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/sched.h>
64 #include <linux/proc_fs.h>
65 #include <linux/ioport.h>
66 #include <linux/uaccess.h>
67 #include <linux/io.h>
68 #include <linux/mm.h>
69 #include <linux/hugetlb.h>
70 #include <linux/pagemap.h>
71 #include <linux/swap.h>
72 #include <linux/smp.h>
73 #include <linux/signal.h>
74 #include <linux/highmem.h>
75 #include <linux/file.h>
76 #include <linux/fdtable.h>
77 #include <linux/times.h>
78 #include <linux/cpuset.h>
79 #include <linux/rcupdate.h>
80 #include <linux/delayacct.h>
81 #include <linux/seq_file.h>
82 #include <linux/pid_namespace.h>
83 #include <linux/ptrace.h>
84 #include <linux/tracehook.h>
85 #include <linux/string_helpers.h>
86 #include <linux/user_namespace.h>
87 #include <linux/fs_struct.h>
88
89 #include <asm/pgtable.h>
90 #include <asm/processor.h>
91 #include "internal.h"
92
task_name(struct seq_file * m,struct task_struct * p)93 static inline void task_name(struct seq_file *m, struct task_struct *p)
94 {
95 char *buf;
96 size_t size;
97 char tcomm[sizeof(p->comm)];
98 int ret;
99
100 get_task_comm(tcomm, p);
101
102 seq_puts(m, "Name:\t");
103
104 size = seq_get_buf(m, &buf);
105 ret = string_escape_str(tcomm, buf, size, ESCAPE_SPACE | ESCAPE_SPECIAL, "\n\\");
106 seq_commit(m, ret < size ? ret : -1);
107
108 seq_putc(m, '\n');
109 }
110
111 /*
112 * The task state array is a strange "bitmap" of
113 * reasons to sleep. Thus "running" is zero, and
114 * you can test for combinations of others with
115 * simple bit tests.
116 */
117 static const char * const task_state_array[] = {
118 "R (running)", /* 0 */
119 "S (sleeping)", /* 1 */
120 "D (disk sleep)", /* 2 */
121 "T (stopped)", /* 4 */
122 "t (tracing stop)", /* 8 */
123 "X (dead)", /* 16 */
124 "Z (zombie)", /* 32 */
125 };
126
get_task_state(struct task_struct * tsk)127 static inline const char *get_task_state(struct task_struct *tsk)
128 {
129 unsigned int state = (tsk->state | tsk->exit_state) & TASK_REPORT;
130
131 /*
132 * Parked tasks do not run; they sit in __kthread_parkme().
133 * Without this check, we would report them as running, which is
134 * clearly wrong, so we report them as sleeping instead.
135 */
136 if (tsk->state == TASK_PARKED)
137 state = TASK_INTERRUPTIBLE;
138
139 BUILD_BUG_ON(1 + ilog2(TASK_REPORT) != ARRAY_SIZE(task_state_array)-1);
140
141 return task_state_array[fls(state)];
142 }
143
get_task_umask(struct task_struct * tsk)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
task_state(struct seq_file * m,struct pid_namespace * ns,struct pid * pid,struct task_struct * p)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
render_sigset_t(struct seq_file * m,const char * header,sigset_t * set)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_printf(m, "%x", x);
250 } while (i >= 4);
251
252 seq_putc(m, '\n');
253 }
254
collect_sigign_sigcatch(struct task_struct * p,sigset_t * ign,sigset_t * catch)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
task_sig(struct seq_file * m,struct task_struct * p)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
render_cap_t(struct seq_file * m,const char * header,kernel_cap_t * a)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
task_cap(struct seq_file * m,struct task_struct * p)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
task_seccomp(struct seq_file * m,struct task_struct * p)344 static inline void task_seccomp(struct seq_file *m, struct task_struct *p)
345 {
346 #ifdef CONFIG_SECCOMP
347 seq_put_decimal_ull(m, "Seccomp:\t", p->seccomp.mode);
348 seq_putc(m, '\n');
349 #endif
350 }
351
task_context_switch_counts(struct seq_file * m,struct task_struct * p)352 static inline void task_context_switch_counts(struct seq_file *m,
353 struct task_struct *p)
354 {
355 seq_put_decimal_ull(m, "voluntary_ctxt_switches:\t", p->nvcsw);
356 seq_put_decimal_ull(m, "\nnonvoluntary_ctxt_switches:\t", p->nivcsw);
357 seq_putc(m, '\n');
358 }
359
task_cpus_allowed(struct seq_file * m,struct task_struct * task)360 static void task_cpus_allowed(struct seq_file *m, struct task_struct *task)
361 {
362 seq_printf(m, "Cpus_allowed:\t%*pb\n",
363 cpumask_pr_args(&task->cpus_allowed));
364 seq_printf(m, "Cpus_allowed_list:\t%*pbl\n",
365 cpumask_pr_args(&task->cpus_allowed));
366 }
367
proc_pid_status(struct seq_file * m,struct pid_namespace * ns,struct pid * pid,struct task_struct * task)368 int proc_pid_status(struct seq_file *m, struct pid_namespace *ns,
369 struct pid *pid, struct task_struct *task)
370 {
371 struct mm_struct *mm = get_task_mm(task);
372
373 task_name(m, task);
374 task_state(m, ns, pid, task);
375
376 if (mm) {
377 task_mem(m, mm);
378 mmput(mm);
379 }
380 task_sig(m, task);
381 task_cap(m, task);
382 task_seccomp(m, task);
383 task_cpus_allowed(m, task);
384 cpuset_task_status_allowed(m, task);
385 task_context_switch_counts(m, task);
386 return 0;
387 }
388
do_task_stat(struct seq_file * m,struct pid_namespace * ns,struct pid * pid,struct task_struct * task,int whole)389 static int do_task_stat(struct seq_file *m, struct pid_namespace *ns,
390 struct pid *pid, struct task_struct *task, int whole)
391 {
392 unsigned long vsize, eip, esp, wchan = 0;
393 int priority, nice;
394 int tty_pgrp = -1, tty_nr = 0;
395 sigset_t sigign, sigcatch;
396 char state;
397 pid_t ppid = 0, pgid = -1, sid = -1;
398 int num_threads = 0;
399 int permitted;
400 struct mm_struct *mm;
401 unsigned long long start_time;
402 unsigned long cmin_flt = 0, cmaj_flt = 0;
403 unsigned long min_flt = 0, maj_flt = 0;
404 cputime_t cutime, cstime, utime, stime;
405 cputime_t cgtime, gtime;
406 unsigned long rsslim = 0;
407 char tcomm[sizeof(task->comm)];
408 unsigned long flags;
409
410 state = *get_task_state(task);
411 vsize = eip = esp = 0;
412 permitted = ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS | PTRACE_MODE_NOAUDIT);
413 mm = get_task_mm(task);
414 if (mm) {
415 vsize = task_vsize(mm);
416 /*
417 * esp and eip are intentionally zeroed out. There is no
418 * non-racy way to read them without freezing the task.
419 * Programs that need reliable values can use ptrace(2).
420 *
421 * The only exception is if the task is core dumping because
422 * a program is not able to use ptrace(2) in that case. It is
423 * safe because the task has stopped executing permanently.
424 */
425 if (permitted && (task->flags & PF_DUMPCORE)) {
426 if (try_get_task_stack(task)) {
427 eip = KSTK_EIP(task);
428 esp = KSTK_ESP(task);
429 put_task_stack(task);
430 }
431 }
432 }
433
434 get_task_comm(tcomm, task);
435
436 sigemptyset(&sigign);
437 sigemptyset(&sigcatch);
438 cutime = cstime = utime = stime = 0;
439 cgtime = gtime = 0;
440
441 if (lock_task_sighand(task, &flags)) {
442 struct signal_struct *sig = task->signal;
443
444 if (sig->tty) {
445 struct pid *pgrp = tty_get_pgrp(sig->tty);
446 tty_pgrp = pid_nr_ns(pgrp, ns);
447 put_pid(pgrp);
448 tty_nr = new_encode_dev(tty_devnum(sig->tty));
449 }
450
451 num_threads = get_nr_threads(task);
452 collect_sigign_sigcatch(task, &sigign, &sigcatch);
453
454 cmin_flt = sig->cmin_flt;
455 cmaj_flt = sig->cmaj_flt;
456 cutime = sig->cutime;
457 cstime = sig->cstime;
458 cgtime = sig->cgtime;
459 rsslim = ACCESS_ONCE(sig->rlim[RLIMIT_RSS].rlim_cur);
460
461 /* add up live thread stats at the group level */
462 if (whole) {
463 struct task_struct *t = task;
464 do {
465 min_flt += t->min_flt;
466 maj_flt += t->maj_flt;
467 gtime += task_gtime(t);
468 } while_each_thread(task, t);
469
470 min_flt += sig->min_flt;
471 maj_flt += sig->maj_flt;
472 thread_group_cputime_adjusted(task, &utime, &stime);
473 gtime += sig->gtime;
474 }
475
476 sid = task_session_nr_ns(task, ns);
477 ppid = task_tgid_nr_ns(task->real_parent, ns);
478 pgid = task_pgrp_nr_ns(task, ns);
479
480 unlock_task_sighand(task, &flags);
481 }
482
483 if (permitted && (!whole || num_threads < 2))
484 wchan = get_wchan(task);
485 if (!whole) {
486 min_flt = task->min_flt;
487 maj_flt = task->maj_flt;
488 task_cputime_adjusted(task, &utime, &stime);
489 gtime = task_gtime(task);
490 }
491
492 /* scale priority and nice values from timeslices to -20..20 */
493 /* to make it look like a "normal" Unix priority/nice value */
494 priority = task_prio(task);
495 nice = task_nice(task);
496
497 /* convert nsec -> ticks */
498 start_time = nsec_to_clock_t(task->real_start_time);
499
500 seq_printf(m, "%d (%s) %c", pid_nr_ns(pid, ns), tcomm, state);
501 seq_put_decimal_ll(m, " ", ppid);
502 seq_put_decimal_ll(m, " ", pgid);
503 seq_put_decimal_ll(m, " ", sid);
504 seq_put_decimal_ll(m, " ", tty_nr);
505 seq_put_decimal_ll(m, " ", tty_pgrp);
506 seq_put_decimal_ull(m, " ", task->flags);
507 seq_put_decimal_ull(m, " ", min_flt);
508 seq_put_decimal_ull(m, " ", cmin_flt);
509 seq_put_decimal_ull(m, " ", maj_flt);
510 seq_put_decimal_ull(m, " ", cmaj_flt);
511 seq_put_decimal_ull(m, " ", cputime_to_clock_t(utime));
512 seq_put_decimal_ull(m, " ", cputime_to_clock_t(stime));
513 seq_put_decimal_ll(m, " ", cputime_to_clock_t(cutime));
514 seq_put_decimal_ll(m, " ", cputime_to_clock_t(cstime));
515 seq_put_decimal_ll(m, " ", priority);
516 seq_put_decimal_ll(m, " ", nice);
517 seq_put_decimal_ll(m, " ", num_threads);
518 seq_put_decimal_ull(m, " ", 0);
519 seq_put_decimal_ull(m, " ", start_time);
520 seq_put_decimal_ull(m, " ", vsize);
521 seq_put_decimal_ull(m, " ", mm ? get_mm_rss(mm) : 0);
522 seq_put_decimal_ull(m, " ", rsslim);
523 seq_put_decimal_ull(m, " ", mm ? (permitted ? mm->start_code : 1) : 0);
524 seq_put_decimal_ull(m, " ", mm ? (permitted ? mm->end_code : 1) : 0);
525 seq_put_decimal_ull(m, " ", (permitted && mm) ? mm->start_stack : 0);
526 seq_put_decimal_ull(m, " ", esp);
527 seq_put_decimal_ull(m, " ", eip);
528 /* The signal information here is obsolete.
529 * It must be decimal for Linux 2.0 compatibility.
530 * Use /proc/#/status for real-time signals.
531 */
532 seq_put_decimal_ull(m, " ", task->pending.signal.sig[0] & 0x7fffffffUL);
533 seq_put_decimal_ull(m, " ", task->blocked.sig[0] & 0x7fffffffUL);
534 seq_put_decimal_ull(m, " ", sigign.sig[0] & 0x7fffffffUL);
535 seq_put_decimal_ull(m, " ", sigcatch.sig[0] & 0x7fffffffUL);
536
537 /*
538 * We used to output the absolute kernel address, but that's an
539 * information leak - so instead we show a 0/1 flag here, to signal
540 * to user-space whether there's a wchan field in /proc/PID/wchan.
541 *
542 * This works with older implementations of procps as well.
543 */
544 if (wchan)
545 seq_puts(m, " 1");
546 else
547 seq_puts(m, " 0");
548
549 seq_put_decimal_ull(m, " ", 0);
550 seq_put_decimal_ull(m, " ", 0);
551 seq_put_decimal_ll(m, " ", task->exit_signal);
552 seq_put_decimal_ll(m, " ", task_cpu(task));
553 seq_put_decimal_ull(m, " ", task->rt_priority);
554 seq_put_decimal_ull(m, " ", task->policy);
555 seq_put_decimal_ull(m, " ", delayacct_blkio_ticks(task));
556 seq_put_decimal_ull(m, " ", cputime_to_clock_t(gtime));
557 seq_put_decimal_ll(m, " ", cputime_to_clock_t(cgtime));
558
559 if (mm && permitted) {
560 seq_put_decimal_ull(m, " ", mm->start_data);
561 seq_put_decimal_ull(m, " ", mm->end_data);
562 seq_put_decimal_ull(m, " ", mm->start_brk);
563 seq_put_decimal_ull(m, " ", mm->arg_start);
564 seq_put_decimal_ull(m, " ", mm->arg_end);
565 seq_put_decimal_ull(m, " ", mm->env_start);
566 seq_put_decimal_ull(m, " ", mm->env_end);
567 } else
568 seq_puts(m, " 0 0 0 0 0 0 0");
569
570 if (permitted)
571 seq_put_decimal_ll(m, " ", task->exit_code);
572 else
573 seq_puts(m, " 0");
574
575 seq_putc(m, '\n');
576 if (mm)
577 mmput(mm);
578 return 0;
579 }
580
proc_tid_stat(struct seq_file * m,struct pid_namespace * ns,struct pid * pid,struct task_struct * task)581 int proc_tid_stat(struct seq_file *m, struct pid_namespace *ns,
582 struct pid *pid, struct task_struct *task)
583 {
584 return do_task_stat(m, ns, pid, task, 0);
585 }
586
proc_tgid_stat(struct seq_file * m,struct pid_namespace * ns,struct pid * pid,struct task_struct * task)587 int proc_tgid_stat(struct seq_file *m, struct pid_namespace *ns,
588 struct pid *pid, struct task_struct *task)
589 {
590 return do_task_stat(m, ns, pid, task, 1);
591 }
592
proc_pid_statm(struct seq_file * m,struct pid_namespace * ns,struct pid * pid,struct task_struct * task)593 int proc_pid_statm(struct seq_file *m, struct pid_namespace *ns,
594 struct pid *pid, struct task_struct *task)
595 {
596 unsigned long size = 0, resident = 0, shared = 0, text = 0, data = 0;
597 struct mm_struct *mm = get_task_mm(task);
598
599 if (mm) {
600 size = task_statm(mm, &shared, &text, &data, &resident);
601 mmput(mm);
602 }
603 /*
604 * For quick read, open code by putting numbers directly
605 * expected format is
606 * seq_printf(m, "%lu %lu %lu %lu 0 %lu 0\n",
607 * size, resident, shared, text, data);
608 */
609 seq_put_decimal_ull(m, "", size);
610 seq_put_decimal_ull(m, " ", resident);
611 seq_put_decimal_ull(m, " ", shared);
612 seq_put_decimal_ull(m, " ", text);
613 seq_put_decimal_ull(m, " ", 0);
614 seq_put_decimal_ull(m, " ", data);
615 seq_put_decimal_ull(m, " ", 0);
616 seq_putc(m, '\n');
617
618 return 0;
619 }
620
621 #ifdef CONFIG_PROC_CHILDREN
622 static struct pid *
get_children_pid(struct inode * inode,struct pid * pid_prev,loff_t pos)623 get_children_pid(struct inode *inode, struct pid *pid_prev, loff_t pos)
624 {
625 struct task_struct *start, *task;
626 struct pid *pid = NULL;
627
628 read_lock(&tasklist_lock);
629
630 start = pid_task(proc_pid(inode), PIDTYPE_PID);
631 if (!start)
632 goto out;
633
634 /*
635 * Lets try to continue searching first, this gives
636 * us significant speedup on children-rich processes.
637 */
638 if (pid_prev) {
639 task = pid_task(pid_prev, PIDTYPE_PID);
640 if (task && task->real_parent == start &&
641 !(list_empty(&task->sibling))) {
642 if (list_is_last(&task->sibling, &start->children))
643 goto out;
644 task = list_first_entry(&task->sibling,
645 struct task_struct, sibling);
646 pid = get_pid(task_pid(task));
647 goto out;
648 }
649 }
650
651 /*
652 * Slow search case.
653 *
654 * We might miss some children here if children
655 * are exited while we were not holding the lock,
656 * but it was never promised to be accurate that
657 * much.
658 *
659 * "Just suppose that the parent sleeps, but N children
660 * exit after we printed their tids. Now the slow paths
661 * skips N extra children, we miss N tasks." (c)
662 *
663 * So one need to stop or freeze the leader and all
664 * its children to get a precise result.
665 */
666 list_for_each_entry(task, &start->children, sibling) {
667 if (pos-- == 0) {
668 pid = get_pid(task_pid(task));
669 break;
670 }
671 }
672
673 out:
674 read_unlock(&tasklist_lock);
675 return pid;
676 }
677
children_seq_show(struct seq_file * seq,void * v)678 static int children_seq_show(struct seq_file *seq, void *v)
679 {
680 struct inode *inode = seq->private;
681 pid_t pid;
682
683 pid = pid_nr_ns(v, inode->i_sb->s_fs_info);
684 seq_printf(seq, "%d ", pid);
685
686 return 0;
687 }
688
children_seq_start(struct seq_file * seq,loff_t * pos)689 static void *children_seq_start(struct seq_file *seq, loff_t *pos)
690 {
691 return get_children_pid(seq->private, NULL, *pos);
692 }
693
children_seq_next(struct seq_file * seq,void * v,loff_t * pos)694 static void *children_seq_next(struct seq_file *seq, void *v, loff_t *pos)
695 {
696 struct pid *pid;
697
698 pid = get_children_pid(seq->private, v, *pos + 1);
699 put_pid(v);
700
701 ++*pos;
702 return pid;
703 }
704
children_seq_stop(struct seq_file * seq,void * v)705 static void children_seq_stop(struct seq_file *seq, void *v)
706 {
707 put_pid(v);
708 }
709
710 static const struct seq_operations children_seq_ops = {
711 .start = children_seq_start,
712 .next = children_seq_next,
713 .stop = children_seq_stop,
714 .show = children_seq_show,
715 };
716
children_seq_open(struct inode * inode,struct file * file)717 static int children_seq_open(struct inode *inode, struct file *file)
718 {
719 struct seq_file *m;
720 int ret;
721
722 ret = seq_open(file, &children_seq_ops);
723 if (ret)
724 return ret;
725
726 m = file->private_data;
727 m->private = inode;
728
729 return ret;
730 }
731
children_seq_release(struct inode * inode,struct file * file)732 int children_seq_release(struct inode *inode, struct file *file)
733 {
734 seq_release(inode, file);
735 return 0;
736 }
737
738 const struct file_operations proc_tid_children_operations = {
739 .open = children_seq_open,
740 .read = seq_read,
741 .llseek = seq_lseek,
742 .release = children_seq_release,
743 };
744 #endif /* CONFIG_PROC_CHILDREN */
745