On Tue, Aug 20, 2019 at 04:58:34PM -0700, David Miller wrote:
From: Matthew Wilcox <willy@infradead.org>
Date: Tue, 20 Aug 2019 15:32:50 -0700
quoted
- idr_replace(&head->handle_idr, fnew, fnew->handle);
+ xa_store(&head->filters, fnew->handle, fnew, 0);
Passing a gfp_t of zero? :-)
Yes! We know we'll never do an allocation here because we're replacing
an entry that already exists. It wouldn't harm us to pass a real
GFP flag, so I'll probably just change that. It might help a future
implementation, and it will definitely save confusion.