Re: [PATCH] vfs: move dentry shrinking outside the inode lock in 'rmdir()'

[Al Viro <viro@zeniv.linux.org.uk>]

On Sun, May 12, 2024 at 11:45:44AM -0400, James Bottomley wrote:
> On Sat, 2024-05-11 at 20:28 +0100, Al Viro wrote:
> > On Sat, May 11, 2024 at 11:42:34AM -0700, Linus Torvalds wrote:
> > 
> > > so we have another level of locking going on, and my patch only
> > > moved
> > > the dcache pruning outside the lock of the directory we're removing
> > > (not outside the lock of the directory that contains the removed
> > > directory).
> > > 
> > > And that outside lock is the much more important one, I bet.
> > 
> > ... and _that_ is where taking d_delete outside of the lock might
> > take an unpleasant analysis of a lot of code.
> 
> Couldn't you obviate this by doing it from a workqueue?  Even if the
> directory is recreated, the chances are most of the negative dentries
> that were under it will still exist and be removable by the time the
> workqueue runs.

Eviction in general - sure
shrink_dcache_parent() in particular... not really - you'd need to keep
dentry pinned for that and that'd cause all kinds of fun for umount
d_delete() - even worse (you don't want dcache lookups to find that
sucker after rmdir(2) returned success to userland).