1 files changed, 196 insertions, 182 deletions

diff --git a/fs/btrfs/subpage.c b/fs/btrfs/subpage.c
index 0e49dab8dad2..722acf768396 100644
--- a/fs/btrfs/subpage.c
+++ b/fs/btrfs/subpage.c
@@ -64,6 +64,7 @@
  *   This means a slightly higher tree locking latency.
  */
 
+#if PAGE_SIZE > SZ_4K
 bool btrfs_is_subpage(const struct btrfs_fs_info *fs_info, struct address_space *mapping)
 {
 	if (fs_info->sectorsize >= PAGE_SIZE)
@@ -74,7 +75,7 @@ bool btrfs_is_subpage(const struct btrfs_fs_info *fs_info, struct address_space
 	 * mapping. And if page->mapping->host is data inode, it's subpage.
 	 * As we have ruled our sectorsize >= PAGE_SIZE case already.
 	 */
-	if (!mapping || !mapping->host || is_data_inode(mapping->host))
+	if (!mapping || !mapping->host || is_data_inode(BTRFS_I(mapping->host)))
 		return true;
 
 	/*
@@ -85,34 +86,7 @@ bool btrfs_is_subpage(const struct btrfs_fs_info *fs_info, struct address_space
 		return true;
 	return false;
 }
-
-void btrfs_init_subpage_info(struct btrfs_subpage_info *subpage_info, u32 sectorsize)
-{
-	unsigned int cur = 0;
-	unsigned int nr_bits;
-
-	ASSERT(IS_ALIGNED(PAGE_SIZE, sectorsize));
-
-	nr_bits = PAGE_SIZE / sectorsize;
-	subpage_info->bitmap_nr_bits = nr_bits;
-
-	subpage_info->uptodate_offset = cur;
-	cur += nr_bits;
-
-	subpage_info->dirty_offset = cur;
-	cur += nr_bits;
-
-	subpage_info->writeback_offset = cur;
-	cur += nr_bits;
-
-	subpage_info->ordered_offset = cur;
-	cur += nr_bits;
-
-	subpage_info->checked_offset = cur;
-	cur += nr_bits;
-
-	subpage_info->total_nr_bits = cur;
-}
+#endif
 
 int btrfs_attach_subpage(const struct btrfs_fs_info *fs_info,
 			 struct folio *folio, enum btrfs_subpage_type type)
@@ -160,18 +134,16 @@ struct btrfs_subpage *btrfs_alloc_subpage(const struct btrfs_fs_info *fs_info,
 	ASSERT(fs_info->sectorsize < PAGE_SIZE);
 
 	real_size = struct_size(ret, bitmaps,
-			BITS_TO_LONGS(fs_info->subpage_info->total_nr_bits));
+			BITS_TO_LONGS(btrfs_bitmap_nr_max * fs_info->sectors_per_page));
 	ret = kzalloc(real_size, GFP_NOFS);
 	if (!ret)
 		return ERR_PTR(-ENOMEM);
 
 	spin_lock_init(&ret->lock);
-	if (type == BTRFS_SUBPAGE_METADATA) {
+	if (type == BTRFS_SUBPAGE_METADATA)
 		atomic_set(&ret->eb_refs, 0);
-	} else {
-		atomic_set(&ret->readers, 0);
-		atomic_set(&ret->writers, 0);
-	}
+	else
+		atomic_set(&ret->nr_locked, 0);
 	return ret;
 }
 
@@ -237,40 +209,15 @@ static void btrfs_subpage_assert(const struct btrfs_fs_info *fs_info,
 		       start + len <= folio_pos(folio) + PAGE_SIZE);
 }
 
-void btrfs_subpage_start_reader(const struct btrfs_fs_info *fs_info,
-				struct folio *folio, u64 start, u32 len)
-{
-	struct btrfs_subpage *subpage = folio_get_private(folio);
-	const int nbits = len >> fs_info->sectorsize_bits;
-
-	btrfs_subpage_assert(fs_info, folio, start, len);
-
-	atomic_add(nbits, &subpage->readers);
-}
-
-void btrfs_subpage_end_reader(const struct btrfs_fs_info *fs_info,
-			      struct folio *folio, u64 start, u32 len)
-{
-	struct btrfs_subpage *subpage = folio_get_private(folio);
-	const int nbits = len >> fs_info->sectorsize_bits;
-	bool is_data;
-	bool last;
-
-	btrfs_subpage_assert(fs_info, folio, start, len);
-	is_data = is_data_inode(folio->mapping->host);
-	ASSERT(atomic_read(&subpage->readers) >= nbits);
-	last = atomic_sub_and_test(nbits, &subpage->readers);
-
-	/*
-	 * For data we need to unlock the page if the last read has finished.
-	 *
-	 * And please don't replace @last with atomic_sub_and_test() call
-	 * inside if () condition.
-	 * As we want the atomic_sub_and_test() to be always executed.
-	 */
-	if (is_data && last)
-		folio_unlock(folio);
-}
+#define subpage_calc_start_bit(fs_info, folio, name, start, len)	\
+({									\
+	unsigned int __start_bit;						\
+									\
+	btrfs_subpage_assert(fs_info, folio, start, len);		\
+	__start_bit = offset_in_page(start) >> fs_info->sectorsize_bits; \
+	__start_bit += fs_info->sectors_per_page * btrfs_bitmap_nr_##name; \
+	__start_bit;							\
+})
 
 static void btrfs_subpage_clamp_range(struct folio *folio, u64 *start, u32 *len)
 {
@@ -290,100 +237,130 @@ static void btrfs_subpage_clamp_range(struct folio *folio, u64 *start, u32 *len)
 			     orig_start + orig_len) - *start;
 }
 
-void btrfs_subpage_start_writer(const struct btrfs_fs_info *fs_info,
-				struct folio *folio, u64 start, u32 len)
-{
-	struct btrfs_subpage *subpage = folio_get_private(folio);
-	const int nbits = (len >> fs_info->sectorsize_bits);
-	int ret;
-
-	btrfs_subpage_assert(fs_info, folio, start, len);
-
-	ASSERT(atomic_read(&subpage->readers) == 0);
-	ret = atomic_add_return(nbits, &subpage->writers);
-	ASSERT(ret == nbits);
-}
-
-bool btrfs_subpage_end_and_test_writer(const struct btrfs_fs_info *fs_info,
-				       struct folio *folio, u64 start, u32 len)
+static bool btrfs_subpage_end_and_test_lock(const struct btrfs_fs_info *fs_info,
+					    struct folio *folio, u64 start, u32 len)
 {
 	struct btrfs_subpage *subpage = folio_get_private(folio);
+	const int start_bit = subpage_calc_start_bit(fs_info, folio, locked, start, len);
 	const int nbits = (len >> fs_info->sectorsize_bits);
+	unsigned long flags;
+	unsigned int cleared = 0;
+	int bit = start_bit;
+	bool last;
 
 	btrfs_subpage_assert(fs_info, folio, start, len);
 
+	spin_lock_irqsave(&subpage->lock, flags);
 	/*
 	 * We have call sites passing @lock_page into
 	 * extent_clear_unlock_delalloc() for compression path.
 	 *
 	 * This @locked_page is locked by plain lock_page(), thus its
-	 * subpage::writers is 0.  Handle them in a special way.
+	 * subpage::locked is 0.  Handle them in a special way.
 	 */
-	if (atomic_read(&subpage->writers) == 0)
+	if (atomic_read(&subpage->nr_locked) == 0) {
+		spin_unlock_irqrestore(&subpage->lock, flags);
 		return true;
+	}
 
-	ASSERT(atomic_read(&subpage->writers) >= nbits);
-	return atomic_sub_and_test(nbits, &subpage->writers);
+	for_each_set_bit_from(bit, subpage->bitmaps, start_bit + nbits) {
+		clear_bit(bit, subpage->bitmaps);
+		cleared++;
+	}
+	ASSERT(atomic_read(&subpage->nr_locked) >= cleared);
+	last = atomic_sub_and_test(cleared, &subpage->nr_locked);
+	spin_unlock_irqrestore(&subpage->lock, flags);
+	return last;
 }
 
 /*
- * Lock a folio for delalloc page writeback.
+ * Handle different locked folios:
  *
- * Return -EAGAIN if the page is not properly initialized.
- * Return 0 with the page locked, and writer counter updated.
+ * - Non-subpage folio
+ *   Just unlock it.
  *
- * Even with 0 returned, the page still need extra check to make sure
- * it's really the correct page, as the caller is using
- * filemap_get_folios_contig(), which can race with page invalidating.
+ * - folio locked but without any subpage locked
+ *   This happens either before writepage_delalloc() or the delalloc range is
+ *   already handled by previous folio.
+ *   We can simple unlock it.
+ *
+ * - folio locked with subpage range locked.
+ *   We go through the locked sectors inside the range and clear their locked
+ *   bitmap, reduce the writer lock number, and unlock the page if that's
+ *   the last locked range.
  */
-int btrfs_folio_start_writer_lock(const struct btrfs_fs_info *fs_info,
-				  struct folio *folio, u64 start, u32 len)
+void btrfs_folio_end_lock(const struct btrfs_fs_info *fs_info,
+			  struct folio *folio, u64 start, u32 len)
 {
+	struct btrfs_subpage *subpage = folio_get_private(folio);
+
+	ASSERT(folio_test_locked(folio));
+
 	if (unlikely(!fs_info) || !btrfs_is_subpage(fs_info, folio->mapping)) {
-		folio_lock(folio);
-		return 0;
+		folio_unlock(folio);
+		return;
 	}
-	folio_lock(folio);
-	if (!folio_test_private(folio) || !folio_get_private(folio)) {
+
+	/*
+	 * For subpage case, there are two types of locked page.  With or
+	 * without locked number.
+	 *
+	 * Since we own the page lock, no one else could touch subpage::locked
+	 * and we are safe to do several atomic operations without spinlock.
+	 */
+	if (atomic_read(&subpage->nr_locked) == 0) {
+		/* No subpage lock, locked by plain lock_page(). */
 		folio_unlock(folio);
-		return -EAGAIN;
+		return;
 	}
+
 	btrfs_subpage_clamp_range(folio, &start, &len);
-	btrfs_subpage_start_writer(fs_info, folio, start, len);
-	return 0;
+	if (btrfs_subpage_end_and_test_lock(fs_info, folio, start, len))
+		folio_unlock(folio);
 }
 
-void btrfs_folio_end_writer_lock(const struct btrfs_fs_info *fs_info,
-				 struct folio *folio, u64 start, u32 len)
+void btrfs_folio_end_lock_bitmap(const struct btrfs_fs_info *fs_info,
+				 struct folio *folio, unsigned long bitmap)
 {
-	if (unlikely(!fs_info) || !btrfs_is_subpage(fs_info, folio->mapping)) {
+	struct btrfs_subpage *subpage = folio_get_private(folio);
+	const int start_bit = fs_info->sectors_per_page * btrfs_bitmap_nr_locked;
+	unsigned long flags;
+	bool last = false;
+	int cleared = 0;
+	int bit;
+
+	if (!btrfs_is_subpage(fs_info, folio->mapping)) {
 		folio_unlock(folio);
 		return;
 	}
-	btrfs_subpage_clamp_range(folio, &start, &len);
-	if (btrfs_subpage_end_and_test_writer(fs_info, folio, start, len))
+
+	if (atomic_read(&subpage->nr_locked) == 0) {
+		/* No subpage lock, locked by plain lock_page(). */
 		folio_unlock(folio);
-}
+		return;
+	}
 
-#define subpage_calc_start_bit(fs_info, folio, name, start, len)	\
-({									\
-	unsigned int start_bit;						\
-									\
-	btrfs_subpage_assert(fs_info, folio, start, len);		\
-	start_bit = offset_in_page(start) >> fs_info->sectorsize_bits;	\
-	start_bit += fs_info->subpage_info->name##_offset;		\
-	start_bit;							\
-})
+	spin_lock_irqsave(&subpage->lock, flags);
+	for_each_set_bit(bit, &bitmap, fs_info->sectors_per_page) {
+		if (test_and_clear_bit(bit + start_bit, subpage->bitmaps))
+			cleared++;
+	}
+	ASSERT(atomic_read(&subpage->nr_locked) >= cleared);
+	last = atomic_sub_and_test(cleared, &subpage->nr_locked);
+	spin_unlock_irqrestore(&subpage->lock, flags);
+	if (last)
+		folio_unlock(folio);
+}
 
 #define subpage_test_bitmap_all_set(fs_info, subpage, name)		\
 	bitmap_test_range_all_set(subpage->bitmaps,			\
-			fs_info->subpage_info->name##_offset,		\
-			fs_info->subpage_info->bitmap_nr_bits)
+			fs_info->sectors_per_page * btrfs_bitmap_nr_##name, \
+			fs_info->sectors_per_page)
 
 #define subpage_test_bitmap_all_zero(fs_info, subpage, name)		\
 	bitmap_test_range_all_zero(subpage->bitmaps,			\
-			fs_info->subpage_info->name##_offset,		\
-			fs_info->subpage_info->bitmap_nr_bits)
+			fs_info->sectors_per_page * btrfs_bitmap_nr_##name, \
+			fs_info->sectors_per_page)
 
 void btrfs_subpage_set_uptodate(const struct btrfs_fs_info *fs_info,
 				struct folio *folio, u64 start, u32 len)
@@ -658,109 +635,146 @@ IMPLEMENT_BTRFS_PAGE_OPS(ordered, folio_set_ordered, folio_clear_ordered,
 IMPLEMENT_BTRFS_PAGE_OPS(checked, folio_set_checked, folio_clear_checked,
 			 folio_test_checked);
 
+#define GET_SUBPAGE_BITMAP(subpage, fs_info, name, dst)			\
+{									\
+	const int sectors_per_page = fs_info->sectors_per_page;		\
+									\
+	ASSERT(sectors_per_page < BITS_PER_LONG);			\
+	*dst = bitmap_read(subpage->bitmaps,				\
+			   sectors_per_page * btrfs_bitmap_nr_##name,	\
+			   sectors_per_page);				\
+}
+
+#define SUBPAGE_DUMP_BITMAP(fs_info, folio, name, start, len)		\
+{									\
+	const struct btrfs_subpage *subpage = folio_get_private(folio);	\
+	unsigned long bitmap;						\
+									\
+	GET_SUBPAGE_BITMAP(subpage, fs_info, name, &bitmap);		\
+	btrfs_warn(fs_info,						\
+	"dumpping bitmap start=%llu len=%u folio=%llu " #name "_bitmap=%*pbl", \
+		   start, len, folio_pos(folio),			\
+		   fs_info->sectors_per_page, &bitmap);			\
+}
+
 /*
  * Make sure not only the page dirty bit is cleared, but also subpage dirty bit
  * is cleared.
  */
-void btrfs_folio_assert_not_dirty(const struct btrfs_fs_info *fs_info, struct folio *folio)
+void btrfs_folio_assert_not_dirty(const struct btrfs_fs_info *fs_info,
+				  struct folio *folio, u64 start, u32 len)
 {
-	struct btrfs_subpage *subpage = folio_get_private(folio);
+	struct btrfs_subpage *subpage;
+	unsigned int start_bit;
+	unsigned int nbits;
+	unsigned long flags;
 
 	if (!IS_ENABLED(CONFIG_BTRFS_ASSERT))
 		return;
 
-	ASSERT(!folio_test_dirty(folio));
-	if (!btrfs_is_subpage(fs_info, folio->mapping))
+	if (!btrfs_is_subpage(fs_info, folio->mapping)) {
+		ASSERT(!folio_test_dirty(folio));
 		return;
+	}
 
-	ASSERT(folio_test_private(folio) && folio_get_private(folio));
-	ASSERT(subpage_test_bitmap_all_zero(fs_info, subpage, dirty));
+	start_bit = subpage_calc_start_bit(fs_info, folio, dirty, start, len);
+	nbits = len >> fs_info->sectorsize_bits;
+	subpage = folio_get_private(folio);
+	ASSERT(subpage);
+	spin_lock_irqsave(&subpage->lock, flags);
+	if (unlikely(!bitmap_test_range_all_zero(subpage->bitmaps, start_bit, nbits))) {
+		SUBPAGE_DUMP_BITMAP(fs_info, folio, dirty, start, len);
+		ASSERT(bitmap_test_range_all_zero(subpage->bitmaps, start_bit, nbits));
+	}
+	ASSERT(bitmap_test_range_all_zero(subpage->bitmaps, start_bit, nbits));
+	spin_unlock_irqrestore(&subpage->lock, flags);
 }
 
 /*
- * Handle different locked pages with different page sizes:
- *
- * - Page locked by plain lock_page()
- *   It should not have any subpage::writers count.
- *   Can be unlocked by unlock_page().
- *   This is the most common locked page for __extent_writepage() called
- *   inside extent_write_cache_pages().
- *   Rarer cases include the @locked_page from extent_write_locked_range().
+ * This is for folio already locked by plain lock_page()/folio_lock(), which
+ * doesn't have any subpage awareness.
  *
- * - Page locked by lock_delalloc_pages()
- *   There is only one caller, all pages except @locked_page for
- *   extent_write_locked_range().
- *   In this case, we have to call subpage helper to handle the case.
+ * This populates the involved subpage ranges so that subpage helpers can
+ * properly unlock them.
  */
-void btrfs_folio_unlock_writer(struct btrfs_fs_info *fs_info,
-			       struct folio *folio, u64 start, u32 len)
+void btrfs_folio_set_lock(const struct btrfs_fs_info *fs_info,
+			  struct folio *folio, u64 start, u32 len)
 {
 	struct btrfs_subpage *subpage;
+	unsigned long flags;
+	unsigned int start_bit;
+	unsigned int nbits;
+	int ret;
 
 	ASSERT(folio_test_locked(folio));
-	/* For non-subpage case, we just unlock the page */
-	if (!btrfs_is_subpage(fs_info, folio->mapping)) {
-		folio_unlock(folio);
+	if (unlikely(!fs_info) || !btrfs_is_subpage(fs_info, folio->mapping))
 		return;
-	}
 
-	ASSERT(folio_test_private(folio) && folio_get_private(folio));
 	subpage = folio_get_private(folio);
-
-	/*
-	 * For subpage case, there are two types of locked page.  With or
-	 * without writers number.
-	 *
-	 * Since we own the page lock, no one else could touch subpage::writers
-	 * and we are safe to do several atomic operations without spinlock.
-	 */
-	if (atomic_read(&subpage->writers) == 0) {
-		/* No writers, locked by plain lock_page() */
-		folio_unlock(folio);
-		return;
+	start_bit = subpage_calc_start_bit(fs_info, folio, locked, start, len);
+	nbits = len >> fs_info->sectorsize_bits;
+	spin_lock_irqsave(&subpage->lock, flags);
+	/* Target range should not yet be locked. */
+	if (unlikely(!bitmap_test_range_all_zero(subpage->bitmaps, start_bit, nbits))) {
+		SUBPAGE_DUMP_BITMAP(fs_info, folio, locked, start, len);
+		ASSERT(bitmap_test_range_all_zero(subpage->bitmaps, start_bit, nbits));
 	}
-
-	/* Have writers, use proper subpage helper to end it */
-	btrfs_folio_end_writer_lock(fs_info, folio, start, len);
+	bitmap_set(subpage->bitmaps, start_bit, nbits);
+	ret = atomic_add_return(nbits, &subpage->nr_locked);
+	ASSERT(ret <= fs_info->sectors_per_page);
+	spin_unlock_irqrestore(&subpage->lock, flags);
 }
 
-#define GET_SUBPAGE_BITMAP(subpage, subpage_info, name, dst)		\
-	bitmap_cut(dst, subpage->bitmaps, 0,				\
-		   subpage_info->name##_offset, subpage_info->bitmap_nr_bits)
-
 void __cold btrfs_subpage_dump_bitmap(const struct btrfs_fs_info *fs_info,
 				      struct folio *folio, u64 start, u32 len)
 {
-	struct btrfs_subpage_info *subpage_info = fs_info->subpage_info;
 	struct btrfs_subpage *subpage;
+	const u32 sectors_per_page = fs_info->sectors_per_page;
 	unsigned long uptodate_bitmap;
-	unsigned long error_bitmap;
 	unsigned long dirty_bitmap;
 	unsigned long writeback_bitmap;
 	unsigned long ordered_bitmap;
 	unsigned long checked_bitmap;
+	unsigned long locked_bitmap;
 	unsigned long flags;
 
 	ASSERT(folio_test_private(folio) && folio_get_private(folio));
-	ASSERT(subpage_info);
+	ASSERT(sectors_per_page > 1);
 	subpage = folio_get_private(folio);
 
 	spin_lock_irqsave(&subpage->lock, flags);
-	GET_SUBPAGE_BITMAP(subpage, subpage_info, uptodate, &uptodate_bitmap);
-	GET_SUBPAGE_BITMAP(subpage, subpage_info, dirty, &dirty_bitmap);
-	GET_SUBPAGE_BITMAP(subpage, subpage_info, writeback, &writeback_bitmap);
-	GET_SUBPAGE_BITMAP(subpage, subpage_info, ordered, &ordered_bitmap);
-	GET_SUBPAGE_BITMAP(subpage, subpage_info, checked, &checked_bitmap);
+	GET_SUBPAGE_BITMAP(subpage, fs_info, uptodate, &uptodate_bitmap);
+	GET_SUBPAGE_BITMAP(subpage, fs_info, dirty, &dirty_bitmap);
+	GET_SUBPAGE_BITMAP(subpage, fs_info, writeback, &writeback_bitmap);
+	GET_SUBPAGE_BITMAP(subpage, fs_info, ordered, &ordered_bitmap);
+	GET_SUBPAGE_BITMAP(subpage, fs_info, checked, &checked_bitmap);
+	GET_SUBPAGE_BITMAP(subpage, fs_info, locked, &locked_bitmap);
 	spin_unlock_irqrestore(&subpage->lock, flags);
 
 	dump_page(folio_page(folio, 0), "btrfs subpage dump");
 	btrfs_warn(fs_info,
-"start=%llu len=%u page=%llu, bitmaps uptodate=%*pbl error=%*pbl dirty=%*pbl writeback=%*pbl ordered=%*pbl checked=%*pbl",
+"start=%llu len=%u page=%llu, bitmaps uptodate=%*pbl dirty=%*pbl locked=%*pbl writeback=%*pbl ordered=%*pbl checked=%*pbl",
 		    start, len, folio_pos(folio),
-		    subpage_info->bitmap_nr_bits, &uptodate_bitmap,
-		    subpage_info->bitmap_nr_bits, &error_bitmap,
-		    subpage_info->bitmap_nr_bits, &dirty_bitmap,
-		    subpage_info->bitmap_nr_bits, &writeback_bitmap,
-		    subpage_info->bitmap_nr_bits, &ordered_bitmap,
-		    subpage_info->bitmap_nr_bits, &checked_bitmap);
+		    sectors_per_page, &uptodate_bitmap,
+		    sectors_per_page, &dirty_bitmap,
+		    sectors_per_page, &locked_bitmap,
+		    sectors_per_page, &writeback_bitmap,
+		    sectors_per_page, &ordered_bitmap,
+		    sectors_per_page, &checked_bitmap);
+}
+
+void btrfs_get_subpage_dirty_bitmap(struct btrfs_fs_info *fs_info,
+				    struct folio *folio,
+				    unsigned long *ret_bitmap)
+{
+	struct btrfs_subpage *subpage;
+	unsigned long flags;
+
+	ASSERT(folio_test_private(folio) && folio_get_private(folio));
+	ASSERT(fs_info->sectors_per_page > 1);
+	subpage = folio_get_private(folio);
+
+	spin_lock_irqsave(&subpage->lock, flags);
+	GET_SUBPAGE_BITMAP(subpage, fs_info, dirty, ret_bitmap);
+	spin_unlock_irqrestore(&subpage->lock, flags);
 }