Skip to contentSkip to conversation

btrfs: migrate extent_buffer::pages[] to folio and more cleanups

PATCH In review Qu Wenruo 2 participants 5 messages Dec 6, 2023 – Dec 7, 2023 Btrfs file system

lore
Share & tools
lore

Qu Wenruo

author
#0 lore raw

v3 cover btrfs: migrate extent_buffer::pages[] to folio and more cleanups

[CHANGELOG]
v2:
- Add a new patch to do more cleanups for metadata page pointers usage

v3:
- Fix a bug in alloc_eb_folio_array() which accesses uninitialized stack It turns out that I have CONFIG_INIT_STACK_ALL_ZERO, which prevents me from exposing the bug. Now changed to CONFIG_INIT_STACK_ALL_PATTERN.

This patchset would migrate extent_buffer::pages[] to folios[], then cleanup the existing metadata page pointer usage to proper folios ones.

This cleanup would help future higher order folios usage for metadata in the following aspects:

- No more need to iterate through the remaining pages for page flags We just call folio_set/mark/start_*() helpers, for the single folio. The helper would only set the flag (mostly for the leading page).

- Single bio_add_folio() call for the whole eb

- Better filio helpers naming PageUptodate() compared to folio_test_uptodate().

The first patch would do a very simple conversion, then the 2nd patch do the prepartion for the higher order folio situation.

There are two locations which won't be converted to folios yet:

- Subpage code
There is no meaning to support higher order folio for subpage. The two conditions are just conflicting with each other.

- Data page pointers
That would be more useful in the future, before we going to support multi-page sectorsize.

However the 2nd one would also add a new corner case:

- Order mismatch in filemap and eb folios
Unforatunately I don't have a better plan other than re-allocate the folios to the same order.
Maybe in the future we would have better ways to handle it? Like migrating the pages to the higher order one?

Qu Wenruo (2):
btrfs: migrate extent_buffer::pages[] to folio
btrfs: cleanup metadata page pointer usage

 fs/btrfs/accessors.c             |  20 +-
 fs/btrfs/accessors.h             |   4 +-
 fs/btrfs/ctree.c                 |   2 +-
 fs/btrfs/disk-io.c               |  25 +-
 fs/btrfs/extent_io.c             | 402 ++++++++++++++++++-------------
 fs/btrfs/extent_io.h             |  21 +-
 fs/btrfs/inode.c                 |   2 +-
 fs/btrfs/subpage.c               |  60 ++---
 fs/btrfs/subpage.h               |  11 +-
 fs/btrfs/tests/extent-io-tests.c |   4 +-
 10 files changed, 322 insertions(+), 229 deletions(-)
Signature · 1 line
2.43.0
author
#1 lore raw

PATCH v3 1/2 btrfs: migrate extent_buffer::pages[] to folio

For now extent_buffer::pages[] are still only accept single page pointer, thus we can migrate to folios pretty easily.

As for single page, page and folio are 1:1 mapped, including their page flags.

This patch would just do the conversion from struct page to struct folio, providing the first step to higher order folio in the future.

This conversion is pretty simple:

- extent_buffer::pages[] -> extent_buffer::folios[]

- page_address(eb->pages[i]) -> folio_address(eb->pages[i])

- eb->pages[i] -> folio_page(eb->folios[i], 0)

There would be more specific cleanups preparing for the incoming higher order folio support.

---
 fs/btrfs/accessors.c             |  20 ++---
 fs/btrfs/accessors.h             |   4 +-
 fs/btrfs/ctree.c                 |   2 +-
 fs/btrfs/disk-io.c               |  19 ++---
 fs/btrfs/extent_io.c             | 123 ++++++++++++++++++-------------
 fs/btrfs/extent_io.h             |   7 +-
 fs/btrfs/tests/extent-io-tests.c |   4 +-
 7 files changed, 103 insertions(+), 76 deletions(-)
fs/btrfs/accessors.c +10 -10
index 206cf1612c1d..8f7cbb7154d4 100644
@@ -27,7 +27,7 @@ static bool check_setget_bounds(const struct extent_buffer *eb,
 void btrfs_init_map_token(struct btrfs_map_token *token, struct extent_buffer *eb)
 {
 	token->eb = eb;
-	token->kaddr = page_address(eb->pages[0]);
+	token->kaddr = folio_address(eb->folios[0]);
 	token->offset = 0;
 }
 
@@ -50,7 +50,7 @@ void btrfs_init_map_token(struct btrfs_map_token *token, struct extent_buffer *e
  * an offset into the extent buffer page array, cast to a specific type.  This
  * gives us all the type checking.
  *
- * The extent buffer pages stored in the array pages do not form a contiguous
+ * The extent buffer pages stored in the array folios may not form a contiguous
  * phyusical range, but the API functions assume the linear offset to the range
  * from 0 to metadata node size.
  */
@@ -74,13 +74,13 @@ u##bits btrfs_get_token_##bits(struct btrfs_map_token *token,		\
 	    member_offset + size <= token->offset + PAGE_SIZE) {	\
 		return get_unaligned_le##bits(token->kaddr + oip);	\
 	}								\
-	token->kaddr = page_address(token->eb->pages[idx]);		\
+	token->kaddr = folio_address(token->eb->folios[idx]);		\
 	token->offset = idx << PAGE_SHIFT;				\
 	if (INLINE_EXTENT_BUFFER_PAGES == 1 || oip + size <= PAGE_SIZE ) \
 		return get_unaligned_le##bits(token->kaddr + oip);	\
 									\
 	memcpy(lebytes, token->kaddr + oip, part);			\
-	token->kaddr = page_address(token->eb->pages[idx + 1]);		\
+	token->kaddr = folio_address(token->eb->folios[idx + 1]);	\
 	token->offset = (idx + 1) << PAGE_SHIFT;			\
 	memcpy(lebytes + part, token->kaddr, size - part);		\
 	return get_unaligned_le##bits(lebytes);				\
@@ -91,7 +91,7 @@ u##bits btrfs_get_##bits(const struct extent_buffer *eb,		\
 	const unsigned long member_offset = (unsigned long)ptr + off;	\
 	const unsigned long oip = get_eb_offset_in_page(eb, member_offset); \
 	const unsigned long idx = get_eb_page_index(member_offset);	\
-	char *kaddr = page_address(eb->pages[idx]);			\
+	char *kaddr = folio_address(eb->folios[idx]);			\
 	const int size = sizeof(u##bits);				\
 	const int part = PAGE_SIZE - oip;				\
 	u8 lebytes[sizeof(u##bits)];					\
@@ -101,7 +101,7 @@ u##bits btrfs_get_##bits(const struct extent_buffer *eb,		\
 		return get_unaligned_le##bits(kaddr + oip);		\
 									\
 	memcpy(lebytes, kaddr + oip, part);				\
-	kaddr = page_address(eb->pages[idx + 1]);			\
+	kaddr = folio_address(eb->folios[idx + 1]);			\
 	memcpy(lebytes + part, kaddr, size - part);			\
 	return get_unaligned_le##bits(lebytes);				\
 }									\
@@ -125,7 +125,7 @@ void btrfs_set_token_##bits(struct btrfs_map_token *token,		\
 		put_unaligned_le##bits(val, token->kaddr + oip);	\
 		return;							\
 	}								\
-	token->kaddr = page_address(token->eb->pages[idx]);		\
+	token->kaddr = folio_address(token->eb->folios[idx]);		\
 	token->offset = idx << PAGE_SHIFT;				\
 	if (INLINE_EXTENT_BUFFER_PAGES == 1 || oip + size <= PAGE_SIZE) { \
 		put_unaligned_le##bits(val, token->kaddr + oip);	\
@@ -133,7 +133,7 @@ void btrfs_set_token_##bits(struct btrfs_map_token *token,		\
 	}								\
 	put_unaligned_le##bits(val, lebytes);				\
 	memcpy(token->kaddr + oip, lebytes, part);			\
-	token->kaddr = page_address(token->eb->pages[idx + 1]);		\
+	token->kaddr = folio_address(token->eb->folios[idx + 1]);	\
 	token->offset = (idx + 1) << PAGE_SHIFT;			\
 	memcpy(token->kaddr, lebytes + part, size - part);		\
 }									\
@@ -143,7 +143,7 @@ void btrfs_set_##bits(const struct extent_buffer *eb, void *ptr,	\
 	const unsigned long member_offset = (unsigned long)ptr + off;	\
 	const unsigned long oip = get_eb_offset_in_page(eb, member_offset); \
 	const unsigned long idx = get_eb_page_index(member_offset);	\
-	char *kaddr = page_address(eb->pages[idx]);			\
+	char *kaddr = folio_address(eb->folios[idx]);			\
 	const int size = sizeof(u##bits);				\
 	const int part = PAGE_SIZE - oip;				\
 	u8 lebytes[sizeof(u##bits)];					\
@@ -156,7 +156,7 @@ void btrfs_set_##bits(const struct extent_buffer *eb, void *ptr,	\
 									\
 	put_unaligned_le##bits(val, lebytes);				\
 	memcpy(kaddr + oip, lebytes, part);				\
-	kaddr = page_address(eb->pages[idx + 1]);			\
+	kaddr = folio_address(eb->folios[idx + 1]);			\
 	memcpy(kaddr, lebytes + part, size - part);			\
 }
 
fs/btrfs/accessors.h +2 -2
index aa0844535644..ed7aa32972ad 100644
@@ -90,14 +90,14 @@ static inline void btrfs_set_token_##name(struct btrfs_map_token *token,\
 #define BTRFS_SETGET_HEADER_FUNCS(name, type, member, bits)		\
 static inline u##bits btrfs_##name(const struct extent_buffer *eb)	\
 {									\
-	const type *p = page_address(eb->pages[0]) +			\
+	const type *p = folio_address(eb->folios[0]) +			\
 			offset_in_page(eb->start);			\
 	return get_unaligned_le##bits(&p->member);			\
 }									\
 static inline void btrfs_set_##name(const struct extent_buffer *eb,	\
 				    u##bits val)			\
 {									\
-	type *p = page_address(eb->pages[0]) + offset_in_page(eb->start); \
+	type *p = folio_address(eb->folios[0]) + offset_in_page(eb->start); \
 	put_unaligned_le##bits(val, &p->member);			\
 }
 
fs/btrfs/ctree.c +1 -1
index 137c4eb24c28..e6c535cf3749 100644
@@ -832,7 +832,7 @@ int btrfs_bin_search(struct extent_buffer *eb, int first_slot,
 
 		if (oip + key_size <= PAGE_SIZE) {
 			const unsigned long idx = get_eb_page_index(offset);
-			char *kaddr = page_address(eb->pages[idx]);
+			char *kaddr = folio_address(eb->folios[idx]);
 
 			oip = get_eb_offset_in_page(eb, offset);
 			tmp = (struct btrfs_disk_key *)(kaddr + oip);
fs/btrfs/disk-io.c +10 -9
index 9317606017e2..78bb85f775f6 100644
@@ -89,7 +89,7 @@ static void csum_tree_block(struct extent_buffer *buf, u8 *result)
 		first_page_part = fs_info->nodesize;
 		num_pages = 1;
 	} else {
-		kaddr = page_address(buf->pages[0]);
+		kaddr = folio_address(buf->folios[0]);
 		first_page_part = min_t(u32, PAGE_SIZE, fs_info->nodesize);
 		num_pages = num_extent_pages(buf);
 	}
@@ -98,7 +98,7 @@ static void csum_tree_block(struct extent_buffer *buf, u8 *result)
 			    first_page_part - BTRFS_CSUM_SIZE);
 
 	for (i = 1; i < num_pages && INLINE_EXTENT_BUFFER_PAGES > 1; i++) {
-		kaddr = page_address(buf->pages[i]);
+		kaddr = folio_address(buf->folios[i]);
 		crypto_shash_update(shash, kaddr, PAGE_SIZE);
 	}
 	memset(result, 0, BTRFS_CSUM_SIZE);
@@ -184,13 +184,14 @@ static int btrfs_repair_eb_io_failure(const struct extent_buffer *eb,
 		return -EROFS;
 
 	for (i = 0; i < num_pages; i++) {
-		struct page *p = eb->pages[i];
-		u64 start = max_t(u64, eb->start, page_offset(p));
-		u64 end = min_t(u64, eb->start + eb->len, page_offset(p) + PAGE_SIZE);
+		u64 start = max_t(u64, eb->start, folio_pos(eb->folios[i]));
+		u64 end = min_t(u64, eb->start + eb->len,
+				folio_pos(eb->folios[i]) + PAGE_SIZE);
 		u32 len = end - start;
 
 		ret = btrfs_repair_io_failure(fs_info, 0, start, len,
-				start, p, offset_in_page(start), mirror_num);
+				start, folio_page(eb->folios[i], 0),
+				offset_in_page(start), mirror_num);
 		if (ret)
 			break;
 	}
@@ -277,8 +278,8 @@ blk_status_t btree_csum_one_bio(struct btrfs_bio *bbio)
 
 	if (WARN_ON_ONCE(found_start != eb->start))
 		return BLK_STS_IOERR;
-	if (WARN_ON(!btrfs_page_test_uptodate(fs_info, eb->pages[0], eb->start,
-					      eb->len)))
+	if (WARN_ON(!btrfs_page_test_uptodate(fs_info, folio_page(eb->folios[0], 0),
+					      eb->start, eb->len)))
 		return BLK_STS_IOERR;
 
 	ASSERT(memcmp_extent_buffer(eb, fs_info->fs_devices->metadata_uuid,
@@ -387,7 +388,7 @@ int btrfs_validate_extent_buffer(struct extent_buffer *eb,
 	}
 
 	csum_tree_block(eb, result);
-	header_csum = page_address(eb->pages[0]) +
+	header_csum = folio_address(eb->folios[0]) +
 		get_eb_offset_in_page(eb, offsetof(struct btrfs_header, csum));
 
 	if (memcmp(result, header_csum, csum_size) != 0) {
fs/btrfs/extent_io.c +72 -51

358 lines not on this page · raw message

fs/btrfs/extent_io.h +6 -1
index c73d53c22ec5..66c2e214b141 100644
@@ -94,7 +94,12 @@ struct extent_buffer {
 
 	struct rw_semaphore lock;
 
-	struct page *pages[INLINE_EXTENT_BUFFER_PAGES];
+	/*
+	 * Pointers to all the folios of the extent buffer.
+	 *
+	 * For now the folio is always order 0 (aka, a single page).
+	 */
+	struct folio *folios[INLINE_EXTENT_BUFFER_PAGES];
 #ifdef CONFIG_BTRFS_DEBUG
 	struct list_head leak_list;
 	pid_t lock_owner;
fs/btrfs/tests/extent-io-tests.c +2 -2
index 1cc86af97dc6..25b3349595e0 100644
@@ -652,7 +652,7 @@ static void dump_eb_and_memory_contents(struct extent_buffer *eb, void *memory,
 					const char *test_name)
 {
 	for (int i = 0; i < eb->len; i++) {
-		struct page *page = eb->pages[i >> PAGE_SHIFT];
+		struct page *page = folio_page(eb->folios[i >> PAGE_SHIFT], 0);
 		void *addr = page_address(page) + offset_in_page(i);
 
 		if (memcmp(addr, memory + i, 1) != 0) {
@@ -668,7 +668,7 @@ static int verify_eb_and_memory(struct extent_buffer *eb, void *memory,
 				const char *test_name)
 {
 	for (int i = 0; i < (eb->len >> PAGE_SHIFT); i++) {
-		void *eb_addr = page_address(eb->pages[i]);
+		void *eb_addr = folio_address(eb->folios[i]);
 
 		if (memcmp(memory + (i << PAGE_SHIFT), eb_addr, PAGE_SIZE) != 0) {
 			dump_eb_and_memory_contents(eb, memory, test_name);
Signature · 1 line
2.43.0
author
#2 lore raw

PATCH v3 2/2 btrfs: cleanup metadata page pointer usage

In reply to Qu Wenruo

Although we have migrated extent_buffer::pages[] to folios[], we're still mostly using the folio_page() help to grab the page.

This patch would do the following cleanups for metadata:

- Introduce num_extent_folios() helper
This is to replace most num_extent_pages() callers.

- Use num_extent_folios() to iterate future large folios This allows us to use things like bio_add_folio()/bio_add_folio_nofail(), and only set the needed flags for the folio (aka the leading/tailing page), which reduces the loop iteration to 1 for large folios.

- Change metadata related functions to use folio pointers Including their function name, involving:
* attach_extent_buffer_page()
* detach_extent_buffer_page()
* page_range_has_eb()
* btrfs_release_extent_buffer_pages()
* btree_clear_page_dirty()
* btrfs_page_inc_eb_refs()
* btrfs_page_dec_eb_refs()

- Change btrfs_is_subpage() to accept an address_space pointer This is to allow both page->mapping and folio->mapping to be utilized. As data is still using the old per-page code, and may keep so for a while.

- Special corner case place holder for future order mismatches between extent buffer and inode filemap
For now it's just a block of comments and a dead ASSERT(), no real handling yet.

The subpage code would still go page, just because subpage and large folio are conflicting conditions, thus we don't need to bother subpage with higher order folios at all. Just folio_page(folio, 0) would be enough.

---
 fs/btrfs/disk-io.c   |   6 +
 fs/btrfs/extent_io.c | 319 ++++++++++++++++++++++++-------------------
 fs/btrfs/extent_io.h |  14 ++
 fs/btrfs/inode.c     |   2 +-
 fs/btrfs/subpage.c   |  60 ++++----
 fs/btrfs/subpage.h   |  11 +-
 6 files changed, 239 insertions(+), 173 deletions(-)
fs/btrfs/disk-io.c +6 -0
index 78bb85f775f6..a5ace9f6e790 100644
@@ -97,6 +97,12 @@ static void csum_tree_block(struct extent_buffer *buf, u8 *result)
 	crypto_shash_update(shash, kaddr + BTRFS_CSUM_SIZE,
 			    first_page_part - BTRFS_CSUM_SIZE);
 
+	/*
+	 * Multiple single-page folios case would reach here.
+	 *
+	 * nodesize <= PAGE_SIZE and large folio all handled by above
+	 * crypto_shash_update() already.
+	 */
 	for (i = 1; i < num_pages && INLINE_EXTENT_BUFFER_PAGES > 1; i++) {
 		kaddr = folio_address(buf->folios[i]);
 		crypto_shash_update(shash, kaddr, PAGE_SIZE);
fs/btrfs/extent_io.c +179 -140

673 lines not on this page · raw message

fs/btrfs/extent_io.h +14 -0
index 66c2e214b141..a5fd5cb20a3c 100644
@@ -243,6 +243,20 @@ static inline int num_extent_pages(const struct extent_buffer *eb)
 	return (eb->len >> PAGE_SHIFT) ?: 1;
 }
 
+/*
+ * This can only be determined at runtime by checking eb::folios[0].
+ *
+ * As we can have either one large folio covering the whole eb
+ * (either nodesize <= PAGE_SIZE, or high order folio), or multiple
+ * single-paged folios.
+ */
+static inline int num_extent_folios(const struct extent_buffer *eb)
+{
+	if (folio_order(eb->folios[0]))
+		return 1;
+	return num_extent_pages(eb);
+}
+
 static inline int extent_buffer_uptodate(const struct extent_buffer *eb)
 {
 	return test_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags);
fs/btrfs/inode.c +1 -1
index f7c0a5ec675f..9ede6aa77fde 100644
@@ -7863,7 +7863,7 @@ static void wait_subpage_spinlock(struct page *page)
 	struct folio *folio = page_folio(page);
 	struct btrfs_subpage *subpage;
 
-	if (!btrfs_is_subpage(fs_info, page))
+	if (!btrfs_is_subpage(fs_info, page->mapping))
 		return;
 
 	ASSERT(folio_test_private(folio) && folio_get_private(folio));
fs/btrfs/subpage.c +33 -27

165 lines not on this page · raw message

fs/btrfs/subpage.h +6 -5
index 5cbf67ccbdeb..93d1c5690faf 100644
@@ -73,7 +73,8 @@ enum btrfs_subpage_type {
 	BTRFS_SUBPAGE_DATA,
 };
 
-bool btrfs_is_subpage(const struct btrfs_fs_info *fs_info, struct page *page);
+bool btrfs_is_subpage(const struct btrfs_fs_info *fs_info,
+		      struct address_space *mapping);
 
 void btrfs_init_subpage_info(struct btrfs_subpage_info *subpage_info, u32 sectorsize);
 int btrfs_attach_subpage(const struct btrfs_fs_info *fs_info,
@@ -86,10 +87,10 @@ struct btrfs_subpage *btrfs_alloc_subpage(const struct btrfs_fs_info *fs_info,
 					  enum btrfs_subpage_type type);
 void btrfs_free_subpage(struct btrfs_subpage *subpage);
 
-void btrfs_page_inc_eb_refs(const struct btrfs_fs_info *fs_info,
-			    struct page *page);
-void btrfs_page_dec_eb_refs(const struct btrfs_fs_info *fs_info,
-			    struct page *page);
+void btrfs_folio_inc_eb_refs(const struct btrfs_fs_info *fs_info,
+			     struct folio *folio);
+void btrfs_folio_dec_eb_refs(const struct btrfs_fs_info *fs_info,
+			     struct folio *folio);
 
 void btrfs_subpage_start_reader(const struct btrfs_fs_info *fs_info,
 		struct page *page, u64 start, u32 len);
Signature · 1 line
2.43.0
#3 lore raw

Re: [PATCH v3 2/2] btrfs: cleanup metadata page pointer usage

Quoted 4 lines · Qu Wenruo

/* Release all pages attached to the extent buffer */ static void btrfs_release_extent_buffer_pages(struct extent_buffer *eb) {

fs/btrfs/extent_io.c +0 -0
 	int i;

fs/btrfs/extent_io.c static void detach_extent_buffer_page(struct extent_buffer *eb, struct page *pag · line 3185

Quoted 10 lines
fs/btrfs/extent_io.c +3 -4
-	int num_pages;
+	int num_folios;
 
 	ASSERT(!extent_buffer_under_io(eb));
 
-	num_pages = num_extent_pages(eb);
-	for (i = 0; i < num_pages; i++) {
-		struct page *page = folio_page(eb->folios[i], 0);
+	num_folios = num_extent_folios(eb);
+	for (i = 0; i < num_folios; i++) {

I could not resist and updated the style in a few functions to

	int num_folios = num_extent_folios(eb);

	for (int i = 0; ...) {
	}

i.e. move num_extent_folios() to the declarataions unless there's some 'return/return' between them and moved the global 'int i' to the for() loop.

It's a cleanup but the lines are changed anyway and we'll probabaly not touch this code anytime soon.

#4 lore raw
In reply to Qu Wenruo
Quoted 8 lines · Qu Wenruo

[CHANGELOG]
v2:
- Add a new patch to do more cleanups for metadata page pointers usage

v3:
- Fix a bug in alloc_eb_folio_array() which accesses uninitialized stack It turns out that I have CONFIG_INIT_STACK_ALL_ZERO, which prevents me from exposing the bug. Now changed to CONFIG_INIT_STACK_ALL_PATTERN.

Seems to be fixed, thanks. I had CONFIG_INIT_STACK_NONE=y so it got caught due to random data on the stack but with CONFIG_INIT_STACK_ALL_PATTERN the pattern makes it clear that it's due to that.

Added to misc-next, thanks.