On Fri, 2021-06-11 at 15:07 +0200, Miklos Szeredi wrote:

On Wed, 9 Jun 2021 at 10:50, Ian Kent [off-list ref] wrote:

quoted

If there are many lookups for non-existent paths these negative
lookups
can lead to a lot of overhead during path walks.

The VFS allows dentries to be created as negative and hashed, and
caches
them so they can be used to reduce the fairly high overhead
alloc/free
cycle that occurs during these lookups.

Use the kernfs node parent revision to identify if a change has
been
made to the containing directory so that the negative dentry can be
discarded and the lookup redone.

Signed-off-by: Ian Kent <raven@themaw.net>
---
 fs/kernfs/dir.c |   52 ++++++++++++++++++++++++++++++++-----------
---------
 1 file changed, 32 insertions(+), 20 deletions(-)

diff --git a/fs/kernfs/dir.c b/fs/kernfs/dir.c
index b3d1bc0f317d0..4f037456a8e17 100644
--- a/fs/kernfs/dir.c
+++ b/fs/kernfs/dir.c

@@ -1039,9 +1039,28 @@ static int kernfs_dop_revalidate(struct

dentry *dentry, unsigned int flags)
        if (flags & LOOKUP_RCU)
                return -ECHILD;

-       /* Always perform fresh lookup for negatives */
-       if (d_really_is_negative(dentry))
-               goto out_bad_unlocked;
+       /* Negative hashed dentry? */
+       if (d_really_is_negative(dentry)) {
+               struct dentry *d_parent = dget_parent(dentry);
+               struct kernfs_node *parent;
+
+               /* If the kernfs parent node has changed discard
and
+                * proceed to ->lookup.
+                */
+               parent = kernfs_dentry_node(d_parent);
+               if (parent) {
+                       if (kernfs_dir_changed(parent, dentry)) {

Perhaps add a note about this being dependent on parent of a negative
dentry never changing.

Which of course it it can change, at any time.

If this was backported to a kernel where this assumption doesn't
hold,
there would be a mathematical chance of a false negative.

Isn't this a cunning way of saying "in thinking about the move case
you've forgotten about the obvious common case, just put back taking
the read lock already, at least for the check"?

Ian