Re: [PATCH 06 of 66] alter compound get_page/put_page

[Andrea Arcangeli]

On Thu, Nov 18, 2010 at 12:37:05PM +0000, Mel Gorman wrote:
> On Wed, Nov 03, 2010 at 04:27:41PM +0100, Andrea Arcangeli wrote:
> > From: Andrea Arcangeli <aarcange@xxxxxxxxxx>
> > 
> > Alter compound get_page/put_page to keep references on subpages too, in order
> > to allow __split_huge_page_refcount to split an hugepage even while subpages
> > have been pinned by one of the get_user_pages() variants.
> > 
> > Signed-off-by: Andrea Arcangeli <aarcange@xxxxxxxxxx>
> > Acked-by: Rik van Riel <riel@xxxxxxxxxx>
> > ---
> > 
> > diff --git a/arch/powerpc/mm/gup.c b/arch/powerpc/mm/gup.c
> > --- a/arch/powerpc/mm/gup.c
> > +++ b/arch/powerpc/mm/gup.c
> > @@ -16,6 +16,16 @@
> >  
> >  #ifdef __HAVE_ARCH_PTE_SPECIAL
> >  
> > +static inline void pin_huge_page_tail(struct page *page)
> > +{
> 
> Minor nit, but get_huge_page_tail?
> 
> Even though "pin" is what it does, pin isn't used elsewhere in naming.

Agreed.

diff --git a/arch/powerpc/mm/gup.c b/arch/powerpc/mm/gup.c
--- a/arch/powerpc/mm/gup.c
+++ b/arch/powerpc/mm/gup.c
@@ -16,7 +16,7 @@
 
 #ifdef __HAVE_ARCH_PTE_SPECIAL
 
-static inline void pin_huge_page_tail(struct page *page)
+static inline void get_huge_page_tail(struct page *page)
 {
 	/*
 	 * __split_huge_page_refcount() cannot run
@@ -58,7 +58,7 @@ static noinline int gup_pte_range(pmd_t 
 			return 0;
 		}
 		if (PageTail(page))
-			pin_huge_page_tail(page);
+			get_huge_page_tail(page);
 		pages[*nr] = page;
 		(*nr)++;
 
diff --git a/arch/x86/mm/gup.c b/arch/x86/mm/gup.c
--- a/arch/x86/mm/gup.c
+++ b/arch/x86/mm/gup.c
@@ -105,7 +105,7 @@ static inline void get_head_page_multipl
 	atomic_add(nr, &page->_count);
 }
 
-static inline void pin_huge_page_tail(struct page *page)
+static inline void get_huge_page_tail(struct page *page)
 {
 	/*
 	 * __split_huge_page_refcount() cannot run
@@ -139,7 +139,7 @@ static noinline int gup_huge_pmd(pmd_t p
 		VM_BUG_ON(compound_head(page) != head);
 		pages[*nr] = page;
 		if (PageTail(page))
-			pin_huge_page_tail(page);
+			get_huge_page_tail(page);
 		(*nr)++;
 		page++;
 		refs++;


> > diff --git a/arch/x86/mm/gup.c b/arch/x86/mm/gup.c
> > --- a/arch/x86/mm/gup.c
> > +++ b/arch/x86/mm/gup.c
> > @@ -105,6 +105,16 @@ static inline void get_head_page_multipl
> >  	atomic_add(nr, &page->_count);
> >  }
> >  
> > +static inline void pin_huge_page_tail(struct page *page)
> > +{
> > +	/*
> > +	 * __split_huge_page_refcount() cannot run
> > +	 * from under us.
> > +	 */
> > +	VM_BUG_ON(atomic_read(&page->_count) < 0);
> > +	atomic_inc(&page->_count);
> > +}
> > +
> 
> This is identical to the x86 implementation. Any possibility they can be
> shared?

There is no place for me today to put gup_fast "equal" bits so I'd
need to create it and just doing it for a single inline function
sounds overkill. I could add a asm-generic/gup_fast.h add move that
function there and include the asm-generic/gup_fast.h from
asm*/include/gup_fast.h, is that what we want just for one function?
With all #ifdefs included I would end up writing more code than the
function itself. No problem with me in doing it though.

> >  static noinline int gup_huge_pmd(pmd_t pmd, unsigned long addr,
> >  		unsigned long end, int write, struct page **pages, int *nr)
> >  {
> > @@ -128,6 +138,8 @@ static noinline int gup_huge_pmd(pmd_t p
> >  	do {
> >  		VM_BUG_ON(compound_head(page) != head);
> >  		pages[*nr] = page;
> > +		if (PageTail(page))
> > +			pin_huge_page_tail(page);
> >  		(*nr)++;
> >  		page++;
> >  		refs++;
> > diff --git a/include/linux/mm.h b/include/linux/mm.h
> > --- a/include/linux/mm.h
> > +++ b/include/linux/mm.h
> > @@ -351,9 +351,17 @@ static inline int page_count(struct page
> >  
> >  static inline void get_page(struct page *page)
> >  {
> > -	page = compound_head(page);
> > -	VM_BUG_ON(atomic_read(&page->_count) == 0);
> > +	VM_BUG_ON(atomic_read(&page->_count) < !PageTail(page));
> 
> Oof, this might need a comment. It's saying that getting a normal page or the
> head of a compound page must already have an elevated reference count. If
> we are getting a tail page, the reference count is stored in both the head
> and the tail so the BUG check does not apply.

Ok.

diff --git a/include/linux/mm.h b/include/linux/mm.h
--- a/include/linux/mm.h
+++ b/include/linux/mm.h
@@ -353,8 +353,20 @@ static inline int page_count(struct page
 
 static inline void get_page(struct page *page)
 {
+	/*
+	 * Getting a normal page or the head of a compound page
+	 * requires to already have an elevated page->_count. Only if
+	 * we're getting a tail page, the elevated page->_count is
+	 * required only in the head page, so for tail pages the
+	 * bugcheck only verifies that the page->_count isn't
+	 * negative.
+	 */
 	VM_BUG_ON(atomic_read(&page->_count) < !PageTail(page));
 	atomic_inc(&page->_count);
+	/*
+	 * Getting a tail page will elevate both the head and tail
+	 * page->_count(s).
+	 */
 	if (unlikely(PageTail(page))) {
 		/*
 		 * This is safe only because

> > diff --git a/mm/swap.c b/mm/swap.c
> > --- a/mm/swap.c
> > +++ b/mm/swap.c
> > @@ -56,17 +56,83 @@ static void __page_cache_release(struct 
> >  		del_page_from_lru(zone, page);
> >  		spin_unlock_irqrestore(&zone->lru_lock, flags);
> >  	}
> > +}
> > +
> > +static void __put_single_page(struct page *page)
> > +{
> > +	__page_cache_release(page);
> >  	free_hot_cold_page(page, 0);
> >  }
> >  
> > +static void __put_compound_page(struct page *page)
> > +{
> > +	compound_page_dtor *dtor;
> > +
> > +	__page_cache_release(page);
> > +	dtor = get_compound_page_dtor(page);
> > +	(*dtor)(page);
> > +}
> > +
> >  static void put_compound_page(struct page *page)
> >  {
> > -	page = compound_head(page);
> > -	if (put_page_testzero(page)) {
> > -		compound_page_dtor *dtor;
> > -
> > -		dtor = get_compound_page_dtor(page);
> > -		(*dtor)(page);
> > +	if (unlikely(PageTail(page))) {
> > +		/* __split_huge_page_refcount can run under us */
> 
> So what? The fact you check PageTail twice is a hint as to what is
> happening and that we are depending on the order of when the head and
> tails bits get cleared but it's hard to be certain of that.

This is correct, we depend on the split_huge_page_refcount to clear
PageTail before overwriting first_page.

+	 	/*
+		 * 1) clear PageTail before overwriting first_page
+		 * 2) clear PageTail before clearing PageHead for VM_BUG_ON
+		 */
+		smp_wmb();

The first PageTail check is just to define the fast path and skip
reading page->first_page and doing smb_rmb() for the much more common
head pages (we're in unlikely branch).

So then we read first_page, we smb_rmb() and if pagetail is still set
after that, we can be sure first_page isn't a dangling pointer.

> 
> > +		struct page *page_head = page->first_page;
> > +		smp_rmb();
> > +		if (likely(PageTail(page) && get_page_unless_zero(page_head))) {
> > +			unsigned long flags;
> > +			if (unlikely(!PageHead(page_head))) {
> > +				/* PageHead is cleared after PageTail */
> > +				smp_rmb();
> > +				VM_BUG_ON(PageTail(page));
> > +				goto out_put_head;
> > +			}
> > +			/*
> > +			 * Only run compound_lock on a valid PageHead,
> > +			 * after having it pinned with
> > +			 * get_page_unless_zero() above.
> > +			 */
> > +			smp_mb();
> > +			/* page_head wasn't a dangling pointer */
> > +			compound_lock_irqsave(page_head, &flags);
> > +			if (unlikely(!PageTail(page))) {
> > +				/* __split_huge_page_refcount run before us */
> > +				compound_unlock_irqrestore(page_head, flags);
> > +				VM_BUG_ON(PageHead(page_head));
> > +			out_put_head:
> > +				if (put_page_testzero(page_head))
> > +					__put_single_page(page_head);
> > +			out_put_single:
> > +				if (put_page_testzero(page))
> > +					__put_single_page(page);
> > +				return;
> > +			}
> > +			VM_BUG_ON(page_head != page->first_page);
> > +			/*
> > +			 * We can release the refcount taken by
> > +			 * get_page_unless_zero now that
> > +			 * split_huge_page_refcount is blocked on the
> > +			 * compound_lock.
> > +			 */
> > +			if (put_page_testzero(page_head))
> > +				VM_BUG_ON(1);
> > +			/* __split_huge_page_refcount will wait now */
> > +			VM_BUG_ON(atomic_read(&page->_count) <= 0);
> > +			atomic_dec(&page->_count);
> > +			VM_BUG_ON(atomic_read(&page_head->_count) <= 0);
> > +			compound_unlock_irqrestore(page_head, flags);
> > +			if (put_page_testzero(page_head))
> > +				__put_compound_page(page_head);
> > +		} else {
> > +			/* page_head is a dangling pointer */
> > +			VM_BUG_ON(PageTail(page));
> > +			goto out_put_single;
> > +		}
> > +	} else if (put_page_testzero(page)) {
> > +		if (PageHead(page))
> > +			__put_compound_page(page);
> > +		else
> > +			__put_single_page(page);
> >  	}
> >  }
> >  
> > @@ -75,7 +141,7 @@ void put_page(struct page *page)
> >  	if (unlikely(PageCompound(page)))
> >  		put_compound_page(page);
> >  	else if (put_page_testzero(page))
> > -		__page_cache_release(page);
> > +		__put_single_page(page);
> >  }
> >  EXPORT_SYMBOL(put_page);
> >  
> 
> Functionally, I don't see a problem so
> 
> Acked-by: Mel Gorman <mel@xxxxxxxxx>
> 
> but some expansion on the leader and the comment, even if done as a
> follow-on patch, would be nice.

diff --git a/mm/swap.c b/mm/swap.c
--- a/mm/swap.c
+++ b/mm/swap.c
@@ -79,8 +79,18 @@ static void put_compound_page(struct pag
 		/* __split_huge_page_refcount can run under us */
 		struct page *page_head = page->first_page;
 		smp_rmb();
+		/*
+		 * If PageTail is still set after smp_rmb() we can be sure
+		 * that the page->first_page we read wasn't a dangling pointer.
+		 * See __split_huge_page_refcount() smp_wmb().
+		 */
 		if (likely(PageTail(page) && get_page_unless_zero(page_head))) {
 			unsigned long flags;
+			/*
+			 * Verify that our page_head wasn't converted
+			 * to a a regular page before we got a
+			 * reference on it.
+			 */
 			if (unlikely(!PageHead(page_head))) {
 				/* PageHead is cleared after PageTail */
 				smp_rmb();
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/