Re: [PATCH] tracing/user_events: Run BPF program if attached

On Wed, May 17, 2023 at 12:26 PM Linus Torvalds
[off-list ref] wrote:

Also note that this does NOT fix the incorrect RCU walks.

.. this is the patch that I think should go on top of it to fix the
misleading "safe" and the incorrect RCU walk.

NOTE! This adds that

        lockdep_assert_held(&event_mutex);

to user_event_enabler_update() too. It's already there in
user_event_enabler_write(), but I'm not actually convinced this has
gotten enough coverage checking, so I also did it in that caller.

Some callers obviously hold that mutex. Others are much less obvious,
eg that user_event_reg() -> update_enable_bit_for() chain. I *assume*
all the 'class->reg()' callers get the event mutex, but I did not in
any way check that it is true.

So that lockdep annotation should be actually *tested* with lockdep
enabled and somebody doing all these operations.

Final note: I do not know this code *AT*ALL*. I'm literally just going
by "this is the only correct coding pattern to use", not by some
deeper understanding of what the code actually wants to do.

                   Linus