2 * linux/fs/proc/array.c
4 * Copyright (C) 1992 by Linus Torvalds
5 * based on ideas by Darren Senn
8 * Michael. K. Johnson: stat,statm extensions.
9 * <johnsonm@stolaf.edu>
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>
17 * Danny ter Haar : added cpuinfo
20 * Alessandro Rubini : profile extension.
21 * <rubini@ipvvis.unipv.it>
23 * Jeff Tranter : added BogoMips field to cpuinfo
24 * <Jeff_Tranter@Mitel.COM>
26 * Bruno Haible : remove 4K limit for the maps file
27 * <haible@ma2s2.mathematik.uni-karlsruhe.de>
29 * Yves Arrouye : remove removal of trailing spaces in get_array.
30 * <Yves.Arrouye@marin.fdn.fr>
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>
39 * aeb@cwi.nl : /proc/partitions
42 * Alan Cox : security fixes.
43 * <alan@lxorguk.ukuu.org.uk>
45 * Al Viro : safe handling of mm_struct
47 * Gerhard Wichert : added BIGMEM support
48 * Siemens AG <Gerhard.Wichert@pdb.siemens.de>
50 * Al Viro & Jeff Garzik : moved most of the thing into base.c and
51 * : proc_misc.c. The rest may eventually go into
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>
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 #include <linux/string_helpers.h>
85 #include <linux/user_namespace.h>
87 #include <asm/pgtable.h>
88 #include <asm/processor.h>
91 static inline void task_name(struct seq_file
*m
, struct task_struct
*p
)
94 char tcomm
[sizeof(p
->comm
)];
96 get_task_comm(tcomm
, p
);
98 seq_puts(m
, "Name:\t");
99 buf
= m
->buf
+ m
->count
;
101 /* Ignore error for now */
102 buf
+= string_escape_str(tcomm
, buf
, m
->size
- m
->count
,
103 ESCAPE_SPACE
| ESCAPE_SPECIAL
, "\n\\");
105 m
->count
= buf
- m
->buf
;
110 * The task state array is a strange "bitmap" of
111 * reasons to sleep. Thus "running" is zero, and
112 * you can test for combinations of others with
115 static const char * const task_state_array
[] = {
116 "R (running)", /* 0 */
117 "S (sleeping)", /* 1 */
118 "D (disk sleep)", /* 2 */
119 "T (stopped)", /* 4 */
120 "t (tracing stop)", /* 8 */
122 "Z (zombie)", /* 32 */
125 static inline const char *get_task_state(struct task_struct
*tsk
)
127 unsigned int state
= (tsk
->state
| tsk
->exit_state
) & TASK_REPORT
;
130 * Parked tasks do not run; they sit in __kthread_parkme().
131 * Without this check, we would report them as running, which is
132 * clearly wrong, so we report them as sleeping instead.
134 if (tsk
->state
== TASK_PARKED
)
135 state
= TASK_INTERRUPTIBLE
;
137 BUILD_BUG_ON(1 + ilog2(TASK_REPORT
) != ARRAY_SIZE(task_state_array
)-1);
139 return task_state_array
[fls(state
)];
142 static inline void task_state(struct seq_file
*m
, struct pid_namespace
*ns
,
143 struct pid
*pid
, struct task_struct
*p
)
145 struct user_namespace
*user_ns
= seq_user_ns(m
);
146 struct group_info
*group_info
;
148 struct task_struct
*tracer
;
149 const struct cred
*cred
;
150 pid_t ppid
, tpid
= 0, tgid
, ngid
;
151 unsigned int max_fds
= 0;
154 ppid
= pid_alive(p
) ?
155 task_tgid_nr_ns(rcu_dereference(p
->real_parent
), ns
) : 0;
157 tracer
= ptrace_parent(p
);
159 tpid
= task_pid_nr_ns(tracer
, ns
);
161 tgid
= task_tgid_nr_ns(p
, ns
);
162 ngid
= task_numa_group_id(p
);
163 cred
= get_task_cred(p
);
167 max_fds
= files_fdtable(p
->files
)->max_fds
;
178 "Uid:\t%d\t%d\t%d\t%d\n"
179 "Gid:\t%d\t%d\t%d\t%d\n"
180 "FDSize:\t%d\nGroups:\t",
182 tgid
, ngid
, pid_nr_ns(pid
, ns
), ppid
, tpid
,
183 from_kuid_munged(user_ns
, cred
->uid
),
184 from_kuid_munged(user_ns
, cred
->euid
),
185 from_kuid_munged(user_ns
, cred
->suid
),
186 from_kuid_munged(user_ns
, cred
->fsuid
),
187 from_kgid_munged(user_ns
, cred
->gid
),
188 from_kgid_munged(user_ns
, cred
->egid
),
189 from_kgid_munged(user_ns
, cred
->sgid
),
190 from_kgid_munged(user_ns
, cred
->fsgid
),
193 group_info
= cred
->group_info
;
194 for (g
= 0; g
< group_info
->ngroups
; g
++)
196 from_kgid_munged(user_ns
, GROUP_AT(group_info
, g
)));
200 seq_puts(m
, "\nNStgid:");
201 for (g
= ns
->level
; g
<= pid
->level
; g
++)
202 seq_printf(m
, "\t%d",
203 task_tgid_nr_ns(p
, pid
->numbers
[g
].ns
));
204 seq_puts(m
, "\nNSpid:");
205 for (g
= ns
->level
; g
<= pid
->level
; g
++)
206 seq_printf(m
, "\t%d",
207 task_pid_nr_ns(p
, pid
->numbers
[g
].ns
));
208 seq_puts(m
, "\nNSpgid:");
209 for (g
= ns
->level
; g
<= pid
->level
; g
++)
210 seq_printf(m
, "\t%d",
211 task_pgrp_nr_ns(p
, pid
->numbers
[g
].ns
));
212 seq_puts(m
, "\nNSsid:");
213 for (g
= ns
->level
; g
<= pid
->level
; g
++)
214 seq_printf(m
, "\t%d",
215 task_session_nr_ns(p
, pid
->numbers
[g
].ns
));
220 void render_sigset_t(struct seq_file
*m
, const char *header
,
232 if (sigismember(set
, i
+1)) x
|= 1;
233 if (sigismember(set
, i
+2)) x
|= 2;
234 if (sigismember(set
, i
+3)) x
|= 4;
235 if (sigismember(set
, i
+4)) x
|= 8;
236 seq_printf(m
, "%x", x
);
242 static void collect_sigign_sigcatch(struct task_struct
*p
, sigset_t
*ign
,
245 struct k_sigaction
*k
;
248 k
= p
->sighand
->action
;
249 for (i
= 1; i
<= _NSIG
; ++i
, ++k
) {
250 if (k
->sa
.sa_handler
== SIG_IGN
)
252 else if (k
->sa
.sa_handler
!= SIG_DFL
)
257 static inline void task_sig(struct seq_file
*m
, struct task_struct
*p
)
260 sigset_t pending
, shpending
, blocked
, ignored
, caught
;
262 unsigned long qsize
= 0;
263 unsigned long qlim
= 0;
265 sigemptyset(&pending
);
266 sigemptyset(&shpending
);
267 sigemptyset(&blocked
);
268 sigemptyset(&ignored
);
269 sigemptyset(&caught
);
271 if (lock_task_sighand(p
, &flags
)) {
272 pending
= p
->pending
.signal
;
273 shpending
= p
->signal
->shared_pending
.signal
;
274 blocked
= p
->blocked
;
275 collect_sigign_sigcatch(p
, &ignored
, &caught
);
276 num_threads
= get_nr_threads(p
);
277 rcu_read_lock(); /* FIXME: is this correct? */
278 qsize
= atomic_read(&__task_cred(p
)->user
->sigpending
);
280 qlim
= task_rlimit(p
, RLIMIT_SIGPENDING
);
281 unlock_task_sighand(p
, &flags
);
284 seq_printf(m
, "Threads:\t%d\n", num_threads
);
285 seq_printf(m
, "SigQ:\t%lu/%lu\n", qsize
, qlim
);
287 /* render them all */
288 render_sigset_t(m
, "SigPnd:\t", &pending
);
289 render_sigset_t(m
, "ShdPnd:\t", &shpending
);
290 render_sigset_t(m
, "SigBlk:\t", &blocked
);
291 render_sigset_t(m
, "SigIgn:\t", &ignored
);
292 render_sigset_t(m
, "SigCgt:\t", &caught
);
295 static void render_cap_t(struct seq_file
*m
, const char *header
,
301 CAP_FOR_EACH_U32(__capi
) {
302 seq_printf(m
, "%08x",
303 a
->cap
[CAP_LAST_U32
- __capi
]);
308 static inline void task_cap(struct seq_file
*m
, struct task_struct
*p
)
310 const struct cred
*cred
;
311 kernel_cap_t cap_inheritable
, cap_permitted
, cap_effective
, cap_bset
;
314 cred
= __task_cred(p
);
315 cap_inheritable
= cred
->cap_inheritable
;
316 cap_permitted
= cred
->cap_permitted
;
317 cap_effective
= cred
->cap_effective
;
318 cap_bset
= cred
->cap_bset
;
321 render_cap_t(m
, "CapInh:\t", &cap_inheritable
);
322 render_cap_t(m
, "CapPrm:\t", &cap_permitted
);
323 render_cap_t(m
, "CapEff:\t", &cap_effective
);
324 render_cap_t(m
, "CapBnd:\t", &cap_bset
);
327 static inline void task_seccomp(struct seq_file
*m
, struct task_struct
*p
)
329 #ifdef CONFIG_SECCOMP
330 seq_printf(m
, "Seccomp:\t%d\n", p
->seccomp
.mode
);
334 static inline void task_context_switch_counts(struct seq_file
*m
,
335 struct task_struct
*p
)
337 seq_printf(m
, "voluntary_ctxt_switches:\t%lu\n"
338 "nonvoluntary_ctxt_switches:\t%lu\n",
343 static void task_cpus_allowed(struct seq_file
*m
, struct task_struct
*task
)
345 seq_printf(m
, "Cpus_allowed:\t%*pb\n",
346 cpumask_pr_args(&task
->cpus_allowed
));
347 seq_printf(m
, "Cpus_allowed_list:\t%*pbl\n",
348 cpumask_pr_args(&task
->cpus_allowed
));
351 int proc_pid_status(struct seq_file
*m
, struct pid_namespace
*ns
,
352 struct pid
*pid
, struct task_struct
*task
)
354 struct mm_struct
*mm
= get_task_mm(task
);
357 task_state(m
, ns
, pid
, task
);
365 task_seccomp(m
, task
);
366 task_cpus_allowed(m
, task
);
367 cpuset_task_status_allowed(m
, task
);
368 task_context_switch_counts(m
, task
);
372 static int do_task_stat(struct seq_file
*m
, struct pid_namespace
*ns
,
373 struct pid
*pid
, struct task_struct
*task
, int whole
)
375 unsigned long vsize
, eip
, esp
, wchan
= ~0UL;
377 int tty_pgrp
= -1, tty_nr
= 0;
378 sigset_t sigign
, sigcatch
;
380 pid_t ppid
= 0, pgid
= -1, sid
= -1;
383 struct mm_struct
*mm
;
384 unsigned long long start_time
;
385 unsigned long cmin_flt
= 0, cmaj_flt
= 0;
386 unsigned long min_flt
= 0, maj_flt
= 0;
387 cputime_t cutime
, cstime
, utime
, stime
;
388 cputime_t cgtime
, gtime
;
389 unsigned long rsslim
= 0;
390 char tcomm
[sizeof(task
->comm
)];
393 state
= *get_task_state(task
);
394 vsize
= eip
= esp
= 0;
395 permitted
= ptrace_may_access(task
, PTRACE_MODE_READ
| PTRACE_MODE_NOAUDIT
);
396 mm
= get_task_mm(task
);
398 vsize
= task_vsize(mm
);
400 eip
= KSTK_EIP(task
);
401 esp
= KSTK_ESP(task
);
405 get_task_comm(tcomm
, task
);
407 sigemptyset(&sigign
);
408 sigemptyset(&sigcatch
);
409 cutime
= cstime
= utime
= stime
= 0;
412 if (lock_task_sighand(task
, &flags
)) {
413 struct signal_struct
*sig
= task
->signal
;
416 struct pid
*pgrp
= tty_get_pgrp(sig
->tty
);
417 tty_pgrp
= pid_nr_ns(pgrp
, ns
);
419 tty_nr
= new_encode_dev(tty_devnum(sig
->tty
));
422 num_threads
= get_nr_threads(task
);
423 collect_sigign_sigcatch(task
, &sigign
, &sigcatch
);
425 cmin_flt
= sig
->cmin_flt
;
426 cmaj_flt
= sig
->cmaj_flt
;
427 cutime
= sig
->cutime
;
428 cstime
= sig
->cstime
;
429 cgtime
= sig
->cgtime
;
430 rsslim
= ACCESS_ONCE(sig
->rlim
[RLIMIT_RSS
].rlim_cur
);
432 /* add up live thread stats at the group level */
434 struct task_struct
*t
= task
;
436 min_flt
+= t
->min_flt
;
437 maj_flt
+= t
->maj_flt
;
438 gtime
+= task_gtime(t
);
439 } while_each_thread(task
, t
);
441 min_flt
+= sig
->min_flt
;
442 maj_flt
+= sig
->maj_flt
;
443 thread_group_cputime_adjusted(task
, &utime
, &stime
);
447 sid
= task_session_nr_ns(task
, ns
);
448 ppid
= task_tgid_nr_ns(task
->real_parent
, ns
);
449 pgid
= task_pgrp_nr_ns(task
, ns
);
451 unlock_task_sighand(task
, &flags
);
454 if (permitted
&& (!whole
|| num_threads
< 2))
455 wchan
= get_wchan(task
);
457 min_flt
= task
->min_flt
;
458 maj_flt
= task
->maj_flt
;
459 task_cputime_adjusted(task
, &utime
, &stime
);
460 gtime
= task_gtime(task
);
463 /* scale priority and nice values from timeslices to -20..20 */
464 /* to make it look like a "normal" Unix priority/nice value */
465 priority
= task_prio(task
);
466 nice
= task_nice(task
);
468 /* convert nsec -> ticks */
469 start_time
= nsec_to_clock_t(task
->real_start_time
);
471 seq_printf(m
, "%d (%s) %c", pid_nr_ns(pid
, ns
), tcomm
, state
);
472 seq_put_decimal_ll(m
, ' ', ppid
);
473 seq_put_decimal_ll(m
, ' ', pgid
);
474 seq_put_decimal_ll(m
, ' ', sid
);
475 seq_put_decimal_ll(m
, ' ', tty_nr
);
476 seq_put_decimal_ll(m
, ' ', tty_pgrp
);
477 seq_put_decimal_ull(m
, ' ', task
->flags
);
478 seq_put_decimal_ull(m
, ' ', min_flt
);
479 seq_put_decimal_ull(m
, ' ', cmin_flt
);
480 seq_put_decimal_ull(m
, ' ', maj_flt
);
481 seq_put_decimal_ull(m
, ' ', cmaj_flt
);
482 seq_put_decimal_ull(m
, ' ', cputime_to_clock_t(utime
));
483 seq_put_decimal_ull(m
, ' ', cputime_to_clock_t(stime
));
484 seq_put_decimal_ll(m
, ' ', cputime_to_clock_t(cutime
));
485 seq_put_decimal_ll(m
, ' ', cputime_to_clock_t(cstime
));
486 seq_put_decimal_ll(m
, ' ', priority
);
487 seq_put_decimal_ll(m
, ' ', nice
);
488 seq_put_decimal_ll(m
, ' ', num_threads
);
489 seq_put_decimal_ull(m
, ' ', 0);
490 seq_put_decimal_ull(m
, ' ', start_time
);
491 seq_put_decimal_ull(m
, ' ', vsize
);
492 seq_put_decimal_ull(m
, ' ', mm
? get_mm_rss(mm
) : 0);
493 seq_put_decimal_ull(m
, ' ', rsslim
);
494 seq_put_decimal_ull(m
, ' ', mm
? (permitted
? mm
->start_code
: 1) : 0);
495 seq_put_decimal_ull(m
, ' ', mm
? (permitted
? mm
->end_code
: 1) : 0);
496 seq_put_decimal_ull(m
, ' ', (permitted
&& mm
) ? mm
->start_stack
: 0);
497 seq_put_decimal_ull(m
, ' ', esp
);
498 seq_put_decimal_ull(m
, ' ', eip
);
499 /* The signal information here is obsolete.
500 * It must be decimal for Linux 2.0 compatibility.
501 * Use /proc/#/status for real-time signals.
503 seq_put_decimal_ull(m
, ' ', task
->pending
.signal
.sig
[0] & 0x7fffffffUL
);
504 seq_put_decimal_ull(m
, ' ', task
->blocked
.sig
[0] & 0x7fffffffUL
);
505 seq_put_decimal_ull(m
, ' ', sigign
.sig
[0] & 0x7fffffffUL
);
506 seq_put_decimal_ull(m
, ' ', sigcatch
.sig
[0] & 0x7fffffffUL
);
507 seq_put_decimal_ull(m
, ' ', wchan
);
508 seq_put_decimal_ull(m
, ' ', 0);
509 seq_put_decimal_ull(m
, ' ', 0);
510 seq_put_decimal_ll(m
, ' ', task
->exit_signal
);
511 seq_put_decimal_ll(m
, ' ', task_cpu(task
));
512 seq_put_decimal_ull(m
, ' ', task
->rt_priority
);
513 seq_put_decimal_ull(m
, ' ', task
->policy
);
514 seq_put_decimal_ull(m
, ' ', delayacct_blkio_ticks(task
));
515 seq_put_decimal_ull(m
, ' ', cputime_to_clock_t(gtime
));
516 seq_put_decimal_ll(m
, ' ', cputime_to_clock_t(cgtime
));
518 if (mm
&& permitted
) {
519 seq_put_decimal_ull(m
, ' ', mm
->start_data
);
520 seq_put_decimal_ull(m
, ' ', mm
->end_data
);
521 seq_put_decimal_ull(m
, ' ', mm
->start_brk
);
522 seq_put_decimal_ull(m
, ' ', mm
->arg_start
);
523 seq_put_decimal_ull(m
, ' ', mm
->arg_end
);
524 seq_put_decimal_ull(m
, ' ', mm
->env_start
);
525 seq_put_decimal_ull(m
, ' ', mm
->env_end
);
527 seq_printf(m
, " 0 0 0 0 0 0 0");
530 seq_put_decimal_ll(m
, ' ', task
->exit_code
);
532 seq_put_decimal_ll(m
, ' ', 0);
540 int proc_tid_stat(struct seq_file
*m
, struct pid_namespace
*ns
,
541 struct pid
*pid
, struct task_struct
*task
)
543 return do_task_stat(m
, ns
, pid
, task
, 0);
546 int proc_tgid_stat(struct seq_file
*m
, struct pid_namespace
*ns
,
547 struct pid
*pid
, struct task_struct
*task
)
549 return do_task_stat(m
, ns
, pid
, task
, 1);
552 int proc_pid_statm(struct seq_file
*m
, struct pid_namespace
*ns
,
553 struct pid
*pid
, struct task_struct
*task
)
555 unsigned long size
= 0, resident
= 0, shared
= 0, text
= 0, data
= 0;
556 struct mm_struct
*mm
= get_task_mm(task
);
559 size
= task_statm(mm
, &shared
, &text
, &data
, &resident
);
563 * For quick read, open code by putting numbers directly
565 * seq_printf(m, "%lu %lu %lu %lu 0 %lu 0\n",
566 * size, resident, shared, text, data);
568 seq_put_decimal_ull(m
, 0, size
);
569 seq_put_decimal_ull(m
, ' ', resident
);
570 seq_put_decimal_ull(m
, ' ', shared
);
571 seq_put_decimal_ull(m
, ' ', text
);
572 seq_put_decimal_ull(m
, ' ', 0);
573 seq_put_decimal_ull(m
, ' ', data
);
574 seq_put_decimal_ull(m
, ' ', 0);
580 #ifdef CONFIG_PROC_CHILDREN
582 get_children_pid(struct inode
*inode
, struct pid
*pid_prev
, loff_t pos
)
584 struct task_struct
*start
, *task
;
585 struct pid
*pid
= NULL
;
587 read_lock(&tasklist_lock
);
589 start
= pid_task(proc_pid(inode
), PIDTYPE_PID
);
594 * Lets try to continue searching first, this gives
595 * us significant speedup on children-rich processes.
598 task
= pid_task(pid_prev
, PIDTYPE_PID
);
599 if (task
&& task
->real_parent
== start
&&
600 !(list_empty(&task
->sibling
))) {
601 if (list_is_last(&task
->sibling
, &start
->children
))
603 task
= list_first_entry(&task
->sibling
,
604 struct task_struct
, sibling
);
605 pid
= get_pid(task_pid(task
));
613 * We might miss some children here if children
614 * are exited while we were not holding the lock,
615 * but it was never promised to be accurate that
618 * "Just suppose that the parent sleeps, but N children
619 * exit after we printed their tids. Now the slow paths
620 * skips N extra children, we miss N tasks." (c)
622 * So one need to stop or freeze the leader and all
623 * its children to get a precise result.
625 list_for_each_entry(task
, &start
->children
, sibling
) {
627 pid
= get_pid(task_pid(task
));
633 read_unlock(&tasklist_lock
);
637 static int children_seq_show(struct seq_file
*seq
, void *v
)
639 struct inode
*inode
= seq
->private;
642 pid
= pid_nr_ns(v
, inode
->i_sb
->s_fs_info
);
643 seq_printf(seq
, "%d ", pid
);
648 static void *children_seq_start(struct seq_file
*seq
, loff_t
*pos
)
650 return get_children_pid(seq
->private, NULL
, *pos
);
653 static void *children_seq_next(struct seq_file
*seq
, void *v
, loff_t
*pos
)
657 pid
= get_children_pid(seq
->private, v
, *pos
+ 1);
664 static void children_seq_stop(struct seq_file
*seq
, void *v
)
669 static const struct seq_operations children_seq_ops
= {
670 .start
= children_seq_start
,
671 .next
= children_seq_next
,
672 .stop
= children_seq_stop
,
673 .show
= children_seq_show
,
676 static int children_seq_open(struct inode
*inode
, struct file
*file
)
681 ret
= seq_open(file
, &children_seq_ops
);
685 m
= file
->private_data
;
691 int children_seq_release(struct inode
*inode
, struct file
*file
)
693 seq_release(inode
, file
);
697 const struct file_operations proc_tid_children_operations
= {
698 .open
= children_seq_open
,
701 .release
= children_seq_release
,
703 #endif /* CONFIG_PROC_CHILDREN */