perf/x86: Fix lockdep warning in for_each_sibling_event() on SPR

On SPR, the load latency event needs an auxiliary event in the same
group to work properly.  There's a check in intel_pmu_hw_config()
for this to iterate sibling events and find a mem-loads-aux event.

The for_each_sibling_event() has a lockdep assert to make sure if it
disabled hardirq or hold leader->ctx->mutex.  This works well if the
given event has a separate leader event since perf_try_init_event()
grabs the leader->ctx->mutex to protect the sibling list.  But it can
cause a problem when the event itself is a leader since the event is
not initialized yet and there's no ctx for the event.

Actually I got a lockdep warning when I run the below command on SPR,
but I guess it could be a NULL pointer dereference.

  $ perf record -d -e cpu/mem-loads/uP true

The code path to the warning is:

  sys_perf_event_open()
    perf_event_alloc()
      perf_init_event()
        perf_try_init_event()
          x86_pmu_event_init()
            hsw_hw_config()
              intel_pmu_hw_config()
                for_each_sibling_event()
                  lockdep_assert_event_ctx()

We don't need for_each_sibling_event() when it's a standalone event.
Let's return the error code directly.

Fixes: f3c0eba ("perf: Add a few assertions")
Reported-by: Greg Thelen <gthelen@google.com>
Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: stable@vger.kernel.org
Link: https://lkml.kernel.org/r/20230704181516.3293665-1-namhyung@kernel.org

Author: namhyung

arch/x86/events/intel/core.c

@@ -3993,6 +3993,13 @@ static int intel_pmu_hw_config(struct perf_event *event)
 		struct perf_event *leader = event->group_leader;
 		struct perf_event *sibling = NULL;
 
+		/*
+		 * When this memload event is also the first event (no group
+		 * exists yet), then there is no aux event before it.
+		 */
+		if (leader == event)
+			return -ENODATA;
+
 		if (!is_mem_loads_aux_event(leader)) {
 			for_each_sibling_event(sibling, leader) {
 				if (is_mem_loads_aux_event(sibling))