Re: [RFC][PATCH v4 15/69] new step_into() flag: WALK_NOFOLLOW

[Al Viro]

On Fri, Mar 13, 2020 at 05:32:32PM -0700, Linus Torvalds wrote:
> I mentioned this last time (perhaps for a different sequence):
> 
> On Fri, Mar 13, 2020 at 4:54 PM Al Viro <viro@xxxxxxxxxxxxxxxxxx> wrote:
> >
> >         if (likely(!d_is_symlink(path->dentry)) ||
> > -          !(flags & WALK_FOLLOW || nd->flags & LOOKUP_FOLLOW)) {
> > +          !(flags & WALK_FOLLOW || nd->flags & LOOKUP_FOLLOW) ||
> > +          flags & WALK_NOFOLLOW) {
> 
> Yes, I know that bitwise operations have higher precedence than the
> logical ones. And I know & (and &&) have higher precedence than | (and
> ||).
> 
> But I have to _think_ about it every time I see code like this.
> 
> I'd really prefer to see
> 
>    if ((a & BIT) || (b & ANOTHER_BIT))
> 
> over the "equivalent" and shorter
> 
>    if (a & BIT || b & ANOTHER_BIT)
> 
> Please make it explicit. It wasn't before either, but it _could_ be.

Not a problem (actually, I'd done that several commits later when I was
rewriting the expression anyway).  Folded the following into it now:

diff --git a/fs/namei.c b/fs/namei.c
index e47b376cf442..79f06be7f5d4 100644
--- a/fs/namei.c
+++ b/fs/namei.c
@@ -1839,8 +1839,8 @@ static inline int step_into(struct nameidata *nd, struct path *path,
 			    int flags, struct inode *inode, unsigned seq)
 {
 	if (likely(!d_is_symlink(path->dentry)) ||
-	   !(flags & WALK_FOLLOW || nd->flags & LOOKUP_FOLLOW) ||
-	   flags & WALK_NOFOLLOW) {
+	   !((flags & WALK_FOLLOW) || (nd->flags & LOOKUP_FOLLOW)) ||
+	   (flags & WALK_NOFOLLOW)) {
 		/* not a symlink or should not follow */
 		path_to_nameidata(path, nd);
 		nd->inode = inode;