Loading...
1/* Pseudo NMI support on sparc64 systems.
2 *
3 * Copyright (C) 2009 David S. Miller <davem@davemloft.net>
4 *
5 * The NMI watchdog support and infrastructure is based almost
6 * entirely upon the x86 NMI support code.
7 */
8#include <linux/kernel.h>
9#include <linux/param.h>
10#include <linux/init.h>
11#include <linux/percpu.h>
12#include <linux/nmi.h>
13#include <linux/export.h>
14#include <linux/kprobes.h>
15#include <linux/kernel_stat.h>
16#include <linux/reboot.h>
17#include <linux/slab.h>
18#include <linux/kdebug.h>
19#include <linux/delay.h>
20#include <linux/smp.h>
21
22#include <asm/perf_event.h>
23#include <asm/ptrace.h>
24#include <asm/pcr.h>
25
26#include "kstack.h"
27
28/* We don't have a real NMI on sparc64, but we can fake one
29 * up using profiling counter overflow interrupts and interrupt
30 * levels.
31 *
32 * The profile overflow interrupts at level 15, so we use
33 * level 14 as our IRQ off level.
34 */
35
36static int panic_on_timeout;
37
38/* nmi_active:
39 * >0: the NMI watchdog is active, but can be disabled
40 * <0: the NMI watchdog has not been set up, and cannot be enabled
41 * 0: the NMI watchdog is disabled, but can be enabled
42 */
43atomic_t nmi_active = ATOMIC_INIT(0); /* oprofile uses this */
44EXPORT_SYMBOL(nmi_active);
45
46static unsigned int nmi_hz = HZ;
47static DEFINE_PER_CPU(short, wd_enabled);
48static int endflag __initdata;
49
50static DEFINE_PER_CPU(unsigned int, last_irq_sum);
51static DEFINE_PER_CPU(long, alert_counter);
52static DEFINE_PER_CPU(int, nmi_touch);
53
54void touch_nmi_watchdog(void)
55{
56 if (atomic_read(&nmi_active)) {
57 int cpu;
58
59 for_each_present_cpu(cpu) {
60 if (per_cpu(nmi_touch, cpu) != 1)
61 per_cpu(nmi_touch, cpu) = 1;
62 }
63 }
64
65 touch_softlockup_watchdog();
66}
67EXPORT_SYMBOL(touch_nmi_watchdog);
68
69static void die_nmi(const char *str, struct pt_regs *regs, int do_panic)
70{
71 int this_cpu = smp_processor_id();
72
73 if (notify_die(DIE_NMIWATCHDOG, str, regs, 0,
74 pt_regs_trap_type(regs), SIGINT) == NOTIFY_STOP)
75 return;
76
77 if (do_panic || panic_on_oops)
78 panic("Watchdog detected hard LOCKUP on cpu %d", this_cpu);
79 else
80 WARN(1, "Watchdog detected hard LOCKUP on cpu %d", this_cpu);
81}
82
83notrace __kprobes void perfctr_irq(int irq, struct pt_regs *regs)
84{
85 unsigned int sum, touched = 0;
86 void *orig_sp;
87
88 clear_softint(1 << irq);
89
90 local_cpu_data().__nmi_count++;
91
92 nmi_enter();
93
94 orig_sp = set_hardirq_stack();
95
96 if (notify_die(DIE_NMI, "nmi", regs, 0,
97 pt_regs_trap_type(regs), SIGINT) == NOTIFY_STOP)
98 touched = 1;
99 else
100 pcr_ops->write_pcr(0, pcr_ops->pcr_nmi_disable);
101
102 sum = local_cpu_data().irq0_irqs;
103 if (__get_cpu_var(nmi_touch)) {
104 __get_cpu_var(nmi_touch) = 0;
105 touched = 1;
106 }
107 if (!touched && __get_cpu_var(last_irq_sum) == sum) {
108 __this_cpu_inc(alert_counter);
109 if (__this_cpu_read(alert_counter) == 30 * nmi_hz)
110 die_nmi("BUG: NMI Watchdog detected LOCKUP",
111 regs, panic_on_timeout);
112 } else {
113 __get_cpu_var(last_irq_sum) = sum;
114 __this_cpu_write(alert_counter, 0);
115 }
116 if (__get_cpu_var(wd_enabled)) {
117 pcr_ops->write_pic(0, pcr_ops->nmi_picl_value(nmi_hz));
118 pcr_ops->write_pcr(0, pcr_ops->pcr_nmi_enable);
119 }
120
121 restore_hardirq_stack(orig_sp);
122
123 nmi_exit();
124}
125
126static inline unsigned int get_nmi_count(int cpu)
127{
128 return cpu_data(cpu).__nmi_count;
129}
130
131static __init void nmi_cpu_busy(void *data)
132{
133 local_irq_enable_in_hardirq();
134 while (endflag == 0)
135 mb();
136}
137
138static void report_broken_nmi(int cpu, int *prev_nmi_count)
139{
140 printk(KERN_CONT "\n");
141
142 printk(KERN_WARNING
143 "WARNING: CPU#%d: NMI appears to be stuck (%d->%d)!\n",
144 cpu, prev_nmi_count[cpu], get_nmi_count(cpu));
145
146 printk(KERN_WARNING
147 "Please report this to bugzilla.kernel.org,\n");
148 printk(KERN_WARNING
149 "and attach the output of the 'dmesg' command.\n");
150
151 per_cpu(wd_enabled, cpu) = 0;
152 atomic_dec(&nmi_active);
153}
154
155void stop_nmi_watchdog(void *unused)
156{
157 pcr_ops->write_pcr(0, pcr_ops->pcr_nmi_disable);
158 __get_cpu_var(wd_enabled) = 0;
159 atomic_dec(&nmi_active);
160}
161
162static int __init check_nmi_watchdog(void)
163{
164 unsigned int *prev_nmi_count;
165 int cpu, err;
166
167 if (!atomic_read(&nmi_active))
168 return 0;
169
170 prev_nmi_count = kmalloc(nr_cpu_ids * sizeof(unsigned int), GFP_KERNEL);
171 if (!prev_nmi_count) {
172 err = -ENOMEM;
173 goto error;
174 }
175
176 printk(KERN_INFO "Testing NMI watchdog ... ");
177
178 smp_call_function(nmi_cpu_busy, (void *)&endflag, 0);
179
180 for_each_possible_cpu(cpu)
181 prev_nmi_count[cpu] = get_nmi_count(cpu);
182 local_irq_enable();
183 mdelay((20 * 1000) / nmi_hz); /* wait 20 ticks */
184
185 for_each_online_cpu(cpu) {
186 if (!per_cpu(wd_enabled, cpu))
187 continue;
188 if (get_nmi_count(cpu) - prev_nmi_count[cpu] <= 5)
189 report_broken_nmi(cpu, prev_nmi_count);
190 }
191 endflag = 1;
192 if (!atomic_read(&nmi_active)) {
193 kfree(prev_nmi_count);
194 atomic_set(&nmi_active, -1);
195 err = -ENODEV;
196 goto error;
197 }
198 printk("OK.\n");
199
200 nmi_hz = 1;
201
202 kfree(prev_nmi_count);
203 return 0;
204error:
205 on_each_cpu(stop_nmi_watchdog, NULL, 1);
206 return err;
207}
208
209void start_nmi_watchdog(void *unused)
210{
211 __get_cpu_var(wd_enabled) = 1;
212 atomic_inc(&nmi_active);
213
214 pcr_ops->write_pcr(0, pcr_ops->pcr_nmi_disable);
215 pcr_ops->write_pic(0, pcr_ops->nmi_picl_value(nmi_hz));
216
217 pcr_ops->write_pcr(0, pcr_ops->pcr_nmi_enable);
218}
219
220static void nmi_adjust_hz_one(void *unused)
221{
222 if (!__get_cpu_var(wd_enabled))
223 return;
224
225 pcr_ops->write_pcr(0, pcr_ops->pcr_nmi_disable);
226 pcr_ops->write_pic(0, pcr_ops->nmi_picl_value(nmi_hz));
227
228 pcr_ops->write_pcr(0, pcr_ops->pcr_nmi_enable);
229}
230
231void nmi_adjust_hz(unsigned int new_hz)
232{
233 nmi_hz = new_hz;
234 on_each_cpu(nmi_adjust_hz_one, NULL, 1);
235}
236EXPORT_SYMBOL_GPL(nmi_adjust_hz);
237
238static int nmi_shutdown(struct notifier_block *nb, unsigned long cmd, void *p)
239{
240 on_each_cpu(stop_nmi_watchdog, NULL, 1);
241 return 0;
242}
243
244static struct notifier_block nmi_reboot_notifier = {
245 .notifier_call = nmi_shutdown,
246};
247
248int __init nmi_init(void)
249{
250 int err;
251
252 on_each_cpu(start_nmi_watchdog, NULL, 1);
253
254 err = check_nmi_watchdog();
255 if (!err) {
256 err = register_reboot_notifier(&nmi_reboot_notifier);
257 if (err) {
258 on_each_cpu(stop_nmi_watchdog, NULL, 1);
259 atomic_set(&nmi_active, -1);
260 }
261 }
262
263 return err;
264}
265
266static int __init setup_nmi_watchdog(char *str)
267{
268 if (!strncmp(str, "panic", 5))
269 panic_on_timeout = 1;
270
271 return 0;
272}
273__setup("nmi_watchdog=", setup_nmi_watchdog);
1/* Pseudo NMI support on sparc64 systems.
2 *
3 * Copyright (C) 2009 David S. Miller <davem@davemloft.net>
4 *
5 * The NMI watchdog support and infrastructure is based almost
6 * entirely upon the x86 NMI support code.
7 */
8#include <linux/kernel.h>
9#include <linux/param.h>
10#include <linux/init.h>
11#include <linux/percpu.h>
12#include <linux/nmi.h>
13#include <linux/export.h>
14#include <linux/kprobes.h>
15#include <linux/kernel_stat.h>
16#include <linux/reboot.h>
17#include <linux/slab.h>
18#include <linux/kdebug.h>
19#include <linux/delay.h>
20#include <linux/smp.h>
21
22#include <asm/perf_event.h>
23#include <asm/ptrace.h>
24#include <asm/pcr.h>
25
26#include "kstack.h"
27
28/* We don't have a real NMI on sparc64, but we can fake one
29 * up using profiling counter overflow interrupts and interrupt
30 * levels.
31 *
32 * The profile overflow interrupts at level 15, so we use
33 * level 14 as our IRQ off level.
34 */
35
36static int panic_on_timeout;
37
38/* nmi_active:
39 * >0: the NMI watchdog is active, but can be disabled
40 * <0: the NMI watchdog has not been set up, and cannot be enabled
41 * 0: the NMI watchdog is disabled, but can be enabled
42 */
43atomic_t nmi_active = ATOMIC_INIT(0); /* oprofile uses this */
44EXPORT_SYMBOL(nmi_active);
45
46static unsigned int nmi_hz = HZ;
47static DEFINE_PER_CPU(short, wd_enabled);
48static int endflag __initdata;
49
50static DEFINE_PER_CPU(unsigned int, last_irq_sum);
51static DEFINE_PER_CPU(long, alert_counter);
52static DEFINE_PER_CPU(int, nmi_touch);
53
54void touch_nmi_watchdog(void)
55{
56 if (atomic_read(&nmi_active)) {
57 int cpu;
58
59 for_each_present_cpu(cpu) {
60 if (per_cpu(nmi_touch, cpu) != 1)
61 per_cpu(nmi_touch, cpu) = 1;
62 }
63 }
64
65 touch_softlockup_watchdog();
66}
67EXPORT_SYMBOL(touch_nmi_watchdog);
68
69static void die_nmi(const char *str, struct pt_regs *regs, int do_panic)
70{
71 int this_cpu = smp_processor_id();
72
73 if (notify_die(DIE_NMIWATCHDOG, str, regs, 0,
74 pt_regs_trap_type(regs), SIGINT) == NOTIFY_STOP)
75 return;
76
77 if (do_panic || panic_on_oops)
78 panic("Watchdog detected hard LOCKUP on cpu %d", this_cpu);
79 else
80 WARN(1, "Watchdog detected hard LOCKUP on cpu %d", this_cpu);
81}
82
83notrace __kprobes void perfctr_irq(int irq, struct pt_regs *regs)
84{
85 unsigned int sum, touched = 0;
86 void *orig_sp;
87
88 clear_softint(1 << irq);
89
90 local_cpu_data().__nmi_count++;
91
92 nmi_enter();
93
94 orig_sp = set_hardirq_stack();
95
96 if (notify_die(DIE_NMI, "nmi", regs, 0,
97 pt_regs_trap_type(regs), SIGINT) == NOTIFY_STOP)
98 touched = 1;
99 else
100 pcr_ops->write_pcr(0, pcr_ops->pcr_nmi_disable);
101
102 sum = local_cpu_data().irq0_irqs;
103 if (__this_cpu_read(nmi_touch)) {
104 __this_cpu_write(nmi_touch, 0);
105 touched = 1;
106 }
107 if (!touched && __this_cpu_read(last_irq_sum) == sum) {
108 __this_cpu_inc(alert_counter);
109 if (__this_cpu_read(alert_counter) == 30 * nmi_hz)
110 die_nmi("BUG: NMI Watchdog detected LOCKUP",
111 regs, panic_on_timeout);
112 } else {
113 __this_cpu_write(last_irq_sum, sum);
114 __this_cpu_write(alert_counter, 0);
115 }
116 if (__this_cpu_read(wd_enabled)) {
117 pcr_ops->write_pic(0, pcr_ops->nmi_picl_value(nmi_hz));
118 pcr_ops->write_pcr(0, pcr_ops->pcr_nmi_enable);
119 }
120
121 restore_hardirq_stack(orig_sp);
122
123 nmi_exit();
124}
125
126static inline unsigned int get_nmi_count(int cpu)
127{
128 return cpu_data(cpu).__nmi_count;
129}
130
131static __init void nmi_cpu_busy(void *data)
132{
133 while (endflag == 0)
134 mb();
135}
136
137static void report_broken_nmi(int cpu, int *prev_nmi_count)
138{
139 printk(KERN_CONT "\n");
140
141 printk(KERN_WARNING
142 "WARNING: CPU#%d: NMI appears to be stuck (%d->%d)!\n",
143 cpu, prev_nmi_count[cpu], get_nmi_count(cpu));
144
145 printk(KERN_WARNING
146 "Please report this to bugzilla.kernel.org,\n");
147 printk(KERN_WARNING
148 "and attach the output of the 'dmesg' command.\n");
149
150 per_cpu(wd_enabled, cpu) = 0;
151 atomic_dec(&nmi_active);
152}
153
154void stop_nmi_watchdog(void *unused)
155{
156 pcr_ops->write_pcr(0, pcr_ops->pcr_nmi_disable);
157 __this_cpu_write(wd_enabled, 0);
158 atomic_dec(&nmi_active);
159}
160
161static int __init check_nmi_watchdog(void)
162{
163 unsigned int *prev_nmi_count;
164 int cpu, err;
165
166 if (!atomic_read(&nmi_active))
167 return 0;
168
169 prev_nmi_count = kmalloc(nr_cpu_ids * sizeof(unsigned int), GFP_KERNEL);
170 if (!prev_nmi_count) {
171 err = -ENOMEM;
172 goto error;
173 }
174
175 printk(KERN_INFO "Testing NMI watchdog ... ");
176
177 smp_call_function(nmi_cpu_busy, (void *)&endflag, 0);
178
179 for_each_possible_cpu(cpu)
180 prev_nmi_count[cpu] = get_nmi_count(cpu);
181 local_irq_enable();
182 mdelay((20 * 1000) / nmi_hz); /* wait 20 ticks */
183
184 for_each_online_cpu(cpu) {
185 if (!per_cpu(wd_enabled, cpu))
186 continue;
187 if (get_nmi_count(cpu) - prev_nmi_count[cpu] <= 5)
188 report_broken_nmi(cpu, prev_nmi_count);
189 }
190 endflag = 1;
191 if (!atomic_read(&nmi_active)) {
192 kfree(prev_nmi_count);
193 atomic_set(&nmi_active, -1);
194 err = -ENODEV;
195 goto error;
196 }
197 printk("OK.\n");
198
199 nmi_hz = 1;
200
201 kfree(prev_nmi_count);
202 return 0;
203error:
204 on_each_cpu(stop_nmi_watchdog, NULL, 1);
205 return err;
206}
207
208void start_nmi_watchdog(void *unused)
209{
210 __this_cpu_write(wd_enabled, 1);
211 atomic_inc(&nmi_active);
212
213 pcr_ops->write_pcr(0, pcr_ops->pcr_nmi_disable);
214 pcr_ops->write_pic(0, pcr_ops->nmi_picl_value(nmi_hz));
215
216 pcr_ops->write_pcr(0, pcr_ops->pcr_nmi_enable);
217}
218
219static void nmi_adjust_hz_one(void *unused)
220{
221 if (!__this_cpu_read(wd_enabled))
222 return;
223
224 pcr_ops->write_pcr(0, pcr_ops->pcr_nmi_disable);
225 pcr_ops->write_pic(0, pcr_ops->nmi_picl_value(nmi_hz));
226
227 pcr_ops->write_pcr(0, pcr_ops->pcr_nmi_enable);
228}
229
230void nmi_adjust_hz(unsigned int new_hz)
231{
232 nmi_hz = new_hz;
233 on_each_cpu(nmi_adjust_hz_one, NULL, 1);
234}
235EXPORT_SYMBOL_GPL(nmi_adjust_hz);
236
237static int nmi_shutdown(struct notifier_block *nb, unsigned long cmd, void *p)
238{
239 on_each_cpu(stop_nmi_watchdog, NULL, 1);
240 return 0;
241}
242
243static struct notifier_block nmi_reboot_notifier = {
244 .notifier_call = nmi_shutdown,
245};
246
247int __init nmi_init(void)
248{
249 int err;
250
251 on_each_cpu(start_nmi_watchdog, NULL, 1);
252
253 err = check_nmi_watchdog();
254 if (!err) {
255 err = register_reboot_notifier(&nmi_reboot_notifier);
256 if (err) {
257 on_each_cpu(stop_nmi_watchdog, NULL, 1);
258 atomic_set(&nmi_active, -1);
259 }
260 }
261
262 return err;
263}
264
265static int __init setup_nmi_watchdog(char *str)
266{
267 if (!strncmp(str, "panic", 5))
268 panic_on_timeout = 1;
269
270 return 0;
271}
272__setup("nmi_watchdog=", setup_nmi_watchdog);