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/times.h>
  77 #include <linux/cpuset.h>
  78 #include <linux/rcupdate.h>
  79 #include <linux/delayacct.h>
  80 #include <linux/seq_file.h>
  81 #include <linux/pid_namespace.h>
  82
  83 #include <asm/pgtable.h>
  84 #include <asm/processor.h>
  85 #include "internal.h"
  86
  87 /* Gcc optimizes away "strlen(x)" for constant x */
  88 #define ADDBUF(buffer, string) \
  89 do { memcpy(buffer, string, strlen(string)); \
  90      buffer += strlen(string); } while (0)
  91
  92 static inline void task_name(struct seq_file *m, struct task_struct *p)
  93 {
  94         int i;
  95         char *buf, *end;
  96         char *name;
  97         char tcomm[sizeof(p->comm)];
  98
  99         get_task_comm(tcomm, p);
 100
 101         seq_printf(m, "Name:\t");
 102         end = m->buf + m->size;
 103         buf = m->buf + m->count;
 104         name = tcomm;
 105         i = sizeof(tcomm);
 106         while (i && (buf < end)) {
 107                 unsigned char c = *name;
 108                 name++;
 109                 i--;
 110                 *buf = c;
 111                 if (!c)
 112                         break;
 113                 if (c == '\\') {
 114                         buf++;
 115                         if (buf < end)
 116                                 *buf++ = c;
 117                         continue;
 118                 }
 119                 if (c == '\n') {
 120                         *buf++ = '\\';
 121                         if (buf < end)
 122                                 *buf++ = 'n';
 123                         continue;
 124                 }
 125                 buf++;
 126         }
 127         m->count = buf - m->buf;
 128         seq_printf(m, "\n");
 129 }
 130
 131 /*
 132  * The task state array is a strange "bitmap" of
 133  * reasons to sleep. Thus "running" is zero, and
 134  * you can test for combinations of others with
 135  * simple bit tests.
 136  */
 137 static const char *task_state_array[] = {
 138         "R (running)",          /*  0 */
 139         "S (sleeping)",         /*  1 */
 140         "D (disk sleep)",       /*  2 */
 141         "T (stopped)",          /*  4 */
 142         "T (tracing stop)",     /*  8 */
 143         "Z (zombie)",           /* 16 */
 144         "X (dead)"              /* 32 */
 145 };
 146
 147 static inline const char *get_task_state(struct task_struct *tsk)
 148 {
 149         unsigned int state = (tsk->state & TASK_REPORT) | tsk->exit_state;
 150         const char **p = &task_state_array[0];
 151
 152         while (state) {
 153                 p++;
 154                 state >>= 1;
 155         }
 156         return *p;
 157 }
 158
 159 static inline void task_state(struct seq_file *m, struct pid_namespace *ns,
 160                                 struct pid *pid, struct task_struct *p)
 161 {
 162         struct group_info *group_info;
 163         int g;
 164         struct fdtable *fdt = NULL;
 165         pid_t ppid, tpid;
 166
 167         rcu_read_lock();
 168         ppid = pid_alive(p) ?
 169                 task_tgid_nr_ns(rcu_dereference(p->real_parent), ns) : 0;
 170         tpid = pid_alive(p) && p->ptrace ?
 171                 task_pid_nr_ns(rcu_dereference(p->parent), ns) : 0;
 172         seq_printf(m,
 173                 "State:\t%s\n"
 174                 "Tgid:\t%d\n"
 175                 "Pid:\t%d\n"
 176                 "PPid:\t%d\n"
 177                 "TracerPid:\t%d\n"
 178                 "Uid:\t%d\t%d\t%d\t%d\n"
 179                 "Gid:\t%d\t%d\t%d\t%d\n",
 180                 get_task_state(p),
 181                 task_tgid_nr_ns(p, ns),
 182                 pid_nr_ns(pid, ns),
 183                 ppid, tpid,
 184                 p->uid, p->euid, p->suid, p->fsuid,
 185                 p->gid, p->egid, p->sgid, p->fsgid);
 186
 187         task_lock(p);
 188         if (p->files)
 189                 fdt = files_fdtable(p->files);
 190         seq_printf(m,
 191                 "FDSize:\t%d\n"
 192                 "Groups:\t",
 193                 fdt ? fdt->max_fds : 0);
 194         rcu_read_unlock();
 195
 196         group_info = p->group_info;
 197         get_group_info(group_info);
 198         task_unlock(p);
 199
 200         for (g = 0; g < min(group_info->ngroups, NGROUPS_SMALL); g++)
 201                 seq_printf(m, "%d ", GROUP_AT(group_info, g));
 202         put_group_info(group_info);
 203
 204         seq_printf(m, "\n");
 205 }
 206
 207 static void render_sigset_t(struct seq_file *m, const char *header,
 208                                 sigset_t *set)
 209 {
 210         int i;
 211
 212         seq_printf(m, "%s", header);
 213
 214         i = _NSIG;
 215         do {
 216                 int x = 0;
 217
 218                 i -= 4;
 219                 if (sigismember(set, i+1)) x |= 1;
 220                 if (sigismember(set, i+2)) x |= 2;
 221                 if (sigismember(set, i+3)) x |= 4;
 222                 if (sigismember(set, i+4)) x |= 8;
 223                 seq_printf(m, "%x", x);
 224         } while (i >= 4);
 225
 226         seq_printf(m, "\n");
 227 }
 228
 229 static void collect_sigign_sigcatch(struct task_struct *p, sigset_t *ign,
 230                                     sigset_t *catch)
 231 {
 232         struct k_sigaction *k;
 233         int i;
 234
 235         k = p->sighand->action;
 236         for (i = 1; i <= _NSIG; ++i, ++k) {
 237                 if (k->sa.sa_handler == SIG_IGN)
 238                         sigaddset(ign, i);
 239                 else if (k->sa.sa_handler != SIG_DFL)
 240                         sigaddset(catch, i);
 241         }
 242 }
 243
 244 static inline void task_sig(struct seq_file *m, struct task_struct *p)
 245 {
 246         unsigned long flags;
 247         sigset_t pending, shpending, blocked, ignored, caught;
 248         int num_threads = 0;
 249         unsigned long qsize = 0;
 250         unsigned long qlim = 0;
 251
 252         sigemptyset(&pending);
 253         sigemptyset(&shpending);
 254         sigemptyset(&blocked);
 255         sigemptyset(&ignored);
 256         sigemptyset(&caught);
 257
 258         rcu_read_lock();
 259         if (lock_task_sighand(p, &flags)) {
 260                 pending = p->pending.signal;
 261                 shpending = p->signal->shared_pending.signal;
 262                 blocked = p->blocked;
 263                 collect_sigign_sigcatch(p, &ignored, &caught);
 264                 num_threads = atomic_read(&p->signal->count);
 265                 qsize = atomic_read(&p->user->sigpending);
 266                 qlim = p->signal->rlim[RLIMIT_SIGPENDING].rlim_cur;
 267                 unlock_task_sighand(p, &flags);
 268         }
 269         rcu_read_unlock();
 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[(_LINUX_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         render_cap_t(m, "CapInh:\t", &p->cap_inheritable);
 298         render_cap_t(m, "CapPrm:\t", &p->cap_permitted);
 299         render_cap_t(m, "CapEff:\t", &p->cap_effective);
 300 }
 301
 302 static inline void task_context_switch_counts(struct seq_file *m,
 303                                                 struct task_struct *p)
 304 {
 305         seq_printf(m,   "voluntary_ctxt_switches:\t%lu\n"
 306                         "nonvoluntary_ctxt_switches:\t%lu\n",
 307                         p->nvcsw,
 308                         p->nivcsw);
 309 }
 310
 311 int proc_pid_status(struct seq_file *m, struct pid_namespace *ns,
 312                         struct pid *pid, struct task_struct *task)
 313 {
 314         struct mm_struct *mm = get_task_mm(task);
 315
 316         task_name(m, task);
 317         task_state(m, ns, pid, task);
 318
 319         if (mm) {
 320                 task_mem(m, mm);
 321                 mmput(mm);
 322         }
 323         task_sig(m, task);
 324         task_cap(m, task);
 325         cpuset_task_status_allowed(m, task);
 326 #if defined(CONFIG_S390)
 327         task_show_regs(m, task);
 328 #endif
 329         task_context_switch_counts(m, task);
 330         return 0;
 331 }
 332
 333 /*
 334  * Use precise platform statistics if available:
 335  */
 336 #ifdef CONFIG_VIRT_CPU_ACCOUNTING
 337 static cputime_t task_utime(struct task_struct *p)
 338 {
 339         return p->utime;
 340 }
 341
 342 static cputime_t task_stime(struct task_struct *p)
 343 {
 344         return p->stime;
 345 }
 346 #else
 347 static cputime_t task_utime(struct task_struct *p)
 348 {
 349         clock_t utime = cputime_to_clock_t(p->utime),
 350                 total = utime + cputime_to_clock_t(p->stime);
 351         u64 temp;
 352
 353         /*
 354          * Use CFS's precise accounting:
 355          */
 356         temp = (u64)nsec_to_clock_t(p->se.sum_exec_runtime);
 357
 358         if (total) {
 359                 temp *= utime;
 360                 do_div(temp, total);
 361         }
 362         utime = (clock_t)temp;
 363
 364         p->prev_utime = max(p->prev_utime, clock_t_to_cputime(utime));
 365         return p->prev_utime;
 366 }
 367
 368 static cputime_t task_stime(struct task_struct *p)
 369 {
 370         clock_t stime;
 371
 372         /*
 373          * Use CFS's precise accounting. (we subtract utime from
 374          * the total, to make sure the total observed by userspace
 375          * grows monotonically - apps rely on that):
 376          */
 377         stime = nsec_to_clock_t(p->se.sum_exec_runtime) -
 378                         cputime_to_clock_t(task_utime(p));
 379
 380         if (stime >= 0)
 381                 p->prev_stime = max(p->prev_stime, clock_t_to_cputime(stime));
 382
 383         return p->prev_stime;
 384 }
 385 #endif
 386
 387 static cputime_t task_gtime(struct task_struct *p)
 388 {
 389         return p->gtime;
 390 }
 391
 392 static int do_task_stat(struct seq_file *m, struct pid_namespace *ns,
 393                         struct pid *pid, struct task_struct *task, int whole)
 394 {
 395         unsigned long vsize, eip, esp, wchan = ~0UL;
 396         long priority, nice;
 397         int tty_pgrp = -1, tty_nr = 0;
 398         sigset_t sigign, sigcatch;
 399         char state;
 400         pid_t ppid = 0, pgid = -1, sid = -1;
 401         int num_threads = 0;
 402         struct mm_struct *mm;
 403         unsigned long long start_time;
 404         unsigned long cmin_flt = 0, cmaj_flt = 0;
 405         unsigned long  min_flt = 0,  maj_flt = 0;
 406         cputime_t cutime, cstime, utime, stime;
 407         cputime_t cgtime, gtime;
 408         unsigned long rsslim = 0;
 409         char tcomm[sizeof(task->comm)];
 410         unsigned long flags;
 411
 412         state = *get_task_state(task);
 413         vsize = eip = esp = 0;
 414         mm = get_task_mm(task);
 415         if (mm) {
 416                 vsize = task_vsize(mm);
 417                 eip = KSTK_EIP(task);
 418                 esp = KSTK_ESP(task);
 419         }
 420
 421         get_task_comm(tcomm, task);
 422
 423         sigemptyset(&sigign);
 424         sigemptyset(&sigcatch);
 425         cutime = cstime = utime = stime = cputime_zero;
 426         cgtime = gtime = cputime_zero;
 427
 428         rcu_read_lock();
 429         if (lock_task_sighand(task, &flags)) {
 430                 struct signal_struct *sig = task->signal;
 431
 432                 if (sig->tty) {
 433                         tty_pgrp = pid_nr_ns(sig->tty->pgrp, ns);
 434                         tty_nr = new_encode_dev(tty_devnum(sig->tty));
 435                 }
 436
 437                 num_threads = atomic_read(&sig->count);
 438                 collect_sigign_sigcatch(task, &sigign, &sigcatch);
 439
 440                 cmin_flt = sig->cmin_flt;
 441                 cmaj_flt = sig->cmaj_flt;
 442                 cutime = sig->cutime;
 443                 cstime = sig->cstime;
 444                 cgtime = sig->cgtime;
 445                 rsslim = sig->rlim[RLIMIT_RSS].rlim_cur;
 446
 447                 /* add up live thread stats at the group level */
 448                 if (whole) {
 449                         struct task_struct *t = task;
 450                         do {
 451                                 min_flt += t->min_flt;
 452                                 maj_flt += t->maj_flt;
 453                                 utime = cputime_add(utime, task_utime(t));
 454                                 stime = cputime_add(stime, task_stime(t));
 455                                 gtime = cputime_add(gtime, task_gtime(t));
 456                                 t = next_thread(t);
 457                         } while (t != task);
 458
 459                         min_flt += sig->min_flt;
 460                         maj_flt += sig->maj_flt;
 461                         utime = cputime_add(utime, sig->utime);
 462                         stime = cputime_add(stime, sig->stime);
 463                         gtime = cputime_add(gtime, sig->gtime);
 464                 }
 465
 466                 sid = task_session_nr_ns(task, ns);
 467                 ppid = task_tgid_nr_ns(task->real_parent, ns);
 468                 pgid = task_pgrp_nr_ns(task, ns);
 469
 470                 unlock_task_sighand(task, &flags);
 471         }
 472         rcu_read_unlock();
 473
 474         if (!whole || num_threads < 2)
 475                 wchan = get_wchan(task);
 476         if (!whole) {
 477                 min_flt = task->min_flt;
 478                 maj_flt = task->maj_flt;
 479                 utime = task_utime(task);
 480                 stime = task_stime(task);
 481                 gtime = task_gtime(task);
 482         }
 483
 484         /* scale priority and nice values from timeslices to -20..20 */
 485         /* to make it look like a "normal" Unix priority/nice value  */
 486         priority = task_prio(task);
 487         nice = task_nice(task);
 488
 489         /* Temporary variable needed for gcc-2.96 */
 490         /* convert timespec -> nsec*/
 491         start_time =
 492                 (unsigned long long)task->real_start_time.tv_sec * NSEC_PER_SEC
 493                                 + task->real_start_time.tv_nsec;
 494         /* convert nsec -> ticks */
 495         start_time = nsec_to_clock_t(start_time);
 496
 497         seq_printf(m, "%d (%s) %c %d %d %d %d %d %u %lu \
 498 %lu %lu %lu %lu %lu %ld %ld %ld %ld %d 0 %llu %lu %ld %lu %lu %lu %lu %lu \
 499 %lu %lu %lu %lu %lu %lu %lu %lu %d %d %u %u %llu %lu %ld\n",
 500                 pid_nr_ns(pid, ns),
 501                 tcomm,
 502                 state,
 503                 ppid,
 504                 pgid,
 505                 sid,
 506                 tty_nr,
 507                 tty_pgrp,
 508                 task->flags,
 509                 min_flt,
 510                 cmin_flt,
 511                 maj_flt,
 512                 cmaj_flt,
 513                 cputime_to_clock_t(utime),
 514                 cputime_to_clock_t(stime),
 515                 cputime_to_clock_t(cutime),
 516                 cputime_to_clock_t(cstime),
 517                 priority,
 518                 nice,
 519                 num_threads,
 520                 start_time,
 521                 vsize,
 522                 mm ? get_mm_rss(mm) : 0,
 523                 rsslim,
 524                 mm ? mm->start_code : 0,
 525                 mm ? mm->end_code : 0,
 526                 mm ? mm->start_stack : 0,
 527                 esp,
 528                 eip,
 529                 /* The signal information here is obsolete.
 530                  * It must be decimal for Linux 2.0 compatibility.
 531                  * Use /proc/#/status for real-time signals.
 532                  */
 533                 task->pending.signal.sig[0] & 0x7fffffffUL,
 534                 task->blocked.sig[0] & 0x7fffffffUL,
 535                 sigign      .sig[0] & 0x7fffffffUL,
 536                 sigcatch    .sig[0] & 0x7fffffffUL,
 537                 wchan,
 538                 0UL,
 539                 0UL,
 540                 task->exit_signal,
 541                 task_cpu(task),
 542                 task->rt_priority,
 543                 task->policy,
 544                 (unsigned long long)delayacct_blkio_ticks(task),
 545                 cputime_to_clock_t(gtime),
 546                 cputime_to_clock_t(cgtime));
 547         if (mm)
 548                 mmput(mm);
 549         return 0;
 550 }
 551
 552 int proc_tid_stat(struct seq_file *m, struct pid_namespace *ns,
 553                         struct pid *pid, struct task_struct *task)
 554 {
 555         return do_task_stat(m, ns, pid, task, 0);
 556 }
 557
 558 int proc_tgid_stat(struct seq_file *m, struct pid_namespace *ns,
 559                         struct pid *pid, struct task_struct *task)
 560 {
 561         return do_task_stat(m, ns, pid, task, 1);
 562 }
 563
 564 int proc_pid_statm(struct seq_file *m, struct pid_namespace *ns,
 565                         struct pid *pid, struct task_struct *task)
 566 {
 567         int size = 0, resident = 0, shared = 0, text = 0, lib = 0, data = 0;
 568         struct mm_struct *mm = get_task_mm(task);
 569
 570         if (mm) {
 571                 size = task_statm(mm, &shared, &text, &data, &resident);
 572                 mmput(mm);
 573         }
 574         seq_printf(m, "%d %d %d %d %d %d %d\n",
 575                         size, resident, shared, text, lib, data, 0);
 576
 577         return 0;
 578 }