xfs: rmap: only XFS_DATA_FORK field would be set shared

[PATCH -next] xfs: rmap: only XFS_DATA_FORK field would be set shared

Posted by Zeng Heng 3 years, 7 months ago

whichfork could be one of the following 3 types:
XFS_DATA_FORK
XFS_ATTR_FORK
XFS_COW_FORK

XFS_COW_FORK is in-memory field instead of on-disk
field, which doesn't need to update in rmap_btree
(already preclude by `xfs_rmap_update_is_needed` ahead).

XFS_ATTR_FORK field doesn't support reflink share.

So here propose to set "whichfork == XFS_DATA_FORK"
as condition.

Signed-off-by: Zeng Heng <zengheng4@huawei.com>
---
 fs/xfs/libxfs/xfs_rmap.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/fs/xfs/libxfs/xfs_rmap.c b/fs/xfs/libxfs/xfs_rmap.c
index 094dfc897ebc..74b34a331cf0 100644
--- a/fs/xfs/libxfs/xfs_rmap.c
+++ b/fs/xfs/libxfs/xfs_rmap.c
@@ -2549,7 +2549,7 @@ xfs_rmap_map_extent(
 	if (!xfs_rmap_update_is_needed(tp->t_mountp, whichfork))
 		return;
 
-	if (whichfork != XFS_ATTR_FORK && xfs_is_reflink_inode(ip))
+	if (whichfork == XFS_DATA_FORK && xfs_is_reflink_inode(ip))
 		type = XFS_RMAP_MAP_SHARED;
 
 	__xfs_rmap_add(tp, type, ip->i_ino, whichfork, PREV);
@@ -2568,7 +2568,7 @@ xfs_rmap_unmap_extent(
 	if (!xfs_rmap_update_is_needed(tp->t_mountp, whichfork))
 		return;
 
-	if (whichfork != XFS_ATTR_FORK && xfs_is_reflink_inode(ip))
+	if (whichfork == XFS_DATA_FORK && xfs_is_reflink_inode(ip))
 		type = XFS_RMAP_UNMAP_SHARED;
 
 	__xfs_rmap_add(tp, type, ip->i_ino, whichfork, PREV);
@@ -2593,7 +2593,7 @@ xfs_rmap_convert_extent(
 	if (!xfs_rmap_update_is_needed(mp, whichfork))
 		return;
 
-	if (whichfork != XFS_ATTR_FORK && xfs_is_reflink_inode(ip))
+	if (whichfork == XFS_DATA_FORK && xfs_is_reflink_inode(ip))
 		type = XFS_RMAP_CONVERT_SHARED;
 
 	__xfs_rmap_add(tp, type, ip->i_ino, whichfork, PREV);
-- 
2.25.1

Re: [PATCH -next] xfs: rmap: only XFS_DATA_FORK field would be set shared

Posted by Darrick J. Wong 3 years, 7 months ago

On Tue, Sep 06, 2022 at 05:35:54PM +0800, Zeng Heng wrote:
> whichfork could be one of the following 3 types:
> XFS_DATA_FORK
> XFS_ATTR_FORK
> XFS_COW_FORK
> 
> XFS_COW_FORK is in-memory field instead of on-disk
> field, which doesn't need to update in rmap_btree
> (already preclude by `xfs_rmap_update_is_needed` ahead).
> 
> XFS_ATTR_FORK field doesn't support reflink share.
> 
> So here propose to set "whichfork == XFS_DATA_FORK"
> as condition.

As you already state, xfs_rmap_update_is_needed() filters out
whichfork==XFS_COW_FORK.  Why is this change necessary?  It looks like
pointless churn to me.

--D

> 
> Signed-off-by: Zeng Heng <zengheng4@huawei.com>
> ---
>  fs/xfs/libxfs/xfs_rmap.c | 6 +++---
>  1 file changed, 3 insertions(+), 3 deletions(-)
> 
> diff --git a/fs/xfs/libxfs/xfs_rmap.c b/fs/xfs/libxfs/xfs_rmap.c
> index 094dfc897ebc..74b34a331cf0 100644
> --- a/fs/xfs/libxfs/xfs_rmap.c
> +++ b/fs/xfs/libxfs/xfs_rmap.c
> @@ -2549,7 +2549,7 @@ xfs_rmap_map_extent(
>  	if (!xfs_rmap_update_is_needed(tp->t_mountp, whichfork))
>  		return;
>  
> -	if (whichfork != XFS_ATTR_FORK && xfs_is_reflink_inode(ip))
> +	if (whichfork == XFS_DATA_FORK && xfs_is_reflink_inode(ip))
>  		type = XFS_RMAP_MAP_SHARED;
>  
>  	__xfs_rmap_add(tp, type, ip->i_ino, whichfork, PREV);
> @@ -2568,7 +2568,7 @@ xfs_rmap_unmap_extent(
>  	if (!xfs_rmap_update_is_needed(tp->t_mountp, whichfork))
>  		return;
>  
> -	if (whichfork != XFS_ATTR_FORK && xfs_is_reflink_inode(ip))
> +	if (whichfork == XFS_DATA_FORK && xfs_is_reflink_inode(ip))
>  		type = XFS_RMAP_UNMAP_SHARED;
>  
>  	__xfs_rmap_add(tp, type, ip->i_ino, whichfork, PREV);
> @@ -2593,7 +2593,7 @@ xfs_rmap_convert_extent(
>  	if (!xfs_rmap_update_is_needed(mp, whichfork))
>  		return;
>  
> -	if (whichfork != XFS_ATTR_FORK && xfs_is_reflink_inode(ip))
> +	if (whichfork == XFS_DATA_FORK && xfs_is_reflink_inode(ip))
>  		type = XFS_RMAP_CONVERT_SHARED;
>  
>  	__xfs_rmap_add(tp, type, ip->i_ino, whichfork, PREV);
> -- 
> 2.25.1
>