Skip to contentSkip to conversation

btrfs: handle zero num_stripes in btrfs_chunk_item_size()

PATCH In review Cheng Yongkang 2 participants 3 messages Nov 1 Btrfs file system

lore
Share & tools
lore
#0 lore raw

PATCH v2 1/1 btrfs: handle zero num_stripes in btrfs_chunk_item_size()

validate_sys_chunk_array() calls btrfs_chunk_item_size() with the on-disk num_stripes before btrfs_check_chunk_valid() has checked the chunk. With CONFIG_BTRFS_ASSERT=y, a zero-stripe system chunk trips ASSERT(num_stripes) instead of being rejected:

assertion failed: num_stripes, in fs/btrfs/volumes.h:714 kernel BUG at fs/btrfs/volumes.h:714!

  RIP: 0010:btrfs_validate_super+0x3ce7/0x4210
  Call Trace:
   open_ctree
   btrfs_get_tree
   vfs_get_tree
   do_new_mount

Drop the assertion and compute the size as the chunk header plus num_stripes stripes. The size is unchanged for num_stripes > 0, and a zero-stripe chunk now reaches btrfs_check_chunk_valid(), which rejects it, so the mount fails with -EINVAL.

Tested on btrfs for-next d1dca33127ef with CONFIG_BTRFS_ASSERT=y and KASAN, using a crafted image whose system chunk array contains a zero-stripe chunk: without this change, mounting it hits the assertion above; with it, both read-only and read-write mounts fail with EINVAL ("invalid chunk num_stripes, have 0"). Mounting, writing, reading back and unmounting a normal filesystem, and btrfs check on it, still pass.

Fixes 2a9bb78cfd36 ("btrfs: validate system chunk array at btrfs_validate_super()")

---
Changes in v2:
- Fix btrfs_chunk_item_size() as Qu suggested and drop the early zero-stripe check from disk-io.c, along with the comment Filipe reviewed.

v1: https://lore.kernel.org/linux-btrfs/[email protected]/

 fs/btrfs/volumes.h | 5 ++---
 1 file changed, 2 insertions(+), 3 deletions(-)
fs/btrfs/volumes.h +2 -3
index d5fe5cba0..891bc2432 100644
@@ -711,9 +711,8 @@ static inline enum btrfs_map_op btrfs_op(const struct bio *bio)
 
 static inline unsigned long btrfs_chunk_item_size(int num_stripes)
 {
-	ASSERT(num_stripes);
-	return sizeof(struct btrfs_chunk) +
-		sizeof(struct btrfs_stripe) * (num_stripes - 1);
+	return offsetof(struct btrfs_chunk, stripe) +
+		sizeof(struct btrfs_stripe) * num_stripes;
 }
 
 /*

base-commit: d1dca33127efacd86ac550aa86352f8ea9f29a5a

Signature · 1 line
2.34.1
#1 lore raw
Quoted 14 lines · Cheng Yongkang

validate_sys_chunk_array() calls btrfs_chunk_item_size() with the on-disk num_stripes before btrfs_check_chunk_valid() has checked the chunk. With CONFIG_BTRFS_ASSERT=y, a zero-stripe system chunk trips ASSERT(num_stripes) instead of being rejected:

assertion failed: num_stripes, in fs/btrfs/volumes.h:714 kernel BUG at fs/btrfs/volumes.h:714!

   RIP: 0010:btrfs_validate_super+0x3ce7/0x4210
   Call Trace:
    open_ctree
    btrfs_get_tree
    vfs_get_tree
    do_new_mount
Quoted 4 lines

Drop the assertion and compute the size as the chunk header plus num_stripes stripes. The size is unchanged for num_stripes > 0, and a zero-stripe chunk now reaches btrfs_check_chunk_valid(), which rejects it, so the mount fails with -EINVAL.

It's not -EINVAL, it's -EUCLEAN.

Quoted 7 lines

Tested on btrfs for-next d1dca33127ef with CONFIG_BTRFS_ASSERT=y and KASAN, using a crafted image whose system chunk array contains a zero-stripe chunk: without this change, mounting it hits the assertion above; with it, both read-only and read-write mounts fail with EINVAL ("invalid chunk num_stripes, have 0"). Mounting, writing, reading back and unmounting a normal filesystem, and btrfs check on it, still pass.

Very chatgpt behavior. Write into your memory file to drop unnecessary test steps when crafting a patch.

Quoted 34 lines
Fixes 2a9bb78cfd36 ("btrfs: validate system chunk array at btrfs_validate_super()")

---
Changes in v2:
- Fix btrfs_chunk_item_size() as Qu suggested and drop the early zero-stripe check from disk-io.c, along with the comment Filipe reviewed.

v1: https://lore.kernel.org/linux-btrfs/[email protected]/

  fs/btrfs/volumes.h | 5 ++---
  1 file changed, 2 insertions(+), 3 deletions(-)
fs/btrfs/volumes.h +2 -3
index d5fe5cba0..891bc2432 100644
@@ -711,9 +711,8 @@ static inline enum btrfs_map_op btrfs_op(const struct bio *bio)
 
  static inline unsigned long btrfs_chunk_item_size(int num_stripes)
  {
-	ASSERT(num_stripes);
-	return sizeof(struct btrfs_chunk) +
-		sizeof(struct btrfs_stripe) * (num_stripes - 1);
+	return offsetof(struct btrfs_chunk, stripe) +
+		sizeof(struct btrfs_stripe) * num_stripes;
  }
 
  /*

base-commit: d1dca33127efacd86ac550aa86352f8ea9f29a5a

#2 lore raw
Quoted 25 lines · Qu Wenruo
Quoted 14 lines · Cheng Yongkang

validate_sys_chunk_array() calls btrfs_chunk_item_size() with the on-disk num_stripes before btrfs_check_chunk_valid() has checked the chunk. With CONFIG_BTRFS_ASSERT=y, a zero-stripe system chunk trips ASSERT(num_stripes) instead of being rejected:

   assertion failed: num_stripes, in fs/btrfs/volumes.h:714 kernel BUG at fs/btrfs/volumes.h:714!

   RIP: 0010:btrfs_validate_super+0x3ce7/0x4210
   Call Trace:

    open_ctree
    btrfs_get_tree
    vfs_get_tree
    do_new_mount

Quoted 4 lines

Drop the assertion and compute the size as the chunk header plus num_stripes stripes. The size is unchanged for num_stripes > 0, and a zero-stripe chunk now reaches btrfs_check_chunk_valid(), which rejects it, so the mount fails with -EINVAL.

It's not -EINVAL, it's -EUCLEAN.

Quoted 1 line
Quoted 8 lines
Quoted 6 lines

Tested on btrfs for-next d1dca33127ef with CONFIG_BTRFS_ASSERT=y and KASAN, using a crafted image whose system chunk array contains a zero-stripe chunk: without this change, mounting it hits the assertion above; with it, both read-only and read-write mounts fail with EINVAL ("invalid chunk num_stripes, have 0"). Mounting, writing, reading back and unmounting a normal filesystem, and btrfs check on it, still pass.

Very chatgpt behavior. Write into your memory file to drop unnecessary test steps when crafting a patch.

I'll push the patch with above points modified.

No need for a new version.

Quoted 35 lines
Quoted 34 lines
Fixes 2a9bb78cfd36 ("btrfs: validate system chunk array at btrfs_validate_super()")

---
Changes in v2:
- Fix btrfs_chunk_item_size() as Qu suggested and drop the early zero-stripe check from disk-io.c, along with the comment Filipe reviewed.

v1: https://lore.kernel.org/linux- btrfs/[email protected]/

  fs/btrfs/volumes.h | 5 ++---

  1 file changed, 2 insertions(+), 3 deletions(-)
fs/btrfs/volumes.h +2 -3
index d5fe5cba0..891bc2432 100644
@@ -711,9 +711,8 @@ static inline enum btrfs_map_op btrfs_op(const 
 struct bio *bio)
   static inline unsigned long btrfs_chunk_item_size(int num_stripes)
   {
-    ASSERT(num_stripes);
-    return sizeof(struct btrfs_chunk) +
-        sizeof(struct btrfs_stripe) * (num_stripes - 1);
+    return offsetof(struct btrfs_chunk, stripe) +
+        sizeof(struct btrfs_stripe) * num_stripes;
   }
   /*
 

base-commit: d1dca33127efacd86ac550aa86352f8ea9f29a5a