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
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}
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/kernel.h>
60#include <linux/kernel_stat.h>
61#include <linux/tty.h>
62#include <linux/string.h>
63#include <linux/mman.h>
64#include <linux/sched/mm.h>
65#include <linux/sched/numa_balancing.h>
66#include <linux/sched/task_stack.h>
67#include <linux/sched/task.h>
68#include <linux/sched/cputime.h>
69#include <linux/proc_fs.h>
70#include <linux/ioport.h>
71#include <linux/uaccess.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/tracehook.h>
91#include <linux/string_helpers.h>
92#include <linux/user_namespace.h>
93#include <linux/fs_struct.h>
94
95#include <asm/pgtable.h>
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 *ign,
252 sigset_t *catch)
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(ign, i);
261 else if (k->sa.sa_handler != SIG_DFL)
262 sigaddset(catch, 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 = atomic_read(&__task_cred(p)->user->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#endif
346 seq_puts(m, "\nSpeculation_Store_Bypass:\t");
347 switch (arch_prctl_spec_ctrl_get(p, PR_SPEC_STORE_BYPASS)) {
348 case -EINVAL:
349 seq_puts(m, "unknown");
350 break;
351 case PR_SPEC_NOT_AFFECTED:
352 seq_puts(m, "not vulnerable");
353 break;
354 case PR_SPEC_PRCTL | PR_SPEC_FORCE_DISABLE:
355 seq_puts(m, "thread force mitigated");
356 break;
357 case PR_SPEC_PRCTL | PR_SPEC_DISABLE:
358 seq_puts(m, "thread mitigated");
359 break;
360 case PR_SPEC_PRCTL | PR_SPEC_ENABLE:
361 seq_puts(m, "thread vulnerable");
362 break;
363 case PR_SPEC_DISABLE:
364 seq_puts(m, "globally mitigated");
365 break;
366 default:
367 seq_puts(m, "vulnerable");
368 break;
369 }
370 seq_putc(m, '\n');
371}
372
373static inline void task_context_switch_counts(struct seq_file *m,
374 struct task_struct *p)
375{
376 seq_put_decimal_ull(m, "voluntary_ctxt_switches:\t", p->nvcsw);
377 seq_put_decimal_ull(m, "\nnonvoluntary_ctxt_switches:\t", p->nivcsw);
378 seq_putc(m, '\n');
379}
380
381static void task_cpus_allowed(struct seq_file *m, struct task_struct *task)
382{
383 seq_printf(m, "Cpus_allowed:\t%*pb\n",
384 cpumask_pr_args(task->cpus_ptr));
385 seq_printf(m, "Cpus_allowed_list:\t%*pbl\n",
386 cpumask_pr_args(task->cpus_ptr));
387}
388
389static inline void task_core_dumping(struct seq_file *m, struct mm_struct *mm)
390{
391 seq_put_decimal_ull(m, "CoreDumping:\t", !!mm->core_state);
392 seq_putc(m, '\n');
393}
394
395static inline void task_thp_status(struct seq_file *m, struct mm_struct *mm)
396{
397 bool thp_enabled = IS_ENABLED(CONFIG_TRANSPARENT_HUGEPAGE);
398
399 if (thp_enabled)
400 thp_enabled = !test_bit(MMF_DISABLE_THP, &mm->flags);
401 seq_printf(m, "THP_enabled:\t%d\n", thp_enabled);
402}
403
404int proc_pid_status(struct seq_file *m, struct pid_namespace *ns,
405 struct pid *pid, struct task_struct *task)
406{
407 struct mm_struct *mm = get_task_mm(task);
408
409 seq_puts(m, "Name:\t");
410 proc_task_name(m, task, true);
411 seq_putc(m, '\n');
412
413 task_state(m, ns, pid, task);
414
415 if (mm) {
416 task_mem(m, mm);
417 task_core_dumping(m, mm);
418 task_thp_status(m, mm);
419 mmput(mm);
420 }
421 task_sig(m, task);
422 task_cap(m, task);
423 task_seccomp(m, task);
424 task_cpus_allowed(m, task);
425 cpuset_task_status_allowed(m, task);
426 task_context_switch_counts(m, task);
427 return 0;
428}
429
430static int do_task_stat(struct seq_file *m, struct pid_namespace *ns,
431 struct pid *pid, struct task_struct *task, int whole)
432{
433 unsigned long vsize, eip, esp, wchan = 0;
434 int priority, nice;
435 int tty_pgrp = -1, tty_nr = 0;
436 sigset_t sigign, sigcatch;
437 char state;
438 pid_t ppid = 0, pgid = -1, sid = -1;
439 int num_threads = 0;
440 int permitted;
441 struct mm_struct *mm;
442 unsigned long long start_time;
443 unsigned long cmin_flt = 0, cmaj_flt = 0;
444 unsigned long min_flt = 0, maj_flt = 0;
445 u64 cutime, cstime, utime, stime;
446 u64 cgtime, gtime;
447 unsigned long rsslim = 0;
448 unsigned long flags;
449
450 state = *get_task_state(task);
451 vsize = eip = esp = 0;
452 permitted = ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS | PTRACE_MODE_NOAUDIT);
453 mm = get_task_mm(task);
454 if (mm) {
455 vsize = task_vsize(mm);
456 /*
457 * esp and eip are intentionally zeroed out. There is no
458 * non-racy way to read them without freezing the task.
459 * Programs that need reliable values can use ptrace(2).
460 *
461 * The only exception is if the task is core dumping because
462 * a program is not able to use ptrace(2) in that case. It is
463 * safe because the task has stopped executing permanently.
464 */
465 if (permitted && (task->flags & (PF_EXITING|PF_DUMPCORE))) {
466 if (try_get_task_stack(task)) {
467 eip = KSTK_EIP(task);
468 esp = KSTK_ESP(task);
469 put_task_stack(task);
470 }
471 }
472 }
473
474 sigemptyset(&sigign);
475 sigemptyset(&sigcatch);
476 cutime = cstime = utime = stime = 0;
477 cgtime = gtime = 0;
478
479 if (lock_task_sighand(task, &flags)) {
480 struct signal_struct *sig = task->signal;
481
482 if (sig->tty) {
483 struct pid *pgrp = tty_get_pgrp(sig->tty);
484 tty_pgrp = pid_nr_ns(pgrp, ns);
485 put_pid(pgrp);
486 tty_nr = new_encode_dev(tty_devnum(sig->tty));
487 }
488
489 num_threads = get_nr_threads(task);
490 collect_sigign_sigcatch(task, &sigign, &sigcatch);
491
492 cmin_flt = sig->cmin_flt;
493 cmaj_flt = sig->cmaj_flt;
494 cutime = sig->cutime;
495 cstime = sig->cstime;
496 cgtime = sig->cgtime;
497 rsslim = READ_ONCE(sig->rlim[RLIMIT_RSS].rlim_cur);
498
499 /* add up live thread stats at the group level */
500 if (whole) {
501 struct task_struct *t = task;
502 do {
503 min_flt += t->min_flt;
504 maj_flt += t->maj_flt;
505 gtime += task_gtime(t);
506 } while_each_thread(task, t);
507
508 min_flt += sig->min_flt;
509 maj_flt += sig->maj_flt;
510 thread_group_cputime_adjusted(task, &utime, &stime);
511 gtime += sig->gtime;
512 }
513
514 sid = task_session_nr_ns(task, ns);
515 ppid = task_tgid_nr_ns(task->real_parent, ns);
516 pgid = task_pgrp_nr_ns(task, ns);
517
518 unlock_task_sighand(task, &flags);
519 }
520
521 if (permitted && (!whole || num_threads < 2))
522 wchan = get_wchan(task);
523 if (!whole) {
524 min_flt = task->min_flt;
525 maj_flt = task->maj_flt;
526 task_cputime_adjusted(task, &utime, &stime);
527 gtime = task_gtime(task);
528 }
529
530 /* scale priority and nice values from timeslices to -20..20 */
531 /* to make it look like a "normal" Unix priority/nice value */
532 priority = task_prio(task);
533 nice = task_nice(task);
534
535 /* convert nsec -> ticks */
536 start_time = nsec_to_clock_t(task->real_start_time);
537
538 seq_put_decimal_ull(m, "", pid_nr_ns(pid, ns));
539 seq_puts(m, " (");
540 proc_task_name(m, task, false);
541 seq_puts(m, ") ");
542 seq_putc(m, state);
543 seq_put_decimal_ll(m, " ", ppid);
544 seq_put_decimal_ll(m, " ", pgid);
545 seq_put_decimal_ll(m, " ", sid);
546 seq_put_decimal_ll(m, " ", tty_nr);
547 seq_put_decimal_ll(m, " ", tty_pgrp);
548 seq_put_decimal_ull(m, " ", task->flags);
549 seq_put_decimal_ull(m, " ", min_flt);
550 seq_put_decimal_ull(m, " ", cmin_flt);
551 seq_put_decimal_ull(m, " ", maj_flt);
552 seq_put_decimal_ull(m, " ", cmaj_flt);
553 seq_put_decimal_ull(m, " ", nsec_to_clock_t(utime));
554 seq_put_decimal_ull(m, " ", nsec_to_clock_t(stime));
555 seq_put_decimal_ll(m, " ", nsec_to_clock_t(cutime));
556 seq_put_decimal_ll(m, " ", nsec_to_clock_t(cstime));
557 seq_put_decimal_ll(m, " ", priority);
558 seq_put_decimal_ll(m, " ", nice);
559 seq_put_decimal_ll(m, " ", num_threads);
560 seq_put_decimal_ull(m, " ", 0);
561 seq_put_decimal_ull(m, " ", start_time);
562 seq_put_decimal_ull(m, " ", vsize);
563 seq_put_decimal_ull(m, " ", mm ? get_mm_rss(mm) : 0);
564 seq_put_decimal_ull(m, " ", rsslim);
565 seq_put_decimal_ull(m, " ", mm ? (permitted ? mm->start_code : 1) : 0);
566 seq_put_decimal_ull(m, " ", mm ? (permitted ? mm->end_code : 1) : 0);
567 seq_put_decimal_ull(m, " ", (permitted && mm) ? mm->start_stack : 0);
568 seq_put_decimal_ull(m, " ", esp);
569 seq_put_decimal_ull(m, " ", eip);
570 /* The signal information here is obsolete.
571 * It must be decimal for Linux 2.0 compatibility.
572 * Use /proc/#/status for real-time signals.
573 */
574 seq_put_decimal_ull(m, " ", task->pending.signal.sig[0] & 0x7fffffffUL);
575 seq_put_decimal_ull(m, " ", task->blocked.sig[0] & 0x7fffffffUL);
576 seq_put_decimal_ull(m, " ", sigign.sig[0] & 0x7fffffffUL);
577 seq_put_decimal_ull(m, " ", sigcatch.sig[0] & 0x7fffffffUL);
578
579 /*
580 * We used to output the absolute kernel address, but that's an
581 * information leak - so instead we show a 0/1 flag here, to signal
582 * to user-space whether there's a wchan field in /proc/PID/wchan.
583 *
584 * This works with older implementations of procps as well.
585 */
586 if (wchan)
587 seq_puts(m, " 1");
588 else
589 seq_puts(m, " 0");
590
591 seq_put_decimal_ull(m, " ", 0);
592 seq_put_decimal_ull(m, " ", 0);
593 seq_put_decimal_ll(m, " ", task->exit_signal);
594 seq_put_decimal_ll(m, " ", task_cpu(task));
595 seq_put_decimal_ull(m, " ", task->rt_priority);
596 seq_put_decimal_ull(m, " ", task->policy);
597 seq_put_decimal_ull(m, " ", delayacct_blkio_ticks(task));
598 seq_put_decimal_ull(m, " ", nsec_to_clock_t(gtime));
599 seq_put_decimal_ll(m, " ", nsec_to_clock_t(cgtime));
600
601 if (mm && permitted) {
602 seq_put_decimal_ull(m, " ", mm->start_data);
603 seq_put_decimal_ull(m, " ", mm->end_data);
604 seq_put_decimal_ull(m, " ", mm->start_brk);
605 seq_put_decimal_ull(m, " ", mm->arg_start);
606 seq_put_decimal_ull(m, " ", mm->arg_end);
607 seq_put_decimal_ull(m, " ", mm->env_start);
608 seq_put_decimal_ull(m, " ", mm->env_end);
609 } else
610 seq_puts(m, " 0 0 0 0 0 0 0");
611
612 if (permitted)
613 seq_put_decimal_ll(m, " ", task->exit_code);
614 else
615 seq_puts(m, " 0");
616
617 seq_putc(m, '\n');
618 if (mm)
619 mmput(mm);
620 return 0;
621}
622
623int proc_tid_stat(struct seq_file *m, struct pid_namespace *ns,
624 struct pid *pid, struct task_struct *task)
625{
626 return do_task_stat(m, ns, pid, task, 0);
627}
628
629int proc_tgid_stat(struct seq_file *m, struct pid_namespace *ns,
630 struct pid *pid, struct task_struct *task)
631{
632 return do_task_stat(m, ns, pid, task, 1);
633}
634
635int proc_pid_statm(struct seq_file *m, struct pid_namespace *ns,
636 struct pid *pid, struct task_struct *task)
637{
638 unsigned long size = 0, resident = 0, shared = 0, text = 0, data = 0;
639 struct mm_struct *mm = get_task_mm(task);
640
641 if (mm) {
642 size = task_statm(mm, &shared, &text, &data, &resident);
643 mmput(mm);
644 }
645 /*
646 * For quick read, open code by putting numbers directly
647 * expected format is
648 * seq_printf(m, "%lu %lu %lu %lu 0 %lu 0\n",
649 * size, resident, shared, text, data);
650 */
651 seq_put_decimal_ull(m, "", size);
652 seq_put_decimal_ull(m, " ", resident);
653 seq_put_decimal_ull(m, " ", shared);
654 seq_put_decimal_ull(m, " ", text);
655 seq_put_decimal_ull(m, " ", 0);
656 seq_put_decimal_ull(m, " ", data);
657 seq_put_decimal_ull(m, " ", 0);
658 seq_putc(m, '\n');
659
660 return 0;
661}
662
663#ifdef CONFIG_PROC_CHILDREN
664static struct pid *
665get_children_pid(struct inode *inode, struct pid *pid_prev, loff_t pos)
666{
667 struct task_struct *start, *task;
668 struct pid *pid = NULL;
669
670 read_lock(&tasklist_lock);
671
672 start = pid_task(proc_pid(inode), PIDTYPE_PID);
673 if (!start)
674 goto out;
675
676 /*
677 * Lets try to continue searching first, this gives
678 * us significant speedup on children-rich processes.
679 */
680 if (pid_prev) {
681 task = pid_task(pid_prev, PIDTYPE_PID);
682 if (task && task->real_parent == start &&
683 !(list_empty(&task->sibling))) {
684 if (list_is_last(&task->sibling, &start->children))
685 goto out;
686 task = list_first_entry(&task->sibling,
687 struct task_struct, sibling);
688 pid = get_pid(task_pid(task));
689 goto out;
690 }
691 }
692
693 /*
694 * Slow search case.
695 *
696 * We might miss some children here if children
697 * are exited while we were not holding the lock,
698 * but it was never promised to be accurate that
699 * much.
700 *
701 * "Just suppose that the parent sleeps, but N children
702 * exit after we printed their tids. Now the slow paths
703 * skips N extra children, we miss N tasks." (c)
704 *
705 * So one need to stop or freeze the leader and all
706 * its children to get a precise result.
707 */
708 list_for_each_entry(task, &start->children, sibling) {
709 if (pos-- == 0) {
710 pid = get_pid(task_pid(task));
711 break;
712 }
713 }
714
715out:
716 read_unlock(&tasklist_lock);
717 return pid;
718}
719
720static int children_seq_show(struct seq_file *seq, void *v)
721{
722 struct inode *inode = file_inode(seq->file);
723
724 seq_printf(seq, "%d ", pid_nr_ns(v, proc_pid_ns(inode)));
725 return 0;
726}
727
728static void *children_seq_start(struct seq_file *seq, loff_t *pos)
729{
730 return get_children_pid(file_inode(seq->file), NULL, *pos);
731}
732
733static void *children_seq_next(struct seq_file *seq, void *v, loff_t *pos)
734{
735 struct pid *pid;
736
737 pid = get_children_pid(file_inode(seq->file), v, *pos + 1);
738 put_pid(v);
739
740 ++*pos;
741 return pid;
742}
743
744static void children_seq_stop(struct seq_file *seq, void *v)
745{
746 put_pid(v);
747}
748
749static const struct seq_operations children_seq_ops = {
750 .start = children_seq_start,
751 .next = children_seq_next,
752 .stop = children_seq_stop,
753 .show = children_seq_show,
754};
755
756static int children_seq_open(struct inode *inode, struct file *file)
757{
758 return seq_open(file, &children_seq_ops);
759}
760
761const struct file_operations proc_tid_children_operations = {
762 .open = children_seq_open,
763 .read = seq_read,
764 .llseek = seq_lseek,
765 .release = seq_release,
766};
767#endif /* CONFIG_PROC_CHILDREN */