On Tue, May 07, 2019 at 08:26:06AM -0700, Linus Torvalds wrote: > On Mon, May 6, 2019 at 9:15 PM Al Viro <viro@xxxxxxxxxxxxxxxxxx> wrote: > > > > Umm... Where would you put the cutoff for try_dget()? 1G? Because > > 2G-<something relatively small> is risky - have it reached, then > > get the rest of the way to 2G by normal dget() and you've got trouble. > > I'd make the limit be 2G exactly like the page count. Negative counts > are fine - they work exactly like large integers. It's only 0 that is > special. Negative ->d_lockref.count are used for "lockref is dead"... > - add the "limit negative dentries" patches that were already written > for other reasons by Waiman Long. Irrelevant here, IMO - negative or not (and evictable or pinned, for that matter) it's easier solved by having d_alloc() fail on parent's ->d_count overflow. And I'm pretty sure we can hit that crap without creating any negative dentries.
