1 // SPDX-License-Identifier: GPL-2.0
3 * linux/fs/proc/array.c
5 * Copyright (C) 1992 by Linus Torvalds
6 * based on ideas by Darren Senn
9 * Michael. K. Johnson: stat,statm extensions.
10 * <johnsonm@stolaf.edu>
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>
18 * Danny ter Haar : added cpuinfo
21 * Alessandro Rubini : profile extension.
22 * <rubini@ipvvis.unipv.it>
24 * Jeff Tranter : added BogoMips field to cpuinfo
25 * <Jeff_Tranter@Mitel.COM>
27 * Bruno Haible : remove 4K limit for the maps file
28 * <haible@ma2s2.mathematik.uni-karlsruhe.de>
30 * Yves Arrouye : remove removal of trailing spaces in get_array.
31 * <Yves.Arrouye@marin.fdn.fr>
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>
40 * aeb@cwi.nl : /proc/partitions
43 * Alan Cox : security fixes.
44 * <alan@lxorguk.ukuu.org.uk>
46 * Al Viro : safe handling of mm_struct
48 * Gerhard Wichert : added BIGMEM support
49 * Siemens AG <Gerhard.Wichert@pdb.siemens.de>
51 * Al Viro & Jeff Garzik : moved most of the thing into base.c and
52 * : proc_misc.c. The rest may eventually go into
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>
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>
95 #include <asm/pgtable.h>
96 #include <asm/processor.h>
99 void proc_task_name(struct seq_file
*m
, struct task_struct
*p
, bool escape
)
106 if (p
->flags
& PF_WQ_WORKER
)
107 wq_worker_comm(tcomm
, sizeof(tcomm
), p
);
109 __get_task_comm(tcomm
, sizeof(tcomm
), p
);
111 size
= seq_get_buf(m
, &buf
);
113 ret
= string_escape_str(tcomm
, buf
, size
,
114 ESCAPE_SPACE
| ESCAPE_SPECIAL
, "\n\\");
118 ret
= strscpy(buf
, tcomm
, size
);
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
130 static const char * const task_state_array
[] = {
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 */
142 /* states beyond TASK_REPORT: */
143 "I (idle)", /* 0x80 */
146 static inline const char *get_task_state(struct task_struct
*tsk
)
148 BUILD_BUG_ON(1 + ilog2(TASK_REPORT_MAX
) != ARRAY_SIZE(task_state_array
));
149 return task_state_array
[task_state_index(tsk
)];
152 static inline void task_state(struct seq_file
*m
, struct pid_namespace
*ns
,
153 struct pid
*pid
, struct task_struct
*p
)
155 struct user_namespace
*user_ns
= seq_user_ns(m
);
156 struct group_info
*group_info
;
158 struct task_struct
*tracer
;
159 const struct cred
*cred
;
160 pid_t ppid
, tpid
= 0, tgid
, ngid
;
161 unsigned int max_fds
= 0;
164 ppid
= pid_alive(p
) ?
165 task_tgid_nr_ns(rcu_dereference(p
->real_parent
), ns
) : 0;
167 tracer
= ptrace_parent(p
);
169 tpid
= task_pid_nr_ns(tracer
, ns
);
171 tgid
= task_tgid_nr_ns(p
, ns
);
172 ngid
= task_numa_group_id(p
);
173 cred
= get_task_cred(p
);
177 umask
= p
->fs
->umask
;
179 max_fds
= files_fdtable(p
->files
)->max_fds
;
184 seq_printf(m
, "Umask:\t%#04o\n", umask
);
185 seq_puts(m
, "State:\t");
186 seq_puts(m
, get_task_state(p
));
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
);
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
]));
209 /* Trailing space shouldn't have been added in the first place. */
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
));
229 void render_sigset_t(struct seq_file
*m
, const char *header
,
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
]);
251 static void collect_sigign_sigcatch(struct task_struct
*p
, sigset_t
*ign
,
254 struct k_sigaction
*k
;
257 k
= p
->sighand
->action
;
258 for (i
= 1; i
<= _NSIG
; ++i
, ++k
) {
259 if (k
->sa
.sa_handler
== SIG_IGN
)
261 else if (k
->sa
.sa_handler
!= SIG_DFL
)
266 static inline void task_sig(struct seq_file
*m
, struct task_struct
*p
)
269 sigset_t pending
, shpending
, blocked
, ignored
, caught
;
271 unsigned int qsize
= 0;
272 unsigned long qlim
= 0;
274 sigemptyset(&pending
);
275 sigemptyset(&shpending
);
276 sigemptyset(&blocked
);
277 sigemptyset(&ignored
);
278 sigemptyset(&caught
);
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
);
289 qlim
= task_rlimit(p
, RLIMIT_SIGPENDING
);
290 unlock_task_sighand(p
, &flags
);
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
);
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
);
305 static void render_cap_t(struct seq_file
*m
, const char *header
,
311 CAP_FOR_EACH_U32(__capi
) {
312 seq_put_hex_ll(m
, NULL
,
313 a
->cap
[CAP_LAST_U32
- __capi
], 8);
318 static inline void task_cap(struct seq_file
*m
, struct task_struct
*p
)
320 const struct cred
*cred
;
321 kernel_cap_t cap_inheritable
, cap_permitted
, cap_effective
,
322 cap_bset
, cap_ambient
;
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
;
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
);
340 static inline void task_seccomp(struct seq_file
*m
, struct task_struct
*p
)
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
);
346 seq_printf(m
, "\nSpeculation_Store_Bypass:\t");
347 switch (arch_prctl_spec_ctrl_get(p
, PR_SPEC_STORE_BYPASS
)) {
349 seq_printf(m
, "unknown");
351 case PR_SPEC_NOT_AFFECTED
:
352 seq_printf(m
, "not vulnerable");
354 case PR_SPEC_PRCTL
| PR_SPEC_FORCE_DISABLE
:
355 seq_printf(m
, "thread force mitigated");
357 case PR_SPEC_PRCTL
| PR_SPEC_DISABLE
:
358 seq_printf(m
, "thread mitigated");
360 case PR_SPEC_PRCTL
| PR_SPEC_ENABLE
:
361 seq_printf(m
, "thread vulnerable");
363 case PR_SPEC_DISABLE
:
364 seq_printf(m
, "globally mitigated");
367 seq_printf(m
, "vulnerable");
373 static inline void task_context_switch_counts(struct seq_file
*m
,
374 struct task_struct
*p
)
376 seq_put_decimal_ull(m
, "voluntary_ctxt_switches:\t", p
->nvcsw
);
377 seq_put_decimal_ull(m
, "\nnonvoluntary_ctxt_switches:\t", p
->nivcsw
);
381 static void task_cpus_allowed(struct seq_file
*m
, struct task_struct
*task
)
383 seq_printf(m
, "Cpus_allowed:\t%*pb\n",
384 cpumask_pr_args(&task
->cpus_allowed
));
385 seq_printf(m
, "Cpus_allowed_list:\t%*pbl\n",
386 cpumask_pr_args(&task
->cpus_allowed
));
389 static inline void task_core_dumping(struct seq_file
*m
, struct mm_struct
*mm
)
391 seq_put_decimal_ull(m
, "CoreDumping:\t", !!mm
->core_state
);
395 int proc_pid_status(struct seq_file
*m
, struct pid_namespace
*ns
,
396 struct pid
*pid
, struct task_struct
*task
)
398 struct mm_struct
*mm
= get_task_mm(task
);
400 seq_puts(m
, "Name:\t");
401 proc_task_name(m
, task
, true);
404 task_state(m
, ns
, pid
, task
);
408 task_core_dumping(m
, mm
);
413 task_seccomp(m
, task
);
414 task_cpus_allowed(m
, task
);
415 cpuset_task_status_allowed(m
, task
);
416 task_context_switch_counts(m
, task
);
420 static int do_task_stat(struct seq_file
*m
, struct pid_namespace
*ns
,
421 struct pid
*pid
, struct task_struct
*task
, int whole
)
423 unsigned long vsize
, eip
, esp
, wchan
= 0;
425 int tty_pgrp
= -1, tty_nr
= 0;
426 sigset_t sigign
, sigcatch
;
428 pid_t ppid
= 0, pgid
= -1, sid
= -1;
431 struct mm_struct
*mm
;
432 unsigned long long start_time
;
433 unsigned long cmin_flt
= 0, cmaj_flt
= 0;
434 unsigned long min_flt
= 0, maj_flt
= 0;
435 u64 cutime
, cstime
, utime
, stime
;
437 unsigned long rsslim
= 0;
440 state
= *get_task_state(task
);
441 vsize
= eip
= esp
= 0;
442 permitted
= ptrace_may_access(task
, PTRACE_MODE_READ_FSCREDS
| PTRACE_MODE_NOAUDIT
);
443 mm
= get_task_mm(task
);
445 vsize
= task_vsize(mm
);
447 * esp and eip are intentionally zeroed out. There is no
448 * non-racy way to read them without freezing the task.
449 * Programs that need reliable values can use ptrace(2).
451 * The only exception is if the task is core dumping because
452 * a program is not able to use ptrace(2) in that case. It is
453 * safe because the task has stopped executing permanently.
455 if (permitted
&& (task
->flags
& (PF_EXITING
|PF_DUMPCORE
))) {
456 if (try_get_task_stack(task
)) {
457 eip
= KSTK_EIP(task
);
458 esp
= KSTK_ESP(task
);
459 put_task_stack(task
);
464 sigemptyset(&sigign
);
465 sigemptyset(&sigcatch
);
466 cutime
= cstime
= utime
= stime
= 0;
469 if (lock_task_sighand(task
, &flags
)) {
470 struct signal_struct
*sig
= task
->signal
;
473 struct pid
*pgrp
= tty_get_pgrp(sig
->tty
);
474 tty_pgrp
= pid_nr_ns(pgrp
, ns
);
476 tty_nr
= new_encode_dev(tty_devnum(sig
->tty
));
479 num_threads
= get_nr_threads(task
);
480 collect_sigign_sigcatch(task
, &sigign
, &sigcatch
);
482 cmin_flt
= sig
->cmin_flt
;
483 cmaj_flt
= sig
->cmaj_flt
;
484 cutime
= sig
->cutime
;
485 cstime
= sig
->cstime
;
486 cgtime
= sig
->cgtime
;
487 rsslim
= READ_ONCE(sig
->rlim
[RLIMIT_RSS
].rlim_cur
);
489 /* add up live thread stats at the group level */
491 struct task_struct
*t
= task
;
493 min_flt
+= t
->min_flt
;
494 maj_flt
+= t
->maj_flt
;
495 gtime
+= task_gtime(t
);
496 } while_each_thread(task
, t
);
498 min_flt
+= sig
->min_flt
;
499 maj_flt
+= sig
->maj_flt
;
500 thread_group_cputime_adjusted(task
, &utime
, &stime
);
504 sid
= task_session_nr_ns(task
, ns
);
505 ppid
= task_tgid_nr_ns(task
->real_parent
, ns
);
506 pgid
= task_pgrp_nr_ns(task
, ns
);
508 unlock_task_sighand(task
, &flags
);
511 if (permitted
&& (!whole
|| num_threads
< 2))
512 wchan
= get_wchan(task
);
514 min_flt
= task
->min_flt
;
515 maj_flt
= task
->maj_flt
;
516 task_cputime_adjusted(task
, &utime
, &stime
);
517 gtime
= task_gtime(task
);
520 /* scale priority and nice values from timeslices to -20..20 */
521 /* to make it look like a "normal" Unix priority/nice value */
522 priority
= task_prio(task
);
523 nice
= task_nice(task
);
525 /* convert nsec -> ticks */
526 start_time
= nsec_to_clock_t(task
->real_start_time
);
528 seq_put_decimal_ull(m
, "", pid_nr_ns(pid
, ns
));
530 proc_task_name(m
, task
, false);
533 seq_put_decimal_ll(m
, " ", ppid
);
534 seq_put_decimal_ll(m
, " ", pgid
);
535 seq_put_decimal_ll(m
, " ", sid
);
536 seq_put_decimal_ll(m
, " ", tty_nr
);
537 seq_put_decimal_ll(m
, " ", tty_pgrp
);
538 seq_put_decimal_ull(m
, " ", task
->flags
);
539 seq_put_decimal_ull(m
, " ", min_flt
);
540 seq_put_decimal_ull(m
, " ", cmin_flt
);
541 seq_put_decimal_ull(m
, " ", maj_flt
);
542 seq_put_decimal_ull(m
, " ", cmaj_flt
);
543 seq_put_decimal_ull(m
, " ", nsec_to_clock_t(utime
));
544 seq_put_decimal_ull(m
, " ", nsec_to_clock_t(stime
));
545 seq_put_decimal_ll(m
, " ", nsec_to_clock_t(cutime
));
546 seq_put_decimal_ll(m
, " ", nsec_to_clock_t(cstime
));
547 seq_put_decimal_ll(m
, " ", priority
);
548 seq_put_decimal_ll(m
, " ", nice
);
549 seq_put_decimal_ll(m
, " ", num_threads
);
550 seq_put_decimal_ull(m
, " ", 0);
551 seq_put_decimal_ull(m
, " ", start_time
);
552 seq_put_decimal_ull(m
, " ", vsize
);
553 seq_put_decimal_ull(m
, " ", mm
? get_mm_rss(mm
) : 0);
554 seq_put_decimal_ull(m
, " ", rsslim
);
555 seq_put_decimal_ull(m
, " ", mm
? (permitted
? mm
->start_code
: 1) : 0);
556 seq_put_decimal_ull(m
, " ", mm
? (permitted
? mm
->end_code
: 1) : 0);
557 seq_put_decimal_ull(m
, " ", (permitted
&& mm
) ? mm
->start_stack
: 0);
558 seq_put_decimal_ull(m
, " ", esp
);
559 seq_put_decimal_ull(m
, " ", eip
);
560 /* The signal information here is obsolete.
561 * It must be decimal for Linux 2.0 compatibility.
562 * Use /proc/#/status for real-time signals.
564 seq_put_decimal_ull(m
, " ", task
->pending
.signal
.sig
[0] & 0x7fffffffUL
);
565 seq_put_decimal_ull(m
, " ", task
->blocked
.sig
[0] & 0x7fffffffUL
);
566 seq_put_decimal_ull(m
, " ", sigign
.sig
[0] & 0x7fffffffUL
);
567 seq_put_decimal_ull(m
, " ", sigcatch
.sig
[0] & 0x7fffffffUL
);
570 * We used to output the absolute kernel address, but that's an
571 * information leak - so instead we show a 0/1 flag here, to signal
572 * to user-space whether there's a wchan field in /proc/PID/wchan.
574 * This works with older implementations of procps as well.
581 seq_put_decimal_ull(m
, " ", 0);
582 seq_put_decimal_ull(m
, " ", 0);
583 seq_put_decimal_ll(m
, " ", task
->exit_signal
);
584 seq_put_decimal_ll(m
, " ", task_cpu(task
));
585 seq_put_decimal_ull(m
, " ", task
->rt_priority
);
586 seq_put_decimal_ull(m
, " ", task
->policy
);
587 seq_put_decimal_ull(m
, " ", delayacct_blkio_ticks(task
));
588 seq_put_decimal_ull(m
, " ", nsec_to_clock_t(gtime
));
589 seq_put_decimal_ll(m
, " ", nsec_to_clock_t(cgtime
));
591 if (mm
&& permitted
) {
592 seq_put_decimal_ull(m
, " ", mm
->start_data
);
593 seq_put_decimal_ull(m
, " ", mm
->end_data
);
594 seq_put_decimal_ull(m
, " ", mm
->start_brk
);
595 seq_put_decimal_ull(m
, " ", mm
->arg_start
);
596 seq_put_decimal_ull(m
, " ", mm
->arg_end
);
597 seq_put_decimal_ull(m
, " ", mm
->env_start
);
598 seq_put_decimal_ull(m
, " ", mm
->env_end
);
600 seq_puts(m
, " 0 0 0 0 0 0 0");
603 seq_put_decimal_ll(m
, " ", task
->exit_code
);
613 int proc_tid_stat(struct seq_file
*m
, struct pid_namespace
*ns
,
614 struct pid
*pid
, struct task_struct
*task
)
616 return do_task_stat(m
, ns
, pid
, task
, 0);
619 int proc_tgid_stat(struct seq_file
*m
, struct pid_namespace
*ns
,
620 struct pid
*pid
, struct task_struct
*task
)
622 return do_task_stat(m
, ns
, pid
, task
, 1);
625 int proc_pid_statm(struct seq_file
*m
, struct pid_namespace
*ns
,
626 struct pid
*pid
, struct task_struct
*task
)
628 unsigned long size
= 0, resident
= 0, shared
= 0, text
= 0, data
= 0;
629 struct mm_struct
*mm
= get_task_mm(task
);
632 size
= task_statm(mm
, &shared
, &text
, &data
, &resident
);
636 * For quick read, open code by putting numbers directly
638 * seq_printf(m, "%lu %lu %lu %lu 0 %lu 0\n",
639 * size, resident, shared, text, data);
641 seq_put_decimal_ull(m
, "", size
);
642 seq_put_decimal_ull(m
, " ", resident
);
643 seq_put_decimal_ull(m
, " ", shared
);
644 seq_put_decimal_ull(m
, " ", text
);
645 seq_put_decimal_ull(m
, " ", 0);
646 seq_put_decimal_ull(m
, " ", data
);
647 seq_put_decimal_ull(m
, " ", 0);
653 #ifdef CONFIG_PROC_CHILDREN
655 get_children_pid(struct inode
*inode
, struct pid
*pid_prev
, loff_t pos
)
657 struct task_struct
*start
, *task
;
658 struct pid
*pid
= NULL
;
660 read_lock(&tasklist_lock
);
662 start
= pid_task(proc_pid(inode
), PIDTYPE_PID
);
667 * Lets try to continue searching first, this gives
668 * us significant speedup on children-rich processes.
671 task
= pid_task(pid_prev
, PIDTYPE_PID
);
672 if (task
&& task
->real_parent
== start
&&
673 !(list_empty(&task
->sibling
))) {
674 if (list_is_last(&task
->sibling
, &start
->children
))
676 task
= list_first_entry(&task
->sibling
,
677 struct task_struct
, sibling
);
678 pid
= get_pid(task_pid(task
));
686 * We might miss some children here if children
687 * are exited while we were not holding the lock,
688 * but it was never promised to be accurate that
691 * "Just suppose that the parent sleeps, but N children
692 * exit after we printed their tids. Now the slow paths
693 * skips N extra children, we miss N tasks." (c)
695 * So one need to stop or freeze the leader and all
696 * its children to get a precise result.
698 list_for_each_entry(task
, &start
->children
, sibling
) {
700 pid
= get_pid(task_pid(task
));
706 read_unlock(&tasklist_lock
);
710 static int children_seq_show(struct seq_file
*seq
, void *v
)
712 struct inode
*inode
= file_inode(seq
->file
);
714 seq_printf(seq
, "%d ", pid_nr_ns(v
, proc_pid_ns(inode
)));
718 static void *children_seq_start(struct seq_file
*seq
, loff_t
*pos
)
720 return get_children_pid(file_inode(seq
->file
), NULL
, *pos
);
723 static void *children_seq_next(struct seq_file
*seq
, void *v
, loff_t
*pos
)
727 pid
= get_children_pid(file_inode(seq
->file
), v
, *pos
+ 1);
734 static void children_seq_stop(struct seq_file
*seq
, void *v
)
739 static const struct seq_operations children_seq_ops
= {
740 .start
= children_seq_start
,
741 .next
= children_seq_next
,
742 .stop
= children_seq_stop
,
743 .show
= children_seq_show
,
746 static int children_seq_open(struct inode
*inode
, struct file
*file
)
748 return seq_open(file
, &children_seq_ops
);
751 const struct file_operations proc_tid_children_operations
= {
752 .open
= children_seq_open
,
755 .release
= seq_release
,
757 #endif /* CONFIG_PROC_CHILDREN */