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