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/config.h>
  56 #include <linux/types.h>
  57 #include <linux/errno.h>
  58 #include <linux/time.h>
  59 #include <linux/kernel.h>
  60 #include <linux/kernel_stat.h>
  61 #include <linux/tty.h>
  62 #include <linux/string.h>
  63 #include <linux/mman.h>
  64 #include <linux/proc_fs.h>
  65 #include <linux/ioport.h>
  66 #include <linux/mm.h>
  67 #include <linux/hugetlb.h>
  68 #include <linux/pagemap.h>
  69 #include <linux/swap.h>
  70 #include <linux/slab.h>
  71 #include <linux/smp.h>
  72 #include <linux/signal.h>
  73 #include <linux/highmem.h>
  74 #include <linux/file.h>
  75 #include <linux/times.h>
  76 #include <linux/cpuset.h>
  77 #include <linux/vs_context.h>
  78 #include <linux/vs_network.h>
  79 #include <linux/vs_cvirt.h>
  80
  81 #include <asm/uaccess.h>
  82 #include <asm/pgtable.h>
  83 #include <asm/io.h>
  84 #include <asm/processor.h>
  85 #include "internal.h"
  86
  87 /* Gcc optimizes away "strlen(x)" for constant x */
  88 #define ADDBUF(buffer, string) \
  89 do { memcpy(buffer, string, strlen(string)); \
  90      buffer += strlen(string); } while (0)
  91
  92 static inline char * task_name(struct task_struct *p, char * buf)
  93 {
  94         int i;
  95         char * name;
  96         char tcomm[sizeof(p->comm)];
  97
  98         get_task_comm(tcomm, p);
  99
 100         ADDBUF(buf, "Name:\t");
 101         name = tcomm;
 102         i = sizeof(tcomm);
 103         do {
 104                 unsigned char c = *name;
 105                 name++;
 106                 i--;
 107                 *buf = c;
 108                 if (!c)
 109                         break;
 110                 if (c == '\\') {
 111                         buf[1] = c;
 112                         buf += 2;
 113                         continue;
 114                 }
 115                 if (c == '\n') {
 116                         buf[0] = '\\';
 117                         buf[1] = 'n';
 118                         buf += 2;
 119                         continue;
 120                 }
 121                 buf++;
 122         } while (i);
 123         *buf = '\n';
 124         return buf+1;
 125 }
 126
 127 /*
 128  * The task state array is a strange "bitmap" of
 129  * reasons to sleep. Thus "running" is zero, and
 130  * you can test for combinations of others with
 131  * simple bit tests.
 132  */
 133 static const char *task_state_array[] = {
 134         "R (running)",          /*  0 */
 135         "S (sleeping)",         /*  1 */
 136         "D (disk sleep)",       /*  2 */
 137         "T (stopped)",          /*  4 */
 138         "T (tracing stop)",     /*  8 */
 139         "Z (zombie)",           /* 16 */
 140         "X (dead)",             /* 32 */
 141         "H (on hold)"           /* 64 */
 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 group_info *group_info;
 166         int g;
 167         pid_t pid, ptgid, tppid, tgid;
 168
 169         read_lock(&tasklist_lock);
 170         tgid = vx_map_tgid(p->tgid);
 171         pid = vx_map_pid(p->pid);
 172         ptgid = vx_map_pid(p->group_leader->real_parent->tgid);
 173         tppid = vx_map_pid(p->parent->pid);
 174         buffer += sprintf(buffer,
 175                 "State:\t%s\n"
 176                 "SleepAVG:\t%lu%%\n"
 177                 "Tgid:\t%d\n"
 178                 "Pid:\t%d\n"
 179                 "PPid:\t%d\n"
 180                 "TracerPid:\t%d\n"
 181                 "Uid:\t%d\t%d\t%d\t%d\n"
 182                 "Gid:\t%d\t%d\t%d\t%d\n",
 183                 get_task_state(p),
 184                 (p->sleep_avg/1024)*100/(1020000000/1024),
 185                 tgid, pid, (pid > 1) ? ptgid : 0,
 186                 pid_alive(p) && p->ptrace ? tppid : 0,
 187                 p->uid, p->euid, p->suid, p->fsuid,
 188                 p->gid, p->egid, p->sgid, p->fsgid);
 189         read_unlock(&tasklist_lock);
 190         task_lock(p);
 191         buffer += sprintf(buffer,
 192                 "FDSize:\t%d\n"
 193                 "Groups:\t",
 194                 p->files ? p->files->max_fds : 0);
 195
 196         group_info = p->group_info;
 197         get_group_info(group_info);
 198         task_unlock(p);
 199
 200         for (g = 0; g < min(group_info->ngroups,NGROUPS_SMALL); g++)
 201                 buffer += sprintf(buffer, "%d ", GROUP_AT(group_info,g));
 202         put_group_info(group_info);
 203
 204         buffer += sprintf(buffer, "\n");
 205         return buffer;
 206 }
 207
 208 static char * render_sigset_t(const char *header, sigset_t *set, char *buffer)
 209 {
 210         int i, len;
 211
 212         len = strlen(header);
 213         memcpy(buffer, header, len);
 214         buffer += len;
 215
 216         i = _NSIG;
 217         do {
 218                 int x = 0;
 219
 220                 i -= 4;
 221                 if (sigismember(set, i+1)) x |= 1;
 222                 if (sigismember(set, i+2)) x |= 2;
 223                 if (sigismember(set, i+3)) x |= 4;
 224                 if (sigismember(set, i+4)) x |= 8;
 225                 *buffer++ = (x < 10 ? '0' : 'a' - 10) + x;
 226         } while (i >= 4);
 227
 228         *buffer++ = '\n';
 229         *buffer = 0;
 230         return buffer;
 231 }
 232
 233 static void collect_sigign_sigcatch(struct task_struct *p, sigset_t *ign,
 234                                     sigset_t *catch)
 235 {
 236         struct k_sigaction *k;
 237         int i;
 238
 239         k = p->sighand->action;
 240         for (i = 1; i <= _NSIG; ++i, ++k) {
 241                 if (k->sa.sa_handler == SIG_IGN)
 242                         sigaddset(ign, i);
 243                 else if (k->sa.sa_handler != SIG_DFL)
 244                         sigaddset(catch, i);
 245         }
 246 }
 247
 248 static inline char * task_sig(struct task_struct *p, char *buffer)
 249 {
 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         /* Gather all the data with the appropriate locks held */
 262         read_lock(&tasklist_lock);
 263         if (p->sighand) {
 264                 spin_lock_irq(&p->sighand->siglock);
 265                 pending = p->pending.signal;
 266                 shpending = p->signal->shared_pending.signal;
 267                 blocked = p->blocked;
 268                 collect_sigign_sigcatch(p, &ignored, &caught);
 269                 num_threads = atomic_read(&p->signal->count);
 270                 qsize = atomic_read(&p->user->sigpending);
 271                 qlim = p->signal->rlim[RLIMIT_SIGPENDING].rlim_cur;
 272                 spin_unlock_irq(&p->sighand->siglock);
 273         }
 274         read_unlock(&tasklist_lock);
 275
 276         buffer += sprintf(buffer, "Threads:\t%d\n", num_threads);
 277         buffer += sprintf(buffer, "SigQ:\t%lu/%lu\n", qsize, qlim);
 278
 279         /* render them all */
 280         buffer = render_sigset_t("SigPnd:\t", &pending, buffer);
 281         buffer = render_sigset_t("ShdPnd:\t", &shpending, buffer);
 282         buffer = render_sigset_t("SigBlk:\t", &blocked, buffer);
 283         buffer = render_sigset_t("SigIgn:\t", &ignored, buffer);
 284         buffer = render_sigset_t("SigCgt:\t", &caught, buffer);
 285
 286         return buffer;
 287 }
 288
 289 static inline char *task_cap(struct task_struct *p, char *buffer)
 290 {
 291     return buffer + sprintf(buffer, "CapInh:\t%016x\n"
 292                             "CapPrm:\t%016x\n"
 293                             "CapEff:\t%016x\n",
 294                             cap_t(p->cap_inheritable),
 295                             cap_t(p->cap_permitted),
 296                             cap_t(p->cap_effective));
 297 }
 298
 299 int proc_pid_status(struct task_struct *task, char * buffer)
 300 {
 301         char * orig = buffer;
 302 #ifdef  CONFIG_VSERVER_LEGACY
 303         struct vx_info *vxi;
 304 #endif
 305 #ifdef  CONFIG_VSERVER_LEGACYNET
 306         struct nx_info *nxi;
 307 #endif
 308         struct mm_struct *mm = get_task_mm(task);
 309
 310         buffer = task_name(task, buffer);
 311         buffer = task_state(task, buffer);
 312
 313         if (mm) {
 314                 buffer = task_mem(mm, buffer);
 315                 mmput(mm);
 316         }
 317         buffer = task_sig(task, buffer);
 318         buffer = task_cap(task, buffer);
 319         buffer = cpuset_task_status_allowed(task, buffer);
 320
 321         if (task_vx_flags(task, VXF_INFO_HIDE, 0))
 322                 goto skip;
 323 #ifdef  CONFIG_VSERVER_LEGACY
 324         buffer += sprintf (buffer,"s_context: %d\n", vx_task_xid(task));
 325         vxi = task_get_vx_info(task);
 326         if (vxi) {
 327                 buffer += sprintf (buffer,"ctxflags: %08llx\n"
 328                         ,(unsigned long long)vxi->vx_flags);
 329                 buffer += sprintf (buffer,"initpid: %d\n"
 330                         ,vxi->vx_initpid);
 331         } else {
 332                 buffer += sprintf (buffer,"ctxflags: none\n");
 333                 buffer += sprintf (buffer,"initpid: none\n");
 334         }
 335         put_vx_info(vxi);
 336 #else
 337         buffer += sprintf (buffer,"VxID: %d\n", vx_task_xid(task));
 338 #endif
 339 #ifdef  CONFIG_VSERVER_LEGACYNET
 340         nxi = task_get_nx_info(task);
 341         if (nxi) {
 342                 int i;
 343
 344                 buffer += sprintf (buffer,"ipv4root:");
 345                 for (i=0; i<nxi->nbipv4; i++){
 346                         buffer += sprintf (buffer," %08x/%08x"
 347                                 ,nxi->ipv4[i]
 348                                 ,nxi->mask[i]);
 349                 }
 350                 *buffer++ = '\n';
 351                 buffer += sprintf (buffer,"ipv4root_bcast: %08x\n"
 352                         ,nxi->v4_bcast);
 353         } else {
 354                 buffer += sprintf (buffer,"ipv4root: 0\n");
 355                 buffer += sprintf (buffer,"ipv4root_bcast: 0\n");
 356         }
 357         put_nx_info(nxi);
 358 #endif
 359 skip:
 360 #if defined(CONFIG_ARCH_S390)
 361         buffer = task_show_regs(task, buffer);
 362 #endif
 363         return buffer - orig;
 364 }
 365
 366 static int do_task_stat(struct task_struct *task, char * buffer, int whole)
 367 {
 368         unsigned long vsize, eip, esp, wchan = ~0UL;
 369         long priority, nice;
 370         int tty_pgrp = -1, tty_nr = 0;
 371         sigset_t sigign, sigcatch;
 372         char state;
 373         int res;
 374         pid_t pid, ppid, pgid = -1, sid = -1;
 375         int num_threads = 0;
 376         struct mm_struct *mm;
 377         unsigned long long start_time;
 378         unsigned long cmin_flt = 0, cmaj_flt = 0;
 379         unsigned long  min_flt = 0,  maj_flt = 0;
 380         cputime_t cutime, cstime, utime, stime;
 381         unsigned long rsslim = 0;
 382         unsigned long it_real_value = 0;
 383         struct task_struct *t;
 384         char tcomm[sizeof(task->comm)];
 385
 386         state = *get_task_state(task);
 387         vsize = eip = esp = 0;
 388         mm = get_task_mm(task);
 389         if (mm) {
 390                 vsize = task_vsize(mm);
 391                 eip = KSTK_EIP(task);
 392                 esp = KSTK_ESP(task);
 393         }
 394
 395         get_task_comm(tcomm, task);
 396
 397         sigemptyset(&sigign);
 398         sigemptyset(&sigcatch);
 399         cutime = cstime = utime = stime = cputime_zero;
 400         read_lock(&tasklist_lock);
 401         if (task->sighand) {
 402                 spin_lock_irq(&task->sighand->siglock);
 403                 num_threads = atomic_read(&task->signal->count);
 404                 collect_sigign_sigcatch(task, &sigign, &sigcatch);
 405
 406                 /* add up live thread stats at the group level */
 407                 if (whole) {
 408                         t = task;
 409                         do {
 410                                 min_flt += t->min_flt;
 411                                 maj_flt += t->maj_flt;
 412                                 utime = cputime_add(utime, t->utime);
 413                                 stime = cputime_add(stime, t->stime);
 414                                 t = next_thread(t);
 415                         } while (t != task);
 416                 }
 417
 418                 spin_unlock_irq(&task->sighand->siglock);
 419         }
 420         if (task->signal) {
 421                 if (task->signal->tty) {
 422                         tty_pgrp = task->signal->tty->pgrp;
 423                         tty_nr = new_encode_dev(tty_devnum(task->signal->tty));
 424                 }
 425                 pgid = process_group(task);
 426                 sid = task->signal->session;
 427                 cmin_flt = task->signal->cmin_flt;
 428                 cmaj_flt = task->signal->cmaj_flt;
 429                 cutime = task->signal->cutime;
 430                 cstime = task->signal->cstime;
 431                 rsslim = task->signal->rlim[RLIMIT_RSS].rlim_cur;
 432                 if (whole) {
 433                         min_flt += task->signal->min_flt;
 434                         maj_flt += task->signal->maj_flt;
 435                         utime = cputime_add(utime, task->signal->utime);
 436                         stime = cputime_add(stime, task->signal->stime);
 437                 }
 438                 it_real_value = task->signal->it_real_value;
 439         }
 440         pid = vx_info_map_pid(task->vx_info, pid_alive(task) ? task->pid : 0);
 441         ppid = (!(pid > 1)) ? 0 : vx_info_map_tgid(task->vx_info,
 442                 task->group_leader->real_parent->tgid);
 443         pgid = vx_info_map_pid(task->vx_info, pgid);
 444
 445         read_unlock(&tasklist_lock);
 446
 447         if (!whole || num_threads<2)
 448                 wchan = get_wchan(task);
 449         if (!whole) {
 450                 min_flt = task->min_flt;
 451                 maj_flt = task->maj_flt;
 452                 utime = task->utime;
 453                 stime = task->stime;
 454         }
 455
 456         /* scale priority and nice values from timeslices to -20..20 */
 457         /* to make it look like a "normal" Unix priority/nice value  */
 458         priority = task_prio(task);
 459         nice = task_nice(task);
 460
 461         /* Temporary variable needed for gcc-2.96 */
 462         /* convert timespec -> nsec*/
 463         start_time = (unsigned long long)task->start_time.tv_sec * NSEC_PER_SEC
 464                                 + task->start_time.tv_nsec;
 465
 466         /* convert nsec -> ticks */
 467         start_time = nsec_to_clock_t(start_time);
 468
 469         /* fixup start time for virt uptime */
 470         if (vx_flags(VXF_VIRT_UPTIME, 0)) {
 471                 unsigned long long bias =
 472                         current->vx_info->cvirt.bias_clock;
 473
 474                 if (start_time > bias)
 475                         start_time -= bias;
 476                 else
 477                         start_time = 0;
 478         }
 479
 480         res = sprintf(buffer,"%d (%s) %c %d %d %d %d %d %lu %lu \
 481 %lu %lu %lu %lu %lu %ld %ld %ld %ld %d %ld %llu %lu %ld %lu %lu %lu %lu %lu \
 482 %lu %lu %lu %lu %lu %lu %lu %lu %d %d %lu %lu\n",
 483                 pid,
 484                 tcomm,
 485                 state,
 486                 ppid,
 487                 pgid,
 488                 sid,
 489                 tty_nr,
 490                 tty_pgrp,
 491                 task->flags,
 492                 min_flt,
 493                 cmin_flt,
 494                 maj_flt,
 495                 cmaj_flt,
 496                 cputime_to_clock_t(utime),
 497                 cputime_to_clock_t(stime),
 498                 cputime_to_clock_t(cutime),
 499                 cputime_to_clock_t(cstime),
 500                 priority,
 501                 nice,
 502                 num_threads,
 503                 jiffies_to_clock_t(it_real_value),
 504                 start_time,
 505                 vsize,
 506                 mm ? get_mm_counter(mm, rss) : 0, /* you might want to shift this left 3 */
 507                 rsslim,
 508                 mm ? mm->start_code : 0,
 509                 mm ? mm->end_code : 0,
 510                 mm ? mm->start_stack : 0,
 511                 esp,
 512                 eip,
 513                 /* The signal information here is obsolete.
 514                  * It must be decimal for Linux 2.0 compatibility.
 515                  * Use /proc/#/status for real-time signals.
 516                  */
 517                 task->pending.signal.sig[0] & 0x7fffffffUL,
 518                 task->blocked.sig[0] & 0x7fffffffUL,
 519                 sigign      .sig[0] & 0x7fffffffUL,
 520                 sigcatch    .sig[0] & 0x7fffffffUL,
 521                 wchan,
 522                 0UL,
 523                 0UL,
 524                 task->exit_signal,
 525                 task_cpu(task),
 526                 task->rt_priority,
 527                 task->policy);
 528         if(mm)
 529                 mmput(mm);
 530         return res;
 531 }
 532
 533 int proc_tid_stat(struct task_struct *task, char * buffer)
 534 {
 535         return do_task_stat(task, buffer, 0);
 536 }
 537
 538 int proc_tgid_stat(struct task_struct *task, char * buffer)
 539 {
 540         return do_task_stat(task, buffer, 1);
 541 }
 542
 543 int proc_pid_statm(struct task_struct *task, char *buffer)
 544 {
 545         int size = 0, resident = 0, shared = 0, text = 0, lib = 0, data = 0;
 546         struct mm_struct *mm = get_task_mm(task);
 547
 548         if (mm) {
 549                 size = task_statm(mm, &shared, &text, &data, &resident);
 550                 mmput(mm);
 551         }
 552
 553         return sprintf(buffer,"%d %d %d %d %d %d %d\n",
 554                        size, resident, shared, text, lib, data, 0);
 555 }