Re: [PATCH v2] btrfs: emit BTRFS_SEND_C_FILEATTR during full send

[Qu Wenruo <wqu@suse.com>]

在 2026/6/17 16:26, Mateusz Poliwczak 写道:
> Currently BTRFS_SEND_C_FILEATTR is only emitted during incremental
> send when fileattr change between snapshots. This change makes sure
> we also emit BTRFS_SEND_C_FILEATTR when sending the full snapshot.
> 
> Without this there is no way to gain access to the fileattr of a file
> e.g. the FS_NOCOW_FL flag without doing a separate FS_IOC_GETFLAGS
> syscall, which might not be even possible if the stream is send over
> network or such.

Unfortunately, btrfs receive will just ignore those fileattrs.

It's a known limit:
https://github.com/kdave/btrfs-progs/issues/1057#issuecomment-3492918443

The reason is that different file attrs will require completely 
different handling.
 From NODATACOW/NODATASUM which can only be set for empty file, to 
immutable flags handling.

So this patch will not do any difference, and the whole situation is a 
very deep rabbit hole.

If you want to address this problem, please check this issue first:

https://github.com/kdave/btrfs-progs/issues/526

> 
> Signed-off-by: Mateusz Poliwczak <mpoliwczak34@gmail.com>
> ---
>   fs/btrfs/send.c | 4 +++-
>   1 file changed, 3 insertions(+), 1 deletion(-)
> 
> diff --git a/fs/btrfs/send.c b/fs/btrfs/send.c
> index 3ae480c7474b..effbe7ac1a81 100644
> --- a/fs/btrfs/send.c
> +++ b/fs/btrfs/send.c
> @@ -6575,8 +6575,10 @@ static int finish_inode_if_needed(struct send_ctx *sctx, bool at_end)
>   
>   	if (!sctx->parent_root || sctx->cur_inode_new) {
>   		need_chown = true;
> -		if (!S_ISLNK(sctx->cur_inode_mode))
> +		if (!S_ISLNK(sctx->cur_inode_mode)) {
>   			need_chmod = true;
> +			need_fileattr = true;
> +		}
>   		if (sctx->cur_inode_next_write_offset == sctx->cur_inode_size)
>   			need_truncate = false;
>   	} else {