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