96361e8fa3a8a6117a0201263ecfeece53e297c6
[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         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                 qsize = atomic_read(&__task_cred(p)->user->sigpending);
272                 qlim = p->signal->rlim[RLIMIT_SIGPENDING].rlim_cur;
273                 unlock_task_sighand(p, &flags);
274         }
275
276         seq_printf(m, "Threads:\t%d\n", num_threads);
277         seq_printf(m, "SigQ:\t%lu/%lu\n", qsize, qlim);
278
279         /* render them all */
280         render_sigset_t(m, "SigPnd:\t", &pending);
281         render_sigset_t(m, "ShdPnd:\t", &shpending);
282         render_sigset_t(m, "SigBlk:\t", &blocked);
283         render_sigset_t(m, "SigIgn:\t", &ignored);
284         render_sigset_t(m, "SigCgt:\t", &caught);
285 }
286
287 static void render_cap_t(struct seq_file *m, const char *header,
288                         kernel_cap_t *a)
289 {
290         unsigned __capi;
291
292         seq_printf(m, "%s", header);
293         CAP_FOR_EACH_U32(__capi) {
294                 seq_printf(m, "%08x",
295                            a->cap[(_KERNEL_CAPABILITY_U32S-1) - __capi]);
296         }
297         seq_printf(m, "\n");
298 }
299
300 static inline void task_cap(struct seq_file *m, struct task_struct *p)
301 {
302         const struct cred *cred;
303         kernel_cap_t cap_inheritable, cap_permitted, cap_effective, cap_bset;
304
305         rcu_read_lock();
306         cred = __task_cred(p);
307         cap_inheritable = cred->cap_inheritable;
308         cap_permitted   = cred->cap_permitted;
309         cap_effective   = cred->cap_effective;
310         cap_bset        = cred->cap_bset;
311         rcu_read_unlock();
312
313         render_cap_t(m, "CapInh:\t", &cap_inheritable);
314         render_cap_t(m, "CapPrm:\t", &cap_permitted);
315         render_cap_t(m, "CapEff:\t", &cap_effective);
316         render_cap_t(m, "CapBnd:\t", &cap_bset);
317 }
318
319 static inline void task_context_switch_counts(struct seq_file *m,
320                                                 struct task_struct *p)
321 {
322         seq_printf(m,   "voluntary_ctxt_switches:\t%lu\n"
323                         "nonvoluntary_ctxt_switches:\t%lu\n",
324                         p->nvcsw,
325                         p->nivcsw);
326 }
327
328 #ifdef CONFIG_MMU
329
330 struct stack_stats {
331         struct vm_area_struct *vma;
332         unsigned long   startpage;
333         unsigned long   usage;
334 };
335
336 static int stack_usage_pte_range(pmd_t *pmd, unsigned long addr,
337                                 unsigned long end, struct mm_walk *walk)
338 {
339         struct stack_stats *ss = walk->private;
340         struct vm_area_struct *vma = ss->vma;
341         pte_t *pte, ptent;
342         spinlock_t *ptl;
343         int ret = 0;
344
345         pte = pte_offset_map_lock(vma->vm_mm, pmd, addr, &ptl);
346         for (; addr != end; pte++, addr += PAGE_SIZE) {
347                 ptent = *pte;
348
349 #ifdef CONFIG_STACK_GROWSUP
350                 if (pte_present(ptent) || is_swap_pte(ptent))
351                         ss->usage = addr - ss->startpage + PAGE_SIZE;
352 #else
353                 if (pte_present(ptent) || is_swap_pte(ptent)) {
354                         ss->usage = ss->startpage - addr + PAGE_SIZE;
355                         pte++;
356                         ret = 1;
357                         break;
358                 }
359 #endif
360         }
361         pte_unmap_unlock(pte - 1, ptl);
362         cond_resched();
363         return ret;
364 }
365
366 static inline unsigned long get_stack_usage_in_bytes(struct vm_area_struct *vma,
367                                 struct task_struct *task)
368 {
369         struct stack_stats ss;
370         struct mm_walk stack_walk = {
371                 .pmd_entry = stack_usage_pte_range,
372                 .mm = vma->vm_mm,
373                 .private = &ss,
374         };
375
376         if (!vma->vm_mm || is_vm_hugetlb_page(vma))
377                 return 0;
378
379         ss.vma = vma;
380         ss.startpage = task->stack_start & PAGE_MASK;
381         ss.usage = 0;
382
383 #ifdef CONFIG_STACK_GROWSUP
384         walk_page_range(KSTK_ESP(task) & PAGE_MASK, vma->vm_end,
385                 &stack_walk);
386 #else
387         walk_page_range(vma->vm_start, (KSTK_ESP(task) & PAGE_MASK) + PAGE_SIZE,
388                 &stack_walk);
389 #endif
390         return ss.usage;
391 }
392
393 static inline void task_show_stack_usage(struct seq_file *m,
394                                                 struct task_struct *task)
395 {
396         struct vm_area_struct   *vma;
397         struct mm_struct        *mm = get_task_mm(task);
398
399         if (mm) {
400                 down_read(&mm->mmap_sem);
401                 vma = find_vma(mm, task->stack_start);
402                 if (vma)
403                         seq_printf(m, "Stack usage:\t%lu kB\n",
404                                 get_stack_usage_in_bytes(vma, task) >> 10);
405
406                 up_read(&mm->mmap_sem);
407                 mmput(mm);
408         }
409 }
410 #else
411 static void task_show_stack_usage(struct seq_file *m, struct task_struct *task)
412 {
413 }
414 #endif          /* CONFIG_MMU */
415
416 static void task_cpus_allowed(struct seq_file *m, struct task_struct *task)
417 {
418         seq_printf(m, "Cpus_allowed:\t");
419         seq_cpumask(m, &task->cpus_allowed);
420         seq_printf(m, "\n");
421         seq_printf(m, "Cpus_allowed_list:\t");
422         seq_cpumask_list(m, &task->cpus_allowed);
423         seq_printf(m, "\n");
424 }
425
426 int proc_pid_status(struct seq_file *m, struct pid_namespace *ns,
427                         struct pid *pid, struct task_struct *task)
428 {
429         struct mm_struct *mm = get_task_mm(task);
430
431         task_name(m, task);
432         task_state(m, ns, pid, task);
433
434         if (mm) {
435                 task_mem(m, mm);
436                 mmput(mm);
437         }
438         task_sig(m, task);
439         task_cap(m, task);
440         task_cpus_allowed(m, task);
441         cpuset_task_status_allowed(m, task);
442 #if defined(CONFIG_S390)
443         task_show_regs(m, task);
444 #endif
445         task_context_switch_counts(m, task);
446         task_show_stack_usage(m, task);
447         return 0;
448 }
449
450 static int do_task_stat(struct seq_file *m, struct pid_namespace *ns,
451                         struct pid *pid, struct task_struct *task, int whole)
452 {
453         unsigned long vsize, eip, esp, wchan = ~0UL;
454         long priority, nice;
455         int tty_pgrp = -1, tty_nr = 0;
456         sigset_t sigign, sigcatch;
457         char state;
458         pid_t ppid = 0, pgid = -1, sid = -1;
459         int num_threads = 0;
460         int permitted;
461         struct mm_struct *mm;
462         unsigned long long start_time;
463         unsigned long cmin_flt = 0, cmaj_flt = 0;
464         unsigned long  min_flt = 0,  maj_flt = 0;
465         cputime_t cutime, cstime, utime, stime;
466         cputime_t cgtime, gtime;
467         unsigned long rsslim = 0;
468         char tcomm[sizeof(task->comm)];
469         unsigned long flags;
470
471         state = *get_task_state(task);
472         vsize = eip = esp = 0;
473         permitted = ptrace_may_access(task, PTRACE_MODE_READ);
474         mm = get_task_mm(task);
475         if (mm) {
476                 vsize = task_vsize(mm);
477                 if (permitted) {
478                         eip = KSTK_EIP(task);
479                         esp = KSTK_ESP(task);
480                 }
481         }
482
483         get_task_comm(tcomm, task);
484
485         sigemptyset(&sigign);
486         sigemptyset(&sigcatch);
487         cutime = cstime = utime = stime = cputime_zero;
488         cgtime = gtime = cputime_zero;
489
490         if (lock_task_sighand(task, &flags)) {
491                 struct signal_struct *sig = task->signal;
492
493                 if (sig->tty) {
494                         struct pid *pgrp = tty_get_pgrp(sig->tty);
495                         tty_pgrp = pid_nr_ns(pgrp, ns);
496                         put_pid(pgrp);
497                         tty_nr = new_encode_dev(tty_devnum(sig->tty));
498                 }
499
500                 num_threads = atomic_read(&sig->count);
501                 collect_sigign_sigcatch(task, &sigign, &sigcatch);
502
503                 cmin_flt = sig->cmin_flt;
504                 cmaj_flt = sig->cmaj_flt;
505                 cutime = sig->cutime;
506                 cstime = sig->cstime;
507                 cgtime = sig->cgtime;
508                 rsslim = sig->rlim[RLIMIT_RSS].rlim_cur;
509
510                 /* add up live thread stats at the group level */
511                 if (whole) {
512                         struct task_struct *t = task;
513                         do {
514                                 min_flt += t->min_flt;
515                                 maj_flt += t->maj_flt;
516                                 gtime = cputime_add(gtime, t->gtime);
517                                 t = next_thread(t);
518                         } while (t != task);
519
520                         min_flt += sig->min_flt;
521                         maj_flt += sig->maj_flt;
522                         thread_group_times(task, &utime, &stime);
523                         gtime = cputime_add(gtime, sig->gtime);
524                 }
525
526                 sid = task_session_nr_ns(task, ns);
527                 ppid = task_tgid_nr_ns(task->real_parent, ns);
528                 pgid = task_pgrp_nr_ns(task, ns);
529
530                 unlock_task_sighand(task, &flags);
531         }
532
533         if (permitted && (!whole || num_threads < 2))
534                 wchan = get_wchan(task);
535         if (!whole) {
536                 min_flt = task->min_flt;
537                 maj_flt = task->maj_flt;
538                 task_times(task, &utime, &stime);
539                 gtime = task->gtime;
540         }
541
542         /* scale priority and nice values from timeslices to -20..20 */
543         /* to make it look like a "normal" Unix priority/nice value  */
544         priority = task_prio(task);
545         nice = task_nice(task);
546
547         /* Temporary variable needed for gcc-2.96 */
548         /* convert timespec -> nsec*/
549         start_time =
550                 (unsigned long long)task->real_start_time.tv_sec * NSEC_PER_SEC
551                                 + task->real_start_time.tv_nsec;
552         /* convert nsec -> ticks */
553         start_time = nsec_to_clock_t(start_time);
554
555         seq_printf(m, "%d (%s) %c %d %d %d %d %d %u %lu \
556 %lu %lu %lu %lu %lu %ld %ld %ld %ld %d 0 %llu %lu %ld %lu %lu %lu %lu %lu \
557 %lu %lu %lu %lu %lu %lu %lu %lu %d %d %u %u %llu %lu %ld\n",
558                 pid_nr_ns(pid, ns),
559                 tcomm,
560                 state,
561                 ppid,
562                 pgid,
563                 sid,
564                 tty_nr,
565                 tty_pgrp,
566                 task->flags,
567                 min_flt,
568                 cmin_flt,
569                 maj_flt,
570                 cmaj_flt,
571                 cputime_to_clock_t(utime),
572                 cputime_to_clock_t(stime),
573                 cputime_to_clock_t(cutime),
574                 cputime_to_clock_t(cstime),
575                 priority,
576                 nice,
577                 num_threads,
578                 start_time,
579                 vsize,
580                 mm ? get_mm_rss(mm) : 0,
581                 rsslim,
582                 mm ? mm->start_code : 0,
583                 mm ? mm->end_code : 0,
584                 (permitted && mm) ? task->stack_start : 0,
585                 esp,
586                 eip,
587                 /* The signal information here is obsolete.
588                  * It must be decimal for Linux 2.0 compatibility.
589                  * Use /proc/#/status for real-time signals.
590                  */
591                 task->pending.signal.sig[0] & 0x7fffffffUL,
592                 task->blocked.sig[0] & 0x7fffffffUL,
593                 sigign      .sig[0] & 0x7fffffffUL,
594                 sigcatch    .sig[0] & 0x7fffffffUL,
595                 wchan,
596                 0UL,
597                 0UL,
598                 task->exit_signal,
599                 task_cpu(task),
600                 task->rt_priority,
601                 task->policy,
602                 (unsigned long long)delayacct_blkio_ticks(task),
603                 cputime_to_clock_t(gtime),
604                 cputime_to_clock_t(cgtime));
605         if (mm)
606                 mmput(mm);
607         return 0;
608 }
609
610 int proc_tid_stat(struct seq_file *m, struct pid_namespace *ns,
611                         struct pid *pid, struct task_struct *task)
612 {
613         return do_task_stat(m, ns, pid, task, 0);
614 }
615
616 int proc_tgid_stat(struct seq_file *m, struct pid_namespace *ns,
617                         struct pid *pid, struct task_struct *task)
618 {
619         return do_task_stat(m, ns, pid, task, 1);
620 }
621
622 int proc_pid_statm(struct seq_file *m, struct pid_namespace *ns,
623                         struct pid *pid, struct task_struct *task)
624 {
625         int size = 0, resident = 0, shared = 0, text = 0, lib = 0, data = 0;
626         struct mm_struct *mm = get_task_mm(task);
627
628         if (mm) {
629                 size = task_statm(mm, &shared, &text, &data, &resident);
630                 mmput(mm);
631         }
632         seq_printf(m, "%d %d %d %d %d %d %d\n",
633                         size, resident, shared, text, lib, data, 0);
634
635         return 0;
636 }