e51f2ec2c5e5c53809cacac73c544daaec76a579
[linux-2.6.git] / fs / proc / array.c
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.
43  *                      <alan@lxorguk.ukuu.org.uk>
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
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>
66 #include <linux/io.h>
67 #include <linux/mm.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/swapops.h>
85
86 #include <asm/pgtable.h>
87 #include <asm/processor.h>
88 #include "internal.h"
89
90 static inline void task_name(struct seq_file *m, struct task_struct *p)
91 {
92         int i;
93         char *buf, *end;
94         char *name;
95         char tcomm[sizeof(p->comm)];
96
97         get_task_comm(tcomm, p);
98
99         seq_printf(m, "Name:\t");
100         end = m->buf + m->size;
101         buf = m->buf + m->count;
102         name = tcomm;
103         i = sizeof(tcomm);
104         while (i && (buf < end)) {
105                 unsigned char c = *name;
106                 name++;
107                 i--;
108                 *buf = c;
109                 if (!c)
110                         break;
111                 if (c == '\\') {
112                         buf++;
113                         if (buf < end)
114                                 *buf++ = c;
115                         continue;
116                 }
117                 if (c == '\n') {
118                         *buf++ = '\\';
119                         if (buf < end)
120                                 *buf++ = 'n';
121                         continue;
122                 }
123                 buf++;
124         }
125         m->count = buf - m->buf;
126         seq_printf(m, "\n");
127 }
128
129 /*
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
133  * simple bit tests.
134  */
135 static const char *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 */
142         "X (dead)",             /*  32 */
143         "x (dead)",             /*  64 */
144         "K (wakekill)",         /* 128 */
145         "W (waking)",           /* 256 */
146 };
147
148 static inline const char *get_task_state(struct task_struct *tsk)
149 {
150         unsigned int state = (tsk->state & TASK_REPORT) | tsk->exit_state;
151         const char **p = &task_state_array[0];
152
153         BUILD_BUG_ON(1 + ilog2(TASK_STATE_MAX) != ARRAY_SIZE(task_state_array));
154
155         while (state) {
156                 p++;
157                 state >>= 1;
158         }
159         return *p;
160 }
161
162 static inline void task_state(struct seq_file *m, struct pid_namespace *ns,
163                                 struct pid *pid, struct task_struct *p)
164 {
165         struct group_info *group_info;
166         int g;
167         struct fdtable *fdt = NULL;
168         const struct cred *cred;
169         pid_t ppid, tpid;
170
171         rcu_read_lock();
172         ppid = pid_alive(p) ?
173                 task_tgid_nr_ns(rcu_dereference(p->real_parent), ns) : 0;
174         tpid = 0;
175         if (pid_alive(p)) {
176                 struct task_struct *tracer = tracehook_tracer_task(p);
177                 if (tracer)
178                         tpid = task_pid_nr_ns(tracer, ns);
179         }
180         cred = get_cred((struct cred *) __task_cred(p));
181         seq_printf(m,
182                 "State:\t%s\n"
183                 "Tgid:\t%d\n"
184                 "Pid:\t%d\n"
185                 "PPid:\t%d\n"
186                 "TracerPid:\t%d\n"
187                 "Uid:\t%d\t%d\t%d\t%d\n"
188                 "Gid:\t%d\t%d\t%d\t%d\n",
189                 get_task_state(p),
190                 task_tgid_nr_ns(p, ns),
191                 pid_nr_ns(pid, ns),
192                 ppid, tpid,
193                 cred->uid, cred->euid, cred->suid, cred->fsuid,
194                 cred->gid, cred->egid, cred->sgid, cred->fsgid);
195
196         task_lock(p);
197         if (p->files)
198                 fdt = files_fdtable(p->files);
199         seq_printf(m,
200                 "FDSize:\t%d\n"
201                 "Groups:\t",
202                 fdt ? fdt->max_fds : 0);
203         rcu_read_unlock();
204
205         group_info = cred->group_info;
206         task_unlock(p);
207
208         for (g = 0; g < min(group_info->ngroups, NGROUPS_SMALL); g++)
209                 seq_printf(m, "%d ", GROUP_AT(group_info, g));
210         put_cred(cred);
211
212         seq_printf(m, "\n");
213 }
214
215 static void render_sigset_t(struct seq_file *m, const char *header,
216                                 sigset_t *set)
217 {
218         int i;
219
220         seq_printf(m, "%s", header);
221
222         i = _NSIG;
223         do {
224                 int x = 0;
225
226                 i -= 4;
227                 if (sigismember(set, i+1)) x |= 1;
228                 if (sigismember(set, i+2)) x |= 2;
229                 if (sigismember(set, i+3)) x |= 4;
230                 if (sigismember(set, i+4)) x |= 8;
231                 seq_printf(m, "%x", x);
232         } while (i >= 4);
233
234         seq_printf(m, "\n");
235 }
236
237 static void collect_sigign_sigcatch(struct task_struct *p, sigset_t *ign,
238                                     sigset_t *catch)
239 {
240         struct k_sigaction *k;
241         int i;
242
243         k = p->sighand->action;
244         for (i = 1; i <= _NSIG; ++i, ++k) {
245                 if (k->sa.sa_handler == SIG_IGN)
246                         sigaddset(ign, i);
247                 else if (k->sa.sa_handler != SIG_DFL)
248                         sigaddset(catch, i);
249         }
250 }
251
252 static inline void task_sig(struct seq_file *m, struct task_struct *p)
253 {
254         unsigned long flags;
255         sigset_t pending, shpending, blocked, ignored, caught;
256         int num_threads = 0;
257         unsigned long qsize = 0;
258         unsigned long qlim = 0;
259
260         sigemptyset(&pending);
261         sigemptyset(&shpending);
262         sigemptyset(&blocked);
263         sigemptyset(&ignored);
264         sigemptyset(&caught);
265
266         if (lock_task_sighand(p, &flags)) {
267                 pending = p->pending.signal;
268                 shpending = p->signal->shared_pending.signal;
269                 blocked = p->blocked;
270                 collect_sigign_sigcatch(p, &ignored, &caught);
271                 num_threads = atomic_read(&p->signal->count);
272                 rcu_read_lock();  /* FIXME: is this correct? */
273                 qsize = atomic_read(&__task_cred(p)->user->sigpending);
274                 rcu_read_unlock();
275                 qlim = task_rlimit(p, RLIMIT_SIGPENDING);
276                 unlock_task_sighand(p, &flags);
277         }
278
279         seq_printf(m, "Threads:\t%d\n", num_threads);
280         seq_printf(m, "SigQ:\t%lu/%lu\n", qsize, qlim);
281
282         /* render them all */
283         render_sigset_t(m, "SigPnd:\t", &pending);
284         render_sigset_t(m, "ShdPnd:\t", &shpending);
285         render_sigset_t(m, "SigBlk:\t", &blocked);
286         render_sigset_t(m, "SigIgn:\t", &ignored);
287         render_sigset_t(m, "SigCgt:\t", &caught);
288 }
289
290 static void render_cap_t(struct seq_file *m, const char *header,
291                         kernel_cap_t *a)
292 {
293         unsigned __capi;
294
295         seq_printf(m, "%s", header);
296         CAP_FOR_EACH_U32(__capi) {
297                 seq_printf(m, "%08x",
298                            a->cap[(_KERNEL_CAPABILITY_U32S-1) - __capi]);
299         }
300         seq_printf(m, "\n");
301 }
302
303 static inline void task_cap(struct seq_file *m, struct task_struct *p)
304 {
305         const struct cred *cred;
306         kernel_cap_t cap_inheritable, cap_permitted, cap_effective, cap_bset;
307
308         rcu_read_lock();
309         cred = __task_cred(p);
310         cap_inheritable = cred->cap_inheritable;
311         cap_permitted   = cred->cap_permitted;
312         cap_effective   = cred->cap_effective;
313         cap_bset        = cred->cap_bset;
314         rcu_read_unlock();
315
316         render_cap_t(m, "CapInh:\t", &cap_inheritable);
317         render_cap_t(m, "CapPrm:\t", &cap_permitted);
318         render_cap_t(m, "CapEff:\t", &cap_effective);
319         render_cap_t(m, "CapBnd:\t", &cap_bset);
320 }
321
322 static inline void task_context_switch_counts(struct seq_file *m,
323                                                 struct task_struct *p)
324 {
325         seq_printf(m,   "voluntary_ctxt_switches:\t%lu\n"
326                         "nonvoluntary_ctxt_switches:\t%lu\n",
327                         p->nvcsw,
328                         p->nivcsw);
329 }
330
331 static void task_cpus_allowed(struct seq_file *m, struct task_struct *task)
332 {
333         seq_printf(m, "Cpus_allowed:\t");
334         seq_cpumask(m, &task->cpus_allowed);
335         seq_printf(m, "\n");
336         seq_printf(m, "Cpus_allowed_list:\t");
337         seq_cpumask_list(m, &task->cpus_allowed);
338         seq_printf(m, "\n");
339 }
340
341 int proc_pid_status(struct seq_file *m, struct pid_namespace *ns,
342                         struct pid *pid, struct task_struct *task)
343 {
344         struct mm_struct *mm = get_task_mm(task);
345
346         task_name(m, task);
347         task_state(m, ns, pid, task);
348
349         if (mm) {
350                 task_mem(m, mm);
351                 mmput(mm);
352         }
353         task_sig(m, task);
354         task_cap(m, task);
355         task_cpus_allowed(m, task);
356         cpuset_task_status_allowed(m, task);
357 #if defined(CONFIG_S390)
358         task_show_regs(m, task);
359 #endif
360         task_context_switch_counts(m, task);
361         return 0;
362 }
363
364 static int do_task_stat(struct seq_file *m, struct pid_namespace *ns,
365                         struct pid *pid, struct task_struct *task, int whole)
366 {
367         unsigned long vsize, eip, esp, wchan = ~0UL;
368         long priority, nice;
369         int tty_pgrp = -1, tty_nr = 0;
370         sigset_t sigign, sigcatch;
371         char state;
372         pid_t ppid = 0, pgid = -1, sid = -1;
373         int num_threads = 0;
374         int permitted;
375         struct mm_struct *mm;
376         unsigned long long start_time;
377         unsigned long cmin_flt = 0, cmaj_flt = 0;
378         unsigned long  min_flt = 0,  maj_flt = 0;
379         cputime_t cutime, cstime, utime, stime;
380         cputime_t cgtime, gtime;
381         unsigned long rsslim = 0;
382         char tcomm[sizeof(task->comm)];
383         unsigned long flags;
384
385         state = *get_task_state(task);
386         vsize = eip = esp = 0;
387         permitted = ptrace_may_access(task, PTRACE_MODE_READ);
388         mm = get_task_mm(task);
389         if (mm) {
390                 vsize = task_vsize(mm);
391                 if (permitted) {
392                         eip = KSTK_EIP(task);
393                         esp = KSTK_ESP(task);
394                 }
395         }
396
397         get_task_comm(tcomm, task);
398
399         sigemptyset(&sigign);
400         sigemptyset(&sigcatch);
401         cutime = cstime = utime = stime = cputime_zero;
402         cgtime = gtime = cputime_zero;
403
404         if (lock_task_sighand(task, &flags)) {
405                 struct signal_struct *sig = task->signal;
406
407                 if (sig->tty) {
408                         struct pid *pgrp = tty_get_pgrp(sig->tty);
409                         tty_pgrp = pid_nr_ns(pgrp, ns);
410                         put_pid(pgrp);
411                         tty_nr = new_encode_dev(tty_devnum(sig->tty));
412                 }
413
414                 num_threads = atomic_read(&sig->count);
415                 collect_sigign_sigcatch(task, &sigign, &sigcatch);
416
417                 cmin_flt = sig->cmin_flt;
418                 cmaj_flt = sig->cmaj_flt;
419                 cutime = sig->cutime;
420                 cstime = sig->cstime;
421                 cgtime = sig->cgtime;
422                 rsslim = ACCESS_ONCE(sig->rlim[RLIMIT_RSS].rlim_cur);
423
424                 /* add up live thread stats at the group level */
425                 if (whole) {
426                         struct task_struct *t = task;
427                         do {
428                                 min_flt += t->min_flt;
429                                 maj_flt += t->maj_flt;
430                                 gtime = cputime_add(gtime, t->gtime);
431                                 t = next_thread(t);
432                         } while (t != task);
433
434                         min_flt += sig->min_flt;
435                         maj_flt += sig->maj_flt;
436                         thread_group_times(task, &utime, &stime);
437                         gtime = cputime_add(gtime, sig->gtime);
438                 }
439
440                 sid = task_session_nr_ns(task, ns);
441                 ppid = task_tgid_nr_ns(task->real_parent, ns);
442                 pgid = task_pgrp_nr_ns(task, ns);
443
444                 unlock_task_sighand(task, &flags);
445         }
446
447         if (permitted && (!whole || num_threads < 2))
448                 wchan = get_wchan(task);
449         if (!whole) {
450                 min_flt = task->min_flt;
451                 maj_flt = task->maj_flt;
452                 task_times(task, &utime, &stime);
453                 gtime = task->gtime;
454         }
455
456         /* scale priority and nice values from timeslices to -20..20 */
457         /* to make it look like a "normal" Unix priority/nice value  */
458         priority = task_prio(task);
459         nice = task_nice(task);
460
461         /* Temporary variable needed for gcc-2.96 */
462         /* convert timespec -> nsec*/
463         start_time =
464                 (unsigned long long)task->real_start_time.tv_sec * NSEC_PER_SEC
465                                 + task->real_start_time.tv_nsec;
466         /* convert nsec -> ticks */
467         start_time = nsec_to_clock_t(start_time);
468
469         seq_printf(m, "%d (%s) %c %d %d %d %d %d %u %lu \
470 %lu %lu %lu %lu %lu %ld %ld %ld %ld %d 0 %llu %lu %ld %lu %lu %lu %lu %lu \
471 %lu %lu %lu %lu %lu %lu %lu %lu %d %d %u %u %llu %lu %ld\n",
472                 pid_nr_ns(pid, ns),
473                 tcomm,
474                 state,
475                 ppid,
476                 pgid,
477                 sid,
478                 tty_nr,
479                 tty_pgrp,
480                 task->flags,
481                 min_flt,
482                 cmin_flt,
483                 maj_flt,
484                 cmaj_flt,
485                 cputime_to_clock_t(utime),
486                 cputime_to_clock_t(stime),
487                 cputime_to_clock_t(cutime),
488                 cputime_to_clock_t(cstime),
489                 priority,
490                 nice,
491                 num_threads,
492                 start_time,
493                 vsize,
494                 mm ? get_mm_rss(mm) : 0,
495                 rsslim,
496                 mm ? mm->start_code : 0,
497                 mm ? mm->end_code : 0,
498                 (permitted && mm) ? task->stack_start : 0,
499                 esp,
500                 eip,
501                 /* The signal information here is obsolete.
502                  * It must be decimal for Linux 2.0 compatibility.
503                  * Use /proc/#/status for real-time signals.
504                  */
505                 task->pending.signal.sig[0] & 0x7fffffffUL,
506                 task->blocked.sig[0] & 0x7fffffffUL,
507                 sigign      .sig[0] & 0x7fffffffUL,
508                 sigcatch    .sig[0] & 0x7fffffffUL,
509                 wchan,
510                 0UL,
511                 0UL,
512                 task->exit_signal,
513                 task_cpu(task),
514                 task->rt_priority,
515                 task->policy,
516                 (unsigned long long)delayacct_blkio_ticks(task),
517                 cputime_to_clock_t(gtime),
518                 cputime_to_clock_t(cgtime));
519         if (mm)
520                 mmput(mm);
521         return 0;
522 }
523
524 int proc_tid_stat(struct seq_file *m, struct pid_namespace *ns,
525                         struct pid *pid, struct task_struct *task)
526 {
527         return do_task_stat(m, ns, pid, task, 0);
528 }
529
530 int proc_tgid_stat(struct seq_file *m, struct pid_namespace *ns,
531                         struct pid *pid, struct task_struct *task)
532 {
533         return do_task_stat(m, ns, pid, task, 1);
534 }
535
536 int proc_pid_statm(struct seq_file *m, struct pid_namespace *ns,
537                         struct pid *pid, struct task_struct *task)
538 {
539         int size = 0, resident = 0, shared = 0, text = 0, lib = 0, data = 0;
540         struct mm_struct *mm = get_task_mm(task);
541
542         if (mm) {
543                 size = task_statm(mm, &shared, &text, &data, &resident);
544                 mmput(mm);
545         }
546         seq_printf(m, "%d %d %d %d %d %d %d\n",
547                         size, resident, shared, text, lib, data, 0);
548
549         return 0;
550 }