xref: /openbmc/linux/fs/proc/array.c (revision 643d1f7f) 
1 /*
2  *  linux/fs/proc/array.c
3  *
4  *  Copyright (C) 1992  by Linus Torvalds
5  *  based on ideas by Darren Senn
6  *
7  * Fixes:
8  * Michael. K. Johnson: stat,statm extensions.
9  *                      <johnsonm@stolaf.edu>
10  *
11  * Pauline Middelink :  Made cmdline,envline only break at '\0's, to
12  *                      make sure SET_PROCTITLE works. Also removed
13  *                      bad '!' which forced address recalculation for
14  *                      EVERY character on the current page.
15  *                      <middelin@polyware.iaf.nl>
16  *
17  * Danny ter Haar    :	added cpuinfo
18  *			<dth@cistron.nl>
19  *
20  * Alessandro Rubini :  profile extension.
21  *                      <rubini@ipvvis.unipv.it>
22  *
23  * Jeff Tranter      :  added BogoMips field to cpuinfo
24  *                      <Jeff_Tranter@Mitel.COM>
25  *
26  * Bruno Haible      :  remove 4K limit for the maps file
27  *			<haible@ma2s2.mathematik.uni-karlsruhe.de>
28  *
29  * Yves Arrouye      :  remove removal of trailing spaces in get_array.
30  *			<Yves.Arrouye@marin.fdn.fr>
31  *
32  * Jerome Forissier  :  added per-CPU time information to /proc/stat
33  *                      and /proc/<pid>/cpu extension
34  *                      <forissier@isia.cma.fr>
35  *			- Incorporation and non-SMP safe operation
36  *			of forissier patch in 2.1.78 by
37  *			Hans Marcus <crowbar@concepts.nl>
38  *
39  * aeb@cwi.nl        :  /proc/partitions
40  *
41  *
42  * Alan Cox	     :  security fixes.
43  *			<Alan.Cox@linux.org>
44  *
45  * Al Viro           :  safe handling of mm_struct
46  *
47  * Gerhard Wichert   :  added BIGMEM support
48  * Siemens AG           <Gerhard.Wichert@pdb.siemens.de>
49  *
50  * Al Viro & Jeff Garzik :  moved most of the thing into base.c and
51  *			 :  proc_misc.c. The rest may eventually go into
52  *			 :  base.c too.
53  */
54 
55 #include <linux/types.h>
56 #include <linux/errno.h>
57 #include <linux/time.h>
58 #include <linux/kernel.h>
59 #include <linux/kernel_stat.h>
60 #include <linux/tty.h>
61 #include <linux/string.h>
62 #include <linux/mman.h>
63 #include <linux/proc_fs.h>
64 #include <linux/ioport.h>
65 #include <linux/uaccess.h>
66 #include <linux/io.h>
67 #include <linux/mm.h>
68 #include <linux/hugetlb.h>
69 #include <linux/pagemap.h>
70 #include <linux/swap.h>
71 #include <linux/slab.h>
72 #include <linux/smp.h>
73 #include <linux/signal.h>
74 #include <linux/highmem.h>
75 #include <linux/file.h>
76 #include <linux/times.h>
77 #include <linux/cpuset.h>
78 #include <linux/rcupdate.h>
79 #include <linux/delayacct.h>
80 #include <linux/pid_namespace.h>
81 
82 #include <asm/pgtable.h>
83 #include <asm/processor.h>
84 #include "internal.h"
85 
86 /* Gcc optimizes away "strlen(x)" for constant x */
87 #define ADDBUF(buffer, string) \
88 do { memcpy(buffer, string, strlen(string)); \
89      buffer += strlen(string); } while (0)
90 
91 static inline char *task_name(struct task_struct *p, char *buf)
92 {
93 	int i;
94 	char *name;
95 	char tcomm[sizeof(p->comm)];
96 
97 	get_task_comm(tcomm, p);
98 
99 	ADDBUF(buf, "Name:\t");
100 	name = tcomm;
101 	i = sizeof(tcomm);
102 	do {
103 		unsigned char c = *name;
104 		name++;
105 		i--;
106 		*buf = c;
107 		if (!c)
108 			break;
109 		if (c == '\\') {
110 			buf[1] = c;
111 			buf += 2;
112 			continue;
113 		}
114 		if (c == '\n') {
115 			buf[0] = '\\';
116 			buf[1] = 'n';
117 			buf += 2;
118 			continue;
119 		}
120 		buf++;
121 	} while (i);
122 	*buf = '\n';
123 	return buf+1;
124 }
125 
126 /*
127  * The task state array is a strange "bitmap" of
128  * reasons to sleep. Thus "running" is zero, and
129  * you can test for combinations of others with
130  * simple bit tests.
131  */
132 static const char *task_state_array[] = {
133 	"R (running)",		/*  0 */
134 	"S (sleeping)",		/*  1 */
135 	"D (disk sleep)",	/*  2 */
136 	"T (stopped)",		/*  4 */
137 	"T (tracing stop)",	/*  8 */
138 	"Z (zombie)",		/* 16 */
139 	"X (dead)"		/* 32 */
140 };
141 
142 static inline const char *get_task_state(struct task_struct *tsk)
143 {
144 	unsigned int state = (tsk->state & TASK_REPORT) | tsk->exit_state;
145 	const char **p = &task_state_array[0];
146 
147 	while (state) {
148 		p++;
149 		state >>= 1;
150 	}
151 	return *p;
152 }
153 
154 static inline char *task_state(struct task_struct *p, char *buffer)
155 {
156 	struct group_info *group_info;
157 	int g;
158 	struct fdtable *fdt = NULL;
159 	struct pid_namespace *ns;
160 	pid_t ppid, tpid;
161 
162 	ns = current->nsproxy->pid_ns;
163 	rcu_read_lock();
164 	ppid = pid_alive(p) ?
165 		task_tgid_nr_ns(rcu_dereference(p->real_parent), ns) : 0;
166 	tpid = pid_alive(p) && p->ptrace ?
167 		task_pid_nr_ns(rcu_dereference(p->parent), ns) : 0;
168 	buffer += sprintf(buffer,
169 		"State:\t%s\n"
170 		"Tgid:\t%d\n"
171 		"Pid:\t%d\n"
172 		"PPid:\t%d\n"
173 		"TracerPid:\t%d\n"
174 		"Uid:\t%d\t%d\t%d\t%d\n"
175 		"Gid:\t%d\t%d\t%d\t%d\n",
176 		get_task_state(p),
177 		task_tgid_nr_ns(p, ns),
178 		task_pid_nr_ns(p, ns),
179 		ppid, tpid,
180 		p->uid, p->euid, p->suid, p->fsuid,
181 		p->gid, p->egid, p->sgid, p->fsgid);
182 
183 	task_lock(p);
184 	if (p->files)
185 		fdt = files_fdtable(p->files);
186 	buffer += sprintf(buffer,
187 		"FDSize:\t%d\n"
188 		"Groups:\t",
189 		fdt ? fdt->max_fds : 0);
190 	rcu_read_unlock();
191 
192 	group_info = p->group_info;
193 	get_group_info(group_info);
194 	task_unlock(p);
195 
196 	for (g = 0; g < min(group_info->ngroups, NGROUPS_SMALL); g++)
197 		buffer += sprintf(buffer, "%d ", GROUP_AT(group_info, g));
198 	put_group_info(group_info);
199 
200 	buffer += sprintf(buffer, "\n");
201 	return buffer;
202 }
203 
204 static char *render_sigset_t(const char *header, sigset_t *set, char *buffer)
205 {
206 	int i, len;
207 
208 	len = strlen(header);
209 	memcpy(buffer, header, len);
210 	buffer += len;
211 
212 	i = _NSIG;
213 	do {
214 		int x = 0;
215 
216 		i -= 4;
217 		if (sigismember(set, i+1)) x |= 1;
218 		if (sigismember(set, i+2)) x |= 2;
219 		if (sigismember(set, i+3)) x |= 4;
220 		if (sigismember(set, i+4)) x |= 8;
221 		*buffer++ = (x < 10 ? '0' : 'a' - 10) + x;
222 	} while (i >= 4);
223 
224 	*buffer++ = '\n';
225 	*buffer = 0;
226 	return buffer;
227 }
228 
229 static void collect_sigign_sigcatch(struct task_struct *p, sigset_t *ign,
230 				    sigset_t *catch)
231 {
232 	struct k_sigaction *k;
233 	int i;
234 
235 	k = p->sighand->action;
236 	for (i = 1; i <= _NSIG; ++i, ++k) {
237 		if (k->sa.sa_handler == SIG_IGN)
238 			sigaddset(ign, i);
239 		else if (k->sa.sa_handler != SIG_DFL)
240 			sigaddset(catch, i);
241 	}
242 }
243 
244 static inline char *task_sig(struct task_struct *p, char *buffer)
245 {
246 	unsigned long flags;
247 	sigset_t pending, shpending, blocked, ignored, caught;
248 	int num_threads = 0;
249 	unsigned long qsize = 0;
250 	unsigned long qlim = 0;
251 
252 	sigemptyset(&pending);
253 	sigemptyset(&shpending);
254 	sigemptyset(&blocked);
255 	sigemptyset(&ignored);
256 	sigemptyset(&caught);
257 
258 	rcu_read_lock();
259 	if (lock_task_sighand(p, &flags)) {
260 		pending = p->pending.signal;
261 		shpending = p->signal->shared_pending.signal;
262 		blocked = p->blocked;
263 		collect_sigign_sigcatch(p, &ignored, &caught);
264 		num_threads = atomic_read(&p->signal->count);
265 		qsize = atomic_read(&p->user->sigpending);
266 		qlim = p->signal->rlim[RLIMIT_SIGPENDING].rlim_cur;
267 		unlock_task_sighand(p, &flags);
268 	}
269 	rcu_read_unlock();
270 
271 	buffer += sprintf(buffer, "Threads:\t%d\n", num_threads);
272 	buffer += sprintf(buffer, "SigQ:\t%lu/%lu\n", qsize, qlim);
273 
274 	/* render them all */
275 	buffer = render_sigset_t("SigPnd:\t", &pending, buffer);
276 	buffer = render_sigset_t("ShdPnd:\t", &shpending, buffer);
277 	buffer = render_sigset_t("SigBlk:\t", &blocked, buffer);
278 	buffer = render_sigset_t("SigIgn:\t", &ignored, buffer);
279 	buffer = render_sigset_t("SigCgt:\t", &caught, buffer);
280 
281 	return buffer;
282 }
283 
284 static inline char *task_cap(struct task_struct *p, char *buffer)
285 {
286     return buffer + sprintf(buffer, "CapInh:\t%016x\n"
287 			    "CapPrm:\t%016x\n"
288 			    "CapEff:\t%016x\n",
289 			    cap_t(p->cap_inheritable),
290 			    cap_t(p->cap_permitted),
291 			    cap_t(p->cap_effective));
292 }
293 
294 static inline char *task_context_switch_counts(struct task_struct *p,
295 						char *buffer)
296 {
297 	return buffer + sprintf(buffer, "voluntary_ctxt_switches:\t%lu\n"
298 			    "nonvoluntary_ctxt_switches:\t%lu\n",
299 			    p->nvcsw,
300 			    p->nivcsw);
301 }
302 
303 int proc_pid_status(struct task_struct *task, char *buffer)
304 {
305 	char *orig = buffer;
306 	struct mm_struct *mm = get_task_mm(task);
307 
308 	buffer = task_name(task, buffer);
309 	buffer = task_state(task, buffer);
310 
311 	if (mm) {
312 		buffer = task_mem(mm, buffer);
313 		mmput(mm);
314 	}
315 	buffer = task_sig(task, buffer);
316 	buffer = task_cap(task, buffer);
317 	buffer = cpuset_task_status_allowed(task, buffer);
318 #if defined(CONFIG_S390)
319 	buffer = task_show_regs(task, buffer);
320 #endif
321 	buffer = task_context_switch_counts(task, buffer);
322 	return buffer - orig;
323 }
324 
325 /*
326  * Use precise platform statistics if available:
327  */
328 #ifdef CONFIG_VIRT_CPU_ACCOUNTING
329 static cputime_t task_utime(struct task_struct *p)
330 {
331 	return p->utime;
332 }
333 
334 static cputime_t task_stime(struct task_struct *p)
335 {
336 	return p->stime;
337 }
338 #else
339 static cputime_t task_utime(struct task_struct *p)
340 {
341 	clock_t utime = cputime_to_clock_t(p->utime),
342 		total = utime + cputime_to_clock_t(p->stime);
343 	u64 temp;
344 
345 	/*
346 	 * Use CFS's precise accounting:
347 	 */
348 	temp = (u64)nsec_to_clock_t(p->se.sum_exec_runtime);
349 
350 	if (total) {
351 		temp *= utime;
352 		do_div(temp, total);
353 	}
354 	utime = (clock_t)temp;
355 
356 	p->prev_utime = max(p->prev_utime, clock_t_to_cputime(utime));
357 	return p->prev_utime;
358 }
359 
360 static cputime_t task_stime(struct task_struct *p)
361 {
362 	clock_t stime;
363 
364 	/*
365 	 * Use CFS's precise accounting. (we subtract utime from
366 	 * the total, to make sure the total observed by userspace
367 	 * grows monotonically - apps rely on that):
368 	 */
369 	stime = nsec_to_clock_t(p->se.sum_exec_runtime) -
370 			cputime_to_clock_t(task_utime(p));
371 
372 	if (stime >= 0)
373 		p->prev_stime = max(p->prev_stime, clock_t_to_cputime(stime));
374 
375 	return p->prev_stime;
376 }
377 #endif
378 
379 static cputime_t task_gtime(struct task_struct *p)
380 {
381 	return p->gtime;
382 }
383 
384 static int do_task_stat(struct task_struct *task, char *buffer, int whole)
385 {
386 	unsigned long vsize, eip, esp, wchan = ~0UL;
387 	long priority, nice;
388 	int tty_pgrp = -1, tty_nr = 0;
389 	sigset_t sigign, sigcatch;
390 	char state;
391 	int res;
392 	pid_t ppid = 0, pgid = -1, sid = -1;
393 	int num_threads = 0;
394 	struct mm_struct *mm;
395 	unsigned long long start_time;
396 	unsigned long cmin_flt = 0, cmaj_flt = 0;
397 	unsigned long  min_flt = 0,  maj_flt = 0;
398 	cputime_t cutime, cstime, utime, stime;
399 	cputime_t cgtime, gtime;
400 	unsigned long rsslim = 0;
401 	char tcomm[sizeof(task->comm)];
402 	unsigned long flags;
403 	struct pid_namespace *ns;
404 
405 	ns = current->nsproxy->pid_ns;
406 
407 	state = *get_task_state(task);
408 	vsize = eip = esp = 0;
409 	mm = get_task_mm(task);
410 	if (mm) {
411 		vsize = task_vsize(mm);
412 		eip = KSTK_EIP(task);
413 		esp = KSTK_ESP(task);
414 	}
415 
416 	get_task_comm(tcomm, task);
417 
418 	sigemptyset(&sigign);
419 	sigemptyset(&sigcatch);
420 	cutime = cstime = utime = stime = cputime_zero;
421 	cgtime = gtime = cputime_zero;
422 
423 	rcu_read_lock();
424 	if (lock_task_sighand(task, &flags)) {
425 		struct signal_struct *sig = task->signal;
426 
427 		if (sig->tty) {
428 			tty_pgrp = pid_nr_ns(sig->tty->pgrp, ns);
429 			tty_nr = new_encode_dev(tty_devnum(sig->tty));
430 		}
431 
432 		num_threads = atomic_read(&sig->count);
433 		collect_sigign_sigcatch(task, &sigign, &sigcatch);
434 
435 		cmin_flt = sig->cmin_flt;
436 		cmaj_flt = sig->cmaj_flt;
437 		cutime = sig->cutime;
438 		cstime = sig->cstime;
439 		cgtime = sig->cgtime;
440 		rsslim = sig->rlim[RLIMIT_RSS].rlim_cur;
441 
442 		/* add up live thread stats at the group level */
443 		if (whole) {
444 			struct task_struct *t = task;
445 			do {
446 				min_flt += t->min_flt;
447 				maj_flt += t->maj_flt;
448 				utime = cputime_add(utime, task_utime(t));
449 				stime = cputime_add(stime, task_stime(t));
450 				gtime = cputime_add(gtime, task_gtime(t));
451 				t = next_thread(t);
452 			} while (t != task);
453 
454 			min_flt += sig->min_flt;
455 			maj_flt += sig->maj_flt;
456 			utime = cputime_add(utime, sig->utime);
457 			stime = cputime_add(stime, sig->stime);
458 			gtime = cputime_add(gtime, sig->gtime);
459 		}
460 
461 		sid = task_session_nr_ns(task, ns);
462 		ppid = task_tgid_nr_ns(task->real_parent, ns);
463 		pgid = task_pgrp_nr_ns(task, ns);
464 
465 		unlock_task_sighand(task, &flags);
466 	}
467 	rcu_read_unlock();
468 
469 	if (!whole || num_threads < 2)
470 		wchan = get_wchan(task);
471 	if (!whole) {
472 		min_flt = task->min_flt;
473 		maj_flt = task->maj_flt;
474 		utime = task_utime(task);
475 		stime = task_stime(task);
476 		gtime = task_gtime(task);
477 	}
478 
479 	/* scale priority and nice values from timeslices to -20..20 */
480 	/* to make it look like a "normal" Unix priority/nice value  */
481 	priority = task_prio(task);
482 	nice = task_nice(task);
483 
484 	/* Temporary variable needed for gcc-2.96 */
485 	/* convert timespec -> nsec*/
486 	start_time =
487 		(unsigned long long)task->real_start_time.tv_sec * NSEC_PER_SEC
488 				+ task->real_start_time.tv_nsec;
489 	/* convert nsec -> ticks */
490 	start_time = nsec_to_clock_t(start_time);
491 
492 	res = sprintf(buffer, "%d (%s) %c %d %d %d %d %d %u %lu \
493 %lu %lu %lu %lu %lu %ld %ld %ld %ld %d 0 %llu %lu %ld %lu %lu %lu %lu %lu \
494 %lu %lu %lu %lu %lu %lu %lu %lu %d %d %u %u %llu %lu %ld\n",
495 		task_pid_nr_ns(task, ns),
496 		tcomm,
497 		state,
498 		ppid,
499 		pgid,
500 		sid,
501 		tty_nr,
502 		tty_pgrp,
503 		task->flags,
504 		min_flt,
505 		cmin_flt,
506 		maj_flt,
507 		cmaj_flt,
508 		cputime_to_clock_t(utime),
509 		cputime_to_clock_t(stime),
510 		cputime_to_clock_t(cutime),
511 		cputime_to_clock_t(cstime),
512 		priority,
513 		nice,
514 		num_threads,
515 		start_time,
516 		vsize,
517 		mm ? get_mm_rss(mm) : 0,
518 		rsslim,
519 		mm ? mm->start_code : 0,
520 		mm ? mm->end_code : 0,
521 		mm ? mm->start_stack : 0,
522 		esp,
523 		eip,
524 		/* The signal information here is obsolete.
525 		 * It must be decimal for Linux 2.0 compatibility.
526 		 * Use /proc/#/status for real-time signals.
527 		 */
528 		task->pending.signal.sig[0] & 0x7fffffffUL,
529 		task->blocked.sig[0] & 0x7fffffffUL,
530 		sigign      .sig[0] & 0x7fffffffUL,
531 		sigcatch    .sig[0] & 0x7fffffffUL,
532 		wchan,
533 		0UL,
534 		0UL,
535 		task->exit_signal,
536 		task_cpu(task),
537 		task->rt_priority,
538 		task->policy,
539 		(unsigned long long)delayacct_blkio_ticks(task),
540 		cputime_to_clock_t(gtime),
541 		cputime_to_clock_t(cgtime));
542 	if (mm)
543 		mmput(mm);
544 	return res;
545 }
546 
547 int proc_tid_stat(struct task_struct *task, char *buffer)
548 {
549 	return do_task_stat(task, buffer, 0);
550 }
551 
552 int proc_tgid_stat(struct task_struct *task, char *buffer)
553 {
554 	return do_task_stat(task, buffer, 1);
555 }
556 
557 int proc_pid_statm(struct task_struct *task, char *buffer)
558 {
559 	int size = 0, resident = 0, shared = 0, text = 0, lib = 0, data = 0;
560 	struct mm_struct *mm = get_task_mm(task);
561 
562 	if (mm) {
563 		size = task_statm(mm, &shared, &text, &data, &resident);
564 		mmput(mm);
565 	}
566 
567 	return sprintf(buffer, "%d %d %d %d %d %d %d\n",
568 		       size, resident, shared, text, lib, data, 0);
569 }
570