fs/proc/array.c

/* */
This source file includes following definitions.
read_core
read_profile
write_profile
get_loadavg
get_kstat
get_uptime
get_meminfo
get_version
get_cmdline
get_task
get_phys_addr
get_array
get_env
get_arg
get_wchan
get_stat
statm_pte_range
statm_pmd_range
statm_pgd_range
get_statm
read_maps
get_root_array
get_process_array
fill_array
array_read
arraylong_read
   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    :  Some minor additions for cpuinfo
  18  *                      <danny@ow.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 
  30 #include <linux/types.h>
  31 #include <linux/errno.h>
  32 #include <linux/sched.h>
  33 #include <linux/kernel.h>
  34 #include <linux/kernel_stat.h>
  35 #include <linux/tty.h>
  36 #include <linux/user.h>
  37 #include <linux/a.out.h>
  38 #include <linux/string.h>
  39 #include <linux/mman.h>
  40 #include <linux/proc_fs.h>
  41 #include <linux/ioport.h>
  42 #include <linux/config.h>
  43 #include <linux/mm.h>
  44 #ifdef CONFIG_APM
  45 #include <linux/apm_bios.h>
  46 #endif
  47 
  48 #include <asm/segment.h>
  49 #include <asm/pgtable.h>
  50 #include <asm/io.h>
  51 
  52 #define LOAD_INT(x) ((x) >> FSHIFT)
  53 #define LOAD_FRAC(x) LOAD_INT(((x) & (FIXED_1-1)) * 100)
  54 
  55 #ifdef CONFIG_DEBUG_MALLOC
  56 int get_malloc(char * buffer);
  57 #endif
  58 
  59 
  60 static int read_core(struct inode * inode, struct file * file,char * buf, int count)
     /*  */
  61 {
  62         unsigned long p = file->f_pos, memsize;
  63         int read;
  64         int count1;
  65         char * pnt;
  66         struct user dump;
  67 #ifdef __i386__
  68 #       define FIRST_MAPPED     PAGE_SIZE       /* we don't have page 0 mapped on x86.. */
  69 #else
  70 #       define FIRST_MAPPED     0
  71 #endif
  72 
  73         memset(&dump, 0, sizeof(struct user));
  74         dump.magic = CMAGIC;
  75         dump.u_dsize = MAP_NR(high_memory);
  76 #ifdef __alpha__
  77         dump.start_data = PAGE_OFFSET;
  78 #endif
  79 
  80         if (count < 0)
  81                 return -EINVAL;
  82         memsize = MAP_NR(high_memory + PAGE_SIZE) << PAGE_SHIFT;
  83         if (p >= memsize)
  84                 return 0;
  85         if (count > memsize - p)
  86                 count = memsize - p;
  87         read = 0;
  88 
  89         if (p < sizeof(struct user) && count > 0) {
  90                 count1 = count;
  91                 if (p + count1 > sizeof(struct user))
  92                         count1 = sizeof(struct user)-p;
  93                 pnt = (char *) &dump + p;
  94                 memcpy_tofs(buf,(void *) pnt, count1);
  95                 buf += count1;
  96                 p += count1;
  97                 count -= count1;
  98                 read += count1;
  99         }
 100 
 101         while (count > 0 && p < PAGE_SIZE + FIRST_MAPPED) {
 102                 put_user(0,buf);
 103                 buf++;
 104                 p++;
 105                 count--;
 106                 read++;
 107         }
 108         memcpy_tofs(buf, (void *) (PAGE_OFFSET + p - PAGE_SIZE), count);
 109         read += count;
 110         file->f_pos += read;
 111         return read;
 112 }
 113 
 114 static struct file_operations proc_kcore_operations = {
 115         NULL,           /* lseek */
 116         read_core,
 117 };
 118 
 119 struct inode_operations proc_kcore_inode_operations = {
 120         &proc_kcore_operations, 
 121 };
 122 
 123 
 124 extern unsigned long prof_len;
 125 extern unsigned long * prof_buffer;
 126 extern unsigned long prof_shift;
 127 /*
 128  * This function accesses profiling information. The returned data is
 129  * binary: the sampling step and the actual contents of the profile
 130  * buffer. Use of the program readprofile is recommended in order to
 131  * get meaningful info out of these data.
 132  */
 133 static int read_profile(struct inode *inode, struct file *file, char *buf, int count)
     /*  */
 134 {
 135     unsigned long p = file->f_pos;
 136         int read;
 137         char * pnt;
 138         unsigned long sample_step = 1 << prof_shift;
 139 
 140         if (count < 0)
 141             return -EINVAL;
 142         if (p >= (prof_len+1)*sizeof(unsigned long))
 143             return 0;
 144         if (count > (prof_len+1)*sizeof(unsigned long) - p)
 145             count = (prof_len+1)*sizeof(unsigned long) - p;
 146     read = 0;
 147 
 148     while (p < sizeof(unsigned long) && count > 0) {
 149         put_user(*((char *)(&sample_step)+p),buf);
 150                 buf++; p++; count--; read++;
 151     }
 152     pnt = (char *)prof_buffer + p - sizeof(unsigned long);
 153         memcpy_tofs(buf,(void *)pnt,count);
 154         read += count;
 155         file->f_pos += read;
 156         return read;
 157 }
 158 
 159 /* Writing to /proc/profile resets the counters */
 160 static int write_profile(struct inode * inode, struct file * file, const char * buf, int count)
     /*  */
 161 {
 162     int i=prof_len;
 163 
 164     while (i--)
 165             prof_buffer[i]=0UL;
 166     return count;
 167 }
 168 
 169 static struct file_operations proc_profile_operations = {
 170         NULL,           /* lseek */
 171         read_profile,
 172         write_profile,
 173 };
 174 
 175 struct inode_operations proc_profile_inode_operations = {
 176         &proc_profile_operations, 
 177 };
 178 
 179 
 180 static int get_loadavg(char * buffer)
     /*  */
 181 {
 182         int a, b, c;
 183 
 184         a = avenrun[0] + (FIXED_1/200);
 185         b = avenrun[1] + (FIXED_1/200);
 186         c = avenrun[2] + (FIXED_1/200);
 187         return sprintf(buffer,"%d.%02d %d.%02d %d.%02d %d/%d\n",
 188                 LOAD_INT(a), LOAD_FRAC(a),
 189                 LOAD_INT(b), LOAD_FRAC(b),
 190                 LOAD_INT(c), LOAD_FRAC(c),
 191                 nr_running, nr_tasks);
 192 }
 193 
 194 static int get_kstat(char * buffer)
     /*  */
 195 {
 196         int i, len;
 197         unsigned sum = 0;
 198 
 199         for (i = 0 ; i < NR_IRQS ; i++)
 200                 sum += kstat.interrupts[i];
 201         len = sprintf(buffer,
 202                 "cpu  %u %u %u %lu\n"
 203                 "disk %u %u %u %u\n"
 204                 "disk_rio %u %u %u %u\n"
 205                 "disk_wio %u %u %u %u\n"
 206                 "disk_rblk %u %u %u %u\n"
 207                 "disk_wblk %u %u %u %u\n"
 208                 "page %u %u\n"
 209                 "swap %u %u\n"
 210                 "intr %u",
 211                 kstat.cpu_user,
 212                 kstat.cpu_nice,
 213                 kstat.cpu_system,
 214                 jiffies - (kstat.cpu_user + kstat.cpu_nice + kstat.cpu_system),
 215                 kstat.dk_drive[0], kstat.dk_drive[1],
 216                 kstat.dk_drive[2], kstat.dk_drive[3],
 217                 kstat.dk_drive_rio[0], kstat.dk_drive_rio[1],
 218                 kstat.dk_drive_rio[2], kstat.dk_drive_rio[3],
 219                 kstat.dk_drive_wio[0], kstat.dk_drive_wio[1],
 220                 kstat.dk_drive_wio[2], kstat.dk_drive_wio[3],
 221                 kstat.dk_drive_rblk[0], kstat.dk_drive_rblk[1],
 222                 kstat.dk_drive_rblk[2], kstat.dk_drive_rblk[3],
 223                 kstat.dk_drive_wblk[0], kstat.dk_drive_wblk[1],
 224                 kstat.dk_drive_wblk[2], kstat.dk_drive_wblk[3],
 225                 kstat.pgpgin,
 226                 kstat.pgpgout,
 227                 kstat.pswpin,
 228                 kstat.pswpout,
 229                 sum);
 230         for (i = 0 ; i < NR_IRQS ; i++)
 231                 len += sprintf(buffer + len, " %u", kstat.interrupts[i]);
 232         len += sprintf(buffer + len,
 233                 "\nctxt %u\n"
 234                 "btime %lu\n",
 235                 kstat.context_swtch,
 236                 xtime.tv_sec - jiffies / HZ);
 237         return len;
 238 }
 239 
 240 
 241 static int get_uptime(char * buffer)
     /*  */
 242 {
 243         unsigned long uptime;
 244         unsigned long idle;
 245 
 246         uptime = jiffies;
 247         idle = task[0]->utime + task[0]->stime;
 248 
 249         /* The formula for the fraction parts really is ((t * 100) / HZ) % 100, but
 250            that would overflow about every five days at HZ == 100.
 251            Therefore the identity a = (a / b) * b + a % b is used so that it is
 252            calculated as (((t / HZ) * 100) + ((t % HZ) * 100) / HZ) % 100.
 253            The part in front of the '+' always evaluates as 0 (mod 100). All divisions
 254            in the above formulas are truncating. For HZ being a power of 10, the
 255            calculations simplify to the version in the #else part (if the printf
 256            format is adapted to the same number of digits as zeroes in HZ.
 257          */
 258 #if HZ!=100
 259         return sprintf(buffer,"%lu.%02lu %lu.%02lu\n",
 260                 uptime / HZ,
 261                 (((uptime % HZ) * 100) / HZ) % 100,
 262                 idle / HZ,
 263                 (((idle % HZ) * 100) / HZ) % 100);
 264 #else
 265         return sprintf(buffer,"%lu.%02lu %lu.%02lu\n",
 266                 uptime / HZ,
 267                 uptime % HZ,
 268                 idle / HZ,
 269                 idle % HZ);
 270 #endif
 271 }
 272 
 273 static int get_meminfo(char * buffer)
     /*  */
 274 {
 275         struct sysinfo i;
 276 
 277         si_meminfo(&i);
 278         si_swapinfo(&i);
 279         return sprintf(buffer, "        total:   used:    free:   shared:  buffers:\n"
 280                 "Mem:  %8lu %8lu %8lu %8lu %8lu\n"
 281                 "Swap: %8lu %8lu %8lu\n",
 282                 i.totalram, i.totalram-i.freeram, i.freeram, i.sharedram, i.bufferram,
 283                 i.totalswap, i.totalswap-i.freeswap, i.freeswap);
 284 }
 285 
 286 static int get_version(char * buffer)
     /*  */
 287 {
 288         extern char *linux_banner;
 289 
 290         strcpy(buffer, linux_banner);
 291         return strlen(buffer);
 292 }
 293 
 294 static int get_cmdline(char * buffer)
     /*  */
 295 {
 296         extern char saved_command_line[];
 297 
 298         return sprintf(buffer, "%s\n", saved_command_line);
 299 }
 300 
 301 static struct task_struct ** get_task(pid_t pid)
     /*  */
 302 {
 303         struct task_struct ** p;
 304 
 305         p = task;
 306         while (++p < task+NR_TASKS) {
 307                 if (*p && (*p)->pid == pid)
 308                         return p;
 309         }
 310         return NULL;
 311 }
 312 
 313 static unsigned long get_phys_addr(struct task_struct * p, unsigned long ptr)
     /*  */
 314 {
 315         pgd_t *page_dir;
 316         pmd_t *page_middle;
 317         pte_t pte;
 318 
 319         if (!p || !p->mm || ptr >= TASK_SIZE)
 320                 return 0;
 321         page_dir = pgd_offset(p->mm,ptr);
 322         if (pgd_none(*page_dir))
 323                 return 0;
 324         if (pgd_bad(*page_dir)) {
 325                 printk("bad page directory entry %08lx\n", pgd_val(*page_dir));
 326                 pgd_clear(page_dir);
 327                 return 0;
 328         }
 329         page_middle = pmd_offset(page_dir,ptr);
 330         if (pmd_none(*page_middle))
 331                 return 0;
 332         if (pmd_bad(*page_middle)) {
 333                 printk("bad page middle entry %08lx\n", pmd_val(*page_middle));
 334                 pmd_clear(page_middle);
 335                 return 0;
 336         }
 337         pte = *pte_offset(page_middle,ptr);
 338         if (!pte_present(pte))
 339                 return 0;
 340         return pte_page(pte) + (ptr & ~PAGE_MASK);
 341 }
 342 
 343 static int get_array(struct task_struct ** p, unsigned long start, unsigned long end, char * buffer)
     /*  */
 344 {
 345         unsigned long addr;
 346         int size = 0, result = 0;
 347         char c;
 348 
 349         if (start >= end)
 350                 return result;
 351         for (;;) {
 352                 addr = get_phys_addr(*p, start);
 353                 if (!addr)
 354                         goto ready;
 355                 do {
 356                         c = *(char *) addr;
 357                         if (!c)
 358                                 result = size;
 359                         if (size < PAGE_SIZE)
 360                                 buffer[size++] = c;
 361                         else
 362                                 goto ready;
 363                         addr++;
 364                         start++;
 365                         if (!c && start >= end)
 366                                 goto ready;
 367                 } while (addr & ~PAGE_MASK);
 368         }
 369 ready:
 370         /* remove the trailing blanks, used to fill out argv,envp space */
 371         while (result>0 && buffer[result-1]==' ')
 372                 result--;
 373         return result;
 374 }
 375 
 376 static int get_env(int pid, char * buffer)
     /*  */
 377 {
 378         struct task_struct ** p = get_task(pid);
 379 
 380         if (!p || !*p || !(*p)->mm)
 381                 return 0;
 382         return get_array(p, (*p)->mm->env_start, (*p)->mm->env_end, buffer);
 383 }
 384 
 385 static int get_arg(int pid, char * buffer)
     /*  */
 386 {
 387         struct task_struct ** p = get_task(pid);
 388 
 389         if (!p || !*p || !(*p)->mm)
 390                 return 0;
 391         return get_array(p, (*p)->mm->arg_start, (*p)->mm->arg_end, buffer);
 392 }
 393 
 394 static unsigned long get_wchan(struct task_struct *p)
     /*  */
 395 {
 396         if (!p || p == current || p->state == TASK_RUNNING)
 397                 return 0;
 398 #if defined(__i386__)
 399         {
 400                 unsigned long ebp, eip;
 401                 unsigned long stack_page;
 402                 int count = 0;
 403 
 404                 stack_page = p->kernel_stack_page;
 405                 if (!stack_page)
 406                         return 0;
 407                 ebp = p->tss.ebp;
 408                 do {
 409                         if (ebp < stack_page || ebp >= 4092+stack_page)
 410                                 return 0;
 411                         eip = *(unsigned long *) (ebp+4);
 412                         if ((void *)eip != sleep_on &&
 413                             (void *)eip != interruptible_sleep_on)
 414                                 return eip;
 415                         ebp = *(unsigned long *) ebp;
 416                 } while (count++ < 16);
 417         }
 418 #elif defined(__alpha__)
 419         /*
 420          * This one depends on the frame size of schedule().  Do a
 421          * "disass schedule" in gdb to find the frame size.  Also, the
 422          * code assumes that sleep_on() follows immediately after
 423          * interruptible_sleep_on() and that add_timer() follows
 424          * immediately after interruptible_sleep().  Ugly, isn't it?
 425          * Maybe adding a wchan field to task_struct would be better,
 426          * after all...
 427          */
 428         {
 429             unsigned long schedule_frame;
 430             unsigned long pc;
 431 
 432             pc = thread_saved_pc(&p->tss);
 433             if (pc >= (unsigned long) interruptible_sleep_on && pc < (unsigned long) add_timer) {
 434                 schedule_frame = ((unsigned long *)p->tss.ksp)[6];
 435                 return ((unsigned long *)schedule_frame)[12];
 436             }
 437             return pc;
 438         }
 439 #endif
 440         return 0;
 441 }
 442 
 443 #if defined(__i386__)
 444 # define KSTK_EIP(tsk)  (((unsigned long *)tsk->kernel_stack_page)[1019])
 445 # define KSTK_ESP(tsk)  (((unsigned long *)tsk->kernel_stack_page)[1022])
 446 #elif defined(__alpha__)
 447   /*
 448    * See arch/alpha/kernel/ptrace.c for details.
 449    */
 450 # define PT_REG(reg)            (PAGE_SIZE - sizeof(struct pt_regs)     \
 451                                  + (long)&((struct pt_regs *)0)->reg)
 452 # define KSTK_EIP(tsk)  (*(unsigned long *)(tsk->kernel_stack_page + PT_REG(pc)))
 453 # define KSTK_ESP(tsk)  ((tsk) == current ? rdusp() : (tsk)->tss.usp)
 454 #endif
 455 
 456 static int get_stat(int pid, char * buffer)
     /*  */
 457 {
 458         struct task_struct ** p = get_task(pid), *tsk;
 459         unsigned long sigignore=0, sigcatch=0, wchan;
 460         unsigned long vsize, eip, esp;
 461         long priority, nice;
 462         int i,tty_pgrp;
 463         char state;
 464 
 465         if (!p || (tsk = *p) == NULL)
 466                 return 0;
 467         if (tsk->state < 0 || tsk->state > 5)
 468                 state = '.';
 469         else
 470                 state = "RSDZTW"[tsk->state];
 471         vsize = eip = esp = 0;
 472         if (tsk->mm) {
 473                 struct vm_area_struct *vma = tsk->mm->mmap;
 474                 while (vma) {
 475                         vsize += vma->vm_end - vma->vm_start;
 476                         vma = vma->vm_next;
 477                 }
 478                 if (tsk->kernel_stack_page) {
 479                         eip = KSTK_EIP(tsk);
 480                         esp = KSTK_ESP(tsk);
 481                 }
 482         }
 483         wchan = get_wchan(tsk);
 484         if (tsk->sig) {
 485                 unsigned long bit = 1;
 486                 for(i=0; i<32; ++i) {
 487                         switch((unsigned long) tsk->sig->action[i].sa_handler) {
 488                                 case 0:
 489                                         break;
 490                                 case 1:
 491                                         sigignore |= bit;
 492                                         break;
 493                                 default:
 494                                         sigcatch |= bit;
 495                         }
 496                         bit <<= 1;
 497                 }
 498         }
 499         if (tsk->tty)
 500                 tty_pgrp = tsk->tty->pgrp;
 501         else
 502                 tty_pgrp = -1;
 503 
 504         /* scale priority and nice values from timeslices to -20..20 */
 505         /* to make it look like a "normal" unix priority/nice value  */
 506         priority = tsk->counter;
 507         priority = 20 - (priority * 10 + DEF_PRIORITY / 2) / DEF_PRIORITY;
 508         nice = tsk->priority;
 509         nice = 20 - (nice * 20 + DEF_PRIORITY / 2) / DEF_PRIORITY;
 510 
 511         return sprintf(buffer,"%d (%s) %c %d %d %d %d %d %lu %lu \
 512 %lu %lu %lu %ld %ld %ld %ld %ld %ld %lu %lu %ld %lu %lu %lu %lu %lu %lu %lu %lu %lu \
 513 %lu %lu %lu %lu\n",
 514                 pid,
 515                 tsk->comm,
 516                 state,
 517                 tsk->p_pptr->pid,
 518                 tsk->pgrp,
 519                 tsk->session,
 520                 tsk->tty ? kdev_t_to_nr(tsk->tty->device) : 0,
 521                 tty_pgrp,
 522                 tsk->flags,
 523                 tsk->min_flt,
 524                 tsk->cmin_flt,
 525                 tsk->maj_flt,
 526                 tsk->cmaj_flt,
 527                 tsk->utime,
 528                 tsk->stime,
 529                 tsk->cutime,
 530                 tsk->cstime,
 531                 priority,
 532                 nice,
 533                 tsk->timeout,
 534                 tsk->it_real_value,
 535                 tsk->start_time,
 536                 vsize,
 537                 tsk->mm ? tsk->mm->rss : 0, /* you might want to shift this left 3 */
 538                 tsk->rlim ? tsk->rlim[RLIMIT_RSS].rlim_cur : 0,
 539                 tsk->mm ? tsk->mm->start_code : 0,
 540                 tsk->mm ? tsk->mm->end_code : 0,
 541                 tsk->mm ? tsk->mm->start_stack : 0,
 542                 esp,
 543                 eip,
 544                 tsk->signal,
 545                 tsk->blocked,
 546                 sigignore,
 547                 sigcatch,
 548                 wchan);
 549 }
 550                 
 551 static inline void statm_pte_range(pmd_t * pmd, unsigned long address, unsigned long size,
     /*  */
 552         int * pages, int * shared, int * dirty, int * total)
 553 {
 554         pte_t * pte;
 555         unsigned long end;
 556 
 557         if (pmd_none(*pmd))
 558                 return;
 559         if (pmd_bad(*pmd)) {
 560                 printk("statm_pte_range: bad pmd (%08lx)\n", pmd_val(*pmd));
 561                 pmd_clear(pmd);
 562                 return;
 563         }
 564         pte = pte_offset(pmd, address);
 565         address &= ~PMD_MASK;
 566         end = address + size;
 567         if (end > PMD_SIZE)
 568                 end = PMD_SIZE;
 569         do {
 570                 pte_t page = *pte;
 571 
 572                 address += PAGE_SIZE;
 573                 pte++;
 574                 if (pte_none(page))
 575                         continue;
 576                 ++*total;
 577                 if (!pte_present(page))
 578                         continue;
 579                 ++*pages;
 580                 if (pte_dirty(page))
 581                         ++*dirty;
 582                 if (pte_page(page) >= high_memory)
 583                         continue;
 584                 if (mem_map[MAP_NR(pte_page(page))].count > 1)
 585                         ++*shared;
 586         } while (address < end);
 587 }
 588 
 589 static inline void statm_pmd_range(pgd_t * pgd, unsigned long address, unsigned long size,
     /*  */
 590         int * pages, int * shared, int * dirty, int * total)
 591 {
 592         pmd_t * pmd;
 593         unsigned long end;
 594 
 595         if (pgd_none(*pgd))
 596                 return;
 597         if (pgd_bad(*pgd)) {
 598                 printk("statm_pmd_range: bad pgd (%08lx)\n", pgd_val(*pgd));
 599                 pgd_clear(pgd);
 600                 return;
 601         }
 602         pmd = pmd_offset(pgd, address);
 603         address &= ~PGDIR_MASK;
 604         end = address + size;
 605         if (end > PGDIR_SIZE)
 606                 end = PGDIR_SIZE;
 607         do {
 608                 statm_pte_range(pmd, address, end - address, pages, shared, dirty, total);
 609                 address = (address + PMD_SIZE) & PMD_MASK;
 610                 pmd++;
 611         } while (address < end);
 612 }
 613 
 614 static void statm_pgd_range(pgd_t * pgd, unsigned long address, unsigned long end,
     /*  */
 615         int * pages, int * shared, int * dirty, int * total)
 616 {
 617         while (address < end) {
 618                 statm_pmd_range(pgd, address, end - address, pages, shared, dirty, total);
 619                 address = (address + PGDIR_SIZE) & PGDIR_MASK;
 620                 pgd++;
 621         }
 622 }
 623 
 624 static int get_statm(int pid, char * buffer)
     /*  */
 625 {
 626         struct task_struct ** p = get_task(pid), *tsk;
 627         int size=0, resident=0, share=0, trs=0, lrs=0, drs=0, dt=0;
 628 
 629         if (!p || (tsk = *p) == NULL)
 630                 return 0;
 631         if (tsk->mm) {
 632                 struct vm_area_struct * vma = tsk->mm->mmap;
 633 
 634                 while (vma) {
 635                         pgd_t *pgd = pgd_offset(tsk->mm, vma->vm_start);
 636                         int pages = 0, shared = 0, dirty = 0, total = 0;
 637 
 638                         statm_pgd_range(pgd, vma->vm_start, vma->vm_end, &pages, &shared, &dirty, &total);
 639                         resident += pages;
 640                         share += shared;
 641                         dt += dirty;
 642                         size += total;
 643                         if (vma->vm_flags & VM_EXECUTABLE)
 644                                 trs += pages;   /* text */
 645                         else if (vma->vm_flags & VM_GROWSDOWN)
 646                                 drs += pages;   /* stack */
 647                         else if (vma->vm_end > 0x60000000)
 648                                 lrs += pages;   /* library */
 649                         else
 650                                 drs += pages;
 651                         vma = vma->vm_next;
 652                 }
 653         }
 654         return sprintf(buffer,"%d %d %d %d %d %d %d\n",
 655                        size, resident, share, trs, lrs, drs, dt);
 656 }
 657 
 658 /*
 659  * The way we support synthetic files > 4K
 660  * - without storing their contents in some buffer and
 661  * - without walking through the entire synthetic file until we reach the
 662  *   position of the requested data
 663  * is to cleverly encode the current position in the file's f_pos field.
 664  * There is no requirement that a read() call which returns `count' bytes
 665  * of data increases f_pos by exactly `count'.
 666  *
 667  * This idea is Linus' one. Bruno implemented it.
 668  */
 669 
 670 /*
 671  * For the /proc/<pid>/maps file, we use fixed length records, each containing
 672  * a single line.
 673  */
 674 #define MAPS_LINE_LENGTH        1024
 675 #define MAPS_LINE_SHIFT         10
 676 /*
 677  * f_pos = (number of the vma in the task->mm->mmap list) * MAPS_LINE_LENGTH
 678  *         + (index into the line)
 679  */
 680 #define MAPS_LINE_FORMAT          "%08lx-%08lx %s %08lx %02x:%02x %lu\n"
 681 #define MAPS_LINE_MAX   49 /* sum of 8  1  8  1 4 1 8  1  2 1  2 1 10 1 */
 682 
 683 static int read_maps (int pid, struct file * file, char * buf, int count)
     /*  */
 684 {
 685         struct task_struct ** p = get_task(pid);
 686         char * destptr;
 687         loff_t lineno;
 688         int column;
 689         struct vm_area_struct * map;
 690         int i;
 691 
 692         if (!p || !*p)
 693                 return -EINVAL;
 694 
 695         if (!(*p)->mm || count == 0)
 696                 return 0;
 697 
 698         /* decode f_pos */
 699         lineno = file->f_pos >> MAPS_LINE_SHIFT;
 700         column = file->f_pos & (MAPS_LINE_LENGTH-1);
 701 
 702         /* quickly go to line lineno */
 703         for (map = (*p)->mm->mmap, i = 0; map && (i < lineno); map = map->vm_next, i++)
 704                 continue;
 705 
 706         destptr = buf;
 707 
 708         for ( ; map ; ) {
 709                 /* produce the next line */
 710                 char line[MAPS_LINE_MAX+1];
 711                 char str[5], *cp = str;
 712                 int flags;
 713                 kdev_t dev;
 714                 unsigned long ino;
 715                 int len;
 716 
 717                 flags = map->vm_flags;
 718 
 719                 *cp++ = flags & VM_READ ? 'r' : '-';
 720                 *cp++ = flags & VM_WRITE ? 'w' : '-';
 721                 *cp++ = flags & VM_EXEC ? 'x' : '-';
 722                 *cp++ = flags & VM_MAYSHARE ? 's' : 'p';
 723                 *cp++ = 0;
 724 
 725                 if (map->vm_inode != NULL) {
 726                         dev = map->vm_inode->i_dev;
 727                         ino = map->vm_inode->i_ino;
 728                 } else {
 729                         dev = 0;
 730                         ino = 0;
 731                 }
 732 
 733                 len = sprintf(line, MAPS_LINE_FORMAT,
 734                               map->vm_start, map->vm_end, str, map->vm_offset,
 735                               MAJOR(dev),MINOR(dev), ino);
 736 
 737                 if (column >= len) {
 738                         column = 0; /* continue with next line at column 0 */
 739                         lineno++;
 740                         map = map->vm_next;
 741                         continue;
 742                 }
 743 
 744                 i = len-column;
 745                 if (i > count)
 746                         i = count;
 747                 memcpy_tofs(destptr, line+column, i);
 748                 destptr += i; count -= i;
 749                 column += i;
 750                 if (column >= len) {
 751                         column = 0; /* next time: next line at column 0 */
 752                         lineno++;
 753                         map = map->vm_next;
 754                 }
 755 
 756                 /* done? */
 757                 if (count == 0)
 758                         break;
 759 
 760                 /* By writing to user space, we might have slept.
 761                  * Stop the loop, to avoid a race condition.
 762                  */
 763                 if (*p != current)
 764                         break;
 765         }
 766 
 767         /* encode f_pos */
 768         file->f_pos = (lineno << MAPS_LINE_SHIFT) + column;
 769 
 770         return destptr-buf;
 771 }
 772 
 773 extern int get_module_list(char *);
 774 extern int get_device_list(char *);
 775 extern int get_filesystem_list(char *);
 776 extern int get_ksyms_list(char *, char **, off_t, int);
 777 extern int get_irq_list(char *);
 778 extern int get_dma_list(char *);
 779 extern int get_cpuinfo(char *);
 780 extern int get_pci_list(char*);
 781 
 782 static int get_root_array(char * page, int type, char **start, off_t offset, int length)
     /*  */
 783 {
 784         switch (type) {
 785                 case PROC_LOADAVG:
 786                         return get_loadavg(page);
 787 
 788                 case PROC_UPTIME:
 789                         return get_uptime(page);
 790 
 791                 case PROC_MEMINFO:
 792                         return get_meminfo(page);
 793 
 794 #ifdef CONFIG_PCI
 795                 case PROC_PCI:
 796                         return get_pci_list(page);
 797 #endif
 798                         
 799                 case PROC_CPUINFO:
 800                         return get_cpuinfo(page);
 801 
 802                 case PROC_VERSION:
 803                         return get_version(page);
 804 
 805 #ifdef CONFIG_DEBUG_MALLOC
 806                 case PROC_MALLOC:
 807                         return get_malloc(page);
 808 #endif
 809 
 810                 case PROC_MODULES:
 811                         return get_module_list(page);
 812 
 813                 case PROC_STAT:
 814                         return get_kstat(page);
 815 
 816                 case PROC_DEVICES:
 817                         return get_device_list(page);
 818 
 819                 case PROC_INTERRUPTS:
 820                         return get_irq_list(page);
 821 
 822                 case PROC_FILESYSTEMS:
 823                         return get_filesystem_list(page);
 824 
 825                 case PROC_KSYMS:
 826                         return get_ksyms_list(page, start, offset, length);
 827 
 828                 case PROC_DMA:
 829                         return get_dma_list(page);
 830 
 831                 case PROC_IOPORTS:
 832                         return get_ioport_list(page);
 833 #ifdef CONFIG_APM
 834                 case PROC_APM:
 835                         return apm_proc(page);
 836 #endif
 837                 case PROC_CMDLINE:
 838                         return get_cmdline(page);
 839         }
 840         return -EBADF;
 841 }
 842 
 843 static int get_process_array(char * page, int pid, int type)
     /*  */
 844 {
 845         switch (type) {
 846                 case PROC_PID_ENVIRON:
 847                         return get_env(pid, page);
 848                 case PROC_PID_CMDLINE:
 849                         return get_arg(pid, page);
 850                 case PROC_PID_STAT:
 851                         return get_stat(pid, page);
 852                 case PROC_PID_STATM:
 853                         return get_statm(pid, page);
 854         }
 855         return -EBADF;
 856 }
 857 
 858 
 859 static inline int fill_array(char * page, int pid, int type, char **start, off_t offset, int length)
     /*  */
 860 {
 861         if (pid)
 862                 return get_process_array(page, pid, type);
 863         return get_root_array(page, type, start, offset, length);
 864 }
 865 
 866 #define PROC_BLOCK_SIZE (3*1024)                /* 4K page size but our output routines use some slack for overruns */
 867 
 868 static int array_read(struct inode * inode, struct file * file,char * buf, int count)
     /*  */
 869 {
 870         unsigned long page;
 871         char *start;
 872         int length;
 873         int end;
 874         unsigned int type, pid;
 875 
 876         if (count < 0)
 877                 return -EINVAL;
 878         if (count > PROC_BLOCK_SIZE)
 879                 count = PROC_BLOCK_SIZE;
 880         if (!(page = __get_free_page(GFP_KERNEL)))
 881                 return -ENOMEM;
 882         type = inode->i_ino;
 883         pid = type >> 16;
 884         type &= 0x0000ffff;
 885         start = NULL;
 886         length = fill_array((char *) page, pid, type,
 887                             &start, file->f_pos, count);
 888         if (length < 0) {
 889                 free_page(page);
 890                 return length;
 891         }
 892         if (start != NULL) {
 893                 /* We have had block-adjusting processing! */
 894                 memcpy_tofs(buf, start, length);
 895                 file->f_pos += length;
 896                 count = length;
 897         } else {
 898                 /* Static 4kB (or whatever) block capacity */
 899                 if (file->f_pos >= length) {
 900                         free_page(page);
 901                         return 0;
 902                 }
 903                 if (count + file->f_pos > length)
 904                         count = length - file->f_pos;
 905                 end = count + file->f_pos;
 906                 memcpy_tofs(buf, (char *) page + file->f_pos, count);
 907                 file->f_pos = end;
 908         }
 909         free_page(page);
 910         return count;
 911 }
 912 
 913 static struct file_operations proc_array_operations = {
 914         NULL,           /* array_lseek */
 915         array_read,
 916         NULL,           /* array_write */
 917         NULL,           /* array_readdir */
 918         NULL,           /* array_select */
 919         NULL,           /* array_ioctl */
 920         NULL,           /* mmap */
 921         NULL,           /* no special open code */
 922         NULL,           /* no special release code */
 923         NULL            /* can't fsync */
 924 };
 925 
 926 struct inode_operations proc_array_inode_operations = {
 927         &proc_array_operations, /* default base directory file-ops */
 928         NULL,                   /* create */
 929         NULL,                   /* lookup */
 930         NULL,                   /* link */
 931         NULL,                   /* unlink */
 932         NULL,                   /* symlink */
 933         NULL,                   /* mkdir */
 934         NULL,                   /* rmdir */
 935         NULL,                   /* mknod */
 936         NULL,                   /* rename */
 937         NULL,                   /* readlink */
 938         NULL,                   /* follow_link */
 939         NULL,                   /* readpage */
 940         NULL,                   /* writepage */
 941         NULL,                   /* bmap */
 942         NULL,                   /* truncate */
 943         NULL                    /* permission */
 944 };
 945 
 946 static int arraylong_read (struct inode * inode, struct file * file, char * buf, int count)
     /*  */
 947 {
 948         unsigned int pid = inode->i_ino >> 16;
 949         unsigned int type = inode->i_ino & 0x0000ffff;
 950 
 951         if (count < 0)
 952                 return -EINVAL;
 953 
 954         switch (type) {
 955                 case PROC_PID_MAPS:
 956                         return read_maps(pid, file, buf, count);
 957         }
 958         return -EINVAL;
 959 }
 960 
 961 static struct file_operations proc_arraylong_operations = {
 962         NULL,           /* array_lseek */
 963         arraylong_read,
 964         NULL,           /* array_write */
 965         NULL,           /* array_readdir */
 966         NULL,           /* array_select */
 967         NULL,           /* array_ioctl */
 968         NULL,           /* mmap */
 969         NULL,           /* no special open code */
 970         NULL,           /* no special release code */
 971         NULL            /* can't fsync */
 972 };
 973 
 974 struct inode_operations proc_arraylong_inode_operations = {
 975         &proc_arraylong_operations,     /* default base directory file-ops */
 976         NULL,                   /* create */
 977         NULL,                   /* lookup */
 978         NULL,                   /* link */
 979         NULL,                   /* unlink */
 980         NULL,                   /* symlink */
 981         NULL,                   /* mkdir */
 982         NULL,                   /* rmdir */
 983         NULL,                   /* mknod */
 984         NULL,                   /* rename */
 985         NULL,                   /* readlink */
 986         NULL,                   /* follow_link */
 987         NULL,                   /* readpage */
 988         NULL,                   /* writepage */
 989         NULL,                   /* bmap */
 990         NULL,                   /* truncate */
 991         NULL                    /* permission */
 992 };
/* */
root/fs/proc/array.c

DEFINITIONS
