CRED: Use RCU to access another task's creds and to release a task's own creds
[linux-2.6-block.git] / fs / proc / array.c
CommitLineData
1da177e4
LT
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.
526719ba 43 * <alan@lxorguk.ukuu.org.uk>
1da177e4
LT
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
1da177e4
LT
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>
8ea02606
IM
65#include <linux/uaccess.h>
66#include <linux/io.h>
1da177e4
LT
67#include <linux/mm.h>
68#include <linux/hugetlb.h>
69#include <linux/pagemap.h>
70#include <linux/swap.h>
71#include <linux/slab.h>
72#include <linux/smp.h>
73#include <linux/signal.h>
74#include <linux/highmem.h>
75#include <linux/file.h>
9f3acc31 76#include <linux/fdtable.h>
1da177e4
LT
77#include <linux/times.h>
78#include <linux/cpuset.h>
4fb3a538 79#include <linux/rcupdate.h>
25890454 80#include <linux/delayacct.h>
ee992744 81#include <linux/seq_file.h>
b488893a 82#include <linux/pid_namespace.h>
0d094efe 83#include <linux/tracehook.h>
1da177e4 84
1da177e4 85#include <asm/pgtable.h>
1da177e4
LT
86#include <asm/processor.h>
87#include "internal.h"
88
df5f8314 89static inline void task_name(struct seq_file *m, struct task_struct *p)
1da177e4
LT
90{
91 int i;
df5f8314 92 char *buf, *end;
8ea02606 93 char *name;
1da177e4
LT
94 char tcomm[sizeof(p->comm)];
95
96 get_task_comm(tcomm, p);
97
df5f8314
EB
98 seq_printf(m, "Name:\t");
99 end = m->buf + m->size;
100 buf = m->buf + m->count;
1da177e4
LT
101 name = tcomm;
102 i = sizeof(tcomm);
df5f8314 103 while (i && (buf < end)) {
1da177e4
LT
104 unsigned char c = *name;
105 name++;
106 i--;
107 *buf = c;
108 if (!c)
109 break;
110 if (c == '\\') {
df5f8314
EB
111 buf++;
112 if (buf < end)
113 *buf++ = c;
1da177e4
LT
114 continue;
115 }
116 if (c == '\n') {
df5f8314
EB
117 *buf++ = '\\';
118 if (buf < end)
119 *buf++ = 'n';
1da177e4
LT
120 continue;
121 }
122 buf++;
df5f8314
EB
123 }
124 m->count = buf - m->buf;
125 seq_printf(m, "\n");
1da177e4
LT
126}
127
128/*
129 * The task state array is a strange "bitmap" of
130 * reasons to sleep. Thus "running" is zero, and
131 * you can test for combinations of others with
132 * simple bit tests.
133 */
134static const char *task_state_array[] = {
135 "R (running)", /* 0 */
136 "S (sleeping)", /* 1 */
137 "D (disk sleep)", /* 2 */
138 "T (stopped)", /* 4 */
139 "T (tracing stop)", /* 8 */
140 "Z (zombie)", /* 16 */
141 "X (dead)" /* 32 */
142};
143
8ea02606 144static inline const char *get_task_state(struct task_struct *tsk)
1da177e4 145{
1587e2b1 146 unsigned int state = (tsk->state & TASK_REPORT) | tsk->exit_state;
1da177e4
LT
147 const char **p = &task_state_array[0];
148
149 while (state) {
150 p++;
151 state >>= 1;
152 }
153 return *p;
154}
155
df5f8314
EB
156static inline void task_state(struct seq_file *m, struct pid_namespace *ns,
157 struct pid *pid, struct task_struct *p)
1da177e4
LT
158{
159 struct group_info *group_info;
160 int g;
badf1662 161 struct fdtable *fdt = NULL;
c69e8d9c 162 const struct cred *cred;
b488893a 163 pid_t ppid, tpid;
1da177e4 164
b0fa9db6 165 rcu_read_lock();
b488893a
PE
166 ppid = pid_alive(p) ?
167 task_tgid_nr_ns(rcu_dereference(p->real_parent), ns) : 0;
0d094efe
RM
168 tpid = 0;
169 if (pid_alive(p)) {
170 struct task_struct *tracer = tracehook_tracer_task(p);
171 if (tracer)
172 tpid = task_pid_nr_ns(tracer, ns);
173 }
c69e8d9c 174 cred = get_cred((struct cred *) __task_cred(p));
df5f8314 175 seq_printf(m,
1da177e4 176 "State:\t%s\n"
1da177e4
LT
177 "Tgid:\t%d\n"
178 "Pid:\t%d\n"
179 "PPid:\t%d\n"
180 "TracerPid:\t%d\n"
181 "Uid:\t%d\t%d\t%d\t%d\n"
182 "Gid:\t%d\t%d\t%d\t%d\n",
183 get_task_state(p),
b488893a 184 task_tgid_nr_ns(p, ns),
df5f8314 185 pid_nr_ns(pid, ns),
b488893a 186 ppid, tpid,
c69e8d9c
DH
187 cred->uid, cred->euid, cred->suid, cred->fsuid,
188 cred->gid, cred->egid, cred->sgid, cred->fsgid);
b0fa9db6 189
1da177e4 190 task_lock(p);
badf1662
DS
191 if (p->files)
192 fdt = files_fdtable(p->files);
df5f8314 193 seq_printf(m,
1da177e4
LT
194 "FDSize:\t%d\n"
195 "Groups:\t",
badf1662 196 fdt ? fdt->max_fds : 0);
4fb3a538 197 rcu_read_unlock();
1da177e4 198
c69e8d9c 199 group_info = cred->group_info;
1da177e4
LT
200 task_unlock(p);
201
8ea02606 202 for (g = 0; g < min(group_info->ngroups, NGROUPS_SMALL); g++)
df5f8314 203 seq_printf(m, "%d ", GROUP_AT(group_info, g));
c69e8d9c 204 put_cred(cred);
1da177e4 205
df5f8314 206 seq_printf(m, "\n");
1da177e4
LT
207}
208
df5f8314
EB
209static void render_sigset_t(struct seq_file *m, const char *header,
210 sigset_t *set)
1da177e4 211{
df5f8314 212 int i;
1da177e4 213
df5f8314 214 seq_printf(m, "%s", header);
1da177e4
LT
215
216 i = _NSIG;
217 do {
218 int x = 0;
219
220 i -= 4;
221 if (sigismember(set, i+1)) x |= 1;
222 if (sigismember(set, i+2)) x |= 2;
223 if (sigismember(set, i+3)) x |= 4;
224 if (sigismember(set, i+4)) x |= 8;
df5f8314 225 seq_printf(m, "%x", x);
1da177e4
LT
226 } while (i >= 4);
227
df5f8314 228 seq_printf(m, "\n");
1da177e4
LT
229}
230
231static void collect_sigign_sigcatch(struct task_struct *p, sigset_t *ign,
232 sigset_t *catch)
233{
234 struct k_sigaction *k;
235 int i;
236
237 k = p->sighand->action;
238 for (i = 1; i <= _NSIG; ++i, ++k) {
239 if (k->sa.sa_handler == SIG_IGN)
240 sigaddset(ign, i);
241 else if (k->sa.sa_handler != SIG_DFL)
242 sigaddset(catch, i);
243 }
244}
245
df5f8314 246static inline void task_sig(struct seq_file *m, struct task_struct *p)
1da177e4 247{
5e6b3f42 248 unsigned long flags;
1da177e4
LT
249 sigset_t pending, shpending, blocked, ignored, caught;
250 int num_threads = 0;
251 unsigned long qsize = 0;
252 unsigned long qlim = 0;
253
254 sigemptyset(&pending);
255 sigemptyset(&shpending);
256 sigemptyset(&blocked);
257 sigemptyset(&ignored);
258 sigemptyset(&caught);
259
5e6b3f42 260 if (lock_task_sighand(p, &flags)) {
1da177e4
LT
261 pending = p->pending.signal;
262 shpending = p->signal->shared_pending.signal;
263 blocked = p->blocked;
264 collect_sigign_sigcatch(p, &ignored, &caught);
265 num_threads = atomic_read(&p->signal->count);
c69e8d9c 266 qsize = atomic_read(&__task_cred(p)->user->sigpending);
1da177e4 267 qlim = p->signal->rlim[RLIMIT_SIGPENDING].rlim_cur;
5e6b3f42 268 unlock_task_sighand(p, &flags);
1da177e4 269 }
1da177e4 270
df5f8314
EB
271 seq_printf(m, "Threads:\t%d\n", num_threads);
272 seq_printf(m, "SigQ:\t%lu/%lu\n", qsize, qlim);
1da177e4
LT
273
274 /* render them all */
df5f8314
EB
275 render_sigset_t(m, "SigPnd:\t", &pending);
276 render_sigset_t(m, "ShdPnd:\t", &shpending);
277 render_sigset_t(m, "SigBlk:\t", &blocked);
278 render_sigset_t(m, "SigIgn:\t", &ignored);
279 render_sigset_t(m, "SigCgt:\t", &caught);
1da177e4
LT
280}
281
df5f8314
EB
282static void render_cap_t(struct seq_file *m, const char *header,
283 kernel_cap_t *a)
e338d263
AM
284{
285 unsigned __capi;
286
df5f8314 287 seq_printf(m, "%s", header);
e338d263 288 CAP_FOR_EACH_U32(__capi) {
df5f8314 289 seq_printf(m, "%08x",
ca05a99a 290 a->cap[(_KERNEL_CAPABILITY_U32S-1) - __capi]);
e338d263 291 }
df5f8314 292 seq_printf(m, "\n");
e338d263
AM
293}
294
df5f8314 295static inline void task_cap(struct seq_file *m, struct task_struct *p)
1da177e4 296{
c69e8d9c
DH
297 const struct cred *cred;
298 kernel_cap_t cap_inheritable, cap_permitted, cap_effective, cap_bset;
b6dff3ec 299
c69e8d9c
DH
300 rcu_read_lock();
301 cred = __task_cred(p);
302 cap_inheritable = cred->cap_inheritable;
303 cap_permitted = cred->cap_permitted;
304 cap_effective = cred->cap_effective;
305 cap_bset = cred->cap_bset;
306 rcu_read_unlock();
307
308 render_cap_t(m, "CapInh:\t", &cap_inheritable);
309 render_cap_t(m, "CapPrm:\t", &cap_permitted);
310 render_cap_t(m, "CapEff:\t", &cap_effective);
311 render_cap_t(m, "CapBnd:\t", &cap_bset);
1da177e4
LT
312}
313
df5f8314
EB
314static inline void task_context_switch_counts(struct seq_file *m,
315 struct task_struct *p)
b663a79c 316{
df5f8314
EB
317 seq_printf(m, "voluntary_ctxt_switches:\t%lu\n"
318 "nonvoluntary_ctxt_switches:\t%lu\n",
319 p->nvcsw,
320 p->nivcsw);
b663a79c
MU
321}
322
df5f8314
EB
323int proc_pid_status(struct seq_file *m, struct pid_namespace *ns,
324 struct pid *pid, struct task_struct *task)
1da177e4 325{
1da177e4
LT
326 struct mm_struct *mm = get_task_mm(task);
327
df5f8314
EB
328 task_name(m, task);
329 task_state(m, ns, pid, task);
8ea02606 330
1da177e4 331 if (mm) {
df5f8314 332 task_mem(m, mm);
1da177e4
LT
333 mmput(mm);
334 }
df5f8314
EB
335 task_sig(m, task);
336 task_cap(m, task);
337 cpuset_task_status_allowed(m, task);
347a8dc3 338#if defined(CONFIG_S390)
df5f8314 339 task_show_regs(m, task);
1da177e4 340#endif
df5f8314
EB
341 task_context_switch_counts(m, task);
342 return 0;
1da177e4
LT
343}
344
ee992744
EB
345static int do_task_stat(struct seq_file *m, struct pid_namespace *ns,
346 struct pid *pid, struct task_struct *task, int whole)
1da177e4
LT
347{
348 unsigned long vsize, eip, esp, wchan = ~0UL;
349 long priority, nice;
350 int tty_pgrp = -1, tty_nr = 0;
351 sigset_t sigign, sigcatch;
352 char state;
8ea02606 353 pid_t ppid = 0, pgid = -1, sid = -1;
1da177e4
LT
354 int num_threads = 0;
355 struct mm_struct *mm;
356 unsigned long long start_time;
357 unsigned long cmin_flt = 0, cmaj_flt = 0;
358 unsigned long min_flt = 0, maj_flt = 0;
efe567fc 359 cputime_t cutime, cstime, utime, stime;
9ac52315 360 cputime_t cgtime, gtime;
1da177e4 361 unsigned long rsslim = 0;
1da177e4 362 char tcomm[sizeof(task->comm)];
a593d6ed 363 unsigned long flags;
1da177e4
LT
364
365 state = *get_task_state(task);
366 vsize = eip = esp = 0;
367 mm = get_task_mm(task);
368 if (mm) {
369 vsize = task_vsize(mm);
370 eip = KSTK_EIP(task);
371 esp = KSTK_ESP(task);
372 }
373
374 get_task_comm(tcomm, task);
375
376 sigemptyset(&sigign);
377 sigemptyset(&sigcatch);
efe567fc 378 cutime = cstime = utime = stime = cputime_zero;
9ac52315 379 cgtime = gtime = cputime_zero;
3cfd0885 380
a593d6ed
ON
381 if (lock_task_sighand(task, &flags)) {
382 struct signal_struct *sig = task->signal;
91593504
ON
383
384 if (sig->tty) {
5d0fdf1e
AC
385 struct pid *pgrp = tty_get_pgrp(sig->tty);
386 tty_pgrp = pid_nr_ns(pgrp, ns);
387 put_pid(pgrp);
91593504 388 tty_nr = new_encode_dev(tty_devnum(sig->tty));
a593d6ed
ON
389 }
390
391 num_threads = atomic_read(&sig->count);
1da177e4
LT
392 collect_sigign_sigcatch(task, &sigign, &sigcatch);
393
a593d6ed
ON
394 cmin_flt = sig->cmin_flt;
395 cmaj_flt = sig->cmaj_flt;
396 cutime = sig->cutime;
397 cstime = sig->cstime;
9ac52315 398 cgtime = sig->cgtime;
a593d6ed
ON
399 rsslim = sig->rlim[RLIMIT_RSS].rlim_cur;
400
1da177e4
LT
401 /* add up live thread stats at the group level */
402 if (whole) {
f06febc9 403 struct task_cputime cputime;
a593d6ed 404 struct task_struct *t = task;
1da177e4
LT
405 do {
406 min_flt += t->min_flt;
407 maj_flt += t->maj_flt;
9ac52315 408 gtime = cputime_add(gtime, task_gtime(t));
1da177e4
LT
409 t = next_thread(t);
410 } while (t != task);
1da177e4 411
a593d6ed
ON
412 min_flt += sig->min_flt;
413 maj_flt += sig->maj_flt;
f06febc9
FM
414 thread_group_cputime(task, &cputime);
415 utime = cputime.utime;
416 stime = cputime.stime;
f9e26291 417 gtime = cputime_add(gtime, sig->gtime);
1da177e4 418 }
a593d6ed 419
b488893a 420 sid = task_session_nr_ns(task, ns);
a98fdcef 421 ppid = task_tgid_nr_ns(task->real_parent, ns);
b488893a 422 pgid = task_pgrp_nr_ns(task, ns);
a593d6ed
ON
423
424 unlock_task_sighand(task, &flags);
1da177e4 425 }
1da177e4 426
8ea02606 427 if (!whole || num_threads < 2)
1da177e4
LT
428 wchan = get_wchan(task);
429 if (!whole) {
430 min_flt = task->min_flt;
431 maj_flt = task->maj_flt;
b27f03d4
IM
432 utime = task_utime(task);
433 stime = task_stime(task);
9ac52315 434 gtime = task_gtime(task);
1da177e4
LT
435 }
436
437 /* scale priority and nice values from timeslices to -20..20 */
438 /* to make it look like a "normal" Unix priority/nice value */
439 priority = task_prio(task);
440 nice = task_nice(task);
441
442 /* Temporary variable needed for gcc-2.96 */
443 /* convert timespec -> nsec*/
924b42d5
TJ
444 start_time =
445 (unsigned long long)task->real_start_time.tv_sec * NSEC_PER_SEC
446 + task->real_start_time.tv_nsec;
1da177e4
LT
447 /* convert nsec -> ticks */
448 start_time = nsec_to_clock_t(start_time);
449
ee992744 450 seq_printf(m, "%d (%s) %c %d %d %d %d %d %u %lu \
4dee26b7 451%lu %lu %lu %lu %lu %ld %ld %ld %ld %d 0 %llu %lu %ld %lu %lu %lu %lu %lu \
9ac52315 452%lu %lu %lu %lu %lu %lu %lu %lu %d %d %u %u %llu %lu %ld\n",
ee992744 453 pid_nr_ns(pid, ns),
1da177e4
LT
454 tcomm,
455 state,
456 ppid,
457 pgid,
458 sid,
459 tty_nr,
460 tty_pgrp,
461 task->flags,
462 min_flt,
463 cmin_flt,
464 maj_flt,
465 cmaj_flt,
efe567fc
CB
466 cputime_to_clock_t(utime),
467 cputime_to_clock_t(stime),
1da177e4
LT
468 cputime_to_clock_t(cutime),
469 cputime_to_clock_t(cstime),
470 priority,
471 nice,
472 num_threads,
1da177e4
LT
473 start_time,
474 vsize,
4294621f 475 mm ? get_mm_rss(mm) : 0,
8ea02606 476 rsslim,
1da177e4
LT
477 mm ? mm->start_code : 0,
478 mm ? mm->end_code : 0,
479 mm ? mm->start_stack : 0,
480 esp,
481 eip,
482 /* The signal information here is obsolete.
483 * It must be decimal for Linux 2.0 compatibility.
484 * Use /proc/#/status for real-time signals.
485 */
486 task->pending.signal.sig[0] & 0x7fffffffUL,
487 task->blocked.sig[0] & 0x7fffffffUL,
488 sigign .sig[0] & 0x7fffffffUL,
489 sigcatch .sig[0] & 0x7fffffffUL,
490 wchan,
491 0UL,
492 0UL,
493 task->exit_signal,
494 task_cpu(task),
495 task->rt_priority,
25890454 496 task->policy,
9ac52315
LV
497 (unsigned long long)delayacct_blkio_ticks(task),
498 cputime_to_clock_t(gtime),
499 cputime_to_clock_t(cgtime));
8ea02606 500 if (mm)
1da177e4 501 mmput(mm);
ee992744 502 return 0;
1da177e4
LT
503}
504
ee992744
EB
505int proc_tid_stat(struct seq_file *m, struct pid_namespace *ns,
506 struct pid *pid, struct task_struct *task)
1da177e4 507{
ee992744 508 return do_task_stat(m, ns, pid, task, 0);
1da177e4
LT
509}
510
ee992744
EB
511int proc_tgid_stat(struct seq_file *m, struct pid_namespace *ns,
512 struct pid *pid, struct task_struct *task)
1da177e4 513{
ee992744 514 return do_task_stat(m, ns, pid, task, 1);
1da177e4
LT
515}
516
a56d3fc7
EB
517int proc_pid_statm(struct seq_file *m, struct pid_namespace *ns,
518 struct pid *pid, struct task_struct *task)
1da177e4
LT
519{
520 int size = 0, resident = 0, shared = 0, text = 0, lib = 0, data = 0;
521 struct mm_struct *mm = get_task_mm(task);
8ea02606 522
1da177e4
LT
523 if (mm) {
524 size = task_statm(mm, &shared, &text, &data, &resident);
525 mmput(mm);
526 }
a56d3fc7
EB
527 seq_printf(m, "%d %d %d %d %d %d %d\n",
528 size, resident, shared, text, lib, data, 0);
1da177e4 529
a56d3fc7 530 return 0;
1da177e4 531}