timers: fix itimer/many thread hang
[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.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 #include <linux/tracehook.h>
84
85 #include <asm/pgtable.h>
86 #include <asm/processor.h>
87 #include "internal.h"
88
89 /* Gcc optimizes away "strlen(x)" for constant x */
90 #define ADDBUF(buffer, string) \
91 do { memcpy(buffer, string, strlen(string)); \
92      buffer += strlen(string); } while (0)
93
94 static inline void task_name(struct seq_file *m, struct task_struct *p)
95 {
96         int i;
97         char *buf, *end;
98         char *name;
99         char tcomm[sizeof(p->comm)];
100
101         get_task_comm(tcomm, p);
102
103         seq_printf(m, "Name:\t");
104         end = m->buf + m->size;
105         buf = m->buf + m->count;
106         name = tcomm;
107         i = sizeof(tcomm);
108         while (i && (buf < end)) {
109                 unsigned char c = *name;
110                 name++;
111                 i--;
112                 *buf = c;
113                 if (!c)
114                         break;
115                 if (c == '\\') {
116                         buf++;
117                         if (buf < end)
118                                 *buf++ = c;
119                         continue;
120                 }
121                 if (c == '\n') {
122                         *buf++ = '\\';
123                         if (buf < end)
124                                 *buf++ = 'n';
125                         continue;
126                 }
127                 buf++;
128         }
129         m->count = buf - m->buf;
130         seq_printf(m, "\n");
131 }
132
133 /*
134  * The task state array is a strange "bitmap" of
135  * reasons to sleep. Thus "running" is zero, and
136  * you can test for combinations of others with
137  * simple bit tests.
138  */
139 static const char *task_state_array[] = {
140         "R (running)",          /*  0 */
141         "S (sleeping)",         /*  1 */
142         "D (disk sleep)",       /*  2 */
143         "T (stopped)",          /*  4 */
144         "T (tracing stop)",     /*  8 */
145         "Z (zombie)",           /* 16 */
146         "X (dead)"              /* 32 */
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         pid_t ppid, tpid;
168
169         rcu_read_lock();
170         ppid = pid_alive(p) ?
171                 task_tgid_nr_ns(rcu_dereference(p->real_parent), ns) : 0;
172         tpid = 0;
173         if (pid_alive(p)) {
174                 struct task_struct *tracer = tracehook_tracer_task(p);
175                 if (tracer)
176                         tpid = task_pid_nr_ns(tracer, ns);
177         }
178         seq_printf(m,
179                 "State:\t%s\n"
180                 "Tgid:\t%d\n"
181                 "Pid:\t%d\n"
182                 "PPid:\t%d\n"
183                 "TracerPid:\t%d\n"
184                 "Uid:\t%d\t%d\t%d\t%d\n"
185                 "Gid:\t%d\t%d\t%d\t%d\n",
186                 get_task_state(p),
187                 task_tgid_nr_ns(p, ns),
188                 pid_nr_ns(pid, ns),
189                 ppid, tpid,
190                 p->uid, p->euid, p->suid, p->fsuid,
191                 p->gid, p->egid, p->sgid, p->fsgid);
192
193         task_lock(p);
194         if (p->files)
195                 fdt = files_fdtable(p->files);
196         seq_printf(m,
197                 "FDSize:\t%d\n"
198                 "Groups:\t",
199                 fdt ? fdt->max_fds : 0);
200         rcu_read_unlock();
201
202         group_info = p->group_info;
203         get_group_info(group_info);
204         task_unlock(p);
205
206         for (g = 0; g < min(group_info->ngroups, NGROUPS_SMALL); g++)
207                 seq_printf(m, "%d ", GROUP_AT(group_info, g));
208         put_group_info(group_info);
209
210         seq_printf(m, "\n");
211 }
212
213 static void render_sigset_t(struct seq_file *m, const char *header,
214                                 sigset_t *set)
215 {
216         int i;
217
218         seq_printf(m, "%s", header);
219
220         i = _NSIG;
221         do {
222                 int x = 0;
223
224                 i -= 4;
225                 if (sigismember(set, i+1)) x |= 1;
226                 if (sigismember(set, i+2)) x |= 2;
227                 if (sigismember(set, i+3)) x |= 4;
228                 if (sigismember(set, i+4)) x |= 8;
229                 seq_printf(m, "%x", x);
230         } while (i >= 4);
231
232         seq_printf(m, "\n");
233 }
234
235 static void collect_sigign_sigcatch(struct task_struct *p, sigset_t *ign,
236                                     sigset_t *catch)
237 {
238         struct k_sigaction *k;
239         int i;
240
241         k = p->sighand->action;
242         for (i = 1; i <= _NSIG; ++i, ++k) {
243                 if (k->sa.sa_handler == SIG_IGN)
244                         sigaddset(ign, i);
245                 else if (k->sa.sa_handler != SIG_DFL)
246                         sigaddset(catch, i);
247         }
248 }
249
250 static inline void task_sig(struct seq_file *m, struct task_struct *p)
251 {
252         unsigned long flags;
253         sigset_t pending, shpending, blocked, ignored, caught;
254         int num_threads = 0;
255         unsigned long qsize = 0;
256         unsigned long qlim = 0;
257
258         sigemptyset(&pending);
259         sigemptyset(&shpending);
260         sigemptyset(&blocked);
261         sigemptyset(&ignored);
262         sigemptyset(&caught);
263
264         rcu_read_lock();
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(&p->user->sigpending);
272                 qlim = p->signal->rlim[RLIMIT_SIGPENDING].rlim_cur;
273                 unlock_task_sighand(p, &flags);
274         }
275         rcu_read_unlock();
276
277         seq_printf(m, "Threads:\t%d\n", num_threads);
278         seq_printf(m, "SigQ:\t%lu/%lu\n", qsize, qlim);
279
280         /* render them all */
281         render_sigset_t(m, "SigPnd:\t", &pending);
282         render_sigset_t(m, "ShdPnd:\t", &shpending);
283         render_sigset_t(m, "SigBlk:\t", &blocked);
284         render_sigset_t(m, "SigIgn:\t", &ignored);
285         render_sigset_t(m, "SigCgt:\t", &caught);
286 }
287
288 static void render_cap_t(struct seq_file *m, const char *header,
289                         kernel_cap_t *a)
290 {
291         unsigned __capi;
292
293         seq_printf(m, "%s", header);
294         CAP_FOR_EACH_U32(__capi) {
295                 seq_printf(m, "%08x",
296                            a->cap[(_KERNEL_CAPABILITY_U32S-1) - __capi]);
297         }
298         seq_printf(m, "\n");
299 }
300
301 static inline void task_cap(struct seq_file *m, struct task_struct *p)
302 {
303         render_cap_t(m, "CapInh:\t", &p->cap_inheritable);
304         render_cap_t(m, "CapPrm:\t", &p->cap_permitted);
305         render_cap_t(m, "CapEff:\t", &p->cap_effective);
306         render_cap_t(m, "CapBnd:\t", &p->cap_bset);
307 }
308
309 static inline void task_context_switch_counts(struct seq_file *m,
310                                                 struct task_struct *p)
311 {
312         seq_printf(m,   "voluntary_ctxt_switches:\t%lu\n"
313                         "nonvoluntary_ctxt_switches:\t%lu\n",
314                         p->nvcsw,
315                         p->nivcsw);
316 }
317
318 int proc_pid_status(struct seq_file *m, struct pid_namespace *ns,
319                         struct pid *pid, struct task_struct *task)
320 {
321         struct mm_struct *mm = get_task_mm(task);
322
323         task_name(m, task);
324         task_state(m, ns, pid, task);
325
326         if (mm) {
327                 task_mem(m, mm);
328                 mmput(mm);
329         }
330         task_sig(m, task);
331         task_cap(m, task);
332         cpuset_task_status_allowed(m, task);
333 #if defined(CONFIG_S390)
334         task_show_regs(m, task);
335 #endif
336         task_context_switch_counts(m, task);
337         return 0;
338 }
339
340 static int do_task_stat(struct seq_file *m, struct pid_namespace *ns,
341                         struct pid *pid, struct task_struct *task, int whole)
342 {
343         unsigned long vsize, eip, esp, wchan = ~0UL;
344         long priority, nice;
345         int tty_pgrp = -1, tty_nr = 0;
346         sigset_t sigign, sigcatch;
347         char state;
348         pid_t ppid = 0, pgid = -1, sid = -1;
349         int num_threads = 0;
350         struct mm_struct *mm;
351         unsigned long long start_time;
352         unsigned long cmin_flt = 0, cmaj_flt = 0;
353         unsigned long  min_flt = 0,  maj_flt = 0;
354         cputime_t cutime, cstime, utime, stime;
355         cputime_t cgtime, gtime;
356         unsigned long rsslim = 0;
357         char tcomm[sizeof(task->comm)];
358         unsigned long flags;
359
360         state = *get_task_state(task);
361         vsize = eip = esp = 0;
362         mm = get_task_mm(task);
363         if (mm) {
364                 vsize = task_vsize(mm);
365                 eip = KSTK_EIP(task);
366                 esp = KSTK_ESP(task);
367         }
368
369         get_task_comm(tcomm, task);
370
371         sigemptyset(&sigign);
372         sigemptyset(&sigcatch);
373         cutime = cstime = utime = stime = cputime_zero;
374         cgtime = gtime = cputime_zero;
375
376         if (lock_task_sighand(task, &flags)) {
377                 struct signal_struct *sig = task->signal;
378
379                 if (sig->tty) {
380                         struct pid *pgrp = tty_get_pgrp(sig->tty);
381                         tty_pgrp = pid_nr_ns(pgrp, ns);
382                         put_pid(pgrp);
383                         tty_nr = new_encode_dev(tty_devnum(sig->tty));
384                 }
385
386                 num_threads = atomic_read(&sig->count);
387                 collect_sigign_sigcatch(task, &sigign, &sigcatch);
388
389                 cmin_flt = sig->cmin_flt;
390                 cmaj_flt = sig->cmaj_flt;
391                 cutime = sig->cutime;
392                 cstime = sig->cstime;
393                 cgtime = sig->cgtime;
394                 rsslim = sig->rlim[RLIMIT_RSS].rlim_cur;
395
396                 /* add up live thread stats at the group level */
397                 if (whole) {
398                         struct task_cputime cputime;
399                         struct task_struct *t = task;
400                         do {
401                                 min_flt += t->min_flt;
402                                 maj_flt += t->maj_flt;
403                                 gtime = cputime_add(gtime, task_gtime(t));
404                                 t = next_thread(t);
405                         } while (t != task);
406
407                         min_flt += sig->min_flt;
408                         maj_flt += sig->maj_flt;
409                         thread_group_cputime(task, &cputime);
410                         utime = cputime.utime;
411                         stime = cputime.stime;
412                         gtime = cputime_add(gtime, sig->gtime);
413                 }
414
415                 sid = task_session_nr_ns(task, ns);
416                 ppid = task_tgid_nr_ns(task->real_parent, ns);
417                 pgid = task_pgrp_nr_ns(task, ns);
418
419                 unlock_task_sighand(task, &flags);
420         }
421
422         if (!whole || num_threads < 2)
423                 wchan = get_wchan(task);
424         if (!whole) {
425                 min_flt = task->min_flt;
426                 maj_flt = task->maj_flt;
427                 utime = task_utime(task);
428                 stime = task_stime(task);
429                 gtime = task_gtime(task);
430         }
431
432         /* scale priority and nice values from timeslices to -20..20 */
433         /* to make it look like a "normal" Unix priority/nice value  */
434         priority = task_prio(task);
435         nice = task_nice(task);
436
437         /* Temporary variable needed for gcc-2.96 */
438         /* convert timespec -> nsec*/
439         start_time =
440                 (unsigned long long)task->real_start_time.tv_sec * NSEC_PER_SEC
441                                 + task->real_start_time.tv_nsec;
442         /* convert nsec -> ticks */
443         start_time = nsec_to_clock_t(start_time);
444
445         seq_printf(m, "%d (%s) %c %d %d %d %d %d %u %lu \
446 %lu %lu %lu %lu %lu %ld %ld %ld %ld %d 0 %llu %lu %ld %lu %lu %lu %lu %lu \
447 %lu %lu %lu %lu %lu %lu %lu %lu %d %d %u %u %llu %lu %ld\n",
448                 pid_nr_ns(pid, ns),
449                 tcomm,
450                 state,
451                 ppid,
452                 pgid,
453                 sid,
454                 tty_nr,
455                 tty_pgrp,
456                 task->flags,
457                 min_flt,
458                 cmin_flt,
459                 maj_flt,
460                 cmaj_flt,
461                 cputime_to_clock_t(utime),
462                 cputime_to_clock_t(stime),
463                 cputime_to_clock_t(cutime),
464                 cputime_to_clock_t(cstime),
465                 priority,
466                 nice,
467                 num_threads,
468                 start_time,
469                 vsize,
470                 mm ? get_mm_rss(mm) : 0,
471                 rsslim,
472                 mm ? mm->start_code : 0,
473                 mm ? mm->end_code : 0,
474                 mm ? mm->start_stack : 0,
475                 esp,
476                 eip,
477                 /* The signal information here is obsolete.
478                  * It must be decimal for Linux 2.0 compatibility.
479                  * Use /proc/#/status for real-time signals.
480                  */
481                 task->pending.signal.sig[0] & 0x7fffffffUL,
482                 task->blocked.sig[0] & 0x7fffffffUL,
483                 sigign      .sig[0] & 0x7fffffffUL,
484                 sigcatch    .sig[0] & 0x7fffffffUL,
485                 wchan,
486                 0UL,
487                 0UL,
488                 task->exit_signal,
489                 task_cpu(task),
490                 task->rt_priority,
491                 task->policy,
492                 (unsigned long long)delayacct_blkio_ticks(task),
493                 cputime_to_clock_t(gtime),
494                 cputime_to_clock_t(cgtime));
495         if (mm)
496                 mmput(mm);
497         return 0;
498 }
499
500 int proc_tid_stat(struct seq_file *m, struct pid_namespace *ns,
501                         struct pid *pid, struct task_struct *task)
502 {
503         return do_task_stat(m, ns, pid, task, 0);
504 }
505
506 int proc_tgid_stat(struct seq_file *m, struct pid_namespace *ns,
507                         struct pid *pid, struct task_struct *task)
508 {
509         return do_task_stat(m, ns, pid, task, 1);
510 }
511
512 int proc_pid_statm(struct seq_file *m, struct pid_namespace *ns,
513                         struct pid *pid, struct task_struct *task)
514 {
515         int size = 0, resident = 0, shared = 0, text = 0, lib = 0, data = 0;
516         struct mm_struct *mm = get_task_mm(task);
517
518         if (mm) {
519                 size = task_statm(mm, &shared, &text, &data, &resident);
520                 mmput(mm);
521         }
522         seq_printf(m, "%d %d %d %d %d %d %d\n",
523                         size, resident, shared, text, lib, data, 0);
524
525         return 0;
526 }