2 * linux/fs/proc/array.c
4 * Copyright (C) 1992 by Linus Torvalds
5 * based on ideas by Darren Senn
8 * Michael. K. Johnson: stat,statm extensions.
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.
17 * Danny ter Haar : added cpuinfo
20 * Alessandro Rubini : profile extension.
23 * Jeff Tranter : added BogoMips field to cpuinfo
26 * Bruno Haible : remove 4K limit for the maps file
29 * Yves Arrouye : remove removal of trailing spaces in get_array.
32 * Jerome Forissier : added per-CPU time information to /proc/stat
33 * and /proc/<pid>/cpu extension
35 * - Incorporation and non-SMP safe operation
36 * of forissier patch in 2.1.78 by
42 * Alan Cox : security fixes.
45 * Al Viro : safe handling of mm_struct
47 * Gerhard Wichert : added BIGMEM support
50 * Al Viro & Jeff Garzik : moved most of the thing into base.c and
51 * : proc_misc.c. The rest may eventually go into
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>
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/string_helpers.h>
85 #include <linux/user_namespace.h>
86 #include <linux/fs_struct.h>
88 #include <asm/pgtable.h>
89 #include <asm/processor.h>
92 static inline void task_name(struct seq_file *m, struct task_struct *p)
96 char tcomm[sizeof(p->comm)];
99 get_task_comm(tcomm, p);
101 seq_puts(m, "Name:\t");
103 size = seq_get_buf(m, &buf);
104 ret = string_escape_str(tcomm, buf, size, ESCAPE_SPACE | ESCAPE_SPECIAL, "\n\\");
105 seq_commit(m, ret < size ? ret : -1);
111 * The task state array is a strange "bitmap" of
112 * reasons to sleep. Thus "running" is zero, and
113 * you can test for combinations of others with
116 static const char * const task_state_array[] = {
117 "R (running)", /* 0 */
118 "S (sleeping)", /* 1 */
119 "D (disk sleep)", /* 2 */
120 "T (stopped)", /* 4 */
121 "t (tracing stop)", /* 8 */
123 "Z (zombie)", /* 32 */
126 static inline const char *get_task_state(struct task_struct *tsk)
128 unsigned int state = (tsk->state | tsk->exit_state) & TASK_REPORT;
131 * Parked tasks do not run; they sit in __kthread_parkme().
132 * Without this check, we would report them as running, which is
133 * clearly wrong, so we report them as sleeping instead.
135 if (tsk->state == TASK_PARKED)
136 state = TASK_INTERRUPTIBLE;
138 BUILD_BUG_ON(1 + ilog2(TASK_REPORT) != ARRAY_SIZE(task_state_array)-1);
140 return task_state_array[fls(state)];
143 static inline int get_task_umask(struct task_struct *tsk)
145 struct fs_struct *fs;
156 static inline void task_state(struct seq_file *m, struct pid_namespace *ns,
157 struct pid *pid, struct task_struct *p)
159 struct user_namespace *user_ns = seq_user_ns(m);
160 struct group_info *group_info;
162 struct task_struct *tracer;
163 const struct cred *cred;
164 pid_t ppid, tpid = 0, tgid, ngid;
165 unsigned int max_fds = 0;
168 ppid = pid_alive(p) ?
169 task_tgid_nr_ns(rcu_dereference(p->real_parent), ns) : 0;
171 tracer = ptrace_parent(p);
173 tpid = task_pid_nr_ns(tracer, ns);
175 tgid = task_tgid_nr_ns(p, ns);
176 ngid = task_numa_group_id(p);
177 cred = get_task_cred(p);
179 umask = get_task_umask(p);
181 seq_printf(m, "Umask:\t%#04o\n", umask);
185 max_fds = files_fdtable(p->files)->max_fds;
189 seq_printf(m, "State:\t%s", get_task_state(p));
191 seq_put_decimal_ull(m, "\nTgid:\t", tgid);
192 seq_put_decimal_ull(m, "\nNgid:\t", ngid);
193 seq_put_decimal_ull(m, "\nPid:\t", pid_nr_ns(pid, ns));
194 seq_put_decimal_ull(m, "\nPPid:\t", ppid);
195 seq_put_decimal_ull(m, "\nTracerPid:\t", tpid);
196 seq_put_decimal_ull(m, "\nUid:\t", from_kuid_munged(user_ns, cred->uid));
197 seq_put_decimal_ull(m, "\t", from_kuid_munged(user_ns, cred->euid));
198 seq_put_decimal_ull(m, "\t", from_kuid_munged(user_ns, cred->suid));
199 seq_put_decimal_ull(m, "\t", from_kuid_munged(user_ns, cred->fsuid));
200 seq_put_decimal_ull(m, "\nGid:\t", from_kgid_munged(user_ns, cred->gid));
201 seq_put_decimal_ull(m, "\t", from_kgid_munged(user_ns, cred->egid));
202 seq_put_decimal_ull(m, "\t", from_kgid_munged(user_ns, cred->sgid));
203 seq_put_decimal_ull(m, "\t", from_kgid_munged(user_ns, cred->fsgid));
204 seq_put_decimal_ull(m, "\nFDSize:\t", max_fds);
206 seq_puts(m, "\nGroups:\t");
207 group_info = cred->group_info;
208 for (g = 0; g < group_info->ngroups; g++)
209 seq_put_decimal_ull(m, g ? " " : "",
210 from_kgid_munged(user_ns, group_info->gid[g]));
212 /* Trailing space shouldn't have been added in the first place. */
216 seq_puts(m, "\nNStgid:");
217 for (g = ns->level; g <= pid->level; g++)
218 seq_put_decimal_ull(m, "\t", task_tgid_nr_ns(p, pid->numbers[g].ns));
219 seq_puts(m, "\nNSpid:");
220 for (g = ns->level; g <= pid->level; g++)
221 seq_put_decimal_ull(m, "\t", task_pid_nr_ns(p, pid->numbers[g].ns));
222 seq_puts(m, "\nNSpgid:");
223 for (g = ns->level; g <= pid->level; g++)
224 seq_put_decimal_ull(m, "\t", task_pgrp_nr_ns(p, pid->numbers[g].ns));
225 seq_puts(m, "\nNSsid:");
226 for (g = ns->level; g <= pid->level; g++)
227 seq_put_decimal_ull(m, "\t", task_session_nr_ns(p, pid->numbers[g].ns));
232 void render_sigset_t(struct seq_file *m, const char *header,
244 if (sigismember(set, i+1)) x |= 1;
245 if (sigismember(set, i+2)) x |= 2;
246 if (sigismember(set, i+3)) x |= 4;
247 if (sigismember(set, i+4)) x |= 8;
248 seq_printf(m, "%x", x);
254 static void collect_sigign_sigcatch(struct task_struct *p, sigset_t *ign,
257 struct k_sigaction *k;
260 k = p->sighand->action;
261 for (i = 1; i <= _NSIG; ++i, ++k) {
262 if (k->sa.sa_handler == SIG_IGN)
264 else if (k->sa.sa_handler != SIG_DFL)
269 static inline void task_sig(struct seq_file *m, struct task_struct *p)
272 sigset_t pending, shpending, blocked, ignored, caught;
274 unsigned long qsize = 0;
275 unsigned long qlim = 0;
277 sigemptyset(&pending);
278 sigemptyset(&shpending);
279 sigemptyset(&blocked);
280 sigemptyset(&ignored);
281 sigemptyset(&caught);
283 if (lock_task_sighand(p, &flags)) {
284 pending = p->pending.signal;
285 shpending = p->signal->shared_pending.signal;
286 blocked = p->blocked;
287 collect_sigign_sigcatch(p, &ignored, &caught);
288 num_threads = get_nr_threads(p);
289 rcu_read_lock(); /* FIXME: is this correct? */
290 qsize = atomic_read(&__task_cred(p)->user->sigpending);
292 qlim = task_rlimit(p, RLIMIT_SIGPENDING);
293 unlock_task_sighand(p, &flags);
296 seq_put_decimal_ull(m, "Threads:\t", num_threads);
297 seq_put_decimal_ull(m, "\nSigQ:\t", qsize);
298 seq_put_decimal_ull(m, "/", qlim);
300 /* render them all */
301 render_sigset_t(m, "\nSigPnd:\t", &pending);
302 render_sigset_t(m, "ShdPnd:\t", &shpending);
303 render_sigset_t(m, "SigBlk:\t", &blocked);
304 render_sigset_t(m, "SigIgn:\t", &ignored);
305 render_sigset_t(m, "SigCgt:\t", &caught);
308 static void render_cap_t(struct seq_file *m, const char *header,
314 CAP_FOR_EACH_U32(__capi) {
315 seq_printf(m, "%08x",
316 a->cap[CAP_LAST_U32 - __capi]);
321 static inline void task_cap(struct seq_file *m, struct task_struct *p)
323 const struct cred *cred;
324 kernel_cap_t cap_inheritable, cap_permitted, cap_effective,
325 cap_bset, cap_ambient;
328 cred = __task_cred(p);
329 cap_inheritable = cred->cap_inheritable;
330 cap_permitted = cred->cap_permitted;
331 cap_effective = cred->cap_effective;
332 cap_bset = cred->cap_bset;
333 cap_ambient = cred->cap_ambient;
336 render_cap_t(m, "CapInh:\t", &cap_inheritable);
337 render_cap_t(m, "CapPrm:\t", &cap_permitted);
338 render_cap_t(m, "CapEff:\t", &cap_effective);
339 render_cap_t(m, "CapBnd:\t", &cap_bset);
340 render_cap_t(m, "CapAmb:\t", &cap_ambient);
343 static inline void task_seccomp(struct seq_file *m, struct task_struct *p)
345 #ifdef CONFIG_SECCOMP
346 seq_put_decimal_ull(m, "Seccomp:\t", p->seccomp.mode);
351 static inline void task_context_switch_counts(struct seq_file *m,
352 struct task_struct *p)
354 seq_put_decimal_ull(m, "voluntary_ctxt_switches:\t", p->nvcsw);
355 seq_put_decimal_ull(m, "\nnonvoluntary_ctxt_switches:\t", p->nivcsw);
359 static void task_cpus_allowed(struct seq_file *m, struct task_struct *task)
361 seq_printf(m, "Cpus_allowed:\t%*pb\n",
362 cpumask_pr_args(&task->cpus_allowed));
363 seq_printf(m, "Cpus_allowed_list:\t%*pbl\n",
364 cpumask_pr_args(&task->cpus_allowed));
367 int proc_pid_status(struct seq_file *m, struct pid_namespace *ns,
368 struct pid *pid, struct task_struct *task)
370 struct mm_struct *mm = get_task_mm(task);
373 task_state(m, ns, pid, task);
381 task_seccomp(m, task);
382 task_cpus_allowed(m, task);
383 cpuset_task_status_allowed(m, task);
384 task_context_switch_counts(m, task);
388 static int do_task_stat(struct seq_file *m, struct pid_namespace *ns,
389 struct pid *pid, struct task_struct *task, int whole)
391 unsigned long vsize, eip, esp, wchan = 0;
393 int tty_pgrp = -1, tty_nr = 0;
394 sigset_t sigign, sigcatch;
396 pid_t ppid = 0, pgid = -1, sid = -1;
399 struct mm_struct *mm;
400 unsigned long long start_time;
401 unsigned long cmin_flt = 0, cmaj_flt = 0;
402 unsigned long min_flt = 0, maj_flt = 0;
403 cputime_t cutime, cstime, utime, stime;
404 cputime_t cgtime, gtime;
405 unsigned long rsslim = 0;
406 char tcomm[sizeof(task->comm)];
409 state = *get_task_state(task);
410 vsize = eip = esp = 0;
411 permitted = ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS | PTRACE_MODE_NOAUDIT);
412 mm = get_task_mm(task);
414 vsize = task_vsize(mm);
416 eip = KSTK_EIP(task);
417 esp = KSTK_ESP(task);
421 get_task_comm(tcomm, task);
423 sigemptyset(&sigign);
424 sigemptyset(&sigcatch);
425 cutime = cstime = utime = stime = 0;
428 if (lock_task_sighand(task, &flags)) {
429 struct signal_struct *sig = task->signal;
432 struct pid *pgrp = tty_get_pgrp(sig->tty);
433 tty_pgrp = pid_nr_ns(pgrp, ns);
435 tty_nr = new_encode_dev(tty_devnum(sig->tty));
438 num_threads = get_nr_threads(task);
439 collect_sigign_sigcatch(task, &sigign, &sigcatch);
441 cmin_flt = sig->cmin_flt;
442 cmaj_flt = sig->cmaj_flt;
443 cutime = sig->cutime;
444 cstime = sig->cstime;
445 cgtime = sig->cgtime;
446 rsslim = ACCESS_ONCE(sig->rlim[RLIMIT_RSS].rlim_cur);
448 /* add up live thread stats at the group level */
450 struct task_struct *t = task;
452 min_flt += t->min_flt;
453 maj_flt += t->maj_flt;
454 gtime += task_gtime(t);
455 } while_each_thread(task, t);
457 min_flt += sig->min_flt;
458 maj_flt += sig->maj_flt;
459 thread_group_cputime_adjusted(task, &utime, &stime);
463 sid = task_session_nr_ns(task, ns);
464 ppid = task_tgid_nr_ns(task->real_parent, ns);
465 pgid = task_pgrp_nr_ns(task, ns);
467 unlock_task_sighand(task, &flags);
470 if (permitted && (!whole || num_threads < 2))
471 wchan = get_wchan(task);
473 min_flt = task->min_flt;
474 maj_flt = task->maj_flt;
475 task_cputime_adjusted(task, &utime, &stime);
476 gtime = task_gtime(task);
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);
484 /* convert nsec -> ticks */
485 start_time = nsec_to_clock_t(task->real_start_time);
487 seq_printf(m, "%d (%s) %c", pid_nr_ns(pid, ns), tcomm, state);
488 seq_put_decimal_ll(m, " ", ppid);
489 seq_put_decimal_ll(m, " ", pgid);
490 seq_put_decimal_ll(m, " ", sid);
491 seq_put_decimal_ll(m, " ", tty_nr);
492 seq_put_decimal_ll(m, " ", tty_pgrp);
493 seq_put_decimal_ull(m, " ", task->flags);
494 seq_put_decimal_ull(m, " ", min_flt);
495 seq_put_decimal_ull(m, " ", cmin_flt);
496 seq_put_decimal_ull(m, " ", maj_flt);
497 seq_put_decimal_ull(m, " ", cmaj_flt);
498 seq_put_decimal_ull(m, " ", cputime_to_clock_t(utime));
499 seq_put_decimal_ull(m, " ", cputime_to_clock_t(stime));
500 seq_put_decimal_ll(m, " ", cputime_to_clock_t(cutime));
501 seq_put_decimal_ll(m, " ", cputime_to_clock_t(cstime));
502 seq_put_decimal_ll(m, " ", priority);
503 seq_put_decimal_ll(m, " ", nice);
504 seq_put_decimal_ll(m, " ", num_threads);
505 seq_put_decimal_ull(m, " ", 0);
506 seq_put_decimal_ull(m, " ", start_time);
507 seq_put_decimal_ull(m, " ", vsize);
508 seq_put_decimal_ull(m, " ", mm ? get_mm_rss(mm) : 0);
509 seq_put_decimal_ull(m, " ", rsslim);
510 seq_put_decimal_ull(m, " ", mm ? (permitted ? mm->start_code : 1) : 0);
511 seq_put_decimal_ull(m, " ", mm ? (permitted ? mm->end_code : 1) : 0);
512 seq_put_decimal_ull(m, " ", (permitted && mm) ? mm->start_stack : 0);
513 seq_put_decimal_ull(m, " ", esp);
514 seq_put_decimal_ull(m, " ", eip);
515 /* The signal information here is obsolete.
516 * It must be decimal for Linux 2.0 compatibility.
517 * Use /proc/#/status for real-time signals.
519 seq_put_decimal_ull(m, " ", task->pending.signal.sig[0] & 0x7fffffffUL);
520 seq_put_decimal_ull(m, " ", task->blocked.sig[0] & 0x7fffffffUL);
521 seq_put_decimal_ull(m, " ", sigign.sig[0] & 0x7fffffffUL);
522 seq_put_decimal_ull(m, " ", sigcatch.sig[0] & 0x7fffffffUL);
525 * We used to output the absolute kernel address, but that's an
526 * information leak - so instead we show a 0/1 flag here, to signal
527 * to user-space whether there's a wchan field in /proc/PID/wchan.
529 * This works with older implementations of procps as well.
536 seq_put_decimal_ull(m, " ", 0);
537 seq_put_decimal_ull(m, " ", 0);
538 seq_put_decimal_ll(m, " ", task->exit_signal);
539 seq_put_decimal_ll(m, " ", task_cpu(task));
540 seq_put_decimal_ull(m, " ", task->rt_priority);
541 seq_put_decimal_ull(m, " ", task->policy);
542 seq_put_decimal_ull(m, " ", delayacct_blkio_ticks(task));
543 seq_put_decimal_ull(m, " ", cputime_to_clock_t(gtime));
544 seq_put_decimal_ll(m, " ", cputime_to_clock_t(cgtime));
546 if (mm && permitted) {
547 seq_put_decimal_ull(m, " ", mm->start_data);
548 seq_put_decimal_ull(m, " ", mm->end_data);
549 seq_put_decimal_ull(m, " ", mm->start_brk);
550 seq_put_decimal_ull(m, " ", mm->arg_start);
551 seq_put_decimal_ull(m, " ", mm->arg_end);
552 seq_put_decimal_ull(m, " ", mm->env_start);
553 seq_put_decimal_ull(m, " ", mm->env_end);
555 seq_puts(m, " 0 0 0 0 0 0 0");
558 seq_put_decimal_ll(m, " ", task->exit_code);
568 int proc_tid_stat(struct seq_file *m, struct pid_namespace *ns,
569 struct pid *pid, struct task_struct *task)
571 return do_task_stat(m, ns, pid, task, 0);
574 int proc_tgid_stat(struct seq_file *m, struct pid_namespace *ns,
575 struct pid *pid, struct task_struct *task)
577 return do_task_stat(m, ns, pid, task, 1);
580 int proc_pid_statm(struct seq_file *m, struct pid_namespace *ns,
581 struct pid *pid, struct task_struct *task)
583 unsigned long size = 0, resident = 0, shared = 0, text = 0, data = 0;
584 struct mm_struct *mm = get_task_mm(task);
587 size = task_statm(mm, &shared, &text, &data, &resident);
591 * For quick read, open code by putting numbers directly
593 * seq_printf(m, "%lu %lu %lu %lu 0 %lu 0\n",
594 * size, resident, shared, text, data);
596 seq_put_decimal_ull(m, "", size);
597 seq_put_decimal_ull(m, " ", resident);
598 seq_put_decimal_ull(m, " ", shared);
599 seq_put_decimal_ull(m, " ", text);
600 seq_put_decimal_ull(m, " ", 0);
601 seq_put_decimal_ull(m, " ", data);
602 seq_put_decimal_ull(m, " ", 0);
608 #ifdef CONFIG_PROC_CHILDREN
610 get_children_pid(struct inode *inode, struct pid *pid_prev, loff_t pos)
612 struct task_struct *start, *task;
613 struct pid *pid = NULL;
615 read_lock(&tasklist_lock);
617 start = pid_task(proc_pid(inode), PIDTYPE_PID);
622 * Lets try to continue searching first, this gives
623 * us significant speedup on children-rich processes.
626 task = pid_task(pid_prev, PIDTYPE_PID);
627 if (task && task->real_parent == start &&
628 !(list_empty(&task->sibling))) {
629 if (list_is_last(&task->sibling, &start->children))
631 task = list_first_entry(&task->sibling,
632 struct task_struct, sibling);
633 pid = get_pid(task_pid(task));
641 * We might miss some children here if children
642 * are exited while we were not holding the lock,
643 * but it was never promised to be accurate that
646 * "Just suppose that the parent sleeps, but N children
647 * exit after we printed their tids. Now the slow paths
648 * skips N extra children, we miss N tasks." (c)
650 * So one need to stop or freeze the leader and all
651 * its children to get a precise result.
653 list_for_each_entry(task, &start->children, sibling) {
655 pid = get_pid(task_pid(task));
661 read_unlock(&tasklist_lock);
665 static int children_seq_show(struct seq_file *seq, void *v)
667 struct inode *inode = seq->private;
670 pid = pid_nr_ns(v, inode->i_sb->s_fs_info);
671 seq_printf(seq, "%d ", pid);
676 static void *children_seq_start(struct seq_file *seq, loff_t *pos)
678 return get_children_pid(seq->private, NULL, *pos);
681 static void *children_seq_next(struct seq_file *seq, void *v, loff_t *pos)
685 pid = get_children_pid(seq->private, v, *pos + 1);
692 static void children_seq_stop(struct seq_file *seq, void *v)
697 static const struct seq_operations children_seq_ops = {
698 .start = children_seq_start,
699 .next = children_seq_next,
700 .stop = children_seq_stop,
701 .show = children_seq_show,
704 static int children_seq_open(struct inode *inode, struct file *file)
709 ret = seq_open(file, &children_seq_ops);
713 m = file->private_data;
719 int children_seq_release(struct inode *inode, struct file *file)
721 seq_release(inode, file);
725 const struct file_operations proc_tid_children_operations = {
726 .open = children_seq_open,
729 .release = children_seq_release,
731 #endif /* CONFIG_PROC_CHILDREN */