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/ptrace.h>
  84 #include <linux/tracehook.h>
  85
  86 #include <asm/pgtable.h>
  87 #include <asm/processor.h>
  88 #include "internal.h"
  89
  90 static inline void task_name(struct seq_file *m, struct task_struct *p)
  91 {
  92         int i;
  93         char *buf, *end;
  94         char *name;
  95         char tcomm[sizeof(p->comm)];
  96
  97         get_task_comm(tcomm, p);
  98
  99         seq_printf(m, "Name:\t");
 100         end = m->buf + m->size;
 101         buf = m->buf + m->count;
 102         name = tcomm;
 103         i = sizeof(tcomm);
 104         while (i && (buf < end)) {
 105                 unsigned char c = *name;
 106                 name++;
 107                 i--;
 108                 *buf = c;
 109                 if (!c)
 110                         break;
 111                 if (c == '\\') {
 112                         buf++;
 113                         if (buf < end)
 114                                 *buf++ = c;
 115                         continue;
 116                 }
 117                 if (c == '\n') {
 118                         *buf++ = '\\';
 119                         if (buf < end)
 120                                 *buf++ = 'n';
 121                         continue;
 122                 }
 123                 buf++;
 124         }
 125         m->count = buf - m->buf;
 126         seq_printf(m, "\n");
 127 }
 128
 129 /*
 130  * The task state array is a strange "bitmap" of
 131  * reasons to sleep. Thus "running" is zero, and
 132  * you can test for combinations of others with
 133  * simple bit tests.
 134  */
 135 static const char *task_state_array[] = {
 136         "R (running)",          /*  0 */
 137         "S (sleeping)",         /*  1 */
 138         "D (disk sleep)",       /*  2 */
 139         "T (stopped)",          /*  4 */
 140         "T (tracing stop)",     /*  8 */
 141         "Z (zombie)",           /* 16 */
 142         "X (dead)"              /* 32 */
 143 };
 144
 145 static inline const char *get_task_state(struct task_struct *tsk)
 146 {
 147         unsigned int state = (tsk->state & TASK_REPORT) | tsk->exit_state;
 148         const char **p = &task_state_array[0];
 149
 150         while (state) {
 151                 p++;
 152                 state >>= 1;
 153         }
 154         return *p;
 155 }
 156
 157 static inline void task_state(struct seq_file *m, struct pid_namespace *ns,
 158                                 struct pid *pid, struct task_struct *p)
 159 {
 160         struct group_info *group_info;
 161         int g;
 162         struct fdtable *fdt = NULL;
 163         const struct cred *cred;
 164         pid_t ppid, tpid;
 165
 166         rcu_read_lock();
 167         ppid = pid_alive(p) ?
 168                 task_tgid_nr_ns(rcu_dereference(p->real_parent), ns) : 0;
 169         tpid = 0;
 170         if (pid_alive(p)) {
 171                 struct task_struct *tracer = tracehook_tracer_task(p);
 172                 if (tracer)
 173                         tpid = task_pid_nr_ns(tracer, ns);
 174         }
 175         cred = get_cred((struct cred *) __task_cred(p));
 176         seq_printf(m,
 177                 "State:\t%s\n"
 178                 "Tgid:\t%d\n"
 179                 "Pid:\t%d\n"
 180                 "PPid:\t%d\n"
 181                 "TracerPid:\t%d\n"
 182                 "Uid:\t%d\t%d\t%d\t%d\n"
 183                 "Gid:\t%d\t%d\t%d\t%d\n",
 184                 get_task_state(p),
 185                 task_tgid_nr_ns(p, ns),
 186                 pid_nr_ns(pid, ns),
 187                 ppid, tpid,
 188                 cred->uid, cred->euid, cred->suid, cred->fsuid,
 189                 cred->gid, cred->egid, cred->sgid, cred->fsgid);
 190
 191         task_lock(p);
 192         if (p->files)
 193                 fdt = files_fdtable(p->files);
 194         seq_printf(m,
 195                 "FDSize:\t%d\n"
 196                 "Groups:\t",
 197                 fdt ? fdt->max_fds : 0);
 198         rcu_read_unlock();
 199
 200         group_info = cred->group_info;
 201         task_unlock(p);
 202
 203         for (g = 0; g < min(group_info->ngroups, NGROUPS_SMALL); g++)
 204                 seq_printf(m, "%d ", GROUP_AT(group_info, g));
 205         put_cred(cred);
 206
 207         seq_printf(m, "\n");
 208 }
 209
 210 static void render_sigset_t(struct seq_file *m, const char *header,
 211                                 sigset_t *set)
 212 {
 213         int i;
 214
 215         seq_printf(m, "%s", header);
 216
 217         i = _NSIG;
 218         do {
 219                 int x = 0;
 220
 221                 i -= 4;
 222                 if (sigismember(set, i+1)) x |= 1;
 223                 if (sigismember(set, i+2)) x |= 2;
 224                 if (sigismember(set, i+3)) x |= 4;
 225                 if (sigismember(set, i+4)) x |= 8;
 226                 seq_printf(m, "%x", x);
 227         } while (i >= 4);
 228
 229         seq_printf(m, "\n");
 230 }
 231
 232 static void collect_sigign_sigcatch(struct task_struct *p, sigset_t *ign,
 233                                     sigset_t *catch)
 234 {
 235         struct k_sigaction *k;
 236         int i;
 237
 238         k = p->sighand->action;
 239         for (i = 1; i <= _NSIG; ++i, ++k) {
 240                 if (k->sa.sa_handler == SIG_IGN)
 241                         sigaddset(ign, i);
 242                 else if (k->sa.sa_handler != SIG_DFL)
 243                         sigaddset(catch, i);
 244         }
 245 }
 246
 247 static inline void task_sig(struct seq_file *m, struct task_struct *p)
 248 {
 249         unsigned long flags;
 250         sigset_t pending, shpending, blocked, ignored, caught;
 251         int num_threads = 0;
 252         unsigned long qsize = 0;
 253         unsigned long qlim = 0;
 254
 255         sigemptyset(&pending);
 256         sigemptyset(&shpending);
 257         sigemptyset(&blocked);
 258         sigemptyset(&ignored);
 259         sigemptyset(&caught);
 260
 261         if (lock_task_sighand(p, &flags)) {
 262                 pending = p->pending.signal;
 263                 shpending = p->signal->shared_pending.signal;
 264                 blocked = p->blocked;
 265                 collect_sigign_sigcatch(p, &ignored, &caught);
 266                 num_threads = atomic_read(&p->signal->count);
 267                 qsize = atomic_read(&__task_cred(p)->user->sigpending);
 268                 qlim = p->signal->rlim[RLIMIT_SIGPENDING].rlim_cur;
 269                 unlock_task_sighand(p, &flags);
 270         }
 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         const struct cred *cred;
 299         kernel_cap_t cap_inheritable, cap_permitted, cap_effective, cap_bset;
 300
 301         rcu_read_lock();
 302         cred = __task_cred(p);
 303         cap_inheritable = cred->cap_inheritable;
 304         cap_permitted   = cred->cap_permitted;
 305         cap_effective   = cred->cap_effective;
 306         cap_bset        = cred->cap_bset;
 307         rcu_read_unlock();
 308
 309         render_cap_t(m, "CapInh:\t", &cap_inheritable);
 310         render_cap_t(m, "CapPrm:\t", &cap_permitted);
 311         render_cap_t(m, "CapEff:\t", &cap_effective);
 312         render_cap_t(m, "CapBnd:\t", &cap_bset);
 313 }
 314
 315 static inline void task_context_switch_counts(struct seq_file *m,
 316                                                 struct task_struct *p)
 317 {
 318         seq_printf(m,   "voluntary_ctxt_switches:\t%lu\n"
 319                         "nonvoluntary_ctxt_switches:\t%lu\n",
 320                         p->nvcsw,
 321                         p->nivcsw);
 322 }
 323
 324 int proc_pid_status(struct seq_file *m, struct pid_namespace *ns,
 325                         struct pid *pid, struct task_struct *task)
 326 {
 327         struct mm_struct *mm = get_task_mm(task);
 328
 329         task_name(m, task);
 330         task_state(m, ns, pid, task);
 331
 332         if (mm) {
 333                 task_mem(m, mm);
 334                 mmput(mm);
 335         }
 336         task_sig(m, task);
 337         task_cap(m, task);
 338         cpuset_task_status_allowed(m, task);
 339 #if defined(CONFIG_S390)
 340         task_show_regs(m, task);
 341 #endif
 342         task_context_switch_counts(m, task);
 343         return 0;
 344 }
 345
 346 static int do_task_stat(struct seq_file *m, struct pid_namespace *ns,
 347                         struct pid *pid, struct task_struct *task, int whole)
 348 {
 349         unsigned long vsize, eip, esp, wchan = ~0UL;
 350         long priority, nice;
 351         int tty_pgrp = -1, tty_nr = 0;
 352         sigset_t sigign, sigcatch;
 353         char state;
 354         pid_t ppid = 0, pgid = -1, sid = -1;
 355         int num_threads = 0;
 356         int permitted;
 357         struct mm_struct *mm;
 358         unsigned long long start_time;
 359         unsigned long cmin_flt = 0, cmaj_flt = 0;
 360         unsigned long  min_flt = 0,  maj_flt = 0;
 361         cputime_t cutime, cstime, utime, stime;
 362         cputime_t cgtime, gtime;
 363         unsigned long rsslim = 0;
 364         char tcomm[sizeof(task->comm)];
 365         unsigned long flags;
 366
 367         state = *get_task_state(task);
 368         vsize = eip = esp = 0;
 369         permitted = ptrace_may_access(task, PTRACE_MODE_READ);
 370         mm = get_task_mm(task);
 371         if (mm) {
 372                 vsize = task_vsize(mm);
 373                 if (permitted) {
 374                         eip = KSTK_EIP(task);
 375                         esp = KSTK_ESP(task);
 376                 }
 377         }
 378
 379         get_task_comm(tcomm, task);
 380
 381         sigemptyset(&sigign);
 382         sigemptyset(&sigcatch);
 383         cutime = cstime = utime = stime = cputime_zero;
 384         cgtime = gtime = cputime_zero;
 385
 386         if (lock_task_sighand(task, &flags)) {
 387                 struct signal_struct *sig = task->signal;
 388
 389                 if (sig->tty) {
 390                         struct pid *pgrp = tty_get_pgrp(sig->tty);
 391                         tty_pgrp = pid_nr_ns(pgrp, ns);
 392                         put_pid(pgrp);
 393                         tty_nr = new_encode_dev(tty_devnum(sig->tty));
 394                 }
 395
 396                 num_threads = atomic_read(&sig->count);
 397                 collect_sigign_sigcatch(task, &sigign, &sigcatch);
 398
 399                 cmin_flt = sig->cmin_flt;
 400                 cmaj_flt = sig->cmaj_flt;
 401                 cutime = sig->cutime;
 402                 cstime = sig->cstime;
 403                 cgtime = sig->cgtime;
 404                 rsslim = sig->rlim[RLIMIT_RSS].rlim_cur;
 405
 406                 /* add up live thread stats at the group level */
 407                 if (whole) {
 408                         struct task_cputime cputime;
 409                         struct task_struct *t = task;
 410                         do {
 411                                 min_flt += t->min_flt;
 412                                 maj_flt += t->maj_flt;
 413                                 gtime = cputime_add(gtime, task_gtime(t));
 414                                 t = next_thread(t);
 415                         } while (t != task);
 416
 417                         min_flt += sig->min_flt;
 418                         maj_flt += sig->maj_flt;
 419                         thread_group_cputime(task, &cputime);
 420                         utime = cputime.utime;
 421                         stime = cputime.stime;
 422                         gtime = cputime_add(gtime, sig->gtime);
 423                 }
 424
 425                 sid = task_session_nr_ns(task, ns);
 426                 ppid = task_tgid_nr_ns(task->real_parent, ns);
 427                 pgid = task_pgrp_nr_ns(task, ns);
 428
 429                 unlock_task_sighand(task, &flags);
 430         }
 431
 432         if (permitted && (!whole || num_threads < 2))
 433                 wchan = get_wchan(task);
 434         if (!whole) {
 435                 min_flt = task->min_flt;
 436                 maj_flt = task->maj_flt;
 437                 utime = task_utime(task);
 438                 stime = task_stime(task);
 439                 gtime = task_gtime(task);
 440         }
 441
 442         /* scale priority and nice values from timeslices to -20..20 */
 443         /* to make it look like a "normal" Unix priority/nice value  */
 444         priority = task_prio(task);
 445         nice = task_nice(task);
 446
 447         /* Temporary variable needed for gcc-2.96 */
 448         /* convert timespec -> nsec*/
 449         start_time =
 450                 (unsigned long long)task->real_start_time.tv_sec * NSEC_PER_SEC
 451                                 + task->real_start_time.tv_nsec;
 452         /* convert nsec -> ticks */
 453         start_time = nsec_to_clock_t(start_time);
 454
 455         seq_printf(m, "%d (%s) %c %d %d %d %d %d %u %lu \
 456 %lu %lu %lu %lu %lu %ld %ld %ld %ld %d 0 %llu %lu %ld %lu %lu %lu %lu %lu \
 457 %lu %lu %lu %lu %lu %lu %lu %lu %d %d %u %u %llu %lu %ld\n",
 458                 pid_nr_ns(pid, ns),
 459                 tcomm,
 460                 state,
 461                 ppid,
 462                 pgid,
 463                 sid,
 464                 tty_nr,
 465                 tty_pgrp,
 466                 task->flags,
 467                 min_flt,
 468                 cmin_flt,
 469                 maj_flt,
 470                 cmaj_flt,
 471                 cputime_to_clock_t(utime),
 472                 cputime_to_clock_t(stime),
 473                 cputime_to_clock_t(cutime),
 474                 cputime_to_clock_t(cstime),
 475                 priority,
 476                 nice,
 477                 num_threads,
 478                 start_time,
 479                 vsize,
 480                 mm ? get_mm_rss(mm) : 0,
 481                 rsslim,
 482                 mm ? mm->start_code : 0,
 483                 mm ? mm->end_code : 0,
 484                 (permitted && mm) ? mm->start_stack : 0,
 485                 esp,
 486                 eip,
 487                 /* The signal information here is obsolete.
 488                  * It must be decimal for Linux 2.0 compatibility.
 489                  * Use /proc/#/status for real-time signals.
 490                  */
 491                 task->pending.signal.sig[0] & 0x7fffffffUL,
 492                 task->blocked.sig[0] & 0x7fffffffUL,
 493                 sigign      .sig[0] & 0x7fffffffUL,
 494                 sigcatch    .sig[0] & 0x7fffffffUL,
 495                 wchan,
 496                 0UL,
 497                 0UL,
 498                 task->exit_signal,
 499                 task_cpu(task),
 500                 task->rt_priority,
 501                 task->policy,
 502                 (unsigned long long)delayacct_blkio_ticks(task),
 503                 cputime_to_clock_t(gtime),
 504                 cputime_to_clock_t(cgtime));
 505         if (mm)
 506                 mmput(mm);
 507         return 0;
 508 }
 509
 510 int proc_tid_stat(struct seq_file *m, struct pid_namespace *ns,
 511                         struct pid *pid, struct task_struct *task)
 512 {
 513         return do_task_stat(m, ns, pid, task, 0);
 514 }
 515
 516 int proc_tgid_stat(struct seq_file *m, struct pid_namespace *ns,
 517                         struct pid *pid, struct task_struct *task)
 518 {
 519         return do_task_stat(m, ns, pid, task, 1);
 520 }
 521
 522 int proc_pid_statm(struct seq_file *m, struct pid_namespace *ns,
 523                         struct pid *pid, struct task_struct *task)
 524 {
 525         int size = 0, resident = 0, shared = 0, text = 0, lib = 0, data = 0;
 526         struct mm_struct *mm = get_task_mm(task);
 527
 528         if (mm) {
 529                 size = task_statm(mm, &shared, &text, &data, &resident);
 530                 mmput(mm);
 531         }
 532         seq_printf(m, "%d %d %d %d %d %d %d\n",
 533                         size, resident, shared, text, lib, data, 0);
 534
 535         return 0;
 536 }