1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
|
From: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Date: Tue, 12 Mar 2024 19:01:52 +0100
Subject: [PATCH 4/4] perf: Split __perf_pending_irq() out of
perf_pending_irq()
Origin: https://www.kernel.org/pub/linux/kernel/projects/rt/6.8/older/patches-6.8.2-rt11.tar.xz
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().
Tested-by: Marco Elver <elver@google.com>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Reported-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Link: https://lore.kernel.org/r/20240312180814.3373778-5-bigeasy@linutronix.de
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(-)
--- 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;
--- a/kernel/events/core.c
+++ b/kernel/events/core.c
@@ -2449,7 +2449,7 @@ static void __perf_event_disable(struct
* 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 = 1;
- irq_work_queue(&event->pending_irq);
+ irq_work_queue(&event->pending_disable_irq);
}
#define MAX_INTERRUPTS (~0ULL)
@@ -5175,6 +5175,7 @@ static void perf_addr_filters_splice(str
static void _free_event(struct perf_event *event)
{
irq_work_sync(&event->pending_irq);
+ irq_work_sync(&event->pending_disable_irq);
unaccount_event(event);
@@ -6711,7 +6712,7 @@ static void perf_sigtrap(struct perf_eve
/*
* 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 = READ_ONCE(event->oncpu);
@@ -6749,11 +6750,26 @@ static void __perf_pending_irq(struct pe
* 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 = container_of(entry, struct perf_event, pending_disable_irq);
+ int rctx;
+
+ /*
+ * If we 'fail' here, that's OK, it means recursion is already disabled
+ * and we won't recurse 'further'.
+ */
+ rctx = perf_swevent_get_recursion_context();
+ __perf_pending_disable(event);
+ if (rctx >= 0)
+ perf_swevent_put_recursion_context(rctx);
}
static void perf_pending_irq(struct irq_work *entry)
@@ -6776,8 +6792,6 @@ static void perf_pending_irq(struct irq_
perf_event_wakeup(event);
}
- __perf_pending_irq(event);
-
if (rctx >= 0)
perf_swevent_put_recursion_context(rctx);
}
@@ -9572,7 +9586,7 @@ static int __perf_event_overflow(struct
* is processed.
*/
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
@@ -11912,6 +11926,7 @@ perf_event_alloc(struct perf_event_attr
init_waitqueue_head(&event->waitq);
init_irq_work(&event->pending_irq, perf_pending_irq);
+ event->pending_disable_irq = IRQ_WORK_INIT_HARD(perf_pending_disable);
init_task_work(&event->pending_task, perf_pending_task);
mutex_init(&event->mmap_mutex);
|
