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
  84 #include <asm/pgtable.h>
  85 #include <asm/processor.h>
  86 #include "internal.h"
  87
  88 /* Gcc optimizes away "strlen(x)" for constant x */
  89 #define ADDBUF(buffer, string) \
  90 do { memcpy(buffer, string, strlen(string)); \
  91      buffer += strlen(string); } while (0)
  92
  93 static inline void task_name(struct seq_file *m, struct task_struct *p)
  94 {
  95         int i;
  96         char *buf, *end;
  97         char *name;
  98         char tcomm[sizeof(p->comm)];
  99
 100         get_task_comm(tcomm, p);
 101
 102         seq_printf(m, "Name:\t");
 103         end = m->buf + m->size;
 104         buf = m->buf + m->count;
 105         name = tcomm;
 106         i = sizeof(tcomm);
 107         while (i && (buf < end)) {
 108                 unsigned char c = *name;
 109                 name++;
 110                 i--;
 111                 *buf = c;
 112                 if (!c)
 113                         break;
 114                 if (c == '\\') {
 115                         buf++;
 116                         if (buf < end)
 117                                 *buf++ = c;
 118                         continue;
 119                 }
 120                 if (c == '\n') {
 121                         *buf++ = '\\';
 122                         if (buf < end)
 123                                 *buf++ = 'n';
 124                         continue;
 125                 }
 126                 buf++;
 127         }
 128         m->count = buf - m->buf;
 129         seq_printf(m, "\n");
 130 }
 131
 132 /*
 133  * The task state array is a strange "bitmap" of
 134  * reasons to sleep. Thus "running" is zero, and
 135  * you can test for combinations of others with
 136  * simple bit tests.
 137  */
 138 static const char *task_state_array[] = {
 139         "R (running)",          /*  0 */
 140         "S (sleeping)",         /*  1 */
 141         "D (disk sleep)",       /*  2 */
 142         "T (stopped)",          /*  4 */
 143         "T (tracing stop)",     /*  8 */
 144         "Z (zombie)",           /* 16 */
 145         "X (dead)"              /* 32 */
 146 };
 147
 148 static inline const char *get_task_state(struct task_struct *tsk)
 149 {
 150         unsigned int state = (tsk->state & TASK_REPORT) | tsk->exit_state;
 151         const char **p = &task_state_array[0];
 152
 153         while (state) {
 154                 p++;
 155                 state >>= 1;
 156         }
 157         return *p;
 158 }
 159
 160 static inline void task_state(struct seq_file *m, struct pid_namespace *ns,
 161                                 struct pid *pid, struct task_struct *p)
 162 {
 163         struct group_info *group_info;
 164         int g;
 165         struct fdtable *fdt = NULL;
 166         pid_t ppid, tpid;
 167
 168         rcu_read_lock();
 169         ppid = pid_alive(p) ?
 170                 task_tgid_nr_ns(rcu_dereference(p->real_parent), ns) : 0;
 171         tpid = pid_alive(p) && p->ptrace ?
 172                 task_pid_nr_ns(rcu_dereference(p->parent), ns) : 0;
 173         seq_printf(m,
 174                 "State:\t%s\n"
 175                 "Tgid:\t%d\n"
 176                 "Pid:\t%d\n"
 177                 "PPid:\t%d\n"
 178                 "TracerPid:\t%d\n"
 179                 "Uid:\t%d\t%d\t%d\t%d\n"
 180                 "Gid:\t%d\t%d\t%d\t%d\n",
 181                 get_task_state(p),
 182                 task_tgid_nr_ns(p, ns),
 183                 pid_nr_ns(pid, ns),
 184                 ppid, tpid,
 185                 p->uid, p->euid, p->suid, p->fsuid,
 186                 p->gid, p->egid, p->sgid, p->fsgid);
 187
 188         task_lock(p);
 189         if (p->files)
 190                 fdt = files_fdtable(p->files);
 191         seq_printf(m,
 192                 "FDSize:\t%d\n"
 193                 "Groups:\t",
 194                 fdt ? fdt->max_fds : 0);
 195         rcu_read_unlock();
 196
 197         group_info = p->group_info;
 198         get_group_info(group_info);
 199         task_unlock(p);
 200
 201         for (g = 0; g < min(group_info->ngroups, NGROUPS_SMALL); g++)
 202                 seq_printf(m, "%d ", GROUP_AT(group_info, g));
 203         put_group_info(group_info);
 204
 205         seq_printf(m, "\n");
 206 }
 207
 208 static void render_sigset_t(struct seq_file *m, const char *header,
 209                                 sigset_t *set)
 210 {
 211         int i;
 212
 213         seq_printf(m, "%s", header);
 214
 215         i = _NSIG;
 216         do {
 217                 int x = 0;
 218
 219                 i -= 4;
 220                 if (sigismember(set, i+1)) x |= 1;
 221                 if (sigismember(set, i+2)) x |= 2;
 222                 if (sigismember(set, i+3)) x |= 4;
 223                 if (sigismember(set, i+4)) x |= 8;
 224                 seq_printf(m, "%x", x);
 225         } while (i >= 4);
 226
 227         seq_printf(m, "\n");
 228 }
 229
 230 static void collect_sigign_sigcatch(struct task_struct *p, sigset_t *ign,
 231                                     sigset_t *catch)
 232 {
 233         struct k_sigaction *k;
 234         int i;
 235
 236         k = p->sighand->action;
 237         for (i = 1; i <= _NSIG; ++i, ++k) {
 238                 if (k->sa.sa_handler == SIG_IGN)
 239                         sigaddset(ign, i);
 240                 else if (k->sa.sa_handler != SIG_DFL)
 241                         sigaddset(catch, i);
 242         }
 243 }
 244
 245 static inline void task_sig(struct seq_file *m, struct task_struct *p)
 246 {
 247         unsigned long flags;
 248         sigset_t pending, shpending, blocked, ignored, caught;
 249         int num_threads = 0;
 250         unsigned long qsize = 0;
 251         unsigned long qlim = 0;
 252
 253         sigemptyset(&pending);
 254         sigemptyset(&shpending);
 255         sigemptyset(&blocked);
 256         sigemptyset(&ignored);
 257         sigemptyset(&caught);
 258
 259         rcu_read_lock();
 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(&p->user->sigpending);
 267                 qlim = p->signal->rlim[RLIMIT_SIGPENDING].rlim_cur;
 268                 unlock_task_sighand(p, &flags);
 269         }
 270         rcu_read_unlock();
 271
 272         seq_printf(m, "Threads:\t%d\n", num_threads);
 273         seq_printf(m, "SigQ:\t%lu/%lu\n", qsize, qlim);
 274
 275         /* render them all */
 276         render_sigset_t(m, "SigPnd:\t", &pending);
 277         render_sigset_t(m, "ShdPnd:\t", &shpending);
 278         render_sigset_t(m, "SigBlk:\t", &blocked);
 279         render_sigset_t(m, "SigIgn:\t", &ignored);
 280         render_sigset_t(m, "SigCgt:\t", &caught);
 281 }
 282
 283 static void render_cap_t(struct seq_file *m, const char *header,
 284                         kernel_cap_t *a)
 285 {
 286         unsigned __capi;
 287
 288         seq_printf(m, "%s", header);
 289         CAP_FOR_EACH_U32(__capi) {
 290                 seq_printf(m, "%08x",
 291                            a->cap[(_KERNEL_CAPABILITY_U32S-1) - __capi]);
 292         }
 293         seq_printf(m, "\n");
 294 }
 295
 296 static inline void task_cap(struct seq_file *m, struct task_struct *p)
 297 {
 298         render_cap_t(m, "CapInh:\t", &p->cap_inheritable);
 299         render_cap_t(m, "CapPrm:\t", &p->cap_permitted);
 300         render_cap_t(m, "CapEff:\t", &p->cap_effective);
 301         render_cap_t(m, "CapBnd:\t", &p->cap_bset);
 302 }
 303
 304 static inline void task_context_switch_counts(struct seq_file *m,
 305                                                 struct task_struct *p)
 306 {
 307         seq_printf(m,   "voluntary_ctxt_switches:\t%lu\n"
 308                         "nonvoluntary_ctxt_switches:\t%lu\n",
 309                         p->nvcsw,
 310                         p->nivcsw);
 311 }
 312
 313 int proc_pid_status(struct seq_file *m, struct pid_namespace *ns,
 314                         struct pid *pid, struct task_struct *task)
 315 {
 316         struct mm_struct *mm = get_task_mm(task);
 317
 318         task_name(m, task);
 319         task_state(m, ns, pid, task);
 320
 321         if (mm) {
 322                 task_mem(m, mm);
 323                 mmput(mm);
 324         }
 325         task_sig(m, task);
 326         task_cap(m, task);
 327         cpuset_task_status_allowed(m, task);
 328 #if defined(CONFIG_S390)
 329         task_show_regs(m, task);
 330 #endif
 331         task_context_switch_counts(m, task);
 332         return 0;
 333 }
 334
 335 /*
 336  * Use precise platform statistics if available:
 337  */
 338 #ifdef CONFIG_VIRT_CPU_ACCOUNTING
 339 static cputime_t task_utime(struct task_struct *p)
 340 {
 341         return p->utime;
 342 }
 343
 344 static cputime_t task_stime(struct task_struct *p)
 345 {
 346         return p->stime;
 347 }
 348 #else
 349 static cputime_t task_utime(struct task_struct *p)
 350 {
 351         clock_t utime = cputime_to_clock_t(p->utime),
 352                 total = utime + cputime_to_clock_t(p->stime);
 353         u64 temp;
 354
 355         /*
 356          * Use CFS's precise accounting:
 357          */
 358         temp = (u64)nsec_to_clock_t(p->se.sum_exec_runtime);
 359
 360         if (total) {
 361                 temp *= utime;
 362                 do_div(temp, total);
 363         }
 364         utime = (clock_t)temp;
 365
 366         p->prev_utime = max(p->prev_utime, clock_t_to_cputime(utime));
 367         return p->prev_utime;
 368 }
 369
 370 static cputime_t task_stime(struct task_struct *p)
 371 {
 372         clock_t stime;
 373
 374         /*
 375          * Use CFS's precise accounting. (we subtract utime from
 376          * the total, to make sure the total observed by userspace
 377          * grows monotonically - apps rely on that):
 378          */
 379         stime = nsec_to_clock_t(p->se.sum_exec_runtime) -
 380                         cputime_to_clock_t(task_utime(p));
 381
 382         if (stime >= 0)
 383                 p->prev_stime = max(p->prev_stime, clock_t_to_cputime(stime));
 384
 385         return p->prev_stime;
 386 }
 387 #endif
 388
 389 static cputime_t task_gtime(struct task_struct *p)
 390 {
 391         return p->gtime;
 392 }
 393
 394 static int do_task_stat(struct seq_file *m, struct pid_namespace *ns,
 395                         struct pid *pid, struct task_struct *task, int whole)
 396 {
 397         unsigned long vsize, eip, esp, wchan = ~0UL;
 398         long priority, nice;
 399         int tty_pgrp = -1, tty_nr = 0;
 400         sigset_t sigign, sigcatch;
 401         char state;
 402         pid_t ppid = 0, pgid = -1, sid = -1;
 403         int num_threads = 0;
 404         struct mm_struct *mm;
 405         unsigned long long start_time;
 406         unsigned long cmin_flt = 0, cmaj_flt = 0;
 407         unsigned long  min_flt = 0,  maj_flt = 0;
 408         cputime_t cutime, cstime, utime, stime;
 409         cputime_t cgtime, gtime;
 410         unsigned long rsslim = 0;
 411         char tcomm[sizeof(task->comm)];
 412         unsigned long flags;
 413
 414         state = *get_task_state(task);
 415         vsize = eip = esp = 0;
 416         mm = get_task_mm(task);
 417         if (mm) {
 418                 vsize = task_vsize(mm);
 419                 eip = KSTK_EIP(task);
 420                 esp = KSTK_ESP(task);
 421         }
 422
 423         get_task_comm(tcomm, task);
 424
 425         sigemptyset(&sigign);
 426         sigemptyset(&sigcatch);
 427         cutime = cstime = utime = stime = cputime_zero;
 428         cgtime = gtime = cputime_zero;
 429
 430         if (lock_task_sighand(task, &flags)) {
 431                 struct signal_struct *sig = task->signal;
 432
 433                 if (sig->tty) {
 434                         struct pid *pgrp = tty_get_pgrp(sig->tty);
 435                         tty_pgrp = pid_nr_ns(pgrp, ns);
 436                         put_pid(pgrp);
 437                         tty_nr = new_encode_dev(tty_devnum(sig->tty));
 438                 }
 439
 440                 num_threads = atomic_read(&sig->count);
 441                 collect_sigign_sigcatch(task, &sigign, &sigcatch);
 442
 443                 cmin_flt = sig->cmin_flt;
 444                 cmaj_flt = sig->cmaj_flt;
 445                 cutime = sig->cutime;
 446                 cstime = sig->cstime;
 447                 cgtime = sig->cgtime;
 448                 rsslim = sig->rlim[RLIMIT_RSS].rlim_cur;
 449
 450                 /* add up live thread stats at the group level */
 451                 if (whole) {
 452                         struct task_struct *t = task;
 453                         do {
 454                                 min_flt += t->min_flt;
 455                                 maj_flt += t->maj_flt;
 456                                 utime = cputime_add(utime, task_utime(t));
 457                                 stime = cputime_add(stime, task_stime(t));
 458                                 gtime = cputime_add(gtime, task_gtime(t));
 459                                 t = next_thread(t);
 460                         } while (t != task);
 461
 462                         min_flt += sig->min_flt;
 463                         maj_flt += sig->maj_flt;
 464                         utime = cputime_add(utime, sig->utime);
 465                         stime = cputime_add(stime, sig->stime);
 466                         gtime = cputime_add(gtime, sig->gtime);
 467                 }
 468
 469                 sid = task_session_nr_ns(task, ns);
 470                 ppid = task_tgid_nr_ns(task->real_parent, ns);
 471                 pgid = task_pgrp_nr_ns(task, ns);
 472
 473                 unlock_task_sighand(task, &flags);
 474         }
 475
 476         if (!whole || num_threads < 2)
 477                 wchan = get_wchan(task);
 478         if (!whole) {
 479                 min_flt = task->min_flt;
 480                 maj_flt = task->maj_flt;
 481                 utime = task_utime(task);
 482                 stime = task_stime(task);
 483                 gtime = task_gtime(task);
 484         }
 485
 486         /* scale priority and nice values from timeslices to -20..20 */
 487         /* to make it look like a "normal" Unix priority/nice value  */
 488         priority = task_prio(task);
 489         nice = task_nice(task);
 490
 491         /* Temporary variable needed for gcc-2.96 */
 492         /* convert timespec -> nsec*/
 493         start_time =
 494                 (unsigned long long)task->real_start_time.tv_sec * NSEC_PER_SEC
 495                                 + task->real_start_time.tv_nsec;
 496         /* convert nsec -> ticks */
 497         start_time = nsec_to_clock_t(start_time);
 498
 499         seq_printf(m, "%d (%s) %c %d %d %d %d %d %u %lu \
 500 %lu %lu %lu %lu %lu %ld %ld %ld %ld %d 0 %llu %lu %ld %lu %lu %lu %lu %lu \
 501 %lu %lu %lu %lu %lu %lu %lu %lu %d %d %u %u %llu %lu %ld\n",
 502                 pid_nr_ns(pid, ns),
 503                 tcomm,
 504                 state,
 505                 ppid,
 506                 pgid,
 507                 sid,
 508                 tty_nr,
 509                 tty_pgrp,
 510                 task->flags,
 511                 min_flt,
 512                 cmin_flt,
 513                 maj_flt,
 514                 cmaj_flt,
 515                 cputime_to_clock_t(utime),
 516                 cputime_to_clock_t(stime),
 517                 cputime_to_clock_t(cutime),
 518                 cputime_to_clock_t(cstime),
 519                 priority,
 520                 nice,
 521                 num_threads,
 522                 start_time,
 523                 vsize,
 524                 mm ? get_mm_rss(mm) : 0,
 525                 rsslim,
 526                 mm ? mm->start_code : 0,
 527                 mm ? mm->end_code : 0,
 528                 mm ? mm->start_stack : 0,
 529                 esp,
 530                 eip,
 531                 /* The signal information here is obsolete.
 532                  * It must be decimal for Linux 2.0 compatibility.
 533                  * Use /proc/#/status for real-time signals.
 534                  */
 535                 task->pending.signal.sig[0] & 0x7fffffffUL,
 536                 task->blocked.sig[0] & 0x7fffffffUL,
 537                 sigign      .sig[0] & 0x7fffffffUL,
 538                 sigcatch    .sig[0] & 0x7fffffffUL,
 539                 wchan,
 540                 0UL,
 541                 0UL,
 542                 task->exit_signal,
 543                 task_cpu(task),
 544                 task->rt_priority,
 545                 task->policy,
 546                 (unsigned long long)delayacct_blkio_ticks(task),
 547                 cputime_to_clock_t(gtime),
 548                 cputime_to_clock_t(cgtime));
 549         if (mm)
 550                 mmput(mm);
 551         return 0;
 552 }
 553
 554 int proc_tid_stat(struct seq_file *m, struct pid_namespace *ns,
 555                         struct pid *pid, struct task_struct *task)
 556 {
 557         return do_task_stat(m, ns, pid, task, 0);
 558 }
 559
 560 int proc_tgid_stat(struct seq_file *m, struct pid_namespace *ns,
 561                         struct pid *pid, struct task_struct *task)
 562 {
 563         return do_task_stat(m, ns, pid, task, 1);
 564 }
 565
 566 int proc_pid_statm(struct seq_file *m, struct pid_namespace *ns,
 567                         struct pid *pid, struct task_struct *task)
 568 {
 569         int size = 0, resident = 0, shared = 0, text = 0, lib = 0, data = 0;
 570         struct mm_struct *mm = get_task_mm(task);
 571
 572         if (mm) {
 573                 size = task_statm(mm, &shared, &text, &data, &resident);
 574                 mmput(mm);
 575         }
 576         seq_printf(m, "%d %d %d %d %d %d %d\n",
 577                         size, resident, shared, text, lib, data, 0);
 578
 579         return 0;
 580 }