On Fri, Mar 01, 2019 at 12:34:50PM +0800, Anand Jain wrote: > btrfs_set_prop() accepts transaction pointer as the first argument, > however in ioctl.c for the purpose of setting the compression property, > we call btrfs_set_prop() with NULL transaction pointer. Down in > the call chain btrfs_setxattr() starts transaction to update the > attribute and also to update the inode. So for better clarity, create > btrfs_set_prop_trans() with no transaction pointer as argument, in > preparation to start transaction here instead of doing it down the > call chain at btrfs_setxattr(). > > Also now the btrfs_set_prop() is a static function. > > Signed-off-by: Anand Jain <anand.jain@xxxxxxxxxx> > --- > v5: change log updated, fix c-style. > v4: born > > Its named btrfs_set_prop_trans() instead of btrfs_set_prop_notrans() > because in the next patch I am starting the transaction in > btrfs_set_prop_trans(). That's right, it should be _trans. > I am ok if this and the next patch are squashed > during integration, as it is separated only for the purpose of easy review. No, the patch separation does make it easier to review.
