Re: [PATCH 15/79] bfs: switch to new ctime accessors

[Jan Kara]

On Thu 22-06-23 08:51:58, Jeff Layton wrote:
> On Thu, 2023-06-22 at 14:30 +0200, Jan Kara wrote:
> > On Wed 21-06-23 12:57:19, Jeff Layton wrote:
> > > On Wed, 2023-06-21 at 18:48 +0200, Jan Kara wrote:
> > > > On Wed 21-06-23 10:45:28, Jeff Layton wrote:
> > > > > In later patches, we're going to change how the ctime.tv_nsec field is
> > > > > utilized. Switch to using accessor functions instead of raw accesses of
> > > > > inode->i_ctime.
> > > > > 
> > > > > Signed-off-by: Jeff Layton <jlayton@xxxxxxxxxx>
> > > > 
> > > > ...
> > > > 
> > > > > diff --git a/fs/bfs/inode.c b/fs/bfs/inode.c
> > > > > index 1926bec2c850..c964316be32b 100644
> > > > > --- a/fs/bfs/inode.c
> > > > > +++ b/fs/bfs/inode.c
> > > > > @@ -82,10 +82,10 @@ struct inode *bfs_iget(struct super_block *sb, unsigned long ino)
> > > > >  	inode->i_blocks = BFS_FILEBLOCKS(di);
> > > > >  	inode->i_atime.tv_sec =  le32_to_cpu(di->i_atime);
> > > > >  	inode->i_mtime.tv_sec =  le32_to_cpu(di->i_mtime);
> > > > > -	inode->i_ctime.tv_sec =  le32_to_cpu(di->i_ctime);
> > > > > +	inode_ctime_set_sec(inode, le32_to_cpu(di->i_ctime));
> > > > >  	inode->i_atime.tv_nsec = 0;
> > > > >  	inode->i_mtime.tv_nsec = 0;
> > > > > -	inode->i_ctime.tv_nsec = 0;
> > > > > +	inode_ctime_set_nsec(inode, 0);
> > > > 
> > > > So I'm somewhat wondering here - in other filesystem you construct
> > > > timespec64 and then use inode_ctime_set(). Here you use
> > > > inode_ctime_set_sec() + inode_ctime_set_nsec(). What's the benefit? It
> > > > seems these two functions are not used that much some maybe we could just
> > > > live with just inode_ctime_set() and constructing timespec64 when needed?
> > > > 
> > > > 								Honza
> > > 
> > > The main advantage is that by using that, I didn't need to do quite so
> > > much of this conversion by hand. My coccinelle skills are pretty
> > > primitive. I went with whatever conversion was going to give minimal
> > > changes, to the existing accesses for the most part.
> > > 
> > > We could certainly do it the way you suggest, it just means having to
> > > re-touch a lot of this code by hand, or someone with better coccinelle
> > > chops suggesting a way to declare a temporary variables in place.
> > 
> > Well, maybe temporary variables aren't that convenient but we could provide
> > function setting ctime from sec & nsec value without having to declare
> > temporary timespec64? Attached is a semantic patch that should deal with
> > that - at least it seems to handle all the cases I've found.
> > 
> 
> Ok, let me try respinning this with your cocci script and see how it
> looks.
> 
> Damien also suggested in a reply to the zonefs patch a preference for
> the naming style you have above. Should I also rename these like?
> 
>     inode_ctime_peek -> inode_get_ctime
>     inode_ctime_set -> inode_set_ctime
> 
> This would be the time to change it if that's preferred.

I don't really care much so whatever you decide is better :)

								Honza
-- 
Jan Kara <jack@xxxxxxxx>
SUSE Labs, CR