Re: [PATCH] eventfs: Have inodes have unique inode numbers

[Steven Rostedt <rostedt@xxxxxxxxxxx>]

On Sat, 27 Jan 2024 13:47:32 -0800
Linus Torvalds <torvalds@xxxxxxxxxxxxxxxxxxxx> wrote:

> There are tons of other 'ei->dentry' uses, and I didn't look at those.
> Baby steps. But this *seems* like an obvious cleanup, and many small
> obvious cleanups later and perhaps the 'ei->dentry' pointer (and the
> '->d_children[]' array) can eventually go away. They should all be
> entirely useless - there's really no reason for a filesystem to hold
> on to back-pointers of dentries.

I was working on getting rid of ei->dentry, but then I hit:

void eventfs_remove_dir(struct eventfs_inode *ei)
{
	struct dentry *dentry;

	if (!ei)
		return;

	mutex_lock(&eventfs_mutex);
	dentry = ei->dentry;
	eventfs_remove_rec(ei, 0);
	mutex_unlock(&eventfs_mutex);

	/*
	 * If any of the ei children has a dentry, then the ei itself
	 * must have a dentry.
	 */
	if (dentry)
		simple_recursive_removal(dentry, NULL);
}

Where it deletes the all the existing dentries in a tree. Is this a
valid place to keep ei->dentry? I believe this is what makes the
directory disappear from the user's view. But the ei->dentry is there to
know that it is in the user's view to begin with.

-- Steve