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/mm.h>
  66 #include <linux/hugetlb.h>
  67 #include <linux/pagemap.h>
  68 #include <linux/swap.h>
  69 #include <linux/slab.h>
  70 #include <linux/smp.h>
  71 #include <linux/signal.h>
  72 #include <linux/highmem.h>
  73 #include <linux/file.h>
  74 #include <linux/times.h>
  75 #include <linux/cpuset.h>
  76 #include <linux/tracehook.h>
  77 #include <linux/rcupdate.h>
  78 #include <linux/delayacct.h>
  79
  80 #include <asm/uaccess.h>
  81 #include <asm/pgtable.h>
  82 #include <asm/io.h>
  83 #include <asm/processor.h>
  84 #include "internal.h"
  85
  86 /* Gcc optimizes away "strlen(x)" for constant x */
  87 #define ADDBUF(buffer, string) \
  88 do { memcpy(buffer, string, strlen(string)); \
  89      buffer += strlen(string); } while (0)
  90
  91 static inline char * task_name(struct task_struct *p, char * buf)
  92 {
  93         int i;
  94         char * name;
  95         char tcomm[sizeof(p->comm)];
  96
  97         get_task_comm(tcomm, p);
  98
  99         ADDBUF(buf, "Name:\t");
 100         name = tcomm;
 101         i = sizeof(tcomm);
 102         do {
 103                 unsigned char c = *name;
 104                 name++;
 105                 i--;
 106                 *buf = c;
 107                 if (!c)
 108                         break;
 109                 if (c == '\\') {
 110                         buf[1] = c;
 111                         buf += 2;
 112                         continue;
 113                 }
 114                 if (c == '\n') {
 115                         buf[0] = '\\';
 116                         buf[1] = 'n';
 117                         buf += 2;
 118                         continue;
 119                 }
 120                 buf++;
 121         } while (i);
 122         *buf = '\n';
 123         return buf+1;
 124 }
 125
 126 /*
 127  * The task state array is a strange "bitmap" of
 128  * reasons to sleep. Thus "running" is zero, and
 129  * you can test for combinations of others with
 130  * simple bit tests.
 131  */
 132 static const char *task_state_array[] = {
 133         "R (running)",          /*  0 */
 134         "S (sleeping)",         /*  1 */
 135         "D (disk sleep)",       /*  2 */
 136         "T (stopped)",          /*  4 */
 137         "T (tracing stop)",     /*  8 */
 138         "Z (zombie)",           /* 16 */
 139         "X (dead)",             /* 32 */
 140         "N (noninteractive)",   /* 64 */
 141         "H (on hold)"           /* 128 */
 142 };
 143
 144 static inline const char * get_task_state(struct task_struct *tsk)
 145 {
 146         unsigned int state = (tsk->state & (TASK_RUNNING |
 147                                             TASK_INTERRUPTIBLE |
 148                                             TASK_UNINTERRUPTIBLE |
 149                                             TASK_STOPPED |
 150                                            TASK_TRACED |
 151                                            TASK_ONHOLD)) |
 152                         (tsk->exit_state & (EXIT_ZOMBIE |
 153                                             EXIT_DEAD));
 154         const char **p = &task_state_array[0];
 155
 156         while (state) {
 157                 p++;
 158                 state >>= 1;
 159         }
 160         return *p;
 161 }
 162
 163 static inline char * task_state(struct task_struct *p, char *buffer)
 164 {
 165         struct task_struct *tracer;
 166         pid_t tracer_pid, pid, ptgid, tgid;
 167         struct group_info *group_info;
 168         int g;
 169         struct fdtable *fdt = NULL;
 170
 171         rcu_read_lock();
 172         tracer = tracehook_tracer_task(p);
 173         tracer_pid = tracer == NULL ? 0 : vx_map_pid(tracer->pid);
 174         rcu_read_unlock();
 175
 176         read_lock(&tasklist_lock);
 177         tgid = vx_map_tgid(p->tgid);
 178         pid = vx_map_pid(p->pid);
 179         ptgid = vx_map_pid(p->group_leader->parent->tgid);
 180         buffer += sprintf(buffer,
 181                 "State:\t%s\n"
 182                 "SleepAVG:\t%lu%%\n"
 183                 "Tgid:\t%d\n"
 184                 "Pid:\t%d\n"
 185                 "PPid:\t%d\n"
 186                 "TracerPid:\t%d\n"
 187                 "Uid:\t%d\t%d\t%d\t%d\n"
 188                 "Gid:\t%d\t%d\t%d\t%d\n",
 189                 get_task_state(p),
 190                 (p->sleep_avg/1024)*100/(1020000000/1024),
 191                 tgid, pid, (pid > 1) ? ptgid : 0,
 192                 tracer_pid,
 193                 p->uid, p->euid, p->suid, p->fsuid,
 194                 p->gid, p->egid, p->sgid, p->fsgid);
 195         read_unlock(&tasklist_lock);
 196         task_lock(p);
 197         rcu_read_lock();
 198         if (p->files)
 199                 fdt = files_fdtable(p->files);
 200         buffer += sprintf(buffer,
 201                 "FDSize:\t%d\n"
 202                 "Groups:\t",
 203                 fdt ? fdt->max_fds : 0);
 204         rcu_read_unlock();
 205
 206         group_info = p->group_info;
 207         get_group_info(group_info);
 208         task_unlock(p);
 209
 210         for (g = 0; g < min(group_info->ngroups,NGROUPS_SMALL); g++)
 211                 buffer += sprintf(buffer, "%d ", GROUP_AT(group_info,g));
 212         put_group_info(group_info);
 213
 214         buffer += sprintf(buffer, "\n");
 215         return buffer;
 216 }
 217
 218 static char * render_sigset_t(const char *header, sigset_t *set, char *buffer)
 219 {
 220         int i, len;
 221
 222         len = strlen(header);
 223         memcpy(buffer, header, len);
 224         buffer += len;
 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                 *buffer++ = (x < 10 ? '0' : 'a' - 10) + x;
 236         } while (i >= 4);
 237
 238         *buffer++ = '\n';
 239         *buffer = 0;
 240         return buffer;
 241 }
 242
 243 static void collect_sigign_sigcatch(struct task_struct *p, sigset_t *ign,
 244                                     sigset_t *catch)
 245 {
 246         struct k_sigaction *k;
 247         int i;
 248
 249         k = p->sighand->action;
 250         for (i = 1; i <= _NSIG; ++i, ++k) {
 251                 if (k->sa.sa_handler == SIG_IGN)
 252                         sigaddset(ign, i);
 253                 else if (k->sa.sa_handler != SIG_DFL)
 254                         sigaddset(catch, i);
 255         }
 256 }
 257
 258 static inline char * task_sig(struct task_struct *p, char *buffer)
 259 {
 260         sigset_t pending, shpending, blocked, ignored, caught;
 261         int num_threads = 0;
 262         unsigned long qsize = 0;
 263         unsigned long qlim = 0;
 264
 265         sigemptyset(&pending);
 266         sigemptyset(&shpending);
 267         sigemptyset(&blocked);
 268         sigemptyset(&ignored);
 269         sigemptyset(&caught);
 270
 271         /* Gather all the data with the appropriate locks held */
 272         read_lock(&tasklist_lock);
 273         if (p->sighand) {
 274                 spin_lock_irq(&p->sighand->siglock);
 275                 pending = p->pending.signal;
 276                 shpending = p->signal->shared_pending.signal;
 277                 blocked = p->blocked;
 278                 collect_sigign_sigcatch(p, &ignored, &caught);
 279                 num_threads = atomic_read(&p->signal->count);
 280                 qsize = atomic_read(&p->user->sigpending);
 281                 qlim = p->signal->rlim[RLIMIT_SIGPENDING].rlim_cur;
 282                 spin_unlock_irq(&p->sighand->siglock);
 283         }
 284         read_unlock(&tasklist_lock);
 285
 286         buffer += sprintf(buffer, "Threads:\t%d\n", num_threads);
 287         buffer += sprintf(buffer, "SigQ:\t%lu/%lu\n", qsize, qlim);
 288
 289         /* render them all */
 290         buffer = render_sigset_t("SigPnd:\t", &pending, buffer);
 291         buffer = render_sigset_t("ShdPnd:\t", &shpending, buffer);
 292         buffer = render_sigset_t("SigBlk:\t", &blocked, buffer);
 293         buffer = render_sigset_t("SigIgn:\t", &ignored, buffer);
 294         buffer = render_sigset_t("SigCgt:\t", &caught, buffer);
 295
 296         return buffer;
 297 }
 298
 299 static inline char *task_cap(struct task_struct *p, char *buffer)
 300 {
 301     return buffer + sprintf(buffer, "CapInh:\t%016x\n"
 302                             "CapPrm:\t%016x\n"
 303                             "CapEff:\t%016x\n",
 304                             cap_t(p->cap_inheritable),
 305                             cap_t(p->cap_permitted),
 306                             cap_t(p->cap_effective));
 307 }
 308
 309 int proc_pid_status(struct task_struct *task, char * buffer)
 310 {
 311         char * orig = buffer;
 312 #ifdef  CONFIG_VSERVER_LEGACY
 313         struct vx_info *vxi;
 314 #endif
 315 #ifdef  CONFIG_VSERVER_LEGACYNET
 316         struct nx_info *nxi;
 317 #endif
 318         struct mm_struct *mm = get_task_mm(task);
 319
 320         buffer = task_name(task, buffer);
 321         buffer = task_state(task, buffer);
 322
 323         if (mm) {
 324                 buffer = task_mem(mm, buffer);
 325                 mmput(mm);
 326         }
 327         buffer = task_sig(task, buffer);
 328         buffer = task_cap(task, buffer);
 329         buffer = cpuset_task_status_allowed(task, buffer);
 330
 331         if (task_vx_flags(task, VXF_INFO_HIDE, 0))
 332                 goto skip;
 333 #ifdef  CONFIG_VSERVER_LEGACY
 334         buffer += sprintf (buffer,"s_context: %d\n", vx_task_xid(task));
 335         vxi = task_get_vx_info(task);
 336         if (vxi) {
 337                 buffer += sprintf (buffer,"ctxflags: %08llx\n"
 338                         ,(unsigned long long)vxi->vx_flags);
 339                 buffer += sprintf (buffer,"initpid: %d\n"
 340                         ,vxi->vx_initpid);
 341         } else {
 342                 buffer += sprintf (buffer,"ctxflags: none\n");
 343                 buffer += sprintf (buffer,"initpid: none\n");
 344         }
 345         put_vx_info(vxi);
 346 #else
 347         buffer += sprintf (buffer,"VxID: %d\n", vx_task_xid(task));
 348 #endif
 349 #ifdef  CONFIG_VSERVER_LEGACYNET
 350         nxi = task_get_nx_info(task);
 351         if (nxi) {
 352                 int i;
 353
 354                 buffer += sprintf (buffer,"ipv4root:");
 355                 for (i=0; i<nxi->nbipv4; i++){
 356                         buffer += sprintf (buffer," %08x/%08x"
 357                                 ,nxi->ipv4[i]
 358                                 ,nxi->mask[i]);
 359                 }
 360                 *buffer++ = '\n';
 361                 buffer += sprintf (buffer,"ipv4root_bcast: %08x\n"
 362                         ,nxi->v4_bcast);
 363         } else {
 364                 buffer += sprintf (buffer,"ipv4root: 0\n");
 365                 buffer += sprintf (buffer,"ipv4root_bcast: 0\n");
 366         }
 367         put_nx_info(nxi);
 368 #endif
 369 skip:
 370 #if defined(CONFIG_S390)
 371         buffer = task_show_regs(task, buffer);
 372 #endif
 373         return buffer - orig;
 374 }
 375
 376 static int do_task_stat(struct task_struct *task, char * buffer, int whole)
 377 {
 378         unsigned long vsize, eip, esp, wchan = ~0UL;
 379         long priority, nice;
 380         int tty_pgrp = -1, tty_nr = 0;
 381         sigset_t sigign, sigcatch;
 382         char state;
 383         int res;
 384         pid_t pid, ppid, pgid = -1, sid = -1;
 385         int num_threads = 0;
 386         struct mm_struct *mm;
 387         unsigned long long start_time;
 388         unsigned long cmin_flt = 0, cmaj_flt = 0;
 389         unsigned long  min_flt = 0,  maj_flt = 0;
 390         cputime_t cutime, cstime, utime, stime;
 391         unsigned long rsslim = 0;
 392         struct task_struct *t;
 393         char tcomm[sizeof(task->comm)];
 394
 395         state = *get_task_state(task);
 396         vsize = eip = esp = 0;
 397         mm = get_task_mm(task);
 398         if (mm) {
 399                 vsize = task_vsize(mm);
 400                 eip = KSTK_EIP(task);
 401                 esp = KSTK_ESP(task);
 402         }
 403
 404         get_task_comm(tcomm, task);
 405
 406         sigemptyset(&sigign);
 407         sigemptyset(&sigcatch);
 408         cutime = cstime = utime = stime = cputime_zero;
 409
 410         mutex_lock(&tty_mutex);
 411         read_lock(&tasklist_lock);
 412         if (task->sighand) {
 413                 spin_lock_irq(&task->sighand->siglock);
 414                 num_threads = atomic_read(&task->signal->count);
 415                 collect_sigign_sigcatch(task, &sigign, &sigcatch);
 416
 417                 /* add up live thread stats at the group level */
 418                 if (whole) {
 419                         t = task;
 420                         do {
 421                                 min_flt += t->min_flt;
 422                                 maj_flt += t->maj_flt;
 423                                 utime = cputime_add(utime, t->utime);
 424                                 stime = cputime_add(stime, t->stime);
 425                                 t = next_thread(t);
 426                         } while (t != task);
 427                 }
 428
 429                 spin_unlock_irq(&task->sighand->siglock);
 430         }
 431         if (task->signal) {
 432                 if (task->signal->tty) {
 433                         tty_pgrp = task->signal->tty->pgrp;
 434                         tty_nr = new_encode_dev(tty_devnum(task->signal->tty));
 435                 }
 436                 pgid = process_group(task);
 437                 sid = task->signal->session;
 438                 cmin_flt = task->signal->cmin_flt;
 439                 cmaj_flt = task->signal->cmaj_flt;
 440                 cutime = task->signal->cutime;
 441                 cstime = task->signal->cstime;
 442                 rsslim = task->signal->rlim[RLIMIT_RSS].rlim_cur;
 443                 if (whole) {
 444                         min_flt += task->signal->min_flt;
 445                         maj_flt += task->signal->maj_flt;
 446                         utime = cputime_add(utime, task->signal->utime);
 447                         stime = cputime_add(stime, task->signal->stime);
 448                 }
 449         }
 450         pid = vx_info_map_pid(task->vx_info, pid_alive(task) ? task->pid : 0);
 451         ppid = (!(pid > 1)) ? 0 : vx_info_map_tgid(task->vx_info,
 452                 task->group_leader->parent->tgid);
 453         pgid = vx_info_map_pid(task->vx_info, pgid);
 454
 455         read_unlock(&tasklist_lock);
 456         mutex_unlock(&tty_mutex);
 457
 458         if (!whole || num_threads<2) {
 459                 wchan = 0;
 460                 if (current->uid == task->uid || current->euid == task->uid ||
 461                                 capable(CAP_SYS_NICE))
 462                         wchan = get_wchan(task);
 463         }
 464         if (!whole) {
 465                 min_flt = task->min_flt;
 466                 maj_flt = task->maj_flt;
 467                 utime = task->utime;
 468                 stime = task->stime;
 469         }
 470
 471         /* scale priority and nice values from timeslices to -20..20 */
 472         /* to make it look like a "normal" Unix priority/nice value  */
 473         priority = task_prio(task);
 474         nice = task_nice(task);
 475
 476         /* Temporary variable needed for gcc-2.96 */
 477         /* convert timespec -> nsec*/
 478         start_time = (unsigned long long)task->start_time.tv_sec * NSEC_PER_SEC
 479                                 + task->start_time.tv_nsec;
 480         /* convert nsec -> ticks */
 481         start_time = nsec_to_clock_t(start_time);
 482
 483         /* fixup start time for virt uptime */
 484         if (vx_flags(VXF_VIRT_UPTIME, 0)) {
 485                 unsigned long long bias =
 486                         current->vx_info->cvirt.bias_clock;
 487
 488                 if (start_time > bias)
 489                         start_time -= bias;
 490                 else
 491                         start_time = 0;
 492         }
 493
 494         res = sprintf(buffer,"%d (%s) %c %d %d %d %d %d %lu %lu \
 495 %lu %lu %lu %lu %lu %ld %ld %ld %ld %d 0 %llu %lu %ld %lu %lu %lu %lu %lu \
 496 %lu %lu %lu %lu %lu %lu %lu %lu %d %d %lu %lu %llu\n",
 497                 pid,
 498                 tcomm,
 499                 state,
 500                 ppid,
 501                 pgid,
 502                 sid,
 503                 tty_nr,
 504                 tty_pgrp,
 505                 task->flags,
 506                 min_flt,
 507                 cmin_flt,
 508                 maj_flt,
 509                 cmaj_flt,
 510                 cputime_to_clock_t(utime),
 511                 cputime_to_clock_t(stime),
 512                 cputime_to_clock_t(cutime),
 513                 cputime_to_clock_t(cstime),
 514                 priority,
 515                 nice,
 516                 num_threads,
 517                 start_time,
 518                 vsize,
 519                 mm ? get_mm_rss(mm) : 0,
 520                 rsslim,
 521                 mm ? mm->start_code : 0,
 522                 mm ? mm->end_code : 0,
 523                 mm ? mm->start_stack : 0,
 524                 esp,
 525                 eip,
 526                 /* The signal information here is obsolete.
 527                  * It must be decimal for Linux 2.0 compatibility.
 528                  * Use /proc/#/status for real-time signals.
 529                  */
 530                 task->pending.signal.sig[0] & 0x7fffffffUL,
 531                 task->blocked.sig[0] & 0x7fffffffUL,
 532                 sigign      .sig[0] & 0x7fffffffUL,
 533                 sigcatch    .sig[0] & 0x7fffffffUL,
 534                 wchan,
 535                 0UL,
 536                 0UL,
 537                 task->exit_signal,
 538                 task_cpu(task),
 539                 task->rt_priority,
 540                 task->policy,
 541                 (unsigned long long)delayacct_blkio_ticks(task));
 542         if(mm)
 543                 mmput(mm);
 544         return res;
 545 }
 546
 547 int proc_tid_stat(struct task_struct *task, char * buffer)
 548 {
 549         return do_task_stat(task, buffer, 0);
 550 }
 551
 552 int proc_tgid_stat(struct task_struct *task, char * buffer)
 553 {
 554         return do_task_stat(task, buffer, 1);
 555 }
 556
 557 int proc_pid_statm(struct task_struct *task, char *buffer)
 558 {
 559         int size = 0, resident = 0, shared = 0, text = 0, lib = 0, data = 0;
 560         struct mm_struct *mm = get_task_mm(task);
 561
 562         if (mm) {
 563                 size = task_statm(mm, &shared, &text, &data, &resident);
 564                 mmput(mm);
 565         }
 566
 567         return sprintf(buffer,"%d %d %d %d %d %d %d\n",
 568                        size, resident, shared, text, lib, data, 0);
 569 }