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