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>
86 #include <linux/fs_struct.h>
88 #include <asm/pgtable.h>
89 #include <asm/processor.h>
92 static inline void task_name(struct seq_file
*m
, struct task_struct
*p
)
96 char tcomm
[sizeof(p
->comm
)];
99 get_task_comm(tcomm
, p
);
101 seq_puts(m
, "Name:\t");
103 size
= seq_get_buf(m
, &buf
);
104 ret
= string_escape_str(tcomm
, buf
, size
, ESCAPE_SPACE
| ESCAPE_SPECIAL
, "\n\\");
105 seq_commit(m
, ret
< size
? ret
: -1);
111 * The task state array is a strange "bitmap" of
112 * reasons to sleep. Thus "running" is zero, and
113 * you can test for combinations of others with
116 static const char * const task_state_array
[] = {
117 "R (running)", /* 0 */
118 "S (sleeping)", /* 1 */
119 "D (disk sleep)", /* 2 */
120 "T (stopped)", /* 4 */
121 "t (tracing stop)", /* 8 */
123 "Z (zombie)", /* 32 */
126 static inline const char *get_task_state(struct task_struct
*tsk
)
128 unsigned int state
= (tsk
->state
| tsk
->exit_state
) & TASK_REPORT
;
131 * Parked tasks do not run; they sit in __kthread_parkme().
132 * Without this check, we would report them as running, which is
133 * clearly wrong, so we report them as sleeping instead.
135 if (tsk
->state
== TASK_PARKED
)
136 state
= TASK_INTERRUPTIBLE
;
138 BUILD_BUG_ON(1 + ilog2(TASK_REPORT
) != ARRAY_SIZE(task_state_array
)-1);
140 return task_state_array
[fls(state
)];
143 static inline int get_task_umask(struct task_struct
*tsk
)
145 struct fs_struct
*fs
;
156 static inline void task_state(struct seq_file
*m
, struct pid_namespace
*ns
,
157 struct pid
*pid
, struct task_struct
*p
)
159 struct user_namespace
*user_ns
= seq_user_ns(m
);
160 struct group_info
*group_info
;
162 struct task_struct
*tracer
;
163 const struct cred
*cred
;
164 pid_t ppid
, tpid
= 0, tgid
, ngid
;
165 unsigned int max_fds
= 0;
168 ppid
= pid_alive(p
) ?
169 task_tgid_nr_ns(rcu_dereference(p
->real_parent
), ns
) : 0;
171 tracer
= ptrace_parent(p
);
173 tpid
= task_pid_nr_ns(tracer
, ns
);
175 tgid
= task_tgid_nr_ns(p
, ns
);
176 ngid
= task_numa_group_id(p
);
177 cred
= get_task_cred(p
);
179 umask
= get_task_umask(p
);
181 seq_printf(m
, "Umask:\t%#04o\n", umask
);
185 max_fds
= files_fdtable(p
->files
)->max_fds
;
196 "Uid:\t%d\t%d\t%d\t%d\n"
197 "Gid:\t%d\t%d\t%d\t%d\n"
198 "FDSize:\t%d\nGroups:\t",
200 tgid
, ngid
, pid_nr_ns(pid
, ns
), ppid
, tpid
,
201 from_kuid_munged(user_ns
, cred
->uid
),
202 from_kuid_munged(user_ns
, cred
->euid
),
203 from_kuid_munged(user_ns
, cred
->suid
),
204 from_kuid_munged(user_ns
, cred
->fsuid
),
205 from_kgid_munged(user_ns
, cred
->gid
),
206 from_kgid_munged(user_ns
, cred
->egid
),
207 from_kgid_munged(user_ns
, cred
->sgid
),
208 from_kgid_munged(user_ns
, cred
->fsgid
),
211 group_info
= cred
->group_info
;
212 for (g
= 0; g
< group_info
->ngroups
; g
++)
214 from_kgid_munged(user_ns
, GROUP_AT(group_info
, g
)));
218 seq_puts(m
, "\nNStgid:");
219 for (g
= ns
->level
; g
<= pid
->level
; g
++)
220 seq_printf(m
, "\t%d",
221 task_tgid_nr_ns(p
, pid
->numbers
[g
].ns
));
222 seq_puts(m
, "\nNSpid:");
223 for (g
= ns
->level
; g
<= pid
->level
; g
++)
224 seq_printf(m
, "\t%d",
225 task_pid_nr_ns(p
, pid
->numbers
[g
].ns
));
226 seq_puts(m
, "\nNSpgid:");
227 for (g
= ns
->level
; g
<= pid
->level
; g
++)
228 seq_printf(m
, "\t%d",
229 task_pgrp_nr_ns(p
, pid
->numbers
[g
].ns
));
230 seq_puts(m
, "\nNSsid:");
231 for (g
= ns
->level
; g
<= pid
->level
; g
++)
232 seq_printf(m
, "\t%d",
233 task_session_nr_ns(p
, pid
->numbers
[g
].ns
));
238 void render_sigset_t(struct seq_file
*m
, const char *header
,
250 if (sigismember(set
, i
+1)) x
|= 1;
251 if (sigismember(set
, i
+2)) x
|= 2;
252 if (sigismember(set
, i
+3)) x
|= 4;
253 if (sigismember(set
, i
+4)) x
|= 8;
254 seq_printf(m
, "%x", x
);
260 static void collect_sigign_sigcatch(struct task_struct
*p
, sigset_t
*ign
,
263 struct k_sigaction
*k
;
266 k
= p
->sighand
->action
;
267 for (i
= 1; i
<= _NSIG
; ++i
, ++k
) {
268 if (k
->sa
.sa_handler
== SIG_IGN
)
270 else if (k
->sa
.sa_handler
!= SIG_DFL
)
275 static inline void task_sig(struct seq_file
*m
, struct task_struct
*p
)
278 sigset_t pending
, shpending
, blocked
, ignored
, caught
;
280 unsigned long qsize
= 0;
281 unsigned long qlim
= 0;
283 sigemptyset(&pending
);
284 sigemptyset(&shpending
);
285 sigemptyset(&blocked
);
286 sigemptyset(&ignored
);
287 sigemptyset(&caught
);
289 if (lock_task_sighand(p
, &flags
)) {
290 pending
= p
->pending
.signal
;
291 shpending
= p
->signal
->shared_pending
.signal
;
292 blocked
= p
->blocked
;
293 collect_sigign_sigcatch(p
, &ignored
, &caught
);
294 num_threads
= get_nr_threads(p
);
295 rcu_read_lock(); /* FIXME: is this correct? */
296 qsize
= atomic_read(&__task_cred(p
)->user
->sigpending
);
298 qlim
= task_rlimit(p
, RLIMIT_SIGPENDING
);
299 unlock_task_sighand(p
, &flags
);
302 seq_printf(m
, "Threads:\t%d\n", num_threads
);
303 seq_printf(m
, "SigQ:\t%lu/%lu\n", qsize
, qlim
);
305 /* render them all */
306 render_sigset_t(m
, "SigPnd:\t", &pending
);
307 render_sigset_t(m
, "ShdPnd:\t", &shpending
);
308 render_sigset_t(m
, "SigBlk:\t", &blocked
);
309 render_sigset_t(m
, "SigIgn:\t", &ignored
);
310 render_sigset_t(m
, "SigCgt:\t", &caught
);
313 static void render_cap_t(struct seq_file
*m
, const char *header
,
319 CAP_FOR_EACH_U32(__capi
) {
320 seq_printf(m
, "%08x",
321 a
->cap
[CAP_LAST_U32
- __capi
]);
326 static inline void task_cap(struct seq_file
*m
, struct task_struct
*p
)
328 const struct cred
*cred
;
329 kernel_cap_t cap_inheritable
, cap_permitted
, cap_effective
,
330 cap_bset
, cap_ambient
;
333 cred
= __task_cred(p
);
334 cap_inheritable
= cred
->cap_inheritable
;
335 cap_permitted
= cred
->cap_permitted
;
336 cap_effective
= cred
->cap_effective
;
337 cap_bset
= cred
->cap_bset
;
338 cap_ambient
= cred
->cap_ambient
;
341 render_cap_t(m
, "CapInh:\t", &cap_inheritable
);
342 render_cap_t(m
, "CapPrm:\t", &cap_permitted
);
343 render_cap_t(m
, "CapEff:\t", &cap_effective
);
344 render_cap_t(m
, "CapBnd:\t", &cap_bset
);
345 render_cap_t(m
, "CapAmb:\t", &cap_ambient
);
348 static inline void task_seccomp(struct seq_file
*m
, struct task_struct
*p
)
350 #ifdef CONFIG_SECCOMP
351 seq_printf(m
, "Seccomp:\t%d\n", p
->seccomp
.mode
);
355 static inline void task_context_switch_counts(struct seq_file
*m
,
356 struct task_struct
*p
)
358 seq_printf(m
, "voluntary_ctxt_switches:\t%lu\n"
359 "nonvoluntary_ctxt_switches:\t%lu\n",
364 static void task_cpus_allowed(struct seq_file
*m
, struct task_struct
*task
)
366 seq_printf(m
, "Cpus_allowed:\t%*pb\n",
367 cpumask_pr_args(&task
->cpus_allowed
));
368 seq_printf(m
, "Cpus_allowed_list:\t%*pbl\n",
369 cpumask_pr_args(&task
->cpus_allowed
));
372 int proc_pid_status(struct seq_file
*m
, struct pid_namespace
*ns
,
373 struct pid
*pid
, struct task_struct
*task
)
375 struct mm_struct
*mm
= get_task_mm(task
);
378 task_state(m
, ns
, pid
, task
);
386 task_seccomp(m
, task
);
387 task_cpus_allowed(m
, task
);
388 cpuset_task_status_allowed(m
, task
);
389 task_context_switch_counts(m
, task
);
393 static int do_task_stat(struct seq_file
*m
, struct pid_namespace
*ns
,
394 struct pid
*pid
, struct task_struct
*task
, int whole
)
396 unsigned long vsize
, eip
, esp
, wchan
= 0;
398 int tty_pgrp
= -1, tty_nr
= 0;
399 sigset_t sigign
, sigcatch
;
401 pid_t ppid
= 0, pgid
= -1, sid
= -1;
404 struct mm_struct
*mm
;
405 unsigned long long start_time
;
406 unsigned long cmin_flt
= 0, cmaj_flt
= 0;
407 unsigned long min_flt
= 0, maj_flt
= 0;
408 cputime_t cutime
, cstime
, utime
, stime
;
409 cputime_t cgtime
, gtime
;
410 unsigned long rsslim
= 0;
411 char tcomm
[sizeof(task
->comm
)];
414 state
= *get_task_state(task
);
415 vsize
= eip
= esp
= 0;
416 permitted
= ptrace_may_access(task
, PTRACE_MODE_READ_FSCREDS
| PTRACE_MODE_NOAUDIT
);
417 mm
= get_task_mm(task
);
419 vsize
= task_vsize(mm
);
421 eip
= KSTK_EIP(task
);
422 esp
= KSTK_ESP(task
);
426 get_task_comm(tcomm
, task
);
428 sigemptyset(&sigign
);
429 sigemptyset(&sigcatch
);
430 cutime
= cstime
= utime
= stime
= 0;
433 if (lock_task_sighand(task
, &flags
)) {
434 struct signal_struct
*sig
= task
->signal
;
437 struct pid
*pgrp
= tty_get_pgrp(sig
->tty
);
438 tty_pgrp
= pid_nr_ns(pgrp
, ns
);
440 tty_nr
= new_encode_dev(tty_devnum(sig
->tty
));
443 num_threads
= get_nr_threads(task
);
444 collect_sigign_sigcatch(task
, &sigign
, &sigcatch
);
446 cmin_flt
= sig
->cmin_flt
;
447 cmaj_flt
= sig
->cmaj_flt
;
448 cutime
= sig
->cutime
;
449 cstime
= sig
->cstime
;
450 cgtime
= sig
->cgtime
;
451 rsslim
= ACCESS_ONCE(sig
->rlim
[RLIMIT_RSS
].rlim_cur
);
453 /* add up live thread stats at the group level */
455 struct task_struct
*t
= task
;
457 min_flt
+= t
->min_flt
;
458 maj_flt
+= t
->maj_flt
;
459 gtime
+= task_gtime(t
);
460 } while_each_thread(task
, t
);
462 min_flt
+= sig
->min_flt
;
463 maj_flt
+= sig
->maj_flt
;
464 thread_group_cputime_adjusted(task
, &utime
, &stime
);
468 sid
= task_session_nr_ns(task
, ns
);
469 ppid
= task_tgid_nr_ns(task
->real_parent
, ns
);
470 pgid
= task_pgrp_nr_ns(task
, ns
);
472 unlock_task_sighand(task
, &flags
);
475 if (permitted
&& (!whole
|| num_threads
< 2))
476 wchan
= get_wchan(task
);
478 min_flt
= task
->min_flt
;
479 maj_flt
= task
->maj_flt
;
480 task_cputime_adjusted(task
, &utime
, &stime
);
481 gtime
= task_gtime(task
);
484 /* scale priority and nice values from timeslices to -20..20 */
485 /* to make it look like a "normal" Unix priority/nice value */
486 priority
= task_prio(task
);
487 nice
= task_nice(task
);
489 /* convert nsec -> ticks */
490 start_time
= nsec_to_clock_t(task
->real_start_time
);
492 seq_printf(m
, "%d (%s) %c", pid_nr_ns(pid
, ns
), tcomm
, state
);
493 seq_put_decimal_ll(m
, ' ', ppid
);
494 seq_put_decimal_ll(m
, ' ', pgid
);
495 seq_put_decimal_ll(m
, ' ', sid
);
496 seq_put_decimal_ll(m
, ' ', tty_nr
);
497 seq_put_decimal_ll(m
, ' ', tty_pgrp
);
498 seq_put_decimal_ull(m
, ' ', task
->flags
);
499 seq_put_decimal_ull(m
, ' ', min_flt
);
500 seq_put_decimal_ull(m
, ' ', cmin_flt
);
501 seq_put_decimal_ull(m
, ' ', maj_flt
);
502 seq_put_decimal_ull(m
, ' ', cmaj_flt
);
503 seq_put_decimal_ull(m
, ' ', cputime_to_clock_t(utime
));
504 seq_put_decimal_ull(m
, ' ', cputime_to_clock_t(stime
));
505 seq_put_decimal_ll(m
, ' ', cputime_to_clock_t(cutime
));
506 seq_put_decimal_ll(m
, ' ', cputime_to_clock_t(cstime
));
507 seq_put_decimal_ll(m
, ' ', priority
);
508 seq_put_decimal_ll(m
, ' ', nice
);
509 seq_put_decimal_ll(m
, ' ', num_threads
);
510 seq_put_decimal_ull(m
, ' ', 0);
511 seq_put_decimal_ull(m
, ' ', start_time
);
512 seq_put_decimal_ull(m
, ' ', vsize
);
513 seq_put_decimal_ull(m
, ' ', mm
? get_mm_rss(mm
) : 0);
514 seq_put_decimal_ull(m
, ' ', rsslim
);
515 seq_put_decimal_ull(m
, ' ', mm
? (permitted
? mm
->start_code
: 1) : 0);
516 seq_put_decimal_ull(m
, ' ', mm
? (permitted
? mm
->end_code
: 1) : 0);
517 seq_put_decimal_ull(m
, ' ', (permitted
&& mm
) ? mm
->start_stack
: 0);
518 seq_put_decimal_ull(m
, ' ', esp
);
519 seq_put_decimal_ull(m
, ' ', eip
);
520 /* The signal information here is obsolete.
521 * It must be decimal for Linux 2.0 compatibility.
522 * Use /proc/#/status for real-time signals.
524 seq_put_decimal_ull(m
, ' ', task
->pending
.signal
.sig
[0] & 0x7fffffffUL
);
525 seq_put_decimal_ull(m
, ' ', task
->blocked
.sig
[0] & 0x7fffffffUL
);
526 seq_put_decimal_ull(m
, ' ', sigign
.sig
[0] & 0x7fffffffUL
);
527 seq_put_decimal_ull(m
, ' ', sigcatch
.sig
[0] & 0x7fffffffUL
);
530 * We used to output the absolute kernel address, but that's an
531 * information leak - so instead we show a 0/1 flag here, to signal
532 * to user-space whether there's a wchan field in /proc/PID/wchan.
534 * This works with older implementations of procps as well.
541 seq_put_decimal_ull(m
, ' ', 0);
542 seq_put_decimal_ull(m
, ' ', 0);
543 seq_put_decimal_ll(m
, ' ', task
->exit_signal
);
544 seq_put_decimal_ll(m
, ' ', task_cpu(task
));
545 seq_put_decimal_ull(m
, ' ', task
->rt_priority
);
546 seq_put_decimal_ull(m
, ' ', task
->policy
);
547 seq_put_decimal_ull(m
, ' ', delayacct_blkio_ticks(task
));
548 seq_put_decimal_ull(m
, ' ', cputime_to_clock_t(gtime
));
549 seq_put_decimal_ll(m
, ' ', cputime_to_clock_t(cgtime
));
551 if (mm
&& permitted
) {
552 seq_put_decimal_ull(m
, ' ', mm
->start_data
);
553 seq_put_decimal_ull(m
, ' ', mm
->end_data
);
554 seq_put_decimal_ull(m
, ' ', mm
->start_brk
);
555 seq_put_decimal_ull(m
, ' ', mm
->arg_start
);
556 seq_put_decimal_ull(m
, ' ', mm
->arg_end
);
557 seq_put_decimal_ull(m
, ' ', mm
->env_start
);
558 seq_put_decimal_ull(m
, ' ', mm
->env_end
);
560 seq_printf(m
, " 0 0 0 0 0 0 0");
563 seq_put_decimal_ll(m
, ' ', task
->exit_code
);
565 seq_put_decimal_ll(m
, ' ', 0);
573 int proc_tid_stat(struct seq_file
*m
, struct pid_namespace
*ns
,
574 struct pid
*pid
, struct task_struct
*task
)
576 return do_task_stat(m
, ns
, pid
, task
, 0);
579 int proc_tgid_stat(struct seq_file
*m
, struct pid_namespace
*ns
,
580 struct pid
*pid
, struct task_struct
*task
)
582 return do_task_stat(m
, ns
, pid
, task
, 1);
585 int proc_pid_statm(struct seq_file
*m
, struct pid_namespace
*ns
,
586 struct pid
*pid
, struct task_struct
*task
)
588 unsigned long size
= 0, resident
= 0, shared
= 0, text
= 0, data
= 0;
589 struct mm_struct
*mm
= get_task_mm(task
);
592 size
= task_statm(mm
, &shared
, &text
, &data
, &resident
);
596 * For quick read, open code by putting numbers directly
598 * seq_printf(m, "%lu %lu %lu %lu 0 %lu 0\n",
599 * size, resident, shared, text, data);
601 seq_put_decimal_ull(m
, 0, size
);
602 seq_put_decimal_ull(m
, ' ', resident
);
603 seq_put_decimal_ull(m
, ' ', shared
);
604 seq_put_decimal_ull(m
, ' ', text
);
605 seq_put_decimal_ull(m
, ' ', 0);
606 seq_put_decimal_ull(m
, ' ', data
);
607 seq_put_decimal_ull(m
, ' ', 0);
613 #ifdef CONFIG_PROC_CHILDREN
615 get_children_pid(struct inode
*inode
, struct pid
*pid_prev
, loff_t pos
)
617 struct task_struct
*start
, *task
;
618 struct pid
*pid
= NULL
;
620 read_lock(&tasklist_lock
);
622 start
= pid_task(proc_pid(inode
), PIDTYPE_PID
);
627 * Lets try to continue searching first, this gives
628 * us significant speedup on children-rich processes.
631 task
= pid_task(pid_prev
, PIDTYPE_PID
);
632 if (task
&& task
->real_parent
== start
&&
633 !(list_empty(&task
->sibling
))) {
634 if (list_is_last(&task
->sibling
, &start
->children
))
636 task
= list_first_entry(&task
->sibling
,
637 struct task_struct
, sibling
);
638 pid
= get_pid(task_pid(task
));
646 * We might miss some children here if children
647 * are exited while we were not holding the lock,
648 * but it was never promised to be accurate that
651 * "Just suppose that the parent sleeps, but N children
652 * exit after we printed their tids. Now the slow paths
653 * skips N extra children, we miss N tasks." (c)
655 * So one need to stop or freeze the leader and all
656 * its children to get a precise result.
658 list_for_each_entry(task
, &start
->children
, sibling
) {
660 pid
= get_pid(task_pid(task
));
666 read_unlock(&tasklist_lock
);
670 static int children_seq_show(struct seq_file
*seq
, void *v
)
672 struct inode
*inode
= seq
->private;
675 pid
= pid_nr_ns(v
, inode
->i_sb
->s_fs_info
);
676 seq_printf(seq
, "%d ", pid
);
681 static void *children_seq_start(struct seq_file
*seq
, loff_t
*pos
)
683 return get_children_pid(seq
->private, NULL
, *pos
);
686 static void *children_seq_next(struct seq_file
*seq
, void *v
, loff_t
*pos
)
690 pid
= get_children_pid(seq
->private, v
, *pos
+ 1);
697 static void children_seq_stop(struct seq_file
*seq
, void *v
)
702 static const struct seq_operations children_seq_ops
= {
703 .start
= children_seq_start
,
704 .next
= children_seq_next
,
705 .stop
= children_seq_stop
,
706 .show
= children_seq_show
,
709 static int children_seq_open(struct inode
*inode
, struct file
*file
)
714 ret
= seq_open(file
, &children_seq_ops
);
718 m
= file
->private_data
;
724 int children_seq_release(struct inode
*inode
, struct file
*file
)
726 seq_release(inode
, file
);
730 const struct file_operations proc_tid_children_operations
= {
731 .open
= children_seq_open
,
734 .release
= children_seq_release
,
736 #endif /* CONFIG_PROC_CHILDREN */