On Fri, 21 Jul 2023 09:18:43 -0400
Steven Rostedt [off-list ref] wrote:

OK, I got this working and it appears to pass all my tests. I actually got
it working Wednesday night, but I tried a different approach on Thursday
that got rid of the evenfs_file and only used eventfs_inodes and makes the
files more dynamic. There's still a couple of corner cases that are not
working with this approach (the dentry counters are getting out of sync).
This should not stop this from going in. I'll continue working on that
approach for the next merge cycle. But for now, here's the patch to this
series that works.

Just figured out my bug with my new design. It was in the eventfs_release()
code, where I have a loop that does the dput on the children, but I wasn't
dput(child), I was dput(parent) in that loop!!

Anyway, for this merge window I much rather have this code in, and for the
next merge window I'll add this update.

I can post the new design too, but first let's focus on this.

-- Steve