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.
9 * <johnsonm@stolaf.edu>
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>
17 * Danny ter Haar : added cpuinfo
20 * Alessandro Rubini : profile extension.
21 * <rubini@ipvvis.unipv.it>
23 * Jeff Tranter : added BogoMips field to cpuinfo
24 * <Jeff_Tranter@Mitel.COM>
26 * Bruno Haible : remove 4K limit for the maps file
27 * <haible@ma2s2.mathematik.uni-karlsruhe.de>
29 * Yves Arrouye : remove removal of trailing spaces in get_array.
30 * <Yves.Arrouye@marin.fdn.fr>
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>
39 * aeb@cwi.nl : /proc/partitions
42 * Alan Cox : security fixes.
43 * <alan@lxorguk.ukuu.org.uk>
45 * Al Viro : safe handling of mm_struct
47 * Gerhard Wichert : added BIGMEM support
48 * Siemens AG <Gerhard.Wichert@pdb.siemens.de>
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>
87 #include <asm/pgtable.h>
88 #include <asm/processor.h>
91 static inline void task_name(struct seq_file *m, struct task_struct *p)
94 char tcomm[sizeof(p->comm)];
96 get_task_comm(tcomm, p);
98 seq_puts(m, "Name:\t");
99 buf = m->buf + m->count;
101 /* Ignore error for now */
102 string_escape_str(tcomm, &buf, m->size - m->count,
103 ESCAPE_SPACE | ESCAPE_SPECIAL, "\n\\");
105 m->count = buf - m->buf;
110 * The task state array is a strange "bitmap" of
111 * reasons to sleep. Thus "running" is zero, and
112 * you can test for combinations of others with
115 static const char * const task_state_array[] = {
116 "R (running)", /* 0 */
117 "S (sleeping)", /* 1 */
118 "D (disk sleep)", /* 2 */
119 "T (stopped)", /* 4 */
120 "t (tracing stop)", /* 8 */
122 "Z (zombie)", /* 32 */
125 static inline const char *get_task_state(struct task_struct *tsk)
127 unsigned int state = (tsk->state | tsk->exit_state) & TASK_REPORT;
129 BUILD_BUG_ON(1 + ilog2(TASK_REPORT) != ARRAY_SIZE(task_state_array)-1);
131 return task_state_array[fls(state)];
134 static inline void task_state(struct seq_file *m, struct pid_namespace *ns,
135 struct pid *pid, struct task_struct *p)
137 struct user_namespace *user_ns = seq_user_ns(m);
138 struct group_info *group_info;
140 struct task_struct *tracer;
141 const struct cred *cred;
142 pid_t ppid, tpid = 0, tgid, ngid;
143 unsigned int max_fds = 0;
146 ppid = pid_alive(p) ?
147 task_tgid_nr_ns(rcu_dereference(p->real_parent), ns) : 0;
149 tracer = ptrace_parent(p);
151 tpid = task_pid_nr_ns(tracer, ns);
153 tgid = task_tgid_nr_ns(p, ns);
154 ngid = task_numa_group_id(p);
155 cred = get_task_cred(p);
159 max_fds = files_fdtable(p->files)->max_fds;
170 "Uid:\t%d\t%d\t%d\t%d\n"
171 "Gid:\t%d\t%d\t%d\t%d\n"
172 "FDSize:\t%d\nGroups:\t",
174 tgid, ngid, pid_nr_ns(pid, ns), ppid, tpid,
175 from_kuid_munged(user_ns, cred->uid),
176 from_kuid_munged(user_ns, cred->euid),
177 from_kuid_munged(user_ns, cred->suid),
178 from_kuid_munged(user_ns, cred->fsuid),
179 from_kgid_munged(user_ns, cred->gid),
180 from_kgid_munged(user_ns, cred->egid),
181 from_kgid_munged(user_ns, cred->sgid),
182 from_kgid_munged(user_ns, cred->fsgid),
185 group_info = cred->group_info;
186 for (g = 0; g < group_info->ngroups; g++)
188 from_kgid_munged(user_ns, GROUP_AT(group_info, g)));
194 void render_sigset_t(struct seq_file *m, const char *header,
206 if (sigismember(set, i+1)) x |= 1;
207 if (sigismember(set, i+2)) x |= 2;
208 if (sigismember(set, i+3)) x |= 4;
209 if (sigismember(set, i+4)) x |= 8;
210 seq_printf(m, "%x", x);
216 static void collect_sigign_sigcatch(struct task_struct *p, sigset_t *ign,
219 struct k_sigaction *k;
222 k = p->sighand->action;
223 for (i = 1; i <= _NSIG; ++i, ++k) {
224 if (k->sa.sa_handler == SIG_IGN)
226 else if (k->sa.sa_handler != SIG_DFL)
231 static inline void task_sig(struct seq_file *m, struct task_struct *p)
234 sigset_t pending, shpending, blocked, ignored, caught;
236 unsigned long qsize = 0;
237 unsigned long qlim = 0;
239 sigemptyset(&pending);
240 sigemptyset(&shpending);
241 sigemptyset(&blocked);
242 sigemptyset(&ignored);
243 sigemptyset(&caught);
245 if (lock_task_sighand(p, &flags)) {
246 pending = p->pending.signal;
247 shpending = p->signal->shared_pending.signal;
248 blocked = p->blocked;
249 collect_sigign_sigcatch(p, &ignored, &caught);
250 num_threads = get_nr_threads(p);
251 rcu_read_lock(); /* FIXME: is this correct? */
252 qsize = atomic_read(&__task_cred(p)->user->sigpending);
254 qlim = task_rlimit(p, RLIMIT_SIGPENDING);
255 unlock_task_sighand(p, &flags);
258 seq_printf(m, "Threads:\t%d\n", num_threads);
259 seq_printf(m, "SigQ:\t%lu/%lu\n", qsize, qlim);
261 /* render them all */
262 render_sigset_t(m, "SigPnd:\t", &pending);
263 render_sigset_t(m, "ShdPnd:\t", &shpending);
264 render_sigset_t(m, "SigBlk:\t", &blocked);
265 render_sigset_t(m, "SigIgn:\t", &ignored);
266 render_sigset_t(m, "SigCgt:\t", &caught);
269 static void render_cap_t(struct seq_file *m, const char *header,
275 CAP_FOR_EACH_U32(__capi) {
276 seq_printf(m, "%08x",
277 a->cap[CAP_LAST_U32 - __capi]);
282 static inline void task_cap(struct seq_file *m, struct task_struct *p)
284 const struct cred *cred;
285 kernel_cap_t cap_inheritable, cap_permitted, cap_effective, cap_bset;
288 cred = __task_cred(p);
289 cap_inheritable = cred->cap_inheritable;
290 cap_permitted = cred->cap_permitted;
291 cap_effective = cred->cap_effective;
292 cap_bset = cred->cap_bset;
295 render_cap_t(m, "CapInh:\t", &cap_inheritable);
296 render_cap_t(m, "CapPrm:\t", &cap_permitted);
297 render_cap_t(m, "CapEff:\t", &cap_effective);
298 render_cap_t(m, "CapBnd:\t", &cap_bset);
301 static inline void task_seccomp(struct seq_file *m, struct task_struct *p)
303 #ifdef CONFIG_SECCOMP
304 seq_printf(m, "Seccomp:\t%d\n", p->seccomp.mode);
308 static inline void task_context_switch_counts(struct seq_file *m,
309 struct task_struct *p)
311 seq_printf(m, "voluntary_ctxt_switches:\t%lu\n"
312 "nonvoluntary_ctxt_switches:\t%lu\n",
317 static void task_cpus_allowed(struct seq_file *m, struct task_struct *task)
319 seq_printf(m, "Cpus_allowed:\t%*pb\n",
320 cpumask_pr_args(&task->cpus_allowed));
321 seq_printf(m, "Cpus_allowed_list:\t%*pbl\n",
322 cpumask_pr_args(&task->cpus_allowed));
325 int proc_pid_status(struct seq_file *m, struct pid_namespace *ns,
326 struct pid *pid, struct task_struct *task)
328 struct mm_struct *mm = get_task_mm(task);
331 task_state(m, ns, pid, task);
339 task_seccomp(m, task);
340 task_cpus_allowed(m, task);
341 cpuset_task_status_allowed(m, task);
342 task_context_switch_counts(m, task);
346 static int do_task_stat(struct seq_file *m, struct pid_namespace *ns,
347 struct pid *pid, struct task_struct *task, int whole)
349 unsigned long vsize, eip, esp, wchan = ~0UL;
351 int tty_pgrp = -1, tty_nr = 0;
352 sigset_t sigign, sigcatch;
354 pid_t ppid = 0, pgid = -1, sid = -1;
357 struct mm_struct *mm;
358 unsigned long long start_time;
359 unsigned long cmin_flt = 0, cmaj_flt = 0;
360 unsigned long min_flt = 0, maj_flt = 0;
361 cputime_t cutime, cstime, utime, stime;
362 cputime_t cgtime, gtime;
363 unsigned long rsslim = 0;
364 char tcomm[sizeof(task->comm)];
367 state = *get_task_state(task);
368 vsize = eip = esp = 0;
369 permitted = ptrace_may_access(task, PTRACE_MODE_READ | PTRACE_MODE_NOAUDIT);
370 mm = get_task_mm(task);
372 vsize = task_vsize(mm);
374 eip = KSTK_EIP(task);
375 esp = KSTK_ESP(task);
379 get_task_comm(tcomm, task);
381 sigemptyset(&sigign);
382 sigemptyset(&sigcatch);
383 cutime = cstime = utime = stime = 0;
386 if (lock_task_sighand(task, &flags)) {
387 struct signal_struct *sig = task->signal;
390 struct pid *pgrp = tty_get_pgrp(sig->tty);
391 tty_pgrp = pid_nr_ns(pgrp, ns);
393 tty_nr = new_encode_dev(tty_devnum(sig->tty));
396 num_threads = get_nr_threads(task);
397 collect_sigign_sigcatch(task, &sigign, &sigcatch);
399 cmin_flt = sig->cmin_flt;
400 cmaj_flt = sig->cmaj_flt;
401 cutime = sig->cutime;
402 cstime = sig->cstime;
403 cgtime = sig->cgtime;
404 rsslim = ACCESS_ONCE(sig->rlim[RLIMIT_RSS].rlim_cur);
406 /* add up live thread stats at the group level */
408 struct task_struct *t = task;
410 min_flt += t->min_flt;
411 maj_flt += t->maj_flt;
412 gtime += task_gtime(t);
413 } while_each_thread(task, t);
415 min_flt += sig->min_flt;
416 maj_flt += sig->maj_flt;
417 thread_group_cputime_adjusted(task, &utime, &stime);
421 sid = task_session_nr_ns(task, ns);
422 ppid = task_tgid_nr_ns(task->real_parent, ns);
423 pgid = task_pgrp_nr_ns(task, ns);
425 unlock_task_sighand(task, &flags);
428 if (permitted && (!whole || num_threads < 2))
429 wchan = get_wchan(task);
431 min_flt = task->min_flt;
432 maj_flt = task->maj_flt;
433 task_cputime_adjusted(task, &utime, &stime);
434 gtime = task_gtime(task);
437 /* scale priority and nice values from timeslices to -20..20 */
438 /* to make it look like a "normal" Unix priority/nice value */
439 priority = task_prio(task);
440 nice = task_nice(task);
442 /* convert nsec -> ticks */
443 start_time = nsec_to_clock_t(task->real_start_time);
445 seq_printf(m, "%d (%s) %c", pid_nr_ns(pid, ns), tcomm, state);
446 seq_put_decimal_ll(m, ' ', ppid);
447 seq_put_decimal_ll(m, ' ', pgid);
448 seq_put_decimal_ll(m, ' ', sid);
449 seq_put_decimal_ll(m, ' ', tty_nr);
450 seq_put_decimal_ll(m, ' ', tty_pgrp);
451 seq_put_decimal_ull(m, ' ', task->flags);
452 seq_put_decimal_ull(m, ' ', min_flt);
453 seq_put_decimal_ull(m, ' ', cmin_flt);
454 seq_put_decimal_ull(m, ' ', maj_flt);
455 seq_put_decimal_ull(m, ' ', cmaj_flt);
456 seq_put_decimal_ull(m, ' ', cputime_to_clock_t(utime));
457 seq_put_decimal_ull(m, ' ', cputime_to_clock_t(stime));
458 seq_put_decimal_ll(m, ' ', cputime_to_clock_t(cutime));
459 seq_put_decimal_ll(m, ' ', cputime_to_clock_t(cstime));
460 seq_put_decimal_ll(m, ' ', priority);
461 seq_put_decimal_ll(m, ' ', nice);
462 seq_put_decimal_ll(m, ' ', num_threads);
463 seq_put_decimal_ull(m, ' ', 0);
464 seq_put_decimal_ull(m, ' ', start_time);
465 seq_put_decimal_ull(m, ' ', vsize);
466 seq_put_decimal_ull(m, ' ', mm ? get_mm_rss(mm) : 0);
467 seq_put_decimal_ull(m, ' ', rsslim);
468 seq_put_decimal_ull(m, ' ', mm ? (permitted ? mm->start_code : 1) : 0);
469 seq_put_decimal_ull(m, ' ', mm ? (permitted ? mm->end_code : 1) : 0);
470 seq_put_decimal_ull(m, ' ', (permitted && mm) ? mm->start_stack : 0);
471 seq_put_decimal_ull(m, ' ', esp);
472 seq_put_decimal_ull(m, ' ', eip);
473 /* The signal information here is obsolete.
474 * It must be decimal for Linux 2.0 compatibility.
475 * Use /proc/#/status for real-time signals.
477 seq_put_decimal_ull(m, ' ', task->pending.signal.sig[0] & 0x7fffffffUL);
478 seq_put_decimal_ull(m, ' ', task->blocked.sig[0] & 0x7fffffffUL);
479 seq_put_decimal_ull(m, ' ', sigign.sig[0] & 0x7fffffffUL);
480 seq_put_decimal_ull(m, ' ', sigcatch.sig[0] & 0x7fffffffUL);
481 seq_put_decimal_ull(m, ' ', wchan);
482 seq_put_decimal_ull(m, ' ', 0);
483 seq_put_decimal_ull(m, ' ', 0);
484 seq_put_decimal_ll(m, ' ', task->exit_signal);
485 seq_put_decimal_ll(m, ' ', task_cpu(task));
486 seq_put_decimal_ull(m, ' ', task->rt_priority);
487 seq_put_decimal_ull(m, ' ', task->policy);
488 seq_put_decimal_ull(m, ' ', delayacct_blkio_ticks(task));
489 seq_put_decimal_ull(m, ' ', cputime_to_clock_t(gtime));
490 seq_put_decimal_ll(m, ' ', cputime_to_clock_t(cgtime));
492 if (mm && permitted) {
493 seq_put_decimal_ull(m, ' ', mm->start_data);
494 seq_put_decimal_ull(m, ' ', mm->end_data);
495 seq_put_decimal_ull(m, ' ', mm->start_brk);
496 seq_put_decimal_ull(m, ' ', mm->arg_start);
497 seq_put_decimal_ull(m, ' ', mm->arg_end);
498 seq_put_decimal_ull(m, ' ', mm->env_start);
499 seq_put_decimal_ull(m, ' ', mm->env_end);
501 seq_printf(m, " 0 0 0 0 0 0 0");
504 seq_put_decimal_ll(m, ' ', task->exit_code);
506 seq_put_decimal_ll(m, ' ', 0);
514 int proc_tid_stat(struct seq_file *m, struct pid_namespace *ns,
515 struct pid *pid, struct task_struct *task)
517 return do_task_stat(m, ns, pid, task, 0);
520 int proc_tgid_stat(struct seq_file *m, struct pid_namespace *ns,
521 struct pid *pid, struct task_struct *task)
523 return do_task_stat(m, ns, pid, task, 1);
526 int proc_pid_statm(struct seq_file *m, struct pid_namespace *ns,
527 struct pid *pid, struct task_struct *task)
529 unsigned long size = 0, resident = 0, shared = 0, text = 0, data = 0;
530 struct mm_struct *mm = get_task_mm(task);
533 size = task_statm(mm, &shared, &text, &data, &resident);
537 * For quick read, open code by putting numbers directly
539 * seq_printf(m, "%lu %lu %lu %lu 0 %lu 0\n",
540 * size, resident, shared, text, data);
542 seq_put_decimal_ull(m, 0, size);
543 seq_put_decimal_ull(m, ' ', resident);
544 seq_put_decimal_ull(m, ' ', shared);
545 seq_put_decimal_ull(m, ' ', text);
546 seq_put_decimal_ull(m, ' ', 0);
547 seq_put_decimal_ull(m, ' ', data);
548 seq_put_decimal_ull(m, ' ', 0);
554 #ifdef CONFIG_CHECKPOINT_RESTORE
556 get_children_pid(struct inode *inode, struct pid *pid_prev, loff_t pos)
558 struct task_struct *start, *task;
559 struct pid *pid = NULL;
561 read_lock(&tasklist_lock);
563 start = pid_task(proc_pid(inode), PIDTYPE_PID);
568 * Lets try to continue searching first, this gives
569 * us significant speedup on children-rich processes.
572 task = pid_task(pid_prev, PIDTYPE_PID);
573 if (task && task->real_parent == start &&
574 !(list_empty(&task->sibling))) {
575 if (list_is_last(&task->sibling, &start->children))
577 task = list_first_entry(&task->sibling,
578 struct task_struct, sibling);
579 pid = get_pid(task_pid(task));
587 * We might miss some children here if children
588 * are exited while we were not holding the lock,
589 * but it was never promised to be accurate that
592 * "Just suppose that the parent sleeps, but N children
593 * exit after we printed their tids. Now the slow paths
594 * skips N extra children, we miss N tasks." (c)
596 * So one need to stop or freeze the leader and all
597 * its children to get a precise result.
599 list_for_each_entry(task, &start->children, sibling) {
601 pid = get_pid(task_pid(task));
607 read_unlock(&tasklist_lock);
611 static int children_seq_show(struct seq_file *seq, void *v)
613 struct inode *inode = seq->private;
616 pid = pid_nr_ns(v, inode->i_sb->s_fs_info);
617 return seq_printf(seq, "%d ", pid);
620 static void *children_seq_start(struct seq_file *seq, loff_t *pos)
622 return get_children_pid(seq->private, NULL, *pos);
625 static void *children_seq_next(struct seq_file *seq, void *v, loff_t *pos)
629 pid = get_children_pid(seq->private, v, *pos + 1);
636 static void children_seq_stop(struct seq_file *seq, void *v)
641 static const struct seq_operations children_seq_ops = {
642 .start = children_seq_start,
643 .next = children_seq_next,
644 .stop = children_seq_stop,
645 .show = children_seq_show,
648 static int children_seq_open(struct inode *inode, struct file *file)
653 ret = seq_open(file, &children_seq_ops);
657 m = file->private_data;
663 int children_seq_release(struct inode *inode, struct file *file)
665 seq_release(inode, file);
669 const struct file_operations proc_tid_children_operations = {
670 .open = children_seq_open,
673 .release = children_seq_release,
675 #endif /* CONFIG_CHECKPOINT_RESTORE */