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 #include <linux/vs_context.h>
  80 #include <linux/vs_network.h>
  81
  82 #include <asm/uaccess.h>
  83 #include <asm/pgtable.h>
  84 #include <asm/io.h>
  85 #include <asm/processor.h>
  86 #include "internal.h"
  87
  88 /* Gcc optimizes away "strlen(x)" for constant x */
  89 #define ADDBUF(buffer, string) \
  90 do { memcpy(buffer, string, strlen(string)); \
  91      buffer += strlen(string); } while (0)
  92
  93 static inline char * task_name(struct task_struct *p, char * buf)
  94 {
  95         int i;
  96         char * name;
  97         char tcomm[sizeof(p->comm)];
  98
  99         get_task_comm(tcomm, p);
 100
 101         ADDBUF(buf, "Name:\t");
 102         name = tcomm;
 103         i = sizeof(tcomm);
 104         do {
 105                 unsigned char c = *name;
 106                 name++;
 107                 i--;
 108                 *buf = c;
 109                 if (!c)
 110                         break;
 111                 if (c == '\\') {
 112                         buf[1] = c;
 113                         buf += 2;
 114                         continue;
 115                 }
 116                 if (c == '\n') {
 117                         buf[0] = '\\';
 118                         buf[1] = 'n';
 119                         buf += 2;
 120                         continue;
 121                 }
 122                 buf++;
 123         } while (i);
 124         *buf = '\n';
 125         return buf+1;
 126 }
 127
 128 /*
 129  * The task state array is a strange "bitmap" of
 130  * reasons to sleep. Thus "running" is zero, and
 131  * you can test for combinations of others with
 132  * simple bit tests.
 133  */
 134 static const char *task_state_array[] = {
 135         "R (running)",          /*  0 */
 136         "S (sleeping)",         /*  1 */
 137         "D (disk sleep)",       /*  2 */
 138         "T (stopped)",          /*  4 */
 139         "T (tracing stop)",     /*  8 */
 140         "Z (zombie)",           /* 16 */
 141         "X (dead)",             /* 32 */
 142         "N (noninteractive)",   /* 64 */
 143         "H (on hold)"           /* 128 */
 144 };
 145
 146 static inline const char * get_task_state(struct task_struct *tsk)
 147 {
 148         unsigned int state = (tsk->state & (TASK_RUNNING |
 149                                             TASK_INTERRUPTIBLE |
 150                                             TASK_UNINTERRUPTIBLE |
 151                                             TASK_STOPPED |
 152                                            TASK_TRACED |
 153                                            TASK_ONHOLD)) |
 154                         (tsk->exit_state & (EXIT_ZOMBIE |
 155                                             EXIT_DEAD));
 156         const char **p = &task_state_array[0];
 157
 158         while (state) {
 159                 p++;
 160                 state >>= 1;
 161         }
 162         return *p;
 163 }
 164
 165 static inline char * task_state(struct task_struct *p, char *buffer)
 166 {
 167         struct task_struct *tracer;
 168         pid_t tracer_pid, pid, ptgid, tgid;
 169         struct group_info *group_info;
 170         int g;
 171         struct fdtable *fdt = NULL;
 172
 173         rcu_read_lock();
 174         tracer = tracehook_tracer_task(p);
 175         tracer_pid = tracer == NULL ? 0 : vx_map_pid(tracer->pid);
 176         rcu_read_unlock();
 177
 178         read_lock(&tasklist_lock);
 179         tgid = vx_map_tgid(p->tgid);
 180         pid = vx_map_pid(p->pid);
 181         ptgid = vx_map_pid(p->group_leader->parent->tgid);
 182         buffer += sprintf(buffer,
 183                 "State:\t%s\n"
 184                 "SleepAVG:\t%lu%%\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                 (p->sleep_avg/1024)*100/(1020000000/1024),
 193                 tgid, pid, (pid > 1) ? ptgid : 0,
 194                 tracer_pid,
 195                 p->uid, p->euid, p->suid, p->fsuid,
 196                 p->gid, p->egid, p->sgid, p->fsgid);
 197         read_unlock(&tasklist_lock);
 198         task_lock(p);
 199         rcu_read_lock();
 200         if (p->files)
 201                 fdt = files_fdtable(p->files);
 202         buffer += sprintf(buffer,
 203                 "FDSize:\t%d\n"
 204                 "Groups:\t",
 205                 fdt ? fdt->max_fds : 0);
 206         rcu_read_unlock();
 207
 208         group_info = p->group_info;
 209         get_group_info(group_info);
 210         task_unlock(p);
 211
 212         for (g = 0; g < min(group_info->ngroups,NGROUPS_SMALL); g++)
 213                 buffer += sprintf(buffer, "%d ", GROUP_AT(group_info,g));
 214         put_group_info(group_info);
 215
 216         buffer += sprintf(buffer, "\n");
 217         return buffer;
 218 }
 219
 220 static char * render_sigset_t(const char *header, sigset_t *set, char *buffer)
 221 {
 222         int i, len;
 223
 224         len = strlen(header);
 225         memcpy(buffer, header, len);
 226         buffer += len;
 227
 228         i = _NSIG;
 229         do {
 230                 int x = 0;
 231
 232                 i -= 4;
 233                 if (sigismember(set, i+1)) x |= 1;
 234                 if (sigismember(set, i+2)) x |= 2;
 235                 if (sigismember(set, i+3)) x |= 4;
 236                 if (sigismember(set, i+4)) x |= 8;
 237                 *buffer++ = (x < 10 ? '0' : 'a' - 10) + x;
 238         } while (i >= 4);
 239
 240         *buffer++ = '\n';
 241         *buffer = 0;
 242         return buffer;
 243 }
 244
 245 static void collect_sigign_sigcatch(struct task_struct *p, sigset_t *ign,
 246                                     sigset_t *catch)
 247 {
 248         struct k_sigaction *k;
 249         int i;
 250
 251         k = p->sighand->action;
 252         for (i = 1; i <= _NSIG; ++i, ++k) {
 253                 if (k->sa.sa_handler == SIG_IGN)
 254                         sigaddset(ign, i);
 255                 else if (k->sa.sa_handler != SIG_DFL)
 256                         sigaddset(catch, i);
 257         }
 258 }
 259
 260 static inline char * task_sig(struct task_struct *p, char *buffer)
 261 {
 262         sigset_t pending, shpending, blocked, ignored, caught;
 263         int num_threads = 0;
 264         unsigned long qsize = 0;
 265         unsigned long qlim = 0;
 266
 267         sigemptyset(&pending);
 268         sigemptyset(&shpending);
 269         sigemptyset(&blocked);
 270         sigemptyset(&ignored);
 271         sigemptyset(&caught);
 272
 273         /* Gather all the data with the appropriate locks held */
 274         read_lock(&tasklist_lock);
 275         if (p->sighand) {
 276                 spin_lock_irq(&p->sighand->siglock);
 277                 pending = p->pending.signal;
 278                 shpending = p->signal->shared_pending.signal;
 279                 blocked = p->blocked;
 280                 collect_sigign_sigcatch(p, &ignored, &caught);
 281                 num_threads = atomic_read(&p->signal->count);
 282                 qsize = atomic_read(&p->user->sigpending);
 283                 qlim = p->signal->rlim[RLIMIT_SIGPENDING].rlim_cur;
 284                 spin_unlock_irq(&p->sighand->siglock);
 285         }
 286         read_unlock(&tasklist_lock);
 287
 288         buffer += sprintf(buffer, "Threads:\t%d\n", num_threads);
 289         buffer += sprintf(buffer, "SigQ:\t%lu/%lu\n", qsize, qlim);
 290
 291         /* render them all */
 292         buffer = render_sigset_t("SigPnd:\t", &pending, buffer);
 293         buffer = render_sigset_t("ShdPnd:\t", &shpending, buffer);
 294         buffer = render_sigset_t("SigBlk:\t", &blocked, buffer);
 295         buffer = render_sigset_t("SigIgn:\t", &ignored, buffer);
 296         buffer = render_sigset_t("SigCgt:\t", &caught, buffer);
 297
 298         return buffer;
 299 }
 300
 301 static inline char *task_cap(struct task_struct *p, char *buffer)
 302 {
 303     return buffer + sprintf(buffer, "CapInh:\t%016x\n"
 304                             "CapPrm:\t%016x\n"
 305                             "CapEff:\t%016x\n",
 306                             cap_t(p->cap_inheritable),
 307                             cap_t(p->cap_permitted),
 308                             cap_t(p->cap_effective));
 309 }
 310
 311 int proc_pid_status(struct task_struct *task, char * buffer)
 312 {
 313         char * orig = buffer;
 314 #ifdef  CONFIG_VSERVER_LEGACY
 315         struct vx_info *vxi;
 316 #endif
 317 #ifdef  CONFIG_VSERVER_LEGACYNET
 318         struct nx_info *nxi;
 319 #endif
 320         struct mm_struct *mm = get_task_mm(task);
 321
 322         buffer = task_name(task, buffer);
 323         buffer = task_state(task, buffer);
 324
 325         if (mm) {
 326                 buffer = task_mem(mm, buffer);
 327                 mmput(mm);
 328         }
 329         buffer = task_sig(task, buffer);
 330         buffer = task_cap(task, buffer);
 331         buffer = cpuset_task_status_allowed(task, buffer);
 332
 333         if (task_vx_flags(task, VXF_INFO_HIDE, 0))
 334                 goto skip;
 335 #ifdef  CONFIG_VSERVER_LEGACY
 336         buffer += sprintf (buffer,"s_context: %d\n", vx_task_xid(task));
 337         vxi = task_get_vx_info(task);
 338         if (vxi) {
 339                 buffer += sprintf (buffer,"ctxflags: %08llx\n"
 340                         ,(unsigned long long)vxi->vx_flags);
 341                 buffer += sprintf (buffer,"initpid: %d\n"
 342                         ,vxi->vx_initpid);
 343         } else {
 344                 buffer += sprintf (buffer,"ctxflags: none\n");
 345                 buffer += sprintf (buffer,"initpid: none\n");
 346         }
 347         put_vx_info(vxi);
 348 #else
 349         buffer += sprintf (buffer,"VxID: %d\n", vx_task_xid(task));
 350 #endif
 351 #ifdef  CONFIG_VSERVER_LEGACYNET
 352         nxi = task_get_nx_info(task);
 353         if (nxi) {
 354                 int i;
 355
 356                 buffer += sprintf (buffer,"ipv4root:");
 357                 for (i=0; i<nxi->nbipv4; i++){
 358                         buffer += sprintf (buffer," %08x/%08x"
 359                                 ,nxi->ipv4[i]
 360                                 ,nxi->mask[i]);
 361                 }
 362                 *buffer++ = '\n';
 363                 buffer += sprintf (buffer,"ipv4root_bcast: %08x\n"
 364                         ,nxi->v4_bcast);
 365         } else {
 366                 buffer += sprintf (buffer,"ipv4root: 0\n");
 367                 buffer += sprintf (buffer,"ipv4root_bcast: 0\n");
 368         }
 369         put_nx_info(nxi);
 370 #endif
 371 skip:
 372 #if defined(CONFIG_S390)
 373         buffer = task_show_regs(task, buffer);
 374 #endif
 375         return buffer - orig;
 376 }
 377
 378 static int do_task_stat(struct task_struct *task, char * buffer, int whole)
 379 {
 380         unsigned long vsize, eip, esp, wchan = ~0UL;
 381         long priority, nice;
 382         int tty_pgrp = -1, tty_nr = 0;
 383         sigset_t sigign, sigcatch;
 384         char state;
 385         int res;
 386         pid_t pid, ppid, pgid = -1, sid = -1;
 387         int num_threads = 0;
 388         struct mm_struct *mm;
 389         unsigned long long start_time;
 390         unsigned long cmin_flt = 0, cmaj_flt = 0;
 391         unsigned long  min_flt = 0,  maj_flt = 0;
 392         cputime_t cutime, cstime, utime, stime;
 393         unsigned long rsslim = 0;
 394         struct task_struct *t;
 395         char tcomm[sizeof(task->comm)];
 396
 397         state = *get_task_state(task);
 398         vsize = eip = esp = 0;
 399         mm = get_task_mm(task);
 400         if (mm) {
 401                 vsize = task_vsize(mm);
 402                 eip = KSTK_EIP(task);
 403                 esp = KSTK_ESP(task);
 404         }
 405
 406         get_task_comm(tcomm, task);
 407
 408         sigemptyset(&sigign);
 409         sigemptyset(&sigcatch);
 410         cutime = cstime = utime = stime = cputime_zero;
 411
 412         mutex_lock(&tty_mutex);
 413         read_lock(&tasklist_lock);
 414         if (task->sighand) {
 415                 spin_lock_irq(&task->sighand->siglock);
 416                 num_threads = atomic_read(&task->signal->count);
 417                 collect_sigign_sigcatch(task, &sigign, &sigcatch);
 418
 419                 /* add up live thread stats at the group level */
 420                 if (whole) {
 421                         t = task;
 422                         do {
 423                                 min_flt += t->min_flt;
 424                                 maj_flt += t->maj_flt;
 425                                 utime = cputime_add(utime, t->utime);
 426                                 stime = cputime_add(stime, t->stime);
 427                                 t = next_thread(t);
 428                         } while (t != task);
 429                 }
 430
 431                 spin_unlock_irq(&task->sighand->siglock);
 432         }
 433         if (task->signal) {
 434                 if (task->signal->tty) {
 435                         tty_pgrp = task->signal->tty->pgrp;
 436                         tty_nr = new_encode_dev(tty_devnum(task->signal->tty));
 437                 }
 438                 pgid = process_group(task);
 439                 sid = task->signal->session;
 440                 cmin_flt = task->signal->cmin_flt;
 441                 cmaj_flt = task->signal->cmaj_flt;
 442                 cutime = task->signal->cutime;
 443                 cstime = task->signal->cstime;
 444                 rsslim = task->signal->rlim[RLIMIT_RSS].rlim_cur;
 445                 if (whole) {
 446                         min_flt += task->signal->min_flt;
 447                         maj_flt += task->signal->maj_flt;
 448                         utime = cputime_add(utime, task->signal->utime);
 449                         stime = cputime_add(stime, task->signal->stime);
 450                 }
 451         }
 452         pid = vx_info_map_pid(task->vx_info, pid_alive(task) ? task->pid : 0);
 453         ppid = (!(pid > 1)) ? 0 : vx_info_map_tgid(task->vx_info,
 454                 task->group_leader->parent->tgid);
 455         pgid = vx_info_map_pid(task->vx_info, pgid);
 456
 457         read_unlock(&tasklist_lock);
 458         mutex_unlock(&tty_mutex);
 459
 460         if (!whole || num_threads<2) {
 461                 wchan = 0;
 462                 if (current->uid == task->uid || current->euid == task->uid ||
 463                                 capable(CAP_SYS_NICE))
 464                         wchan = get_wchan(task);
 465         }
 466         if (!whole) {
 467                 min_flt = task->min_flt;
 468                 maj_flt = task->maj_flt;
 469                 utime = task->utime;
 470                 stime = task->stime;
 471         }
 472
 473         /* scale priority and nice values from timeslices to -20..20 */
 474         /* to make it look like a "normal" Unix priority/nice value  */
 475         priority = task_prio(task);
 476         nice = task_nice(task);
 477
 478         /* Temporary variable needed for gcc-2.96 */
 479         /* convert timespec -> nsec*/
 480         start_time = (unsigned long long)task->start_time.tv_sec * NSEC_PER_SEC
 481                                 + task->start_time.tv_nsec;
 482         /* convert nsec -> ticks */
 483         start_time = nsec_to_clock_t(start_time);
 484
 485         /* fixup start time for virt uptime */
 486         if (vx_flags(VXF_VIRT_UPTIME, 0)) {
 487                 unsigned long long bias =
 488                         current->vx_info->cvirt.bias_clock;
 489
 490                 if (start_time > bias)
 491                         start_time -= bias;
 492                 else
 493                         start_time = 0;
 494         }
 495
 496         res = sprintf(buffer,"%d (%s) %c %d %d %d %d %d %lu %lu \
 497 %lu %lu %lu %lu %lu %ld %ld %ld %ld %d 0 %llu %lu %ld %lu %lu %lu %lu %lu \
 498 %lu %lu %lu %lu %lu %lu %lu %lu %d %d %lu %lu %llu\n",
 499                 pid,
 500                 tcomm,
 501                 state,
 502                 ppid,
 503                 pgid,
 504                 sid,
 505                 tty_nr,
 506                 tty_pgrp,
 507                 task->flags,
 508                 min_flt,
 509                 cmin_flt,
 510                 maj_flt,
 511                 cmaj_flt,
 512                 cputime_to_clock_t(utime),
 513                 cputime_to_clock_t(stime),
 514                 cputime_to_clock_t(cutime),
 515                 cputime_to_clock_t(cstime),
 516                 priority,
 517                 nice,
 518                 num_threads,
 519                 start_time,
 520                 vsize,
 521                 mm ? get_mm_rss(mm) : 0,
 522                 rsslim,
 523                 mm ? mm->start_code : 0,
 524                 mm ? mm->end_code : 0,
 525                 mm ? mm->start_stack : 0,
 526                 esp,
 527                 eip,
 528                 /* The signal information here is obsolete.
 529                  * It must be decimal for Linux 2.0 compatibility.
 530                  * Use /proc/#/status for real-time signals.
 531                  */
 532                 task->pending.signal.sig[0] & 0x7fffffffUL,
 533                 task->blocked.sig[0] & 0x7fffffffUL,
 534                 sigign      .sig[0] & 0x7fffffffUL,
 535                 sigcatch    .sig[0] & 0x7fffffffUL,
 536                 wchan,
 537                 0UL,
 538                 0UL,
 539                 task->exit_signal,
 540                 task_cpu(task),
 541                 task->rt_priority,
 542                 task->policy,
 543                 (unsigned long long)delayacct_blkio_ticks(task));
 544         if(mm)
 545                 mmput(mm);
 546         return res;
 547 }
 548
 549 int proc_tid_stat(struct task_struct *task, char * buffer)
 550 {
 551         return do_task_stat(task, buffer, 0);
 552 }
 553
 554 int proc_tgid_stat(struct task_struct *task, char * buffer)
 555 {
 556         return do_task_stat(task, buffer, 1);
 557 }
 558
 559 int proc_pid_statm(struct task_struct *task, char *buffer)
 560 {
 561         int size = 0, resident = 0, shared = 0, text = 0, lib = 0, data = 0;
 562         struct mm_struct *mm = get_task_mm(task);
 563
 564         if (mm) {
 565                 size = task_statm(mm, &shared, &text, &data, &resident);
 566                 mmput(mm);
 567         }
 568
 569         return sprintf(buffer,"%d %d %d %d %d %d %d\n",
 570                        size, resident, shared, text, lib, data, 0);
 571 }