sched: Assert task state bits at build time
[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/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 #include <linux/ptrace.h>
84 #include <linux/tracehook.h>
85 #include <linux/swapops.h>
86
87 #include <asm/pgtable.h>
88 #include <asm/processor.h>
89 #include "internal.h"
90
91 static inline void task_name(struct seq_file *m, struct task_struct *p)
92 {
93         int i;
94         char *buf, *end;
95         char *name;
96         char tcomm[sizeof(p->comm)];
97
98         get_task_comm(tcomm, p);
99
100         seq_printf(m, "Name:\t");
101         end = m->buf + m->size;
102         buf = m->buf + m->count;
103         name = tcomm;
104         i = sizeof(tcomm);
105         while (i && (buf < end)) {
106                 unsigned char c = *name;
107                 name++;
108                 i--;
109                 *buf = c;
110                 if (!c)
111                         break;
112                 if (c == '\\') {
113                         buf++;
114                         if (buf < end)
115                                 *buf++ = c;
116                         continue;
117                 }
118                 if (c == '\n') {
119                         *buf++ = '\\';
120                         if (buf < end)
121                                 *buf++ = 'n';
122                         continue;
123                 }
124                 buf++;
125         }
126         m->count = buf - m->buf;
127         seq_printf(m, "\n");
128 }
129
130 /*
131  * The task state array is a strange "bitmap" of
132  * reasons to sleep. Thus "running" is zero, and
133  * you can test for combinations of others with
134  * simple bit tests.
135  */
136 static const char *task_state_array[] = {
137         "R (running)",          /*   0 */
138         "S (sleeping)",         /*   1 */
139         "D (disk sleep)",       /*   2 */
140         "T (stopped)",          /*   4 */
141         "t (tracing stop)",     /*   8 */
142         "Z (zombie)",           /*  16 */
143         "X (dead)",             /*  32 */
144         "x (dead)",             /*  64 */
145         "K (wakekill)",         /* 128 */
146         "W (waking)",           /* 256 */
147 };
148
149 static inline const char *get_task_state(struct task_struct *tsk)
150 {
151         unsigned int state = (tsk->state & TASK_REPORT) | tsk->exit_state;
152         const char **p = &task_state_array[0];
153
154         BUILD_BUG_ON(1 + ilog2(TASK_STATE_MAX) != ARRAY_SIZE(task_state_array));
155
156         while (state) {
157                 p++;
158                 state >>= 1;
159         }
160         return *p;
161 }
162
163 static inline void task_state(struct seq_file *m, struct pid_namespace *ns,
164                                 struct pid *pid, struct task_struct *p)
165 {
166         struct group_info *group_info;
167         int g;
168         struct fdtable *fdt = NULL;
169         const struct cred *cred;
170         pid_t ppid, tpid;
171
172         rcu_read_lock();
173         ppid = pid_alive(p) ?
174                 task_tgid_nr_ns(rcu_dereference(p->real_parent), ns) : 0;
175         tpid = 0;
176         if (pid_alive(p)) {
177                 struct task_struct *tracer = tracehook_tracer_task(p);
178                 if (tracer)
179                         tpid = task_pid_nr_ns(tracer, ns);
180         }
181         cred = get_cred((struct cred *) __task_cred(p));
182         seq_printf(m,
183                 "State:\t%s\n"
184                 "Tgid:\t%d\n"
185                 "Pid:\t%d\n"
186                 "PPid:\t%d\n"
187                 "TracerPid:\t%d\n"
188                 "Uid:\t%d\t%d\t%d\t%d\n"
189                 "Gid:\t%d\t%d\t%d\t%d\n",
190                 get_task_state(p),
191                 task_tgid_nr_ns(p, ns),
192                 pid_nr_ns(pid, ns),
193                 ppid, tpid,
194                 cred->uid, cred->euid, cred->suid, cred->fsuid,
195                 cred->gid, cred->egid, cred->sgid, cred->fsgid);
196
197         task_lock(p);
198         if (p->files)
199                 fdt = files_fdtable(p->files);
200         seq_printf(m,
201                 "FDSize:\t%d\n"
202                 "Groups:\t",
203                 fdt ? fdt->max_fds : 0);
204         rcu_read_unlock();
205
206         group_info = cred->group_info;
207         task_unlock(p);
208
209         for (g = 0; g < min(group_info->ngroups, NGROUPS_SMALL); g++)
210                 seq_printf(m, "%d ", GROUP_AT(group_info, g));
211         put_cred(cred);
212
213         seq_printf(m, "\n");
214 }
215
216 static void render_sigset_t(struct seq_file *m, const char *header,
217                                 sigset_t *set)
218 {
219         int i;
220
221         seq_printf(m, "%s", header);
222
223         i = _NSIG;
224         do {
225                 int x = 0;
226
227                 i -= 4;
228                 if (sigismember(set, i+1)) x |= 1;
229                 if (sigismember(set, i+2)) x |= 2;
230                 if (sigismember(set, i+3)) x |= 4;
231                 if (sigismember(set, i+4)) x |= 8;
232                 seq_printf(m, "%x", x);
233         } while (i >= 4);
234
235         seq_printf(m, "\n");
236 }
237
238 static void collect_sigign_sigcatch(struct task_struct *p, sigset_t *ign,
239                                     sigset_t *catch)
240 {
241         struct k_sigaction *k;
242         int i;
243
244         k = p->sighand->action;
245         for (i = 1; i <= _NSIG; ++i, ++k) {
246                 if (k->sa.sa_handler == SIG_IGN)
247                         sigaddset(ign, i);
248                 else if (k->sa.sa_handler != SIG_DFL)
249                         sigaddset(catch, i);
250         }
251 }
252
253 static inline void task_sig(struct seq_file *m, struct task_struct *p)
254 {
255         unsigned long flags;
256         sigset_t pending, shpending, blocked, ignored, caught;
257         int num_threads = 0;
258         unsigned long qsize = 0;
259         unsigned long qlim = 0;
260
261         sigemptyset(&pending);
262         sigemptyset(&shpending);
263         sigemptyset(&blocked);
264         sigemptyset(&ignored);
265         sigemptyset(&caught);
266
267         if (lock_task_sighand(p, &flags)) {
268                 pending = p->pending.signal;
269                 shpending = p->signal->shared_pending.signal;
270                 blocked = p->blocked;
271                 collect_sigign_sigcatch(p, &ignored, &caught);
272                 num_threads = atomic_read(&p->signal->count);
273                 qsize = atomic_read(&__task_cred(p)->user->sigpending);
274                 qlim = p->signal->rlim[RLIMIT_SIGPENDING].rlim_cur;
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 #ifdef CONFIG_MMU
331
332 struct stack_stats {
333         struct vm_area_struct *vma;
334         unsigned long   startpage;
335         unsigned long   usage;
336 };
337
338 static int stack_usage_pte_range(pmd_t *pmd, unsigned long addr,
339                                 unsigned long end, struct mm_walk *walk)
340 {
341         struct stack_stats *ss = walk->private;
342         struct vm_area_struct *vma = ss->vma;
343         pte_t *pte, ptent;
344         spinlock_t *ptl;
345         int ret = 0;
346
347         pte = pte_offset_map_lock(vma->vm_mm, pmd, addr, &ptl);
348         for (; addr != end; pte++, addr += PAGE_SIZE) {
349                 ptent = *pte;
350
351 #ifdef CONFIG_STACK_GROWSUP
352                 if (pte_present(ptent) || is_swap_pte(ptent))
353                         ss->usage = addr - ss->startpage + PAGE_SIZE;
354 #else
355                 if (pte_present(ptent) || is_swap_pte(ptent)) {
356                         ss->usage = ss->startpage - addr + PAGE_SIZE;
357                         pte++;
358                         ret = 1;
359                         break;
360                 }
361 #endif
362         }
363         pte_unmap_unlock(pte - 1, ptl);
364         cond_resched();
365         return ret;
366 }
367
368 static inline unsigned long get_stack_usage_in_bytes(struct vm_area_struct *vma,
369                                 struct task_struct *task)
370 {
371         struct stack_stats ss;
372         struct mm_walk stack_walk = {
373                 .pmd_entry = stack_usage_pte_range,
374                 .mm = vma->vm_mm,
375                 .private = &ss,
376         };
377
378         if (!vma->vm_mm || is_vm_hugetlb_page(vma))
379                 return 0;
380
381         ss.vma = vma;
382         ss.startpage = task->stack_start & PAGE_MASK;
383         ss.usage = 0;
384
385 #ifdef CONFIG_STACK_GROWSUP
386         walk_page_range(KSTK_ESP(task) & PAGE_MASK, vma->vm_end,
387                 &stack_walk);
388 #else
389         walk_page_range(vma->vm_start, (KSTK_ESP(task) & PAGE_MASK) + PAGE_SIZE,
390                 &stack_walk);
391 #endif
392         return ss.usage;
393 }
394
395 static inline void task_show_stack_usage(struct seq_file *m,
396                                                 struct task_struct *task)
397 {
398         struct vm_area_struct   *vma;
399         struct mm_struct        *mm = get_task_mm(task);
400
401         if (mm) {
402                 down_read(&mm->mmap_sem);
403                 vma = find_vma(mm, task->stack_start);
404                 if (vma)
405                         seq_printf(m, "Stack usage:\t%lu kB\n",
406                                 get_stack_usage_in_bytes(vma, task) >> 10);
407
408                 up_read(&mm->mmap_sem);
409                 mmput(mm);
410         }
411 }
412 #else
413 static void task_show_stack_usage(struct seq_file *m, struct task_struct *task)
414 {
415 }
416 #endif          /* CONFIG_MMU */
417
418 static void task_cpus_allowed(struct seq_file *m, struct task_struct *task)
419 {
420         seq_printf(m, "Cpus_allowed:\t");
421         seq_cpumask(m, &task->cpus_allowed);
422         seq_printf(m, "\n");
423         seq_printf(m, "Cpus_allowed_list:\t");
424         seq_cpumask_list(m, &task->cpus_allowed);
425         seq_printf(m, "\n");
426 }
427
428 int proc_pid_status(struct seq_file *m, struct pid_namespace *ns,
429                         struct pid *pid, struct task_struct *task)
430 {
431         struct mm_struct *mm = get_task_mm(task);
432
433         task_name(m, task);
434         task_state(m, ns, pid, task);
435
436         if (mm) {
437                 task_mem(m, mm);
438                 mmput(mm);
439         }
440         task_sig(m, task);
441         task_cap(m, task);
442         task_cpus_allowed(m, task);
443         cpuset_task_status_allowed(m, task);
444 #if defined(CONFIG_S390)
445         task_show_regs(m, task);
446 #endif
447         task_context_switch_counts(m, task);
448         task_show_stack_usage(m, task);
449         return 0;
450 }
451
452 static int do_task_stat(struct seq_file *m, struct pid_namespace *ns,
453                         struct pid *pid, struct task_struct *task, int whole)
454 {
455         unsigned long vsize, eip, esp, wchan = ~0UL;
456         long priority, nice;
457         int tty_pgrp = -1, tty_nr = 0;
458         sigset_t sigign, sigcatch;
459         char state;
460         pid_t ppid = 0, pgid = -1, sid = -1;
461         int num_threads = 0;
462         int permitted;
463         struct mm_struct *mm;
464         unsigned long long start_time;
465         unsigned long cmin_flt = 0, cmaj_flt = 0;
466         unsigned long  min_flt = 0,  maj_flt = 0;
467         cputime_t cutime, cstime, utime, stime;
468         cputime_t cgtime, gtime;
469         unsigned long rsslim = 0;
470         char tcomm[sizeof(task->comm)];
471         unsigned long flags;
472
473         state = *get_task_state(task);
474         vsize = eip = esp = 0;
475         permitted = ptrace_may_access(task, PTRACE_MODE_READ);
476         mm = get_task_mm(task);
477         if (mm) {
478                 vsize = task_vsize(mm);
479                 if (permitted) {
480                         eip = KSTK_EIP(task);
481                         esp = KSTK_ESP(task);
482                 }
483         }
484
485         get_task_comm(tcomm, task);
486
487         sigemptyset(&sigign);
488         sigemptyset(&sigcatch);
489         cutime = cstime = utime = stime = cputime_zero;
490         cgtime = gtime = cputime_zero;
491
492         if (lock_task_sighand(task, &flags)) {
493                 struct signal_struct *sig = task->signal;
494
495                 if (sig->tty) {
496                         struct pid *pgrp = tty_get_pgrp(sig->tty);
497                         tty_pgrp = pid_nr_ns(pgrp, ns);
498                         put_pid(pgrp);
499                         tty_nr = new_encode_dev(tty_devnum(sig->tty));
500                 }
501
502                 num_threads = atomic_read(&sig->count);
503                 collect_sigign_sigcatch(task, &sigign, &sigcatch);
504
505                 cmin_flt = sig->cmin_flt;
506                 cmaj_flt = sig->cmaj_flt;
507                 cutime = sig->cutime;
508                 cstime = sig->cstime;
509                 cgtime = sig->cgtime;
510                 rsslim = sig->rlim[RLIMIT_RSS].rlim_cur;
511
512                 /* add up live thread stats at the group level */
513                 if (whole) {
514                         struct task_struct *t = task;
515                         do {
516                                 min_flt += t->min_flt;
517                                 maj_flt += t->maj_flt;
518                                 gtime = cputime_add(gtime, t->gtime);
519                                 t = next_thread(t);
520                         } while (t != task);
521
522                         min_flt += sig->min_flt;
523                         maj_flt += sig->maj_flt;
524                         thread_group_times(task, &utime, &stime);
525                         gtime = cputime_add(gtime, sig->gtime);
526                 }
527
528                 sid = task_session_nr_ns(task, ns);
529                 ppid = task_tgid_nr_ns(task->real_parent, ns);
530                 pgid = task_pgrp_nr_ns(task, ns);
531
532                 unlock_task_sighand(task, &flags);
533         }
534
535         if (permitted && (!whole || num_threads < 2))
536                 wchan = get_wchan(task);
537         if (!whole) {
538                 min_flt = task->min_flt;
539                 maj_flt = task->maj_flt;
540                 task_times(task, &utime, &stime);
541                 gtime = task->gtime;
542         }
543
544         /* scale priority and nice values from timeslices to -20..20 */
545         /* to make it look like a "normal" Unix priority/nice value  */
546         priority = task_prio(task);
547         nice = task_nice(task);
548
549         /* Temporary variable needed for gcc-2.96 */
550         /* convert timespec -> nsec*/
551         start_time =
552                 (unsigned long long)task->real_start_time.tv_sec * NSEC_PER_SEC
553                                 + task->real_start_time.tv_nsec;
554         /* convert nsec -> ticks */
555         start_time = nsec_to_clock_t(start_time);
556
557         seq_printf(m, "%d (%s) %c %d %d %d %d %d %u %lu \
558 %lu %lu %lu %lu %lu %ld %ld %ld %ld %d 0 %llu %lu %ld %lu %lu %lu %lu %lu \
559 %lu %lu %lu %lu %lu %lu %lu %lu %d %d %u %u %llu %lu %ld\n",
560                 pid_nr_ns(pid, ns),
561                 tcomm,
562                 state,
563                 ppid,
564                 pgid,
565                 sid,
566                 tty_nr,
567                 tty_pgrp,
568                 task->flags,
569                 min_flt,
570                 cmin_flt,
571                 maj_flt,
572                 cmaj_flt,
573                 cputime_to_clock_t(utime),
574                 cputime_to_clock_t(stime),
575                 cputime_to_clock_t(cutime),
576                 cputime_to_clock_t(cstime),
577                 priority,
578                 nice,
579                 num_threads,
580                 start_time,
581                 vsize,
582                 mm ? get_mm_rss(mm) : 0,
583                 rsslim,
584                 mm ? mm->start_code : 0,
585                 mm ? mm->end_code : 0,
586                 (permitted && mm) ? task->stack_start : 0,
587                 esp,
588                 eip,
589                 /* The signal information here is obsolete.
590                  * It must be decimal for Linux 2.0 compatibility.
591                  * Use /proc/#/status for real-time signals.
592                  */
593                 task->pending.signal.sig[0] & 0x7fffffffUL,
594                 task->blocked.sig[0] & 0x7fffffffUL,
595                 sigign      .sig[0] & 0x7fffffffUL,
596                 sigcatch    .sig[0] & 0x7fffffffUL,
597                 wchan,
598                 0UL,
599                 0UL,
600                 task->exit_signal,
601                 task_cpu(task),
602                 task->rt_priority,
603                 task->policy,
604                 (unsigned long long)delayacct_blkio_ticks(task),
605                 cputime_to_clock_t(gtime),
606                 cputime_to_clock_t(cgtime));
607         if (mm)
608                 mmput(mm);
609         return 0;
610 }
611
612 int proc_tid_stat(struct seq_file *m, struct pid_namespace *ns,
613                         struct pid *pid, struct task_struct *task)
614 {
615         return do_task_stat(m, ns, pid, task, 0);
616 }
617
618 int proc_tgid_stat(struct seq_file *m, struct pid_namespace *ns,
619                         struct pid *pid, struct task_struct *task)
620 {
621         return do_task_stat(m, ns, pid, task, 1);
622 }
623
624 int proc_pid_statm(struct seq_file *m, struct pid_namespace *ns,
625                         struct pid *pid, struct task_struct *task)
626 {
627         int size = 0, resident = 0, shared = 0, text = 0, lib = 0, data = 0;
628         struct mm_struct *mm = get_task_mm(task);
629
630         if (mm) {
631                 size = task_statm(mm, &shared, &text, &data, &resident);
632                 mmput(mm);
633         }
634         seq_printf(m, "%d %d %d %d %d %d %d\n",
635                         size, resident, shared, text, lib, data, 0);
636
637         return 0;
638 }