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 /*
 341  * Use precise platform statistics if available:
 342  */
 343 #ifdef CONFIG_VIRT_CPU_ACCOUNTING
 344 static cputime_t task_utime(struct task_struct *p)
 345 {
 346         return p->utime;
 347 }
 348
 349 static cputime_t task_stime(struct task_struct *p)
 350 {
 351         return p->stime;
 352 }
 353 #else
 354 static cputime_t task_utime(struct task_struct *p)
 355 {
 356         clock_t utime = cputime_to_clock_t(p->utime),
 357                 total = utime + cputime_to_clock_t(p->stime);
 358         u64 temp;
 359
 360         /*
 361          * Use CFS's precise accounting:
 362          */
 363         temp = (u64)nsec_to_clock_t(p->se.sum_exec_runtime);
 364
 365         if (total) {
 366                 temp *= utime;
 367                 do_div(temp, total);
 368         }
 369         utime = (clock_t)temp;
 370
 371         p->prev_utime = max(p->prev_utime, clock_t_to_cputime(utime));
 372         return p->prev_utime;
 373 }
 374
 375 static cputime_t task_stime(struct task_struct *p)
 376 {
 377         clock_t stime;
 378
 379         /*
 380          * Use CFS's precise accounting. (we subtract utime from
 381          * the total, to make sure the total observed by userspace
 382          * grows monotonically - apps rely on that):
 383          */
 384         stime = nsec_to_clock_t(p->se.sum_exec_runtime) -
 385                         cputime_to_clock_t(task_utime(p));
 386
 387         if (stime >= 0)
 388                 p->prev_stime = max(p->prev_stime, clock_t_to_cputime(stime));
 389
 390         return p->prev_stime;
 391 }
 392 #endif
 393
 394 static cputime_t task_gtime(struct task_struct *p)
 395 {
 396         return p->gtime;
 397 }
 398
 399 static int do_task_stat(struct seq_file *m, struct pid_namespace *ns,
 400                         struct pid *pid, struct task_struct *task, int whole)
 401 {
 402         unsigned long vsize, eip, esp, wchan = ~0UL;
 403         long priority, nice;
 404         int tty_pgrp = -1, tty_nr = 0;
 405         sigset_t sigign, sigcatch;
 406         char state;
 407         pid_t ppid = 0, pgid = -1, sid = -1;
 408         int num_threads = 0;
 409         struct mm_struct *mm;
 410         unsigned long long start_time;
 411         unsigned long cmin_flt = 0, cmaj_flt = 0;
 412         unsigned long  min_flt = 0,  maj_flt = 0;
 413         cputime_t cutime, cstime, utime, stime;
 414         cputime_t cgtime, gtime;
 415         unsigned long rsslim = 0;
 416         char tcomm[sizeof(task->comm)];
 417         unsigned long flags;
 418
 419         state = *get_task_state(task);
 420         vsize = eip = esp = 0;
 421         mm = get_task_mm(task);
 422         if (mm) {
 423                 vsize = task_vsize(mm);
 424                 eip = KSTK_EIP(task);
 425                 esp = KSTK_ESP(task);
 426         }
 427
 428         get_task_comm(tcomm, task);
 429
 430         sigemptyset(&sigign);
 431         sigemptyset(&sigcatch);
 432         cutime = cstime = utime = stime = cputime_zero;
 433         cgtime = gtime = cputime_zero;
 434
 435         if (lock_task_sighand(task, &flags)) {
 436                 struct signal_struct *sig = task->signal;
 437
 438                 if (sig->tty) {
 439                         struct pid *pgrp = tty_get_pgrp(sig->tty);
 440                         tty_pgrp = pid_nr_ns(pgrp, ns);
 441                         put_pid(pgrp);
 442                         tty_nr = new_encode_dev(tty_devnum(sig->tty));
 443                 }
 444
 445                 num_threads = atomic_read(&sig->count);
 446                 collect_sigign_sigcatch(task, &sigign, &sigcatch);
 447
 448                 cmin_flt = sig->cmin_flt;
 449                 cmaj_flt = sig->cmaj_flt;
 450                 cutime = sig->cutime;
 451                 cstime = sig->cstime;
 452                 cgtime = sig->cgtime;
 453                 rsslim = sig->rlim[RLIMIT_RSS].rlim_cur;
 454
 455                 /* add up live thread stats at the group level */
 456                 if (whole) {
 457                         struct task_struct *t = task;
 458                         do {
 459                                 min_flt += t->min_flt;
 460                                 maj_flt += t->maj_flt;
 461                                 utime = cputime_add(utime, task_utime(t));
 462                                 stime = cputime_add(stime, task_stime(t));
 463                                 gtime = cputime_add(gtime, task_gtime(t));
 464                                 t = next_thread(t);
 465                         } while (t != task);
 466
 467                         min_flt += sig->min_flt;
 468                         maj_flt += sig->maj_flt;
 469                         utime = cputime_add(utime, sig->utime);
 470                         stime = cputime_add(stime, sig->stime);
 471                         gtime = cputime_add(gtime, sig->gtime);
 472                 }
 473
 474                 sid = task_session_nr_ns(task, ns);
 475                 ppid = task_tgid_nr_ns(task->real_parent, ns);
 476                 pgid = task_pgrp_nr_ns(task, ns);
 477
 478                 unlock_task_sighand(task, &flags);
 479         }
 480
 481         if (!whole || num_threads < 2)
 482                 wchan = get_wchan(task);
 483         if (!whole) {
 484                 min_flt = task->min_flt;
 485                 maj_flt = task->maj_flt;
 486                 utime = task_utime(task);
 487                 stime = task_stime(task);
 488                 gtime = task_gtime(task);
 489         }
 490
 491         /* scale priority and nice values from timeslices to -20..20 */
 492         /* to make it look like a "normal" Unix priority/nice value  */
 493         priority = task_prio(task);
 494         nice = task_nice(task);
 495
 496         /* Temporary variable needed for gcc-2.96 */
 497         /* convert timespec -> nsec*/
 498         start_time =
 499                 (unsigned long long)task->real_start_time.tv_sec * NSEC_PER_SEC
 500                                 + task->real_start_time.tv_nsec;
 501         /* convert nsec -> ticks */
 502         start_time = nsec_to_clock_t(start_time);
 503
 504         seq_printf(m, "%d (%s) %c %d %d %d %d %d %u %lu \
 505 %lu %lu %lu %lu %lu %ld %ld %ld %ld %d 0 %llu %lu %ld %lu %lu %lu %lu %lu \
 506 %lu %lu %lu %lu %lu %lu %lu %lu %d %d %u %u %llu %lu %ld\n",
 507                 pid_nr_ns(pid, ns),
 508                 tcomm,
 509                 state,
 510                 ppid,
 511                 pgid,
 512                 sid,
 513                 tty_nr,
 514                 tty_pgrp,
 515                 task->flags,
 516                 min_flt,
 517                 cmin_flt,
 518                 maj_flt,
 519                 cmaj_flt,
 520                 cputime_to_clock_t(utime),
 521                 cputime_to_clock_t(stime),
 522                 cputime_to_clock_t(cutime),
 523                 cputime_to_clock_t(cstime),
 524                 priority,
 525                 nice,
 526                 num_threads,
 527                 start_time,
 528                 vsize,
 529                 mm ? get_mm_rss(mm) : 0,
 530                 rsslim,
 531                 mm ? mm->start_code : 0,
 532                 mm ? mm->end_code : 0,
 533                 mm ? mm->start_stack : 0,
 534                 esp,
 535                 eip,
 536                 /* The signal information here is obsolete.
 537                  * It must be decimal for Linux 2.0 compatibility.
 538                  * Use /proc/#/status for real-time signals.
 539                  */
 540                 task->pending.signal.sig[0] & 0x7fffffffUL,
 541                 task->blocked.sig[0] & 0x7fffffffUL,
 542                 sigign      .sig[0] & 0x7fffffffUL,
 543                 sigcatch    .sig[0] & 0x7fffffffUL,
 544                 wchan,
 545                 0UL,
 546                 0UL,
 547                 task->exit_signal,
 548                 task_cpu(task),
 549                 task->rt_priority,
 550                 task->policy,
 551                 (unsigned long long)delayacct_blkio_ticks(task),
 552                 cputime_to_clock_t(gtime),
 553                 cputime_to_clock_t(cgtime));
 554         if (mm)
 555                 mmput(mm);
 556         return 0;
 557 }
 558
 559 int proc_tid_stat(struct seq_file *m, struct pid_namespace *ns,
 560                         struct pid *pid, struct task_struct *task)
 561 {
 562         return do_task_stat(m, ns, pid, task, 0);
 563 }
 564
 565 int proc_tgid_stat(struct seq_file *m, struct pid_namespace *ns,
 566                         struct pid *pid, struct task_struct *task)
 567 {
 568         return do_task_stat(m, ns, pid, task, 1);
 569 }
 570
 571 int proc_pid_statm(struct seq_file *m, struct pid_namespace *ns,
 572                         struct pid *pid, struct task_struct *task)
 573 {
 574         int size = 0, resident = 0, shared = 0, text = 0, lib = 0, data = 0;
 575         struct mm_struct *mm = get_task_mm(task);
 576
 577         if (mm) {
 578                 size = task_statm(mm, &shared, &text, &data, &resident);
 579                 mmput(mm);
 580         }
 581         seq_printf(m, "%d %d %d %d %d %d %d\n",
 582                         size, resident, shared, text, lib, data, 0);
 583
 584         return 0;
 585 }