1 files changed, 279 insertions, 259 deletions

diff --git a/fs/btrfs/subpage.c b/fs/btrfs/subpage.c
index 54736f6238e6..d4f019233493 100644
--- a/fs/btrfs/subpage.c
+++ b/fs/btrfs/subpage.c
@@ -2,12 +2,11 @@
 
 #include <linux/slab.h>
 #include "messages.h"
-#include "ctree.h"
 #include "subpage.h"
 #include "btrfs_inode.h"
 
 /*
- * Subpage (sectorsize < PAGE_SIZE) support overview:
+ * Subpage (block size < folio size) support overview:
  *
  * Limitations:
  *
@@ -64,64 +63,15 @@
  *   This means a slightly higher tree locking latency.
  */
 
-bool btrfs_is_subpage(const struct btrfs_fs_info *fs_info, struct address_space *mapping)
-{
-	if (fs_info->sectorsize >= PAGE_SIZE)
-		return false;
-
-	/*
-	 * Only data pages (either through DIO or compression) can have no
-	 * 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))
-		return true;
-
-	/*
-	 * Now the only remaining case is metadata, which we only go subpage
-	 * routine if nodesize < PAGE_SIZE.
-	 */
-	if (fs_info->nodesize < PAGE_SIZE)
-		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->locked_offset = cur;
-	cur += nr_bits;
-
-	subpage_info->total_nr_bits = cur;
-}
-
 int btrfs_attach_subpage(const struct btrfs_fs_info *fs_info,
 			 struct folio *folio, enum btrfs_subpage_type type)
 {
 	struct btrfs_subpage *subpage;
 
+	/* For metadata we don't support large folio yet. */
+	if (type == BTRFS_SUBPAGE_METADATA)
+		ASSERT(!folio_test_large(folio));
+
 	/*
 	 * We have cases like a dummy extent buffer page, which is not mapped
 	 * and doesn't need to be locked.
@@ -130,10 +80,14 @@ int btrfs_attach_subpage(const struct btrfs_fs_info *fs_info,
 		ASSERT(folio_test_locked(folio));
 
 	/* Either not subpage, or the folio already has private attached. */
-	if (!btrfs_is_subpage(fs_info, folio->mapping) || folio_test_private(folio))
+	if (folio_test_private(folio))
+		return 0;
+	if (type == BTRFS_SUBPAGE_METADATA && !btrfs_meta_is_subpage(fs_info))
+		return 0;
+	if (type == BTRFS_SUBPAGE_DATA && !btrfs_is_subpage(fs_info, folio))
 		return 0;
 
-	subpage = btrfs_alloc_subpage(fs_info, type);
+	subpage = btrfs_alloc_subpage(fs_info, folio_size(folio), type);
 	if (IS_ERR(subpage))
 		return  PTR_ERR(subpage);
 
@@ -141,12 +95,17 @@ int btrfs_attach_subpage(const struct btrfs_fs_info *fs_info,
 	return 0;
 }
 
-void btrfs_detach_subpage(const struct btrfs_fs_info *fs_info, struct folio *folio)
+void btrfs_detach_subpage(const struct btrfs_fs_info *fs_info, struct folio *folio,
+			  enum btrfs_subpage_type type)
 {
 	struct btrfs_subpage *subpage;
 
 	/* Either not subpage, or the folio already has private attached. */
-	if (!btrfs_is_subpage(fs_info, folio->mapping) || !folio_test_private(folio))
+	if (!folio_test_private(folio))
+		return;
+	if (type == BTRFS_SUBPAGE_METADATA && !btrfs_meta_is_subpage(fs_info))
+		return;
+	if (type == BTRFS_SUBPAGE_DATA && !btrfs_is_subpage(fs_info, folio))
 		return;
 
 	subpage = folio_detach_private(folio);
@@ -155,26 +114,25 @@ void btrfs_detach_subpage(const struct btrfs_fs_info *fs_info, struct folio *fol
 }
 
 struct btrfs_subpage *btrfs_alloc_subpage(const struct btrfs_fs_info *fs_info,
-					  enum btrfs_subpage_type type)
+				size_t fsize, enum btrfs_subpage_type type)
 {
 	struct btrfs_subpage *ret;
 	unsigned int real_size;
 
-	ASSERT(fs_info->sectorsize < PAGE_SIZE);
+	ASSERT(fs_info->sectorsize < fsize);
 
 	real_size = struct_size(ret, bitmaps,
-			BITS_TO_LONGS(fs_info->subpage_info->total_nr_bits));
+			BITS_TO_LONGS(btrfs_bitmap_nr_max *
+				      (fsize >> fs_info->sectorsize_bits)));
 	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;
 }
 
@@ -196,7 +154,7 @@ void btrfs_folio_inc_eb_refs(const struct btrfs_fs_info *fs_info, struct folio *
 {
 	struct btrfs_subpage *subpage;
 
-	if (!btrfs_is_subpage(fs_info, folio->mapping))
+	if (!btrfs_meta_is_subpage(fs_info))
 		return;
 
 	ASSERT(folio_test_private(folio) && folio->mapping);
@@ -210,7 +168,7 @@ void btrfs_folio_dec_eb_refs(const struct btrfs_fs_info *fs_info, struct folio *
 {
 	struct btrfs_subpage *subpage;
 
-	if (!btrfs_is_subpage(fs_info, folio->mapping))
+	if (!btrfs_meta_is_subpage(fs_info))
 		return;
 
 	ASSERT(folio_test_private(folio) && folio->mapping);
@@ -224,9 +182,6 @@ void btrfs_folio_dec_eb_refs(const struct btrfs_fs_info *fs_info, struct folio *
 static void btrfs_subpage_assert(const struct btrfs_fs_info *fs_info,
 				 struct folio *folio, u64 start, u32 len)
 {
-	/* For subpage support, the folio must be single page. */
-	ASSERT(folio_order(folio) == 0);
-
 	/* Basic checks */
 	ASSERT(folio_test_private(folio) && folio_get_private(folio));
 	ASSERT(IS_ALIGNED(start, fs_info->sectorsize) &&
@@ -237,75 +192,21 @@ static void btrfs_subpage_assert(const struct btrfs_fs_info *fs_info,
 	 */
 	if (folio->mapping)
 		ASSERT(folio_pos(folio) <= start &&
-		       start + len <= folio_pos(folio) + PAGE_SIZE);
+		       start + len <= folio_pos(folio) + folio_size(folio));
 }
 
 #define subpage_calc_start_bit(fs_info, folio, name, start, len)	\
 ({									\
-	unsigned int start_bit;						\
+	unsigned int __start_bit;					\
+	const unsigned int blocks_per_folio =				\
+			   btrfs_blocks_per_folio(fs_info, folio);	\
 									\
 	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;							\
+	__start_bit = offset_in_folio(folio, start) >> fs_info->sectorsize_bits; \
+	__start_bit += blocks_per_folio * btrfs_bitmap_nr_##name;	\
+	__start_bit;							\
 })
 
-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 start_bit = subpage_calc_start_bit(fs_info, folio, locked, start, len);
-	const int nbits = len >> fs_info->sectorsize_bits;
-	unsigned long flags;
-
-
-	btrfs_subpage_assert(fs_info, folio, start, len);
-
-	spin_lock_irqsave(&subpage->lock, flags);
-	/*
-	 * Even though it's just for reading the page, no one should have
-	 * locked the subpage range.
-	 */
-	ASSERT(bitmap_test_range_all_zero(subpage->bitmaps, start_bit, nbits));
-	bitmap_set(subpage->bitmaps, start_bit, nbits);
-	atomic_add(nbits, &subpage->readers);
-	spin_unlock_irqrestore(&subpage->lock, flags);
-}
-
-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 start_bit = subpage_calc_start_bit(fs_info, folio, locked, start, len);
-	const int nbits = len >> fs_info->sectorsize_bits;
-	unsigned long flags;
-	bool is_data;
-	bool last;
-
-	btrfs_subpage_assert(fs_info, folio, start, len);
-	is_data = is_data_inode(folio->mapping->host);
-
-	spin_lock_irqsave(&subpage->lock, flags);
-
-	/* The range should have already been locked. */
-	ASSERT(bitmap_test_range_all_set(subpage->bitmaps, start_bit, nbits));
-	ASSERT(atomic_read(&subpage->readers) >= nbits);
-
-	bitmap_clear(subpage->bitmaps, start_bit, 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);
-	spin_unlock_irqrestore(&subpage->lock, flags);
-}
-
 static void btrfs_subpage_clamp_range(struct folio *folio, u64 *start, u32 *len)
 {
 	u64 orig_start = *start;
@@ -320,37 +221,19 @@ static void btrfs_subpage_clamp_range(struct folio *folio, u64 *start, u32 *len)
 	if (folio_pos(folio) >= orig_start + orig_len)
 		*len = 0;
 	else
-		*len = min_t(u64, folio_pos(folio) + PAGE_SIZE,
+		*len = min_t(u64, folio_pos(folio) + folio_size(folio),
 			     orig_start + orig_len) - *start;
 }
 
-static 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 start_bit = subpage_calc_start_bit(fs_info, folio, locked, start, len);
-	const int nbits = (len >> fs_info->sectorsize_bits);
-	unsigned long flags;
-	int ret;
-
-	btrfs_subpage_assert(fs_info, folio, start, len);
-
-	spin_lock_irqsave(&subpage->lock, flags);
-	ASSERT(atomic_read(&subpage->readers) == 0);
-	ASSERT(bitmap_test_range_all_zero(subpage->bitmaps, start_bit, nbits));
-	bitmap_set(subpage->bitmaps, start_bit, nbits);
-	ret = atomic_add_return(nbits, &subpage->writers);
-	ASSERT(ret == nbits);
-	spin_unlock_irqrestore(&subpage->lock, flags);
-}
-
-static 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);
@@ -361,70 +244,124 @@ static bool btrfs_subpage_end_and_test_writer(const struct btrfs_fs_info *fs_inf
 	 * 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);
-	/* The target range should have been locked. */
-	ASSERT(bitmap_test_range_all_set(subpage->bitmaps, start_bit, nbits));
-	bitmap_clear(subpage->bitmaps, start_bit, nbits);
-	last = 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:
+ *
+ * - Non-subpage folio
+ *   Just unlock it.
  *
- * Return -EAGAIN if the page is not properly initialized.
- * Return 0 with the page locked, and writer counter updated.
+ * - 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.
  *
- * 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 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)
 {
-	if (unlikely(!fs_info) || !btrfs_is_subpage(fs_info, folio->mapping)) {
-		folio_lock(folio);
-		return 0;
+	struct btrfs_subpage *subpage = folio_get_private(folio);
+
+	ASSERT(folio_test_locked(folio));
+
+	if (unlikely(!fs_info) || !btrfs_is_subpage(fs_info, folio)) {
+		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 unsigned int blocks_per_folio = btrfs_blocks_per_folio(fs_info, folio);
+	const int start_bit = blocks_per_folio * btrfs_bitmap_nr_locked;
+	unsigned long flags;
+	bool last = false;
+	int cleared = 0;
+	int bit;
+
+	if (!btrfs_is_subpage(fs_info, folio)) {
 		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;
+	}
+
+	spin_lock_irqsave(&subpage->lock, flags);
+	for_each_set_bit(bit, &bitmap, blocks_per_folio) {
+		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)		\
+#define subpage_test_bitmap_all_set(fs_info, folio, name)		\
+({									\
+	struct btrfs_subpage *subpage = folio_get_private(folio);	\
+	const unsigned int blocks_per_folio =				\
+				btrfs_blocks_per_folio(fs_info, folio); \
+									\
 	bitmap_test_range_all_set(subpage->bitmaps,			\
-			fs_info->subpage_info->name##_offset,		\
-			fs_info->subpage_info->bitmap_nr_bits)
+			blocks_per_folio * btrfs_bitmap_nr_##name,	\
+			blocks_per_folio);				\
+})
 
-#define subpage_test_bitmap_all_zero(fs_info, subpage, name)		\
+#define subpage_test_bitmap_all_zero(fs_info, folio, name)		\
+({									\
+	struct btrfs_subpage *subpage = folio_get_private(folio);	\
+	const unsigned int blocks_per_folio =				\
+				btrfs_blocks_per_folio(fs_info, folio); \
+									\
 	bitmap_test_range_all_zero(subpage->bitmaps,			\
-			fs_info->subpage_info->name##_offset,		\
-			fs_info->subpage_info->bitmap_nr_bits)
+			blocks_per_folio * btrfs_bitmap_nr_##name,	\
+			blocks_per_folio);				\
+})
 
 void btrfs_subpage_set_uptodate(const struct btrfs_fs_info *fs_info,
 				struct folio *folio, u64 start, u32 len)
@@ -436,7 +373,7 @@ void btrfs_subpage_set_uptodate(const struct btrfs_fs_info *fs_info,
 
 	spin_lock_irqsave(&subpage->lock, flags);
 	bitmap_set(subpage->bitmaps, start_bit, len >> fs_info->sectorsize_bits);
-	if (subpage_test_bitmap_all_set(fs_info, subpage, uptodate))
+	if (subpage_test_bitmap_all_set(fs_info, folio, uptodate))
 		folio_mark_uptodate(folio);
 	spin_unlock_irqrestore(&subpage->lock, flags);
 }
@@ -490,7 +427,7 @@ bool btrfs_subpage_clear_and_test_dirty(const struct btrfs_fs_info *fs_info,
 
 	spin_lock_irqsave(&subpage->lock, flags);
 	bitmap_clear(subpage->bitmaps, start_bit, len >> fs_info->sectorsize_bits);
-	if (subpage_test_bitmap_all_zero(fs_info, subpage, dirty))
+	if (subpage_test_bitmap_all_zero(fs_info, folio, dirty))
 		last = true;
 	spin_unlock_irqrestore(&subpage->lock, flags);
 	return last;
@@ -531,7 +468,7 @@ void btrfs_subpage_clear_writeback(const struct btrfs_fs_info *fs_info,
 
 	spin_lock_irqsave(&subpage->lock, flags);
 	bitmap_clear(subpage->bitmaps, start_bit, len >> fs_info->sectorsize_bits);
-	if (subpage_test_bitmap_all_zero(fs_info, subpage, writeback)) {
+	if (subpage_test_bitmap_all_zero(fs_info, folio, writeback)) {
 		ASSERT(folio_test_writeback(folio));
 		folio_end_writeback(folio);
 	}
@@ -562,7 +499,7 @@ void btrfs_subpage_clear_ordered(const struct btrfs_fs_info *fs_info,
 
 	spin_lock_irqsave(&subpage->lock, flags);
 	bitmap_clear(subpage->bitmaps, start_bit, len >> fs_info->sectorsize_bits);
-	if (subpage_test_bitmap_all_zero(fs_info, subpage, ordered))
+	if (subpage_test_bitmap_all_zero(fs_info, folio, ordered))
 		folio_clear_ordered(folio);
 	spin_unlock_irqrestore(&subpage->lock, flags);
 }
@@ -577,7 +514,7 @@ void btrfs_subpage_set_checked(const struct btrfs_fs_info *fs_info,
 
 	spin_lock_irqsave(&subpage->lock, flags);
 	bitmap_set(subpage->bitmaps, start_bit, len >> fs_info->sectorsize_bits);
-	if (subpage_test_bitmap_all_set(fs_info, subpage, checked))
+	if (subpage_test_bitmap_all_set(fs_info, folio, checked))
 		folio_set_checked(folio);
 	spin_unlock_irqrestore(&subpage->lock, flags);
 }
@@ -633,7 +570,7 @@ void btrfs_folio_set_##name(const struct btrfs_fs_info *fs_info,	\
 			    struct folio *folio, u64 start, u32 len)	\
 {									\
 	if (unlikely(!fs_info) ||					\
-	    !btrfs_is_subpage(fs_info, folio->mapping)) {		\
+	    !btrfs_is_subpage(fs_info, folio)) {			\
 		folio_set_func(folio);					\
 		return;							\
 	}								\
@@ -643,7 +580,7 @@ void btrfs_folio_clear_##name(const struct btrfs_fs_info *fs_info,	\
 			      struct folio *folio, u64 start, u32 len)	\
 {									\
 	if (unlikely(!fs_info) ||					\
-	    !btrfs_is_subpage(fs_info, folio->mapping)) {		\
+	    !btrfs_is_subpage(fs_info, folio)) {			\
 		folio_clear_func(folio);				\
 		return;							\
 	}								\
@@ -653,7 +590,7 @@ bool btrfs_folio_test_##name(const struct btrfs_fs_info *fs_info,	\
 			     struct folio *folio, u64 start, u32 len)	\
 {									\
 	if (unlikely(!fs_info) ||					\
-	    !btrfs_is_subpage(fs_info, folio->mapping))			\
+	    !btrfs_is_subpage(fs_info, folio))				\
 		return folio_test_func(folio);				\
 	return btrfs_subpage_test_##name(fs_info, folio, start, len);	\
 }									\
@@ -661,7 +598,7 @@ void btrfs_folio_clamp_set_##name(const struct btrfs_fs_info *fs_info,	\
 				  struct folio *folio, u64 start, u32 len) \
 {									\
 	if (unlikely(!fs_info) ||					\
-	    !btrfs_is_subpage(fs_info, folio->mapping)) {		\
+	    !btrfs_is_subpage(fs_info, folio)) {			\
 		folio_set_func(folio);					\
 		return;							\
 	}								\
@@ -672,7 +609,7 @@ void btrfs_folio_clamp_clear_##name(const struct btrfs_fs_info *fs_info, \
 				    struct folio *folio, u64 start, u32 len) \
 {									\
 	if (unlikely(!fs_info) ||					\
-	    !btrfs_is_subpage(fs_info, folio->mapping)) {		\
+	    !btrfs_is_subpage(fs_info, folio)) {			\
 		folio_clear_func(folio);				\
 		return;							\
 	}								\
@@ -683,10 +620,32 @@ bool btrfs_folio_clamp_test_##name(const struct btrfs_fs_info *fs_info,	\
 				   struct folio *folio, u64 start, u32 len) \
 {									\
 	if (unlikely(!fs_info) ||					\
-	    !btrfs_is_subpage(fs_info, folio->mapping))			\
+	    !btrfs_is_subpage(fs_info, folio))				\
 		return folio_test_func(folio);				\
 	btrfs_subpage_clamp_range(folio, &start, &len);			\
 	return btrfs_subpage_test_##name(fs_info, folio, start, len);	\
+}									\
+void btrfs_meta_folio_set_##name(struct folio *folio, const struct extent_buffer *eb) \
+{									\
+	if (!btrfs_meta_is_subpage(eb->fs_info)) {			\
+		folio_set_func(folio);					\
+		return;							\
+	}								\
+	btrfs_subpage_set_##name(eb->fs_info, folio, eb->start, eb->len); \
+}									\
+void btrfs_meta_folio_clear_##name(struct folio *folio, const struct extent_buffer *eb) \
+{									\
+	if (!btrfs_meta_is_subpage(eb->fs_info)) {			\
+		folio_clear_func(folio);				\
+		return;							\
+	}								\
+	btrfs_subpage_clear_##name(eb->fs_info, folio, eb->start, eb->len); \
+}									\
+bool btrfs_meta_folio_test_##name(struct folio *folio, const struct extent_buffer *eb) \
+{									\
+	if (!btrfs_meta_is_subpage(eb->fs_info))			\
+		return folio_test_func(folio);				\
+	return btrfs_subpage_test_##name(eb->fs_info, folio, eb->start, eb->len); \
 }
 IMPLEMENT_BTRFS_PAGE_OPS(uptodate, folio_mark_uptodate, folio_clear_uptodate,
 			 folio_test_uptodate);
@@ -699,110 +658,171 @@ 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(fs_info, folio, name, dst)			\
+{									\
+	const unsigned int blocks_per_folio =				\
+				btrfs_blocks_per_folio(fs_info, folio);	\
+	const struct btrfs_subpage *subpage = folio_get_private(folio);	\
+									\
+	ASSERT(blocks_per_folio <= BITS_PER_LONG);			\
+	*dst = bitmap_read(subpage->bitmaps,				\
+			   blocks_per_folio * btrfs_bitmap_nr_##name,	\
+			   blocks_per_folio);				\
+}
+
+#define SUBPAGE_DUMP_BITMAP(fs_info, folio, name, start, len)		\
+{									\
+	unsigned long bitmap;						\
+	const unsigned int blocks_per_folio =				\
+				btrfs_blocks_per_folio(fs_info, folio);	\
+									\
+	GET_SUBPAGE_BITMAP(fs_info, folio, name, &bitmap);		\
+	btrfs_warn(fs_info,						\
+	"dumpping bitmap start=%llu len=%u folio=%llu " #name "_bitmap=%*pbl", \
+		   start, len, folio_pos(folio),			\
+		   blocks_per_folio, &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)) {
+		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:
+ * This is for folio already locked by plain lock_page()/folio_lock(), which
+ * doesn't have any subpage awareness.
  *
- * - 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().
- *
- * - 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))
 		return;
-	}
 
-	ASSERT(folio_test_private(folio) && folio_get_private(folio));
 	subpage = folio_get_private(folio);
+	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));
+	}
+	bitmap_set(subpage->bitmaps, start_bit, nbits);
+	ret = atomic_add_return(nbits, &subpage->nr_locked);
+	ASSERT(ret <= btrfs_blocks_per_folio(fs_info, folio));
+	spin_unlock_irqrestore(&subpage->lock, flags);
+}
 
-	/*
-	 * 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;
+/*
+ * Clear the dirty flag for the folio.
+ *
+ * If the affected folio is no longer dirty, return true. Otherwise return false.
+ */
+bool btrfs_meta_folio_clear_and_test_dirty(struct folio *folio, const struct extent_buffer *eb)
+{
+	bool last;
+
+	if (!btrfs_meta_is_subpage(eb->fs_info)) {
+		folio_clear_dirty_for_io(folio);
+		return true;
 	}
 
-	/* Have writers, use proper subpage helper to end it */
-	btrfs_folio_end_writer_lock(fs_info, folio, start, len);
+	last = btrfs_subpage_clear_and_test_dirty(eb->fs_info, folio, eb->start, eb->len);
+	if (last) {
+		folio_clear_dirty_for_io(folio);
+		return true;
+	}
+	return false;
 }
 
-#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 unsigned int blocks_per_folio = btrfs_blocks_per_folio(fs_info, folio);
 	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(blocks_per_folio > 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, subpage_info, locked, &checked_bitmap);
+	GET_SUBPAGE_BITMAP(fs_info, folio, uptodate, &uptodate_bitmap);
+	GET_SUBPAGE_BITMAP(fs_info, folio, dirty, &dirty_bitmap);
+	GET_SUBPAGE_BITMAP(fs_info, folio, writeback, &writeback_bitmap);
+	GET_SUBPAGE_BITMAP(fs_info, folio, ordered, &ordered_bitmap);
+	GET_SUBPAGE_BITMAP(fs_info, folio, checked, &checked_bitmap);
+	GET_SUBPAGE_BITMAP(fs_info, folio, 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);
+		    blocks_per_folio, &uptodate_bitmap,
+		    blocks_per_folio, &dirty_bitmap,
+		    blocks_per_folio, &locked_bitmap,
+		    blocks_per_folio, &writeback_bitmap,
+		    blocks_per_folio, &ordered_bitmap,
+		    blocks_per_folio, &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(btrfs_blocks_per_folio(fs_info, folio) > 1);
+	subpage = folio_get_private(folio);
+
+	spin_lock_irqsave(&subpage->lock, flags);
+	GET_SUBPAGE_BITMAP(fs_info, folio, dirty, ret_bitmap);
+	spin_unlock_irqrestore(&subpage->lock, flags);
 }