[PATCH] ntfs: handle signal interruption in fallocate

[PATCH] ntfs: handle signal interruption in fallocate

[Hongling Zeng]

The ntfs_attr_fallocate() function checks for pending signals during
allocation loops and exits early via 'out' label. However, when a signal
interrupts the operation with err == 0, the function returns 0 (success)
instead of -EINTR.

The signal_pending() checks at the allocation loops jump to 'out' without
setting err = -EINTR, so the function returns success even when interrupted
by a signal.

Set err = -EINTR when jumping to the signal exit path, and only override
when no other error is pending. This ensures:

- Allocation interrupted by signal returns -EINTR
- Allocation that completed successfully before signal arrived returns 0
- Other errors are preserved and not overwritten by -EINTR

Fixes: 495e90fa3348 ("ntfs: update attrib operations")
Cc: stable@vger.kernel.org
Signed-off-by: Hongling Zeng <zenghongling@kylinos.cn>
---
 fs/ntfs/attrib.c | 8 ++++++--
 1 file changed, 6 insertions(+), 2 deletions(-)

diff --git a/fs/ntfs/attrib.c b/fs/ntfs/attrib.c
index 60264833bb63..c898250f63a7 100644
--- a/fs/ntfs/attrib.c
+++ b/fs/ntfs/attrib.c
@@ -5709,7 +5709,7 @@ int ntfs_attr_fallocate(struct ntfs_inode *ni, loff_t start, loff_t byte_len, bo
 				}
 
 				if (signal_pending(current))
-					goto out;
+					goto signal_out;
 
 				vcn += alloc_cnt;
 				try_alloc_cnt -= alloc_cnt;
@@ -5730,7 +5730,7 @@ int ntfs_attr_fallocate(struct ntfs_inode *ni, loff_t start, loff_t byte_len, bo
 		up_write(&ni->runlist.lock);
 		mutex_unlock(&ni->mrec_lock);
 		if (err || signal_pending(current))
-			goto out;
+			goto signal_out;
 
 		vcn += alloc_cnt;
 		try_alloc_cnt -= alloc_cnt;
@@ -5756,4 +5756,8 @@ int ntfs_attr_fallocate(struct ntfs_inode *ni, loff_t start, loff_t byte_len, bo
 	mutex_unlock(&ni->mrec_lock);
 out:
 	return err >= 0 ? 0 : err;
+signal_out:
+	if (!err)
+		err = -EINTR;
+	goto out;
 }
-- 
2.25.1

Re: [PATCH] ntfs: handle signal interruption in fallocate

[liubaolin]

在 2026/8/26 13:59, Hongling Zeng 写道:
> The ntfs_attr_fallocate() function checks for pending signals during
> allocation loops and exits early via 'out' label. However, when a signal
> interrupts the operation with err == 0, the function returns 0 (success)
> instead of -EINTR.
> 
> The signal_pending() checks at the allocation loops jump to 'out' without
> setting err = -EINTR, so the function returns success even when interrupted
> by a signal.
> 
> Set err = -EINTR when jumping to the signal exit path, and only override
> when no other error is pending. This ensures:
> 
> - Allocation interrupted by signal returns -EINTR
> - Allocation that completed successfully before signal arrived returns 0
> - Other errors are preserved and not overwritten by -EINTR
> 
> Fixes: 495e90fa3348 ("ntfs: update attrib operations")
> Cc: stable@vger.kernel.org
> Signed-off-by: Hongling Zeng <zenghongling@kylinos.cn>
> ---
>   fs/ntfs/attrib.c | 8 ++++++--
>   1 file changed, 6 insertions(+), 2 deletions(-)
> 
> diff --git a/fs/ntfs/attrib.c b/fs/ntfs/attrib.c
> index 60264833bb63..c898250f63a7 100644
> --- a/fs/ntfs/attrib.c
> +++ b/fs/ntfs/attrib.c
> @@ -5709,7 +5709,7 @@ int ntfs_attr_fallocate(struct ntfs_inode *ni, loff_t start, loff_t byte_len, bo
>   				}
>   
>   				if (signal_pending(current))
> -					goto out;
> +					goto signal_out;
>   
>   				vcn += alloc_cnt;
>   				try_alloc_cnt -= alloc_cnt;
> @@ -5730,7 +5730,7 @@ int ntfs_attr_fallocate(struct ntfs_inode *ni, loff_t start, loff_t byte_len, bo
>   		up_write(&ni->runlist.lock);
>   		mutex_unlock(&ni->mrec_lock);
>   		if (err || signal_pending(current))
> -			goto out;
> +			goto signal_out;
>   
>   		vcn += alloc_cnt;
>   		try_alloc_cnt -= alloc_cnt;
> @@ -5756,4 +5756,8 @@ int ntfs_attr_fallocate(struct ntfs_inode *ni, loff_t start, loff_t byte_len, bo
>   	mutex_unlock(&ni->mrec_lock);
>   out:
>   	return err >= 0 ? 0 : err;
> +signal_out:
> +	if (!err)
> +		err = -EINTR;
> +	goto out;
>   }

Looks good to me. 
  

Reviewed-by: Baolin Liu <liubaolin@kylinos.cn>

Re: [PATCH] ntfs: handle signal interruption in fallocate

[Namjae Jeon]

On Wed, Aug 26, 2026 at 3:00 PM Hongling Zeng <zenghongling@kylinos.cn> wrote:
>
> The ntfs_attr_fallocate() function checks for pending signals during
> allocation loops and exits early via 'out' label. However, when a signal
> interrupts the operation with err == 0, the function returns 0 (success)
> instead of -EINTR.
>
> The signal_pending() checks at the allocation loops jump to 'out' without
> setting err = -EINTR, so the function returns success even when interrupted
> by a signal.
>
> Set err = -EINTR when jumping to the signal exit path, and only override
> when no other error is pending. This ensures:
>
> - Allocation interrupted by signal returns -EINTR
> - Allocation that completed successfully before signal arrived returns 0
> - Other errors are preserved and not overwritten by -EINTR
>
> Fixes: 495e90fa3348 ("ntfs: update attrib operations")
> Cc: stable@vger.kernel.org
> Signed-off-by: Hongling Zeng <zenghongling@kylinos.cn>
Applid it to #ntfs-next.
Thanks!