sched: yield fix
[linux-block.git] / kernel / sched_debug.c
CommitLineData
43ae34cb
IM
1/*
2 * kernel/time/sched_debug.c
3 *
4 * Print the CFS rbtree
5 *
6 * Copyright(C) 2007, Red Hat, Inc., Ingo Molnar
7 *
8 * This program is free software; you can redistribute it and/or modify
9 * it under the terms of the GNU General Public License version 2 as
10 * published by the Free Software Foundation.
11 */
12
13#include <linux/proc_fs.h>
14#include <linux/sched.h>
15#include <linux/seq_file.h>
16#include <linux/kallsyms.h>
17#include <linux/utsname.h>
18
19/*
20 * This allows printing both to /proc/sched_debug and
21 * to the console
22 */
23#define SEQ_printf(m, x...) \
24 do { \
25 if (m) \
26 seq_printf(m, x); \
27 else \
28 printk(x); \
29 } while (0)
30
ef83a571
IM
31/*
32 * Ease the printing of nsec fields:
33 */
34static long long nsec_high(long long nsec)
35{
36 if (nsec < 0) {
37 nsec = -nsec;
38 do_div(nsec, 1000000);
39 return -nsec;
40 }
41 do_div(nsec, 1000000);
42
43 return nsec;
44}
45
46static unsigned long nsec_low(long long nsec)
47{
48 if (nsec < 0)
49 nsec = -nsec;
50
51 return do_div(nsec, 1000000);
52}
53
54#define SPLIT_NS(x) nsec_high(x), nsec_low(x)
55
43ae34cb 56static void
a48da48b 57print_task(struct seq_file *m, struct rq *rq, struct task_struct *p)
43ae34cb
IM
58{
59 if (rq->curr == p)
60 SEQ_printf(m, "R");
61 else
62 SEQ_printf(m, " ");
63
ef83a571 64 SEQ_printf(m, "%15s %5d %9Ld.%06ld %9Ld %5d ",
43ae34cb 65 p->comm, p->pid,
ef83a571 66 SPLIT_NS(p->se.vruntime),
43ae34cb 67 (long long)(p->nvcsw + p->nivcsw),
6f605d83 68 p->prio);
6cfb0d5d 69#ifdef CONFIG_SCHEDSTATS
c86da3a3 70 SEQ_printf(m, "%9Ld.%06ld %9Ld.%06ld %9Ld.%06ld\n",
ef83a571
IM
71 SPLIT_NS(p->se.vruntime),
72 SPLIT_NS(p->se.sum_exec_runtime),
73 SPLIT_NS(p->se.sum_sleep_runtime));
6cfb0d5d 74#else
ef83a571
IM
75 SEQ_printf(m, "%15Ld %15Ld %15Ld.%06ld %15Ld.%06ld %15Ld.%06ld\n",
76 0LL, 0LL, 0LL, 0L, 0LL, 0L, 0LL, 0L);
6cfb0d5d 77#endif
43ae34cb
IM
78}
79
a48da48b 80static void print_rq(struct seq_file *m, struct rq *rq, int rq_cpu)
43ae34cb
IM
81{
82 struct task_struct *g, *p;
83
84 SEQ_printf(m,
85 "\nrunnable tasks:\n"
c86da3a3
MG
86 " task PID tree-key switches prio"
87 " exec-runtime sum-exec sum-sleep\n"
1a75b94f 88 "------------------------------------------------------"
c86da3a3 89 "----------------------------------------------------\n");
43ae34cb
IM
90
91 read_lock_irq(&tasklist_lock);
92
93 do_each_thread(g, p) {
94 if (!p->se.on_rq || task_cpu(p) != rq_cpu)
95 continue;
96
a48da48b 97 print_task(m, rq, p);
43ae34cb
IM
98 } while_each_thread(g, p);
99
100 read_unlock_irq(&tasklist_lock);
101}
102
5cef9eca 103void print_cfs_rq(struct seq_file *m, int cpu, struct cfs_rq *cfs_rq)
43ae34cb 104{
86d9560c
IM
105 s64 MIN_vruntime = -1, min_vruntime, max_vruntime = -1,
106 spread, rq0_min_vruntime, spread0;
67e12eac
IM
107 struct rq *rq = &per_cpu(runqueues, cpu);
108 struct sched_entity *last;
109 unsigned long flags;
110
5167e75f 111 SEQ_printf(m, "\ncfs_rq\n");
43ae34cb 112
ef83a571
IM
113 SEQ_printf(m, " .%-30s: %Ld.%06ld\n", "exec_clock",
114 SPLIT_NS(cfs_rq->exec_clock));
67e12eac
IM
115
116 spin_lock_irqsave(&rq->lock, flags);
117 if (cfs_rq->rb_leftmost)
118 MIN_vruntime = (__pick_next_entity(cfs_rq))->vruntime;
119 last = __pick_last_entity(cfs_rq);
120 if (last)
121 max_vruntime = last->vruntime;
86d9560c
IM
122 min_vruntime = rq->cfs.min_vruntime;
123 rq0_min_vruntime = per_cpu(runqueues, 0).cfs.min_vruntime;
67e12eac 124 spin_unlock_irqrestore(&rq->lock, flags);
ef83a571
IM
125 SEQ_printf(m, " .%-30s: %Ld.%06ld\n", "MIN_vruntime",
126 SPLIT_NS(MIN_vruntime));
127 SEQ_printf(m, " .%-30s: %Ld.%06ld\n", "min_vruntime",
128 SPLIT_NS(min_vruntime));
129 SEQ_printf(m, " .%-30s: %Ld.%06ld\n", "max_vruntime",
130 SPLIT_NS(max_vruntime));
67e12eac 131 spread = max_vruntime - MIN_vruntime;
ef83a571
IM
132 SEQ_printf(m, " .%-30s: %Ld.%06ld\n", "spread",
133 SPLIT_NS(spread));
86d9560c 134 spread0 = min_vruntime - rq0_min_vruntime;
ef83a571
IM
135 SEQ_printf(m, " .%-30s: %Ld.%06ld\n", "spread0",
136 SPLIT_NS(spread0));
545f3b18
SV
137 SEQ_printf(m, " .%-30s: %ld\n", "nr_running", cfs_rq->nr_running);
138 SEQ_printf(m, " .%-30s: %ld\n", "load", cfs_rq->load.weight);
fdd71d13 139#ifdef CONFIG_SCHEDSTATS
b8efb561
IM
140 SEQ_printf(m, " .%-30s: %ld\n", "bkl_cnt",
141 rq->bkl_cnt);
fdd71d13 142#endif
ddc97297
PZ
143 SEQ_printf(m, " .%-30s: %ld\n", "nr_spread_over",
144 cfs_rq->nr_spread_over);
43ae34cb
IM
145}
146
a48da48b 147static void print_cpu(struct seq_file *m, int cpu)
43ae34cb
IM
148{
149 struct rq *rq = &per_cpu(runqueues, cpu);
150
151#ifdef CONFIG_X86
152 {
153 unsigned int freq = cpu_khz ? : 1;
154
155 SEQ_printf(m, "\ncpu#%d, %u.%03u MHz\n",
156 cpu, freq / 1000, (freq % 1000));
157 }
158#else
159 SEQ_printf(m, "\ncpu#%d\n", cpu);
160#endif
161
162#define P(x) \
163 SEQ_printf(m, " .%-30s: %Ld\n", #x, (long long)(rq->x))
ef83a571
IM
164#define PN(x) \
165 SEQ_printf(m, " .%-30s: %Ld.%06ld\n", #x, SPLIT_NS(rq->x))
43ae34cb
IM
166
167 P(nr_running);
168 SEQ_printf(m, " .%-30s: %lu\n", "load",
495eca49 169 rq->load.weight);
43ae34cb
IM
170 P(nr_switches);
171 P(nr_load_updates);
172 P(nr_uninterruptible);
173 SEQ_printf(m, " .%-30s: %lu\n", "jiffies", jiffies);
ef83a571 174 PN(next_balance);
43ae34cb 175 P(curr->pid);
ef83a571
IM
176 PN(clock);
177 PN(idle_clock);
178 PN(prev_clock_raw);
43ae34cb
IM
179 P(clock_warps);
180 P(clock_overflows);
2aa44d05 181 P(clock_deep_idle_events);
ef83a571 182 PN(clock_max_delta);
43ae34cb
IM
183 P(cpu_load[0]);
184 P(cpu_load[1]);
185 P(cpu_load[2]);
186 P(cpu_load[3]);
187 P(cpu_load[4]);
188#undef P
ef83a571 189#undef PN
43ae34cb 190
5cef9eca 191 print_cfs_stats(m, cpu);
43ae34cb 192
a48da48b 193 print_rq(m, rq, cpu);
43ae34cb
IM
194}
195
196static int sched_debug_show(struct seq_file *m, void *v)
197{
198 u64 now = ktime_to_ns(ktime_get());
199 int cpu;
200
6cfb0d5d 201 SEQ_printf(m, "Sched Debug Version: v0.05-v20, %s %.*s\n",
43ae34cb
IM
202 init_utsname()->release,
203 (int)strcspn(init_utsname()->version, " "),
204 init_utsname()->version);
205
ef83a571 206 SEQ_printf(m, "now at %Lu.%06ld msecs\n", SPLIT_NS(now));
43ae34cb 207
1aa4731e 208#define P(x) \
d822cece 209 SEQ_printf(m, " .%-40s: %Ld\n", #x, (long long)(x))
1aa4731e 210#define PN(x) \
d822cece 211 SEQ_printf(m, " .%-40s: %Ld.%06ld\n", #x, SPLIT_NS(x))
1aa4731e
IM
212 PN(sysctl_sched_latency);
213 PN(sysctl_sched_min_granularity);
214 PN(sysctl_sched_wakeup_granularity);
215 PN(sysctl_sched_batch_wakeup_granularity);
216 PN(sysctl_sched_child_runs_first);
217 P(sysctl_sched_features);
218#undef PN
219#undef P
220
43ae34cb 221 for_each_online_cpu(cpu)
a48da48b 222 print_cpu(m, cpu);
43ae34cb
IM
223
224 SEQ_printf(m, "\n");
225
226 return 0;
227}
228
f3373461 229static void sysrq_sched_debug_show(void)
43ae34cb
IM
230{
231 sched_debug_show(NULL, NULL);
232}
233
24e377a8
SV
234#ifdef CONFIG_FAIR_USER_SCHED
235
236static DEFINE_MUTEX(root_user_share_mutex);
237
238static int
239root_user_share_read_proc(char *page, char **start, off_t off, int count,
240 int *eof, void *data)
241{
242 int len;
243
244 len = sprintf(page, "%d\n", init_task_grp_load);
245
246 return len;
247}
248
249static int
250root_user_share_write_proc(struct file *file, const char __user *buffer,
251 unsigned long count, void *data)
252{
253 unsigned long shares;
254 char kbuf[sizeof(unsigned long)+1];
255 int rc = 0;
256
257 if (copy_from_user(kbuf, buffer, sizeof(kbuf)))
258 return -EFAULT;
259
260 shares = simple_strtoul(kbuf, NULL, 0);
261
262 if (!shares)
263 shares = NICE_0_LOAD;
264
265 mutex_lock(&root_user_share_mutex);
266
267 init_task_grp_load = shares;
268 rc = sched_group_set_shares(&init_task_grp, shares);
269
270 mutex_unlock(&root_user_share_mutex);
271
272 return (rc < 0 ? rc : count);
273}
274
275#endif /* CONFIG_FAIR_USER_SCHED */
276
43ae34cb
IM
277static int sched_debug_open(struct inode *inode, struct file *filp)
278{
279 return single_open(filp, sched_debug_show, NULL);
280}
281
282static struct file_operations sched_debug_fops = {
283 .open = sched_debug_open,
284 .read = seq_read,
285 .llseek = seq_lseek,
5ea473a1 286 .release = single_release,
43ae34cb
IM
287};
288
289static int __init init_sched_debug_procfs(void)
290{
291 struct proc_dir_entry *pe;
292
293 pe = create_proc_entry("sched_debug", 0644, NULL);
294 if (!pe)
295 return -ENOMEM;
296
297 pe->proc_fops = &sched_debug_fops;
298
24e377a8
SV
299#ifdef CONFIG_FAIR_USER_SCHED
300 pe = create_proc_entry("root_user_share", 0644, NULL);
301 if (!pe)
302 return -ENOMEM;
303
304 pe->read_proc = root_user_share_read_proc;
305 pe->write_proc = root_user_share_write_proc;
306#endif
307
43ae34cb
IM
308 return 0;
309}
310
311__initcall(init_sched_debug_procfs);
312
313void proc_sched_show_task(struct task_struct *p, struct seq_file *m)
314{
315 unsigned long flags;
316 int num_threads = 1;
317
318 rcu_read_lock();
319 if (lock_task_sighand(p, &flags)) {
320 num_threads = atomic_read(&p->signal->count);
321 unlock_task_sighand(p, &flags);
322 }
323 rcu_read_unlock();
324
325 SEQ_printf(m, "%s (%d, #threads: %d)\n", p->comm, p->pid, num_threads);
326 SEQ_printf(m, "----------------------------------------------\n");
327#define P(F) \
328 SEQ_printf(m, "%-25s:%20Ld\n", #F, (long long)p->F)
ef83a571
IM
329#define PN(F) \
330 SEQ_printf(m, "%-25s:%14Ld.%06ld\n", #F, SPLIT_NS((long long)p->F))
43ae34cb 331
ef83a571
IM
332 PN(se.exec_start);
333 PN(se.vruntime);
334 PN(se.sum_exec_runtime);
6cfb0d5d
IM
335
336#ifdef CONFIG_SCHEDSTATS
ef83a571
IM
337 PN(se.wait_start);
338 PN(se.sleep_start);
339 PN(se.block_start);
340 PN(se.sleep_max);
341 PN(se.block_max);
342 PN(se.exec_max);
343 PN(se.slice_max);
344 PN(se.wait_max);
b8efb561 345 P(sched_info.bkl_cnt);
6cfb0d5d 346#endif
43ae34cb
IM
347 SEQ_printf(m, "%-25s:%20Ld\n",
348 "nr_switches", (long long)(p->nvcsw + p->nivcsw));
349 P(se.load.weight);
350 P(policy);
351 P(prio);
352#undef P
ef83a571 353#undef PN
43ae34cb
IM
354
355 {
356 u64 t0, t1;
357
358 t0 = sched_clock();
359 t1 = sched_clock();
360 SEQ_printf(m, "%-25s:%20Ld\n",
361 "clock-delta", (long long)(t1-t0));
362 }
363}
364
365void proc_sched_set_task(struct task_struct *p)
366{
6cfb0d5d 367#ifdef CONFIG_SCHEDSTATS
a4b29ba2
IM
368 p->se.sleep_max = 0;
369 p->se.block_max = 0;
370 p->se.exec_max = 0;
eba1ed4b 371 p->se.slice_max = 0;
a4b29ba2 372 p->se.wait_max = 0;
b8efb561 373 p->sched_info.bkl_cnt = 0;
6cfb0d5d 374#endif
a4b29ba2 375 p->se.sum_exec_runtime = 0;
2491b2b8 376 p->se.prev_sum_exec_runtime = 0;
43ae34cb 377}