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/rcupdate.h>
  77 #include <linux/delayacct.h>
  78
  79 #include <asm/uaccess.h>
  80 #include <asm/pgtable.h>
  81 #include <asm/io.h>
  82 #include <asm/processor.h>
  83 #include "internal.h"
  84
  85 /* Gcc optimizes away "strlen(x)" for constant x */
  86 #define ADDBUF(buffer, string) \
  87 do { memcpy(buffer, string, strlen(string)); \
  88      buffer += strlen(string); } while (0)
  89
  90 static inline char * task_name(struct task_struct *p, char * buf)
  91 {
  92         int i;
  93         char * name;
  94         char tcomm[sizeof(p->comm)];
  95
  96         get_task_comm(tcomm, p);
  97
  98         ADDBUF(buf, "Name:\t");
  99         name = tcomm;
 100         i = sizeof(tcomm);
 101         do {
 102                 unsigned char c = *name;
 103                 name++;
 104                 i--;
 105                 *buf = c;
 106                 if (!c)
 107                         break;
 108                 if (c == '\\') {
 109                         buf[1] = c;
 110                         buf += 2;
 111                         continue;
 112                 }
 113                 if (c == '\n') {
 114                         buf[0] = '\\';
 115                         buf[1] = 'n';
 116                         buf += 2;
 117                         continue;
 118                 }
 119                 buf++;
 120         } while (i);
 121         *buf = '\n';
 122         return buf+1;
 123 }
 124
 125 /*
 126  * The task state array is a strange "bitmap" of
 127  * reasons to sleep. Thus "running" is zero, and
 128  * you can test for combinations of others with
 129  * simple bit tests.
 130  */
 131 static const char *task_state_array[] = {
 132         "R (running)",          /*  0 */
 133         "S (sleeping)",         /*  1 */
 134         "D (disk sleep)",       /*  2 */
 135         "T (stopped)",          /*  4 */
 136         "T (tracing stop)",     /*  8 */
 137         "Z (zombie)",           /* 16 */
 138         "X (dead)"              /* 32 */
 139 };
 140
 141 static inline const char * get_task_state(struct task_struct *tsk)
 142 {
 143         unsigned int state = (tsk->state & (TASK_RUNNING |
 144                                             TASK_INTERRUPTIBLE |
 145                                             TASK_UNINTERRUPTIBLE |
 146                                             TASK_STOPPED |
 147                                             TASK_TRACED)) |
 148                         (tsk->exit_state & (EXIT_ZOMBIE |
 149                                             EXIT_DEAD));
 150         const char **p = &task_state_array[0];
 151
 152         while (state) {
 153                 p++;
 154                 state >>= 1;
 155         }
 156         return *p;
 157 }
 158
 159 static inline char * task_state(struct task_struct *p, char *buffer)
 160 {
 161         struct group_info *group_info;
 162         int g;
 163         struct fdtable *fdt = NULL;
 164
 165         rcu_read_lock();
 166         buffer += sprintf(buffer,
 167                 "State:\t%s\n"
 168                 "SleepAVG:\t%lu%%\n"
 169                 "Tgid:\t%d\n"
 170                 "Pid:\t%d\n"
 171                 "PPid:\t%d\n"
 172                 "TracerPid:\t%d\n"
 173                 "Uid:\t%d\t%d\t%d\t%d\n"
 174                 "Gid:\t%d\t%d\t%d\t%d\n",
 175                 get_task_state(p),
 176                 (p->sleep_avg/1024)*100/(1020000000/1024),
 177                 p->tgid, p->pid,
 178                 pid_alive(p) ? rcu_dereference(p->real_parent)->tgid : 0,
 179                 pid_alive(p) && p->ptrace ? rcu_dereference(p->parent)->pid : 0,
 180                 p->uid, p->euid, p->suid, p->fsuid,
 181                 p->gid, p->egid, p->sgid, p->fsgid);
 182
 183         task_lock(p);
 184         if (p->files)
 185                 fdt = files_fdtable(p->files);
 186         buffer += sprintf(buffer,
 187                 "FDSize:\t%d\n"
 188                 "Groups:\t",
 189                 fdt ? fdt->max_fds : 0);
 190         rcu_read_unlock();
 191
 192         group_info = p->group_info;
 193         get_group_info(group_info);
 194         task_unlock(p);
 195
 196         for (g = 0; g < min(group_info->ngroups,NGROUPS_SMALL); g++)
 197                 buffer += sprintf(buffer, "%d ", GROUP_AT(group_info,g));
 198         put_group_info(group_info);
 199
 200         buffer += sprintf(buffer, "\n");
 201         return buffer;
 202 }
 203
 204 static char * render_sigset_t(const char *header, sigset_t *set, char *buffer)
 205 {
 206         int i, len;
 207
 208         len = strlen(header);
 209         memcpy(buffer, header, len);
 210         buffer += len;
 211
 212         i = _NSIG;
 213         do {
 214                 int x = 0;
 215
 216                 i -= 4;
 217                 if (sigismember(set, i+1)) x |= 1;
 218                 if (sigismember(set, i+2)) x |= 2;
 219                 if (sigismember(set, i+3)) x |= 4;
 220                 if (sigismember(set, i+4)) x |= 8;
 221                 *buffer++ = (x < 10 ? '0' : 'a' - 10) + x;
 222         } while (i >= 4);
 223
 224         *buffer++ = '\n';
 225         *buffer = 0;
 226         return buffer;
 227 }
 228
 229 static void collect_sigign_sigcatch(struct task_struct *p, sigset_t *ign,
 230                                     sigset_t *catch)
 231 {
 232         struct k_sigaction *k;
 233         int i;
 234
 235         k = p->sighand->action;
 236         for (i = 1; i <= _NSIG; ++i, ++k) {
 237                 if (k->sa.sa_handler == SIG_IGN)
 238                         sigaddset(ign, i);
 239                 else if (k->sa.sa_handler != SIG_DFL)
 240                         sigaddset(catch, i);
 241         }
 242 }
 243
 244 static inline char * task_sig(struct task_struct *p, char *buffer)
 245 {
 246         unsigned long flags;
 247         sigset_t pending, shpending, blocked, ignored, caught;
 248         int num_threads = 0;
 249         unsigned long qsize = 0;
 250         unsigned long qlim = 0;
 251
 252         sigemptyset(&pending);
 253         sigemptyset(&shpending);
 254         sigemptyset(&blocked);
 255         sigemptyset(&ignored);
 256         sigemptyset(&caught);
 257
 258         rcu_read_lock();
 259         if (lock_task_sighand(p, &flags)) {
 260                 pending = p->pending.signal;
 261                 shpending = p->signal->shared_pending.signal;
 262                 blocked = p->blocked;
 263                 collect_sigign_sigcatch(p, &ignored, &caught);
 264                 num_threads = atomic_read(&p->signal->count);
 265                 qsize = atomic_read(&p->user->sigpending);
 266                 qlim = p->signal->rlim[RLIMIT_SIGPENDING].rlim_cur;
 267                 unlock_task_sighand(p, &flags);
 268         }
 269         rcu_read_unlock();
 270
 271         buffer += sprintf(buffer, "Threads:\t%d\n", num_threads);
 272         buffer += sprintf(buffer, "SigQ:\t%lu/%lu\n", qsize, qlim);
 273
 274         /* render them all */
 275         buffer = render_sigset_t("SigPnd:\t", &pending, buffer);
 276         buffer = render_sigset_t("ShdPnd:\t", &shpending, buffer);
 277         buffer = render_sigset_t("SigBlk:\t", &blocked, buffer);
 278         buffer = render_sigset_t("SigIgn:\t", &ignored, buffer);
 279         buffer = render_sigset_t("SigCgt:\t", &caught, buffer);
 280
 281         return buffer;
 282 }
 283
 284 static inline char *task_cap(struct task_struct *p, char *buffer)
 285 {
 286     return buffer + sprintf(buffer, "CapInh:\t%016x\n"
 287                             "CapPrm:\t%016x\n"
 288                             "CapEff:\t%016x\n",
 289                             cap_t(p->cap_inheritable),
 290                             cap_t(p->cap_permitted),
 291                             cap_t(p->cap_effective));
 292 }
 293
 294 int proc_pid_status(struct task_struct *task, char * buffer)
 295 {
 296         char * orig = buffer;
 297         struct mm_struct *mm = get_task_mm(task);
 298
 299         buffer = task_name(task, buffer);
 300         buffer = task_state(task, buffer);
 301
 302         if (mm) {
 303                 buffer = task_mem(mm, buffer);
 304                 mmput(mm);
 305         }
 306         buffer = task_sig(task, buffer);
 307         buffer = task_cap(task, buffer);
 308         buffer = cpuset_task_status_allowed(task, buffer);
 309 #if defined(CONFIG_S390)
 310         buffer = task_show_regs(task, buffer);
 311 #endif
 312         return buffer - orig;
 313 }
 314
 315 static int do_task_stat(struct task_struct *task, char * buffer, int whole)
 316 {
 317         unsigned long vsize, eip, esp, wchan = ~0UL;
 318         long priority, nice;
 319         int tty_pgrp = -1, tty_nr = 0;
 320         sigset_t sigign, sigcatch;
 321         char state;
 322         int res;
 323         pid_t ppid = 0, pgid = -1, sid = -1;
 324         int num_threads = 0;
 325         struct mm_struct *mm;
 326         unsigned long long start_time;
 327         unsigned long cmin_flt = 0, cmaj_flt = 0;
 328         unsigned long  min_flt = 0,  maj_flt = 0;
 329         cputime_t cutime, cstime, utime, stime;
 330         unsigned long rsslim = 0;
 331         char tcomm[sizeof(task->comm)];
 332         unsigned long flags;
 333
 334         state = *get_task_state(task);
 335         vsize = eip = esp = 0;
 336         mm = get_task_mm(task);
 337         if (mm) {
 338                 vsize = task_vsize(mm);
 339                 eip = KSTK_EIP(task);
 340                 esp = KSTK_ESP(task);
 341         }
 342
 343         get_task_comm(tcomm, task);
 344
 345         sigemptyset(&sigign);
 346         sigemptyset(&sigcatch);
 347         cutime = cstime = utime = stime = cputime_zero;
 348
 349         rcu_read_lock();
 350         if (lock_task_sighand(task, &flags)) {
 351                 struct signal_struct *sig = task->signal;
 352
 353                 if (sig->tty) {
 354                         tty_pgrp = pid_nr(sig->tty->pgrp);
 355                         tty_nr = new_encode_dev(tty_devnum(sig->tty));
 356                 }
 357
 358                 num_threads = atomic_read(&sig->count);
 359                 collect_sigign_sigcatch(task, &sigign, &sigcatch);
 360
 361                 cmin_flt = sig->cmin_flt;
 362                 cmaj_flt = sig->cmaj_flt;
 363                 cutime = sig->cutime;
 364                 cstime = sig->cstime;
 365                 rsslim = sig->rlim[RLIMIT_RSS].rlim_cur;
 366
 367                 /* add up live thread stats at the group level */
 368                 if (whole) {
 369                         struct task_struct *t = task;
 370                         do {
 371                                 min_flt += t->min_flt;
 372                                 maj_flt += t->maj_flt;
 373                                 utime = cputime_add(utime, t->utime);
 374                                 stime = cputime_add(stime, t->stime);
 375                                 t = next_thread(t);
 376                         } while (t != task);
 377
 378                         min_flt += sig->min_flt;
 379                         maj_flt += sig->maj_flt;
 380                         utime = cputime_add(utime, sig->utime);
 381                         stime = cputime_add(stime, sig->stime);
 382                 }
 383
 384                 sid = signal_session(sig);
 385                 pgid = process_group(task);
 386                 ppid = rcu_dereference(task->real_parent)->tgid;
 387
 388                 unlock_task_sighand(task, &flags);
 389         }
 390         rcu_read_unlock();
 391
 392         if (!whole || num_threads<2)
 393                 wchan = get_wchan(task);
 394         if (!whole) {
 395                 min_flt = task->min_flt;
 396                 maj_flt = task->maj_flt;
 397                 utime = task->utime;
 398                 stime = task->stime;
 399         }
 400
 401         /* scale priority and nice values from timeslices to -20..20 */
 402         /* to make it look like a "normal" Unix priority/nice value  */
 403         priority = task_prio(task);
 404         nice = task_nice(task);
 405
 406         /* Temporary variable needed for gcc-2.96 */
 407         /* convert timespec -> nsec*/
 408         start_time = (unsigned long long)task->start_time.tv_sec * NSEC_PER_SEC
 409                                 + task->start_time.tv_nsec;
 410         /* convert nsec -> ticks */
 411         start_time = nsec_to_clock_t(start_time);
 412
 413         res = sprintf(buffer,"%d (%s) %c %d %d %d %d %d %lu %lu \
 414 %lu %lu %lu %lu %lu %ld %ld %ld %ld %d 0 %llu %lu %ld %lu %lu %lu %lu %lu \
 415 %lu %lu %lu %lu %lu %lu %lu %lu %d %d %lu %lu %llu\n",
 416                 task->pid,
 417                 tcomm,
 418                 state,
 419                 ppid,
 420                 pgid,
 421                 sid,
 422                 tty_nr,
 423                 tty_pgrp,
 424                 task->flags,
 425                 min_flt,
 426                 cmin_flt,
 427                 maj_flt,
 428                 cmaj_flt,
 429                 cputime_to_clock_t(utime),
 430                 cputime_to_clock_t(stime),
 431                 cputime_to_clock_t(cutime),
 432                 cputime_to_clock_t(cstime),
 433                 priority,
 434                 nice,
 435                 num_threads,
 436                 start_time,
 437                 vsize,
 438                 mm ? get_mm_rss(mm) : 0,
 439                 rsslim,
 440                 mm ? mm->start_code : 0,
 441                 mm ? mm->end_code : 0,
 442                 mm ? mm->start_stack : 0,
 443                 esp,
 444                 eip,
 445                 /* The signal information here is obsolete.
 446                  * It must be decimal for Linux 2.0 compatibility.
 447                  * Use /proc/#/status for real-time signals.
 448                  */
 449                 task->pending.signal.sig[0] & 0x7fffffffUL,
 450                 task->blocked.sig[0] & 0x7fffffffUL,
 451                 sigign      .sig[0] & 0x7fffffffUL,
 452                 sigcatch    .sig[0] & 0x7fffffffUL,
 453                 wchan,
 454                 0UL,
 455                 0UL,
 456                 task->exit_signal,
 457                 task_cpu(task),
 458                 task->rt_priority,
 459                 task->policy,
 460                 (unsigned long long)delayacct_blkio_ticks(task));
 461         if(mm)
 462                 mmput(mm);
 463         return res;
 464 }
 465
 466 int proc_tid_stat(struct task_struct *task, char * buffer)
 467 {
 468         return do_task_stat(task, buffer, 0);
 469 }
 470
 471 int proc_tgid_stat(struct task_struct *task, char * buffer)
 472 {
 473         return do_task_stat(task, buffer, 1);
 474 }
 475
 476 int proc_pid_statm(struct task_struct *task, char *buffer)
 477 {
 478         int size = 0, resident = 0, shared = 0, text = 0, lib = 0, data = 0;
 479         struct mm_struct *mm = get_task_mm(task);
 480
 481         if (mm) {
 482                 size = task_statm(mm, &shared, &text, &data, &resident);
 483                 mmput(mm);
 484         }
 485
 486         return sprintf(buffer,"%d %d %d %d %d %d %d\n",
 487                        size, resident, shared, text, lib, data, 0);
 488 }