summaryrefslogtreecommitdiff
path: root/arch/powerpc/kernel/watchdog.c
blob: dfb0677644801da9c7d2b417142fed7ba4440220 (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
/*
 * Watchdog support on powerpc systems.
 *
 * Copyright 2017, IBM Corporation.
 *
 * This uses code from arch/sparc/kernel/nmi.c and kernel/watchdog.c
 */
#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 watchdog has a simple timer that runs on each CPU, once per timer
 * period. This is the heartbeat.
 *
 * Then there are checks to see if the heartbeat has not triggered on a CPU
 * for the panic timeout period. Currently the watchdog only supports an
 * SMP check, so the heartbeat only turns on when we have 2 or more CPUs.
 *
 * This is not an NMI watchdog, but Linux uses that name for a generic
 * watchdog in some cases, so NMI gets used in some places.
 */

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);

/*
 * These are for the SMP checker. CPUs clear their pending bit in their
 * heartbeat. If the bitmask becomes empty, the time is noted and the
 * bitmask is refilled.
 *
 * All CPUs clear their bit in the pending mask every timer period.
 * Once all have cleared, the time is noted and the bits are reset.
 * If the time since all clear was greater than the panic timeout,
 * we can panic with the list of stuck CPUs.
 *
 * This will work best with NMI IPIs for crash code so the stuck CPUs
 * can be pulled out to get their backtraces.
 */
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)
{
	pr_emerg("Watchdog CPU:%d Hard LOCKUP\n", raw_smp_processor_id());
	print_modules();
	print_irqtrace_events(current);
	if (regs)
		show_regs(regs);
	else
		dump_stack();

	if (hardlockup_panic)
		nmi_panic(regs, "Hard LOCKUP");
}

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("Watchdog CPU:%d detected Hard LOCKUP other CPUS:%*pbl\n",
			cpu, cpumask_pr_args(&wd_smp_cpus_pending));

	/*
	 * Try to trigger the stuck CPUs.
	 */
	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))) {
			unsigned long flags;

			pr_emerg("Watchdog CPU:%d became unstuck\n", cpu);
			wd_smp_lock(&flags);
			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) {
		per_cpu(wd_timer_tb, cpu) = 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("Watchdog CPU:%d Hard LOCKUP\n", cpu);
		print_modules();
		print_irqtrace_events(current);
		if (regs)
			show_regs(regs);
		else
			dump_stack();

		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(unsigned long data)
{
	struct timer_list *t = this_cpu_ptr(&wd_timer);
	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();

	if (get_tb() - per_cpu(wd_timer_tb, cpu) >= ticks)
		watchdog_timer_interrupt(cpu);
}
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();

	setup_pinned_timer(t, wd_timer_fn, 0);
	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_reconfigure(bool run)
{
	int cpu;

	cpus_read_lock();
	if (!run) {
		for_each_cpu(cpu, &wd_cpus_enabled)
			stop_wd_on_cpu(cpu);
	} else {
		watchdog_calc_timeouts();
		for_each_cpu_and(cpu, cpu_online_mask, &watchdog_cpumask)
			start_wd_on_cpu(cpu);
	}
	cpus_read_unlock();
}

/*
 * This runs after lockup_detector_init() which sets up watchdog_cpumask.
 */
static int __init powerpc_watchdog_init(void)
{
	int err;

	watchdog_calc_timeouts();

	err = cpuhp_setup_state(CPUHP_AP_ONLINE_DYN, "powerpc/watchdog:online",
				start_wd_on_cpu, stop_wd_on_cpu);
	if (err < 0)
		pr_warn("Watchdog could not be initialized");

	return 0;
}
arch_initcall(powerpc_watchdog_init);

static void handle_backtrace_ipi(struct pt_regs *regs)
{
	nmi_cpu_backtrace(regs);
}

static void raise_backtrace_ipi(cpumask_t *mask)
{
	unsigned int cpu;

	for_each_cpu(cpu, mask) {
		if (cpu == smp_processor_id())
			handle_backtrace_ipi(NULL);
		else
			smp_send_nmi_ipi(cpu, handle_backtrace_ipi, 1000000);
	}
}

void arch_trigger_cpumask_backtrace(const cpumask_t *mask, bool exclude_self)
{
	nmi_trigger_cpumask_backtrace(mask, exclude_self, raise_backtrace_ipi);
}