taskstats.c 14.0 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
/*
 * taskstats.c - Export per-task statistics to userland
 *
 * Copyright (C) Shailabh Nagar, IBM Corp. 2006
 *           (C) Balbir Singh,   IBM Corp. 2006
 *
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License as published by
 * the Free Software Foundation; either version 2 of the License, or
 * (at your option) any later version.
 *
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU General Public License for more details.
 *
 */

#include <linux/kernel.h>
#include <linux/taskstats_kern.h>
21
#include <linux/tsacct_kern.h>
22
#include <linux/delayacct.h>
23 24
#include <linux/cpumask.h>
#include <linux/percpu.h>
25
#include <linux/slab.h>
B
Balbir Singh 已提交
26 27 28 29
#include <linux/cgroupstats.h>
#include <linux/cgroup.h>
#include <linux/fs.h>
#include <linux/file.h>
30 31 32
#include <net/genetlink.h>
#include <asm/atomic.h>

33 34 35 36 37 38
/*
 * Maximum length of a cpumask that can be specified in
 * the TASKSTATS_CMD_ATTR_REGISTER/DEREGISTER_CPUMASK attribute
 */
#define TASKSTATS_CPUMASK_MAXLEN	(100+6*NR_CPUS)

39
static DEFINE_PER_CPU(__u32, taskstats_seqnum);
40
static int family_registered;
41
struct kmem_cache *taskstats_cache;
42 43 44 45 46 47 48 49

static struct genl_family family = {
	.id		= GENL_ID_GENERATE,
	.name		= TASKSTATS_GENL_NAME,
	.version	= TASKSTATS_GENL_VERSION,
	.maxattr	= TASKSTATS_CMD_ATTR_MAX,
};

A
Alexey Dobriyan 已提交
50
static const struct nla_policy taskstats_cmd_get_policy[TASKSTATS_CMD_ATTR_MAX+1] = {
51 52
	[TASKSTATS_CMD_ATTR_PID]  = { .type = NLA_U32 },
	[TASKSTATS_CMD_ATTR_TGID] = { .type = NLA_U32 },
53 54 55
	[TASKSTATS_CMD_ATTR_REGISTER_CPUMASK] = { .type = NLA_STRING },
	[TASKSTATS_CMD_ATTR_DEREGISTER_CPUMASK] = { .type = NLA_STRING },};

A
Alexey Dobriyan 已提交
56
static const struct nla_policy cgroupstats_cmd_get_policy[CGROUPSTATS_CMD_ATTR_MAX+1] = {
B
Balbir Singh 已提交
57 58 59
	[CGROUPSTATS_CMD_ATTR_FD] = { .type = NLA_U32 },
};

60 61 62
struct listener {
	struct list_head list;
	pid_t pid;
63
	char valid;
64 65
};

66 67 68 69 70 71 72 73 74 75 76
struct listener_list {
	struct rw_semaphore sem;
	struct list_head list;
};
static DEFINE_PER_CPU(struct listener_list, listener_array);

enum actions {
	REGISTER,
	DEREGISTER,
	CPU_DONT_CARE
};
77 78

static int prepare_reply(struct genl_info *info, u8 cmd, struct sk_buff **skbp,
79
				size_t size)
80 81 82 83 84 85 86
{
	struct sk_buff *skb;
	void *reply;

	/*
	 * If new attributes are added, please revisit this allocation
	 */
87
	skb = genlmsg_new(size, GFP_KERNEL);
88 89 90 91 92 93 94
	if (!skb)
		return -ENOMEM;

	if (!info) {
		int seq = get_cpu_var(taskstats_seqnum)++;
		put_cpu_var(taskstats_seqnum);

95
		reply = genlmsg_put(skb, 0, seq, &family, 0, cmd);
96
	} else
97
		reply = genlmsg_put_reply(skb, info, &family, 0, cmd);
98 99 100 101 102 103 104 105 106
	if (reply == NULL) {
		nlmsg_free(skb);
		return -EINVAL;
	}

	*skbp = skb;
	return 0;
}

107 108 109
/*
 * Send taskstats data in @skb to listener with nl_pid @pid
 */
J
Johannes Berg 已提交
110
static int send_reply(struct sk_buff *skb, struct genl_info *info)
111
{
112
	struct genlmsghdr *genlhdr = nlmsg_data(nlmsg_hdr(skb));
113
	void *reply = genlmsg_data(genlhdr);
114 115 116 117 118 119 120 121
	int rc;

	rc = genlmsg_end(skb, reply);
	if (rc < 0) {
		nlmsg_free(skb);
		return rc;
	}

J
Johannes Berg 已提交
122
	return genlmsg_reply(skb, info);
123 124
}

125 126 127
/*
 * Send taskstats data in @skb to listeners registered for @cpu's exit data
 */
128 129
static void send_cpu_listeners(struct sk_buff *skb,
					struct listener_list *listeners)
130
{
131
	struct genlmsghdr *genlhdr = nlmsg_data(nlmsg_hdr(skb));
132 133 134
	struct listener *s, *tmp;
	struct sk_buff *skb_next, *skb_cur = skb;
	void *reply = genlmsg_data(genlhdr);
135
	int rc, delcount = 0;
136 137 138 139

	rc = genlmsg_end(skb, reply);
	if (rc < 0) {
		nlmsg_free(skb);
140
		return;
141 142 143
	}

	rc = 0;
144
	down_read(&listeners->sem);
145
	list_for_each_entry(s, &listeners->list, list) {
146 147 148
		skb_next = NULL;
		if (!list_is_last(&s->list, &listeners->list)) {
			skb_next = skb_clone(skb_cur, GFP_KERNEL);
149
			if (!skb_next)
150 151
				break;
		}
J
Johannes Berg 已提交
152
		rc = genlmsg_unicast(&init_net, skb_cur, s->pid);
153
		if (rc == -ECONNREFUSED) {
154 155
			s->valid = 0;
			delcount++;
156 157 158
		}
		skb_cur = skb_next;
	}
159
	up_read(&listeners->sem);
160

161 162 163
	if (skb_cur)
		nlmsg_free(skb_cur);

164
	if (!delcount)
165
		return;
166 167 168 169 170 171 172 173 174 175

	/* Delete invalidated entries */
	down_write(&listeners->sem);
	list_for_each_entry_safe(s, tmp, &listeners->list, list) {
		if (!s->valid) {
			list_del(&s->list);
			kfree(s);
		}
	}
	up_write(&listeners->sem);
176 177
}

178
static int fill_pid(pid_t pid, struct task_struct *tsk,
179 180
		struct taskstats *stats)
{
181
	int rc = 0;
182

183 184
	if (!tsk) {
		rcu_read_lock();
185
		tsk = find_task_by_vpid(pid);
186 187 188 189
		if (tsk)
			get_task_struct(tsk);
		rcu_read_unlock();
		if (!tsk)
190 191 192 193
			return -ESRCH;
	} else
		get_task_struct(tsk);

194
	memset(stats, 0, sizeof(*stats));
195 196 197 198
	/*
	 * Each accounting subsystem adds calls to its functions to
	 * fill in relevant parts of struct taskstsats as follows
	 *
199
	 *	per-task-foo(stats, tsk);
200 201
	 */

202
	delayacct_add_tsk(stats, tsk);
203 204

	/* fill in basic acct fields */
205
	stats->version = TASKSTATS_VERSION;
206 207
	stats->nvcsw = tsk->nvcsw;
	stats->nivcsw = tsk->nivcsw;
208
	bacct_add_tsk(stats, tsk);
209

210 211 212
	/* fill in extended acct fields */
	xacct_add_tsk(stats, tsk);

213
	/* Define err: label here if needed */
214 215 216 217 218
	put_task_struct(tsk);
	return rc;

}

219
static int fill_tgid(pid_t tgid, struct task_struct *first,
220 221
		struct taskstats *stats)
{
222
	struct task_struct *tsk;
223
	unsigned long flags;
224
	int rc = -ESRCH;
225

226 227 228 229
	/*
	 * Add additional stats from live tasks except zombie thread group
	 * leaders who are already counted with the dead tasks
	 */
230 231
	rcu_read_lock();
	if (!first)
232
		first = find_task_by_vpid(tgid);
233

234 235
	if (!first || !lock_task_sighand(first, &flags))
		goto out;
236

237 238
	if (first->signal->stats)
		memcpy(stats, first->signal->stats, sizeof(*stats));
239 240
	else
		memset(stats, 0, sizeof(*stats));
241

242
	tsk = first;
243
	do {
244
		if (tsk->exit_state)
245
			continue;
246
		/*
247
		 * Accounting subsystem can call its functions here to
248 249
		 * fill in relevant parts of struct taskstsats as follows
		 *
250
		 *	per-task-foo(stats, tsk);
251
		 */
252
		delayacct_add_tsk(stats, tsk);
253

254 255
		stats->nvcsw += tsk->nvcsw;
		stats->nivcsw += tsk->nivcsw;
256
	} while_each_thread(first, tsk);
257

258 259 260 261 262 263
	unlock_task_sighand(first, &flags);
	rc = 0;
out:
	rcu_read_unlock();

	stats->version = TASKSTATS_VERSION;
264
	/*
265
	 * Accounting subsystems can also add calls here to modify
266
	 * fields of taskstats.
267
	 */
268
	return rc;
269 270 271 272 273 274 275
}


static void fill_tgid_exit(struct task_struct *tsk)
{
	unsigned long flags;

276
	spin_lock_irqsave(&tsk->sighand->siglock, flags);
277 278 279 280 281 282 283 284 285 286 287
	if (!tsk->signal->stats)
		goto ret;

	/*
	 * Each accounting subsystem calls its functions here to
	 * accumalate its per-task stats for tsk, into the per-tgid structure
	 *
	 *	per-task-foo(tsk->signal->stats, tsk);
	 */
	delayacct_add_tsk(tsk->signal->stats, tsk);
ret:
288
	spin_unlock_irqrestore(&tsk->sighand->siglock, flags);
289
	return;
290 291
}

292
static int add_del_listener(pid_t pid, const struct cpumask *mask, int isadd)
293 294 295 296
{
	struct listener_list *listeners;
	struct listener *s, *tmp;
	unsigned int cpu;
297

298
	if (!cpumask_subset(mask, cpu_possible_mask))
299 300 301
		return -EINVAL;

	if (isadd == REGISTER) {
302
		for_each_cpu(cpu, mask) {
303 304 305 306 307 308
			s = kmalloc_node(sizeof(struct listener), GFP_KERNEL,
					 cpu_to_node(cpu));
			if (!s)
				goto cleanup;
			s->pid = pid;
			INIT_LIST_HEAD(&s->list);
309
			s->valid = 1;
310 311 312 313 314 315 316 317 318 319 320

			listeners = &per_cpu(listener_array, cpu);
			down_write(&listeners->sem);
			list_add(&s->list, &listeners->list);
			up_write(&listeners->sem);
		}
		return 0;
	}

	/* Deregister or cleanup */
cleanup:
321
	for_each_cpu(cpu, mask) {
322 323 324 325 326 327 328 329 330 331 332 333 334 335
		listeners = &per_cpu(listener_array, cpu);
		down_write(&listeners->sem);
		list_for_each_entry_safe(s, tmp, &listeners->list, list) {
			if (s->pid == pid) {
				list_del(&s->list);
				kfree(s);
				break;
			}
		}
		up_write(&listeners->sem);
	}
	return 0;
}

336
static int parse(struct nlattr *na, struct cpumask *mask)
337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352
{
	char *data;
	int len;
	int ret;

	if (na == NULL)
		return 1;
	len = nla_len(na);
	if (len > TASKSTATS_CPUMASK_MAXLEN)
		return -E2BIG;
	if (len < 1)
		return -EINVAL;
	data = kmalloc(len, GFP_KERNEL);
	if (!data)
		return -ENOMEM;
	nla_strlcpy(data, na, len);
353
	ret = cpulist_parse(data, mask);
354 355 356 357
	kfree(data);
	return ret;
}

358
static struct taskstats *mk_reply(struct sk_buff *skb, int type, u32 pid)
359
{
360
	struct nlattr *na, *ret;
361 362
	int aggr;

363 364 365 366 367 368
	/* If we don't pad, we end up with alignment on a 4 byte boundary.
	 * This causes lots of runtime warnings on systems requiring 8 byte
	 * alignment */
	u32 pids[2] = { pid, 0 };
	int pid_size = ALIGN(sizeof(pid), sizeof(long));

369 370 371
	aggr = (type == TASKSTATS_TYPE_PID)
			? TASKSTATS_TYPE_AGGR_PID
			: TASKSTATS_TYPE_AGGR_TGID;
372 373

	na = nla_nest_start(skb, aggr);
374 375
	if (!na)
		goto err;
376
	if (nla_put(skb, type, pid_size, pids) < 0)
377 378 379 380
		goto err;
	ret = nla_reserve(skb, TASKSTATS_TYPE_STATS, sizeof(struct taskstats));
	if (!ret)
		goto err;
381 382
	nla_nest_end(skb, na);

383 384 385
	return nla_data(ret);
err:
	return NULL;
386 387
}

B
Balbir Singh 已提交
388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404
static int cgroupstats_user_cmd(struct sk_buff *skb, struct genl_info *info)
{
	int rc = 0;
	struct sk_buff *rep_skb;
	struct cgroupstats *stats;
	struct nlattr *na;
	size_t size;
	u32 fd;
	struct file *file;
	int fput_needed;

	na = info->attrs[CGROUPSTATS_CMD_ATTR_FD];
	if (!na)
		return -EINVAL;

	fd = nla_get_u32(info->attrs[CGROUPSTATS_CMD_ATTR_FD]);
	file = fget_light(fd, &fput_needed);
405 406
	if (!file)
		return 0;
B
Balbir Singh 已提交
407

408
	size = nla_total_size(sizeof(struct cgroupstats));
B
Balbir Singh 已提交
409

410 411 412 413
	rc = prepare_reply(info, CGROUPSTATS_CMD_NEW, &rep_skb,
				size);
	if (rc < 0)
		goto err;
B
Balbir Singh 已提交
414

415 416 417 418
	na = nla_reserve(rep_skb, CGROUPSTATS_TYPE_CGROUP_STATS,
				sizeof(struct cgroupstats));
	stats = nla_data(na);
	memset(stats, 0, sizeof(*stats));
B
Balbir Singh 已提交
419

420 421 422 423
	rc = cgroupstats_build(stats, file->f_dentry);
	if (rc < 0) {
		nlmsg_free(rep_skb);
		goto err;
B
Balbir Singh 已提交
424 425
	}

J
Johannes Berg 已提交
426
	rc = send_reply(rep_skb, info);
427

B
Balbir Singh 已提交
428
err:
429
	fput_light(file, fput_needed);
B
Balbir Singh 已提交
430 431 432
	return rc;
}

433
static int taskstats_user_cmd(struct sk_buff *skb, struct genl_info *info)
434
{
435
	int rc;
436
	struct sk_buff *rep_skb;
437
	struct taskstats *stats;
438
	size_t size;
439 440 441 442
	cpumask_var_t mask;

	if (!alloc_cpumask_var(&mask, GFP_KERNEL))
		return -ENOMEM;
443

444
	rc = parse(info->attrs[TASKSTATS_CMD_ATTR_REGISTER_CPUMASK], mask);
445
	if (rc < 0)
446 447 448 449 450
		goto free_return_rc;
	if (rc == 0) {
		rc = add_del_listener(info->snd_pid, mask, REGISTER);
		goto free_return_rc;
	}
451

452
	rc = parse(info->attrs[TASKSTATS_CMD_ATTR_DEREGISTER_CPUMASK], mask);
453
	if (rc < 0)
454 455 456 457 458
		goto free_return_rc;
	if (rc == 0) {
		rc = add_del_listener(info->snd_pid, mask, DEREGISTER);
free_return_rc:
		free_cpumask_var(mask);
459
		return rc;
460 461
	}
	free_cpumask_var(mask);
462 463 464 465 466 467 468

	/*
	 * Size includes space for nested attributes
	 */
	size = nla_total_size(sizeof(u32)) +
		nla_total_size(sizeof(struct taskstats)) + nla_total_size(0);

469
	rc = prepare_reply(info, TASKSTATS_CMD_NEW, &rep_skb, size);
470 471 472
	if (rc < 0)
		return rc;

473
	rc = -EINVAL;
474 475
	if (info->attrs[TASKSTATS_CMD_ATTR_PID]) {
		u32 pid = nla_get_u32(info->attrs[TASKSTATS_CMD_ATTR_PID]);
476 477
		stats = mk_reply(rep_skb, TASKSTATS_TYPE_PID, pid);
		if (!stats)
478
			goto err;
479

480 481
		rc = fill_pid(pid, NULL, stats);
		if (rc < 0)
482
			goto err;
483 484
	} else if (info->attrs[TASKSTATS_CMD_ATTR_TGID]) {
		u32 tgid = nla_get_u32(info->attrs[TASKSTATS_CMD_ATTR_TGID]);
485 486
		stats = mk_reply(rep_skb, TASKSTATS_TYPE_TGID, tgid);
		if (!stats)
487
			goto err;
488

489 490
		rc = fill_tgid(tgid, NULL, stats);
		if (rc < 0)
491
			goto err;
492
	} else
493 494
		goto err;

J
Johannes Berg 已提交
495
	return send_reply(rep_skb, info);
496 497 498 499 500
err:
	nlmsg_free(rep_skb);
	return rc;
}

501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524
static struct taskstats *taskstats_tgid_alloc(struct task_struct *tsk)
{
	struct signal_struct *sig = tsk->signal;
	struct taskstats *stats;

	if (sig->stats || thread_group_empty(tsk))
		goto ret;

	/* No problem if kmem_cache_zalloc() fails */
	stats = kmem_cache_zalloc(taskstats_cache, GFP_KERNEL);

	spin_lock_irq(&tsk->sighand->siglock);
	if (!sig->stats) {
		sig->stats = stats;
		stats = NULL;
	}
	spin_unlock_irq(&tsk->sighand->siglock);

	if (stats)
		kmem_cache_free(taskstats_cache, stats);
ret:
	return sig->stats;
}

525
/* Send pid data out on exit */
526
void taskstats_exit(struct task_struct *tsk, int group_dead)
527 528
{
	int rc;
529
	struct listener_list *listeners;
530
	struct taskstats *stats;
531 532 533 534
	struct sk_buff *rep_skb;
	size_t size;
	int is_thread_group;

535
	if (!family_registered)
536 537 538 539 540 541 542 543
		return;

	/*
	 * Size includes space for nested attributes
	 */
	size = nla_total_size(sizeof(u32)) +
		nla_total_size(sizeof(struct taskstats)) + nla_total_size(0);

544
	is_thread_group = !!taskstats_tgid_alloc(tsk);
545 546 547 548 549 550 551
	if (is_thread_group) {
		/* PID + STATS + TGID + STATS */
		size = 2 * size;
		/* fill the tsk->signal->stats structure */
		fill_tgid_exit(tsk);
	}

552 553 554 555
	listeners = &__raw_get_cpu_var(listener_array);
	if (list_empty(&listeners->list))
		return;

556
	rc = prepare_reply(NULL, TASKSTATS_CMD_NEW, &rep_skb, size);
557
	if (rc < 0)
558
		return;
559

560 561
	stats = mk_reply(rep_skb, TASKSTATS_TYPE_PID, tsk->pid);
	if (!stats)
562
		goto err;
563

564
	rc = fill_pid(-1, tsk, stats);
565
	if (rc < 0)
566
		goto err;
567 568

	/*
569
	 * Doesn't matter if tsk is the leader or the last group member leaving
570
	 */
571
	if (!is_thread_group || !group_dead)
572
		goto send;
573

574 575
	stats = mk_reply(rep_skb, TASKSTATS_TYPE_TGID, tsk->tgid);
	if (!stats)
576
		goto err;
577 578

	memcpy(stats, tsk->signal->stats, sizeof(*stats));
579

580
send:
581
	send_cpu_listeners(rep_skb, listeners);
582
	return;
583
err:
584 585 586 587 588
	nlmsg_free(rep_skb);
}

static struct genl_ops taskstats_ops = {
	.cmd		= TASKSTATS_CMD_GET,
589
	.doit		= taskstats_user_cmd,
590 591 592
	.policy		= taskstats_cmd_get_policy,
};

B
Balbir Singh 已提交
593 594 595 596 597 598
static struct genl_ops cgroupstats_ops = {
	.cmd		= CGROUPSTATS_CMD_GET,
	.doit		= cgroupstats_user_cmd,
	.policy		= cgroupstats_cmd_get_policy,
};

599 600 601
/* Needed early in initialization */
void __init taskstats_init_early(void)
{
602 603
	unsigned int i;

604
	taskstats_cache = KMEM_CACHE(taskstats, SLAB_PANIC);
605 606 607 608
	for_each_possible_cpu(i) {
		INIT_LIST_HEAD(&(per_cpu(listener_array, i).list));
		init_rwsem(&(per_cpu(listener_array, i).sem));
	}
609 610 611 612 613 614 615 616 617 618 619 620 621 622
}

static int __init taskstats_init(void)
{
	int rc;

	rc = genl_register_family(&family);
	if (rc)
		return rc;

	rc = genl_register_ops(&family, &taskstats_ops);
	if (rc < 0)
		goto err;

B
Balbir Singh 已提交
623 624 625 626
	rc = genl_register_ops(&family, &cgroupstats_ops);
	if (rc < 0)
		goto err_cgroup_ops;

627
	family_registered = 1;
B
Balbir Singh 已提交
628
	printk("registered taskstats version %d\n", TASKSTATS_GENL_VERSION);
629
	return 0;
B
Balbir Singh 已提交
630 631
err_cgroup_ops:
	genl_unregister_ops(&family, &taskstats_ops);
632 633 634 635 636 637 638 639 640 641
err:
	genl_unregister_family(&family);
	return rc;
}

/*
 * late initcall ensures initialization of statistics collection
 * mechanisms precedes initialization of the taskstats interface
 */
late_initcall(taskstats_init);