Received-SPF: pass (google.com: domain of linux-kernel+bounces-100632-linux.lists.archive=gmail.com@vger.kernel.org designates 2604:1380:45e3:2400::1 as permitted sender) client-ip=2604:1380:45e3:2400::1;
From: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
To: linux-perf-users@vger.kernel.org,
	linux-kernel@vger.kernel.org
Cc: Adrian Hunter <adrian.hunter@intel.com>,
	Alexander Shishkin <alexander.shishkin@linux.intel.com>,
	Arnaldo Carvalho de Melo <acme@kernel.org>,
	Ian Rogers <irogers@google.com>,
	Ingo Molnar <mingo@redhat.com>,
	Jiri Olsa <jolsa@kernel.org>,
	Marco Elver <elver@google.com>,
	Mark Rutland <mark.rutland@arm.com>,
	Namhyung Kim <namhyung@kernel.org>,
	Peter Zijlstra <peterz@infradead.org>,
	Thomas Gleixner <tglx@linutronix.de>,
	Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Subject: [PATCH v2 4/4] perf: Split __perf_pending_irq() out of perf_pending_irq()
Date: Tue, 12 Mar 2024 19:01:52 +0100
Message-ID: <20240312180814.3373778-5-bigeasy@linutronix.de>
In-Reply-To: <20240312180814.3373778-1-bigeasy@linutronix.de>
References: <20240312180814.3373778-1-bigeasy@linutronix.de>
Precedence: bulk
MIME-Version: 1.0
Content-Transfer-Encoding: quoted-printable

perf_pending_irq() invokes perf_event_wakeup() and __perf_pending_irq().
The former is in charge of waking any tasks which wait to be woken up
while the latter disables perf-events.

The irq_work perf_pending_irq(), while this an irq_work, the callback
is invoked in thread context on PREEMPT_RT. This is needed because all
the waking functions (wake_up_all(), kill_fasync()) acquire sleep locks
which must not be used with disabled interrupts.
Disabling events, as done by __perf_pending_irq(), expects a hardirq
context and disabled interrupts. This requirement is not fulfilled on
PREEMPT_RT.

Split functionality based on perf_event::pending_disable into irq_work
named `pending_disable_irq' and invoke it in hardirq context on
PREEMPT_RT. Rename the split out callback to perf_pending_disable().

Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
---
 include/linux/perf_event.h |  1 +
 kernel/events/core.c       | 31 +++++++++++++++++++++++--------
 2 files changed, 24 insertions(+), 8 deletions(-)

diff --git a/include/linux/perf_event.h b/include/linux/perf_event.h
index 24ac6765146c7..c1c6600541657 100644
--- a/include/linux/perf_event.h
+++ b/include/linux/perf_event.h
@@ -783,6 +783,7 @@ struct perf_event {
 	unsigned int			pending_disable;
 	unsigned long			pending_addr;	/* SIGTRAP */
 	struct irq_work			pending_irq;
+	struct irq_work			pending_disable_irq;
 	struct callback_head		pending_task;
 	unsigned int			pending_work;
=20
diff --git a/kernel/events/core.c b/kernel/events/core.c
index 806514d76d8dc..9aafb949fa100 100644
--- a/kernel/events/core.c
+++ b/kernel/events/core.c
@@ -2449,7 +2449,7 @@ static void __perf_event_disable(struct perf_event *e=
vent,
  * hold the top-level event's child_mutex, so any descendant that
  * goes to exit will block in perf_event_exit_event().
  *
- * When called from perf_pending_irq it's OK because event->ctx
+ * When called from perf_pending_disable it's OK because event->ctx
  * is the current context on this CPU and preemption is disabled,
  * hence we can't get into perf_event_task_sched_out for this context.
  */
@@ -2489,7 +2489,7 @@ EXPORT_SYMBOL_GPL(perf_event_disable);
 void perf_event_disable_inatomic(struct perf_event *event)
 {
 	event->pending_disable =3D 1;
-	irq_work_queue(&event->pending_irq);
+	irq_work_queue(&event->pending_disable_irq);
 }
=20
 #define MAX_INTERRUPTS (~0ULL)
@@ -5175,6 +5175,7 @@ static void perf_addr_filters_splice(struct perf_even=
t *event,
 static void _free_event(struct perf_event *event)
 {
 	irq_work_sync(&event->pending_irq);
+	irq_work_sync(&event->pending_disable_irq);
=20
 	unaccount_event(event);
=20
@@ -6711,7 +6712,7 @@ static void perf_sigtrap(struct perf_event *event)
 /*
  * Deliver the pending work in-event-context or follow the context.
  */
-static void __perf_pending_irq(struct perf_event *event)
+static void __perf_pending_disable(struct perf_event *event)
 {
 	int cpu =3D READ_ONCE(event->oncpu);
=20
@@ -6749,11 +6750,26 @@ static void __perf_pending_irq(struct perf_event *e=
vent)
 	 *				  irq_work_queue(); // FAILS
 	 *
 	 *  irq_work_run()
-	 *    perf_pending_irq()
+	 *    perf_pending_disable()
 	 *
 	 * But the event runs on CPU-B and wants disabling there.
 	 */
-	irq_work_queue_on(&event->pending_irq, cpu);
+	irq_work_queue_on(&event->pending_disable_irq, cpu);
+}
+
+static void perf_pending_disable(struct irq_work *entry)
+{
+	struct perf_event *event =3D container_of(entry, struct perf_event, pendi=
ng_disable_irq);
+	int rctx;
+
+	/*
+	 * If we 'fail' here, that's OK, it means recursion is already disabled
+	 * and we won't recurse 'further'.
+	 */
+	rctx =3D perf_swevent_get_recursion_context();
+	__perf_pending_disable(event);
+	if (rctx >=3D 0)
+		perf_swevent_put_recursion_context(rctx);
 }
=20
 static void perf_pending_irq(struct irq_work *entry)
@@ -6776,8 +6792,6 @@ static void perf_pending_irq(struct irq_work *entry)
 		perf_event_wakeup(event);
 	}
=20
-	__perf_pending_irq(event);
-
 	if (rctx >=3D 0)
 		perf_swevent_put_recursion_context(rctx);
 }
@@ -9566,7 +9580,7 @@ static int __perf_event_overflow(struct perf_event *e=
vent,
 			WARN_ON_ONCE(!atomic_long_inc_not_zero(&event->refcount));
 			task_work_add(current, &event->pending_task, TWA_RESUME);
 			if (in_nmi())
-				irq_work_queue(&event->pending_irq);
+				irq_work_queue(&event->pending_disable_irq);
 		} else if (event->attr.exclude_kernel && valid_sample) {
 			/*
 			 * Should not be able to return to user space without
@@ -11906,6 +11920,7 @@ perf_event_alloc(struct perf_event_attr *attr, int =
cpu,
=20
 	init_waitqueue_head(&event->waitq);
 	init_irq_work(&event->pending_irq, perf_pending_irq);
+	event->pending_disable_irq =3D IRQ_WORK_INIT_HARD(perf_pending_disable);
 	init_task_work(&event->pending_task, perf_pending_task);
=20
 	mutex_init(&event->mmap_mutex);
--=20
2.43.0