Re: [RFC PATCH 01/35] vfs: clean up dedup

2018-04-12 Thread Miklos Szeredi 
On Thu, Apr 12, 2018 at 6:25 PM, Matthew Wilcox  wrote:
> On Thu, Apr 12, 2018 at 05:07:52PM +0200, Miklos Szeredi wrote:
>> +ssize_t vfs_dedupe_file_range_one(struct file *src_file, u64 src_pos, u64 
>> len,
>> +   struct file *dst_file, u64 dst_pos)
>
> Why u64 instead of loff_t?

Peculiarity of f_op->dedupe_file_range().

I think a cleanup of copyfile/clone/dedupe interfaces would be good.
Perhaps merge them into one?  Or just merge copyfile/clone?  Even if
not merging it should make sense to make argument types and argument
order the same.

Thanks,
Miklos

Re: [RFC PATCH 01/35] vfs: clean up dedup

2018-04-12 Thread Miklos Szeredi 
On Thu, Apr 12, 2018 at 6:25 PM, Matthew Wilcox  wrote:
> On Thu, Apr 12, 2018 at 05:07:52PM +0200, Miklos Szeredi wrote:
>> +ssize_t vfs_dedupe_file_range_one(struct file *src_file, u64 src_pos, u64 
>> len,
>> +   struct file *dst_file, u64 dst_pos)
>
> Why u64 instead of loff_t?

Peculiarity of f_op->dedupe_file_range().

I think a cleanup of copyfile/clone/dedupe interfaces would be good.
Perhaps merge them into one?  Or just merge copyfile/clone?  Even if
not merging it should make sense to make argument types and argument
order the same.

Thanks,
Miklos

Re: [RFC PATCH 01/35] vfs: clean up dedup

2018-04-12 Thread Matthew Wilcox 
On Thu, Apr 12, 2018 at 05:07:52PM +0200, Miklos Szeredi wrote:
> +ssize_t vfs_dedupe_file_range_one(struct file *src_file, u64 src_pos, u64 
> len,
> +   struct file *dst_file, u64 dst_pos)

Why u64 instead of loff_t?

Re: [RFC PATCH 01/35] vfs: clean up dedup

2018-04-12 Thread Matthew Wilcox 
On Thu, Apr 12, 2018 at 05:07:52PM +0200, Miklos Szeredi wrote:
> +ssize_t vfs_dedupe_file_range_one(struct file *src_file, u64 src_pos, u64 
> len,
> +   struct file *dst_file, u64 dst_pos)

Why u64 instead of loff_t?