On Wed, Jul 08, 2020 at 06:16:34PM +0300, Alexander Shishkin wrote: > Currently, perf requires one file descriptor per event. In large groups, > this may mean running into the limit on open file descriptors. However, > the sibling events in a group only need file descriptors for the initial > configuration stage, after which they may not be needed any more. > > This adds an opt-in flag to the perf_event_open() syscall to retain > sibling events after their file descriptors are closed. In this case, the > actual events will be closed with the group leader.
So having the 1:1 relation with filedesc imposes a resource limit on userspace. This patch breaks that and enables a user to basically DoS the system by creating unbound events.
