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/user_namespace.h>
86 #include <asm/pgtable.h>
87 #include <asm/processor.h>
90 static inline void task_name(struct seq_file
*m
, struct task_struct
*p
)
95 char tcomm
[sizeof(p
->comm
)];
97 get_task_comm(tcomm
, p
);
99 seq_puts(m
, "Name:\t");
100 end
= m
->buf
+ m
->size
;
101 buf
= m
->buf
+ m
->count
;
104 while (i
&& (buf
< end
)) {
105 unsigned char c
= *name
;
125 m
->count
= buf
- m
->buf
;
130 * The task state array is a strange "bitmap" of
131 * reasons to sleep. Thus "running" is zero, and
132 * you can test for combinations of others with
135 static const char * const task_state_array
[] = {
136 "R (running)", /* 0 */
137 "S (sleeping)", /* 1 */
138 "D (disk sleep)", /* 2 */
139 "T (stopped)", /* 4 */
140 "t (tracing stop)", /* 8 */
141 "Z (zombie)", /* 16 */
144 "K (wakekill)", /* 128 */
145 "W (waking)", /* 256 */
148 static inline const char *get_task_state(struct task_struct
*tsk
)
150 unsigned int state
= (tsk
->state
& TASK_REPORT
) | tsk
->exit_state
;
151 const char * const *p
= &task_state_array
[0];
153 BUILD_BUG_ON(1 + ilog2(TASK_STATE_MAX
) != ARRAY_SIZE(task_state_array
));
162 static inline void task_state(struct seq_file
*m
, struct pid_namespace
*ns
,
163 struct pid
*pid
, struct task_struct
*p
)
165 struct user_namespace
*user_ns
= current_user_ns();
166 struct group_info
*group_info
;
168 struct fdtable
*fdt
= NULL
;
169 const struct cred
*cred
;
173 ppid
= pid_alive(p
) ?
174 task_tgid_nr_ns(rcu_dereference(p
->real_parent
), ns
) : 0;
177 struct task_struct
*tracer
= ptrace_parent(p
);
179 tpid
= task_pid_nr_ns(tracer
, ns
);
181 cred
= get_task_cred(p
);
188 "Uid:\t%d\t%d\t%d\t%d\n"
189 "Gid:\t%d\t%d\t%d\t%d\n",
191 task_tgid_nr_ns(p
, ns
),
194 from_kuid_munged(user_ns
, cred
->uid
),
195 from_kuid_munged(user_ns
, cred
->euid
),
196 from_kuid_munged(user_ns
, cred
->suid
),
197 from_kuid_munged(user_ns
, cred
->fsuid
),
198 from_kgid_munged(user_ns
, cred
->gid
),
199 from_kgid_munged(user_ns
, cred
->egid
),
200 from_kgid_munged(user_ns
, cred
->sgid
),
201 from_kgid_munged(user_ns
, cred
->fsgid
));
205 fdt
= files_fdtable(p
->files
);
209 fdt
? fdt
->max_fds
: 0);
212 group_info
= cred
->group_info
;
215 for (g
= 0; g
< min(group_info
->ngroups
, NGROUPS_SMALL
); g
++)
217 from_kgid_munged(user_ns
, GROUP_AT(group_info
, g
)));
223 static void render_sigset_t(struct seq_file
*m
, const char *header
,
235 if (sigismember(set
, i
+1)) x
|= 1;
236 if (sigismember(set
, i
+2)) x
|= 2;
237 if (sigismember(set
, i
+3)) x
|= 4;
238 if (sigismember(set
, i
+4)) x
|= 8;
239 seq_printf(m
, "%x", x
);
245 static void collect_sigign_sigcatch(struct task_struct
*p
, sigset_t
*ign
,
248 struct k_sigaction
*k
;
251 k
= p
->sighand
->action
;
252 for (i
= 1; i
<= _NSIG
; ++i
, ++k
) {
253 if (k
->sa
.sa_handler
== SIG_IGN
)
255 else if (k
->sa
.sa_handler
!= SIG_DFL
)
260 static inline void task_sig(struct seq_file
*m
, struct task_struct
*p
)
263 sigset_t pending
, shpending
, blocked
, ignored
, caught
;
265 unsigned long qsize
= 0;
266 unsigned long qlim
= 0;
268 sigemptyset(&pending
);
269 sigemptyset(&shpending
);
270 sigemptyset(&blocked
);
271 sigemptyset(&ignored
);
272 sigemptyset(&caught
);
274 if (lock_task_sighand(p
, &flags
)) {
275 pending
= p
->pending
.signal
;
276 shpending
= p
->signal
->shared_pending
.signal
;
277 blocked
= p
->blocked
;
278 collect_sigign_sigcatch(p
, &ignored
, &caught
);
279 num_threads
= get_nr_threads(p
);
280 rcu_read_lock(); /* FIXME: is this correct? */
281 qsize
= atomic_read(&__task_cred(p
)->user
->sigpending
);
283 qlim
= task_rlimit(p
, RLIMIT_SIGPENDING
);
284 unlock_task_sighand(p
, &flags
);
287 seq_printf(m
, "Threads:\t%d\n", num_threads
);
288 seq_printf(m
, "SigQ:\t%lu/%lu\n", qsize
, qlim
);
290 /* render them all */
291 render_sigset_t(m
, "SigPnd:\t", &pending
);
292 render_sigset_t(m
, "ShdPnd:\t", &shpending
);
293 render_sigset_t(m
, "SigBlk:\t", &blocked
);
294 render_sigset_t(m
, "SigIgn:\t", &ignored
);
295 render_sigset_t(m
, "SigCgt:\t", &caught
);
298 static void render_cap_t(struct seq_file
*m
, const char *header
,
304 CAP_FOR_EACH_U32(__capi
) {
305 seq_printf(m
, "%08x",
306 a
->cap
[(_KERNEL_CAPABILITY_U32S
-1) - __capi
]);
311 static inline void task_cap(struct seq_file
*m
, struct task_struct
*p
)
313 const struct cred
*cred
;
314 kernel_cap_t cap_inheritable
, cap_permitted
, cap_effective
, cap_bset
;
317 cred
= __task_cred(p
);
318 cap_inheritable
= cred
->cap_inheritable
;
319 cap_permitted
= cred
->cap_permitted
;
320 cap_effective
= cred
->cap_effective
;
321 cap_bset
= cred
->cap_bset
;
324 render_cap_t(m
, "CapInh:\t", &cap_inheritable
);
325 render_cap_t(m
, "CapPrm:\t", &cap_permitted
);
326 render_cap_t(m
, "CapEff:\t", &cap_effective
);
327 render_cap_t(m
, "CapBnd:\t", &cap_bset
);
330 static inline void task_context_switch_counts(struct seq_file
*m
,
331 struct task_struct
*p
)
333 seq_printf(m
, "voluntary_ctxt_switches:\t%lu\n"
334 "nonvoluntary_ctxt_switches:\t%lu\n",
339 static void task_cpus_allowed(struct seq_file
*m
, struct task_struct
*task
)
341 seq_puts(m
, "Cpus_allowed:\t");
342 seq_cpumask(m
, &task
->cpus_allowed
);
344 seq_puts(m
, "Cpus_allowed_list:\t");
345 seq_cpumask_list(m
, &task
->cpus_allowed
);
349 int proc_pid_status(struct seq_file
*m
, struct pid_namespace
*ns
,
350 struct pid
*pid
, struct task_struct
*task
)
352 struct mm_struct
*mm
= get_task_mm(task
);
355 task_state(m
, ns
, pid
, task
);
363 task_cpus_allowed(m
, task
);
364 cpuset_task_status_allowed(m
, task
);
365 task_context_switch_counts(m
, task
);
369 static int do_task_stat(struct seq_file
*m
, struct pid_namespace
*ns
,
370 struct pid
*pid
, struct task_struct
*task
, int whole
)
372 unsigned long vsize
, eip
, esp
, wchan
= ~0UL;
374 int tty_pgrp
= -1, tty_nr
= 0;
375 sigset_t sigign
, sigcatch
;
377 pid_t ppid
= 0, pgid
= -1, sid
= -1;
380 struct mm_struct
*mm
;
381 unsigned long long start_time
;
382 unsigned long cmin_flt
= 0, cmaj_flt
= 0;
383 unsigned long min_flt
= 0, maj_flt
= 0;
384 cputime_t cutime
, cstime
, utime
, stime
;
385 cputime_t cgtime
, gtime
;
386 unsigned long rsslim
= 0;
387 char tcomm
[sizeof(task
->comm
)];
390 state
= *get_task_state(task
);
391 vsize
= eip
= esp
= 0;
392 permitted
= ptrace_may_access(task
, PTRACE_MODE_READ
| PTRACE_MODE_NOAUDIT
);
393 mm
= get_task_mm(task
);
395 vsize
= task_vsize(mm
);
397 eip
= KSTK_EIP(task
);
398 esp
= KSTK_ESP(task
);
402 get_task_comm(tcomm
, task
);
404 sigemptyset(&sigign
);
405 sigemptyset(&sigcatch
);
406 cutime
= cstime
= utime
= stime
= 0;
409 if (lock_task_sighand(task
, &flags
)) {
410 struct signal_struct
*sig
= task
->signal
;
413 struct pid
*pgrp
= tty_get_pgrp(sig
->tty
);
414 tty_pgrp
= pid_nr_ns(pgrp
, ns
);
416 tty_nr
= new_encode_dev(tty_devnum(sig
->tty
));
419 num_threads
= get_nr_threads(task
);
420 collect_sigign_sigcatch(task
, &sigign
, &sigcatch
);
422 cmin_flt
= sig
->cmin_flt
;
423 cmaj_flt
= sig
->cmaj_flt
;
424 cutime
= sig
->cutime
;
425 cstime
= sig
->cstime
;
426 cgtime
= sig
->cgtime
;
427 rsslim
= ACCESS_ONCE(sig
->rlim
[RLIMIT_RSS
].rlim_cur
);
429 /* add up live thread stats at the group level */
431 struct task_struct
*t
= task
;
433 min_flt
+= t
->min_flt
;
434 maj_flt
+= t
->maj_flt
;
439 min_flt
+= sig
->min_flt
;
440 maj_flt
+= sig
->maj_flt
;
441 thread_group_times(task
, &utime
, &stime
);
445 sid
= task_session_nr_ns(task
, ns
);
446 ppid
= task_tgid_nr_ns(task
->real_parent
, ns
);
447 pgid
= task_pgrp_nr_ns(task
, ns
);
449 unlock_task_sighand(task
, &flags
);
452 if (permitted
&& (!whole
|| num_threads
< 2))
453 wchan
= get_wchan(task
);
455 min_flt
= task
->min_flt
;
456 maj_flt
= task
->maj_flt
;
457 task_times(task
, &utime
, &stime
);
461 /* scale priority and nice values from timeslices to -20..20 */
462 /* to make it look like a "normal" Unix priority/nice value */
463 priority
= task_prio(task
);
464 nice
= task_nice(task
);
466 /* Temporary variable needed for gcc-2.96 */
467 /* convert timespec -> nsec*/
469 (unsigned long long)task
->real_start_time
.tv_sec
* NSEC_PER_SEC
470 + task
->real_start_time
.tv_nsec
;
471 /* convert nsec -> ticks */
472 start_time
= nsec_to_clock_t(start_time
);
474 seq_printf(m
, "%d (%s) %c", pid_nr_ns(pid
, ns
), tcomm
, state
);
475 seq_put_decimal_ll(m
, ' ', ppid
);
476 seq_put_decimal_ll(m
, ' ', pgid
);
477 seq_put_decimal_ll(m
, ' ', sid
);
478 seq_put_decimal_ll(m
, ' ', tty_nr
);
479 seq_put_decimal_ll(m
, ' ', tty_pgrp
);
480 seq_put_decimal_ull(m
, ' ', task
->flags
);
481 seq_put_decimal_ull(m
, ' ', min_flt
);
482 seq_put_decimal_ull(m
, ' ', cmin_flt
);
483 seq_put_decimal_ull(m
, ' ', maj_flt
);
484 seq_put_decimal_ull(m
, ' ', cmaj_flt
);
485 seq_put_decimal_ull(m
, ' ', cputime_to_clock_t(utime
));
486 seq_put_decimal_ull(m
, ' ', cputime_to_clock_t(stime
));
487 seq_put_decimal_ll(m
, ' ', cputime_to_clock_t(cutime
));
488 seq_put_decimal_ll(m
, ' ', cputime_to_clock_t(cstime
));
489 seq_put_decimal_ll(m
, ' ', priority
);
490 seq_put_decimal_ll(m
, ' ', nice
);
491 seq_put_decimal_ll(m
, ' ', num_threads
);
492 seq_put_decimal_ull(m
, ' ', 0);
493 seq_put_decimal_ull(m
, ' ', start_time
);
494 seq_put_decimal_ull(m
, ' ', vsize
);
495 seq_put_decimal_ull(m
, ' ', mm
? get_mm_rss(mm
) : 0);
496 seq_put_decimal_ull(m
, ' ', rsslim
);
497 seq_put_decimal_ull(m
, ' ', mm
? (permitted
? mm
->start_code
: 1) : 0);
498 seq_put_decimal_ull(m
, ' ', mm
? (permitted
? mm
->end_code
: 1) : 0);
499 seq_put_decimal_ull(m
, ' ', (permitted
&& mm
) ? mm
->start_stack
: 0);
500 seq_put_decimal_ull(m
, ' ', esp
);
501 seq_put_decimal_ull(m
, ' ', eip
);
502 /* The signal information here is obsolete.
503 * It must be decimal for Linux 2.0 compatibility.
504 * Use /proc/#/status for real-time signals.
506 seq_put_decimal_ull(m
, ' ', task
->pending
.signal
.sig
[0] & 0x7fffffffUL
);
507 seq_put_decimal_ull(m
, ' ', task
->blocked
.sig
[0] & 0x7fffffffUL
);
508 seq_put_decimal_ull(m
, ' ', sigign
.sig
[0] & 0x7fffffffUL
);
509 seq_put_decimal_ull(m
, ' ', sigcatch
.sig
[0] & 0x7fffffffUL
);
510 seq_put_decimal_ull(m
, ' ', wchan
);
511 seq_put_decimal_ull(m
, ' ', 0);
512 seq_put_decimal_ull(m
, ' ', 0);
513 seq_put_decimal_ll(m
, ' ', task
->exit_signal
);
514 seq_put_decimal_ll(m
, ' ', task_cpu(task
));
515 seq_put_decimal_ull(m
, ' ', task
->rt_priority
);
516 seq_put_decimal_ull(m
, ' ', task
->policy
);
517 seq_put_decimal_ull(m
, ' ', delayacct_blkio_ticks(task
));
518 seq_put_decimal_ull(m
, ' ', cputime_to_clock_t(gtime
));
519 seq_put_decimal_ll(m
, ' ', cputime_to_clock_t(cgtime
));
521 if (mm
&& permitted
) {
522 seq_put_decimal_ull(m
, ' ', mm
->start_data
);
523 seq_put_decimal_ull(m
, ' ', mm
->end_data
);
524 seq_put_decimal_ull(m
, ' ', mm
->start_brk
);
525 seq_put_decimal_ull(m
, ' ', mm
->arg_start
);
526 seq_put_decimal_ull(m
, ' ', mm
->arg_end
);
527 seq_put_decimal_ull(m
, ' ', mm
->env_start
);
528 seq_put_decimal_ull(m
, ' ', mm
->env_end
);
530 seq_printf(m
, " 0 0 0 0 0 0 0");
533 seq_put_decimal_ll(m
, ' ', task
->exit_code
);
535 seq_put_decimal_ll(m
, ' ', 0);
543 int proc_tid_stat(struct seq_file
*m
, struct pid_namespace
*ns
,
544 struct pid
*pid
, struct task_struct
*task
)
546 return do_task_stat(m
, ns
, pid
, task
, 0);
549 int proc_tgid_stat(struct seq_file
*m
, struct pid_namespace
*ns
,
550 struct pid
*pid
, struct task_struct
*task
)
552 return do_task_stat(m
, ns
, pid
, task
, 1);
555 int proc_pid_statm(struct seq_file
*m
, struct pid_namespace
*ns
,
556 struct pid
*pid
, struct task_struct
*task
)
558 unsigned long size
= 0, resident
= 0, shared
= 0, text
= 0, data
= 0;
559 struct mm_struct
*mm
= get_task_mm(task
);
562 size
= task_statm(mm
, &shared
, &text
, &data
, &resident
);
566 * For quick read, open code by putting numbers directly
568 * seq_printf(m, "%lu %lu %lu %lu 0 %lu 0\n",
569 * size, resident, shared, text, data);
571 seq_put_decimal_ull(m
, 0, size
);
572 seq_put_decimal_ull(m
, ' ', resident
);
573 seq_put_decimal_ull(m
, ' ', shared
);
574 seq_put_decimal_ull(m
, ' ', text
);
575 seq_put_decimal_ull(m
, ' ', 0);
576 seq_put_decimal_ull(m
, ' ', data
);
577 seq_put_decimal_ull(m
, ' ', 0);
583 #ifdef CONFIG_CHECKPOINT_RESTORE
585 get_children_pid(struct inode
*inode
, struct pid
*pid_prev
, loff_t pos
)
587 struct task_struct
*start
, *task
;
588 struct pid
*pid
= NULL
;
590 read_lock(&tasklist_lock
);
592 start
= pid_task(proc_pid(inode
), PIDTYPE_PID
);
597 * Lets try to continue searching first, this gives
598 * us significant speedup on children-rich processes.
601 task
= pid_task(pid_prev
, PIDTYPE_PID
);
602 if (task
&& task
->real_parent
== start
&&
603 !(list_empty(&task
->sibling
))) {
604 if (list_is_last(&task
->sibling
, &start
->children
))
606 task
= list_first_entry(&task
->sibling
,
607 struct task_struct
, sibling
);
608 pid
= get_pid(task_pid(task
));
616 * We might miss some children here if children
617 * are exited while we were not holding the lock,
618 * but it was never promised to be accurate that
621 * "Just suppose that the parent sleeps, but N children
622 * exit after we printed their tids. Now the slow paths
623 * skips N extra children, we miss N tasks." (c)
625 * So one need to stop or freeze the leader and all
626 * its children to get a precise result.
628 list_for_each_entry(task
, &start
->children
, sibling
) {
630 pid
= get_pid(task_pid(task
));
636 read_unlock(&tasklist_lock
);
640 static int children_seq_show(struct seq_file
*seq
, void *v
)
642 struct inode
*inode
= seq
->private;
645 pid
= pid_nr_ns(v
, inode
->i_sb
->s_fs_info
);
646 return seq_printf(seq
, "%d ", pid
);
649 static void *children_seq_start(struct seq_file
*seq
, loff_t
*pos
)
651 return get_children_pid(seq
->private, NULL
, *pos
);
654 static void *children_seq_next(struct seq_file
*seq
, void *v
, loff_t
*pos
)
658 pid
= get_children_pid(seq
->private, v
, *pos
+ 1);
665 static void children_seq_stop(struct seq_file
*seq
, void *v
)
670 static const struct seq_operations children_seq_ops
= {
671 .start
= children_seq_start
,
672 .next
= children_seq_next
,
673 .stop
= children_seq_stop
,
674 .show
= children_seq_show
,
677 static int children_seq_open(struct inode
*inode
, struct file
*file
)
682 ret
= seq_open(file
, &children_seq_ops
);
686 m
= file
->private_data
;
692 int children_seq_release(struct inode
*inode
, struct file
*file
)
694 seq_release(inode
, file
);
698 const struct file_operations proc_tid_children_operations
= {
699 .open
= children_seq_open
,
702 .release
= children_seq_release
,
704 #endif /* CONFIG_CHECKPOINT_RESTORE */