diff mbox series

[v2,10/13] mm: Free up a word in the first tail page

Message ID 20230816151201.3655946-11-willy@infradead.org (mailing list archive)
State New
Headers show
Series Remove _folio_dtor and _folio_order | expand

Commit Message

Matthew Wilcox Aug. 16, 2023, 3:11 p.m. UTC
Store the folio order in the low byte of the flags word in the first
tail page.  This frees up the word that was being used to store the
order and dtor bytes previously.

Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
---
 include/linux/mm.h         | 10 +++++-----
 include/linux/mm_types.h   |  3 +--
 include/linux/page-flags.h |  7 ++++---
 kernel/crash_core.c        |  1 -
 mm/internal.h              |  2 +-
 5 files changed, 11 insertions(+), 12 deletions(-)

Comments

Mike Kravetz Aug. 22, 2023, 11:17 p.m. UTC | #1
On 08/16/23 16:11, Matthew Wilcox (Oracle) wrote:
> Store the folio order in the low byte of the flags word in the first
> tail page.  This frees up the word that was being used to store the
> order and dtor bytes previously.

hugetlb manually creates and destroys compound pages.  As such it makes
assumptions about struct page layout.  This breaks hugetlb.  The following
will allow fix the breakage.

The hugetlb code is quite fragile when changes like this are made.  I am
open to suggestions on how we can make this more robust.  Perhaps start
with a simple set of APIs to create_folio from a set of contiguous pages
and destroy a folio?
Matthew Wilcox Aug. 23, 2023, 12:29 a.m. UTC | #2
On Tue, Aug 22, 2023 at 04:17:41PM -0700, Mike Kravetz wrote:
> On 08/16/23 16:11, Matthew Wilcox (Oracle) wrote:
> > Store the folio order in the low byte of the flags word in the first
> > tail page.  This frees up the word that was being used to store the
> > order and dtor bytes previously.
> 
> hugetlb manually creates and destroys compound pages.  As such it makes
> assumptions about struct page layout.  This breaks hugetlb.  The following
> will allow fix the breakage.
> 
> The hugetlb code is quite fragile when changes like this are made.  I am
> open to suggestions on how we can make this more robust.  Perhaps start
> with a simple set of APIs to create_folio from a set of contiguous pages
> and destroy a folio?

Thanks; those changes look good to me.

I don't have any _immediate_ plans to make your life easier.  When
we get to a memdesc world, it'll all get much easier.
diff mbox series

Patch

diff --git a/include/linux/mm.h b/include/linux/mm.h
index cf0ae8c51d7f..85568e2b2556 100644
--- a/include/linux/mm.h
+++ b/include/linux/mm.h
@@ -1028,7 +1028,7 @@  struct inode;
  * compound_order() can be called without holding a reference, which means
  * that niceties like page_folio() don't work.  These callers should be
  * prepared to handle wild return values.  For example, PG_head may be
- * set before _folio_order is initialised, or this may be a tail page.
+ * set before the order is initialised, or this may be a tail page.
  * See compaction.c for some good examples.
  */
 static inline unsigned int compound_order(struct page *page)
@@ -1037,7 +1037,7 @@  static inline unsigned int compound_order(struct page *page)
 
 	if (!test_bit(PG_head, &folio->flags))
 		return 0;
-	return folio->_folio_order;
+	return folio->_flags_1 & 0xff;
 }
 
 /**
@@ -1053,7 +1053,7 @@  static inline unsigned int folio_order(struct folio *folio)
 {
 	if (!folio_test_large(folio))
 		return 0;
-	return folio->_folio_order;
+	return folio->_flags_1 & 0xff;
 }
 
 #include <linux/huge_mm.h>
@@ -2025,7 +2025,7 @@  static inline long folio_nr_pages(struct folio *folio)
 #ifdef CONFIG_64BIT
 	return folio->_folio_nr_pages;
 #else
-	return 1L << folio->_folio_order;
+	return 1L << (folio->_flags_1 & 0xff);
 #endif
 }
 
@@ -2043,7 +2043,7 @@  static inline unsigned long compound_nr(struct page *page)
 #ifdef CONFIG_64BIT
 	return folio->_folio_nr_pages;
 #else
-	return 1L << folio->_folio_order;
+	return 1L << (folio->_flags_1 & 0xff);
 #endif
 }
 
diff --git a/include/linux/mm_types.h b/include/linux/mm_types.h
index d45a2b8041e0..659c7b84726c 100644
--- a/include/linux/mm_types.h
+++ b/include/linux/mm_types.h
@@ -282,7 +282,6 @@  static inline struct page *encoded_page_ptr(struct encoded_page *page)
  * @_refcount: Do not access this member directly.  Use folio_ref_count()
  *    to find how many references there are to this folio.
  * @memcg_data: Memory Control Group data.
- * @_folio_order: Do not use directly, call folio_order().
  * @_entire_mapcount: Do not use directly, call folio_entire_mapcount().
  * @_nr_pages_mapped: Do not use directly, call folio_mapcount().
  * @_pincount: Do not use directly, call folio_maybe_dma_pinned().
@@ -334,8 +333,8 @@  struct folio {
 		struct {
 			unsigned long _flags_1;
 			unsigned long _head_1;
+			unsigned long _folio_avail;
 	/* public: */
-			unsigned char _folio_order;
 			atomic_t _entire_mapcount;
 			atomic_t _nr_pages_mapped;
 			atomic_t _pincount;
diff --git a/include/linux/page-flags.h b/include/linux/page-flags.h
index b452fba9bc71..5b466e619f71 100644
--- a/include/linux/page-flags.h
+++ b/include/linux/page-flags.h
@@ -184,7 +184,8 @@  enum pageflags {
 
 	/*
 	 * Flags only valid for compound pages.  Stored in first tail page's
-	 * flags word.
+	 * flags word.  Cannot use the first 8 flags or any flag marked as
+	 * PF_ANY.
 	 */
 
 	/* At least one page in this folio has the hwpoison flag set */
@@ -1065,8 +1066,8 @@  static __always_inline void __ClearPageAnonExclusive(struct page *page)
  * the CHECK_AT_FREE flags above, so need to be cleared.
  */
 #define PAGE_FLAGS_SECOND						\
-	(1UL << PG_has_hwpoisoned	| 1UL << PG_hugetlb |		\
-	 1UL << PG_large_rmappable)
+	(0xffUL /* order */		| 1UL << PG_has_hwpoisoned |	\
+	 1UL << PG_hugetlb		| 1UL << PG_large_rmappable)
 
 #define PAGE_FLAGS_PRIVATE				\
 	(1UL << PG_private | 1UL << PG_private_2)
diff --git a/kernel/crash_core.c b/kernel/crash_core.c
index 934dd86e19f5..693445e1f7f6 100644
--- a/kernel/crash_core.c
+++ b/kernel/crash_core.c
@@ -455,7 +455,6 @@  static int __init crash_save_vmcoreinfo_init(void)
 	VMCOREINFO_OFFSET(page, lru);
 	VMCOREINFO_OFFSET(page, _mapcount);
 	VMCOREINFO_OFFSET(page, private);
-	VMCOREINFO_OFFSET(folio, _folio_order);
 	VMCOREINFO_OFFSET(page, compound_head);
 	VMCOREINFO_OFFSET(pglist_data, node_zones);
 	VMCOREINFO_OFFSET(pglist_data, nr_zones);
diff --git a/mm/internal.h b/mm/internal.h
index 9dc7629ffbc9..5c777b6779fa 100644
--- a/mm/internal.h
+++ b/mm/internal.h
@@ -407,7 +407,7 @@  static inline void folio_set_order(struct folio *folio, unsigned int order)
 	if (WARN_ON_ONCE(!order || !folio_test_large(folio)))
 		return;
 
-	folio->_folio_order = order;
+	folio->_flags_1 = (folio->_flags_1 & ~0xffUL) | order;
 #ifdef CONFIG_64BIT
 	folio->_folio_nr_pages = 1U << order;
 #endif