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/sched.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/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/seq_file.h>
  74
  75 #include <asm/uaccess.h>
  76 #include <asm/pgtable.h>
  77 #include <asm/io.h>
  78 #include <asm/processor.h>
  79
  80 /* Gcc optimizes away "strlen(x)" for constant x */
  81 #define ADDBUF(buffer, string) \
  82 do { memcpy(buffer, string, strlen(string)); \
  83      buffer += strlen(string); } while (0)
  84
  85 static inline char * task_name(struct task_struct *p, char * buf)
  86 {
  87         int i;
  88         char * name;
  89         char tcomm[sizeof(p->comm)];
  90
  91         get_task_comm(tcomm, p);
  92
  93         ADDBUF(buf, "Name:\t");
  94         name = tcomm;
  95         i = sizeof(tcomm);
  96         do {
  97                 unsigned char c = *name;
  98                 name++;
  99                 i--;
 100                 *buf = c;
 101                 if (!c)
 102                         break;
 103                 if (c == '\\') {
 104                         buf[1] = c;
 105                         buf += 2;
 106                         continue;
 107                 }
 108                 if (c == '\n') {
 109                         buf[0] = '\\';
 110                         buf[1] = 'n';
 111                         buf += 2;
 112                         continue;
 113                 }
 114                 buf++;
 115         } while (i);
 116         *buf = '\n';
 117         return buf+1;
 118 }
 119
 120 /*
 121  * The task state array is a strange "bitmap" of
 122  * reasons to sleep. Thus "running" is zero, and
 123  * you can test for combinations of others with
 124  * simple bit tests.
 125  */
 126 static const char *task_state_array[] = {
 127         "R (running)",          /*  0 */
 128         "S (sleeping)",         /*  1 */
 129         "D (disk sleep)",       /*  2 */
 130         "Z (zombie)",           /*  4 */
 131         "T (stopped)",          /*  8 */
 132         "W (paging)"            /* 16 */
 133 };
 134
 135 static inline const char * get_task_state(struct task_struct *tsk)
 136 {
 137         unsigned int state = tsk->state & (TASK_RUNNING |
 138                                            TASK_INTERRUPTIBLE |
 139                                            TASK_UNINTERRUPTIBLE |
 140                                            TASK_ZOMBIE |
 141                                            TASK_STOPPED);
 142         const char **p = &task_state_array[0];
 143
 144         while (state) {
 145                 p++;
 146                 state >>= 1;
 147         }
 148         return *p;
 149 }
 150
 151 static inline char * task_state(struct task_struct *p, char *buffer)
 152 {
 153         int g;
 154
 155         read_lock(&tasklist_lock);
 156         buffer += sprintf(buffer,
 157                 "State:\t%s\n"
 158                 "Tgid:\t%d\n"
 159                 "Pid:\t%d\n"
 160                 "PPid:\t%d\n"
 161                 "TracerPid:\t%d\n"
 162                 "Uid:\t%d\t%d\t%d\t%d\n"
 163                 "Gid:\t%d\t%d\t%d\t%d\n",
 164                 get_task_state(p), p->tgid,
 165                 p->pid, p->pid ? p->p_opptr->pid : 0, 0,
 166                 p->uid, p->euid, p->suid, p->fsuid,
 167                 p->gid, p->egid, p->sgid, p->fsgid);
 168         read_unlock(&tasklist_lock);
 169         task_lock(p);
 170         buffer += sprintf(buffer,
 171                 "FDSize:\t%d\n"
 172                 "Groups:\t",
 173                 p->files ? p->files->max_fds : 0);
 174         task_unlock(p);
 175
 176         for (g = 0; g < p->ngroups; g++)
 177                 buffer += sprintf(buffer, "%d ", p->groups[g]);
 178
 179         buffer += sprintf(buffer, "\n");
 180         return buffer;
 181 }
 182
 183 static inline char * task_mem(struct mm_struct *mm, char *buffer)
 184 {
 185         struct vm_area_struct * vma;
 186         unsigned long data = 0, stack = 0;
 187         unsigned long exec = 0, lib = 0;
 188
 189         down_read(&mm->mmap_sem);
 190         for (vma = mm->mmap; vma; vma = vma->vm_next) {
 191                 unsigned long len = (vma->vm_end - vma->vm_start) >> 10;
 192                 if (!vma->vm_file) {
 193                         data += len;
 194                         if (vma->vm_flags & VM_GROWSDOWN)
 195                                 stack += len;
 196                         continue;
 197                 }
 198                 if (vma->vm_flags & VM_WRITE)
 199                         continue;
 200                 if (vma->vm_flags & VM_EXEC) {
 201                         exec += len;
 202                         if (vma->vm_flags & VM_EXECUTABLE)
 203                                 continue;
 204                         lib += len;
 205                 }
 206         }
 207         buffer += sprintf(buffer,
 208                 "VmSize:\t%8lu kB\n"
 209                 "VmLck:\t%8lu kB\n"
 210                 "VmRSS:\t%8lu kB\n"
 211                 "VmData:\t%8lu kB\n"
 212                 "VmStk:\t%8lu kB\n"
 213                 "VmExe:\t%8lu kB\n"
 214                 "VmLib:\t%8lu kB\n",
 215                 mm->total_vm << (PAGE_SHIFT-10),
 216                 mm->locked_vm << (PAGE_SHIFT-10),
 217                 mm->rss << (PAGE_SHIFT-10),
 218                 data - stack, stack,
 219                 exec - lib, lib);
 220         up_read(&mm->mmap_sem);
 221         return buffer;
 222 }
 223
 224 static void collect_sigign_sigcatch(struct task_struct *p, sigset_t *ign,
 225                                     sigset_t *catch)
 226 {
 227         struct k_sigaction *k;
 228         int i;
 229
 230         sigemptyset(ign);
 231         sigemptyset(catch);
 232
 233         spin_lock_irq(&p->sigmask_lock);
 234
 235         if (p->sig) {
 236                 k = p->sig->action;
 237                 for (i = 1; i <= _NSIG; ++i, ++k) {
 238                         if (k->sa.sa_handler == SIG_IGN)
 239                                 sigaddset(ign, i);
 240                         else if (k->sa.sa_handler != SIG_DFL)
 241                                 sigaddset(catch, i);
 242                 }
 243         }
 244         spin_unlock_irq(&p->sigmask_lock);
 245 }
 246
 247 static inline char * task_sig(struct task_struct *p, char *buffer)
 248 {
 249         sigset_t ign, catch;
 250
 251         buffer += sprintf(buffer, "SigPnd:\t");
 252         buffer = render_sigset_t(&p->pending.signal, buffer);
 253         *buffer++ = '\n';
 254         buffer += sprintf(buffer, "SigBlk:\t");
 255         buffer = render_sigset_t(&p->blocked, buffer);
 256         *buffer++ = '\n';
 257
 258         collect_sigign_sigcatch(p, &ign, &catch);
 259         buffer += sprintf(buffer, "SigIgn:\t");
 260         buffer = render_sigset_t(&ign, buffer);
 261         *buffer++ = '\n';
 262         buffer += sprintf(buffer, "SigCgt:\t"); /* Linux 2.0 uses "SigCgt" */
 263         buffer = render_sigset_t(&catch, buffer);
 264         *buffer++ = '\n';
 265
 266         return buffer;
 267 }
 268
 269 static inline char *task_cap(struct task_struct *p, char *buffer)
 270 {
 271     return buffer + sprintf(buffer, "CapInh:\t%016x\n"
 272                             "CapPrm:\t%016x\n"
 273                             "CapEff:\t%016x\n",
 274                             cap_t(p->cap_inheritable),
 275                             cap_t(p->cap_permitted),
 276                             cap_t(p->cap_effective));
 277 }
 278
 279
 280 int proc_pid_status(struct task_struct *task, char * buffer)
 281 {
 282         char * orig = buffer;
 283         struct mm_struct *mm;
 284
 285         buffer = task_name(task, buffer);
 286         buffer = task_state(task, buffer);
 287         task_lock(task);
 288         mm = task->mm;
 289         if(mm)
 290                 atomic_inc(&mm->mm_users);
 291         task_unlock(task);
 292         if (mm) {
 293                 buffer = task_mem(mm, buffer);
 294                 mmput(mm);
 295         }
 296         buffer = task_sig(task, buffer);
 297         buffer = task_cap(task, buffer);
 298 #if defined(CONFIG_ARCH_S390)
 299         buffer = task_show_regs(task, buffer);
 300 #endif
 301         return buffer - orig;
 302 }
 303
 304 int proc_pid_stat(struct task_struct *task, char * buffer)
 305 {
 306         unsigned long vsize, eip, esp, wchan;
 307         long priority, nice;
 308         int tty_pgrp = -1, tty_nr = 0;
 309         sigset_t sigign, sigcatch;
 310         char state;
 311         int res;
 312         pid_t ppid;
 313         struct mm_struct *mm;
 314         char tcomm[sizeof(task->comm)];
 315
 316         state = *get_task_state(task);
 317         vsize = eip = esp = 0;
 318         task_lock(task);
 319         mm = task->mm;
 320         if(mm)
 321                 atomic_inc(&mm->mm_users);
 322         if (task->tty) {
 323                 tty_pgrp = task->tty->pgrp;
 324                 tty_nr = kdev_t_to_nr(task->tty->device);
 325         }
 326         task_unlock(task);
 327         if (mm) {
 328                 struct vm_area_struct *vma;
 329                 down_read(&mm->mmap_sem);
 330                 vma = mm->mmap;
 331                 while (vma) {
 332                         vsize += vma->vm_end - vma->vm_start;
 333                         vma = vma->vm_next;
 334                 }
 335                 eip = KSTK_EIP(task);
 336                 esp = KSTK_ESP(task);
 337                 up_read(&mm->mmap_sem);
 338         }
 339
 340         get_task_comm(tcomm, task);
 341
 342         wchan = get_wchan(task);
 343
 344         collect_sigign_sigcatch(task, &sigign, &sigcatch);
 345
 346         /* scale priority and nice values from timeslices to -20..20 */
 347         /* to make it look like a "normal" Unix priority/nice value  */
 348         priority = task->counter;
 349         priority = 20 - (priority * 10 + DEF_COUNTER / 2) / DEF_COUNTER;
 350         nice = task->nice;
 351
 352         read_lock(&tasklist_lock);
 353         ppid = task->pid ? task->p_opptr->pid : 0;
 354         read_unlock(&tasklist_lock);
 355         res = sprintf(buffer,"%d (%s) %c %d %d %d %d %d %lu %lu \
 356 %lu %lu %lu %lu %lu %ld %ld %ld %ld %ld %ld %lu %lu %ld %lu %lu %lu %lu %lu \
 357 %lu %lu %lu %lu %lu %lu %lu %lu %d %d\n",
 358                 task->pid,
 359                 tcomm,
 360                 state,
 361                 ppid,
 362                 task->pgrp,
 363                 task->session,
 364                 tty_nr,
 365                 tty_pgrp,
 366                 task->flags,
 367                 task->min_flt,
 368                 task->cmin_flt,
 369                 task->maj_flt,
 370                 task->cmaj_flt,
 371                 task->times.tms_utime,
 372                 task->times.tms_stime,
 373                 task->times.tms_cutime,
 374                 task->times.tms_cstime,
 375                 priority,
 376                 nice,
 377                 0UL /* removed */,
 378                 task->it_real_value,
 379                 task->start_time,
 380                 vsize,
 381                 mm ? mm->rss : 0, /* you might want to shift this left 3 */
 382                 task->rlim[RLIMIT_RSS].rlim_cur,
 383                 mm ? mm->start_code : 0,
 384                 mm ? mm->end_code : 0,
 385                 mm ? mm->start_stack : 0,
 386                 esp,
 387                 eip,
 388                 /* The signal information here is obsolete.
 389                  * It must be decimal for Linux 2.0 compatibility.
 390                  * Use /proc/#/status for real-time signals.
 391                  */
 392                 task->pending.signal.sig[0] & 0x7fffffffUL,
 393                 task->blocked.sig[0] & 0x7fffffffUL,
 394                 sigign      .sig[0] & 0x7fffffffUL,
 395                 sigcatch    .sig[0] & 0x7fffffffUL,
 396                 wchan,
 397                 task->nswap,
 398                 task->cnswap,
 399                 task->exit_signal,
 400                 task->processor);
 401         if(mm)
 402                 mmput(mm);
 403         return res;
 404 }
 405
 406 static inline void statm_pte_range(pmd_t * pmd, unsigned long address, unsigned long size,
 407         int * pages, int * shared, int * dirty, int * total)
 408 {
 409         pte_t * pte;
 410         unsigned long end;
 411
 412         if (pmd_none(*pmd))
 413                 return;
 414         if (pmd_bad(*pmd)) {
 415                 pmd_ERROR(*pmd);
 416                 pmd_clear(pmd);
 417                 return;
 418         }
 419         pte = pte_offset(pmd, address);
 420         address &= ~PMD_MASK;
 421         end = address + size;
 422         if (end > PMD_SIZE)
 423                 end = PMD_SIZE;
 424         do {
 425                 pte_t page = *pte;
 426                 struct page *ptpage;
 427
 428                 address += PAGE_SIZE;
 429                 pte++;
 430                 if (pte_none(page))
 431                         continue;
 432                 ++*total;
 433                 if (!pte_present(page))
 434                         continue;
 435                 ptpage = pte_page(page);
 436                 if ((!VALID_PAGE(ptpage)) || PageReserved(ptpage))
 437                         continue;
 438                 ++*pages;
 439                 if (pte_dirty(page))
 440                         ++*dirty;
 441                 if (page_count(pte_page(page)) > 1)
 442                         ++*shared;
 443         } while (address < end);
 444 }
 445
 446 static inline void statm_pmd_range(pgd_t * pgd, unsigned long address, unsigned long size,
 447         int * pages, int * shared, int * dirty, int * total)
 448 {
 449         pmd_t * pmd;
 450         unsigned long end;
 451
 452         if (pgd_none(*pgd))
 453                 return;
 454         if (pgd_bad(*pgd)) {
 455                 pgd_ERROR(*pgd);
 456                 pgd_clear(pgd);
 457                 return;
 458         }
 459         pmd = pmd_offset(pgd, address);
 460         address &= ~PGDIR_MASK;
 461         end = address + size;
 462         if (end > PGDIR_SIZE)
 463                 end = PGDIR_SIZE;
 464         do {
 465                 statm_pte_range(pmd, address, end - address, pages, shared, dirty, total);
 466                 address = (address + PMD_SIZE) & PMD_MASK;
 467                 pmd++;
 468         } while (address < end);
 469 }
 470
 471 static void statm_pgd_range(pgd_t * pgd, unsigned long address, unsigned long end,
 472         int * pages, int * shared, int * dirty, int * total)
 473 {
 474         while (address < end) {
 475                 statm_pmd_range(pgd, address, end - address, pages, shared, dirty, total);
 476                 address = (address + PGDIR_SIZE) & PGDIR_MASK;
 477                 pgd++;
 478         }
 479 }
 480
 481 int proc_pid_statm(struct task_struct *task, char * buffer)
 482 {
 483         struct mm_struct *mm;
 484         int size=0, resident=0, share=0, trs=0, lrs=0, drs=0, dt=0;
 485
 486         task_lock(task);
 487         mm = task->mm;
 488         if(mm)
 489                 atomic_inc(&mm->mm_users);
 490         task_unlock(task);
 491         if (mm) {
 492                 struct vm_area_struct * vma;
 493                 down_read(&mm->mmap_sem);
 494                 vma = mm->mmap;
 495                 while (vma) {
 496                         pgd_t *pgd = pgd_offset(mm, vma->vm_start);
 497                         int pages = 0, shared = 0, dirty = 0, total = 0;
 498
 499                         statm_pgd_range(pgd, vma->vm_start, vma->vm_end, &pages, &shared, &dirty, &total);
 500                         resident += pages;
 501                         share += shared;
 502                         dt += dirty;
 503                         size += total;
 504                         if (vma->vm_flags & VM_EXECUTABLE)
 505                                 trs += pages;   /* text */
 506                         else if (vma->vm_flags & VM_GROWSDOWN)
 507                                 drs += pages;   /* stack */
 508                         else if (vma->vm_end > 0x60000000)
 509                                 lrs += pages;   /* library */
 510                         else
 511                                 drs += pages;
 512                         vma = vma->vm_next;
 513                 }
 514                 up_read(&mm->mmap_sem);
 515                 mmput(mm);
 516         }
 517         return sprintf(buffer,"%d %d %d %d %d %d %d\n",
 518                        size, resident, share, trs, lrs, drs, dt);
 519 }
 520
 521 static int show_map(struct seq_file *m, void *v)
 522 {
 523         struct vm_area_struct *map = v;
 524         struct file *file = map->vm_file;
 525         int flags = map->vm_flags;
 526         unsigned long ino = 0;
 527         dev_t dev = 0;
 528         int len;
 529
 530         if (file) {
 531                 struct inode *inode = map->vm_file->f_dentry->d_inode;
 532                 dev = kdev_t_to_nr(inode->i_sb->s_dev);
 533                 ino = inode->i_ino;
 534         }
 535
 536         seq_printf(m, "%08lx-%08lx %c%c%c%c %08lx %02x:%02x %lu %n",
 537                         map->vm_start,
 538                         map->vm_end,
 539                         flags & VM_READ ? 'r' : '-',
 540                         flags & VM_WRITE ? 'w' : '-',
 541                         flags & VM_EXEC ? 'x' : '-',
 542                         flags & VM_MAYSHARE ? 's' : 'p',
 543                         map->vm_pgoff << PAGE_SHIFT,
 544                         MAJOR(dev), MINOR(dev), ino, &len);
 545
 546         if (map->vm_file) {
 547                 len = 25 + sizeof(void*) * 6 - len;
 548                 if (len < 1)
 549                         len = 1;
 550                 seq_printf(m, "%*c", len, ' ');
 551                 seq_path(m, file->f_vfsmnt, file->f_dentry, "");
 552         }
 553         seq_putc(m, '\n');
 554         return 0;
 555 }
 556
 557 static void *m_start(struct seq_file *m, loff_t *pos)
 558 {
 559         struct task_struct *task = m->private;
 560         struct mm_struct *mm;
 561         struct vm_area_struct * map;
 562         loff_t l = *pos;
 563
 564         task_lock(task);
 565         mm = task->mm;
 566         if (mm)
 567                 atomic_inc(&mm->mm_users);
 568         task_unlock(task);
 569
 570         if (!mm)
 571                 return NULL;
 572
 573         down_read(&mm->mmap_sem);
 574         map = mm->mmap;
 575         while (l-- && map)
 576                 map = map->vm_next;
 577         if (!map) {
 578                 up_read(&mm->mmap_sem);
 579                 mmput(mm);
 580         }
 581         return map;
 582 }
 583
 584 static void m_stop(struct seq_file *m, void *v)
 585 {
 586         struct vm_area_struct *map = v;
 587         if (map) {
 588                 struct mm_struct *mm = map->vm_mm;
 589                 up_read(&mm->mmap_sem);
 590                 mmput(mm);
 591         }
 592 }
 593
 594 static void *m_next(struct seq_file *m, void *v, loff_t *pos)
 595 {
 596         struct vm_area_struct *map = v;
 597         (*pos)++;
 598         if (map->vm_next)
 599                 return map->vm_next;
 600         m_stop(m, v);
 601         return NULL;
 602 }
 603
 604 struct seq_operations proc_pid_maps_op = {
 605         .start  = m_start,
 606         .next   = m_next,
 607         .stop   = m_stop,
 608         .show   = show_map
 609 };
 610
 611 #ifdef CONFIG_SMP
 612 int proc_pid_cpu(struct task_struct *task, char * buffer)
 613 {
 614         int i, len;
 615
 616         len = sprintf(buffer,
 617                 "cpu  %lu %lu\n",
 618                 task->times.tms_utime,
 619                 task->times.tms_stime);
 620
 621         for (i = 0 ; i < smp_num_cpus; i++)
 622                 len += sprintf(buffer + len, "cpu%d %lu %lu\n",
 623                         i,
 624                         task->per_cpu_utime[cpu_logical_map(i)],
 625                         task->per_cpu_stime[cpu_logical_map(i)]);
 626
 627         return len;
 628 }
 629 #endif