Skip to contentSkip to conversation

btrfs: detect and fix the ram_bytes mismatch

PATCH Superseded Qu Wenruo 3 participants 10 messages Jun 25, 2024 Btrfs file system

v1 of 2 Superseded Latest is v2 →

lore
Share & tools
lore

Qu Wenruo

author
#0 lore raw

cover btrfs: detect and fix the ram_bytes mismatch

There is a long existing mismatch between ram_bytes and disk_num_bytes for regular non-compressed data extents.

It turns out to be caused by truncated ordered extents, which modified ram_bytes unnecessarily.

Thankfully this is not going to cause any data corruption or whatever, kernel can handle it correctly without any extra problem. It's only a small violation on the on-disk format.

This series would fix by:

- Cleanup the @bytenr usage inside btrfs_extent_item_to_extent_map()
- Override the ram_bytes when reading file extent items from disk So that we always get correct extent maps even if the on-disk one is incorrect.
- Add the proper fix for the ram_bytes mismatch
- Add a tree-checker for the ram_bytes mismatch
Since we can have on-disk ram_bytes incorrect already, this check is only for DEBUG and ASSERT builds, and it won't report error but only does a kernel warning for us to catch.

Qu Wenruo (4):
btrfs: cleanup the bytenr usage inside
btrfs_extent_item_to_extent_map()
btrfs: make validate_extent_map() to catch ram_bytes mismatch
btrfs: fix the ram_bytes assignment for truncated ordered extents
btrfs: tree-checker: add extra ram_bytes and disk_num_bytes check

 fs/btrfs/extent_map.c   |  5 +++++
 fs/btrfs/file-item.c    |  9 ++++-----
 fs/btrfs/inode.c        |  4 +---
 fs/btrfs/tree-checker.c | 19 +++++++++++++++++++
 4 files changed, 29 insertions(+), 8 deletions(-)
Signature · 1 line
2.45.2
author
#1 lore raw

PATCH 1/4 btrfs: cleanup the bytenr usage inside btrfs_extent_item_to_extent_map()

[PROBLEMS]
Before commit 85de2be7129c ("btrfs: remove extent_map::block_start member"), we utilized @bytenr variable inside btrfs_extent_item_to_extent_map() to calculate block_start.

But that commit removed block_start completely, we have no need to advance @bytenr at all.

Furthermore with recent enhanced btrfs-progs check for ram_bytes mimsatch, it turns out that for truncated ordered extents, their ram_bytes can be smaller than disk_num_bytes.

[ENHANCEMENT]
Thankfully all above problems are not really going to affect end users, fix them by:

- Declare @bytenr only inside the if branch and make it const So we can remove the unnecessary advance of @bytenr.

- Manually override extent_map::ram_bytes using disk_num_bytes This is for non-compressed regular/preallocated extents.

---
 fs/btrfs/file-item.c | 9 ++++-----
 1 file changed, 4 insertions(+), 5 deletions(-)
fs/btrfs/file-item.c +4 -5
index 55703c833f3d..2cc61c792ee6 100644
@@ -1281,7 +1281,6 @@ void btrfs_extent_item_to_extent_map(struct btrfs_inode *inode,
 	const int slot = path->slots[0];
 	struct btrfs_key key;
 	u64 extent_start;
-	u64 bytenr;
 	u8 type = btrfs_file_extent_type(leaf, fi);
 	int compress_type = btrfs_file_extent_compression(leaf, fi);
 
@@ -1291,22 +1290,22 @@ void btrfs_extent_item_to_extent_map(struct btrfs_inode *inode,
 	em->generation = btrfs_file_extent_generation(leaf, fi);
 	if (type == BTRFS_FILE_EXTENT_REG ||
 	    type == BTRFS_FILE_EXTENT_PREALLOC) {
+		const u64 disk_bytenr = btrfs_file_extent_disk_bytenr(leaf, fi);
+
 		em->start = extent_start;
 		em->len = btrfs_file_extent_end(path) - extent_start;
-		bytenr = btrfs_file_extent_disk_bytenr(leaf, fi);
-		if (bytenr == 0) {
+		if (disk_bytenr == 0) {
 			em->disk_bytenr = EXTENT_MAP_HOLE;
 			em->disk_num_bytes = 0;
 			em->offset = 0;
 			return;
 		}
-		em->disk_bytenr = btrfs_file_extent_disk_bytenr(leaf, fi);
+		em->disk_bytenr = disk_bytenr;
 		em->disk_num_bytes = btrfs_file_extent_disk_num_bytes(leaf, fi);
 		em->offset = btrfs_file_extent_offset(leaf, fi);
 		if (compress_type != BTRFS_COMPRESS_NONE) {
 			extent_map_set_compression(em, compress_type);
 		} else {
-			bytenr += btrfs_file_extent_offset(leaf, fi);
 			if (type == BTRFS_FILE_EXTENT_PREALLOC)
 				em->flags |= EXTENT_FLAG_PREALLOC;
 		}
Signature · 1 line
2.45.2
#2 lore raw

Re: [PATCH 1/4] btrfs: cleanup the bytenr usage inside btrfs_extent_item_to_extent_map()

Quoted 2 lines · Qu Wenruo

[PROBLEMS]

I wouldn't call this "problems", there are no bugs here or anything harmful.

Quoted 9 lines

Before commit 85de2be7129c ("btrfs: remove extent_map::block_start member"), we utilized @bytenr variable inside btrfs_extent_item_to_extent_map() to calculate block_start.

But that commit removed block_start completely, we have no need to advance @bytenr at all.

Furthermore with recent enhanced btrfs-progs check for ram_bytes mimsatch, it turns out that for truncated ordered extents, their

mimsatch -> mismatch

Quoted 11 lines

ram_bytes can be smaller than disk_num_bytes.

[ENHANCEMENT]
Thankfully all above problems are not really going to affect end users, fix them by:

- Declare @bytenr only inside the if branch and make it const So we can remove the unnecessary advance of @bytenr.

- Manually override extent_map::ram_bytes using disk_num_bytes This is for non-compressed regular/preallocated extents.

I don't see anything in the patch changing ram_bytes. Perhaps this is from an early patch version never submitted, or from some other patch?

The code itself looks good.
Thanks.

Quoted 49 lines
---
 fs/btrfs/file-item.c | 9 ++++-----
 1 file changed, 4 insertions(+), 5 deletions(-)
fs/btrfs/file-item.c +4 -6
index 55703c833f3d..2cc61c792ee6 100644
@@ -1281,7 +1281,6 @@ void btrfs_extent_item_to_extent_map(struct btrfs_inode *inode,
        const int slot = path->slots[0];
        struct btrfs_key key;
        u64 extent_start;
-       u64 bytenr;
        u8 type = btrfs_file_extent_type(leaf, fi);
        int compress_type = btrfs_file_extent_compression(leaf, fi);
 
@@ -1291,22 +1290,22 @@ void btrfs_extent_item_to_extent_map(struct btrfs_inode *inode,
        em->generation = btrfs_file_extent_generation(leaf, fi);
        if (type == BTRFS_FILE_EXTENT_REG ||
            type == BTRFS_FILE_EXTENT_PREALLOC) {
+               const u64 disk_bytenr = btrfs_file_extent_disk_bytenr(leaf, fi);
+
                em->start = extent_start;
                em->len = btrfs_file_extent_end(path) - extent_start;
-               bytenr = btrfs_file_extent_disk_bytenr(leaf, fi);
-               if (bytenr == 0) {
+               if (disk_bytenr == 0) {
                        em->disk_bytenr = EXTENT_MAP_HOLE;
                        em->disk_num_bytes = 0;
                        em->offset = 0;
                        return;
                }
-               em->disk_bytenr = btrfs_file_extent_disk_bytenr(leaf, fi);
+               em->disk_bytenr = disk_bytenr;
                em->disk_num_bytes = btrfs_file_extent_disk_num_bytes(leaf, fi);
                em->offset = btrfs_file_extent_offset(leaf, fi);
                if (compress_type != BTRFS_COMPRESS_NONE) {
                        extent_map_set_compression(em, compress_type);
                } else {
-                       bytenr += btrfs_file_extent_offset(leaf, fi);
                        if (type == BTRFS_FILE_EXTENT_PREALLOC)
                                em->flags |= EXTENT_FLAG_PREALLOC;
                }
--

2.45.2

#3 lore raw

Re: [PATCH 1/4] btrfs: cleanup the bytenr usage inside btrfs_extent_item_to_extent_map()

Quoted 27 lines · Filipe Manana
Quoted 2 lines · Qu Wenruo

[PROBLEMS]

I wouldn't call this "problems", there are no bugs here or anything harmful.

Quoted 9 lines

Before commit 85de2be7129c ("btrfs: remove extent_map::block_start member"), we utilized @bytenr variable inside btrfs_extent_item_to_extent_map() to calculate block_start.

But that commit removed block_start completely, we have no need to advance @bytenr at all.

Furthermore with recent enhanced btrfs-progs check for ram_bytes mimsatch, it turns out that for truncated ordered extents, their

mimsatch -> mismatch

Quoted 9 lines

ram_bytes can be smaller than disk_num_bytes.

[ENHANCEMENT]
Thankfully all above problems are not really going to affect end users, fix them by:

- Declare @bytenr only inside the if branch and make it const So we can remove the unnecessary advance of @bytenr.

Quoted 6 lines
Quoted 2 lines

- Manually override extent_map::ram_bytes using disk_num_bytes This is for non-compressed regular/preallocated extents.

I don't see anything in the patch changing ram_bytes. Perhaps this is from an early patch version never submitted, or from some other patch?

My bad, when re-editing the commit message, I got confused with later patches.

I will remove the ram_bytes related part.

Thanks,
Qu

Quoted 54 lines

The code itself looks good.
Thanks.

Quoted 49 lines
---
  fs/btrfs/file-item.c | 9 ++++-----
  1 file changed, 4 insertions(+), 5 deletions(-)
fs/btrfs/file-item.c +4 -6
index 55703c833f3d..2cc61c792ee6 100644
@@ -1281,7 +1281,6 @@ void btrfs_extent_item_to_extent_map(struct btrfs_inode *inode,
         const int slot = path->slots[0];
         struct btrfs_key key;
         u64 extent_start;
-       u64 bytenr;
         u8 type = btrfs_file_extent_type(leaf, fi);
         int compress_type = btrfs_file_extent_compression(leaf, fi);
 
@@ -1291,22 +1290,22 @@ void btrfs_extent_item_to_extent_map(struct btrfs_inode *inode,
         em->generation = btrfs_file_extent_generation(leaf, fi);
         if (type == BTRFS_FILE_EXTENT_REG ||
             type == BTRFS_FILE_EXTENT_PREALLOC) {
+               const u64 disk_bytenr = btrfs_file_extent_disk_bytenr(leaf, fi);
+
                 em->start = extent_start;
                 em->len = btrfs_file_extent_end(path) - extent_start;
-               bytenr = btrfs_file_extent_disk_bytenr(leaf, fi);
-               if (bytenr == 0) {
+               if (disk_bytenr == 0) {
                         em->disk_bytenr = EXTENT_MAP_HOLE;
                         em->disk_num_bytes = 0;
                         em->offset = 0;
                         return;
                 }
-               em->disk_bytenr = btrfs_file_extent_disk_bytenr(leaf, fi);
+               em->disk_bytenr = disk_bytenr;
                 em->disk_num_bytes = btrfs_file_extent_disk_num_bytes(leaf, fi);
                 em->offset = btrfs_file_extent_offset(leaf, fi);
                 if (compress_type != BTRFS_COMPRESS_NONE) {
                         extent_map_set_compression(em, compress_type);
                 } else {
-                       bytenr += btrfs_file_extent_offset(leaf, fi);
                         if (type == BTRFS_FILE_EXTENT_PREALLOC)
                                 em->flags |= EXTENT_FLAG_PREALLOC;
                 }
--

2.45.2

author
#4 lore raw

PATCH 2/4 btrfs: make validate_extent_map() to catch ram_bytes mismatch

In reply to Qu Wenruo

Previously validate_extent_map() is only to catch bugs related to extent_map member cleanups.

But with recent btrfs-check enhancement to catch ram_bytes mismatch with disk_num_bytes, it would be much better to catch such extent maps earlier.

So this patch would add extra ram_bytes validation for extent maps.

Please note that, older filesystems with such mismatch won't trigger this error:

- extent_map::ram_bytes is already fixed when reading from disk At btrfs_extent_item_to_extent_map() we override extent_map::ram_bytes to disk_num_bytes for non-compressed regular/preallocated extents.

So this enhanced sanity checks should not affect end users.

---
 fs/btrfs/extent_map.c | 5 +++++
 1 file changed, 5 insertions(+)
fs/btrfs/extent_map.c +5 -0
index b869a0ee24d2..6961cc73fe3f 100644
@@ -317,6 +317,11 @@ static void validate_extent_map(struct btrfs_fs_info *fs_info, struct extent_map
 		if (em->offset + em->len > em->disk_num_bytes &&
 		    !extent_map_is_compressed(em))
 			dump_extent_map(fs_info, "disk_num_bytes too small", em);
+		if (!extent_map_is_compressed(em) &&
+		    em->ram_bytes != em->disk_num_bytes)
+			dump_extent_map(fs_info,
+		"ram_bytes mismatch with disk_num_bytes for non-compressed em",
+					em);
 	} else if (em->offset) {
 		dump_extent_map(fs_info, "non-zero offset for hole/inline", em);
 	}
Signature · 1 line
2.45.2
#5 lore raw

Re: [PATCH 2/4] btrfs: make validate_extent_map() to catch ram_bytes mismatch

Quoted 9 lines · Qu Wenruo

Previously validate_extent_map() is only to catch bugs related to extent_map member cleanups.

But with recent btrfs-check enhancement to catch ram_bytes mismatch with disk_num_bytes, it would be much better to catch such extent maps earlier.

So this patch would add extra ram_bytes validation for extent maps.

would add -> adds

Quoted 6 lines

Please note that, older filesystems with such mismatch won't trigger this error:

- extent_map::ram_bytes is already fixed when reading from disk At btrfs_extent_item_to_extent_map() we override extent_map::ram_bytes to disk_num_bytes for non-compressed regular/preallocated extents.

At btrfs_extent_item_to_extent_map() we override ram_bytes for any type of extent, it's not conditional on anything.
So this is confusing because it doesn't match the code.

Quoted 2 lines

So this enhanced sanity checks should not affect end users.

checks -> check, otherwise "this" should be "these" (but there's only one check).

The code looks good, thanks.

Quoted 26 lines
---
 fs/btrfs/extent_map.c | 5 +++++
 1 file changed, 5 insertions(+)
fs/btrfs/extent_map.c +5 -1
index b869a0ee24d2..6961cc73fe3f 100644
@@ -317,6 +317,11 @@ static void validate_extent_map(struct btrfs_fs_info *fs_info, struct extent_map
                if (em->offset + em->len > em->disk_num_bytes &&
                    !extent_map_is_compressed(em))
                        dump_extent_map(fs_info, "disk_num_bytes too small", em);
+               if (!extent_map_is_compressed(em) &&
+                   em->ram_bytes != em->disk_num_bytes)
+                       dump_extent_map(fs_info,
+               "ram_bytes mismatch with disk_num_bytes for non-compressed em",
+                                       em);
        } else if (em->offset) {
                dump_extent_map(fs_info, "non-zero offset for hole/inline", em);
        }
--

2.45.2

author
#6 lore raw

PATCH 3/4 btrfs: fix the ram_bytes assignment for truncated ordered extents

In reply to Qu Wenruo

[BUG]
After adding extra checks on btrfs_file_extent_item::ram_bytes to tree-checker, running fsstress with multiple threads can lead to tree-checker warning at write time, as we created file extent items with ram_bytes.

All those offending file extents have offset 0, and ram_bytes matching num_bytes, and smaller than disk_num_bytes.

This would also trigger the recently enhanced btrfs-check, which would catch such mismatch and report them as minor errors.

[CAUSE]
When a folio/page is invalidated and it is part of a submitted OE, we mark the OE truncated just to the beginning of the folio/page.

And for truncated OE, we insert the file extent item with incorrect value for ram_bytes (using num_bytes instead of the usual value).

This is not a big deal for end users, as we do not utilize the ram_bytes field for regular non-compressed extents.
This mismatch is just a small violation against on-disk format.

[FIX]
Fix it by removing the override on btrfs_file_extent_item::ram_bytes.

---
 fs/btrfs/inode.c | 4 +---
 1 file changed, 1 insertion(+), 3 deletions(-)
fs/btrfs/inode.c +1 -3
index d6c43120c5d3..45f77ae8963f 100644
@@ -3018,10 +3018,8 @@ static int insert_ordered_extent_file_extent(struct btrfs_trans_handle *trans,
 	btrfs_set_stack_file_extent_disk_num_bytes(&stack_fi,
 						   oe->disk_num_bytes);
 	btrfs_set_stack_file_extent_offset(&stack_fi, oe->offset);
-	if (test_bit(BTRFS_ORDERED_TRUNCATED, &oe->flags)) {
+	if (test_bit(BTRFS_ORDERED_TRUNCATED, &oe->flags))
 		num_bytes = oe->truncated_len;
-		ram_bytes = num_bytes;
-	}
 	btrfs_set_stack_file_extent_num_bytes(&stack_fi, num_bytes);
 	btrfs_set_stack_file_extent_ram_bytes(&stack_fi, ram_bytes);
 	btrfs_set_stack_file_extent_compression(&stack_fi, oe->compress_type);
Signature · 1 line
2.45.2
Reviewed-by on v1 3/4
#7 lore raw

Re: [PATCH 3/4] btrfs: fix the ram_bytes assignment for truncated ordered extents

Quoted 4 lines · Qu Wenruo

[BUG]
After adding extra checks on btrfs_file_extent_item::ram_bytes to tree-checker, running fsstress with multiple threads can lead to

It's irrelevant to mention multiple threads, that's not necessary to cause the problem.

Quoted 2 lines

tree-checker warning at write time, as we created file extent items with ram_bytes.

This last part of the sentence makes no sense "we created file extent items with ram_bytes" - they must always have ram_bytes.
I think you meant to say "with an invalid ram_bytes value".

Quoted 6 lines

All those offending file extents have offset 0, and ram_bytes matching num_bytes, and smaller than disk_num_bytes.

This would also trigger the recently enhanced btrfs-check, which would catch such mismatch and report them as minor errors.

mismatch -> mismatches

Quoted 25 lines

[CAUSE]
When a folio/page is invalidated and it is part of a submitted OE, we mark the OE truncated just to the beginning of the folio/page.

And for truncated OE, we insert the file extent item with incorrect value for ram_bytes (using num_bytes instead of the usual value).

This is not a big deal for end users, as we do not utilize the ram_bytes field for regular non-compressed extents.
This mismatch is just a small violation against on-disk format.

[FIX]
Fix it by removing the override on btrfs_file_extent_item::ram_bytes.

---
 fs/btrfs/inode.c | 4 +---
 1 file changed, 1 insertion(+), 3 deletions(-)
fs/btrfs/inode.c +0 -0
index d6c43120c5d3..45f77ae8963f 100644

fs/btrfs/inode.c static int insert_ordered_extent_file_extent(struct btrfs_trans_handle *trans, · line 3024

Quoted 8 lines
fs/btrfs/inode.c +1 -3
@@ -3018,10 +3018,8 @@ static int insert_ordered_extent_file_extent(struct btrfs_trans_handle *trans,
        btrfs_set_stack_file_extent_disk_num_bytes(&stack_fi,
                                                   oe->disk_num_bytes);
        btrfs_set_stack_file_extent_offset(&stack_fi, oe->offset);
-       if (test_bit(BTRFS_ORDERED_TRUNCATED, &oe->flags)) {
+       if (test_bit(BTRFS_ORDERED_TRUNCATED, &oe->flags))
                num_bytes = oe->truncated_len;
-               ram_bytes = num_bytes;
-       }

The code looks good, with those updates to the the changelog:

Thanks.

Quoted 7 lines
fs/btrfs/inode.c +0 -1
        btrfs_set_stack_file_extent_num_bytes(&stack_fi, num_bytes);
        btrfs_set_stack_file_extent_ram_bytes(&stack_fi, ram_bytes);
        btrfs_set_stack_file_extent_compression(&stack_fi, oe->compress_type);
--

2.45.2

author
#8 lore raw

PATCH 4/4 btrfs: tree-checker: add extra ram_bytes and disk_num_bytes check

In reply to Qu Wenruo

This is to ensure non-compressed file extents (both regular and prealloc) should have matching ram_bytes and disk_num_bytes.

This is only for CONFIG_BTRFS_DEBUG and CONFIG_BTRFS_ASSERT case, furthermore this will not return error, but just a kernel warning to inform developers.

---
 fs/btrfs/tree-checker.c | 19 +++++++++++++++++++
 1 file changed, 19 insertions(+)
fs/btrfs/tree-checker.c +19 -0
index a2c3651a3d8f..cddabd9a0e37 100644
@@ -340,6 +340,25 @@ static int check_extent_data_item(struct extent_buffer *leaf,
 		}
 	}
 
+	/*
+	 * For non-compressed data extents, ram_bytes should match its disk_bytenr.
+	 * However we do not really utilize ram_bytes in this case, so this check
+	 * is only optional for DEBUG+ASSERT builds for developers to catch the
+	 * unexpected behaviors.
+	 */
+	if (IS_ENABLED(CONFIG_BTRFS_DEBUG) && IS_ENABLED(CONFIG_BTRFS_ASSERT) &&
+	    btrfs_file_extent_compression(leaf, fi) == BTRFS_COMPRESS_NONE &&
+	    btrfs_file_extent_disk_bytenr(leaf, fi)) {
+		if (unlikely(btrfs_file_extent_ram_bytes(leaf, fi) !=
+			     btrfs_file_extent_disk_num_bytes(leaf, fi))) {
+			file_extent_err(leaf, slot,
+"mismatch ram_bytes (%llu) and disk_num_bytes (%llu) for non-compressed extent",
+					btrfs_file_extent_ram_bytes(leaf, fi),
+					btrfs_file_extent_disk_num_bytes(leaf, fi));
+			WARN_ON(1);
+		}
+	}
+
 	return 0;
 }
 
Signature · 1 line
2.45.2
#9 lore raw

Re: [PATCH 4/4] btrfs: tree-checker: add extra ram_bytes and disk_num_bytes check

Quoted 5 lines · Qu Wenruo

This is to ensure non-compressed file extents (both regular and prealloc) should have matching ram_bytes and disk_num_bytes.

This is only for CONFIG_BTRFS_DEBUG and CONFIG_BTRFS_ASSERT case,

I would leave just for DEBUG.

Quoted 22 lines

furthermore this will not return error, but just a kernel warning to inform developers.

---
 fs/btrfs/tree-checker.c | 19 +++++++++++++++++++
 1 file changed, 19 insertions(+)
fs/btrfs/tree-checker.c +6 -0
index a2c3651a3d8f..cddabd9a0e37 100644
@@ -340,6 +340,25 @@ static int check_extent_data_item(struct extent_buffer *leaf,
                }
        }
 
+       /*
+        * For non-compressed data extents, ram_bytes should match its disk_bytenr.
+        * However we do not really utilize ram_bytes in this case, so this check
+        * is only optional for DEBUG+ASSERT builds for developers to catch the
+        * unexpected behaviors.
+        */

fs/btrfs/tree-checker.c static int check_extent_data_item(struct extent_buffer *leaf, · line 358

Quoted 10 lines
fs/btrfs/tree-checker.c +10 -0
+       if (IS_ENABLED(CONFIG_BTRFS_DEBUG) && IS_ENABLED(CONFIG_BTRFS_ASSERT) &&
+           btrfs_file_extent_compression(leaf, fi) == BTRFS_COMPRESS_NONE &&
+           btrfs_file_extent_disk_bytenr(leaf, fi)) {
+               if (unlikely(btrfs_file_extent_ram_bytes(leaf, fi) !=
+                            btrfs_file_extent_disk_num_bytes(leaf, fi))) {
+                       file_extent_err(leaf, slot,
+"mismatch ram_bytes (%llu) and disk_num_bytes (%llu) for non-compressed extent",
+                                       btrfs_file_extent_ram_bytes(leaf, fi),
+                                       btrfs_file_extent_disk_num_bytes(leaf, fi));
+                       WARN_ON(1);

Instead of adding here a WARN_ON(1) and unlikely in the if condition, just make the if condition use WARN_ON:

if (WARN_ON(btrfs_file_extent_ram_bytes(leaf, fi) != btrfs_file_extent_disk_num_bytes(leaf, fi)))

The WARN_ON includes the unlikely, and further this makes it conform to our preference of having error messages after stack traces.

Thanks.

Quoted 10 lines
fs/btrfs/tree-checker.c +3 -1
+               }
+       }
+
        return 0;
 }
 
--

2.45.2