[PATCH v2] ocfs2: clear extent cache after moving/defragmenting extents

[PATCH v2] ocfs2: clear extent cache after moving/defragmenting extents

[Deepanshu Kartikey]

The extent map cache can become stale when extents are moved or
defragmented, causing subsequent operations to see outdated extent
flags. This triggers a BUG_ON in ocfs2_refcount_cal_cow_clusters().

The problem occurs when:
1. copy_file_range() creates a reflinked extent with OCFS2_EXT_REFCOUNTED
2. ioctl(FITRIM) triggers ocfs2_move_extents()
3. __ocfs2_move_extents_range() reads and caches the extent (flags=0x2)
4. ocfs2_move_extent()/ocfs2_defrag_extent() calls __ocfs2_move_extent()
   which clears OCFS2_EXT_REFCOUNTED flag on disk (flags=0x0)
5. The extent map cache is not invalidated after the move
6. Later write() operations read stale cached flags (0x2) but disk has
   updated flags (0x0), causing a mismatch
7. BUG_ON(!(rec->e_flags & OCFS2_EXT_REFCOUNTED)) triggers

Fix by clearing the extent map cache after each extent move/defrag
operation in __ocfs2_move_extents_range(). This ensures subsequent
operations read fresh extent data from disk.

Link: https://lore.kernel.org/all/20251009142917.517229-1-kartikey406@gmail.com/T/
Reported-by: syzbot+6fdd8fa3380730a4b22c@syzkaller.appspotmail.com
Tested-by: syzbot+6fdd8fa3380730a4b22c@syzkaller.appspotmail.com
Closes: https://syzkaller.appspot.com/bug?id=2959889e1f6e216585ce522f7e8bc002b46ad9e7
Signed-off-by: Deepanshu Kartikey <kartikey406@gmail.com>
---
Changes in v2:
  - Fix moved to __ocfs2_move_extents_range() instead of ocfs2_refcount_cow()
  - The real issue is in FITRIM/move_extents code path, not COW path
  - COW path already clears cache at end of ocfs2_refcount_cow_hunk()

 fs/ocfs2/move_extents.c | 5 +++++
 1 file changed, 5 insertions(+)
---
 fs/ocfs2/move_extents.c | 5 +++++
 1 file changed, 5 insertions(+)

diff --git a/fs/ocfs2/move_extents.c b/fs/ocfs2/move_extents.c
index 86f2631e6360..10923bf7c8b8 100644
--- a/fs/ocfs2/move_extents.c
+++ b/fs/ocfs2/move_extents.c
@@ -867,6 +867,11 @@ static int __ocfs2_move_extents_range(struct buffer_head *di_bh,
 			mlog_errno(ret);
 			goto out;
 		}
+		/*
+		 * Invalidate extent cache after moving/defragging to prevent
+		 * stale cached data with outdated extent flags.
+		 */
+		ocfs2_extent_map_trunc(inode, cpos);
 
 		context->clusters_moved += alloc_size;
 next:
-- 
2.43.0

Re: [PATCH v2] ocfs2: clear extent cache after moving/defragmenting extents

[Mark Fasheh]

On Thu, Oct 9, 2025 at 8:49 AM Deepanshu Kartikey <kartikey406@gmail.com> wrote:
>
> The extent map cache can become stale when extents are moved or
> defragmented, causing subsequent operations to see outdated extent
> flags. This triggers a BUG_ON in ocfs2_refcount_cal_cow_clusters().
>
> The problem occurs when:
> 1. copy_file_range() creates a reflinked extent with OCFS2_EXT_REFCOUNTED
> 2. ioctl(FITRIM) triggers ocfs2_move_extents()
> 3. __ocfs2_move_extents_range() reads and caches the extent (flags=0x2)
> 4. ocfs2_move_extent()/ocfs2_defrag_extent() calls __ocfs2_move_extent()
>    which clears OCFS2_EXT_REFCOUNTED flag on disk (flags=0x0)
> 5. The extent map cache is not invalidated after the move
> 6. Later write() operations read stale cached flags (0x2) but disk has
>    updated flags (0x0), causing a mismatch
> 7. BUG_ON(!(rec->e_flags & OCFS2_EXT_REFCOUNTED)) triggers
>
> Fix by clearing the extent map cache after each extent move/defrag
> operation in __ocfs2_move_extents_range(). This ensures subsequent
> operations read fresh extent data from disk.
>
> Link: https://lore.kernel.org/all/20251009142917.517229-1-kartikey406@gmail.com/T/
> Reported-by: syzbot+6fdd8fa3380730a4b22c@syzkaller.appspotmail.com
> Tested-by: syzbot+6fdd8fa3380730a4b22c@syzkaller.appspotmail.com
> Closes: https://syzkaller.appspot.com/bug?id=2959889e1f6e216585ce522f7e8bc002b46ad9e7
> Signed-off-by: Deepanshu Kartikey <kartikey406@gmail.com>

Reviewed-by: Mark Fasheh <mark@fasheh.com>

Re: [PATCH v2] ocfs2: clear extent cache after moving/defragmenting extents

[Joseph Qi]

On 2025/10/9 23:49, Deepanshu Kartikey wrote:
> The extent map cache can become stale when extents are moved or
> defragmented, causing subsequent operations to see outdated extent
> flags. This triggers a BUG_ON in ocfs2_refcount_cal_cow_clusters().
> 
> The problem occurs when:
> 1. copy_file_range() creates a reflinked extent with OCFS2_EXT_REFCOUNTED
> 2. ioctl(FITRIM) triggers ocfs2_move_extents()
> 3. __ocfs2_move_extents_range() reads and caches the extent (flags=0x2)
> 4. ocfs2_move_extent()/ocfs2_defrag_extent() calls __ocfs2_move_extent()
>    which clears OCFS2_EXT_REFCOUNTED flag on disk (flags=0x0)
> 5. The extent map cache is not invalidated after the move
> 6. Later write() operations read stale cached flags (0x2) but disk has
>    updated flags (0x0), causing a mismatch
> 7. BUG_ON(!(rec->e_flags & OCFS2_EXT_REFCOUNTED)) triggers
> 
> Fix by clearing the extent map cache after each extent move/defrag
> operation in __ocfs2_move_extents_range(). This ensures subsequent
> operations read fresh extent data from disk.
> 
> Link: https://lore.kernel.org/all/20251009142917.517229-1-kartikey406@gmail.com/T/
> Reported-by: syzbot+6fdd8fa3380730a4b22c@syzkaller.appspotmail.com
> Tested-by: syzbot+6fdd8fa3380730a4b22c@syzkaller.appspotmail.com
> Closes: https://syzkaller.appspot.com/bug?id=2959889e1f6e216585ce522f7e8bc002b46ad9e7
> Signed-off-by: Deepanshu Kartikey <kartikey406@gmail.com>

Looks fine.
Reviewed-by: Joseph Qi <joseph.qi@linux.alibaba.com>

> ---
> Changes in v2:
>   - Fix moved to __ocfs2_move_extents_range() instead of ocfs2_refcount_cow()
>   - The real issue is in FITRIM/move_extents code path, not COW path
>   - COW path already clears cache at end of ocfs2_refcount_cow_hunk()
> 
>  fs/ocfs2/move_extents.c | 5 +++++
>  1 file changed, 5 insertions(+)
> ---
>  fs/ocfs2/move_extents.c | 5 +++++
>  1 file changed, 5 insertions(+)
> 
> diff --git a/fs/ocfs2/move_extents.c b/fs/ocfs2/move_extents.c
> index 86f2631e6360..10923bf7c8b8 100644
> --- a/fs/ocfs2/move_extents.c
> +++ b/fs/ocfs2/move_extents.c
> @@ -867,6 +867,11 @@ static int __ocfs2_move_extents_range(struct buffer_head *di_bh,
>  			mlog_errno(ret);
>  			goto out;
>  		}
> +		/*
> +		 * Invalidate extent cache after moving/defragging to prevent
> +		 * stale cached data with outdated extent flags.
> +		 */
> +		ocfs2_extent_map_trunc(inode, cpos);
>  
>  		context->clusters_moved += alloc_size;
>  next:

Re: [PATCH v2] ocfs2: clear extent cache after moving/defragmenting extents

[Andrew Morton]

On Thu,  9 Oct 2025 21:19:03 +0530 Deepanshu Kartikey <kartikey406@gmail.com> wrote:

> The extent map cache can become stale when extents are moved or
> defragmented, causing subsequent operations to see outdated extent
> flags. This triggers a BUG_ON in ocfs2_refcount_cal_cow_clusters().

We try to avoid BUG_ONs ;)

>
> ...
>
> The problem occurs when:
>
> ...
>

> --- a/fs/ocfs2/move_extents.c
> +++ b/fs/ocfs2/move_extents.c
> @@ -867,6 +867,11 @@ static int __ocfs2_move_extents_range(struct buffer_head *di_bh,
>  			mlog_errno(ret);
>  			goto out;
>  		}
> +		/*
> +		 * Invalidate extent cache after moving/defragging to prevent
> +		 * stale cached data with outdated extent flags.
> +		 */
> +		ocfs2_extent_map_trunc(inode, cpos);
>  
>  		context->clusters_moved += alloc_size;
>  next:

I assume we should backport this into earlier kernels, so I'll add a
cc:stable to the changelog.

It's very nice (but not essential) to add a Fixes: as well, so -stable
maintainers know how far back in time to backport the fix.

Re: [PATCH v2] ocfs2: clear extent cache after moving/defragmenting extents

[Deepanshu Kartikey]

Hi Andrew,

Thank you for picking this up!

For the Fixes: tag:

Fixes: 53069d4e76954e2e63c1b3c501051c6fbcf7298c ("Ocfs2/move_extents: move/defrag extents within a certain range.")

This commit introduced __ocfs2_move_extents_range() with the while loop that
calls ocfs2_get_clusters() (which caches extent data) followed by
ocfs2_move_extent()/ocfs2_defrag_extent() (which modify disk), but without
invalidating the extent cache afterwards.

Thanks,
Deepanshu

Re: [PATCH v2] ocfs2: clear extent cache after moving/defragmenting extents

[Heming Zhao]

Hi Joseph,

please check my comments in another mail.

https://lore.kernel.org/all/20251009142917.517229-1-kartikey406@gmail.com/T/#m0c02c884701c12d71206d0f9aba4450879b18e0c

- Heming

On 10/10/25 09:21, Joseph Qi wrote:
> 
> 
> On 2025/10/9 23:49, Deepanshu Kartikey wrote:
>> The extent map cache can become stale when extents are moved or
>> defragmented, causing subsequent operations to see outdated extent
>> flags. This triggers a BUG_ON in ocfs2_refcount_cal_cow_clusters().
>>
>> The problem occurs when:
>> 1. copy_file_range() creates a reflinked extent with OCFS2_EXT_REFCOUNTED
>> 2. ioctl(FITRIM) triggers ocfs2_move_extents()
>> 3. __ocfs2_move_extents_range() reads and caches the extent (flags=0x2)
>> 4. ocfs2_move_extent()/ocfs2_defrag_extent() calls __ocfs2_move_extent()
>>     which clears OCFS2_EXT_REFCOUNTED flag on disk (flags=0x0)
>> 5. The extent map cache is not invalidated after the move
>> 6. Later write() operations read stale cached flags (0x2) but disk has
>>     updated flags (0x0), causing a mismatch
>> 7. BUG_ON(!(rec->e_flags & OCFS2_EXT_REFCOUNTED)) triggers
>>
>> Fix by clearing the extent map cache after each extent move/defrag
>> operation in __ocfs2_move_extents_range(). This ensures subsequent
>> operations read fresh extent data from disk.
>>
>> Link: https://lore.kernel.org/all/20251009142917.517229-1-kartikey406@gmail.com/T/
>> Reported-by: syzbot+6fdd8fa3380730a4b22c@syzkaller.appspotmail.com
>> Tested-by: syzbot+6fdd8fa3380730a4b22c@syzkaller.appspotmail.com
>> Closes: https://syzkaller.appspot.com/bug?id=2959889e1f6e216585ce522f7e8bc002b46ad9e7
>> Signed-off-by: Deepanshu Kartikey <kartikey406@gmail.com>
> 
> Looks fine.
> Reviewed-by: Joseph Qi <joseph.qi@linux.alibaba.com>
> 
>> ---
>> Changes in v2:
>>    - Fix moved to __ocfs2_move_extents_range() instead of ocfs2_refcount_cow()
>>    - The real issue is in FITRIM/move_extents code path, not COW path
>>    - COW path already clears cache at end of ocfs2_refcount_cow_hunk()
>>
>>   fs/ocfs2/move_extents.c | 5 +++++
>>   1 file changed, 5 insertions(+)
>> ---
>>   fs/ocfs2/move_extents.c | 5 +++++
>>   1 file changed, 5 insertions(+)
>>
>> diff --git a/fs/ocfs2/move_extents.c b/fs/ocfs2/move_extents.c
>> index 86f2631e6360..10923bf7c8b8 100644
>> --- a/fs/ocfs2/move_extents.c
>> +++ b/fs/ocfs2/move_extents.c
>> @@ -867,6 +867,11 @@ static int __ocfs2_move_extents_range(struct buffer_head *di_bh,
>>   			mlog_errno(ret);
>>   			goto out;
>>   		}
>> +		/*
>> +		 * Invalidate extent cache after moving/defragging to prevent
>> +		 * stale cached data with outdated extent flags.
>> +		 */
>> +		ocfs2_extent_map_trunc(inode, cpos);
>>   
>>   		context->clusters_moved += alloc_size;
>>   next:
> 
>