Skip to contentSkip to conversation

btrfs: Allow btrfs_truncate_block() to fallback to nocow for data space reservation

PATCH Superseded Qu Wenruo 3 participants 3 messages Jun 3, 2020 – Jun 17, 2020 Btrfs file system

v2 of 7 Superseded Latest is v7 →

lore
Share & tools
lore

Qu Wenruo

author
#0 lore raw

PATCH v2 1/1 btrfs: Allow btrfs_truncate_block() to fallback to nocow for data space reservation

[BUG]
When the data space is exhausted, even the inode has NOCOW attribute, btrfs will still refuse to truncate unaligned range due to ENOSPC.

The following script can reproduce it pretty easily:
#!/bin/bash

dev=/dev/test/test
mnt=/mnt/btrfs

umount $dev &> /dev/null
umount $mnt&> /dev/null

mkfs.btrfs -f $dev -b 1G
mount -o nospace_cache $dev $mnt
touch $mnt/foobar
chattr +C $mnt/foobar

xfs_io -f -c "pwrite -b 4k 0 4k" $mnt/foobar > /dev/null xfs_io -f -c "pwrite -b 4k 0 1G" $mnt/padding &> /dev/null sync

xfs_io -c "fpunch 0 2k" $mnt/foobar
umount $mnt

Current btrfs will fail at the fpunch part.

[CAUSE]
Because btrfs_truncate_block() always reserve space without checking the NOCOW attribute.

Since the writeback path follows NOCOW bit, we only need to bother the space reservation code in btrfs_truncate_block().

[FIX]
Make btrfs_truncate_block() to follow btrfs_buffered_write() to try to reserve data space first, and falls back to NOCOW check only when we don't have enough space.

Such always-try-reserve is an optimization introduced in btrfs_buffered_write(), to avoid expensive btrfs_check_can_nocow() call.

Since now check_can_nocow() is needed outside of inode.c, also export it and rename it to btrfs_check_can_nocow().

---
Changelog:
v2:
- Rebased to misc-next
Only one minor conflict in ctree.h

---
 fs/btrfs/ctree.h |  2 ++
 fs/btrfs/file.c  | 10 +++++-----
 fs/btrfs/inode.c | 41 ++++++++++++++++++++++++++++++++++-------
 3 files changed, 41 insertions(+), 12 deletions(-)
fs/btrfs/ctree.h +2 -0
index 161533040978..40e8c8170d39 100644
@@ -2984,6 +2984,8 @@ int btrfs_dirty_pages(struct inode *inode, struct page **pages,
 		      size_t num_pages, loff_t pos, size_t write_bytes,
 		      struct extent_state **cached);
 int btrfs_fdatawrite_range(struct inode *inode, loff_t start, loff_t end);
+int btrfs_check_can_nocow(struct btrfs_inode *inode, loff_t pos,
+			  size_t *write_bytes);
 
 /* tree-defrag.c */
 int btrfs_defrag_leaves(struct btrfs_trans_handle *trans,
fs/btrfs/file.c +5 -5
index fde125616687..a298803e2b08 100644
@@ -1532,8 +1532,8 @@ lock_and_cleanup_extent_if_need(struct btrfs_inode *inode, struct page **pages,
 	return ret;
 }
 
-static noinline int check_can_nocow(struct btrfs_inode *inode, loff_t pos,
-				    size_t *write_bytes)
+int btrfs_check_can_nocow(struct btrfs_inode *inode, loff_t pos,
+			  size_t *write_bytes)
 {
 	struct btrfs_fs_info *fs_info = inode->root->fs_info;
 	struct btrfs_root *root = inode->root;
@@ -1632,8 +1632,8 @@ static noinline ssize_t btrfs_buffered_write(struct kiocb *iocb,
 		if (ret < 0) {
 			if ((BTRFS_I(inode)->flags & (BTRFS_INODE_NODATACOW |
 						      BTRFS_INODE_PREALLOC)) &&
-			    check_can_nocow(BTRFS_I(inode), pos,
-					&write_bytes) > 0) {
+			    btrfs_check_can_nocow(BTRFS_I(inode), pos,
+						  &write_bytes) > 0) {
 				/*
 				 * For nodata cow case, no need to reserve
 				 * data space.
@@ -1950,7 +1950,7 @@ static ssize_t btrfs_file_write_iter(struct kiocb *iocb,
 		 */
 		if (!(BTRFS_I(inode)->flags & (BTRFS_INODE_NODATACOW |
 					      BTRFS_INODE_PREALLOC)) ||
-		    check_can_nocow(BTRFS_I(inode), pos, &count) <= 0) {
+		    btrfs_check_can_nocow(BTRFS_I(inode), pos, &count) <= 0) {
 			inode_unlock(inode);
 			return -EAGAIN;
 		}
fs/btrfs/inode.c +34 -7

66 lines not on this page · raw message

Signature · 1 line
2.26.2
#1 lore raw
Quoted 64 lines · Qu Wenruo

[BUG]
When the data space is exhausted, even the inode has NOCOW attribute, btrfs will still refuse to truncate unaligned range due to ENOSPC.

The following script can reproduce it pretty easily:
#!/bin/bash

dev=/dev/test/test
mnt=/mnt/btrfs

umount $dev &> /dev/null
umount $mnt&> /dev/null

mkfs.btrfs -f $dev -b 1G
mount -o nospace_cache $dev $mnt
touch $mnt/foobar
chattr +C $mnt/foobar

xfs_io -f -c "pwrite -b 4k 0 4k" $mnt/foobar > /dev/null xfs_io -f -c "pwrite -b 4k 0 1G" $mnt/padding &> /dev/null sync

xfs_io -c "fpunch 0 2k" $mnt/foobar
umount $mnt

Current btrfs will fail at the fpunch part.

[CAUSE]
Because btrfs_truncate_block() always reserve space without checking the NOCOW attribute.

Since the writeback path follows NOCOW bit, we only need to bother the space reservation code in btrfs_truncate_block().

[FIX]
Make btrfs_truncate_block() to follow btrfs_buffered_write() to try to reserve data space first, and falls back to NOCOW check only when we don't have enough space.

Such always-try-reserve is an optimization introduced in btrfs_buffered_write(), to avoid expensive btrfs_check_can_nocow() call.

Since now check_can_nocow() is needed outside of inode.c, also export it and rename it to btrfs_check_can_nocow().

---
Changelog:
v2:
- Rebased to misc-next
Only one minor conflict in ctree.h

---
 fs/btrfs/ctree.h |  2 ++
 fs/btrfs/file.c  | 10 +++++-----
 fs/btrfs/inode.c | 41 ++++++++++++++++++++++++++++++++++-------
 3 files changed, 41 insertions(+), 12 deletions(-)
fs/btrfs/ctree.h +0 -0
index 161533040978..40e8c8170d39 100644

fs/btrfs/ctree.h int btrfs_dirty_pages(struct inode *inode, struct page **pages, · line 2988

Quoted 5 lines
fs/btrfs/ctree.h +2 -0
@@ -2984,6 +2984,8 @@ int btrfs_dirty_pages(struct inode *inode, struct page **pages,
 		      size_t num_pages, loff_t pos, size_t write_bytes,
 		      struct extent_state **cached);
 int btrfs_fdatawrite_range(struct inode *inode, loff_t start, loff_t end);
+int btrfs_check_can_nocow(struct btrfs_inode *inode, loff_t pos,
+			  size_t *write_bytes);

This does not apply anymore after Filipe's aio nowait fixes, can you please rebase it on top of current misc-next?

#2 lore raw
In reply to Qu Wenruo
Quoted 88 lines · Qu Wenruo

[BUG]
When the data space is exhausted, even the inode has NOCOW attribute, btrfs will still refuse to truncate unaligned range due to ENOSPC.

The following script can reproduce it pretty easily:
#!/bin/bash

dev=/dev/test/test
mnt=/mnt/btrfs

umount $dev &> /dev/null
umount $mnt&> /dev/null

mkfs.btrfs -f $dev -b 1G
mount -o nospace_cache $dev $mnt
touch $mnt/foobar
chattr +C $mnt/foobar

xfs_io -f -c "pwrite -b 4k 0 4k" $mnt/foobar > /dev/null xfs_io -f -c "pwrite -b 4k 0 1G" $mnt/padding &> /dev/null sync

xfs_io -c "fpunch 0 2k" $mnt/foobar
umount $mnt

Current btrfs will fail at the fpunch part.

[CAUSE]
Because btrfs_truncate_block() always reserve space without checking the NOCOW attribute.

Since the writeback path follows NOCOW bit, we only need to bother the space reservation code in btrfs_truncate_block().

[FIX]
Make btrfs_truncate_block() to follow btrfs_buffered_write() to try to reserve data space first, and falls back to NOCOW check only when we don't have enough space.

Such always-try-reserve is an optimization introduced in btrfs_buffered_write(), to avoid expensive btrfs_check_can_nocow() call.

Since now check_can_nocow() is needed outside of inode.c, also export it and rename it to btrfs_check_can_nocow().

---
Changelog:
v2:
- Rebased to misc-next
Only one minor conflict in ctree.h

---
 fs/btrfs/ctree.h |  2 ++
 fs/btrfs/file.c  | 10 +++++-----
 fs/btrfs/inode.c | 41 ++++++++++++++++++++++++++++++++++-------
 3 files changed, 41 insertions(+), 12 deletions(-)
fs/btrfs/ctree.h +2 -0
index 161533040978..40e8c8170d39 100644
@@ -2984,6 +2984,8 @@ int btrfs_dirty_pages(struct inode *inode, struct page **pages,
 		      size_t num_pages, loff_t pos, size_t write_bytes,
 		      struct extent_state **cached);
 int btrfs_fdatawrite_range(struct inode *inode, loff_t start, loff_t end);
+int btrfs_check_can_nocow(struct btrfs_inode *inode, loff_t pos,
+			  size_t *write_bytes);
 
 /* tree-defrag.c */
 int btrfs_defrag_leaves(struct btrfs_trans_handle *trans,
fs/btrfs/file.c +2 -2
index fde125616687..a298803e2b08 100644
@@ -1532,8 +1532,8 @@ lock_and_cleanup_extent_if_need(struct btrfs_inode *inode, struct page **pages,
 	return ret;
 }
 
-static noinline int check_can_nocow(struct btrfs_inode *inode, loff_t pos,
-				    size_t *write_bytes)
+int btrfs_check_can_nocow(struct btrfs_inode *inode, loff_t pos,
+			  size_t *write_bytes)
 {
 	struct btrfs_fs_info *fs_info = inode->root->fs_info;
 	struct btrfs_root *root = inode->root;

fs/btrfs/file.c static noinline ssize_t btrfs_buffered_write(struct kiocb *iocb, · line 1636

Quoted 7 lines
fs/btrfs/file.c +2 -2
@@ -1632,8 +1632,8 @@ static noinline ssize_t btrfs_buffered_write(struct kiocb *iocb,
 		if (ret < 0) {
 			if ((BTRFS_I(inode)->flags & (BTRFS_INODE_NODATACOW |
 						      BTRFS_INODE_PREALLOC)) &&
-			    check_can_nocow(BTRFS_I(inode), pos,
-					&write_bytes) > 0) {
+			    btrfs_check_can_nocow(BTRFS_I(inode), pos,
+						  &write_bytes) > 0) {

This function acquires DREW lock and I don't see this lock being released in the caller. David already raised concerns about check_can_nocow obscuring this locking. Let's refactor this function before introducing new callers.

Quoted 86 lines
fs/btrfs/file.c +1 -1
 				/*
 				 * For nodata cow case, no need to reserve
 				 * data space.
@@ -1950,7 +1950,7 @@ static ssize_t btrfs_file_write_iter(struct kiocb *iocb,
 		 */
 		if (!(BTRFS_I(inode)->flags & (BTRFS_INODE_NODATACOW |
 					      BTRFS_INODE_PREALLOC)) ||
-		    check_can_nocow(BTRFS_I(inode), pos, &count) <= 0) {
+		    btrfs_check_can_nocow(BTRFS_I(inode), pos, &count) <= 0) {
 			inode_unlock(inode);
 			return -EAGAIN;
 		}
fs/btrfs/inode.c +34 -7
index 1242d0aa108d..6b94ec7369c6 100644
@@ -4493,11 +4493,13 @@ int btrfs_truncate_block(struct inode *inode, loff_t from, loff_t len,
 	struct extent_state *cached_state = NULL;
 	struct extent_changeset *data_reserved = NULL;
 	char *kaddr;
+	bool only_release_metadata = false;
 	u32 blocksize = fs_info->sectorsize;
 	pgoff_t index = from >> PAGE_SHIFT;
 	unsigned offset = from & (blocksize - 1);
 	struct page *page;
 	gfp_t mask = btrfs_alloc_write_mask(mapping);
+	size_t write_bytes = blocksize;
 	int ret = 0;
 	u64 block_start;
 	u64 block_end;
@@ -4509,11 +4511,26 @@ int btrfs_truncate_block(struct inode *inode, loff_t from, loff_t len,
 	block_start = round_down(from, blocksize);
 	block_end = block_start + blocksize - 1;
 
-	ret = btrfs_delalloc_reserve_space(inode, &data_reserved,
-					   block_start, blocksize);
-	if (ret)
+	ret = btrfs_check_data_free_space(inode, &data_reserved, block_start,
+					  blocksize);
+	if (ret < 0) {
+		if ((BTRFS_I(inode)->flags & (BTRFS_INODE_NODATACOW |
+					      BTRFS_INODE_PREALLOC)) &&
+		    btrfs_check_can_nocow(BTRFS_I(inode), block_start,
+					  &write_bytes) > 0) {
+			/* For nocow case, no need to reserve data space. */
+			only_release_metadata = true;
+		} else {
+			goto out;
+		}
+	}
+	ret = btrfs_delalloc_reserve_metadata(BTRFS_I(inode), blocksize);
+	if (ret < 0) {
+		if (!only_release_metadata)
+			btrfs_free_reserved_data_space(inode, data_reserved,
+					block_start, blocksize);
 		goto out;
-
+	}
 again:
 	page = find_or_create_page(mapping, index, mask);
 	if (!page) {
@@ -4582,10 +4599,20 @@ int btrfs_truncate_block(struct inode *inode, loff_t from, loff_t len,
 	set_page_dirty(page);
 	unlock_extent_cached(io_tree, block_start, block_end, &cached_state);
 
+	if (only_release_metadata)
+		set_extent_bit(&BTRFS_I(inode)->io_tree, block_start,
+				block_end, EXTENT_NORESERVE, NULL, NULL,
+				GFP_NOFS);
+
 out_unlock:
-	if (ret)
-		btrfs_delalloc_release_space(inode, data_reserved, block_start,
-					     blocksize, true);
+	if (ret) {
+		if (!only_release_metadata)
+			btrfs_delalloc_release_space(inode, data_reserved,
+					block_start, blocksize, true);
+		else
+			btrfs_delalloc_release_metadata(BTRFS_I(inode),
+					blocksize, true);
+	}
 	btrfs_delalloc_release_extents(BTRFS_I(inode), blocksize);
 	unlock_page(page);
 	put_page(page);