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/vs_context.h>
  77 #include <linux/vs_network.h>
  78 #include <linux/vs_cvirt.h>
  79
  80 #include <asm/uaccess.h>
  81 #include <asm/pgtable.h>
  82 #include <asm/io.h>
  83 #include <asm/processor.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         "H (on hold)"           /* 64 */
 140 };
 141
 142 static inline const char * get_task_state(struct task_struct *tsk)
 143 {
 144         unsigned int state = (tsk->state & (TASK_RUNNING |
 145                                             TASK_INTERRUPTIBLE |
 146                                             TASK_UNINTERRUPTIBLE |
 147                                             TASK_STOPPED |
 148                                            TASK_TRACED |
 149                                            TASK_ONHOLD)) |
 150                         (tsk->exit_state & (EXIT_ZOMBIE |
 151                                             EXIT_DEAD));
 152         const char **p = &task_state_array[0];
 153
 154         while (state) {
 155                 p++;
 156                 state >>= 1;
 157         }
 158         return *p;
 159 }
 160
 161 static inline char * task_state(struct task_struct *p, char *buffer)
 162 {
 163         struct group_info *group_info;
 164         int g;
 165         pid_t pid, ptgid, tppid, tgid;
 166
 167         read_lock(&tasklist_lock);
 168         tgid = vx_map_tgid(p->tgid);
 169         pid = vx_map_pid(p->pid);
 170         ptgid = vx_map_pid(p->group_leader->real_parent->tgid);
 171         tppid = vx_map_pid(p->parent->pid);
 172         buffer += sprintf(buffer,
 173                 "State:\t%s\n"
 174                 "SleepAVG:\t%lu%%\n"
 175                 "Tgid:\t%d\n"
 176                 "Pid:\t%d\n"
 177                 "PPid:\t%d\n"
 178                 "TracerPid:\t%d\n"
 179                 "Uid:\t%d\t%d\t%d\t%d\n"
 180                 "Gid:\t%d\t%d\t%d\t%d\n",
 181                 get_task_state(p),
 182                 (p->sleep_avg/1024)*100/(1020000000/1024),
 183                 tgid, pid, (pid > 1) ? ptgid : 0,
 184                 pid_alive(p) && p->ptrace ? tppid : 0,
 185                 p->uid, p->euid, p->suid, p->fsuid,
 186                 p->gid, p->egid, p->sgid, p->fsgid);
 187         read_unlock(&tasklist_lock);
 188         task_lock(p);
 189         buffer += sprintf(buffer,
 190                 "FDSize:\t%d\n"
 191                 "Groups:\t",
 192                 p->files ? p->files->max_fds : 0);
 193
 194         group_info = p->group_info;
 195         get_group_info(group_info);
 196         task_unlock(p);
 197
 198         for (g = 0; g < min(group_info->ngroups,NGROUPS_SMALL); g++)
 199                 buffer += sprintf(buffer, "%d ", GROUP_AT(group_info,g));
 200         put_group_info(group_info);
 201
 202         buffer += sprintf(buffer, "\n");
 203         return buffer;
 204 }
 205
 206 static char * render_sigset_t(const char *header, sigset_t *set, char *buffer)
 207 {
 208         int i, len;
 209
 210         len = strlen(header);
 211         memcpy(buffer, header, len);
 212         buffer += len;
 213
 214         i = _NSIG;
 215         do {
 216                 int x = 0;
 217
 218                 i -= 4;
 219                 if (sigismember(set, i+1)) x |= 1;
 220                 if (sigismember(set, i+2)) x |= 2;
 221                 if (sigismember(set, i+3)) x |= 4;
 222                 if (sigismember(set, i+4)) x |= 8;
 223                 *buffer++ = (x < 10 ? '0' : 'a' - 10) + x;
 224         } while (i >= 4);
 225
 226         *buffer++ = '\n';
 227         *buffer = 0;
 228         return buffer;
 229 }
 230
 231 static void collect_sigign_sigcatch(struct task_struct *p, sigset_t *ign,
 232                                     sigset_t *catch)
 233 {
 234         struct k_sigaction *k;
 235         int i;
 236
 237         k = p->sighand->action;
 238         for (i = 1; i <= _NSIG; ++i, ++k) {
 239                 if (k->sa.sa_handler == SIG_IGN)
 240                         sigaddset(ign, i);
 241                 else if (k->sa.sa_handler != SIG_DFL)
 242                         sigaddset(catch, i);
 243         }
 244 }
 245
 246 static inline char * task_sig(struct task_struct *p, char *buffer)
 247 {
 248         sigset_t pending, shpending, blocked, ignored, caught;
 249         int num_threads = 0;
 250
 251         sigemptyset(&pending);
 252         sigemptyset(&shpending);
 253         sigemptyset(&blocked);
 254         sigemptyset(&ignored);
 255         sigemptyset(&caught);
 256
 257         /* Gather all the data with the appropriate locks held */
 258         read_lock(&tasklist_lock);
 259         if (p->sighand) {
 260                 spin_lock_irq(&p->sighand->siglock);
 261                 pending = p->pending.signal;
 262                 shpending = p->signal->shared_pending.signal;
 263                 blocked = p->blocked;
 264                 collect_sigign_sigcatch(p, &ignored, &caught);
 265                 num_threads = atomic_read(&p->signal->count);
 266                 spin_unlock_irq(&p->sighand->siglock);
 267         }
 268         read_unlock(&tasklist_lock);
 269
 270         buffer += sprintf(buffer, "Threads:\t%d\n", num_threads);
 271
 272         /* render them all */
 273         buffer = render_sigset_t("SigPnd:\t", &pending, buffer);
 274         buffer = render_sigset_t("ShdPnd:\t", &shpending, buffer);
 275         buffer = render_sigset_t("SigBlk:\t", &blocked, buffer);
 276         buffer = render_sigset_t("SigIgn:\t", &ignored, buffer);
 277         buffer = render_sigset_t("SigCgt:\t", &caught, buffer);
 278
 279         return buffer;
 280 }
 281
 282 static inline char *task_cap(struct task_struct *p, char *buffer)
 283 {
 284     return buffer + sprintf(buffer, "CapInh:\t%016x\n"
 285                             "CapPrm:\t%016x\n"
 286                             "CapEff:\t%016x\n",
 287                             cap_t(p->cap_inheritable),
 288                             cap_t(p->cap_permitted),
 289                             cap_t(p->cap_effective));
 290 }
 291
 292 int proc_pid_status(struct task_struct *task, char * buffer)
 293 {
 294         char * orig = buffer;
 295 #ifdef  CONFIG_VSERVER_LEGACY
 296         struct vx_info *vxi;
 297         struct nx_info *nxi;
 298 #endif
 299         struct mm_struct *mm = get_task_mm(task);
 300
 301         buffer = task_name(task, buffer);
 302         buffer = task_state(task, buffer);
 303
 304         if (mm) {
 305                 buffer = task_mem(mm, buffer);
 306                 mmput(mm);
 307         }
 308         buffer = task_sig(task, buffer);
 309         buffer = task_cap(task, buffer);
 310
 311 #ifdef  CONFIG_VSERVER_LEGACY
 312         buffer += sprintf (buffer,"s_context: %d\n", vx_task_xid(task));
 313         vxi = task_get_vx_info(task);
 314         if (vxi) {
 315                 buffer += sprintf (buffer,"ctxflags: %08llx\n"
 316                         ,(unsigned long long)vxi->vx_flags);
 317                 buffer += sprintf (buffer,"initpid: %d\n"
 318                         ,vxi->vx_initpid);
 319         } else {
 320                 buffer += sprintf (buffer,"ctxflags: none\n");
 321                 buffer += sprintf (buffer,"initpid: none\n");
 322         }
 323         put_vx_info(vxi);
 324         nxi = task_get_nx_info(task);
 325         if (nxi) {
 326                 int i;
 327
 328                 buffer += sprintf (buffer,"ipv4root:");
 329                 for (i=0; i<nxi->nbipv4; i++){
 330                         buffer += sprintf (buffer," %08x/%08x"
 331                                 ,nxi->ipv4[i]
 332                                 ,nxi->mask[i]);
 333                 }
 334                 *buffer++ = '\n';
 335                 buffer += sprintf (buffer,"ipv4root_bcast: %08x\n"
 336                         ,nxi->v4_bcast);
 337         } else {
 338                 buffer += sprintf (buffer,"ipv4root: 0\n");
 339                 buffer += sprintf (buffer,"ipv4root_bcast: 0\n");
 340         }
 341         put_nx_info(nxi);
 342 #endif
 343 #if defined(CONFIG_ARCH_S390)
 344         buffer = task_show_regs(task, buffer);
 345 #endif
 346         return buffer - orig;
 347 }
 348
 349 static int do_task_stat(struct task_struct *task, char * buffer, int whole)
 350 {
 351         unsigned long vsize, eip, esp, wchan = ~0UL;
 352         long priority, nice;
 353         unsigned long long bias_uptime = 0;
 354         int tty_pgrp = -1, tty_nr = 0;
 355         sigset_t sigign, sigcatch;
 356         char state;
 357         int res;
 358         pid_t pid, ppid, pgid = -1, sid = -1;
 359         int num_threads = 0;
 360         struct mm_struct *mm;
 361         unsigned long long start_time;
 362         unsigned long cmin_flt = 0, cmaj_flt = 0, cutime = 0, cstime = 0;
 363         unsigned long  min_flt = 0,  maj_flt = 0,  utime = 0,  stime = 0;
 364         unsigned long rsslim = 0;
 365         struct task_struct *t;
 366         char tcomm[sizeof(task->comm)];
 367
 368         state = *get_task_state(task);
 369         vsize = eip = esp = 0;
 370         mm = get_task_mm(task);
 371         if (mm) {
 372                 vsize = task_vsize(mm);
 373                 eip = KSTK_EIP(task);
 374                 esp = KSTK_ESP(task);
 375         }
 376
 377         get_task_comm(tcomm, task);
 378
 379         sigemptyset(&sigign);
 380         sigemptyset(&sigcatch);
 381         read_lock(&tasklist_lock);
 382         if (task->sighand) {
 383                 spin_lock_irq(&task->sighand->siglock);
 384                 num_threads = atomic_read(&task->signal->count);
 385                 collect_sigign_sigcatch(task, &sigign, &sigcatch);
 386
 387                 /* add up live thread stats at the group level */
 388                 if (whole) {
 389                         t = task;
 390                         do {
 391                                 min_flt += t->min_flt;
 392                                 maj_flt += t->maj_flt;
 393                                 utime += t->utime;
 394                                 stime += t->stime;
 395                                 t = next_thread(t);
 396                         } while (t != task);
 397                 }
 398
 399                 spin_unlock_irq(&task->sighand->siglock);
 400         }
 401         if (task->signal) {
 402                 if (task->signal->tty) {
 403                         tty_pgrp = task->signal->tty->pgrp;
 404                         tty_nr = new_encode_dev(tty_devnum(task->signal->tty));
 405                 }
 406                 pgid = process_group(task);
 407                 sid = task->signal->session;
 408                 cmin_flt = task->signal->cmin_flt;
 409                 cmaj_flt = task->signal->cmaj_flt;
 410                 cutime = task->signal->cutime;
 411                 cstime = task->signal->cstime;
 412                 rsslim = task->signal->rlim[RLIMIT_RSS].rlim_cur;
 413                 if (whole) {
 414                         min_flt += task->signal->min_flt;
 415                         maj_flt += task->signal->maj_flt;
 416                         utime += task->signal->utime;
 417                         stime += task->signal->stime;
 418                 }
 419         }
 420         pid = vx_info_map_pid(task->vx_info, pid_alive(task) ? task->pid : 0);
 421         ppid = (!(pid > 1)) ? 0 : vx_info_map_tgid(task->vx_info,
 422                 task->group_leader->real_parent->tgid);
 423         pgid = vx_info_map_pid(task->vx_info, pgid);
 424
 425         read_unlock(&tasklist_lock);
 426
 427         if (!whole || num_threads<2) {
 428                 wchan = 0;
 429                 if (current->uid == task->uid || current->euid == task->uid ||
 430                                                 capable(CAP_SYS_NICE))
 431                         wchan = get_wchan(task);
 432         }
 433         if (!whole) {
 434                 min_flt = task->min_flt;
 435                 maj_flt = task->maj_flt;
 436                 utime = task->utime;
 437                 stime = task->stime;
 438         }
 439
 440         /* scale priority and nice values from timeslices to -20..20 */
 441         /* to make it look like a "normal" Unix priority/nice value  */
 442         priority = task_prio(task);
 443         nice = task_nice(task);
 444
 445         read_lock(&tasklist_lock);
 446         pid = vx_info_map_pid(task->vx_info, task->pid);
 447         ppid = (!(pid > 1)) ? 0 :
 448                 vx_info_map_pid(task->vx_info, task->real_parent->pid);
 449         pgid = vx_info_map_pid(task->vx_info, pgid);
 450         read_unlock(&tasklist_lock);
 451
 452         /* Temporary variable needed for gcc-2.96 */
 453         /* convert timespec -> nsec*/
 454         start_time = (unsigned long long)task->start_time.tv_sec * NSEC_PER_SEC
 455                                 + task->start_time.tv_nsec;
 456
 457         /* convert nsec -> ticks */
 458         start_time = nsec_to_clock_t(start_time - bias_uptime);
 459
 460         /* fixup start time for virt uptime */
 461         if (vx_flags(VXF_VIRT_UPTIME, 0)) {
 462                 unsigned long long bias =
 463                         current->vx_info->cvirt.bias_clock;
 464
 465                 if (start_time > bias)
 466                         start_time -= bias;
 467                 else
 468                         start_time = 0;
 469         }
 470
 471         res = sprintf(buffer,"%d (%s) %c %d %d %d %d %d %lu %lu \
 472 %lu %lu %lu %lu %lu %ld %ld %ld %ld %d %ld %llu %lu %ld %lu %lu %lu %lu %lu \
 473 %lu %lu %lu %lu %lu %lu %lu %lu %d %d %lu %lu\n",
 474                 pid,
 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                 jiffies_to_clock_t(utime),
 488                 jiffies_to_clock_t(stime),
 489                 jiffies_to_clock_t(cutime),
 490                 jiffies_to_clock_t(cstime),
 491                 priority,
 492                 nice,
 493                 num_threads,
 494                 jiffies_to_clock_t(task->it_real_value),
 495                 start_time,
 496                 vsize,
 497                 mm ? mm->rss : 0, /* you might want to shift this left 3 */
 498                 rsslim,
 499                 mm ? mm->start_code : 0,
 500                 mm ? mm->end_code : 0,
 501                 mm ? mm->start_stack : 0,
 502                 esp,
 503                 eip,
 504                 /* The signal information here is obsolete.
 505                  * It must be decimal for Linux 2.0 compatibility.
 506                  * Use /proc/#/status for real-time signals.
 507                  */
 508                 task->pending.signal.sig[0] & 0x7fffffffUL,
 509                 task->blocked.sig[0] & 0x7fffffffUL,
 510                 sigign      .sig[0] & 0x7fffffffUL,
 511                 sigcatch    .sig[0] & 0x7fffffffUL,
 512                 wchan,
 513                 0UL,
 514                 0UL,
 515                 task->exit_signal,
 516                 task_cpu(task),
 517                 task->rt_priority,
 518                 task->policy);
 519         if(mm)
 520                 mmput(mm);
 521         return res;
 522 }
 523
 524 int proc_tid_stat(struct task_struct *task, char * buffer)
 525 {
 526         return do_task_stat(task, buffer, 0);
 527 }
 528
 529 int proc_tgid_stat(struct task_struct *task, char * buffer)
 530 {
 531         return do_task_stat(task, buffer, 1);
 532 }
 533
 534 int proc_pid_statm(struct task_struct *task, char *buffer)
 535 {
 536         int size = 0, resident = 0, shared = 0, text = 0, lib = 0, data = 0;
 537         struct mm_struct *mm = get_task_mm(task);
 538
 539         if (mm) {
 540                 size = task_statm(mm, &shared, &text, &data, &resident);
 541                 mmput(mm);
 542         }
 543
 544         return sprintf(buffer,"%d %d %d %d %d %d %d\n",
 545                        size, resident, shared, text, lib, data, 0);
 546 }