Loading...
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@lxorguk.ukuu.org.uk>
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/smp.h>
72#include <linux/signal.h>
73#include <linux/highmem.h>
74#include <linux/file.h>
75#include <linux/fdtable.h>
76#include <linux/times.h>
77#include <linux/cpuset.h>
78#include <linux/rcupdate.h>
79#include <linux/delayacct.h>
80#include <linux/seq_file.h>
81#include <linux/pid_namespace.h>
82#include <linux/ptrace.h>
83#include <linux/tracehook.h>
84#include <linux/user_namespace.h>
85
86#include <asm/pgtable.h>
87#include <asm/processor.h>
88#include "internal.h"
89
90static inline void task_name(struct seq_file *m, struct task_struct *p)
91{
92 int i;
93 char *buf, *end;
94 char *name;
95 char tcomm[sizeof(p->comm)];
96
97 get_task_comm(tcomm, p);
98
99 seq_puts(m, "Name:\t");
100 end = m->buf + m->size;
101 buf = m->buf + m->count;
102 name = tcomm;
103 i = sizeof(tcomm);
104 while (i && (buf < end)) {
105 unsigned char c = *name;
106 name++;
107 i--;
108 *buf = c;
109 if (!c)
110 break;
111 if (c == '\\') {
112 buf++;
113 if (buf < end)
114 *buf++ = c;
115 continue;
116 }
117 if (c == '\n') {
118 *buf++ = '\\';
119 if (buf < end)
120 *buf++ = 'n';
121 continue;
122 }
123 buf++;
124 }
125 m->count = buf - m->buf;
126 seq_putc(m, '\n');
127}
128
129/*
130 * The task state array is a strange "bitmap" of
131 * reasons to sleep. Thus "running" is zero, and
132 * you can test for combinations of others with
133 * simple bit tests.
134 */
135static const char * const task_state_array[] = {
136 "R (running)", /* 0 */
137 "S (sleeping)", /* 1 */
138 "D (disk sleep)", /* 2 */
139 "T (stopped)", /* 4 */
140 "t (tracing stop)", /* 8 */
141 "Z (zombie)", /* 16 */
142 "X (dead)", /* 32 */
143 "x (dead)", /* 64 */
144 "K (wakekill)", /* 128 */
145 "W (waking)", /* 256 */
146};
147
148static inline const char *get_task_state(struct task_struct *tsk)
149{
150 unsigned int state = (tsk->state & TASK_REPORT) | tsk->exit_state;
151 const char * const *p = &task_state_array[0];
152
153 BUILD_BUG_ON(1 + ilog2(TASK_STATE_MAX) != ARRAY_SIZE(task_state_array));
154
155 while (state) {
156 p++;
157 state >>= 1;
158 }
159 return *p;
160}
161
162static inline void task_state(struct seq_file *m, struct pid_namespace *ns,
163 struct pid *pid, struct task_struct *p)
164{
165 struct user_namespace *user_ns = current_user_ns();
166 struct group_info *group_info;
167 int g;
168 struct fdtable *fdt = NULL;
169 const struct cred *cred;
170 pid_t ppid, tpid;
171
172 rcu_read_lock();
173 ppid = pid_alive(p) ?
174 task_tgid_nr_ns(rcu_dereference(p->real_parent), ns) : 0;
175 tpid = 0;
176 if (pid_alive(p)) {
177 struct task_struct *tracer = ptrace_parent(p);
178 if (tracer)
179 tpid = task_pid_nr_ns(tracer, ns);
180 }
181 cred = get_task_cred(p);
182 seq_printf(m,
183 "State:\t%s\n"
184 "Tgid:\t%d\n"
185 "Pid:\t%d\n"
186 "PPid:\t%d\n"
187 "TracerPid:\t%d\n"
188 "Uid:\t%d\t%d\t%d\t%d\n"
189 "Gid:\t%d\t%d\t%d\t%d\n",
190 get_task_state(p),
191 task_tgid_nr_ns(p, ns),
192 pid_nr_ns(pid, ns),
193 ppid, tpid,
194 from_kuid_munged(user_ns, cred->uid),
195 from_kuid_munged(user_ns, cred->euid),
196 from_kuid_munged(user_ns, cred->suid),
197 from_kuid_munged(user_ns, cred->fsuid),
198 from_kgid_munged(user_ns, cred->gid),
199 from_kgid_munged(user_ns, cred->egid),
200 from_kgid_munged(user_ns, cred->sgid),
201 from_kgid_munged(user_ns, cred->fsgid));
202
203 task_lock(p);
204 if (p->files)
205 fdt = files_fdtable(p->files);
206 seq_printf(m,
207 "FDSize:\t%d\n"
208 "Groups:\t",
209 fdt ? fdt->max_fds : 0);
210 rcu_read_unlock();
211
212 group_info = cred->group_info;
213 task_unlock(p);
214
215 for (g = 0; g < min(group_info->ngroups, NGROUPS_SMALL); g++)
216 seq_printf(m, "%d ",
217 from_kgid_munged(user_ns, GROUP_AT(group_info, g)));
218 put_cred(cred);
219
220 seq_putc(m, '\n');
221}
222
223static void render_sigset_t(struct seq_file *m, const char *header,
224 sigset_t *set)
225{
226 int i;
227
228 seq_puts(m, header);
229
230 i = _NSIG;
231 do {
232 int x = 0;
233
234 i -= 4;
235 if (sigismember(set, i+1)) x |= 1;
236 if (sigismember(set, i+2)) x |= 2;
237 if (sigismember(set, i+3)) x |= 4;
238 if (sigismember(set, i+4)) x |= 8;
239 seq_printf(m, "%x", x);
240 } while (i >= 4);
241
242 seq_putc(m, '\n');
243}
244
245static void collect_sigign_sigcatch(struct task_struct *p, sigset_t *ign,
246 sigset_t *catch)
247{
248 struct k_sigaction *k;
249 int i;
250
251 k = p->sighand->action;
252 for (i = 1; i <= _NSIG; ++i, ++k) {
253 if (k->sa.sa_handler == SIG_IGN)
254 sigaddset(ign, i);
255 else if (k->sa.sa_handler != SIG_DFL)
256 sigaddset(catch, i);
257 }
258}
259
260static inline void task_sig(struct seq_file *m, struct task_struct *p)
261{
262 unsigned long flags;
263 sigset_t pending, shpending, blocked, ignored, caught;
264 int num_threads = 0;
265 unsigned long qsize = 0;
266 unsigned long qlim = 0;
267
268 sigemptyset(&pending);
269 sigemptyset(&shpending);
270 sigemptyset(&blocked);
271 sigemptyset(&ignored);
272 sigemptyset(&caught);
273
274 if (lock_task_sighand(p, &flags)) {
275 pending = p->pending.signal;
276 shpending = p->signal->shared_pending.signal;
277 blocked = p->blocked;
278 collect_sigign_sigcatch(p, &ignored, &caught);
279 num_threads = get_nr_threads(p);
280 rcu_read_lock(); /* FIXME: is this correct? */
281 qsize = atomic_read(&__task_cred(p)->user->sigpending);
282 rcu_read_unlock();
283 qlim = task_rlimit(p, RLIMIT_SIGPENDING);
284 unlock_task_sighand(p, &flags);
285 }
286
287 seq_printf(m, "Threads:\t%d\n", num_threads);
288 seq_printf(m, "SigQ:\t%lu/%lu\n", qsize, qlim);
289
290 /* render them all */
291 render_sigset_t(m, "SigPnd:\t", &pending);
292 render_sigset_t(m, "ShdPnd:\t", &shpending);
293 render_sigset_t(m, "SigBlk:\t", &blocked);
294 render_sigset_t(m, "SigIgn:\t", &ignored);
295 render_sigset_t(m, "SigCgt:\t", &caught);
296}
297
298static void render_cap_t(struct seq_file *m, const char *header,
299 kernel_cap_t *a)
300{
301 unsigned __capi;
302
303 seq_puts(m, header);
304 CAP_FOR_EACH_U32(__capi) {
305 seq_printf(m, "%08x",
306 a->cap[(_KERNEL_CAPABILITY_U32S-1) - __capi]);
307 }
308 seq_putc(m, '\n');
309}
310
311static inline void task_cap(struct seq_file *m, struct task_struct *p)
312{
313 const struct cred *cred;
314 kernel_cap_t cap_inheritable, cap_permitted, cap_effective, cap_bset;
315
316 rcu_read_lock();
317 cred = __task_cred(p);
318 cap_inheritable = cred->cap_inheritable;
319 cap_permitted = cred->cap_permitted;
320 cap_effective = cred->cap_effective;
321 cap_bset = cred->cap_bset;
322 rcu_read_unlock();
323
324 render_cap_t(m, "CapInh:\t", &cap_inheritable);
325 render_cap_t(m, "CapPrm:\t", &cap_permitted);
326 render_cap_t(m, "CapEff:\t", &cap_effective);
327 render_cap_t(m, "CapBnd:\t", &cap_bset);
328}
329
330static inline void task_context_switch_counts(struct seq_file *m,
331 struct task_struct *p)
332{
333 seq_printf(m, "voluntary_ctxt_switches:\t%lu\n"
334 "nonvoluntary_ctxt_switches:\t%lu\n",
335 p->nvcsw,
336 p->nivcsw);
337}
338
339static void task_cpus_allowed(struct seq_file *m, struct task_struct *task)
340{
341 seq_puts(m, "Cpus_allowed:\t");
342 seq_cpumask(m, &task->cpus_allowed);
343 seq_putc(m, '\n');
344 seq_puts(m, "Cpus_allowed_list:\t");
345 seq_cpumask_list(m, &task->cpus_allowed);
346 seq_putc(m, '\n');
347}
348
349int proc_pid_status(struct seq_file *m, struct pid_namespace *ns,
350 struct pid *pid, struct task_struct *task)
351{
352 struct mm_struct *mm = get_task_mm(task);
353
354 task_name(m, task);
355 task_state(m, ns, pid, task);
356
357 if (mm) {
358 task_mem(m, mm);
359 mmput(mm);
360 }
361 task_sig(m, task);
362 task_cap(m, task);
363 task_cpus_allowed(m, task);
364 cpuset_task_status_allowed(m, task);
365 task_context_switch_counts(m, task);
366 return 0;
367}
368
369static int do_task_stat(struct seq_file *m, struct pid_namespace *ns,
370 struct pid *pid, struct task_struct *task, int whole)
371{
372 unsigned long vsize, eip, esp, wchan = ~0UL;
373 int priority, nice;
374 int tty_pgrp = -1, tty_nr = 0;
375 sigset_t sigign, sigcatch;
376 char state;
377 pid_t ppid = 0, pgid = -1, sid = -1;
378 int num_threads = 0;
379 int permitted;
380 struct mm_struct *mm;
381 unsigned long long start_time;
382 unsigned long cmin_flt = 0, cmaj_flt = 0;
383 unsigned long min_flt = 0, maj_flt = 0;
384 cputime_t cutime, cstime, utime, stime;
385 cputime_t cgtime, gtime;
386 unsigned long rsslim = 0;
387 char tcomm[sizeof(task->comm)];
388 unsigned long flags;
389
390 state = *get_task_state(task);
391 vsize = eip = esp = 0;
392 permitted = ptrace_may_access(task, PTRACE_MODE_READ | PTRACE_MODE_NOAUDIT);
393 mm = get_task_mm(task);
394 if (mm) {
395 vsize = task_vsize(mm);
396 if (permitted) {
397 eip = KSTK_EIP(task);
398 esp = KSTK_ESP(task);
399 }
400 }
401
402 get_task_comm(tcomm, task);
403
404 sigemptyset(&sigign);
405 sigemptyset(&sigcatch);
406 cutime = cstime = utime = stime = 0;
407 cgtime = gtime = 0;
408
409 if (lock_task_sighand(task, &flags)) {
410 struct signal_struct *sig = task->signal;
411
412 if (sig->tty) {
413 struct pid *pgrp = tty_get_pgrp(sig->tty);
414 tty_pgrp = pid_nr_ns(pgrp, ns);
415 put_pid(pgrp);
416 tty_nr = new_encode_dev(tty_devnum(sig->tty));
417 }
418
419 num_threads = get_nr_threads(task);
420 collect_sigign_sigcatch(task, &sigign, &sigcatch);
421
422 cmin_flt = sig->cmin_flt;
423 cmaj_flt = sig->cmaj_flt;
424 cutime = sig->cutime;
425 cstime = sig->cstime;
426 cgtime = sig->cgtime;
427 rsslim = ACCESS_ONCE(sig->rlim[RLIMIT_RSS].rlim_cur);
428
429 /* add up live thread stats at the group level */
430 if (whole) {
431 struct task_struct *t = task;
432 do {
433 min_flt += t->min_flt;
434 maj_flt += t->maj_flt;
435 gtime += t->gtime;
436 t = next_thread(t);
437 } while (t != task);
438
439 min_flt += sig->min_flt;
440 maj_flt += sig->maj_flt;
441 thread_group_times(task, &utime, &stime);
442 gtime += sig->gtime;
443 }
444
445 sid = task_session_nr_ns(task, ns);
446 ppid = task_tgid_nr_ns(task->real_parent, ns);
447 pgid = task_pgrp_nr_ns(task, ns);
448
449 unlock_task_sighand(task, &flags);
450 }
451
452 if (permitted && (!whole || num_threads < 2))
453 wchan = get_wchan(task);
454 if (!whole) {
455 min_flt = task->min_flt;
456 maj_flt = task->maj_flt;
457 task_times(task, &utime, &stime);
458 gtime = task->gtime;
459 }
460
461 /* scale priority and nice values from timeslices to -20..20 */
462 /* to make it look like a "normal" Unix priority/nice value */
463 priority = task_prio(task);
464 nice = task_nice(task);
465
466 /* Temporary variable needed for gcc-2.96 */
467 /* convert timespec -> nsec*/
468 start_time =
469 (unsigned long long)task->real_start_time.tv_sec * NSEC_PER_SEC
470 + task->real_start_time.tv_nsec;
471 /* convert nsec -> ticks */
472 start_time = nsec_to_clock_t(start_time);
473
474 seq_printf(m, "%d (%s) %c", pid_nr_ns(pid, ns), tcomm, state);
475 seq_put_decimal_ll(m, ' ', ppid);
476 seq_put_decimal_ll(m, ' ', pgid);
477 seq_put_decimal_ll(m, ' ', sid);
478 seq_put_decimal_ll(m, ' ', tty_nr);
479 seq_put_decimal_ll(m, ' ', tty_pgrp);
480 seq_put_decimal_ull(m, ' ', task->flags);
481 seq_put_decimal_ull(m, ' ', min_flt);
482 seq_put_decimal_ull(m, ' ', cmin_flt);
483 seq_put_decimal_ull(m, ' ', maj_flt);
484 seq_put_decimal_ull(m, ' ', cmaj_flt);
485 seq_put_decimal_ull(m, ' ', cputime_to_clock_t(utime));
486 seq_put_decimal_ull(m, ' ', cputime_to_clock_t(stime));
487 seq_put_decimal_ll(m, ' ', cputime_to_clock_t(cutime));
488 seq_put_decimal_ll(m, ' ', cputime_to_clock_t(cstime));
489 seq_put_decimal_ll(m, ' ', priority);
490 seq_put_decimal_ll(m, ' ', nice);
491 seq_put_decimal_ll(m, ' ', num_threads);
492 seq_put_decimal_ull(m, ' ', 0);
493 seq_put_decimal_ull(m, ' ', start_time);
494 seq_put_decimal_ull(m, ' ', vsize);
495 seq_put_decimal_ull(m, ' ', mm ? get_mm_rss(mm) : 0);
496 seq_put_decimal_ull(m, ' ', rsslim);
497 seq_put_decimal_ull(m, ' ', mm ? (permitted ? mm->start_code : 1) : 0);
498 seq_put_decimal_ull(m, ' ', mm ? (permitted ? mm->end_code : 1) : 0);
499 seq_put_decimal_ull(m, ' ', (permitted && mm) ? mm->start_stack : 0);
500 seq_put_decimal_ull(m, ' ', esp);
501 seq_put_decimal_ull(m, ' ', eip);
502 /* The signal information here is obsolete.
503 * It must be decimal for Linux 2.0 compatibility.
504 * Use /proc/#/status for real-time signals.
505 */
506 seq_put_decimal_ull(m, ' ', task->pending.signal.sig[0] & 0x7fffffffUL);
507 seq_put_decimal_ull(m, ' ', task->blocked.sig[0] & 0x7fffffffUL);
508 seq_put_decimal_ull(m, ' ', sigign.sig[0] & 0x7fffffffUL);
509 seq_put_decimal_ull(m, ' ', sigcatch.sig[0] & 0x7fffffffUL);
510 seq_put_decimal_ull(m, ' ', wchan);
511 seq_put_decimal_ull(m, ' ', 0);
512 seq_put_decimal_ull(m, ' ', 0);
513 seq_put_decimal_ll(m, ' ', task->exit_signal);
514 seq_put_decimal_ll(m, ' ', task_cpu(task));
515 seq_put_decimal_ull(m, ' ', task->rt_priority);
516 seq_put_decimal_ull(m, ' ', task->policy);
517 seq_put_decimal_ull(m, ' ', delayacct_blkio_ticks(task));
518 seq_put_decimal_ull(m, ' ', cputime_to_clock_t(gtime));
519 seq_put_decimal_ll(m, ' ', cputime_to_clock_t(cgtime));
520
521 if (mm && permitted) {
522 seq_put_decimal_ull(m, ' ', mm->start_data);
523 seq_put_decimal_ull(m, ' ', mm->end_data);
524 seq_put_decimal_ull(m, ' ', mm->start_brk);
525 seq_put_decimal_ull(m, ' ', mm->arg_start);
526 seq_put_decimal_ull(m, ' ', mm->arg_end);
527 seq_put_decimal_ull(m, ' ', mm->env_start);
528 seq_put_decimal_ull(m, ' ', mm->env_end);
529 } else
530 seq_printf(m, " 0 0 0 0 0 0 0");
531
532 if (permitted)
533 seq_put_decimal_ll(m, ' ', task->exit_code);
534 else
535 seq_put_decimal_ll(m, ' ', 0);
536
537 seq_putc(m, '\n');
538 if (mm)
539 mmput(mm);
540 return 0;
541}
542
543int proc_tid_stat(struct seq_file *m, struct pid_namespace *ns,
544 struct pid *pid, struct task_struct *task)
545{
546 return do_task_stat(m, ns, pid, task, 0);
547}
548
549int proc_tgid_stat(struct seq_file *m, struct pid_namespace *ns,
550 struct pid *pid, struct task_struct *task)
551{
552 return do_task_stat(m, ns, pid, task, 1);
553}
554
555int proc_pid_statm(struct seq_file *m, struct pid_namespace *ns,
556 struct pid *pid, struct task_struct *task)
557{
558 unsigned long size = 0, resident = 0, shared = 0, text = 0, data = 0;
559 struct mm_struct *mm = get_task_mm(task);
560
561 if (mm) {
562 size = task_statm(mm, &shared, &text, &data, &resident);
563 mmput(mm);
564 }
565 /*
566 * For quick read, open code by putting numbers directly
567 * expected format is
568 * seq_printf(m, "%lu %lu %lu %lu 0 %lu 0\n",
569 * size, resident, shared, text, data);
570 */
571 seq_put_decimal_ull(m, 0, size);
572 seq_put_decimal_ull(m, ' ', resident);
573 seq_put_decimal_ull(m, ' ', shared);
574 seq_put_decimal_ull(m, ' ', text);
575 seq_put_decimal_ull(m, ' ', 0);
576 seq_put_decimal_ull(m, ' ', data);
577 seq_put_decimal_ull(m, ' ', 0);
578 seq_putc(m, '\n');
579
580 return 0;
581}
582
583#ifdef CONFIG_CHECKPOINT_RESTORE
584static struct pid *
585get_children_pid(struct inode *inode, struct pid *pid_prev, loff_t pos)
586{
587 struct task_struct *start, *task;
588 struct pid *pid = NULL;
589
590 read_lock(&tasklist_lock);
591
592 start = pid_task(proc_pid(inode), PIDTYPE_PID);
593 if (!start)
594 goto out;
595
596 /*
597 * Lets try to continue searching first, this gives
598 * us significant speedup on children-rich processes.
599 */
600 if (pid_prev) {
601 task = pid_task(pid_prev, PIDTYPE_PID);
602 if (task && task->real_parent == start &&
603 !(list_empty(&task->sibling))) {
604 if (list_is_last(&task->sibling, &start->children))
605 goto out;
606 task = list_first_entry(&task->sibling,
607 struct task_struct, sibling);
608 pid = get_pid(task_pid(task));
609 goto out;
610 }
611 }
612
613 /*
614 * Slow search case.
615 *
616 * We might miss some children here if children
617 * are exited while we were not holding the lock,
618 * but it was never promised to be accurate that
619 * much.
620 *
621 * "Just suppose that the parent sleeps, but N children
622 * exit after we printed their tids. Now the slow paths
623 * skips N extra children, we miss N tasks." (c)
624 *
625 * So one need to stop or freeze the leader and all
626 * its children to get a precise result.
627 */
628 list_for_each_entry(task, &start->children, sibling) {
629 if (pos-- == 0) {
630 pid = get_pid(task_pid(task));
631 break;
632 }
633 }
634
635out:
636 read_unlock(&tasklist_lock);
637 return pid;
638}
639
640static int children_seq_show(struct seq_file *seq, void *v)
641{
642 struct inode *inode = seq->private;
643 pid_t pid;
644
645 pid = pid_nr_ns(v, inode->i_sb->s_fs_info);
646 return seq_printf(seq, "%d ", pid);
647}
648
649static void *children_seq_start(struct seq_file *seq, loff_t *pos)
650{
651 return get_children_pid(seq->private, NULL, *pos);
652}
653
654static void *children_seq_next(struct seq_file *seq, void *v, loff_t *pos)
655{
656 struct pid *pid;
657
658 pid = get_children_pid(seq->private, v, *pos + 1);
659 put_pid(v);
660
661 ++*pos;
662 return pid;
663}
664
665static void children_seq_stop(struct seq_file *seq, void *v)
666{
667 put_pid(v);
668}
669
670static const struct seq_operations children_seq_ops = {
671 .start = children_seq_start,
672 .next = children_seq_next,
673 .stop = children_seq_stop,
674 .show = children_seq_show,
675};
676
677static int children_seq_open(struct inode *inode, struct file *file)
678{
679 struct seq_file *m;
680 int ret;
681
682 ret = seq_open(file, &children_seq_ops);
683 if (ret)
684 return ret;
685
686 m = file->private_data;
687 m->private = inode;
688
689 return ret;
690}
691
692int children_seq_release(struct inode *inode, struct file *file)
693{
694 seq_release(inode, file);
695 return 0;
696}
697
698const struct file_operations proc_tid_children_operations = {
699 .open = children_seq_open,
700 .read = seq_read,
701 .llseek = seq_lseek,
702 .release = children_seq_release,
703};
704#endif /* CONFIG_CHECKPOINT_RESTORE */
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@lxorguk.ukuu.org.uk>
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/smp.h>
72#include <linux/signal.h>
73#include <linux/highmem.h>
74#include <linux/file.h>
75#include <linux/fdtable.h>
76#include <linux/times.h>
77#include <linux/cpuset.h>
78#include <linux/rcupdate.h>
79#include <linux/delayacct.h>
80#include <linux/seq_file.h>
81#include <linux/pid_namespace.h>
82#include <linux/ptrace.h>
83#include <linux/tracehook.h>
84
85#include <asm/pgtable.h>
86#include <asm/processor.h>
87#include "internal.h"
88
89static inline void task_name(struct seq_file *m, struct task_struct *p)
90{
91 int i;
92 char *buf, *end;
93 char *name;
94 char tcomm[sizeof(p->comm)];
95
96 get_task_comm(tcomm, p);
97
98 seq_puts(m, "Name:\t");
99 end = m->buf + m->size;
100 buf = m->buf + m->count;
101 name = tcomm;
102 i = sizeof(tcomm);
103 while (i && (buf < end)) {
104 unsigned char c = *name;
105 name++;
106 i--;
107 *buf = c;
108 if (!c)
109 break;
110 if (c == '\\') {
111 buf++;
112 if (buf < end)
113 *buf++ = c;
114 continue;
115 }
116 if (c == '\n') {
117 *buf++ = '\\';
118 if (buf < end)
119 *buf++ = 'n';
120 continue;
121 }
122 buf++;
123 }
124 m->count = buf - m->buf;
125 seq_putc(m, '\n');
126}
127
128/*
129 * The task state array is a strange "bitmap" of
130 * reasons to sleep. Thus "running" is zero, and
131 * you can test for combinations of others with
132 * simple bit tests.
133 */
134static const char * const task_state_array[] = {
135 "R (running)", /* 0 */
136 "S (sleeping)", /* 1 */
137 "D (disk sleep)", /* 2 */
138 "T (stopped)", /* 4 */
139 "t (tracing stop)", /* 8 */
140 "Z (zombie)", /* 16 */
141 "X (dead)", /* 32 */
142 "x (dead)", /* 64 */
143 "K (wakekill)", /* 128 */
144 "W (waking)", /* 256 */
145};
146
147static inline const char *get_task_state(struct task_struct *tsk)
148{
149 unsigned int state = (tsk->state & TASK_REPORT) | tsk->exit_state;
150 const char * const *p = &task_state_array[0];
151
152 BUILD_BUG_ON(1 + ilog2(TASK_STATE_MAX) != ARRAY_SIZE(task_state_array));
153
154 while (state) {
155 p++;
156 state >>= 1;
157 }
158 return *p;
159}
160
161static inline void task_state(struct seq_file *m, struct pid_namespace *ns,
162 struct pid *pid, struct task_struct *p)
163{
164 struct group_info *group_info;
165 int g;
166 struct fdtable *fdt = NULL;
167 const struct cred *cred;
168 pid_t ppid, tpid;
169
170 rcu_read_lock();
171 ppid = pid_alive(p) ?
172 task_tgid_nr_ns(rcu_dereference(p->real_parent), ns) : 0;
173 tpid = 0;
174 if (pid_alive(p)) {
175 struct task_struct *tracer = ptrace_parent(p);
176 if (tracer)
177 tpid = task_pid_nr_ns(tracer, ns);
178 }
179 cred = get_task_cred(p);
180 seq_printf(m,
181 "State:\t%s\n"
182 "Tgid:\t%d\n"
183 "Pid:\t%d\n"
184 "PPid:\t%d\n"
185 "TracerPid:\t%d\n"
186 "Uid:\t%d\t%d\t%d\t%d\n"
187 "Gid:\t%d\t%d\t%d\t%d\n",
188 get_task_state(p),
189 task_tgid_nr_ns(p, ns),
190 pid_nr_ns(pid, ns),
191 ppid, tpid,
192 cred->uid, cred->euid, cred->suid, cred->fsuid,
193 cred->gid, cred->egid, cred->sgid, cred->fsgid);
194
195 task_lock(p);
196 if (p->files)
197 fdt = files_fdtable(p->files);
198 seq_printf(m,
199 "FDSize:\t%d\n"
200 "Groups:\t",
201 fdt ? fdt->max_fds : 0);
202 rcu_read_unlock();
203
204 group_info = cred->group_info;
205 task_unlock(p);
206
207 for (g = 0; g < min(group_info->ngroups, NGROUPS_SMALL); g++)
208 seq_printf(m, "%d ", GROUP_AT(group_info, g));
209 put_cred(cred);
210
211 seq_putc(m, '\n');
212}
213
214static void render_sigset_t(struct seq_file *m, const char *header,
215 sigset_t *set)
216{
217 int i;
218
219 seq_puts(m, header);
220
221 i = _NSIG;
222 do {
223 int x = 0;
224
225 i -= 4;
226 if (sigismember(set, i+1)) x |= 1;
227 if (sigismember(set, i+2)) x |= 2;
228 if (sigismember(set, i+3)) x |= 4;
229 if (sigismember(set, i+4)) x |= 8;
230 seq_printf(m, "%x", x);
231 } while (i >= 4);
232
233 seq_putc(m, '\n');
234}
235
236static void collect_sigign_sigcatch(struct task_struct *p, sigset_t *ign,
237 sigset_t *catch)
238{
239 struct k_sigaction *k;
240 int i;
241
242 k = p->sighand->action;
243 for (i = 1; i <= _NSIG; ++i, ++k) {
244 if (k->sa.sa_handler == SIG_IGN)
245 sigaddset(ign, i);
246 else if (k->sa.sa_handler != SIG_DFL)
247 sigaddset(catch, i);
248 }
249}
250
251static inline void task_sig(struct seq_file *m, struct task_struct *p)
252{
253 unsigned long flags;
254 sigset_t pending, shpending, blocked, ignored, caught;
255 int num_threads = 0;
256 unsigned long qsize = 0;
257 unsigned long qlim = 0;
258
259 sigemptyset(&pending);
260 sigemptyset(&shpending);
261 sigemptyset(&blocked);
262 sigemptyset(&ignored);
263 sigemptyset(&caught);
264
265 if (lock_task_sighand(p, &flags)) {
266 pending = p->pending.signal;
267 shpending = p->signal->shared_pending.signal;
268 blocked = p->blocked;
269 collect_sigign_sigcatch(p, &ignored, &caught);
270 num_threads = get_nr_threads(p);
271 rcu_read_lock(); /* FIXME: is this correct? */
272 qsize = atomic_read(&__task_cred(p)->user->sigpending);
273 rcu_read_unlock();
274 qlim = task_rlimit(p, RLIMIT_SIGPENDING);
275 unlock_task_sighand(p, &flags);
276 }
277
278 seq_printf(m, "Threads:\t%d\n", num_threads);
279 seq_printf(m, "SigQ:\t%lu/%lu\n", qsize, qlim);
280
281 /* render them all */
282 render_sigset_t(m, "SigPnd:\t", &pending);
283 render_sigset_t(m, "ShdPnd:\t", &shpending);
284 render_sigset_t(m, "SigBlk:\t", &blocked);
285 render_sigset_t(m, "SigIgn:\t", &ignored);
286 render_sigset_t(m, "SigCgt:\t", &caught);
287}
288
289static void render_cap_t(struct seq_file *m, const char *header,
290 kernel_cap_t *a)
291{
292 unsigned __capi;
293
294 seq_puts(m, header);
295 CAP_FOR_EACH_U32(__capi) {
296 seq_printf(m, "%08x",
297 a->cap[(_KERNEL_CAPABILITY_U32S-1) - __capi]);
298 }
299 seq_putc(m, '\n');
300}
301
302static inline void task_cap(struct seq_file *m, struct task_struct *p)
303{
304 const struct cred *cred;
305 kernel_cap_t cap_inheritable, cap_permitted, cap_effective, cap_bset;
306
307 rcu_read_lock();
308 cred = __task_cred(p);
309 cap_inheritable = cred->cap_inheritable;
310 cap_permitted = cred->cap_permitted;
311 cap_effective = cred->cap_effective;
312 cap_bset = cred->cap_bset;
313 rcu_read_unlock();
314
315 render_cap_t(m, "CapInh:\t", &cap_inheritable);
316 render_cap_t(m, "CapPrm:\t", &cap_permitted);
317 render_cap_t(m, "CapEff:\t", &cap_effective);
318 render_cap_t(m, "CapBnd:\t", &cap_bset);
319}
320
321static inline void task_context_switch_counts(struct seq_file *m,
322 struct task_struct *p)
323{
324 seq_printf(m, "voluntary_ctxt_switches:\t%lu\n"
325 "nonvoluntary_ctxt_switches:\t%lu\n",
326 p->nvcsw,
327 p->nivcsw);
328}
329
330static void task_cpus_allowed(struct seq_file *m, struct task_struct *task)
331{
332 seq_puts(m, "Cpus_allowed:\t");
333 seq_cpumask(m, &task->cpus_allowed);
334 seq_putc(m, '\n');
335 seq_puts(m, "Cpus_allowed_list:\t");
336 seq_cpumask_list(m, &task->cpus_allowed);
337 seq_putc(m, '\n');
338}
339
340int proc_pid_status(struct seq_file *m, struct pid_namespace *ns,
341 struct pid *pid, struct task_struct *task)
342{
343 struct mm_struct *mm = get_task_mm(task);
344
345 task_name(m, task);
346 task_state(m, ns, pid, task);
347
348 if (mm) {
349 task_mem(m, mm);
350 mmput(mm);
351 }
352 task_sig(m, task);
353 task_cap(m, task);
354 task_cpus_allowed(m, task);
355 cpuset_task_status_allowed(m, task);
356 task_context_switch_counts(m, task);
357 return 0;
358}
359
360static int do_task_stat(struct seq_file *m, struct pid_namespace *ns,
361 struct pid *pid, struct task_struct *task, int whole)
362{
363 unsigned long vsize, eip, esp, wchan = ~0UL;
364 long priority, nice;
365 int tty_pgrp = -1, tty_nr = 0;
366 sigset_t sigign, sigcatch;
367 char state;
368 pid_t ppid = 0, pgid = -1, sid = -1;
369 int num_threads = 0;
370 int permitted;
371 struct mm_struct *mm;
372 unsigned long long start_time;
373 unsigned long cmin_flt = 0, cmaj_flt = 0;
374 unsigned long min_flt = 0, maj_flt = 0;
375 cputime_t cutime, cstime, utime, stime;
376 cputime_t cgtime, gtime;
377 unsigned long rsslim = 0;
378 char tcomm[sizeof(task->comm)];
379 unsigned long flags;
380
381 state = *get_task_state(task);
382 vsize = eip = esp = 0;
383 permitted = ptrace_may_access(task, PTRACE_MODE_READ);
384 mm = get_task_mm(task);
385 if (mm) {
386 vsize = task_vsize(mm);
387 if (permitted) {
388 eip = KSTK_EIP(task);
389 esp = KSTK_ESP(task);
390 }
391 }
392
393 get_task_comm(tcomm, task);
394
395 sigemptyset(&sigign);
396 sigemptyset(&sigcatch);
397 cutime = cstime = utime = stime = cputime_zero;
398 cgtime = gtime = cputime_zero;
399
400 if (lock_task_sighand(task, &flags)) {
401 struct signal_struct *sig = task->signal;
402
403 if (sig->tty) {
404 struct pid *pgrp = tty_get_pgrp(sig->tty);
405 tty_pgrp = pid_nr_ns(pgrp, ns);
406 put_pid(pgrp);
407 tty_nr = new_encode_dev(tty_devnum(sig->tty));
408 }
409
410 num_threads = get_nr_threads(task);
411 collect_sigign_sigcatch(task, &sigign, &sigcatch);
412
413 cmin_flt = sig->cmin_flt;
414 cmaj_flt = sig->cmaj_flt;
415 cutime = sig->cutime;
416 cstime = sig->cstime;
417 cgtime = sig->cgtime;
418 rsslim = ACCESS_ONCE(sig->rlim[RLIMIT_RSS].rlim_cur);
419
420 /* add up live thread stats at the group level */
421 if (whole) {
422 struct task_struct *t = task;
423 do {
424 min_flt += t->min_flt;
425 maj_flt += t->maj_flt;
426 gtime = cputime_add(gtime, t->gtime);
427 t = next_thread(t);
428 } while (t != task);
429
430 min_flt += sig->min_flt;
431 maj_flt += sig->maj_flt;
432 thread_group_times(task, &utime, &stime);
433 gtime = cputime_add(gtime, sig->gtime);
434 }
435
436 sid = task_session_nr_ns(task, ns);
437 ppid = task_tgid_nr_ns(task->real_parent, ns);
438 pgid = task_pgrp_nr_ns(task, ns);
439
440 unlock_task_sighand(task, &flags);
441 }
442
443 if (permitted && (!whole || num_threads < 2))
444 wchan = get_wchan(task);
445 if (!whole) {
446 min_flt = task->min_flt;
447 maj_flt = task->maj_flt;
448 task_times(task, &utime, &stime);
449 gtime = task->gtime;
450 }
451
452 /* scale priority and nice values from timeslices to -20..20 */
453 /* to make it look like a "normal" Unix priority/nice value */
454 priority = task_prio(task);
455 nice = task_nice(task);
456
457 /* Temporary variable needed for gcc-2.96 */
458 /* convert timespec -> nsec*/
459 start_time =
460 (unsigned long long)task->real_start_time.tv_sec * NSEC_PER_SEC
461 + task->real_start_time.tv_nsec;
462 /* convert nsec -> ticks */
463 start_time = nsec_to_clock_t(start_time);
464
465 seq_printf(m, "%d (%s) %c %d %d %d %d %d %u %lu \
466%lu %lu %lu %lu %lu %ld %ld %ld %ld %d 0 %llu %lu %ld %lu %lu %lu %lu %lu \
467%lu %lu %lu %lu %lu %lu %lu %lu %d %d %u %u %llu %lu %ld\n",
468 pid_nr_ns(pid, ns),
469 tcomm,
470 state,
471 ppid,
472 pgid,
473 sid,
474 tty_nr,
475 tty_pgrp,
476 task->flags,
477 min_flt,
478 cmin_flt,
479 maj_flt,
480 cmaj_flt,
481 cputime_to_clock_t(utime),
482 cputime_to_clock_t(stime),
483 cputime_to_clock_t(cutime),
484 cputime_to_clock_t(cstime),
485 priority,
486 nice,
487 num_threads,
488 start_time,
489 vsize,
490 mm ? get_mm_rss(mm) : 0,
491 rsslim,
492 mm ? (permitted ? mm->start_code : 1) : 0,
493 mm ? (permitted ? mm->end_code : 1) : 0,
494 (permitted && mm) ? mm->start_stack : 0,
495 esp,
496 eip,
497 /* The signal information here is obsolete.
498 * It must be decimal for Linux 2.0 compatibility.
499 * Use /proc/#/status for real-time signals.
500 */
501 task->pending.signal.sig[0] & 0x7fffffffUL,
502 task->blocked.sig[0] & 0x7fffffffUL,
503 sigign .sig[0] & 0x7fffffffUL,
504 sigcatch .sig[0] & 0x7fffffffUL,
505 wchan,
506 0UL,
507 0UL,
508 task->exit_signal,
509 task_cpu(task),
510 task->rt_priority,
511 task->policy,
512 (unsigned long long)delayacct_blkio_ticks(task),
513 cputime_to_clock_t(gtime),
514 cputime_to_clock_t(cgtime));
515 if (mm)
516 mmput(mm);
517 return 0;
518}
519
520int proc_tid_stat(struct seq_file *m, struct pid_namespace *ns,
521 struct pid *pid, struct task_struct *task)
522{
523 return do_task_stat(m, ns, pid, task, 0);
524}
525
526int proc_tgid_stat(struct seq_file *m, struct pid_namespace *ns,
527 struct pid *pid, struct task_struct *task)
528{
529 return do_task_stat(m, ns, pid, task, 1);
530}
531
532int proc_pid_statm(struct seq_file *m, struct pid_namespace *ns,
533 struct pid *pid, struct task_struct *task)
534{
535 unsigned long size = 0, resident = 0, shared = 0, text = 0, data = 0;
536 struct mm_struct *mm = get_task_mm(task);
537
538 if (mm) {
539 size = task_statm(mm, &shared, &text, &data, &resident);
540 mmput(mm);
541 }
542 seq_printf(m, "%lu %lu %lu %lu 0 %lu 0\n",
543 size, resident, shared, text, data);
544
545 return 0;
546}