Loading...
1// SPDX-License-Identifier: GPL-2.0
2/*
3 * linux/fs/proc/array.c
4 *
5 * Copyright (C) 1992 by Linus Torvalds
6 * based on ideas by Darren Senn
7 *
8 * Fixes:
9 * Michael. K. Johnson: stat,statm extensions.
10 * <johnsonm@stolaf.edu>
11 *
12 * Pauline Middelink : Made cmdline,envline only break at '\0's, to
13 * make sure SET_PROCTITLE works. Also removed
14 * bad '!' which forced address recalculation for
15 * EVERY character on the current page.
16 * <middelin@polyware.iaf.nl>
17 *
18 * Danny ter Haar : added cpuinfo
19 * <dth@cistron.nl>
20 *
21 * Alessandro Rubini : profile extension.
22 * <rubini@ipvvis.unipv.it>
23 *
24 * Jeff Tranter : added BogoMips field to cpuinfo
25 * <Jeff_Tranter@Mitel.COM>
26 *
27 * Bruno Haible : remove 4K limit for the maps file
28 * <haible@ma2s2.mathematik.uni-karlsruhe.de>
29 *
30 * Yves Arrouye : remove removal of trailing spaces in get_array.
31 * <Yves.Arrouye@marin.fdn.fr>
32 *
33 * Jerome Forissier : added per-CPU time information to /proc/stat
34 * and /proc/<pid>/cpu extension
35 * <forissier@isia.cma.fr>
36 * - Incorporation and non-SMP safe operation
37 * of forissier patch in 2.1.78 by
38 * Hans Marcus <crowbar@concepts.nl>
39 *
40 * aeb@cwi.nl : /proc/partitions
41 *
42 *
43 * Alan Cox : security fixes.
44 * <alan@lxorguk.ukuu.org.uk>
45 *
46 * Al Viro : safe handling of mm_struct
47 *
48 * Gerhard Wichert : added BIGMEM support
49 * Siemens AG <Gerhard.Wichert@pdb.siemens.de>
50 *
51 * Al Viro & Jeff Garzik : moved most of the thing into base.c and
52 * : proc_misc.c. The rest may eventually go into
53 * : base.c too.
54 */
55
56#include <linux/types.h>
57#include <linux/errno.h>
58#include <linux/time.h>
59#include <linux/time_namespace.h>
60#include <linux/kernel.h>
61#include <linux/kernel_stat.h>
62#include <linux/tty.h>
63#include <linux/string.h>
64#include <linux/mman.h>
65#include <linux/sched/mm.h>
66#include <linux/sched/numa_balancing.h>
67#include <linux/sched/task_stack.h>
68#include <linux/sched/task.h>
69#include <linux/sched/cputime.h>
70#include <linux/proc_fs.h>
71#include <linux/ioport.h>
72#include <linux/io.h>
73#include <linux/mm.h>
74#include <linux/hugetlb.h>
75#include <linux/pagemap.h>
76#include <linux/swap.h>
77#include <linux/smp.h>
78#include <linux/signal.h>
79#include <linux/highmem.h>
80#include <linux/file.h>
81#include <linux/fdtable.h>
82#include <linux/times.h>
83#include <linux/cpuset.h>
84#include <linux/rcupdate.h>
85#include <linux/delayacct.h>
86#include <linux/seq_file.h>
87#include <linux/pid_namespace.h>
88#include <linux/prctl.h>
89#include <linux/ptrace.h>
90#include <linux/string_helpers.h>
91#include <linux/user_namespace.h>
92#include <linux/fs_struct.h>
93#include <linux/kthread.h>
94
95#include <asm/processor.h>
96#include "internal.h"
97
98void proc_task_name(struct seq_file *m, struct task_struct *p, bool escape)
99{
100 char tcomm[64];
101
102 /*
103 * Test before PF_KTHREAD because all workqueue worker threads are
104 * kernel threads.
105 */
106 if (p->flags & PF_WQ_WORKER)
107 wq_worker_comm(tcomm, sizeof(tcomm), p);
108 else if (p->flags & PF_KTHREAD)
109 get_kthread_comm(tcomm, sizeof(tcomm), p);
110 else
111 __get_task_comm(tcomm, sizeof(tcomm), p);
112
113 if (escape)
114 seq_escape_str(m, tcomm, ESCAPE_SPACE | ESCAPE_SPECIAL, "\n\\");
115 else
116 seq_printf(m, "%.64s", tcomm);
117}
118
119/*
120 * The task state array is a strange "bitmap" of
121 * reasons to sleep. Thus "running" is zero, and
122 * you can test for combinations of others with
123 * simple bit tests.
124 */
125static const char * const task_state_array[] = {
126
127 /* states in TASK_REPORT: */
128 "R (running)", /* 0x00 */
129 "S (sleeping)", /* 0x01 */
130 "D (disk sleep)", /* 0x02 */
131 "T (stopped)", /* 0x04 */
132 "t (tracing stop)", /* 0x08 */
133 "X (dead)", /* 0x10 */
134 "Z (zombie)", /* 0x20 */
135 "P (parked)", /* 0x40 */
136
137 /* states beyond TASK_REPORT: */
138 "I (idle)", /* 0x80 */
139};
140
141static inline const char *get_task_state(struct task_struct *tsk)
142{
143 BUILD_BUG_ON(1 + ilog2(TASK_REPORT_MAX) != ARRAY_SIZE(task_state_array));
144 return task_state_array[task_state_index(tsk)];
145}
146
147static inline void task_state(struct seq_file *m, struct pid_namespace *ns,
148 struct pid *pid, struct task_struct *p)
149{
150 struct user_namespace *user_ns = seq_user_ns(m);
151 struct group_info *group_info;
152 int g, umask = -1;
153 struct task_struct *tracer;
154 const struct cred *cred;
155 pid_t ppid, tpid = 0, tgid, ngid;
156 unsigned int max_fds = 0;
157
158 rcu_read_lock();
159 ppid = pid_alive(p) ?
160 task_tgid_nr_ns(rcu_dereference(p->real_parent), ns) : 0;
161
162 tracer = ptrace_parent(p);
163 if (tracer)
164 tpid = task_pid_nr_ns(tracer, ns);
165
166 tgid = task_tgid_nr_ns(p, ns);
167 ngid = task_numa_group_id(p);
168 cred = get_task_cred(p);
169
170 task_lock(p);
171 if (p->fs)
172 umask = p->fs->umask;
173 if (p->files)
174 max_fds = files_fdtable(p->files)->max_fds;
175 task_unlock(p);
176 rcu_read_unlock();
177
178 if (umask >= 0)
179 seq_printf(m, "Umask:\t%#04o\n", umask);
180 seq_puts(m, "State:\t");
181 seq_puts(m, get_task_state(p));
182
183 seq_put_decimal_ull(m, "\nTgid:\t", tgid);
184 seq_put_decimal_ull(m, "\nNgid:\t", ngid);
185 seq_put_decimal_ull(m, "\nPid:\t", pid_nr_ns(pid, ns));
186 seq_put_decimal_ull(m, "\nPPid:\t", ppid);
187 seq_put_decimal_ull(m, "\nTracerPid:\t", tpid);
188 seq_put_decimal_ull(m, "\nUid:\t", from_kuid_munged(user_ns, cred->uid));
189 seq_put_decimal_ull(m, "\t", from_kuid_munged(user_ns, cred->euid));
190 seq_put_decimal_ull(m, "\t", from_kuid_munged(user_ns, cred->suid));
191 seq_put_decimal_ull(m, "\t", from_kuid_munged(user_ns, cred->fsuid));
192 seq_put_decimal_ull(m, "\nGid:\t", from_kgid_munged(user_ns, cred->gid));
193 seq_put_decimal_ull(m, "\t", from_kgid_munged(user_ns, cred->egid));
194 seq_put_decimal_ull(m, "\t", from_kgid_munged(user_ns, cred->sgid));
195 seq_put_decimal_ull(m, "\t", from_kgid_munged(user_ns, cred->fsgid));
196 seq_put_decimal_ull(m, "\nFDSize:\t", max_fds);
197
198 seq_puts(m, "\nGroups:\t");
199 group_info = cred->group_info;
200 for (g = 0; g < group_info->ngroups; g++)
201 seq_put_decimal_ull(m, g ? " " : "",
202 from_kgid_munged(user_ns, group_info->gid[g]));
203 put_cred(cred);
204 /* Trailing space shouldn't have been added in the first place. */
205 seq_putc(m, ' ');
206
207#ifdef CONFIG_PID_NS
208 seq_puts(m, "\nNStgid:");
209 for (g = ns->level; g <= pid->level; g++)
210 seq_put_decimal_ull(m, "\t", task_tgid_nr_ns(p, pid->numbers[g].ns));
211 seq_puts(m, "\nNSpid:");
212 for (g = ns->level; g <= pid->level; g++)
213 seq_put_decimal_ull(m, "\t", task_pid_nr_ns(p, pid->numbers[g].ns));
214 seq_puts(m, "\nNSpgid:");
215 for (g = ns->level; g <= pid->level; g++)
216 seq_put_decimal_ull(m, "\t", task_pgrp_nr_ns(p, pid->numbers[g].ns));
217 seq_puts(m, "\nNSsid:");
218 for (g = ns->level; g <= pid->level; g++)
219 seq_put_decimal_ull(m, "\t", task_session_nr_ns(p, pid->numbers[g].ns));
220#endif
221 seq_putc(m, '\n');
222}
223
224void render_sigset_t(struct seq_file *m, const char *header,
225 sigset_t *set)
226{
227 int i;
228
229 seq_puts(m, header);
230
231 i = _NSIG;
232 do {
233 int x = 0;
234
235 i -= 4;
236 if (sigismember(set, i+1)) x |= 1;
237 if (sigismember(set, i+2)) x |= 2;
238 if (sigismember(set, i+3)) x |= 4;
239 if (sigismember(set, i+4)) x |= 8;
240 seq_putc(m, hex_asc[x]);
241 } while (i >= 4);
242
243 seq_putc(m, '\n');
244}
245
246static void collect_sigign_sigcatch(struct task_struct *p, sigset_t *sigign,
247 sigset_t *sigcatch)
248{
249 struct k_sigaction *k;
250 int i;
251
252 k = p->sighand->action;
253 for (i = 1; i <= _NSIG; ++i, ++k) {
254 if (k->sa.sa_handler == SIG_IGN)
255 sigaddset(sigign, i);
256 else if (k->sa.sa_handler != SIG_DFL)
257 sigaddset(sigcatch, i);
258 }
259}
260
261static inline void task_sig(struct seq_file *m, struct task_struct *p)
262{
263 unsigned long flags;
264 sigset_t pending, shpending, blocked, ignored, caught;
265 int num_threads = 0;
266 unsigned int qsize = 0;
267 unsigned long qlim = 0;
268
269 sigemptyset(&pending);
270 sigemptyset(&shpending);
271 sigemptyset(&blocked);
272 sigemptyset(&ignored);
273 sigemptyset(&caught);
274
275 if (lock_task_sighand(p, &flags)) {
276 pending = p->pending.signal;
277 shpending = p->signal->shared_pending.signal;
278 blocked = p->blocked;
279 collect_sigign_sigcatch(p, &ignored, &caught);
280 num_threads = get_nr_threads(p);
281 rcu_read_lock(); /* FIXME: is this correct? */
282 qsize = get_rlimit_value(task_ucounts(p), UCOUNT_RLIMIT_SIGPENDING);
283 rcu_read_unlock();
284 qlim = task_rlimit(p, RLIMIT_SIGPENDING);
285 unlock_task_sighand(p, &flags);
286 }
287
288 seq_put_decimal_ull(m, "Threads:\t", num_threads);
289 seq_put_decimal_ull(m, "\nSigQ:\t", qsize);
290 seq_put_decimal_ull(m, "/", qlim);
291
292 /* render them all */
293 render_sigset_t(m, "\nSigPnd:\t", &pending);
294 render_sigset_t(m, "ShdPnd:\t", &shpending);
295 render_sigset_t(m, "SigBlk:\t", &blocked);
296 render_sigset_t(m, "SigIgn:\t", &ignored);
297 render_sigset_t(m, "SigCgt:\t", &caught);
298}
299
300static void render_cap_t(struct seq_file *m, const char *header,
301 kernel_cap_t *a)
302{
303 unsigned __capi;
304
305 seq_puts(m, header);
306 CAP_FOR_EACH_U32(__capi) {
307 seq_put_hex_ll(m, NULL,
308 a->cap[CAP_LAST_U32 - __capi], 8);
309 }
310 seq_putc(m, '\n');
311}
312
313static inline void task_cap(struct seq_file *m, struct task_struct *p)
314{
315 const struct cred *cred;
316 kernel_cap_t cap_inheritable, cap_permitted, cap_effective,
317 cap_bset, cap_ambient;
318
319 rcu_read_lock();
320 cred = __task_cred(p);
321 cap_inheritable = cred->cap_inheritable;
322 cap_permitted = cred->cap_permitted;
323 cap_effective = cred->cap_effective;
324 cap_bset = cred->cap_bset;
325 cap_ambient = cred->cap_ambient;
326 rcu_read_unlock();
327
328 render_cap_t(m, "CapInh:\t", &cap_inheritable);
329 render_cap_t(m, "CapPrm:\t", &cap_permitted);
330 render_cap_t(m, "CapEff:\t", &cap_effective);
331 render_cap_t(m, "CapBnd:\t", &cap_bset);
332 render_cap_t(m, "CapAmb:\t", &cap_ambient);
333}
334
335static inline void task_seccomp(struct seq_file *m, struct task_struct *p)
336{
337 seq_put_decimal_ull(m, "NoNewPrivs:\t", task_no_new_privs(p));
338#ifdef CONFIG_SECCOMP
339 seq_put_decimal_ull(m, "\nSeccomp:\t", p->seccomp.mode);
340#ifdef CONFIG_SECCOMP_FILTER
341 seq_put_decimal_ull(m, "\nSeccomp_filters:\t",
342 atomic_read(&p->seccomp.filter_count));
343#endif
344#endif
345 seq_puts(m, "\nSpeculation_Store_Bypass:\t");
346 switch (arch_prctl_spec_ctrl_get(p, PR_SPEC_STORE_BYPASS)) {
347 case -EINVAL:
348 seq_puts(m, "unknown");
349 break;
350 case PR_SPEC_NOT_AFFECTED:
351 seq_puts(m, "not vulnerable");
352 break;
353 case PR_SPEC_PRCTL | PR_SPEC_FORCE_DISABLE:
354 seq_puts(m, "thread force mitigated");
355 break;
356 case PR_SPEC_PRCTL | PR_SPEC_DISABLE:
357 seq_puts(m, "thread mitigated");
358 break;
359 case PR_SPEC_PRCTL | PR_SPEC_ENABLE:
360 seq_puts(m, "thread vulnerable");
361 break;
362 case PR_SPEC_DISABLE:
363 seq_puts(m, "globally mitigated");
364 break;
365 default:
366 seq_puts(m, "vulnerable");
367 break;
368 }
369
370 seq_puts(m, "\nSpeculationIndirectBranch:\t");
371 switch (arch_prctl_spec_ctrl_get(p, PR_SPEC_INDIRECT_BRANCH)) {
372 case -EINVAL:
373 seq_puts(m, "unsupported");
374 break;
375 case PR_SPEC_NOT_AFFECTED:
376 seq_puts(m, "not affected");
377 break;
378 case PR_SPEC_PRCTL | PR_SPEC_FORCE_DISABLE:
379 seq_puts(m, "conditional force disabled");
380 break;
381 case PR_SPEC_PRCTL | PR_SPEC_DISABLE:
382 seq_puts(m, "conditional disabled");
383 break;
384 case PR_SPEC_PRCTL | PR_SPEC_ENABLE:
385 seq_puts(m, "conditional enabled");
386 break;
387 case PR_SPEC_ENABLE:
388 seq_puts(m, "always enabled");
389 break;
390 case PR_SPEC_DISABLE:
391 seq_puts(m, "always disabled");
392 break;
393 default:
394 seq_puts(m, "unknown");
395 break;
396 }
397 seq_putc(m, '\n');
398}
399
400static inline void task_context_switch_counts(struct seq_file *m,
401 struct task_struct *p)
402{
403 seq_put_decimal_ull(m, "voluntary_ctxt_switches:\t", p->nvcsw);
404 seq_put_decimal_ull(m, "\nnonvoluntary_ctxt_switches:\t", p->nivcsw);
405 seq_putc(m, '\n');
406}
407
408static void task_cpus_allowed(struct seq_file *m, struct task_struct *task)
409{
410 seq_printf(m, "Cpus_allowed:\t%*pb\n",
411 cpumask_pr_args(&task->cpus_mask));
412 seq_printf(m, "Cpus_allowed_list:\t%*pbl\n",
413 cpumask_pr_args(&task->cpus_mask));
414}
415
416static inline void task_core_dumping(struct seq_file *m, struct task_struct *task)
417{
418 seq_put_decimal_ull(m, "CoreDumping:\t", !!task->signal->core_state);
419 seq_putc(m, '\n');
420}
421
422static inline void task_thp_status(struct seq_file *m, struct mm_struct *mm)
423{
424 bool thp_enabled = IS_ENABLED(CONFIG_TRANSPARENT_HUGEPAGE);
425
426 if (thp_enabled)
427 thp_enabled = !test_bit(MMF_DISABLE_THP, &mm->flags);
428 seq_printf(m, "THP_enabled:\t%d\n", thp_enabled);
429}
430
431int proc_pid_status(struct seq_file *m, struct pid_namespace *ns,
432 struct pid *pid, struct task_struct *task)
433{
434 struct mm_struct *mm = get_task_mm(task);
435
436 seq_puts(m, "Name:\t");
437 proc_task_name(m, task, true);
438 seq_putc(m, '\n');
439
440 task_state(m, ns, pid, task);
441
442 if (mm) {
443 task_mem(m, mm);
444 task_core_dumping(m, task);
445 task_thp_status(m, mm);
446 mmput(mm);
447 }
448 task_sig(m, task);
449 task_cap(m, task);
450 task_seccomp(m, task);
451 task_cpus_allowed(m, task);
452 cpuset_task_status_allowed(m, task);
453 task_context_switch_counts(m, task);
454 return 0;
455}
456
457static int do_task_stat(struct seq_file *m, struct pid_namespace *ns,
458 struct pid *pid, struct task_struct *task, int whole)
459{
460 unsigned long vsize, eip, esp, wchan = 0;
461 int priority, nice;
462 int tty_pgrp = -1, tty_nr = 0;
463 sigset_t sigign, sigcatch;
464 char state;
465 pid_t ppid = 0, pgid = -1, sid = -1;
466 int num_threads = 0;
467 int permitted;
468 struct mm_struct *mm;
469 unsigned long long start_time;
470 unsigned long cmin_flt = 0, cmaj_flt = 0;
471 unsigned long min_flt = 0, maj_flt = 0;
472 u64 cutime, cstime, utime, stime;
473 u64 cgtime, gtime;
474 unsigned long rsslim = 0;
475 unsigned long flags;
476 int exit_code = task->exit_code;
477
478 state = *get_task_state(task);
479 vsize = eip = esp = 0;
480 permitted = ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS | PTRACE_MODE_NOAUDIT);
481 mm = get_task_mm(task);
482 if (mm) {
483 vsize = task_vsize(mm);
484 /*
485 * esp and eip are intentionally zeroed out. There is no
486 * non-racy way to read them without freezing the task.
487 * Programs that need reliable values can use ptrace(2).
488 *
489 * The only exception is if the task is core dumping because
490 * a program is not able to use ptrace(2) in that case. It is
491 * safe because the task has stopped executing permanently.
492 */
493 if (permitted && (task->flags & (PF_EXITING|PF_DUMPCORE))) {
494 if (try_get_task_stack(task)) {
495 eip = KSTK_EIP(task);
496 esp = KSTK_ESP(task);
497 put_task_stack(task);
498 }
499 }
500 }
501
502 sigemptyset(&sigign);
503 sigemptyset(&sigcatch);
504 cutime = cstime = utime = stime = 0;
505 cgtime = gtime = 0;
506
507 if (lock_task_sighand(task, &flags)) {
508 struct signal_struct *sig = task->signal;
509
510 if (sig->tty) {
511 struct pid *pgrp = tty_get_pgrp(sig->tty);
512 tty_pgrp = pid_nr_ns(pgrp, ns);
513 put_pid(pgrp);
514 tty_nr = new_encode_dev(tty_devnum(sig->tty));
515 }
516
517 num_threads = get_nr_threads(task);
518 collect_sigign_sigcatch(task, &sigign, &sigcatch);
519
520 cmin_flt = sig->cmin_flt;
521 cmaj_flt = sig->cmaj_flt;
522 cutime = sig->cutime;
523 cstime = sig->cstime;
524 cgtime = sig->cgtime;
525 rsslim = READ_ONCE(sig->rlim[RLIMIT_RSS].rlim_cur);
526
527 /* add up live thread stats at the group level */
528 if (whole) {
529 struct task_struct *t = task;
530 do {
531 min_flt += t->min_flt;
532 maj_flt += t->maj_flt;
533 gtime += task_gtime(t);
534 } while_each_thread(task, t);
535
536 min_flt += sig->min_flt;
537 maj_flt += sig->maj_flt;
538 thread_group_cputime_adjusted(task, &utime, &stime);
539 gtime += sig->gtime;
540
541 if (sig->flags & (SIGNAL_GROUP_EXIT | SIGNAL_STOP_STOPPED))
542 exit_code = sig->group_exit_code;
543 }
544
545 sid = task_session_nr_ns(task, ns);
546 ppid = task_tgid_nr_ns(task->real_parent, ns);
547 pgid = task_pgrp_nr_ns(task, ns);
548
549 unlock_task_sighand(task, &flags);
550 }
551
552 if (permitted && (!whole || num_threads < 2))
553 wchan = !task_is_running(task);
554 if (!whole) {
555 min_flt = task->min_flt;
556 maj_flt = task->maj_flt;
557 task_cputime_adjusted(task, &utime, &stime);
558 gtime = task_gtime(task);
559 }
560
561 /* scale priority and nice values from timeslices to -20..20 */
562 /* to make it look like a "normal" Unix priority/nice value */
563 priority = task_prio(task);
564 nice = task_nice(task);
565
566 /* apply timens offset for boottime and convert nsec -> ticks */
567 start_time =
568 nsec_to_clock_t(timens_add_boottime_ns(task->start_boottime));
569
570 seq_put_decimal_ull(m, "", pid_nr_ns(pid, ns));
571 seq_puts(m, " (");
572 proc_task_name(m, task, false);
573 seq_puts(m, ") ");
574 seq_putc(m, state);
575 seq_put_decimal_ll(m, " ", ppid);
576 seq_put_decimal_ll(m, " ", pgid);
577 seq_put_decimal_ll(m, " ", sid);
578 seq_put_decimal_ll(m, " ", tty_nr);
579 seq_put_decimal_ll(m, " ", tty_pgrp);
580 seq_put_decimal_ull(m, " ", task->flags);
581 seq_put_decimal_ull(m, " ", min_flt);
582 seq_put_decimal_ull(m, " ", cmin_flt);
583 seq_put_decimal_ull(m, " ", maj_flt);
584 seq_put_decimal_ull(m, " ", cmaj_flt);
585 seq_put_decimal_ull(m, " ", nsec_to_clock_t(utime));
586 seq_put_decimal_ull(m, " ", nsec_to_clock_t(stime));
587 seq_put_decimal_ll(m, " ", nsec_to_clock_t(cutime));
588 seq_put_decimal_ll(m, " ", nsec_to_clock_t(cstime));
589 seq_put_decimal_ll(m, " ", priority);
590 seq_put_decimal_ll(m, " ", nice);
591 seq_put_decimal_ll(m, " ", num_threads);
592 seq_put_decimal_ull(m, " ", 0);
593 seq_put_decimal_ull(m, " ", start_time);
594 seq_put_decimal_ull(m, " ", vsize);
595 seq_put_decimal_ull(m, " ", mm ? get_mm_rss(mm) : 0);
596 seq_put_decimal_ull(m, " ", rsslim);
597 seq_put_decimal_ull(m, " ", mm ? (permitted ? mm->start_code : 1) : 0);
598 seq_put_decimal_ull(m, " ", mm ? (permitted ? mm->end_code : 1) : 0);
599 seq_put_decimal_ull(m, " ", (permitted && mm) ? mm->start_stack : 0);
600 seq_put_decimal_ull(m, " ", esp);
601 seq_put_decimal_ull(m, " ", eip);
602 /* The signal information here is obsolete.
603 * It must be decimal for Linux 2.0 compatibility.
604 * Use /proc/#/status for real-time signals.
605 */
606 seq_put_decimal_ull(m, " ", task->pending.signal.sig[0] & 0x7fffffffUL);
607 seq_put_decimal_ull(m, " ", task->blocked.sig[0] & 0x7fffffffUL);
608 seq_put_decimal_ull(m, " ", sigign.sig[0] & 0x7fffffffUL);
609 seq_put_decimal_ull(m, " ", sigcatch.sig[0] & 0x7fffffffUL);
610
611 /*
612 * We used to output the absolute kernel address, but that's an
613 * information leak - so instead we show a 0/1 flag here, to signal
614 * to user-space whether there's a wchan field in /proc/PID/wchan.
615 *
616 * This works with older implementations of procps as well.
617 */
618 seq_put_decimal_ull(m, " ", wchan);
619
620 seq_put_decimal_ull(m, " ", 0);
621 seq_put_decimal_ull(m, " ", 0);
622 seq_put_decimal_ll(m, " ", task->exit_signal);
623 seq_put_decimal_ll(m, " ", task_cpu(task));
624 seq_put_decimal_ull(m, " ", task->rt_priority);
625 seq_put_decimal_ull(m, " ", task->policy);
626 seq_put_decimal_ull(m, " ", delayacct_blkio_ticks(task));
627 seq_put_decimal_ull(m, " ", nsec_to_clock_t(gtime));
628 seq_put_decimal_ll(m, " ", nsec_to_clock_t(cgtime));
629
630 if (mm && permitted) {
631 seq_put_decimal_ull(m, " ", mm->start_data);
632 seq_put_decimal_ull(m, " ", mm->end_data);
633 seq_put_decimal_ull(m, " ", mm->start_brk);
634 seq_put_decimal_ull(m, " ", mm->arg_start);
635 seq_put_decimal_ull(m, " ", mm->arg_end);
636 seq_put_decimal_ull(m, " ", mm->env_start);
637 seq_put_decimal_ull(m, " ", mm->env_end);
638 } else
639 seq_puts(m, " 0 0 0 0 0 0 0");
640
641 if (permitted)
642 seq_put_decimal_ll(m, " ", exit_code);
643 else
644 seq_puts(m, " 0");
645
646 seq_putc(m, '\n');
647 if (mm)
648 mmput(mm);
649 return 0;
650}
651
652int proc_tid_stat(struct seq_file *m, struct pid_namespace *ns,
653 struct pid *pid, struct task_struct *task)
654{
655 return do_task_stat(m, ns, pid, task, 0);
656}
657
658int proc_tgid_stat(struct seq_file *m, struct pid_namespace *ns,
659 struct pid *pid, struct task_struct *task)
660{
661 return do_task_stat(m, ns, pid, task, 1);
662}
663
664int proc_pid_statm(struct seq_file *m, struct pid_namespace *ns,
665 struct pid *pid, struct task_struct *task)
666{
667 struct mm_struct *mm = get_task_mm(task);
668
669 if (mm) {
670 unsigned long size;
671 unsigned long resident = 0;
672 unsigned long shared = 0;
673 unsigned long text = 0;
674 unsigned long data = 0;
675
676 size = task_statm(mm, &shared, &text, &data, &resident);
677 mmput(mm);
678
679 /*
680 * For quick read, open code by putting numbers directly
681 * expected format is
682 * seq_printf(m, "%lu %lu %lu %lu 0 %lu 0\n",
683 * size, resident, shared, text, data);
684 */
685 seq_put_decimal_ull(m, "", size);
686 seq_put_decimal_ull(m, " ", resident);
687 seq_put_decimal_ull(m, " ", shared);
688 seq_put_decimal_ull(m, " ", text);
689 seq_put_decimal_ull(m, " ", 0);
690 seq_put_decimal_ull(m, " ", data);
691 seq_put_decimal_ull(m, " ", 0);
692 seq_putc(m, '\n');
693 } else {
694 seq_write(m, "0 0 0 0 0 0 0\n", 14);
695 }
696 return 0;
697}
698
699#ifdef CONFIG_PROC_CHILDREN
700static struct pid *
701get_children_pid(struct inode *inode, struct pid *pid_prev, loff_t pos)
702{
703 struct task_struct *start, *task;
704 struct pid *pid = NULL;
705
706 read_lock(&tasklist_lock);
707
708 start = pid_task(proc_pid(inode), PIDTYPE_PID);
709 if (!start)
710 goto out;
711
712 /*
713 * Lets try to continue searching first, this gives
714 * us significant speedup on children-rich processes.
715 */
716 if (pid_prev) {
717 task = pid_task(pid_prev, PIDTYPE_PID);
718 if (task && task->real_parent == start &&
719 !(list_empty(&task->sibling))) {
720 if (list_is_last(&task->sibling, &start->children))
721 goto out;
722 task = list_first_entry(&task->sibling,
723 struct task_struct, sibling);
724 pid = get_pid(task_pid(task));
725 goto out;
726 }
727 }
728
729 /*
730 * Slow search case.
731 *
732 * We might miss some children here if children
733 * are exited while we were not holding the lock,
734 * but it was never promised to be accurate that
735 * much.
736 *
737 * "Just suppose that the parent sleeps, but N children
738 * exit after we printed their tids. Now the slow paths
739 * skips N extra children, we miss N tasks." (c)
740 *
741 * So one need to stop or freeze the leader and all
742 * its children to get a precise result.
743 */
744 list_for_each_entry(task, &start->children, sibling) {
745 if (pos-- == 0) {
746 pid = get_pid(task_pid(task));
747 break;
748 }
749 }
750
751out:
752 read_unlock(&tasklist_lock);
753 return pid;
754}
755
756static int children_seq_show(struct seq_file *seq, void *v)
757{
758 struct inode *inode = file_inode(seq->file);
759
760 seq_printf(seq, "%d ", pid_nr_ns(v, proc_pid_ns(inode->i_sb)));
761 return 0;
762}
763
764static void *children_seq_start(struct seq_file *seq, loff_t *pos)
765{
766 return get_children_pid(file_inode(seq->file), NULL, *pos);
767}
768
769static void *children_seq_next(struct seq_file *seq, void *v, loff_t *pos)
770{
771 struct pid *pid;
772
773 pid = get_children_pid(file_inode(seq->file), v, *pos + 1);
774 put_pid(v);
775
776 ++*pos;
777 return pid;
778}
779
780static void children_seq_stop(struct seq_file *seq, void *v)
781{
782 put_pid(v);
783}
784
785static const struct seq_operations children_seq_ops = {
786 .start = children_seq_start,
787 .next = children_seq_next,
788 .stop = children_seq_stop,
789 .show = children_seq_show,
790};
791
792static int children_seq_open(struct inode *inode, struct file *file)
793{
794 return seq_open(file, &children_seq_ops);
795}
796
797const struct file_operations proc_tid_children_operations = {
798 .open = children_seq_open,
799 .read = seq_read,
800 .llseek = seq_lseek,
801 .release = seq_release,
802};
803#endif /* CONFIG_PROC_CHILDREN */
1// SPDX-License-Identifier: GPL-2.0
2/*
3 * linux/fs/proc/array.c
4 *
5 * Copyright (C) 1992 by Linus Torvalds
6 * based on ideas by Darren Senn
7 *
8 * Fixes:
9 * Michael. K. Johnson: stat,statm extensions.
10 * <johnsonm@stolaf.edu>
11 *
12 * Pauline Middelink : Made cmdline,envline only break at '\0's, to
13 * make sure SET_PROCTITLE works. Also removed
14 * bad '!' which forced address recalculation for
15 * EVERY character on the current page.
16 * <middelin@polyware.iaf.nl>
17 *
18 * Danny ter Haar : added cpuinfo
19 * <dth@cistron.nl>
20 *
21 * Alessandro Rubini : profile extension.
22 * <rubini@ipvvis.unipv.it>
23 *
24 * Jeff Tranter : added BogoMips field to cpuinfo
25 * <Jeff_Tranter@Mitel.COM>
26 *
27 * Bruno Haible : remove 4K limit for the maps file
28 * <haible@ma2s2.mathematik.uni-karlsruhe.de>
29 *
30 * Yves Arrouye : remove removal of trailing spaces in get_array.
31 * <Yves.Arrouye@marin.fdn.fr>
32 *
33 * Jerome Forissier : added per-CPU time information to /proc/stat
34 * and /proc/<pid>/cpu extension
35 * <forissier@isia.cma.fr>
36 * - Incorporation and non-SMP safe operation
37 * of forissier patch in 2.1.78 by
38 * Hans Marcus <crowbar@concepts.nl>
39 *
40 * aeb@cwi.nl : /proc/partitions
41 *
42 *
43 * Alan Cox : security fixes.
44 * <alan@lxorguk.ukuu.org.uk>
45 *
46 * Al Viro : safe handling of mm_struct
47 *
48 * Gerhard Wichert : added BIGMEM support
49 * Siemens AG <Gerhard.Wichert@pdb.siemens.de>
50 *
51 * Al Viro & Jeff Garzik : moved most of the thing into base.c and
52 * : proc_misc.c. The rest may eventually go into
53 * : base.c too.
54 */
55
56#include <linux/types.h>
57#include <linux/errno.h>
58#include <linux/time.h>
59#include <linux/time_namespace.h>
60#include <linux/kernel.h>
61#include <linux/kernel_stat.h>
62#include <linux/tty.h>
63#include <linux/string.h>
64#include <linux/mman.h>
65#include <linux/sched/mm.h>
66#include <linux/sched/numa_balancing.h>
67#include <linux/sched/task_stack.h>
68#include <linux/sched/task.h>
69#include <linux/sched/cputime.h>
70#include <linux/proc_fs.h>
71#include <linux/ioport.h>
72#include <linux/uaccess.h>
73#include <linux/io.h>
74#include <linux/mm.h>
75#include <linux/hugetlb.h>
76#include <linux/pagemap.h>
77#include <linux/swap.h>
78#include <linux/smp.h>
79#include <linux/signal.h>
80#include <linux/highmem.h>
81#include <linux/file.h>
82#include <linux/fdtable.h>
83#include <linux/times.h>
84#include <linux/cpuset.h>
85#include <linux/rcupdate.h>
86#include <linux/delayacct.h>
87#include <linux/seq_file.h>
88#include <linux/pid_namespace.h>
89#include <linux/prctl.h>
90#include <linux/ptrace.h>
91#include <linux/tracehook.h>
92#include <linux/string_helpers.h>
93#include <linux/user_namespace.h>
94#include <linux/fs_struct.h>
95
96#include <asm/processor.h>
97#include "internal.h"
98
99void proc_task_name(struct seq_file *m, struct task_struct *p, bool escape)
100{
101 char *buf;
102 size_t size;
103 char tcomm[64];
104 int ret;
105
106 if (p->flags & PF_WQ_WORKER)
107 wq_worker_comm(tcomm, sizeof(tcomm), p);
108 else
109 __get_task_comm(tcomm, sizeof(tcomm), p);
110
111 size = seq_get_buf(m, &buf);
112 if (escape) {
113 ret = string_escape_str(tcomm, buf, size,
114 ESCAPE_SPACE | ESCAPE_SPECIAL, "\n\\");
115 if (ret >= size)
116 ret = -1;
117 } else {
118 ret = strscpy(buf, tcomm, size);
119 }
120
121 seq_commit(m, ret);
122}
123
124/*
125 * The task state array is a strange "bitmap" of
126 * reasons to sleep. Thus "running" is zero, and
127 * you can test for combinations of others with
128 * simple bit tests.
129 */
130static const char * const task_state_array[] = {
131
132 /* states in TASK_REPORT: */
133 "R (running)", /* 0x00 */
134 "S (sleeping)", /* 0x01 */
135 "D (disk sleep)", /* 0x02 */
136 "T (stopped)", /* 0x04 */
137 "t (tracing stop)", /* 0x08 */
138 "X (dead)", /* 0x10 */
139 "Z (zombie)", /* 0x20 */
140 "P (parked)", /* 0x40 */
141
142 /* states beyond TASK_REPORT: */
143 "I (idle)", /* 0x80 */
144};
145
146static inline const char *get_task_state(struct task_struct *tsk)
147{
148 BUILD_BUG_ON(1 + ilog2(TASK_REPORT_MAX) != ARRAY_SIZE(task_state_array));
149 return task_state_array[task_state_index(tsk)];
150}
151
152static inline void task_state(struct seq_file *m, struct pid_namespace *ns,
153 struct pid *pid, struct task_struct *p)
154{
155 struct user_namespace *user_ns = seq_user_ns(m);
156 struct group_info *group_info;
157 int g, umask = -1;
158 struct task_struct *tracer;
159 const struct cred *cred;
160 pid_t ppid, tpid = 0, tgid, ngid;
161 unsigned int max_fds = 0;
162
163 rcu_read_lock();
164 ppid = pid_alive(p) ?
165 task_tgid_nr_ns(rcu_dereference(p->real_parent), ns) : 0;
166
167 tracer = ptrace_parent(p);
168 if (tracer)
169 tpid = task_pid_nr_ns(tracer, ns);
170
171 tgid = task_tgid_nr_ns(p, ns);
172 ngid = task_numa_group_id(p);
173 cred = get_task_cred(p);
174
175 task_lock(p);
176 if (p->fs)
177 umask = p->fs->umask;
178 if (p->files)
179 max_fds = files_fdtable(p->files)->max_fds;
180 task_unlock(p);
181 rcu_read_unlock();
182
183 if (umask >= 0)
184 seq_printf(m, "Umask:\t%#04o\n", umask);
185 seq_puts(m, "State:\t");
186 seq_puts(m, get_task_state(p));
187
188 seq_put_decimal_ull(m, "\nTgid:\t", tgid);
189 seq_put_decimal_ull(m, "\nNgid:\t", ngid);
190 seq_put_decimal_ull(m, "\nPid:\t", pid_nr_ns(pid, ns));
191 seq_put_decimal_ull(m, "\nPPid:\t", ppid);
192 seq_put_decimal_ull(m, "\nTracerPid:\t", tpid);
193 seq_put_decimal_ull(m, "\nUid:\t", from_kuid_munged(user_ns, cred->uid));
194 seq_put_decimal_ull(m, "\t", from_kuid_munged(user_ns, cred->euid));
195 seq_put_decimal_ull(m, "\t", from_kuid_munged(user_ns, cred->suid));
196 seq_put_decimal_ull(m, "\t", from_kuid_munged(user_ns, cred->fsuid));
197 seq_put_decimal_ull(m, "\nGid:\t", from_kgid_munged(user_ns, cred->gid));
198 seq_put_decimal_ull(m, "\t", from_kgid_munged(user_ns, cred->egid));
199 seq_put_decimal_ull(m, "\t", from_kgid_munged(user_ns, cred->sgid));
200 seq_put_decimal_ull(m, "\t", from_kgid_munged(user_ns, cred->fsgid));
201 seq_put_decimal_ull(m, "\nFDSize:\t", max_fds);
202
203 seq_puts(m, "\nGroups:\t");
204 group_info = cred->group_info;
205 for (g = 0; g < group_info->ngroups; g++)
206 seq_put_decimal_ull(m, g ? " " : "",
207 from_kgid_munged(user_ns, group_info->gid[g]));
208 put_cred(cred);
209 /* Trailing space shouldn't have been added in the first place. */
210 seq_putc(m, ' ');
211
212#ifdef CONFIG_PID_NS
213 seq_puts(m, "\nNStgid:");
214 for (g = ns->level; g <= pid->level; g++)
215 seq_put_decimal_ull(m, "\t", task_tgid_nr_ns(p, pid->numbers[g].ns));
216 seq_puts(m, "\nNSpid:");
217 for (g = ns->level; g <= pid->level; g++)
218 seq_put_decimal_ull(m, "\t", task_pid_nr_ns(p, pid->numbers[g].ns));
219 seq_puts(m, "\nNSpgid:");
220 for (g = ns->level; g <= pid->level; g++)
221 seq_put_decimal_ull(m, "\t", task_pgrp_nr_ns(p, pid->numbers[g].ns));
222 seq_puts(m, "\nNSsid:");
223 for (g = ns->level; g <= pid->level; g++)
224 seq_put_decimal_ull(m, "\t", task_session_nr_ns(p, pid->numbers[g].ns));
225#endif
226 seq_putc(m, '\n');
227}
228
229void render_sigset_t(struct seq_file *m, const char *header,
230 sigset_t *set)
231{
232 int i;
233
234 seq_puts(m, header);
235
236 i = _NSIG;
237 do {
238 int x = 0;
239
240 i -= 4;
241 if (sigismember(set, i+1)) x |= 1;
242 if (sigismember(set, i+2)) x |= 2;
243 if (sigismember(set, i+3)) x |= 4;
244 if (sigismember(set, i+4)) x |= 8;
245 seq_putc(m, hex_asc[x]);
246 } while (i >= 4);
247
248 seq_putc(m, '\n');
249}
250
251static void collect_sigign_sigcatch(struct task_struct *p, sigset_t *sigign,
252 sigset_t *sigcatch)
253{
254 struct k_sigaction *k;
255 int i;
256
257 k = p->sighand->action;
258 for (i = 1; i <= _NSIG; ++i, ++k) {
259 if (k->sa.sa_handler == SIG_IGN)
260 sigaddset(sigign, i);
261 else if (k->sa.sa_handler != SIG_DFL)
262 sigaddset(sigcatch, i);
263 }
264}
265
266static inline void task_sig(struct seq_file *m, struct task_struct *p)
267{
268 unsigned long flags;
269 sigset_t pending, shpending, blocked, ignored, caught;
270 int num_threads = 0;
271 unsigned int qsize = 0;
272 unsigned long qlim = 0;
273
274 sigemptyset(&pending);
275 sigemptyset(&shpending);
276 sigemptyset(&blocked);
277 sigemptyset(&ignored);
278 sigemptyset(&caught);
279
280 if (lock_task_sighand(p, &flags)) {
281 pending = p->pending.signal;
282 shpending = p->signal->shared_pending.signal;
283 blocked = p->blocked;
284 collect_sigign_sigcatch(p, &ignored, &caught);
285 num_threads = get_nr_threads(p);
286 rcu_read_lock(); /* FIXME: is this correct? */
287 qsize = get_ucounts_value(task_ucounts(p), UCOUNT_RLIMIT_SIGPENDING);
288 rcu_read_unlock();
289 qlim = task_rlimit(p, RLIMIT_SIGPENDING);
290 unlock_task_sighand(p, &flags);
291 }
292
293 seq_put_decimal_ull(m, "Threads:\t", num_threads);
294 seq_put_decimal_ull(m, "\nSigQ:\t", qsize);
295 seq_put_decimal_ull(m, "/", qlim);
296
297 /* render them all */
298 render_sigset_t(m, "\nSigPnd:\t", &pending);
299 render_sigset_t(m, "ShdPnd:\t", &shpending);
300 render_sigset_t(m, "SigBlk:\t", &blocked);
301 render_sigset_t(m, "SigIgn:\t", &ignored);
302 render_sigset_t(m, "SigCgt:\t", &caught);
303}
304
305static void render_cap_t(struct seq_file *m, const char *header,
306 kernel_cap_t *a)
307{
308 unsigned __capi;
309
310 seq_puts(m, header);
311 CAP_FOR_EACH_U32(__capi) {
312 seq_put_hex_ll(m, NULL,
313 a->cap[CAP_LAST_U32 - __capi], 8);
314 }
315 seq_putc(m, '\n');
316}
317
318static inline void task_cap(struct seq_file *m, struct task_struct *p)
319{
320 const struct cred *cred;
321 kernel_cap_t cap_inheritable, cap_permitted, cap_effective,
322 cap_bset, cap_ambient;
323
324 rcu_read_lock();
325 cred = __task_cred(p);
326 cap_inheritable = cred->cap_inheritable;
327 cap_permitted = cred->cap_permitted;
328 cap_effective = cred->cap_effective;
329 cap_bset = cred->cap_bset;
330 cap_ambient = cred->cap_ambient;
331 rcu_read_unlock();
332
333 render_cap_t(m, "CapInh:\t", &cap_inheritable);
334 render_cap_t(m, "CapPrm:\t", &cap_permitted);
335 render_cap_t(m, "CapEff:\t", &cap_effective);
336 render_cap_t(m, "CapBnd:\t", &cap_bset);
337 render_cap_t(m, "CapAmb:\t", &cap_ambient);
338}
339
340static inline void task_seccomp(struct seq_file *m, struct task_struct *p)
341{
342 seq_put_decimal_ull(m, "NoNewPrivs:\t", task_no_new_privs(p));
343#ifdef CONFIG_SECCOMP
344 seq_put_decimal_ull(m, "\nSeccomp:\t", p->seccomp.mode);
345#ifdef CONFIG_SECCOMP_FILTER
346 seq_put_decimal_ull(m, "\nSeccomp_filters:\t",
347 atomic_read(&p->seccomp.filter_count));
348#endif
349#endif
350 seq_puts(m, "\nSpeculation_Store_Bypass:\t");
351 switch (arch_prctl_spec_ctrl_get(p, PR_SPEC_STORE_BYPASS)) {
352 case -EINVAL:
353 seq_puts(m, "unknown");
354 break;
355 case PR_SPEC_NOT_AFFECTED:
356 seq_puts(m, "not vulnerable");
357 break;
358 case PR_SPEC_PRCTL | PR_SPEC_FORCE_DISABLE:
359 seq_puts(m, "thread force mitigated");
360 break;
361 case PR_SPEC_PRCTL | PR_SPEC_DISABLE:
362 seq_puts(m, "thread mitigated");
363 break;
364 case PR_SPEC_PRCTL | PR_SPEC_ENABLE:
365 seq_puts(m, "thread vulnerable");
366 break;
367 case PR_SPEC_DISABLE:
368 seq_puts(m, "globally mitigated");
369 break;
370 default:
371 seq_puts(m, "vulnerable");
372 break;
373 }
374
375 seq_puts(m, "\nSpeculationIndirectBranch:\t");
376 switch (arch_prctl_spec_ctrl_get(p, PR_SPEC_INDIRECT_BRANCH)) {
377 case -EINVAL:
378 seq_puts(m, "unsupported");
379 break;
380 case PR_SPEC_NOT_AFFECTED:
381 seq_puts(m, "not affected");
382 break;
383 case PR_SPEC_PRCTL | PR_SPEC_FORCE_DISABLE:
384 seq_puts(m, "conditional force disabled");
385 break;
386 case PR_SPEC_PRCTL | PR_SPEC_DISABLE:
387 seq_puts(m, "conditional disabled");
388 break;
389 case PR_SPEC_PRCTL | PR_SPEC_ENABLE:
390 seq_puts(m, "conditional enabled");
391 break;
392 case PR_SPEC_ENABLE:
393 seq_puts(m, "always enabled");
394 break;
395 case PR_SPEC_DISABLE:
396 seq_puts(m, "always disabled");
397 break;
398 default:
399 seq_puts(m, "unknown");
400 break;
401 }
402 seq_putc(m, '\n');
403}
404
405static inline void task_context_switch_counts(struct seq_file *m,
406 struct task_struct *p)
407{
408 seq_put_decimal_ull(m, "voluntary_ctxt_switches:\t", p->nvcsw);
409 seq_put_decimal_ull(m, "\nnonvoluntary_ctxt_switches:\t", p->nivcsw);
410 seq_putc(m, '\n');
411}
412
413static void task_cpus_allowed(struct seq_file *m, struct task_struct *task)
414{
415 seq_printf(m, "Cpus_allowed:\t%*pb\n",
416 cpumask_pr_args(&task->cpus_mask));
417 seq_printf(m, "Cpus_allowed_list:\t%*pbl\n",
418 cpumask_pr_args(&task->cpus_mask));
419}
420
421static inline void task_core_dumping(struct seq_file *m, struct mm_struct *mm)
422{
423 seq_put_decimal_ull(m, "CoreDumping:\t", !!mm->core_state);
424 seq_putc(m, '\n');
425}
426
427static inline void task_thp_status(struct seq_file *m, struct mm_struct *mm)
428{
429 bool thp_enabled = IS_ENABLED(CONFIG_TRANSPARENT_HUGEPAGE);
430
431 if (thp_enabled)
432 thp_enabled = !test_bit(MMF_DISABLE_THP, &mm->flags);
433 seq_printf(m, "THP_enabled:\t%d\n", thp_enabled);
434}
435
436int proc_pid_status(struct seq_file *m, struct pid_namespace *ns,
437 struct pid *pid, struct task_struct *task)
438{
439 struct mm_struct *mm = get_task_mm(task);
440
441 seq_puts(m, "Name:\t");
442 proc_task_name(m, task, true);
443 seq_putc(m, '\n');
444
445 task_state(m, ns, pid, task);
446
447 if (mm) {
448 task_mem(m, mm);
449 task_core_dumping(m, mm);
450 task_thp_status(m, mm);
451 mmput(mm);
452 }
453 task_sig(m, task);
454 task_cap(m, task);
455 task_seccomp(m, task);
456 task_cpus_allowed(m, task);
457 cpuset_task_status_allowed(m, task);
458 task_context_switch_counts(m, task);
459 return 0;
460}
461
462static int do_task_stat(struct seq_file *m, struct pid_namespace *ns,
463 struct pid *pid, struct task_struct *task, int whole)
464{
465 unsigned long vsize, eip, esp, wchan = 0;
466 int priority, nice;
467 int tty_pgrp = -1, tty_nr = 0;
468 sigset_t sigign, sigcatch;
469 char state;
470 pid_t ppid = 0, pgid = -1, sid = -1;
471 int num_threads = 0;
472 int permitted;
473 struct mm_struct *mm;
474 unsigned long long start_time;
475 unsigned long cmin_flt = 0, cmaj_flt = 0;
476 unsigned long min_flt = 0, maj_flt = 0;
477 u64 cutime, cstime, utime, stime;
478 u64 cgtime, gtime;
479 unsigned long rsslim = 0;
480 unsigned long flags;
481
482 state = *get_task_state(task);
483 vsize = eip = esp = 0;
484 permitted = ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS | PTRACE_MODE_NOAUDIT);
485 mm = get_task_mm(task);
486 if (mm) {
487 vsize = task_vsize(mm);
488 /*
489 * esp and eip are intentionally zeroed out. There is no
490 * non-racy way to read them without freezing the task.
491 * Programs that need reliable values can use ptrace(2).
492 *
493 * The only exception is if the task is core dumping because
494 * a program is not able to use ptrace(2) in that case. It is
495 * safe because the task has stopped executing permanently.
496 */
497 if (permitted && (task->flags & (PF_EXITING|PF_DUMPCORE))) {
498 if (try_get_task_stack(task)) {
499 eip = KSTK_EIP(task);
500 esp = KSTK_ESP(task);
501 put_task_stack(task);
502 }
503 }
504 }
505
506 sigemptyset(&sigign);
507 sigemptyset(&sigcatch);
508 cutime = cstime = utime = stime = 0;
509 cgtime = gtime = 0;
510
511 if (lock_task_sighand(task, &flags)) {
512 struct signal_struct *sig = task->signal;
513
514 if (sig->tty) {
515 struct pid *pgrp = tty_get_pgrp(sig->tty);
516 tty_pgrp = pid_nr_ns(pgrp, ns);
517 put_pid(pgrp);
518 tty_nr = new_encode_dev(tty_devnum(sig->tty));
519 }
520
521 num_threads = get_nr_threads(task);
522 collect_sigign_sigcatch(task, &sigign, &sigcatch);
523
524 cmin_flt = sig->cmin_flt;
525 cmaj_flt = sig->cmaj_flt;
526 cutime = sig->cutime;
527 cstime = sig->cstime;
528 cgtime = sig->cgtime;
529 rsslim = READ_ONCE(sig->rlim[RLIMIT_RSS].rlim_cur);
530
531 /* add up live thread stats at the group level */
532 if (whole) {
533 struct task_struct *t = task;
534 do {
535 min_flt += t->min_flt;
536 maj_flt += t->maj_flt;
537 gtime += task_gtime(t);
538 } while_each_thread(task, t);
539
540 min_flt += sig->min_flt;
541 maj_flt += sig->maj_flt;
542 thread_group_cputime_adjusted(task, &utime, &stime);
543 gtime += sig->gtime;
544 }
545
546 sid = task_session_nr_ns(task, ns);
547 ppid = task_tgid_nr_ns(task->real_parent, ns);
548 pgid = task_pgrp_nr_ns(task, ns);
549
550 unlock_task_sighand(task, &flags);
551 }
552
553 if (permitted && (!whole || num_threads < 2))
554 wchan = get_wchan(task);
555 if (!whole) {
556 min_flt = task->min_flt;
557 maj_flt = task->maj_flt;
558 task_cputime_adjusted(task, &utime, &stime);
559 gtime = task_gtime(task);
560 }
561
562 /* scale priority and nice values from timeslices to -20..20 */
563 /* to make it look like a "normal" Unix priority/nice value */
564 priority = task_prio(task);
565 nice = task_nice(task);
566
567 /* apply timens offset for boottime and convert nsec -> ticks */
568 start_time =
569 nsec_to_clock_t(timens_add_boottime_ns(task->start_boottime));
570
571 seq_put_decimal_ull(m, "", pid_nr_ns(pid, ns));
572 seq_puts(m, " (");
573 proc_task_name(m, task, false);
574 seq_puts(m, ") ");
575 seq_putc(m, state);
576 seq_put_decimal_ll(m, " ", ppid);
577 seq_put_decimal_ll(m, " ", pgid);
578 seq_put_decimal_ll(m, " ", sid);
579 seq_put_decimal_ll(m, " ", tty_nr);
580 seq_put_decimal_ll(m, " ", tty_pgrp);
581 seq_put_decimal_ull(m, " ", task->flags);
582 seq_put_decimal_ull(m, " ", min_flt);
583 seq_put_decimal_ull(m, " ", cmin_flt);
584 seq_put_decimal_ull(m, " ", maj_flt);
585 seq_put_decimal_ull(m, " ", cmaj_flt);
586 seq_put_decimal_ull(m, " ", nsec_to_clock_t(utime));
587 seq_put_decimal_ull(m, " ", nsec_to_clock_t(stime));
588 seq_put_decimal_ll(m, " ", nsec_to_clock_t(cutime));
589 seq_put_decimal_ll(m, " ", nsec_to_clock_t(cstime));
590 seq_put_decimal_ll(m, " ", priority);
591 seq_put_decimal_ll(m, " ", nice);
592 seq_put_decimal_ll(m, " ", num_threads);
593 seq_put_decimal_ull(m, " ", 0);
594 seq_put_decimal_ull(m, " ", start_time);
595 seq_put_decimal_ull(m, " ", vsize);
596 seq_put_decimal_ull(m, " ", mm ? get_mm_rss(mm) : 0);
597 seq_put_decimal_ull(m, " ", rsslim);
598 seq_put_decimal_ull(m, " ", mm ? (permitted ? mm->start_code : 1) : 0);
599 seq_put_decimal_ull(m, " ", mm ? (permitted ? mm->end_code : 1) : 0);
600 seq_put_decimal_ull(m, " ", (permitted && mm) ? mm->start_stack : 0);
601 seq_put_decimal_ull(m, " ", esp);
602 seq_put_decimal_ull(m, " ", eip);
603 /* The signal information here is obsolete.
604 * It must be decimal for Linux 2.0 compatibility.
605 * Use /proc/#/status for real-time signals.
606 */
607 seq_put_decimal_ull(m, " ", task->pending.signal.sig[0] & 0x7fffffffUL);
608 seq_put_decimal_ull(m, " ", task->blocked.sig[0] & 0x7fffffffUL);
609 seq_put_decimal_ull(m, " ", sigign.sig[0] & 0x7fffffffUL);
610 seq_put_decimal_ull(m, " ", sigcatch.sig[0] & 0x7fffffffUL);
611
612 /*
613 * We used to output the absolute kernel address, but that's an
614 * information leak - so instead we show a 0/1 flag here, to signal
615 * to user-space whether there's a wchan field in /proc/PID/wchan.
616 *
617 * This works with older implementations of procps as well.
618 */
619 if (wchan)
620 seq_puts(m, " 1");
621 else
622 seq_puts(m, " 0");
623
624 seq_put_decimal_ull(m, " ", 0);
625 seq_put_decimal_ull(m, " ", 0);
626 seq_put_decimal_ll(m, " ", task->exit_signal);
627 seq_put_decimal_ll(m, " ", task_cpu(task));
628 seq_put_decimal_ull(m, " ", task->rt_priority);
629 seq_put_decimal_ull(m, " ", task->policy);
630 seq_put_decimal_ull(m, " ", delayacct_blkio_ticks(task));
631 seq_put_decimal_ull(m, " ", nsec_to_clock_t(gtime));
632 seq_put_decimal_ll(m, " ", nsec_to_clock_t(cgtime));
633
634 if (mm && permitted) {
635 seq_put_decimal_ull(m, " ", mm->start_data);
636 seq_put_decimal_ull(m, " ", mm->end_data);
637 seq_put_decimal_ull(m, " ", mm->start_brk);
638 seq_put_decimal_ull(m, " ", mm->arg_start);
639 seq_put_decimal_ull(m, " ", mm->arg_end);
640 seq_put_decimal_ull(m, " ", mm->env_start);
641 seq_put_decimal_ull(m, " ", mm->env_end);
642 } else
643 seq_puts(m, " 0 0 0 0 0 0 0");
644
645 if (permitted)
646 seq_put_decimal_ll(m, " ", task->exit_code);
647 else
648 seq_puts(m, " 0");
649
650 seq_putc(m, '\n');
651 if (mm)
652 mmput(mm);
653 return 0;
654}
655
656int proc_tid_stat(struct seq_file *m, struct pid_namespace *ns,
657 struct pid *pid, struct task_struct *task)
658{
659 return do_task_stat(m, ns, pid, task, 0);
660}
661
662int proc_tgid_stat(struct seq_file *m, struct pid_namespace *ns,
663 struct pid *pid, struct task_struct *task)
664{
665 return do_task_stat(m, ns, pid, task, 1);
666}
667
668int proc_pid_statm(struct seq_file *m, struct pid_namespace *ns,
669 struct pid *pid, struct task_struct *task)
670{
671 struct mm_struct *mm = get_task_mm(task);
672
673 if (mm) {
674 unsigned long size;
675 unsigned long resident = 0;
676 unsigned long shared = 0;
677 unsigned long text = 0;
678 unsigned long data = 0;
679
680 size = task_statm(mm, &shared, &text, &data, &resident);
681 mmput(mm);
682
683 /*
684 * For quick read, open code by putting numbers directly
685 * expected format is
686 * seq_printf(m, "%lu %lu %lu %lu 0 %lu 0\n",
687 * size, resident, shared, text, data);
688 */
689 seq_put_decimal_ull(m, "", size);
690 seq_put_decimal_ull(m, " ", resident);
691 seq_put_decimal_ull(m, " ", shared);
692 seq_put_decimal_ull(m, " ", text);
693 seq_put_decimal_ull(m, " ", 0);
694 seq_put_decimal_ull(m, " ", data);
695 seq_put_decimal_ull(m, " ", 0);
696 seq_putc(m, '\n');
697 } else {
698 seq_write(m, "0 0 0 0 0 0 0\n", 14);
699 }
700 return 0;
701}
702
703#ifdef CONFIG_PROC_CHILDREN
704static struct pid *
705get_children_pid(struct inode *inode, struct pid *pid_prev, loff_t pos)
706{
707 struct task_struct *start, *task;
708 struct pid *pid = NULL;
709
710 read_lock(&tasklist_lock);
711
712 start = pid_task(proc_pid(inode), PIDTYPE_PID);
713 if (!start)
714 goto out;
715
716 /*
717 * Lets try to continue searching first, this gives
718 * us significant speedup on children-rich processes.
719 */
720 if (pid_prev) {
721 task = pid_task(pid_prev, PIDTYPE_PID);
722 if (task && task->real_parent == start &&
723 !(list_empty(&task->sibling))) {
724 if (list_is_last(&task->sibling, &start->children))
725 goto out;
726 task = list_first_entry(&task->sibling,
727 struct task_struct, sibling);
728 pid = get_pid(task_pid(task));
729 goto out;
730 }
731 }
732
733 /*
734 * Slow search case.
735 *
736 * We might miss some children here if children
737 * are exited while we were not holding the lock,
738 * but it was never promised to be accurate that
739 * much.
740 *
741 * "Just suppose that the parent sleeps, but N children
742 * exit after we printed their tids. Now the slow paths
743 * skips N extra children, we miss N tasks." (c)
744 *
745 * So one need to stop or freeze the leader and all
746 * its children to get a precise result.
747 */
748 list_for_each_entry(task, &start->children, sibling) {
749 if (pos-- == 0) {
750 pid = get_pid(task_pid(task));
751 break;
752 }
753 }
754
755out:
756 read_unlock(&tasklist_lock);
757 return pid;
758}
759
760static int children_seq_show(struct seq_file *seq, void *v)
761{
762 struct inode *inode = file_inode(seq->file);
763
764 seq_printf(seq, "%d ", pid_nr_ns(v, proc_pid_ns(inode->i_sb)));
765 return 0;
766}
767
768static void *children_seq_start(struct seq_file *seq, loff_t *pos)
769{
770 return get_children_pid(file_inode(seq->file), NULL, *pos);
771}
772
773static void *children_seq_next(struct seq_file *seq, void *v, loff_t *pos)
774{
775 struct pid *pid;
776
777 pid = get_children_pid(file_inode(seq->file), v, *pos + 1);
778 put_pid(v);
779
780 ++*pos;
781 return pid;
782}
783
784static void children_seq_stop(struct seq_file *seq, void *v)
785{
786 put_pid(v);
787}
788
789static const struct seq_operations children_seq_ops = {
790 .start = children_seq_start,
791 .next = children_seq_next,
792 .stop = children_seq_stop,
793 .show = children_seq_show,
794};
795
796static int children_seq_open(struct inode *inode, struct file *file)
797{
798 return seq_open(file, &children_seq_ops);
799}
800
801const struct file_operations proc_tid_children_operations = {
802 .open = children_seq_open,
803 .read = seq_read,
804 .llseek = seq_lseek,
805 .release = seq_release,
806};
807#endif /* CONFIG_PROC_CHILDREN */