summaryrefslogtreecommitdiff
path: root/arch/powerpc/kernel/watchdog.c
blob: 1d82274f7e9f7bcb80a47d08bf0ebb4a168df88a (plain)
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
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
// SPDX-License-Identifier: GPL-2.0
/*
 * Watchdog support on powerpc systems.
 *
 * Copyright 2017, IBM Corporation.
 *
 * This uses code from arch/sparc/kernel/nmi.c and kernel/watchdog.c
 */

#define pr_fmt(fmt) "watchdog: " fmt

#include <linux/kernel.h>
#include <linux/param.h>
#include <linux/init.h>
#include <linux/percpu.h>
#include <linux/cpu.h>
#include <linux/nmi.h>
#include <linux/module.h>
#include <linux/export.h>
#include <linux/kprobes.h>
#include <linux/hardirq.h>
#include <linux/reboot.h>
#include <linux/slab.h>
#include <linux/kdebug.h>
#include <linux/sched/debug.h>
#include <linux/delay.h>
#include <linux/smp.h>

#include <asm/paca.h>

/*
 * The powerpc watchdog ensures that each CPU is able to service timers.
 * The watchdog sets up a simple timer on each CPU to run once per timer
 * period, and updates a per-cpu timestamp and a "pending" cpumask. This is
 * the heartbeat.
 *
 * Then there are two systems to check that the heartbeat is still running.
 * The local soft-NMI, and the SMP checker.
 *
 * The soft-NMI checker can detect lockups on the local CPU. When interrupts
 * are disabled with local_irq_disable(), platforms that use soft-masking
 * can leave hardware interrupts enabled and handle them with a masked
 * interrupt handler. The masked handler can send the timer interrupt to the
 * watchdog's soft_nmi_interrupt(), which appears to Linux as an NMI
 * interrupt, and can be used to detect CPUs stuck with IRQs disabled.
 *
 * The soft-NMI checker will compare the heartbeat timestamp for this CPU
 * with the current time, and take action if the difference exceeds the
 * watchdog threshold.
 *
 * The limitation of the soft-NMI watchdog is that it does not work when
 * interrupts are hard disabled or otherwise not being serviced. This is
 * solved by also having a SMP watchdog where all CPUs check all other
 * CPUs heartbeat.
 *
 * The SMP checker can detect lockups on other CPUs. A gobal "pending"
 * cpumask is kept, containing all CPUs which enable the watchdog. Each
 * CPU clears their pending bit in their heartbeat timer. When the bitmask
 * becomes empty, the last CPU to clear its pending bit updates a global
 * timestamp and refills the pending bitmask.
 *
 * In the heartbeat timer, if any CPU notices that the global timestamp has
 * not been updated for a period exceeding the watchdog threshold, then it
 * means the CPU(s) with their bit still set in the pending mask have had
 * their heartbeat stop, and action is taken.
 *
 * Some platforms implement true NMI IPIs, which can be used by the SMP
 * watchdog to detect an unresponsive CPU and pull it out of its stuck
 * state with the NMI IPI, to get crash/debug data from it. This way the
 * SMP watchdog can detect hardware interrupts off lockups.
 */

static cpumask_t wd_cpus_enabled __read_mostly;

static u64 wd_panic_timeout_tb __read_mostly; /* timebase ticks until panic */
static u64 wd_smp_panic_timeout_tb __read_mostly; /* panic other CPUs */

static u64 wd_timer_period_ms __read_mostly;  /* interval between heartbeat */

static DEFINE_PER_CPU(struct timer_list, wd_timer);
static DEFINE_PER_CPU(u64, wd_timer_tb);

/* SMP checker bits */
static unsigned long __wd_smp_lock;
static cpumask_t wd_smp_cpus_pending;
static cpumask_t wd_smp_cpus_stuck;
static u64 wd_smp_last_reset_tb;

static inline void wd_smp_lock(unsigned long *flags)
{
	/*
	 * Avoid locking layers if possible.
	 * This may be called from low level interrupt handlers at some
	 * point in future.
	 */
	raw_local_irq_save(*flags);
	hard_irq_disable(); /* Make it soft-NMI safe */
	while (unlikely(test_and_set_bit_lock(0, &__wd_smp_lock))) {
		raw_local_irq_restore(*flags);
		spin_until_cond(!test_bit(0, &__wd_smp_lock));
		raw_local_irq_save(*flags);
		hard_irq_disable();
	}
}

static inline void wd_smp_unlock(unsigned long *flags)
{
	clear_bit_unlock(0, &__wd_smp_lock);
	raw_local_irq_restore(*flags);
}

static void wd_lockup_ipi(struct pt_regs *regs)
{
	int cpu = raw_smp_processor_id();
	u64 tb = get_tb();

	pr_emerg("CPU %d Hard LOCKUP\n", cpu);
	pr_emerg("CPU %d TB:%lld, last heartbeat TB:%lld (%lldms ago)\n",
		 cpu, tb, per_cpu(wd_timer_tb, cpu),
		 tb_to_ns(tb - per_cpu(wd_timer_tb, cpu)) / 1000000);
	print_modules();
	print_irqtrace_events(current);
	if (regs)
		show_regs(regs);
	else
		dump_stack();

	/* Do not panic from here because that can recurse into NMI IPI layer */
}

static void set_cpumask_stuck(const struct cpumask *cpumask, u64 tb)
{
	cpumask_or(&wd_smp_cpus_stuck, &wd_smp_cpus_stuck, cpumask);
	cpumask_andnot(&wd_smp_cpus_pending, &wd_smp_cpus_pending, cpumask);
	if (cpumask_empty(&wd_smp_cpus_pending)) {
		wd_smp_last_reset_tb = tb;
		cpumask_andnot(&wd_smp_cpus_pending,
				&wd_cpus_enabled,
				&wd_smp_cpus_stuck);
	}
}
static void set_cpu_stuck(int cpu, u64 tb)
{
	set_cpumask_stuck(cpumask_of(cpu), tb);
}

static void watchdog_smp_panic(int cpu, u64 tb)
{
	unsigned long flags;
	int c;

	wd_smp_lock(&flags);
	/* Double check some things under lock */
	if ((s64)(tb - wd_smp_last_reset_tb) < (s64)wd_smp_panic_timeout_tb)
		goto out;
	if (cpumask_test_cpu(cpu, &wd_smp_cpus_pending))
		goto out;
	if (cpumask_weight(&wd_smp_cpus_pending) == 0)
		goto out;

	pr_emerg("CPU %d detected hard LOCKUP on other CPUs %*pbl\n",
		 cpu, cpumask_pr_args(&wd_smp_cpus_pending));
	pr_emerg("CPU %d TB:%lld, last SMP heartbeat TB:%lld (%lldms ago)\n",
		 cpu, tb, wd_smp_last_reset_tb,
		 tb_to_ns(tb - wd_smp_last_reset_tb) / 1000000);

	if (!sysctl_hardlockup_all_cpu_backtrace) {
		/*
		 * Try to trigger the stuck CPUs, unless we are going to
		 * get a backtrace on all of them anyway.
		 */
		for_each_cpu(c, &wd_smp_cpus_pending) {
			if (c == cpu)
				continue;
			smp_send_nmi_ipi(c, wd_lockup_ipi, 1000000);
		}
		smp_flush_nmi_ipi(1000000);
	}

	/* Take the stuck CPUs out of the watch group */
	set_cpumask_stuck(&wd_smp_cpus_pending, tb);

	wd_smp_unlock(&flags);

	printk_safe_flush();
	/*
	 * printk_safe_flush() seems to require another print
	 * before anything actually goes out to console.
	 */
	if (sysctl_hardlockup_all_cpu_backtrace)
		trigger_allbutself_cpu_backtrace();

	if (hardlockup_panic)
		nmi_panic(NULL, "Hard LOCKUP");

	return;

out:
	wd_smp_unlock(&flags);
}

static void wd_smp_clear_cpu_pending(int cpu, u64 tb)
{
	if (!cpumask_test_cpu(cpu, &wd_smp_cpus_pending)) {
		if (unlikely(cpumask_test_cpu(cpu, &wd_smp_cpus_stuck))) {
			struct pt_regs *regs = get_irq_regs();
			unsigned long flags;

			wd_smp_lock(&flags);

			pr_emerg("CPU %d became unstuck TB:%lld\n",
				 cpu, tb);
			print_irqtrace_events(current);
			if (regs)
				show_regs(regs);
			else
				dump_stack();

			cpumask_clear_cpu(cpu, &wd_smp_cpus_stuck);
			wd_smp_unlock(&flags);
		}
		return;
	}
	cpumask_clear_cpu(cpu, &wd_smp_cpus_pending);
	if (cpumask_empty(&wd_smp_cpus_pending)) {
		unsigned long flags;

		wd_smp_lock(&flags);
		if (cpumask_empty(&wd_smp_cpus_pending)) {
			wd_smp_last_reset_tb = tb;
			cpumask_andnot(&wd_smp_cpus_pending,
					&wd_cpus_enabled,
					&wd_smp_cpus_stuck);
		}
		wd_smp_unlock(&flags);
	}
}

static void watchdog_timer_interrupt(int cpu)
{
	u64 tb = get_tb();

	per_cpu(wd_timer_tb, cpu) = tb;

	wd_smp_clear_cpu_pending(cpu, tb);

	if ((s64)(tb - wd_smp_last_reset_tb) >= (s64)wd_smp_panic_timeout_tb)
		watchdog_smp_panic(cpu, tb);
}

void soft_nmi_interrupt(struct pt_regs *regs)
{
	unsigned long flags;
	int cpu = raw_smp_processor_id();
	u64 tb;

	if (!cpumask_test_cpu(cpu, &wd_cpus_enabled))
		return;

	nmi_enter();

	__this_cpu_inc(irq_stat.soft_nmi_irqs);

	tb = get_tb();
	if (tb - per_cpu(wd_timer_tb, cpu) >= wd_panic_timeout_tb) {
		wd_smp_lock(&flags);
		if (cpumask_test_cpu(cpu, &wd_smp_cpus_stuck)) {
			wd_smp_unlock(&flags);
			goto out;
		}
		set_cpu_stuck(cpu, tb);

		pr_emerg("CPU %d self-detected hard LOCKUP @ %pS\n",
			 cpu, (void *)regs->nip);
		pr_emerg("CPU %d TB:%lld, last heartbeat TB:%lld (%lldms ago)\n",
			 cpu, tb, per_cpu(wd_timer_tb, cpu),
			 tb_to_ns(tb - per_cpu(wd_timer_tb, cpu)) / 1000000);
		print_modules();
		print_irqtrace_events(current);
		show_regs(regs);

		wd_smp_unlock(&flags);

		if (sysctl_hardlockup_all_cpu_backtrace)
			trigger_allbutself_cpu_backtrace();

		if (hardlockup_panic)
			nmi_panic(regs, "Hard LOCKUP");
	}
	if (wd_panic_timeout_tb < 0x7fffffff)
		mtspr(SPRN_DEC, wd_panic_timeout_tb);

out:
	nmi_exit();
}

static void wd_timer_reset(unsigned int cpu, struct timer_list *t)
{
	t->expires = jiffies + msecs_to_jiffies(wd_timer_period_ms);
	if (wd_timer_period_ms > 1000)
		t->expires = __round_jiffies_up(t->expires, cpu);
	add_timer_on(t, cpu);
}

static void wd_timer_fn(struct timer_list *t)
{
	int cpu = smp_processor_id();

	watchdog_timer_interrupt(cpu);

	wd_timer_reset(cpu, t);
}

void arch_touch_nmi_watchdog(void)
{
	unsigned long ticks = tb_ticks_per_usec * wd_timer_period_ms * 1000;
	int cpu = smp_processor_id();
	u64 tb = get_tb();

	if (tb - per_cpu(wd_timer_tb, cpu) >= ticks) {
		per_cpu(wd_timer_tb, cpu) = tb;
		wd_smp_clear_cpu_pending(cpu, tb);
	}
}
EXPORT_SYMBOL(arch_touch_nmi_watchdog);

static void start_watchdog_timer_on(unsigned int cpu)
{
	struct timer_list *t = per_cpu_ptr(&wd_timer, cpu);

	per_cpu(wd_timer_tb, cpu) = get_tb();

	timer_setup(t, wd_timer_fn, TIMER_PINNED);
	wd_timer_reset(cpu, t);
}

static void stop_watchdog_timer_on(unsigned int cpu)
{
	struct timer_list *t = per_cpu_ptr(&wd_timer, cpu);

	del_timer_sync(t);
}

static int start_wd_on_cpu(unsigned int cpu)
{
	unsigned long flags;

	if (cpumask_test_cpu(cpu, &wd_cpus_enabled)) {
		WARN_ON(1);
		return 0;
	}

	if (!(watchdog_enabled & NMI_WATCHDOG_ENABLED))
		return 0;

	if (!cpumask_test_cpu(cpu, &watchdog_cpumask))
		return 0;

	wd_smp_lock(&flags);
	cpumask_set_cpu(cpu, &wd_cpus_enabled);
	if (cpumask_weight(&wd_cpus_enabled) == 1) {
		cpumask_set_cpu(cpu, &wd_smp_cpus_pending);
		wd_smp_last_reset_tb = get_tb();
	}
	wd_smp_unlock(&flags);

	start_watchdog_timer_on(cpu);

	return 0;
}

static int stop_wd_on_cpu(unsigned int cpu)
{
	unsigned long flags;

	if (!cpumask_test_cpu(cpu, &wd_cpus_enabled))
		return 0; /* Can happen in CPU unplug case */

	stop_watchdog_timer_on(cpu);

	wd_smp_lock(&flags);
	cpumask_clear_cpu(cpu, &wd_cpus_enabled);
	wd_smp_unlock(&flags);

	wd_smp_clear_cpu_pending(cpu, get_tb());

	return 0;
}

static void watchdog_calc_timeouts(void)
{
	wd_panic_timeout_tb = watchdog_thresh * ppc_tb_freq;

	/* Have the SMP detector trigger a bit later */
	wd_smp_panic_timeout_tb = wd_panic_timeout_tb * 3 / 2;

	/* 2/5 is the factor that the perf based detector uses */
	wd_timer_period_ms = watchdog_thresh * 1000 * 2 / 5;
}

void watchdog_nmi_stop(void)
{
	int cpu;

	for_each_cpu(cpu, &wd_cpus_enabled)
		stop_wd_on_cpu(cpu);
}

void watchdog_nmi_start(void)
{
	int cpu;

	watchdog_calc_timeouts();
	for_each_cpu_and(cpu, cpu_online_mask, &watchdog_cpumask)
		start_wd_on_cpu(cpu);
}

/*
 * Invoked from core watchdog init.
 */
int __init watchdog_nmi_probe(void)
{
	int err;

	err = cpuhp_setup_state_nocalls(CPUHP_AP_ONLINE_DYN,
					"powerpc/watchdog:online",
					start_wd_on_cpu, stop_wd_on_cpu);
	if (err < 0) {
		pr_warn("could not be initialized");
		return err;
	}
	return 0;
}