Merge branch 'upstream-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mfashe...
[linux-2.6] / 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.Cox@linux.org>
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/slab.h>
72 #include <linux/smp.h>
73 #include <linux/signal.h>
74 #include <linux/highmem.h>
75 #include <linux/file.h>
76 #include <linux/fdtable.h>
77 #include <linux/times.h>
78 #include <linux/cpuset.h>
79 #include <linux/rcupdate.h>
80 #include <linux/delayacct.h>
81 #include <linux/seq_file.h>
82 #include <linux/pid_namespace.h>
83
84 #include <asm/pgtable.h>
85 #include <asm/processor.h>
86 #include "internal.h"
87
88 /* Gcc optimizes away "strlen(x)" for constant x */
89 #define ADDBUF(buffer, string) \
90 do { memcpy(buffer, string, strlen(string)); \
91      buffer += strlen(string); } while (0)
92
93 static inline void task_name(struct seq_file *m, struct task_struct *p)
94 {
95         int i;
96         char *buf, *end;
97         char *name;
98         char tcomm[sizeof(p->comm)];
99
100         get_task_comm(tcomm, p);
101
102         seq_printf(m, "Name:\t");
103         end = m->buf + m->size;
104         buf = m->buf + m->count;
105         name = tcomm;
106         i = sizeof(tcomm);
107         while (i && (buf < end)) {
108                 unsigned char c = *name;
109                 name++;
110                 i--;
111                 *buf = c;
112                 if (!c)
113                         break;
114                 if (c == '\\') {
115                         buf++;
116                         if (buf < end)
117                                 *buf++ = c;
118                         continue;
119                 }
120                 if (c == '\n') {
121                         *buf++ = '\\';
122                         if (buf < end)
123                                 *buf++ = 'n';
124                         continue;
125                 }
126                 buf++;
127         }
128         m->count = buf - m->buf;
129         seq_printf(m, "\n");
130 }
131
132 /*
133  * The task state array is a strange "bitmap" of
134  * reasons to sleep. Thus "running" is zero, and
135  * you can test for combinations of others with
136  * simple bit tests.
137  */
138 static const char *task_state_array[] = {
139         "R (running)",          /*  0 */
140         "S (sleeping)",         /*  1 */
141         "D (disk sleep)",       /*  2 */
142         "T (stopped)",          /*  4 */
143         "T (tracing stop)",     /*  8 */
144         "Z (zombie)",           /* 16 */
145         "X (dead)"              /* 32 */
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         while (state) {
154                 p++;
155                 state >>= 1;
156         }
157         return *p;
158 }
159
160 static inline void task_state(struct seq_file *m, struct pid_namespace *ns,
161                                 struct pid *pid, struct task_struct *p)
162 {
163         struct group_info *group_info;
164         int g;
165         struct fdtable *fdt = NULL;
166         pid_t ppid, tpid;
167
168         rcu_read_lock();
169         ppid = pid_alive(p) ?
170                 task_tgid_nr_ns(rcu_dereference(p->real_parent), ns) : 0;
171         tpid = pid_alive(p) && p->ptrace ?
172                 task_pid_nr_ns(rcu_dereference(p->parent), ns) : 0;
173         seq_printf(m,
174                 "State:\t%s\n"
175                 "Tgid:\t%d\n"
176                 "Pid:\t%d\n"
177                 "PPid:\t%d\n"
178                 "TracerPid:\t%d\n"
179                 "Uid:\t%d\t%d\t%d\t%d\n"
180                 "Gid:\t%d\t%d\t%d\t%d\n",
181                 get_task_state(p),
182                 task_tgid_nr_ns(p, ns),
183                 pid_nr_ns(pid, ns),
184                 ppid, tpid,
185                 p->uid, p->euid, p->suid, p->fsuid,
186                 p->gid, p->egid, p->sgid, p->fsgid);
187
188         task_lock(p);
189         if (p->files)
190                 fdt = files_fdtable(p->files);
191         seq_printf(m,
192                 "FDSize:\t%d\n"
193                 "Groups:\t",
194                 fdt ? fdt->max_fds : 0);
195         rcu_read_unlock();
196
197         group_info = p->group_info;
198         get_group_info(group_info);
199         task_unlock(p);
200
201         for (g = 0; g < min(group_info->ngroups, NGROUPS_SMALL); g++)
202                 seq_printf(m, "%d ", GROUP_AT(group_info, g));
203         put_group_info(group_info);
204
205         seq_printf(m, "\n");
206 }
207
208 static void render_sigset_t(struct seq_file *m, const char *header,
209                                 sigset_t *set)
210 {
211         int i;
212
213         seq_printf(m, "%s", header);
214
215         i = _NSIG;
216         do {
217                 int x = 0;
218
219                 i -= 4;
220                 if (sigismember(set, i+1)) x |= 1;
221                 if (sigismember(set, i+2)) x |= 2;
222                 if (sigismember(set, i+3)) x |= 4;
223                 if (sigismember(set, i+4)) x |= 8;
224                 seq_printf(m, "%x", x);
225         } while (i >= 4);
226
227         seq_printf(m, "\n");
228 }
229
230 static void collect_sigign_sigcatch(struct task_struct *p, sigset_t *ign,
231                                     sigset_t *catch)
232 {
233         struct k_sigaction *k;
234         int i;
235
236         k = p->sighand->action;
237         for (i = 1; i <= _NSIG; ++i, ++k) {
238                 if (k->sa.sa_handler == SIG_IGN)
239                         sigaddset(ign, i);
240                 else if (k->sa.sa_handler != SIG_DFL)
241                         sigaddset(catch, i);
242         }
243 }
244
245 static inline void task_sig(struct seq_file *m, struct task_struct *p)
246 {
247         unsigned long flags;
248         sigset_t pending, shpending, blocked, ignored, caught;
249         int num_threads = 0;
250         unsigned long qsize = 0;
251         unsigned long qlim = 0;
252
253         sigemptyset(&pending);
254         sigemptyset(&shpending);
255         sigemptyset(&blocked);
256         sigemptyset(&ignored);
257         sigemptyset(&caught);
258
259         rcu_read_lock();
260         if (lock_task_sighand(p, &flags)) {
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);
266                 qsize = atomic_read(&p->user->sigpending);
267                 qlim = p->signal->rlim[RLIMIT_SIGPENDING].rlim_cur;
268                 unlock_task_sighand(p, &flags);
269         }
270         rcu_read_unlock();
271
272         seq_printf(m, "Threads:\t%d\n", num_threads);
273         seq_printf(m, "SigQ:\t%lu/%lu\n", qsize, qlim);
274
275         /* render them all */
276         render_sigset_t(m, "SigPnd:\t", &pending);
277         render_sigset_t(m, "ShdPnd:\t", &shpending);
278         render_sigset_t(m, "SigBlk:\t", &blocked);
279         render_sigset_t(m, "SigIgn:\t", &ignored);
280         render_sigset_t(m, "SigCgt:\t", &caught);
281 }
282
283 static void render_cap_t(struct seq_file *m, const char *header,
284                         kernel_cap_t *a)
285 {
286         unsigned __capi;
287
288         seq_printf(m, "%s", header);
289         CAP_FOR_EACH_U32(__capi) {
290                 seq_printf(m, "%08x",
291                            a->cap[(_LINUX_CAPABILITY_U32S-1) - __capi]);
292         }
293         seq_printf(m, "\n");
294 }
295
296 static inline void task_cap(struct seq_file *m, struct task_struct *p)
297 {
298         render_cap_t(m, "CapInh:\t", &p->cap_inheritable);
299         render_cap_t(m, "CapPrm:\t", &p->cap_permitted);
300         render_cap_t(m, "CapEff:\t", &p->cap_effective);
301 }
302
303 static inline void task_context_switch_counts(struct seq_file *m,
304                                                 struct task_struct *p)
305 {
306         seq_printf(m,   "voluntary_ctxt_switches:\t%lu\n"
307                         "nonvoluntary_ctxt_switches:\t%lu\n",
308                         p->nvcsw,
309                         p->nivcsw);
310 }
311
312 int proc_pid_status(struct seq_file *m, struct pid_namespace *ns,
313                         struct pid *pid, struct task_struct *task)
314 {
315         struct mm_struct *mm = get_task_mm(task);
316
317         task_name(m, task);
318         task_state(m, ns, pid, task);
319
320         if (mm) {
321                 task_mem(m, mm);
322                 mmput(mm);
323         }
324         task_sig(m, task);
325         task_cap(m, task);
326         cpuset_task_status_allowed(m, task);
327 #if defined(CONFIG_S390)
328         task_show_regs(m, task);
329 #endif
330         task_context_switch_counts(m, task);
331         return 0;
332 }
333
334 /*
335  * Use precise platform statistics if available:
336  */
337 #ifdef CONFIG_VIRT_CPU_ACCOUNTING
338 static cputime_t task_utime(struct task_struct *p)
339 {
340         return p->utime;
341 }
342
343 static cputime_t task_stime(struct task_struct *p)
344 {
345         return p->stime;
346 }
347 #else
348 static cputime_t task_utime(struct task_struct *p)
349 {
350         clock_t utime = cputime_to_clock_t(p->utime),
351                 total = utime + cputime_to_clock_t(p->stime);
352         u64 temp;
353
354         /*
355          * Use CFS's precise accounting:
356          */
357         temp = (u64)nsec_to_clock_t(p->se.sum_exec_runtime);
358
359         if (total) {
360                 temp *= utime;
361                 do_div(temp, total);
362         }
363         utime = (clock_t)temp;
364
365         p->prev_utime = max(p->prev_utime, clock_t_to_cputime(utime));
366         return p->prev_utime;
367 }
368
369 static cputime_t task_stime(struct task_struct *p)
370 {
371         clock_t stime;
372
373         /*
374          * Use CFS's precise accounting. (we subtract utime from
375          * the total, to make sure the total observed by userspace
376          * grows monotonically - apps rely on that):
377          */
378         stime = nsec_to_clock_t(p->se.sum_exec_runtime) -
379                         cputime_to_clock_t(task_utime(p));
380
381         if (stime >= 0)
382                 p->prev_stime = max(p->prev_stime, clock_t_to_cputime(stime));
383
384         return p->prev_stime;
385 }
386 #endif
387
388 static cputime_t task_gtime(struct task_struct *p)
389 {
390         return p->gtime;
391 }
392
393 static int do_task_stat(struct seq_file *m, struct pid_namespace *ns,
394                         struct pid *pid, struct task_struct *task, int whole)
395 {
396         unsigned long vsize, eip, esp, wchan = ~0UL;
397         long priority, nice;
398         int tty_pgrp = -1, tty_nr = 0;
399         sigset_t sigign, sigcatch;
400         char state;
401         pid_t ppid = 0, pgid = -1, sid = -1;
402         int num_threads = 0;
403         struct mm_struct *mm;
404         unsigned long long start_time;
405         unsigned long cmin_flt = 0, cmaj_flt = 0;
406         unsigned long  min_flt = 0,  maj_flt = 0;
407         cputime_t cutime, cstime, utime, stime;
408         cputime_t cgtime, gtime;
409         unsigned long rsslim = 0;
410         char tcomm[sizeof(task->comm)];
411         unsigned long flags;
412
413         state = *get_task_state(task);
414         vsize = eip = esp = 0;
415         mm = get_task_mm(task);
416         if (mm) {
417                 vsize = task_vsize(mm);
418                 eip = KSTK_EIP(task);
419                 esp = KSTK_ESP(task);
420         }
421
422         get_task_comm(tcomm, task);
423
424         sigemptyset(&sigign);
425         sigemptyset(&sigcatch);
426         cutime = cstime = utime = stime = cputime_zero;
427         cgtime = gtime = cputime_zero;
428
429         if (lock_task_sighand(task, &flags)) {
430                 struct signal_struct *sig = task->signal;
431
432                 if (sig->tty) {
433                         struct pid *pgrp = tty_get_pgrp(sig->tty);
434                         tty_pgrp = pid_nr_ns(pgrp, ns);
435                         put_pid(pgrp);
436                         tty_nr = new_encode_dev(tty_devnum(sig->tty));
437                 }
438
439                 num_threads = atomic_read(&sig->count);
440                 collect_sigign_sigcatch(task, &sigign, &sigcatch);
441
442                 cmin_flt = sig->cmin_flt;
443                 cmaj_flt = sig->cmaj_flt;
444                 cutime = sig->cutime;
445                 cstime = sig->cstime;
446                 cgtime = sig->cgtime;
447                 rsslim = sig->rlim[RLIMIT_RSS].rlim_cur;
448
449                 /* add up live thread stats at the group level */
450                 if (whole) {
451                         struct task_struct *t = task;
452                         do {
453                                 min_flt += t->min_flt;
454                                 maj_flt += t->maj_flt;
455                                 utime = cputime_add(utime, task_utime(t));
456                                 stime = cputime_add(stime, task_stime(t));
457                                 gtime = cputime_add(gtime, task_gtime(t));
458                                 t = next_thread(t);
459                         } while (t != task);
460
461                         min_flt += sig->min_flt;
462                         maj_flt += sig->maj_flt;
463                         utime = cputime_add(utime, sig->utime);
464                         stime = cputime_add(stime, sig->stime);
465                         gtime = cputime_add(gtime, sig->gtime);
466                 }
467
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);
471
472                 unlock_task_sighand(task, &flags);
473         }
474
475         if (!whole || num_threads < 2)
476                 wchan = get_wchan(task);
477         if (!whole) {
478                 min_flt = task->min_flt;
479                 maj_flt = task->maj_flt;
480                 utime = task_utime(task);
481                 stime = task_stime(task);
482                 gtime = task_gtime(task);
483         }
484
485         /* scale priority and nice values from timeslices to -20..20 */
486         /* to make it look like a "normal" Unix priority/nice value  */
487         priority = task_prio(task);
488         nice = task_nice(task);
489
490         /* Temporary variable needed for gcc-2.96 */
491         /* convert timespec -> nsec*/
492         start_time =
493                 (unsigned long long)task->real_start_time.tv_sec * NSEC_PER_SEC
494                                 + task->real_start_time.tv_nsec;
495         /* convert nsec -> ticks */
496         start_time = nsec_to_clock_t(start_time);
497
498         seq_printf(m, "%d (%s) %c %d %d %d %d %d %u %lu \
499 %lu %lu %lu %lu %lu %ld %ld %ld %ld %d 0 %llu %lu %ld %lu %lu %lu %lu %lu \
500 %lu %lu %lu %lu %lu %lu %lu %lu %d %d %u %u %llu %lu %ld\n",
501                 pid_nr_ns(pid, ns),
502                 tcomm,
503                 state,
504                 ppid,
505                 pgid,
506                 sid,
507                 tty_nr,
508                 tty_pgrp,
509                 task->flags,
510                 min_flt,
511                 cmin_flt,
512                 maj_flt,
513                 cmaj_flt,
514                 cputime_to_clock_t(utime),
515                 cputime_to_clock_t(stime),
516                 cputime_to_clock_t(cutime),
517                 cputime_to_clock_t(cstime),
518                 priority,
519                 nice,
520                 num_threads,
521                 start_time,
522                 vsize,
523                 mm ? get_mm_rss(mm) : 0,
524                 rsslim,
525                 mm ? mm->start_code : 0,
526                 mm ? mm->end_code : 0,
527                 mm ? mm->start_stack : 0,
528                 esp,
529                 eip,
530                 /* The signal information here is obsolete.
531                  * It must be decimal for Linux 2.0 compatibility.
532                  * Use /proc/#/status for real-time signals.
533                  */
534                 task->pending.signal.sig[0] & 0x7fffffffUL,
535                 task->blocked.sig[0] & 0x7fffffffUL,
536                 sigign      .sig[0] & 0x7fffffffUL,
537                 sigcatch    .sig[0] & 0x7fffffffUL,
538                 wchan,
539                 0UL,
540                 0UL,
541                 task->exit_signal,
542                 task_cpu(task),
543                 task->rt_priority,
544                 task->policy,
545                 (unsigned long long)delayacct_blkio_ticks(task),
546                 cputime_to_clock_t(gtime),
547                 cputime_to_clock_t(cgtime));
548         if (mm)
549                 mmput(mm);
550         return 0;
551 }
552
553 int proc_tid_stat(struct seq_file *m, struct pid_namespace *ns,
554                         struct pid *pid, struct task_struct *task)
555 {
556         return do_task_stat(m, ns, pid, task, 0);
557 }
558
559 int proc_tgid_stat(struct seq_file *m, struct pid_namespace *ns,
560                         struct pid *pid, struct task_struct *task)
561 {
562         return do_task_stat(m, ns, pid, task, 1);
563 }
564
565 int proc_pid_statm(struct seq_file *m, struct pid_namespace *ns,
566                         struct pid *pid, struct task_struct *task)
567 {
568         int size = 0, resident = 0, shared = 0, text = 0, lib = 0, data = 0;
569         struct mm_struct *mm = get_task_mm(task);
570
571         if (mm) {
572                 size = task_statm(mm, &shared, &text, &data, &resident);
573                 mmput(mm);
574         }
575         seq_printf(m, "%d %d %d %d %d %d %d\n",
576                         size, resident, shared, text, lib, data, 0);
577
578         return 0;
579 }