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 #include <linux/utrace.h>
  86 #include <linux/swapops.h>
  87
  88 #include <asm/pgtable.h>
  89 #include <asm/processor.h>
  90 #include "internal.h"
  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         "x (dead)",             /*  64 */
 146         "K (wakekill)",         /* 128 */
 147         "W (waking)",           /* 256 */
 148 };
 149
 150 static inline const char *get_task_state(struct task_struct *tsk)
 151 {
 152         unsigned int state = (tsk->state & TASK_REPORT) | tsk->exit_state;
 153         const char **p = &task_state_array[0];
 154
 155         BUILD_BUG_ON(1 + ilog2(TASK_STATE_MAX) != ARRAY_SIZE(task_state_array));
 156
 157         while (state) {
 158                 p++;
 159                 state >>= 1;
 160         }
 161         return *p;
 162 }
 163
 164 static inline void task_state(struct seq_file *m, struct pid_namespace *ns,
 165                                 struct pid *pid, struct task_struct *p)
 166 {
 167         struct group_info *group_info;
 168         int g;
 169         struct fdtable *fdt = NULL;
 170         const struct cred *cred;
 171         pid_t ppid, tpid;
 172
 173         rcu_read_lock();
 174         ppid = pid_alive(p) ?
 175                 task_tgid_nr_ns(rcu_dereference(p->real_parent), ns) : 0;
 176         tpid = 0;
 177         if (pid_alive(p)) {
 178                 struct task_struct *tracer = tracehook_tracer_task(p);
 179                 if (tracer)
 180                         tpid = task_pid_nr_ns(tracer, ns);
 181         }
 182         cred = get_cred((struct cred *) __task_cred(p));
 183         seq_printf(m,
 184                 "State:\t%s\n"
 185                 "Tgid:\t%d\n"
 186                 "Pid:\t%d\n"
 187                 "PPid:\t%d\n"
 188                 "TracerPid:\t%d\n"
 189                 "Uid:\t%d\t%d\t%d\t%d\n"
 190                 "Gid:\t%d\t%d\t%d\t%d\n",
 191                 get_task_state(p),
 192                 task_tgid_nr_ns(p, ns),
 193                 pid_nr_ns(pid, ns),
 194                 ppid, tpid,
 195                 cred->uid, cred->euid, cred->suid, cred->fsuid,
 196                 cred->gid, cred->egid, cred->sgid, cred->fsgid);
 197
 198         task_utrace_proc_status(m, p);
 199
 200         task_lock(p);
 201         if (p->files)
 202                 fdt = files_fdtable(p->files);
 203         seq_printf(m,
 204                 "FDSize:\t%d\n"
 205                 "Groups:\t",
 206                 fdt ? fdt->max_fds : 0);
 207         rcu_read_unlock();
 208
 209         group_info = cred->group_info;
 210         task_unlock(p);
 211
 212         for (g = 0; g < min(group_info->ngroups, NGROUPS_SMALL); g++)
 213                 seq_printf(m, "%d ", GROUP_AT(group_info, g));
 214         put_cred(cred);
 215
 216         seq_printf(m, "\n");
 217 }
 218
 219 static void render_sigset_t(struct seq_file *m, const char *header,
 220                                 sigset_t *set)
 221 {
 222         int i;
 223
 224         seq_printf(m, "%s", header);
 225
 226         i = _NSIG;
 227         do {
 228                 int x = 0;
 229
 230                 i -= 4;
 231                 if (sigismember(set, i+1)) x |= 1;
 232                 if (sigismember(set, i+2)) x |= 2;
 233                 if (sigismember(set, i+3)) x |= 4;
 234                 if (sigismember(set, i+4)) x |= 8;
 235                 seq_printf(m, "%x", x);
 236         } while (i >= 4);
 237
 238         seq_printf(m, "\n");
 239 }
 240
 241 static void collect_sigign_sigcatch(struct task_struct *p, sigset_t *ign,
 242                                     sigset_t *catch)
 243 {
 244         struct k_sigaction *k;
 245         int i;
 246
 247         k = p->sighand->action;
 248         for (i = 1; i <= _NSIG; ++i, ++k) {
 249                 if (k->sa.sa_handler == SIG_IGN)
 250                         sigaddset(ign, i);
 251                 else if (k->sa.sa_handler != SIG_DFL)
 252                         sigaddset(catch, i);
 253         }
 254 }
 255
 256 static inline void task_sig(struct seq_file *m, struct task_struct *p)
 257 {
 258         unsigned long flags;
 259         sigset_t pending, shpending, blocked, ignored, caught;
 260         int num_threads = 0;
 261         unsigned long qsize = 0;
 262         unsigned long qlim = 0;
 263
 264         sigemptyset(&pending);
 265         sigemptyset(&shpending);
 266         sigemptyset(&blocked);
 267         sigemptyset(&ignored);
 268         sigemptyset(&caught);
 269
 270         if (lock_task_sighand(p, &flags)) {
 271                 pending = p->pending.signal;
 272                 shpending = p->signal->shared_pending.signal;
 273                 blocked = p->blocked;
 274                 collect_sigign_sigcatch(p, &ignored, &caught);
 275                 num_threads = atomic_read(&p->signal->count);
 276                 qsize = atomic_read(&__task_cred(p)->user->sigpending);
 277                 qlim = task_rlimit(p, RLIMIT_SIGPENDING);
 278                 unlock_task_sighand(p, &flags);
 279         }
 280
 281         seq_printf(m, "Threads:\t%d\n", num_threads);
 282         seq_printf(m, "SigQ:\t%lu/%lu\n", qsize, qlim);
 283
 284         /* render them all */
 285         render_sigset_t(m, "SigPnd:\t", &pending);
 286         render_sigset_t(m, "ShdPnd:\t", &shpending);
 287         render_sigset_t(m, "SigBlk:\t", &blocked);
 288         render_sigset_t(m, "SigIgn:\t", &ignored);
 289         render_sigset_t(m, "SigCgt:\t", &caught);
 290 }
 291
 292 static void render_cap_t(struct seq_file *m, const char *header,
 293                         kernel_cap_t *a)
 294 {
 295         unsigned __capi;
 296
 297         seq_printf(m, "%s", header);
 298         CAP_FOR_EACH_U32(__capi) {
 299                 seq_printf(m, "%08x",
 300                            a->cap[(_KERNEL_CAPABILITY_U32S-1) - __capi]);
 301         }
 302         seq_printf(m, "\n");
 303 }
 304
 305 static inline void task_cap(struct seq_file *m, struct task_struct *p)
 306 {
 307         const struct cred *cred;
 308         kernel_cap_t cap_inheritable, cap_permitted, cap_effective, cap_bset;
 309
 310         rcu_read_lock();
 311         cred = __task_cred(p);
 312         cap_inheritable = cred->cap_inheritable;
 313         cap_permitted   = cred->cap_permitted;
 314         cap_effective   = cred->cap_effective;
 315         cap_bset        = cred->cap_bset;
 316         rcu_read_unlock();
 317
 318         render_cap_t(m, "CapInh:\t", &cap_inheritable);
 319         render_cap_t(m, "CapPrm:\t", &cap_permitted);
 320         render_cap_t(m, "CapEff:\t", &cap_effective);
 321         render_cap_t(m, "CapBnd:\t", &cap_bset);
 322 }
 323
 324 static inline void task_context_switch_counts(struct seq_file *m,
 325                                                 struct task_struct *p)
 326 {
 327         seq_printf(m,   "voluntary_ctxt_switches:\t%lu\n"
 328                         "nonvoluntary_ctxt_switches:\t%lu\n",
 329                         p->nvcsw,
 330                         p->nivcsw);
 331 }
 332
 333 static void task_cpus_allowed(struct seq_file *m, struct task_struct *task)
 334 {
 335         seq_printf(m, "Cpus_allowed:\t");
 336         seq_cpumask(m, &task->cpus_allowed);
 337         seq_printf(m, "\n");
 338         seq_printf(m, "Cpus_allowed_list:\t");
 339         seq_cpumask_list(m, &task->cpus_allowed);
 340         seq_printf(m, "\n");
 341 }
 342
 343 int proc_pid_status(struct seq_file *m, struct pid_namespace *ns,
 344                         struct pid *pid, struct task_struct *task)
 345 {
 346         struct mm_struct *mm = get_task_mm(task);
 347
 348         task_name(m, task);
 349         task_state(m, ns, pid, task);
 350
 351         if (mm) {
 352                 task_mem(m, mm);
 353                 mmput(mm);
 354         }
 355         task_sig(m, task);
 356         task_cap(m, task);
 357         task_cpus_allowed(m, task);
 358         cpuset_task_status_allowed(m, task);
 359 #if defined(CONFIG_S390)
 360         task_show_regs(m, task);
 361 #endif
 362         task_context_switch_counts(m, task);
 363         return 0;
 364 }
 365
 366 static int do_task_stat(struct seq_file *m, struct pid_namespace *ns,
 367                         struct pid *pid, struct task_struct *task, int whole)
 368 {
 369         unsigned long vsize, eip, esp, wchan = ~0UL;
 370         long priority, nice;
 371         int tty_pgrp = -1, tty_nr = 0;
 372         sigset_t sigign, sigcatch;
 373         char state;
 374         pid_t ppid = 0, pgid = -1, sid = -1;
 375         int num_threads = 0;
 376         int permitted;
 377         struct mm_struct *mm;
 378         unsigned long long start_time;
 379         unsigned long cmin_flt = 0, cmaj_flt = 0;
 380         unsigned long  min_flt = 0,  maj_flt = 0;
 381         cputime_t cutime, cstime, utime, stime;
 382         cputime_t cgtime, gtime;
 383         unsigned long rsslim = 0;
 384         char tcomm[sizeof(task->comm)];
 385         unsigned long flags;
 386
 387         state = *get_task_state(task);
 388         vsize = eip = esp = 0;
 389         permitted = ptrace_may_access(task, PTRACE_MODE_READ);
 390         mm = get_task_mm(task);
 391         if (mm) {
 392                 vsize = task_vsize(mm);
 393                 if (permitted) {
 394                         eip = KSTK_EIP(task);
 395                         esp = KSTK_ESP(task);
 396                 }
 397         }
 398
 399         get_task_comm(tcomm, task);
 400
 401         sigemptyset(&sigign);
 402         sigemptyset(&sigcatch);
 403         cutime = cstime = utime = stime = cputime_zero;
 404         cgtime = gtime = cputime_zero;
 405
 406         if (lock_task_sighand(task, &flags)) {
 407                 struct signal_struct *sig = task->signal;
 408
 409                 if (sig->tty) {
 410                         struct pid *pgrp = tty_get_pgrp(sig->tty);
 411                         tty_pgrp = pid_nr_ns(pgrp, ns);
 412                         put_pid(pgrp);
 413                         tty_nr = new_encode_dev(tty_devnum(sig->tty));
 414                 }
 415
 416                 num_threads = atomic_read(&sig->count);
 417                 collect_sigign_sigcatch(task, &sigign, &sigcatch);
 418
 419                 cmin_flt = sig->cmin_flt;
 420                 cmaj_flt = sig->cmaj_flt;
 421                 cutime = sig->cutime;
 422                 cstime = sig->cstime;
 423                 cgtime = sig->cgtime;
 424                 rsslim = ACCESS_ONCE(sig->rlim[RLIMIT_RSS].rlim_cur);
 425
 426                 /* add up live thread stats at the group level */
 427                 if (whole) {
 428                         struct task_struct *t = task;
 429                         do {
 430                                 min_flt += t->min_flt;
 431                                 maj_flt += t->maj_flt;
 432                                 gtime = cputime_add(gtime, t->gtime);
 433                                 t = next_thread(t);
 434                         } while (t != task);
 435
 436                         min_flt += sig->min_flt;
 437                         maj_flt += sig->maj_flt;
 438                         thread_group_times(task, &utime, &stime);
 439                         gtime = cputime_add(gtime, sig->gtime);
 440                 }
 441
 442                 sid = task_session_nr_ns(task, ns);
 443                 ppid = task_tgid_nr_ns(task->real_parent, ns);
 444                 pgid = task_pgrp_nr_ns(task, ns);
 445
 446                 unlock_task_sighand(task, &flags);
 447         }
 448
 449         if (permitted && (!whole || num_threads < 2))
 450                 wchan = get_wchan(task);
 451         if (!whole) {
 452                 min_flt = task->min_flt;
 453                 maj_flt = task->maj_flt;
 454                 task_times(task, &utime, &stime);
 455                 gtime = task->gtime;
 456         }
 457
 458         /* scale priority and nice values from timeslices to -20..20 */
 459         /* to make it look like a "normal" Unix priority/nice value  */
 460         priority = task_prio(task);
 461         nice = task_nice(task);
 462
 463         /* Temporary variable needed for gcc-2.96 */
 464         /* convert timespec -> nsec*/
 465         start_time =
 466                 (unsigned long long)task->real_start_time.tv_sec * NSEC_PER_SEC
 467                                 + task->real_start_time.tv_nsec;
 468         /* convert nsec -> ticks */
 469         start_time = nsec_to_clock_t(start_time);
 470
 471         seq_printf(m, "%d (%s) %c %d %d %d %d %d %u %lu \
 472 %lu %lu %lu %lu %lu %ld %ld %ld %ld %d 0 %llu %lu %ld %lu %lu %lu %lu %lu \
 473 %lu %lu %lu %lu %lu %lu %lu %lu %d %d %u %u %llu %lu %ld\n",
 474                 pid_nr_ns(pid, ns),
 475                 tcomm,
 476                 state,
 477                 ppid,
 478                 pgid,
 479                 sid,
 480                 tty_nr,
 481                 tty_pgrp,
 482                 task->flags,
 483                 min_flt,
 484                 cmin_flt,
 485                 maj_flt,
 486                 cmaj_flt,
 487                 cputime_to_clock_t(utime),
 488                 cputime_to_clock_t(stime),
 489                 cputime_to_clock_t(cutime),
 490                 cputime_to_clock_t(cstime),
 491                 priority,
 492                 nice,
 493                 num_threads,
 494                 start_time,
 495                 vsize,
 496                 mm ? get_mm_rss(mm) : 0,
 497                 rsslim,
 498                 mm ? mm->start_code : 0,
 499                 mm ? mm->end_code : 0,
 500                 (permitted && mm) ? task->stack_start : 0,
 501                 esp,
 502                 eip,
 503                 /* The signal information here is obsolete.
 504                  * It must be decimal for Linux 2.0 compatibility.
 505                  * Use /proc/#/status for real-time signals.
 506                  */
 507                 task->pending.signal.sig[0] & 0x7fffffffUL,
 508                 task->blocked.sig[0] & 0x7fffffffUL,
 509                 sigign      .sig[0] & 0x7fffffffUL,
 510                 sigcatch    .sig[0] & 0x7fffffffUL,
 511                 wchan,
 512                 0UL,
 513                 0UL,
 514                 task->exit_signal,
 515                 task_cpu(task),
 516                 task->rt_priority,
 517                 task->policy,
 518                 (unsigned long long)delayacct_blkio_ticks(task),
 519                 cputime_to_clock_t(gtime),
 520                 cputime_to_clock_t(cgtime));
 521         if (mm)
 522                 mmput(mm);
 523         return 0;
 524 }
 525
 526 int proc_tid_stat(struct seq_file *m, struct pid_namespace *ns,
 527                         struct pid *pid, struct task_struct *task)
 528 {
 529         return do_task_stat(m, ns, pid, task, 0);
 530 }
 531
 532 int proc_tgid_stat(struct seq_file *m, struct pid_namespace *ns,
 533                         struct pid *pid, struct task_struct *task)
 534 {
 535         return do_task_stat(m, ns, pid, task, 1);
 536 }
 537
 538 int proc_pid_statm(struct seq_file *m, struct pid_namespace *ns,
 539                         struct pid *pid, struct task_struct *task)
 540 {
 541         int size = 0, resident = 0, shared = 0, text = 0, lib = 0, data = 0;
 542         struct mm_struct *mm = get_task_mm(task);
 543
 544         if (mm) {
 545                 size = task_statm(mm, &shared, &text, &data, &resident);
 546                 mmput(mm);
 547         }
 548         seq_printf(m, "%d %d %d %d %d %d %d\n",
 549                         size, resident, shared, text, lib, data, 0);
 550
 551         return 0;
 552 }