On Sun, Oct 04, 2020 at 11:08:11AM -0700, Linus Torvalds wrote: > On Sat, Oct 3, 2020 at 7:36 PM Al Viro <viro@xxxxxxxxxxxxxxxxxx> wrote: > > > > Locking and especially control flow in fs/eventpoll.c is > > overcomplicated. As the result, the code has been hard to follow > > and easy to fuck up while modifying. > > Scanning through the patches they all look superficially ok to me, but > I'm wondering how much test coverage you have (because I'm wondering > how much test coverage we have in general for epoll). Besides the in-tree one (tools/testing/selftests/filesystems/epoll) and LTP stuff (testcases/kernel/syscalls/epoll) - only davidel's epoll_test.c. Plus slapped together "let's try to make it go through that codepath" stuff (combined with printks in fs/eventpoll.c)...
