[PATCH] fuse: Move same-superblock check to fuse_copy_file_range

[PATCH] fuse: Move same-superblock check to fuse_copy_file_range

[Chunsheng Luo]

The copy_file_range COPY_FILE_SPLICE capability allows filesystems to
handle cross-superblock copy. However, in the current fuse implementation,
__fuse_copy_file_range accesses src_file->private_data under the assumption
that it points to a fuse_file structure. When the source file belongs to a
non-FUSE filesystem, it will leads to kernel panics.

To resolve this, move the same-superblock check from __fuse_copy_file_range
to fuse_copy_file_range to ensure both files belong to the same fuse
superblock before accessing private_data.

Signed-off-by: Chunsheng Luo <luochunsheng@ustc.edu>
---
 fs/fuse/file.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/fs/fuse/file.c b/fs/fuse/file.c
index 95275a1e2f54..a29f1b84f11b 100644
--- a/fs/fuse/file.c
+++ b/fs/fuse/file.c
@@ -2984,9 +2984,6 @@ static ssize_t __fuse_copy_file_range(struct file *file_in, loff_t pos_in,
 	if (fc->no_copy_file_range)
 		return -EOPNOTSUPP;
 
-	if (file_inode(file_in)->i_sb != file_inode(file_out)->i_sb)
-		return -EXDEV;
-
 	inode_lock(inode_in);
 	err = fuse_writeback_range(inode_in, pos_in, pos_in + len - 1);
 	inode_unlock(inode_in);
@@ -3066,9 +3063,12 @@ static ssize_t fuse_copy_file_range(struct file *src_file, loff_t src_off,
 {
 	ssize_t ret;
 
+	if (file_inode(src_file)->i_sb != file_inode(dst_file)->i_sb)
+		return splice_copy_file_range(src_file, src_off, dst_file,
+					     dst_off, len);
+
 	ret = __fuse_copy_file_range(src_file, src_off, dst_file, dst_off,
 				     len, flags);
-
 	if (ret == -EOPNOTSUPP || ret == -EXDEV)
 		ret = splice_copy_file_range(src_file, src_off, dst_file,
 					     dst_off, len);
-- 
2.43.0

Re: [PATCH] fuse: Move same-superblock check to fuse_copy_file_range

[Bernd Schubert]

On 8/6/25 15:52, Chunsheng Luo wrote:
> The copy_file_range COPY_FILE_SPLICE capability allows filesystems to
> handle cross-superblock copy. However, in the current fuse implementation,
> __fuse_copy_file_range accesses src_file->private_data under the assumption
> that it points to a fuse_file structure. When the source file belongs to a
> non-FUSE filesystem, it will leads to kernel panics.
> 
> To resolve this, move the same-superblock check from __fuse_copy_file_range
> to fuse_copy_file_range to ensure both files belong to the same fuse
> superblock before accessing private_data.
> 
> Signed-off-by: Chunsheng Luo <luochunsheng@ustc.edu>
> ---
>  fs/fuse/file.c | 8 ++++----
>  1 file changed, 4 insertions(+), 4 deletions(-)
> 
> diff --git a/fs/fuse/file.c b/fs/fuse/file.c
> index 95275a1e2f54..a29f1b84f11b 100644
> --- a/fs/fuse/file.c
> +++ b/fs/fuse/file.c
> @@ -2984,9 +2984,6 @@ static ssize_t __fuse_copy_file_range(struct file *file_in, loff_t pos_in,
>  	if (fc->no_copy_file_range)
>  		return -EOPNOTSUPP;
>  
> -	if (file_inode(file_in)->i_sb != file_inode(file_out)->i_sb)
> -		return -EXDEV;
> -
>  	inode_lock(inode_in);
>  	err = fuse_writeback_range(inode_in, pos_in, pos_in + len - 1);
>  	inode_unlock(inode_in);
> @@ -3066,9 +3063,12 @@ static ssize_t fuse_copy_file_range(struct file *src_file, loff_t src_off,
>  {
>  	ssize_t ret;
>  
> +	if (file_inode(src_file)->i_sb != file_inode(dst_file)->i_sb)
> +		return splice_copy_file_range(src_file, src_off, dst_file,
> +					     dst_off, len);
> +
>  	ret = __fuse_copy_file_range(src_file, src_off, dst_file, dst_off,
>  				     len, flags);
> -
>  	if (ret == -EOPNOTSUPP || ret == -EXDEV)
>  		ret = splice_copy_file_range(src_file, src_off, dst_file,
>  					     dst_off, len);

I guess you can remove the check EXDEV here?

Re: [PATCH] fuse: Move same-superblock check to fuse_copy_file_range

[Luis Henriques]

On Wed, Aug 06 2025, Chunsheng Luo wrote:

> The copy_file_range COPY_FILE_SPLICE capability allows filesystems to
> handle cross-superblock copy. However, in the current fuse implementation,
> __fuse_copy_file_range accesses src_file->private_data under the assumption
> that it points to a fuse_file structure. When the source file belongs to a
> non-FUSE filesystem, it will leads to kernel panics.

I wonder if you have actually seen this kernel panic happening.  It seems
like the code you're moving into fuse_copy_file_range() shouldn't be
needed as the same check is already done in generic_copy_file_checks()
(which is called from vfs_copy_file_range()).

Either way, I think your change to fuse_copy_file_range() could be
simplified with something like:

	ssize_t ret = -EXDEV;

	if (file_inode(src_file)->i_sb == file_inode(dst_file)->i_sb)
		ret = __fuse_copy_file_range(src_file, src_off, dst_file, dst_off,
					     len, flags);

	if (ret == -EOPNOTSUPP || ret == -EXDEV)
		ret = splice_copy_file_range(src_file, src_off, dst_file,
					     dst_off, len);

But again, my understanding is that this should never happen in practice
and that the superblock check could even be removed from
__fuse_copy_file_range().

Cheers,
-- 
Luís

>
> To resolve this, move the same-superblock check from __fuse_copy_file_range
> to fuse_copy_file_range to ensure both files belong to the same fuse
> superblock before accessing private_data.
>
> Signed-off-by: Chunsheng Luo <luochunsheng@ustc.edu>
> ---
>  fs/fuse/file.c | 8 ++++----
>  1 file changed, 4 insertions(+), 4 deletions(-)
>
> diff --git a/fs/fuse/file.c b/fs/fuse/file.c
> index 95275a1e2f54..a29f1b84f11b 100644
> --- a/fs/fuse/file.c
> +++ b/fs/fuse/file.c
> @@ -2984,9 +2984,6 @@ static ssize_t __fuse_copy_file_range(struct file *file_in, loff_t pos_in,
>  	if (fc->no_copy_file_range)
>  		return -EOPNOTSUPP;
>  
> -	if (file_inode(file_in)->i_sb != file_inode(file_out)->i_sb)
> -		return -EXDEV;
> -
>  	inode_lock(inode_in);
>  	err = fuse_writeback_range(inode_in, pos_in, pos_in + len - 1);
>  	inode_unlock(inode_in);
> @@ -3066,9 +3063,12 @@ static ssize_t fuse_copy_file_range(struct file *src_file, loff_t src_off,
>  {
>  	ssize_t ret;
>  
> +	if (file_inode(src_file)->i_sb != file_inode(dst_file)->i_sb)
> +		return splice_copy_file_range(src_file, src_off, dst_file,
> +					     dst_off, len);
> +
>  	ret = __fuse_copy_file_range(src_file, src_off, dst_file, dst_off,
>  				     len, flags);
> -
>  	if (ret == -EOPNOTSUPP || ret == -EXDEV)
>  		ret = splice_copy_file_range(src_file, src_off, dst_file,
>  					     dst_off, len);
> -- 
> 2.43.0
>

Re: [PATCH] fuse: Move same-superblock check to fuse_copy_file_range

[Chunsheng Luo]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain; charset=y, Size: 4226 bytes --]

On Thu, Aug 07 2025,  Luis Henriques wrote:

>> The copy_file_range COPY_FILE_SPLICE capability allows filesystems to
>> handle cross-superblock copy. However, in the current fuse implementation,
>> __fuse_copy_file_range accesses src_file->private_data under the assumption
>> that it points to a fuse_file structure. When the source file belongs to a
>> non-FUSE filesystem, it will leads to kernel panics.
>
> I wonder if you have actually seen this kernel panic happening.  It seems
> like the code you're moving into fuse_copy_file_range() shouldn't be
> needed as the same check is already done in generic_copy_file_checks()
> (which is called from vfs_copy_file_range()).
> 
> Either way, I think your change to fuse_copy_file_range() could be
> simplified with something like:
> 
> 	ssize_t ret = -EXDEV;
> 
> 	if (file_inode(src_file)->i_sb == file_inode(dst_file)->i_sb)
> 		ret = __fuse_copy_file_range(src_file, src_off, dst_file, dst_off,
> 					     len, flags);
> 
> 	if (ret == -EOPNOTSUPP || ret == -EXDEV)
> 		ret = splice_copy_file_range(src_file, src_off, dst_file,
> 					     dst_off, len);
> 
> But again, my understanding is that this should never happen in practice
> and that the superblock check could even be removed from
> __fuse_copy_file_range().
> 
> Cheers,
> -- 
> Luís
>

Yes, now copy_file_range won't crash.

generic_copy_file_checks:
	/*
	 * We allow some filesystems to handle cross sb copy, but passing
	 * a file of the wrong filesystem type to filesystem driver can result
	 * in an attempt to dereference the wrong type of ->private_data, so
	 * avoid doing that until we really have a good reason.
	 *
	 * nfs and cifs define several different file_system_type structures
	 * and several different sets of file_operations, but they all end up
	 * using the same ->copy_file_range() function pointer.
	 */
	if (flags & COPY_FILE_SPLICE) {
		/* cross sb splice is allowed */
	} else if (file_out->f_op->copy_file_range) {
		if (file_in->f_op->copy_file_range !=
		    file_out->f_op->copy_file_range)
			return -EXDEV;
	} else if (file_inode(file_in)->i_sb != file_inode(file_out)->i_sb) {
		return -EXDEV;
	}

generic_copy_file_checks mentions that now allows some filesystems to handle cross-sb copy.

code: 
	} else if (file_out->f_op->copy_file_range) {
		if (file_in->f_op->copy_file_range !=
		    file_out->f_op->copy_file_range)
			return -EXDEV;
			
If the same filesystem is satisfied but the sb is not same, it will go to fuse_copy_file_range,
so fuse_copy_file_range needs to handle this situation.

Sorry, there is an mistake with my patch log description. __fuse_copy_file_range does not exist that
the source file is a NON-Fuse filesystem, so It can safely use ->private_data.

Therefore, this patch does not need.

Thanks
Chunsheng Luo

>>
>> To resolve this, move the same-superblock check from __fuse_copy_file_range
>> to fuse_copy_file_range to ensure both files belong to the same fuse
>> superblock before accessing private_data.
>>
>> Signed-off-by: Chunsheng Luo <luochunsheng@ustc.edu>
>> ---
>>  fs/fuse/file.c | 8 ++++----
>>  1 file changed, 4 insertions(+), 4 deletions(-)
>>
>> diff --git a/fs/fuse/file.c b/fs/fuse/file.c
>> index 95275a1e2f54..a29f1b84f11b 100644
>> --- a/fs/fuse/file.c
>> +++ b/fs/fuse/file.c
>> @@ -2984,9 +2984,6 @@ static ssize_t __fuse_copy_file_range(struct file *file_in, loff_t pos_in,
>>  	if (fc->no_copy_file_range)
>>  		return -EOPNOTSUPP;
>>  
>> -	if (file_inode(file_in)->i_sb != file_inode(file_out)->i_sb)
>> -		return -EXDEV;
>> -
>>  	inode_lock(inode_in);
>>  	err = fuse_writeback_range(inode_in, pos_in, pos_in + len - 1);
>>  	inode_unlock(inode_in);
>> @@ -3066,9 +3063,12 @@ static ssize_t fuse_copy_file_range(struct file *src_file, loff_t src_off,
>>  {
>>  	ssize_t ret;
>>  
>> +	if (file_inode(src_file)->i_sb != file_inode(dst_file)->i_sb)
>> +		return splice_copy_file_range(src_file, src_off, dst_file,
>> +					     dst_off, len);
>> +
>>  	ret = __fuse_copy_file_range(src_file, src_off, dst_file, dst_off,
>>  				     len, flags);
>> -
>>  	if (ret == -EOPNOTSUPP || ret == -EXDEV)
>>  		ret = splice_copy_file_range(src_file, src_off, dst_file,
>>  					     dst_off, len);
>> -- 
>> 2.43.0
>>

Re: [PATCH] fuse: Move same-superblock check to fuse_copy_file_range

[Amir Goldstein]

On Thu, Aug 7, 2025 at 2:49 PM Chunsheng Luo <luochunsheng@ustc.edu> wrote:
>
> On Thu, Aug 07 2025,  Luis Henriques wrote:
>
> >> The copy_file_range COPY_FILE_SPLICE capability allows filesystems to
> >> handle cross-superblock copy. However, in the current fuse implementation,
> >> __fuse_copy_file_range accesses src_file->private_data under the assumption
> >> that it points to a fuse_file structure. When the source file belongs to a
> >> non-FUSE filesystem, it will leads to kernel panics.
> >
> > I wonder if you have actually seen this kernel panic happening.  It seems
> > like the code you're moving into fuse_copy_file_range() shouldn't be
> > needed as the same check is already done in generic_copy_file_checks()
> > (which is called from vfs_copy_file_range()).
> >
> > Either way, I think your change to fuse_copy_file_range() could be
> > simplified with something like:
> >
> >       ssize_t ret = -EXDEV;
> >
> >       if (file_inode(src_file)->i_sb == file_inode(dst_file)->i_sb)
> >               ret = __fuse_copy_file_range(src_file, src_off, dst_file, dst_off,
> >                                            len, flags);
> >
> >       if (ret == -EOPNOTSUPP || ret == -EXDEV)
> >               ret = splice_copy_file_range(src_file, src_off, dst_file,
> >                                            dst_off, len);
> >
> > But again, my understanding is that this should never happen in practice
> > and that the superblock check could even be removed from
> > __fuse_copy_file_range().
> >
> > Cheers,
> > --
> > Luís
> >
>
> Yes, now copy_file_range won't crash.
>
> generic_copy_file_checks:
>         /*
>          * We allow some filesystems to handle cross sb copy, but passing
>          * a file of the wrong filesystem type to filesystem driver can result
>          * in an attempt to dereference the wrong type of ->private_data, so
>          * avoid doing that until we really have a good reason.
>          *
>          * nfs and cifs define several different file_system_type structures
>          * and several different sets of file_operations, but they all end up
>          * using the same ->copy_file_range() function pointer.
>          */
>         if (flags & COPY_FILE_SPLICE) {
>                 /* cross sb splice is allowed */
>         } else if (file_out->f_op->copy_file_range) {
>                 if (file_in->f_op->copy_file_range !=
>                     file_out->f_op->copy_file_range)
>                         return -EXDEV;
>         } else if (file_inode(file_in)->i_sb != file_inode(file_out)->i_sb) {
>                 return -EXDEV;
>         }
>
> generic_copy_file_checks mentions that now allows some filesystems to handle cross-sb copy.
>
> code:
>         } else if (file_out->f_op->copy_file_range) {
>                 if (file_in->f_op->copy_file_range !=
>                     file_out->f_op->copy_file_range)
>                         return -EXDEV;
>
> If the same filesystem is satisfied but the sb is not same, it will go to fuse_copy_file_range,
> so fuse_copy_file_range needs to handle this situation.
>
> Sorry, there is an mistake with my patch log description. __fuse_copy_file_range does not exist that
> the source file is a NON-Fuse filesystem, so It can safely use ->private_data.
>
> Therefore, this patch does not need.

Indeed, this patch makes no sense and does not change any logic at all.

Thanks,
Amir.