms/pseries/cmm.c
> > > +++ b/arch/powerpc/platforms/pseries/cmm.c
> > > @@ -532,7 +532,6 @@ static int cmm_migratepage(struct balloon_dev_info *b_dev_info,
> > >
> > >   	spin_lock_irqsave(&b_dev_info->pages_lock, flags);
> > >   	balloon_page_insert(b_dev_info, newpage);
> > > -	balloon_page_delete(page);
> >
>
> Hi Lorenzo,
>
> as always, thanks for the detailed review!

You're welcome :>)

>
> > We seem to just be removing this and not replacing with finalize, is this right?
>
> See below.
>
> >
> > >   	b_dev_info->isolated_pages--;
> > >   	spin_unlock_irqrestore(&b_dev_info->pages_lock, flags);
> > >
> > > @@ -542,6 +541,7 @@ static int cmm_migratepage(struct balloon_dev_info *b_dev_info,
> > >   	 */
> > >   	plpar_page_set_active(page);
> > >
> > > +	balloon_page_finalize(page);
>
> ^ here it is, next to the put_page() just like for the other cases.

OK so it's just moved to a different place for consistency.

>
> Or did you mean something else?

No this is what I meant :)

>
> > >   	/* balloon page list reference */
> > >   	put_page(page);
> > >
> > > diff --git a/drivers/misc/vmw_balloon.c b/drivers/misc/vmw_balloon.c
> > > index c817d8c216413..6653fc53c951c 100644
> > > --- a/drivers/misc/vmw_balloon.c
> > > +++ b/drivers/misc/vmw_balloon.c
> > > @@ -1778,8 +1778,7 @@ static int vmballoon_migratepage(struct balloon_dev_info *b_dev_info,
> > >   	 * @pages_lock . We keep holding @comm_lock since we will need it in a
> > >   	 * second.
> > >   	 */
> > > -	balloon_page_delete(page);
> > > -
> > > +	balloon_page_finalize(page);
> > >   	put_page(page);
> > >
>
>
> [...]
>
> > > -/*
> > > - * balloon_page_delete - delete a page from balloon's page list and clear
> > > - *			 the page->private assignement accordingly.
> > > - * @page    : page to be released from balloon's page list
> > > - *
> > > - * Caller must ensure the page is locked and the spin_lock protecting balloon
> > > - * pages list is held before deleting a page from the balloon device.
> > > - */
> > > -static inline void balloon_page_delete(struct page *page)
> > > -{
> > > -	__ClearPageOffline(page);
> > > -	__ClearPageMovable(page);
> > > -	set_page_private(page, 0);
> > > -	/*
> > > -	 * No touch page.lru field once @page has been isolated
> > > -	 * because VM is using the field.
> > > -	 */
> > > -	if (!PageIsolated(page))
> > > -		list_del(&page->lru);
> >
> > I don't see this check elsewhere, is it because, as per the 1/xx of this series,
> > because by the time we do the finalize
>
> balloon_page_delete() was used on two paths
>
> 1) Removing a page from the balloon for deflation through
> balloon_page_list_dequeue()
>
> 2) Removing an isolated page from the balloon for migration in the
> per-driver migration handlers. Isolated pages were already removed from the
> balloon list during ... isolation.
>
> With this change, 1) does the list_del(&page->lru) manually and 2) only
> calls balloon_page_finalize().
>
> During 1) the same reasoning as in 1/xx applies: isolated pages cannot be in
> the balloon list.

Right yeah this is what I thought, thanks!

>
> >
> > > -}
> > > -
> > >   /*
> > >    * balloon_page_device - get the b_dev_info descriptor for the balloon device
> > >    *			 that enqueues the given page.
> > > @@ -141,12 +120,6 @@ static inline void balloon_page_insert(struct balloon_dev_info *balloon,
> > >   	list_add(&page->lru, &balloon->pages);
> > >   }
> > >
> > > -static inline void balloon_page_delete(struct page *page)
> > > -{
> > > -	__ClearPageOffline(page);
> > > -	list_del(&page->lru);
> > > -}
> > > -
> > >   static inline gfp_t balloon_mapping_gfp_mask(void)
> > >   {
> > >   	return GFP_HIGHUSER;
> > > @@ -154,6 +127,22 @@ static inline gfp_t balloon_mapping_gfp_mask(void)
> > >
> > >   #endif /* CONFIG_BALLOON_COMPACTION */
> > >
> > > +/*
> > > + * balloon_page_finalize - prepare a balloon page that was removed from the
> > > + *			   balloon list for release to the page allocator
> > > + * @page: page to be released to the page allocator
> > > + *
> > > + * Caller must ensure that the page is locked.
> >
> > Can we assert this?
>
> We could, but I'm planning on removing the page lock next (see patch
> description), so not too keen to create more code around that.
>
> Maybe mention that the balloon lock should not be held?
>
> Not a limitation. It could be called with it, just not a requirement today.
>
> I suspect that once we remove the page lock, that we might use the balloon
> lock and rework balloon_page_migrate() to take the lock. TBD.

OK fair enough!

>
> > >> + */
> > > +static inline void balloon_page_finalize(struct page *page)
> > > +{
> > > +	if (IS_ENABLED(CONFIG_BALLOON_COMPACTION)) {
> > > +		__ClearPageMovable(page);
> > > +		set_page_private(page, 0);
> > > +	}
> >
> > Why do we check this? Is this function called from anywhere where that config won't be set?
>
> Sure. balloon_page_list_dequeue() is called from balloon_page_dequeue(),
> which resides outside the CONFIG_BALLOON_COMPACTION ifdef in
> mm/balloon_compaction.c.
>
> At some point (not in this series) we should probably rename
>
> balloon_compaction.c -> balloon.c
>
> To match CONFIG_MEMORY_BALLOON.
>
> Because the compaction part is just one extra bit in there. (an important
> one, but still, you can use the balloon infrastructure without
> compaction/page migration)

Yeah this would be nice!