?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Merge tag 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mst/vhost
[cris-mirror.git] / fs / proc / array.c
blob598803576e4c04445b56bb5e7f4ae564c764f7a3
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/ptrace.h>
89 #include <linux/tracehook.h>
90 #include <linux/string_helpers.h>
91 #include <linux/user_namespace.h>
92 #include <linux/fs_struct.h>
93
94 #include <asm/pgtable.h>
95 #include <asm/processor.h>
96 #include "internal.h"
97
98 static inline void task_name(struct seq_file *m, struct task_struct *p)
99 {
100 char *buf;
101 size_t size;
102 char tcomm[sizeof(p->comm)];
103 int ret;
104
105 get_task_comm(tcomm, p);
106
107 seq_puts(m, "Name:\t");
108
109 size = seq_get_buf(m, &buf);
110 ret = string_escape_str(tcomm, buf, size, ESCAPE_SPACE | ESCAPE_SPECIAL, "\n\\");
111 seq_commit(m, ret < size ? ret : -1);
112
113 seq_putc(m, '\n');
114 }
115
116 /*
117 * The task state array is a strange "bitmap" of
118 * reasons to sleep. Thus "running" is zero, and
119 * you can test for combinations of others with
120 * simple bit tests.
121 */
122 static const char * const task_state_array[] = {
123
124 /* states in TASK_REPORT: */
125 "R (running)", /* 0x00 */
126 "S (sleeping)", /* 0x01 */
127 "D (disk sleep)", /* 0x02 */
128 "T (stopped)", /* 0x04 */
129 "t (tracing stop)", /* 0x08 */
130 "X (dead)", /* 0x10 */
131 "Z (zombie)", /* 0x20 */
132 "P (parked)", /* 0x40 */
133
134 /* states beyond TASK_REPORT: */
135 "I (idle)", /* 0x80 */
136 };
137
138 static inline const char *get_task_state(struct task_struct *tsk)
139 {
140 BUILD_BUG_ON(1 + ilog2(TASK_REPORT_MAX) != ARRAY_SIZE(task_state_array));
141 return task_state_array[task_state_index(tsk)];
142 }
143
144 static inline int get_task_umask(struct task_struct *tsk)
145 {
146 struct fs_struct *fs;
147 int umask = -ENOENT;
148
149 task_lock(tsk);
150 fs = tsk->fs;
151 if (fs)
152 umask = fs->umask;
153 task_unlock(tsk);
154 return umask;
155 }
156
157 static inline void task_state(struct seq_file *m, struct pid_namespace *ns,
158 struct pid *pid, struct task_struct *p)
159 {
160 struct user_namespace *user_ns = seq_user_ns(m);
161 struct group_info *group_info;
162 int g, umask;
163 struct task_struct *tracer;
164 const struct cred *cred;
165 pid_t ppid, tpid = 0, tgid, ngid;
166 unsigned int max_fds = 0;
167
168 rcu_read_lock();
169 ppid = pid_alive(p) ?
170 task_tgid_nr_ns(rcu_dereference(p->real_parent), ns) : 0;
171
172 tracer = ptrace_parent(p);
173 if (tracer)
174 tpid = task_pid_nr_ns(tracer, ns);
175
176 tgid = task_tgid_nr_ns(p, ns);
177 ngid = task_numa_group_id(p);
178 cred = get_task_cred(p);
179
180 umask = get_task_umask(p);
181 if (umask >= 0)
182 seq_printf(m, "Umask:\t%#04o\n", umask);
183
184 task_lock(p);
185 if (p->files)
186 max_fds = files_fdtable(p->files)->max_fds;
187 task_unlock(p);
188 rcu_read_unlock();
189
190 seq_printf(m, "State:\t%s", get_task_state(p));
191
192 seq_put_decimal_ull(m, "\nTgid:\t", tgid);
193 seq_put_decimal_ull(m, "\nNgid:\t", ngid);
194 seq_put_decimal_ull(m, "\nPid:\t", pid_nr_ns(pid, ns));
195 seq_put_decimal_ull(m, "\nPPid:\t", ppid);
196 seq_put_decimal_ull(m, "\nTracerPid:\t", tpid);
197 seq_put_decimal_ull(m, "\nUid:\t", from_kuid_munged(user_ns, cred->uid));
198 seq_put_decimal_ull(m, "\t", from_kuid_munged(user_ns, cred->euid));
199 seq_put_decimal_ull(m, "\t", from_kuid_munged(user_ns, cred->suid));
200 seq_put_decimal_ull(m, "\t", from_kuid_munged(user_ns, cred->fsuid));
201 seq_put_decimal_ull(m, "\nGid:\t", from_kgid_munged(user_ns, cred->gid));
202 seq_put_decimal_ull(m, "\t", from_kgid_munged(user_ns, cred->egid));
203 seq_put_decimal_ull(m, "\t", from_kgid_munged(user_ns, cred->sgid));
204 seq_put_decimal_ull(m, "\t", from_kgid_munged(user_ns, cred->fsgid));
205 seq_put_decimal_ull(m, "\nFDSize:\t", max_fds);
206
207 seq_puts(m, "\nGroups:\t");
208 group_info = cred->group_info;
209 for (g = 0; g < group_info->ngroups; g++)
210 seq_put_decimal_ull(m, g ? " " : "",
211 from_kgid_munged(user_ns, group_info->gid[g]));
212 put_cred(cred);
213 /* Trailing space shouldn't have been added in the first place. */
214 seq_putc(m, ' ');
215
216 #ifdef CONFIG_PID_NS
217 seq_puts(m, "\nNStgid:");
218 for (g = ns->level; g <= pid->level; g++)
219 seq_put_decimal_ull(m, "\t", task_tgid_nr_ns(p, pid->numbers[g].ns));
220 seq_puts(m, "\nNSpid:");
221 for (g = ns->level; g <= pid->level; g++)
222 seq_put_decimal_ull(m, "\t", task_pid_nr_ns(p, pid->numbers[g].ns));
223 seq_puts(m, "\nNSpgid:");
224 for (g = ns->level; g <= pid->level; g++)
225 seq_put_decimal_ull(m, "\t", task_pgrp_nr_ns(p, pid->numbers[g].ns));
226 seq_puts(m, "\nNSsid:");
227 for (g = ns->level; g <= pid->level; g++)
228 seq_put_decimal_ull(m, "\t", task_session_nr_ns(p, pid->numbers[g].ns));
229 #endif
230 seq_putc(m, '\n');
231 }
232
233 void render_sigset_t(struct seq_file *m, const char *header,
234 sigset_t *set)
235 {
236 int i;
237
238 seq_puts(m, header);
239
240 i = _NSIG;
241 do {
242 int x = 0;
243
244 i -= 4;
245 if (sigismember(set, i+1)) x |= 1;
246 if (sigismember(set, i+2)) x |= 2;
247 if (sigismember(set, i+3)) x |= 4;
248 if (sigismember(set, i+4)) x |= 8;
249 seq_putc(m, hex_asc[x]);
250 } while (i >= 4);
251
252 seq_putc(m, '\n');
253 }
254
255 static void collect_sigign_sigcatch(struct task_struct *p, sigset_t *ign,
256 sigset_t *catch)
257 {
258 struct k_sigaction *k;
259 int i;
260
261 k = p->sighand->action;
262 for (i = 1; i <= _NSIG; ++i, ++k) {
263 if (k->sa.sa_handler == SIG_IGN)
264 sigaddset(ign, i);
265 else if (k->sa.sa_handler != SIG_DFL)
266 sigaddset(catch, i);
267 }
268 }
269
270 static inline void task_sig(struct seq_file *m, struct task_struct *p)
271 {
272 unsigned long flags;
273 sigset_t pending, shpending, blocked, ignored, caught;
274 int num_threads = 0;
275 unsigned long qsize = 0;
276 unsigned long qlim = 0;
277
278 sigemptyset(&pending);
279 sigemptyset(&shpending);
280 sigemptyset(&blocked);
281 sigemptyset(&ignored);
282 sigemptyset(&caught);
283
284 if (lock_task_sighand(p, &flags)) {
285 pending = p->pending.signal;
286 shpending = p->signal->shared_pending.signal;
287 blocked = p->blocked;
288 collect_sigign_sigcatch(p, &ignored, &caught);
289 num_threads = get_nr_threads(p);
290 rcu_read_lock(); /* FIXME: is this correct? */
291 qsize = atomic_read(&__task_cred(p)->user->sigpending);
292 rcu_read_unlock();
293 qlim = task_rlimit(p, RLIMIT_SIGPENDING);
294 unlock_task_sighand(p, &flags);
295 }
296
297 seq_put_decimal_ull(m, "Threads:\t", num_threads);
298 seq_put_decimal_ull(m, "\nSigQ:\t", qsize);
299 seq_put_decimal_ull(m, "/", qlim);
300
301 /* render them all */
302 render_sigset_t(m, "\nSigPnd:\t", &pending);
303 render_sigset_t(m, "ShdPnd:\t", &shpending);
304 render_sigset_t(m, "SigBlk:\t", &blocked);
305 render_sigset_t(m, "SigIgn:\t", &ignored);
306 render_sigset_t(m, "SigCgt:\t", &caught);
307 }
308
309 static void render_cap_t(struct seq_file *m, const char *header,
310 kernel_cap_t *a)
311 {
312 unsigned __capi;
313
314 seq_puts(m, header);
315 CAP_FOR_EACH_U32(__capi) {
316 seq_printf(m, "%08x",
317 a->cap[CAP_LAST_U32 - __capi]);
318 }
319 seq_putc(m, '\n');
320 }
321
322 static inline void task_cap(struct seq_file *m, struct task_struct *p)
323 {
324 const struct cred *cred;
325 kernel_cap_t cap_inheritable, cap_permitted, cap_effective,
326 cap_bset, cap_ambient;
327
328 rcu_read_lock();
329 cred = __task_cred(p);
330 cap_inheritable = cred->cap_inheritable;
331 cap_permitted = cred->cap_permitted;
332 cap_effective = cred->cap_effective;
333 cap_bset = cred->cap_bset;
334 cap_ambient = cred->cap_ambient;
335 rcu_read_unlock();
336
337 render_cap_t(m, "CapInh:\t", &cap_inheritable);
338 render_cap_t(m, "CapPrm:\t", &cap_permitted);
339 render_cap_t(m, "CapEff:\t", &cap_effective);
340 render_cap_t(m, "CapBnd:\t", &cap_bset);
341 render_cap_t(m, "CapAmb:\t", &cap_ambient);
342 }
343
344 static inline void task_seccomp(struct seq_file *m, struct task_struct *p)
345 {
346 seq_put_decimal_ull(m, "NoNewPrivs:\t", task_no_new_privs(p));
347 #ifdef CONFIG_SECCOMP
348 seq_put_decimal_ull(m, "\nSeccomp:\t", p->seccomp.mode);
349 #endif
350 seq_putc(m, '\n');
351 }
352
353 static inline void task_context_switch_counts(struct seq_file *m,
354 struct task_struct *p)
355 {
356 seq_put_decimal_ull(m, "voluntary_ctxt_switches:\t", p->nvcsw);
357 seq_put_decimal_ull(m, "\nnonvoluntary_ctxt_switches:\t", p->nivcsw);
358 seq_putc(m, '\n');
359 }
360
361 static void task_cpus_allowed(struct seq_file *m, struct task_struct *task)
362 {
363 seq_printf(m, "Cpus_allowed:\t%*pb\n",
364 cpumask_pr_args(&task->cpus_allowed));
365 seq_printf(m, "Cpus_allowed_list:\t%*pbl\n",
366 cpumask_pr_args(&task->cpus_allowed));
367 }
368
369 static inline void task_core_dumping(struct seq_file *m, struct mm_struct *mm)
370 {
371 seq_printf(m, "CoreDumping:\t%d\n", !!mm->core_state);
372 }
373
374 int proc_pid_status(struct seq_file *m, struct pid_namespace *ns,
375 struct pid *pid, struct task_struct *task)
376 {
377 struct mm_struct *mm = get_task_mm(task);
378
379 task_name(m, task);
380 task_state(m, ns, pid, task);
381
382 if (mm) {
383 task_mem(m, mm);
384 task_core_dumping(m, mm);
385 mmput(mm);
386 }
387 task_sig(m, task);
388 task_cap(m, task);
389 task_seccomp(m, task);
390 task_cpus_allowed(m, task);
391 cpuset_task_status_allowed(m, task);
392 task_context_switch_counts(m, task);
393 return 0;
394 }
395
396 static int do_task_stat(struct seq_file *m, struct pid_namespace *ns,
397 struct pid *pid, struct task_struct *task, int whole)
398 {
399 unsigned long vsize, eip, esp, wchan = 0;
400 int priority, nice;
401 int tty_pgrp = -1, tty_nr = 0;
402 sigset_t sigign, sigcatch;
403 char state;
404 pid_t ppid = 0, pgid = -1, sid = -1;
405 int num_threads = 0;
406 int permitted;
407 struct mm_struct *mm;
408 unsigned long long start_time;
409 unsigned long cmin_flt = 0, cmaj_flt = 0;
410 unsigned long min_flt = 0, maj_flt = 0;
411 u64 cutime, cstime, utime, stime;
412 u64 cgtime, gtime;
413 unsigned long rsslim = 0;
414 char tcomm[sizeof(task->comm)];
415 unsigned long flags;
416
417 state = *get_task_state(task);
418 vsize = eip = esp = 0;
419 permitted = ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS | PTRACE_MODE_NOAUDIT);
420 mm = get_task_mm(task);
421 if (mm) {
422 vsize = task_vsize(mm);
423 /*
424 * esp and eip are intentionally zeroed out. There is no
425 * non-racy way to read them without freezing the task.
426 * Programs that need reliable values can use ptrace(2).
427 *
428 * The only exception is if the task is core dumping because
429 * a program is not able to use ptrace(2) in that case. It is
430 * safe because the task has stopped executing permanently.
431 */
432 if (permitted && (task->flags & PF_DUMPCORE)) {
433 if (try_get_task_stack(task)) {
434 eip = KSTK_EIP(task);
435 esp = KSTK_ESP(task);
436 put_task_stack(task);
437 }
438 }
439 }
440
441 get_task_comm(tcomm, task);
442
443 sigemptyset(&sigign);
444 sigemptyset(&sigcatch);
445 cutime = cstime = utime = stime = 0;
446 cgtime = gtime = 0;
447
448 if (lock_task_sighand(task, &flags)) {
449 struct signal_struct *sig = task->signal;
450
451 if (sig->tty) {
452 struct pid *pgrp = tty_get_pgrp(sig->tty);
453 tty_pgrp = pid_nr_ns(pgrp, ns);
454 put_pid(pgrp);
455 tty_nr = new_encode_dev(tty_devnum(sig->tty));
456 }
457
458 num_threads = get_nr_threads(task);
459 collect_sigign_sigcatch(task, &sigign, &sigcatch);
460
461 cmin_flt = sig->cmin_flt;
462 cmaj_flt = sig->cmaj_flt;
463 cutime = sig->cutime;
464 cstime = sig->cstime;
465 cgtime = sig->cgtime;
466 rsslim = READ_ONCE(sig->rlim[RLIMIT_RSS].rlim_cur);
467
468 /* add up live thread stats at the group level */
469 if (whole) {
470 struct task_struct *t = task;
471 do {
472 min_flt += t->min_flt;
473 maj_flt += t->maj_flt;
474 gtime += task_gtime(t);
475 } while_each_thread(task, t);
476
477 min_flt += sig->min_flt;
478 maj_flt += sig->maj_flt;
479 thread_group_cputime_adjusted(task, &utime, &stime);
480 gtime += sig->gtime;
481 }
482
483 sid = task_session_nr_ns(task, ns);
484 ppid = task_tgid_nr_ns(task->real_parent, ns);
485 pgid = task_pgrp_nr_ns(task, ns);
486
487 unlock_task_sighand(task, &flags);
488 }
489
490 if (permitted && (!whole || num_threads < 2))
491 wchan = get_wchan(task);
492 if (!whole) {
493 min_flt = task->min_flt;
494 maj_flt = task->maj_flt;
495 task_cputime_adjusted(task, &utime, &stime);
496 gtime = task_gtime(task);
497 }
498
499 /* scale priority and nice values from timeslices to -20..20 */
500 /* to make it look like a "normal" Unix priority/nice value */
501 priority = task_prio(task);
502 nice = task_nice(task);
503
504 /* convert nsec -> ticks */
505 start_time = nsec_to_clock_t(task->real_start_time);
506
507 seq_printf(m, "%d (%s) %c", pid_nr_ns(pid, ns), tcomm, state);
508 seq_put_decimal_ll(m, " ", ppid);
509 seq_put_decimal_ll(m, " ", pgid);
510 seq_put_decimal_ll(m, " ", sid);
511 seq_put_decimal_ll(m, " ", tty_nr);
512 seq_put_decimal_ll(m, " ", tty_pgrp);
513 seq_put_decimal_ull(m, " ", task->flags);
514 seq_put_decimal_ull(m, " ", min_flt);
515 seq_put_decimal_ull(m, " ", cmin_flt);
516 seq_put_decimal_ull(m, " ", maj_flt);
517 seq_put_decimal_ull(m, " ", cmaj_flt);
518 seq_put_decimal_ull(m, " ", nsec_to_clock_t(utime));
519 seq_put_decimal_ull(m, " ", nsec_to_clock_t(stime));
520 seq_put_decimal_ll(m, " ", nsec_to_clock_t(cutime));
521 seq_put_decimal_ll(m, " ", nsec_to_clock_t(cstime));
522 seq_put_decimal_ll(m, " ", priority);
523 seq_put_decimal_ll(m, " ", nice);
524 seq_put_decimal_ll(m, " ", num_threads);
525 seq_put_decimal_ull(m, " ", 0);
526 seq_put_decimal_ull(m, " ", start_time);
527 seq_put_decimal_ull(m, " ", vsize);
528 seq_put_decimal_ull(m, " ", mm ? get_mm_rss(mm) : 0);
529 seq_put_decimal_ull(m, " ", rsslim);
530 seq_put_decimal_ull(m, " ", mm ? (permitted ? mm->start_code : 1) : 0);
531 seq_put_decimal_ull(m, " ", mm ? (permitted ? mm->end_code : 1) : 0);
532 seq_put_decimal_ull(m, " ", (permitted && mm) ? mm->start_stack : 0);
533 seq_put_decimal_ull(m, " ", esp);
534 seq_put_decimal_ull(m, " ", eip);
535 /* The signal information here is obsolete.
536 * It must be decimal for Linux 2.0 compatibility.
537 * Use /proc/#/status for real-time signals.
538 */
539 seq_put_decimal_ull(m, " ", task->pending.signal.sig[0] & 0x7fffffffUL);
540 seq_put_decimal_ull(m, " ", task->blocked.sig[0] & 0x7fffffffUL);
541 seq_put_decimal_ull(m, " ", sigign.sig[0] & 0x7fffffffUL);
542 seq_put_decimal_ull(m, " ", sigcatch.sig[0] & 0x7fffffffUL);
543
544 /*
545 * We used to output the absolute kernel address, but that's an
546 * information leak - so instead we show a 0/1 flag here, to signal
547 * to user-space whether there's a wchan field in /proc/PID/wchan.
548 *
549 * This works with older implementations of procps as well.
550 */
551 if (wchan)
552 seq_puts(m, " 1");
553 else
554 seq_puts(m, " 0");
555
556 seq_put_decimal_ull(m, " ", 0);
557 seq_put_decimal_ull(m, " ", 0);
558 seq_put_decimal_ll(m, " ", task->exit_signal);
559 seq_put_decimal_ll(m, " ", task_cpu(task));
560 seq_put_decimal_ull(m, " ", task->rt_priority);
561 seq_put_decimal_ull(m, " ", task->policy);
562 seq_put_decimal_ull(m, " ", delayacct_blkio_ticks(task));
563 seq_put_decimal_ull(m, " ", nsec_to_clock_t(gtime));
564 seq_put_decimal_ll(m, " ", nsec_to_clock_t(cgtime));
565
566 if (mm && permitted) {
567 seq_put_decimal_ull(m, " ", mm->start_data);
568 seq_put_decimal_ull(m, " ", mm->end_data);
569 seq_put_decimal_ull(m, " ", mm->start_brk);
570 seq_put_decimal_ull(m, " ", mm->arg_start);
571 seq_put_decimal_ull(m, " ", mm->arg_end);
572 seq_put_decimal_ull(m, " ", mm->env_start);
573 seq_put_decimal_ull(m, " ", mm->env_end);
574 } else
575 seq_puts(m, " 0 0 0 0 0 0 0");
576
577 if (permitted)
578 seq_put_decimal_ll(m, " ", task->exit_code);
579 else
580 seq_puts(m, " 0");
581
582 seq_putc(m, '\n');
583 if (mm)
584 mmput(mm);
585 return 0;
586 }
587
588 int proc_tid_stat(struct seq_file *m, struct pid_namespace *ns,
589 struct pid *pid, struct task_struct *task)
590 {
591 return do_task_stat(m, ns, pid, task, 0);
592 }
593
594 int proc_tgid_stat(struct seq_file *m, struct pid_namespace *ns,
595 struct pid *pid, struct task_struct *task)
596 {
597 return do_task_stat(m, ns, pid, task, 1);
598 }
599
600 int proc_pid_statm(struct seq_file *m, struct pid_namespace *ns,
601 struct pid *pid, struct task_struct *task)
602 {
603 unsigned long size = 0, resident = 0, shared = 0, text = 0, data = 0;
604 struct mm_struct *mm = get_task_mm(task);
605
606 if (mm) {
607 size = task_statm(mm, &shared, &text, &data, &resident);
608 mmput(mm);
609 }
610 /*
611 * For quick read, open code by putting numbers directly
612 * expected format is
613 * seq_printf(m, "%lu %lu %lu %lu 0 %lu 0\n",
614 * size, resident, shared, text, data);
615 */
616 seq_put_decimal_ull(m, "", size);
617 seq_put_decimal_ull(m, " ", resident);
618 seq_put_decimal_ull(m, " ", shared);
619 seq_put_decimal_ull(m, " ", text);
620 seq_put_decimal_ull(m, " ", 0);
621 seq_put_decimal_ull(m, " ", data);
622 seq_put_decimal_ull(m, " ", 0);
623 seq_putc(m, '\n');
624
625 return 0;
626 }
627
628 #ifdef CONFIG_PROC_CHILDREN
629 static struct pid *
630 get_children_pid(struct inode *inode, struct pid *pid_prev, loff_t pos)
631 {
632 struct task_struct *start, *task;
633 struct pid *pid = NULL;
634
635 read_lock(&tasklist_lock);
636
637 start = pid_task(proc_pid(inode), PIDTYPE_PID);
638 if (!start)
639 goto out;
640
641 /*
642 * Lets try to continue searching first, this gives
643 * us significant speedup on children-rich processes.
644 */
645 if (pid_prev) {
646 task = pid_task(pid_prev, PIDTYPE_PID);
647 if (task && task->real_parent == start &&
648 !(list_empty(&task->sibling))) {
649 if (list_is_last(&task->sibling, &start->children))
650 goto out;
651 task = list_first_entry(&task->sibling,
652 struct task_struct, sibling);
653 pid = get_pid(task_pid(task));
654 goto out;
655 }
656 }
657
658 /*
659 * Slow search case.
660 *
661 * We might miss some children here if children
662 * are exited while we were not holding the lock,
663 * but it was never promised to be accurate that
664 * much.
665 *
666 * "Just suppose that the parent sleeps, but N children
667 * exit after we printed their tids. Now the slow paths
668 * skips N extra children, we miss N tasks." (c)
669 *
670 * So one need to stop or freeze the leader and all
671 * its children to get a precise result.
672 */
673 list_for_each_entry(task, &start->children, sibling) {
674 if (pos-- == 0) {
675 pid = get_pid(task_pid(task));
676 break;
677 }
678 }
679
680 out:
681 read_unlock(&tasklist_lock);
682 return pid;
683 }
684
685 static int children_seq_show(struct seq_file *seq, void *v)
686 {
687 struct inode *inode = seq->private;
688 pid_t pid;
689
690 pid = pid_nr_ns(v, inode->i_sb->s_fs_info);
691 seq_printf(seq, "%d ", pid);
692
693 return 0;
694 }
695
696 static void *children_seq_start(struct seq_file *seq, loff_t *pos)
697 {
698 return get_children_pid(seq->private, NULL, *pos);
699 }
700
701 static void *children_seq_next(struct seq_file *seq, void *v, loff_t *pos)
702 {
703 struct pid *pid;
704
705 pid = get_children_pid(seq->private, v, *pos + 1);
706 put_pid(v);
707
708 ++*pos;
709 return pid;
710 }
711
712 static void children_seq_stop(struct seq_file *seq, void *v)
713 {
714 put_pid(v);
715 }
716
717 static const struct seq_operations children_seq_ops = {
718 .start = children_seq_start,
719 .next = children_seq_next,
720 .stop = children_seq_stop,
721 .show = children_seq_show,
722 };
723
724 static int children_seq_open(struct inode *inode, struct file *file)
725 {
726 struct seq_file *m;
727 int ret;
728
729 ret = seq_open(file, &children_seq_ops);
730 if (ret)
731 return ret;
732
733 m = file->private_data;
734 m->private = inode;
735
736 return ret;
737 }
738
739 const struct file_operations proc_tid_children_operations = {
740 .open = children_seq_open,
741 .read = seq_read,
742 .llseek = seq_lseek,
743 .release = seq_release,
744 };
745 #endif /* CONFIG_PROC_CHILDREN */
CRIS linux kernel tree