On Mon, Mar 14, 2016 at 12:08:31AM +0100, Andreas Gruenbacher wrote:

The xattr representation is the same on disk and at the xattr syscall
layer, and so richacl_from_xattr is used for converting into the
in-memory representation in both cases. The error codes are not the
same when a user supplies an invalid value via setxattr or NFS and
when an invalid xattr is read from disk though. I'll add a parameter
to richacl_from_xattr to make this more explicit.

Better add a wrapper instead of a parameter.

quoted

quoted

+static int
+__ext4_set_richacl(handle_t *handle, struct inode *inode, struct richacl *acl)
+{
+     const int name_index = EXT4_XATTR_INDEX_RICHACL;
+     umode_t mode = inode->i_mode;
+     int retval, size;
+     void *value;
+
+     if (richacl_equiv_mode(acl, &mode) == 0) {
+             inode->i_ctime = ext4_current_time(inode);
+             inode->i_mode = mode;
+             ext4_mark_inode_dirty(handle, inode);
+             return __ext4_remove_richacl(handle, inode);
+     }

Should this check for a NULL acl instead of special casing that
in ext4_set_richacl?

I'm not sure I understand what you mean. When the

ext4_set_richacl checks for a NULL acl pointer and then calls into
__ext4_remove_richacl.  I'd rather have that special casing in one
place.

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs