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 #include <linux/pagemap.h>
  45 #ifdef CONFIG_APM
  46 #include <linux/apm_bios.h>
  47 #endif
  48 #include <linux/swap.h>
  49 
  50 #include <asm/segment.h>
  51 #include <asm/pgtable.h>
  52 #include <asm/io.h>
  53 
  54 #define LOAD_INT(x) ((x) >> FSHIFT)
  55 #define LOAD_FRAC(x) LOAD_INT(((x) & (FIXED_1-1)) * 100)
  56 
  57 #ifdef CONFIG_DEBUG_MALLOC
  58 int get_malloc(char * buffer);
  59 #endif
  60 
  61 
  62 static int read_core(struct inode * inode, struct file * file,char * buf, int count)
     /*  */
  63 {
  64         unsigned long p = file->f_pos, memsize;
  65         int read;
  66         int count1;
  67         char * pnt;
  68         struct user dump;
  69 #ifdef __i386__
  70 #       define FIRST_MAPPED     PAGE_SIZE       /* we don't have page 0 mapped on x86.. */
  71 #else
  72 #       define FIRST_MAPPED     0
  73 #endif
  74 
  75         memset(&dump, 0, sizeof(struct user));
  76         dump.magic = CMAGIC;
  77         dump.u_dsize = MAP_NR(high_memory);
  78 #ifdef __alpha__
  79         dump.start_data = PAGE_OFFSET;
  80 #endif
  81 
  82         if (count < 0)
  83                 return -EINVAL;
  84         memsize = MAP_NR(high_memory + PAGE_SIZE) << PAGE_SHIFT;
  85         if (p >= memsize)
  86                 return 0;
  87         if (count > memsize - p)
  88                 count = memsize - p;
  89         read = 0;
  90 
  91         if (p < sizeof(struct user) && count > 0) {
  92                 count1 = count;
  93                 if (p + count1 > sizeof(struct user))
  94                         count1 = sizeof(struct user)-p;
  95                 pnt = (char *) &dump + p;
  96                 memcpy_tofs(buf,(void *) pnt, count1);
  97                 buf += count1;
  98                 p += count1;
  99                 count -= count1;
 100                 read += count1;
 101         }
 102 
 103         while (count > 0 && p < PAGE_SIZE + FIRST_MAPPED) {
 104                 put_user(0,buf);
 105                 buf++;
 106                 p++;
 107                 count--;
 108                 read++;
 109         }
 110         memcpy_tofs(buf, (void *) (PAGE_OFFSET + p - PAGE_SIZE), count);
 111         read += count;
 112         file->f_pos += read;
 113         return read;
 114 }
 115 
 116 static struct file_operations proc_kcore_operations = {
 117         NULL,           /* lseek */
 118         read_core,
 119 };
 120 
 121 struct inode_operations proc_kcore_inode_operations = {
 122         &proc_kcore_operations, 
 123 };
 124 
 125 
 126 extern unsigned long prof_len;
 127 extern unsigned long * prof_buffer;
 128 extern unsigned long prof_shift;
 129 /*
 130  * This function accesses profiling information. The returned data is
 131  * binary: the sampling step and the actual contents of the profile
 132  * buffer. Use of the program readprofile is recommended in order to
 133  * get meaningful info out of these data.
 134  */
 135 static int read_profile(struct inode *inode, struct file *file, char *buf, int count)
     /*  */
 136 {
 137         unsigned long p = file->f_pos;
 138         int read;
 139         char * pnt;
 140         unsigned long sample_step = 1 << prof_shift;
 141 
 142         if (count < 0)
 143                 return -EINVAL;
 144         if (p >= (prof_len+1)*sizeof(unsigned long))
 145                 return 0;
 146         if (count > (prof_len+1)*sizeof(unsigned long) - p)
 147                 count = (prof_len+1)*sizeof(unsigned long) - p;
 148         read = 0;
 149 
 150         while (p < sizeof(unsigned long) && count > 0) {
 151                 put_user(*((char *)(&sample_step)+p),buf);
 152                 buf++; p++; count--; read++;
 153         }
 154         pnt = (char *)prof_buffer + p - sizeof(unsigned long);
 155         memcpy_tofs(buf,(void *)pnt,count);
 156         read += count;
 157         file->f_pos += read;
 158         return read;
 159 }
 160 
 161 /* Writing to /proc/profile resets the counters */
 162 static int write_profile(struct inode * inode, struct file * file, const char * buf, int count)
     /*  */
 163 {
 164     int i=prof_len;
 165 
 166     while (i--)
 167             prof_buffer[i]=0UL;
 168     return count;
 169 }
 170 
 171 static struct file_operations proc_profile_operations = {
 172         NULL,           /* lseek */
 173         read_profile,
 174         write_profile,
 175 };
 176 
 177 struct inode_operations proc_profile_inode_operations = {
 178         &proc_profile_operations, 
 179 };
 180 
 181 
 182 static int get_loadavg(char * buffer)
     /*  */
 183 {
 184         int a, b, c;
 185 
 186         a = avenrun[0] + (FIXED_1/200);
 187         b = avenrun[1] + (FIXED_1/200);
 188         c = avenrun[2] + (FIXED_1/200);
 189         return sprintf(buffer,"%d.%02d %d.%02d %d.%02d %d/%d\n",
 190                 LOAD_INT(a), LOAD_FRAC(a),
 191                 LOAD_INT(b), LOAD_FRAC(b),
 192                 LOAD_INT(c), LOAD_FRAC(c),
 193                 nr_running, nr_tasks);
 194 }
 195 
 196 static int get_kstat(char * buffer)
     /*  */
 197 {
 198         int i, len;
 199         unsigned sum = 0;
 200 
 201         for (i = 0 ; i < NR_IRQS ; i++)
 202                 sum += kstat.interrupts[i];
 203         len = sprintf(buffer,
 204                 "cpu  %u %u %u %lu\n"
 205                 "disk %u %u %u %u\n"
 206                 "disk_rio %u %u %u %u\n"
 207                 "disk_wio %u %u %u %u\n"
 208                 "disk_rblk %u %u %u %u\n"
 209                 "disk_wblk %u %u %u %u\n"
 210                 "page %u %u\n"
 211                 "swap %u %u\n"
 212                 "intr %u",
 213                 kstat.cpu_user,
 214                 kstat.cpu_nice,
 215                 kstat.cpu_system,
 216                 jiffies - (kstat.cpu_user + kstat.cpu_nice + kstat.cpu_system),
 217                 kstat.dk_drive[0], kstat.dk_drive[1],
 218                 kstat.dk_drive[2], kstat.dk_drive[3],
 219                 kstat.dk_drive_rio[0], kstat.dk_drive_rio[1],
 220                 kstat.dk_drive_rio[2], kstat.dk_drive_rio[3],
 221                 kstat.dk_drive_wio[0], kstat.dk_drive_wio[1],
 222                 kstat.dk_drive_wio[2], kstat.dk_drive_wio[3],
 223                 kstat.dk_drive_rblk[0], kstat.dk_drive_rblk[1],
 224                 kstat.dk_drive_rblk[2], kstat.dk_drive_rblk[3],
 225                 kstat.dk_drive_wblk[0], kstat.dk_drive_wblk[1],
 226                 kstat.dk_drive_wblk[2], kstat.dk_drive_wblk[3],
 227                 kstat.pgpgin,
 228                 kstat.pgpgout,
 229                 kstat.pswpin,
 230                 kstat.pswpout,
 231                 sum);
 232         for (i = 0 ; i < NR_IRQS ; i++)
 233                 len += sprintf(buffer + len, " %u", kstat.interrupts[i]);
 234         len += sprintf(buffer + len,
 235                 "\nctxt %u\n"
 236                 "btime %lu\n",
 237                 kstat.context_swtch,
 238                 xtime.tv_sec - jiffies / HZ);
 239         return len;
 240 }
 241 
 242 
 243 static int get_uptime(char * buffer)
     /*  */
 244 {
 245         unsigned long uptime;
 246         unsigned long idle;
 247 
 248         uptime = jiffies;
 249         idle = task[0]->utime + task[0]->stime;
 250 
 251         /* The formula for the fraction parts really is ((t * 100) / HZ) % 100, but
 252            that would overflow about every five days at HZ == 100.
 253            Therefore the identity a = (a / b) * b + a % b is used so that it is
 254            calculated as (((t / HZ) * 100) + ((t % HZ) * 100) / HZ) % 100.
 255            The part in front of the '+' always evaluates as 0 (mod 100). All divisions
 256            in the above formulas are truncating. For HZ being a power of 10, the
 257            calculations simplify to the version in the #else part (if the printf
 258            format is adapted to the same number of digits as zeroes in HZ.
 259          */
 260 #if HZ!=100
 261         return sprintf(buffer,"%lu.%02lu %lu.%02lu\n",
 262                 uptime / HZ,
 263                 (((uptime % HZ) * 100) / HZ) % 100,
 264                 idle / HZ,
 265                 (((idle % HZ) * 100) / HZ) % 100);
 266 #else
 267         return sprintf(buffer,"%lu.%02lu %lu.%02lu\n",
 268                 uptime / HZ,
 269                 uptime % HZ,
 270                 idle / HZ,
 271                 idle % HZ);
 272 #endif
 273 }
 274 
 275 static int get_meminfo(char * buffer)
     /*  */
 276 {
 277         struct sysinfo i;
 278 
 279         si_meminfo(&i);
 280         si_swapinfo(&i);
 281         return sprintf(buffer, "        total:    used:    free:  shared: buffers:  cached:\n"
 282                 "Mem:  %8lu %8lu %8lu %8lu %8lu %8lu\n"
 283                 "Swap: %8lu %8lu %8lu\n",
 284                 i.totalram, i.totalram-i.freeram, i.freeram, i.sharedram, i.bufferram, page_cache_size*PAGE_SIZE,
 285                 i.totalswap, i.totalswap-i.freeswap, i.freeswap);
 286 }
 287 
 288 static int get_version(char * buffer)
     /*  */
 289 {
 290         extern char *linux_banner;
 291 
 292         strcpy(buffer, linux_banner);
 293         return strlen(buffer);
 294 }
 295 
 296 static int get_cmdline(char * buffer)
     /*  */
 297 {
 298         extern char saved_command_line[];
 299 
 300         return sprintf(buffer, "%s\n", saved_command_line);
 301 }
 302 
 303 static struct task_struct ** get_task(pid_t pid)
     /*  */
 304 {
 305         struct task_struct ** p;
 306 
 307         p = task;
 308         while (++p < task+NR_TASKS) {
 309                 if (*p && (*p)->pid == pid)
 310                         return p;
 311         }
 312         return NULL;
 313 }
 314 
 315 static unsigned long get_phys_addr(struct task_struct * p, unsigned long ptr)
     /*  */
 316 {
 317         pgd_t *page_dir;
 318         pmd_t *page_middle;
 319         pte_t pte;
 320 
 321         if (!p || !p->mm || ptr >= TASK_SIZE)
 322                 return 0;
 323         page_dir = pgd_offset(p->mm,ptr);
 324         if (pgd_none(*page_dir))
 325                 return 0;
 326         if (pgd_bad(*page_dir)) {
 327                 printk("bad page directory entry %08lx\n", pgd_val(*page_dir));
 328                 pgd_clear(page_dir);
 329                 return 0;
 330         }
 331         page_middle = pmd_offset(page_dir,ptr);
 332         if (pmd_none(*page_middle))
 333                 return 0;
 334         if (pmd_bad(*page_middle)) {
 335                 printk("bad page middle entry %08lx\n", pmd_val(*page_middle));
 336                 pmd_clear(page_middle);
 337                 return 0;
 338         }
 339         pte = *pte_offset(page_middle,ptr);
 340         if (!pte_present(pte))
 341                 return 0;
 342         return pte_page(pte) + (ptr & ~PAGE_MASK);
 343 }
 344 
 345 static int get_array(struct task_struct ** p, unsigned long start, unsigned long end, char * buffer)
     /*  */
 346 {
 347         unsigned long addr;
 348         int size = 0, result = 0;
 349         char c;
 350 
 351         if (start >= end)
 352                 return result;
 353         for (;;) {
 354                 addr = get_phys_addr(*p, start);
 355                 if (!addr)
 356                         goto ready;
 357                 do {
 358                         c = *(char *) addr;
 359                         if (!c)
 360                                 result = size;
 361                         if (size < PAGE_SIZE)
 362                                 buffer[size++] = c;
 363                         else
 364                                 goto ready;
 365                         addr++;
 366                         start++;
 367                         if (!c && start >= end)
 368                                 goto ready;
 369                 } while (addr & ~PAGE_MASK);
 370         }
 371 ready:
 372         /* remove the trailing blanks, used to fill out argv,envp space */
 373         while (result>0 && buffer[result-1]==' ')
 374                 result--;
 375         return result;
 376 }
 377 
 378 static int get_env(int pid, char * buffer)
     /*  */
 379 {
 380         struct task_struct ** p = get_task(pid);
 381 
 382         if (!p || !*p || !(*p)->mm)
 383                 return 0;
 384         return get_array(p, (*p)->mm->env_start, (*p)->mm->env_end, buffer);
 385 }
 386 
 387 static int get_arg(int pid, char * buffer)
     /*  */
 388 {
 389         struct task_struct ** p = get_task(pid);
 390 
 391         if (!p || !*p || !(*p)->mm)
 392                 return 0;
 393         return get_array(p, (*p)->mm->arg_start, (*p)->mm->arg_end, buffer);
 394 }
 395 
 396 static unsigned long get_wchan(struct task_struct *p)
     /*  */
 397 {
 398         if (!p || p == current || p->state == TASK_RUNNING)
 399                 return 0;
 400 #if defined(__i386__)
 401         {
 402                 unsigned long ebp, eip;
 403                 unsigned long stack_page;
 404                 int count = 0;
 405 
 406                 stack_page = p->kernel_stack_page;
 407                 if (!stack_page)
 408                         return 0;
 409                 ebp = p->tss.ebp;
 410                 do {
 411                         if (ebp < stack_page || ebp >= 4092+stack_page)
 412                                 return 0;
 413                         eip = *(unsigned long *) (ebp+4);
 414                         if ((void *)eip != sleep_on &&
 415                             (void *)eip != interruptible_sleep_on)
 416                                 return eip;
 417                         ebp = *(unsigned long *) ebp;
 418                 } while (count++ < 16);
 419         }
 420 #elif defined(__alpha__)
 421         /*
 422          * This one depends on the frame size of schedule().  Do a
 423          * "disass schedule" in gdb to find the frame size.  Also, the
 424          * code assumes that sleep_on() follows immediately after
 425          * interruptible_sleep_on() and that add_timer() follows
 426          * immediately after interruptible_sleep().  Ugly, isn't it?
 427          * Maybe adding a wchan field to task_struct would be better,
 428          * after all...
 429          */
 430         {
 431             unsigned long schedule_frame;
 432             unsigned long pc;
 433 
 434             pc = thread_saved_pc(&p->tss);
 435             if (pc >= (unsigned long) interruptible_sleep_on && pc < (unsigned long) add_timer) {
 436                 schedule_frame = ((unsigned long *)p->tss.ksp)[6];
 437                 return ((unsigned long *)schedule_frame)[12];
 438             }
 439             return pc;
 440         }
 441 #endif
 442         return 0;
 443 }
 444 
 445 #if defined(__i386__)
 446 # define KSTK_EIP(tsk)  (((unsigned long *)tsk->kernel_stack_page)[1019])
 447 # define KSTK_ESP(tsk)  (((unsigned long *)tsk->kernel_stack_page)[1022])
 448 #elif defined(__alpha__)
 449   /*
 450    * See arch/alpha/kernel/ptrace.c for details.
 451    */
 452 # define PT_REG(reg)            (PAGE_SIZE - sizeof(struct pt_regs)     \
 453                                  + (long)&((struct pt_regs *)0)->reg)
 454 # define KSTK_EIP(tsk)  (*(unsigned long *)(tsk->kernel_stack_page + PT_REG(pc)))
 455 # define KSTK_ESP(tsk)  ((tsk) == current ? rdusp() : (tsk)->tss.usp)
 456 #endif
 457 
 458 static int get_stat(int pid, char * buffer)
     /*  */
 459 {
 460         struct task_struct ** p = get_task(pid), *tsk;
 461         unsigned long sigignore=0, sigcatch=0, wchan;
 462         unsigned long vsize, eip, esp;
 463         long priority, nice;
 464         int i,tty_pgrp;
 465         char state;
 466 
 467         if (!p || (tsk = *p) == NULL)
 468                 return 0;
 469         if (tsk->state < 0 || tsk->state > 5)
 470                 state = '.';
 471         else
 472                 state = "RSDZTW"[tsk->state];
 473         vsize = eip = esp = 0;
 474         if (tsk->mm && tsk->mm != &init_mm) {
 475                 struct vm_area_struct *vma = tsk->mm->mmap;
 476                 while (vma) {
 477                         vsize += vma->vm_end - vma->vm_start;
 478                         vma = vma->vm_next;
 479                 }
 480                 if (tsk->kernel_stack_page) {
 481                         eip = KSTK_EIP(tsk);
 482                         esp = KSTK_ESP(tsk);
 483                 }
 484         }
 485         wchan = get_wchan(tsk);
 486         if (tsk->sig) {
 487                 unsigned long bit = 1;
 488                 for(i=0; i<32; ++i) {
 489                         switch((unsigned long) tsk->sig->action[i].sa_handler) {
 490                                 case 0:
 491                                         break;
 492                                 case 1:
 493                                         sigignore |= bit;
 494                                         break;
 495                                 default:
 496                                         sigcatch |= bit;
 497                         }
 498                         bit <<= 1;
 499                 }
 500         }
 501         if (tsk->tty)
 502                 tty_pgrp = tsk->tty->pgrp;
 503         else
 504                 tty_pgrp = -1;
 505 
 506         /* scale priority and nice values from timeslices to -20..20 */
 507         /* to make it look like a "normal" unix priority/nice value  */
 508         priority = tsk->counter;
 509         priority = 20 - (priority * 10 + DEF_PRIORITY / 2) / DEF_PRIORITY;
 510         nice = tsk->priority;
 511         nice = 20 - (nice * 20 + DEF_PRIORITY / 2) / DEF_PRIORITY;
 512 
 513         return sprintf(buffer,"%d (%s) %c %d %d %d %d %d %lu %lu \
 514 %lu %lu %lu %ld %ld %ld %ld %ld %ld %lu %lu %ld %lu %lu %lu %lu %lu %lu %lu %lu %lu \
 515 %lu %lu %lu %lu\n",
 516                 pid,
 517                 tsk->comm,
 518                 state,
 519                 tsk->p_pptr->pid,
 520                 tsk->pgrp,
 521                 tsk->session,
 522                 tsk->tty ? kdev_t_to_nr(tsk->tty->device) : 0,
 523                 tty_pgrp,
 524                 tsk->flags,
 525                 tsk->min_flt,
 526                 tsk->cmin_flt,
 527                 tsk->maj_flt,
 528                 tsk->cmaj_flt,
 529                 tsk->utime,
 530                 tsk->stime,
 531                 tsk->cutime,
 532                 tsk->cstime,
 533                 priority,
 534                 nice,
 535                 tsk->timeout,
 536                 tsk->it_real_value,
 537                 tsk->start_time,
 538                 vsize,
 539                 tsk->mm ? tsk->mm->rss : 0, /* you might want to shift this left 3 */
 540                 tsk->rlim ? tsk->rlim[RLIMIT_RSS].rlim_cur : 0,
 541                 tsk->mm ? tsk->mm->start_code : 0,
 542                 tsk->mm ? tsk->mm->end_code : 0,
 543                 tsk->mm ? tsk->mm->start_stack : 0,
 544                 esp,
 545                 eip,
 546                 tsk->signal,
 547                 tsk->blocked,
 548                 sigignore,
 549                 sigcatch,
 550                 wchan);
 551 }
 552                 
 553 static inline void statm_pte_range(pmd_t * pmd, unsigned long address, unsigned long size,
     /*  */
 554         int * pages, int * shared, int * dirty, int * total)
 555 {
 556         pte_t * pte;
 557         unsigned long end;
 558 
 559         if (pmd_none(*pmd))
 560                 return;
 561         if (pmd_bad(*pmd)) {
 562                 printk("statm_pte_range: bad pmd (%08lx)\n", pmd_val(*pmd));
 563                 pmd_clear(pmd);
 564                 return;
 565         }
 566         pte = pte_offset(pmd, address);
 567         address &= ~PMD_MASK;
 568         end = address + size;
 569         if (end > PMD_SIZE)
 570                 end = PMD_SIZE;
 571         do {
 572                 pte_t page = *pte;
 573 
 574                 address += PAGE_SIZE;
 575                 pte++;
 576                 if (pte_none(page))
 577                         continue;
 578                 ++*total;
 579                 if (!pte_present(page))
 580                         continue;
 581                 ++*pages;
 582                 if (pte_dirty(page))
 583                         ++*dirty;
 584                 if (pte_page(page) >= high_memory)
 585                         continue;
 586                 if (mem_map[MAP_NR(pte_page(page))].count > 1)
 587                         ++*shared;
 588         } while (address < end);
 589 }
 590 
 591 static inline void statm_pmd_range(pgd_t * pgd, unsigned long address, unsigned long size,
     /*  */
 592         int * pages, int * shared, int * dirty, int * total)
 593 {
 594         pmd_t * pmd;
 595         unsigned long end;
 596 
 597         if (pgd_none(*pgd))
 598                 return;
 599         if (pgd_bad(*pgd)) {
 600                 printk("statm_pmd_range: bad pgd (%08lx)\n", pgd_val(*pgd));
 601                 pgd_clear(pgd);
 602                 return;
 603         }
 604         pmd = pmd_offset(pgd, address);
 605         address &= ~PGDIR_MASK;
 606         end = address + size;
 607         if (end > PGDIR_SIZE)
 608                 end = PGDIR_SIZE;
 609         do {
 610                 statm_pte_range(pmd, address, end - address, pages, shared, dirty, total);
 611                 address = (address + PMD_SIZE) & PMD_MASK;
 612                 pmd++;
 613         } while (address < end);
 614 }
 615 
 616 static void statm_pgd_range(pgd_t * pgd, unsigned long address, unsigned long end,
     /*  */
 617         int * pages, int * shared, int * dirty, int * total)
 618 {
 619         while (address < end) {
 620                 statm_pmd_range(pgd, address, end - address, pages, shared, dirty, total);
 621                 address = (address + PGDIR_SIZE) & PGDIR_MASK;
 622                 pgd++;
 623         }
 624 }
 625 
 626 static int get_statm(int pid, char * buffer)
     /*  */
 627 {
 628         struct task_struct ** p = get_task(pid), *tsk;
 629         int size=0, resident=0, share=0, trs=0, lrs=0, drs=0, dt=0;
 630 
 631         if (!p || (tsk = *p) == NULL)
 632                 return 0;
 633         if (tsk->mm && tsk->mm != &init_mm) {
 634                 struct vm_area_struct * vma = tsk->mm->mmap;
 635 
 636                 while (vma) {
 637                         pgd_t *pgd = pgd_offset(tsk->mm, vma->vm_start);
 638                         int pages = 0, shared = 0, dirty = 0, total = 0;
 639 
 640                         statm_pgd_range(pgd, vma->vm_start, vma->vm_end, &pages, &shared, &dirty, &total);
 641                         resident += pages;
 642                         share += shared;
 643                         dt += dirty;
 644                         size += total;
 645                         if (vma->vm_flags & VM_EXECUTABLE)
 646                                 trs += pages;   /* text */
 647                         else if (vma->vm_flags & VM_GROWSDOWN)
 648                                 drs += pages;   /* stack */
 649                         else if (vma->vm_end > 0x60000000)
 650                                 lrs += pages;   /* library */
 651                         else
 652                                 drs += pages;
 653                         vma = vma->vm_next;
 654                 }
 655         }
 656         return sprintf(buffer,"%d %d %d %d %d %d %d\n",
 657                        size, resident, share, trs, lrs, drs, dt);
 658 }
 659 
 660 /*
 661  * The way we support synthetic files > 4K
 662  * - without storing their contents in some buffer and
 663  * - without walking through the entire synthetic file until we reach the
 664  *   position of the requested data
 665  * is to cleverly encode the current position in the file's f_pos field.
 666  * There is no requirement that a read() call which returns `count' bytes
 667  * of data increases f_pos by exactly `count'.
 668  *
 669  * This idea is Linus' one. Bruno implemented it.
 670  */
 671 
 672 /*
 673  * For the /proc/<pid>/maps file, we use fixed length records, each containing
 674  * a single line.
 675  */
 676 #define MAPS_LINE_LENGTH        1024
 677 #define MAPS_LINE_SHIFT         10
 678 /*
 679  * f_pos = (number of the vma in the task->mm->mmap list) * MAPS_LINE_LENGTH
 680  *         + (index into the line)
 681  */
 682 #define MAPS_LINE_FORMAT          "%08lx-%08lx %s %08lx %02x:%02x %lu\n"
 683 #define MAPS_LINE_MAX   49 /* sum of 8  1  8  1 4 1 8  1  2 1  2 1 10 1 */
 684 
 685 static int read_maps (int pid, struct file * file, char * buf, int count)
     /*  */
 686 {
 687         struct task_struct ** p = get_task(pid);
 688         char * destptr;
 689         loff_t lineno;
 690         int column;
 691         struct vm_area_struct * map;
 692         int i;
 693 
 694         if (!p || !*p)
 695                 return -EINVAL;
 696 
 697         if (!(*p)->mm || (*p)->mm == &init_mm || count == 0)
 698                 return 0;
 699 
 700         /* decode f_pos */
 701         lineno = file->f_pos >> MAPS_LINE_SHIFT;
 702         column = file->f_pos & (MAPS_LINE_LENGTH-1);
 703 
 704         /* quickly go to line lineno */
 705         for (map = (*p)->mm->mmap, i = 0; map && (i < lineno); map = map->vm_next, i++)
 706                 continue;
 707 
 708         destptr = buf;
 709 
 710         for ( ; map ; ) {
 711                 /* produce the next line */
 712                 char line[MAPS_LINE_MAX+1];
 713                 char str[5], *cp = str;
 714                 int flags;
 715                 kdev_t dev;
 716                 unsigned long ino;
 717                 int len;
 718 
 719                 flags = map->vm_flags;
 720 
 721                 *cp++ = flags & VM_READ ? 'r' : '-';
 722                 *cp++ = flags & VM_WRITE ? 'w' : '-';
 723                 *cp++ = flags & VM_EXEC ? 'x' : '-';
 724                 *cp++ = flags & VM_MAYSHARE ? 's' : 'p';
 725                 *cp++ = 0;
 726 
 727                 if (map->vm_inode != NULL) {
 728                         dev = map->vm_inode->i_dev;
 729                         ino = map->vm_inode->i_ino;
 730                 } else {
 731                         dev = 0;
 732                         ino = 0;
 733                 }
 734 
 735                 len = sprintf(line, MAPS_LINE_FORMAT,
 736                               map->vm_start, map->vm_end, str, map->vm_offset,
 737                               MAJOR(dev),MINOR(dev), ino);
 738 
 739                 if (column >= len) {
 740                         column = 0; /* continue with next line at column 0 */
 741                         lineno++;
 742                         map = map->vm_next;
 743                         continue;
 744                 }
 745 
 746                 i = len-column;
 747                 if (i > count)
 748                         i = count;
 749                 memcpy_tofs(destptr, line+column, i);
 750                 destptr += i; count -= i;
 751                 column += i;
 752                 if (column >= len) {
 753                         column = 0; /* next time: next line at column 0 */
 754                         lineno++;
 755                         map = map->vm_next;
 756                 }
 757 
 758                 /* done? */
 759                 if (count == 0)
 760                         break;
 761 
 762                 /* By writing to user space, we might have slept.
 763                  * Stop the loop, to avoid a race condition.
 764                  */
 765                 if (*p != current)
 766                         break;
 767         }
 768 
 769         /* encode f_pos */
 770         file->f_pos = (lineno << MAPS_LINE_SHIFT) + column;
 771 
 772         return destptr-buf;
 773 }
 774 
 775 #ifdef CONFIG_MODULES
 776 extern int get_module_list(char *);
 777 extern int get_ksyms_list(char *, char **, off_t, int);
 778 #endif
 779 extern int get_device_list(char *);
 780 extern int get_filesystem_list(char *);
 781 extern int get_irq_list(char *);
 782 extern int get_dma_list(char *);
 783 extern int get_cpuinfo(char *);
 784 extern int get_pci_list(char*);
 785 #ifdef __SMP_PROF__
 786 extern int get_smp_prof_list(char *);
 787 #endif
 788 
 789 static int get_root_array(char * page, int type, char **start, off_t offset, int length)
     /*  */
 790 {
 791         switch (type) {
 792                 case PROC_LOADAVG:
 793                         return get_loadavg(page);
 794 
 795                 case PROC_UPTIME:
 796                         return get_uptime(page);
 797 
 798                 case PROC_MEMINFO:
 799                         return get_meminfo(page);
 800 
 801 #ifdef CONFIG_PCI
 802                 case PROC_PCI:
 803                         return get_pci_list(page);
 804 #endif
 805                         
 806                 case PROC_CPUINFO:
 807                         return get_cpuinfo(page);
 808 
 809                 case PROC_VERSION:
 810                         return get_version(page);
 811 
 812 #ifdef CONFIG_DEBUG_MALLOC
 813                 case PROC_MALLOC:
 814                         return get_malloc(page);
 815 #endif
 816 
 817 #ifdef CONFIG_MODULES
 818                 case PROC_MODULES:
 819                         return get_module_list(page);
 820 
 821                 case PROC_KSYMS:
 822                         return get_ksyms_list(page, start, offset, length);
 823 #endif
 824 
 825                 case PROC_STAT:
 826                         return get_kstat(page);
 827 
 828                 case PROC_DEVICES:
 829                         return get_device_list(page);
 830 
 831                 case PROC_INTERRUPTS:
 832                         return get_irq_list(page);
 833 
 834                 case PROC_FILESYSTEMS:
 835                         return get_filesystem_list(page);
 836 
 837                 case PROC_DMA:
 838                         return get_dma_list(page);
 839 
 840                 case PROC_IOPORTS:
 841                         return get_ioport_list(page);
 842 #ifdef CONFIG_APM
 843                 case PROC_APM:
 844                         return apm_proc(page);
 845 #endif
 846 #ifdef __SMP_PROF__
 847                 case PROC_SMP_PROF:
 848                         return get_smp_prof_list(page);
 849 #endif
 850                 case PROC_CMDLINE:
 851                         return get_cmdline(page);
 852         }
 853         return -EBADF;
 854 }
 855 
 856 static int get_process_array(char * page, int pid, int type)
     /*  */
 857 {
 858         switch (type) {
 859                 case PROC_PID_ENVIRON:
 860                         return get_env(pid, page);
 861                 case PROC_PID_CMDLINE:
 862                         return get_arg(pid, page);
 863                 case PROC_PID_STAT:
 864                         return get_stat(pid, page);
 865                 case PROC_PID_STATM:
 866                         return get_statm(pid, page);
 867         }
 868         return -EBADF;
 869 }
 870 
 871 
 872 static inline int fill_array(char * page, int pid, int type, char **start, off_t offset, int length)
     /*  */
 873 {
 874         if (pid)
 875                 return get_process_array(page, pid, type);
 876         return get_root_array(page, type, start, offset, length);
 877 }
 878 
 879 #define PROC_BLOCK_SIZE (3*1024)                /* 4K page size but our output routines use some slack for overruns */
 880 
 881 static int array_read(struct inode * inode, struct file * file,char * buf, int count)
     /*  */
 882 {
 883         unsigned long page;
 884         char *start;
 885         int length;
 886         int end;
 887         unsigned int type, pid;
 888 
 889         if (count < 0)
 890                 return -EINVAL;
 891         if (count > PROC_BLOCK_SIZE)
 892                 count = PROC_BLOCK_SIZE;
 893         if (!(page = __get_free_page(GFP_KERNEL)))
 894                 return -ENOMEM;
 895         type = inode->i_ino;
 896         pid = type >> 16;
 897         type &= 0x0000ffff;
 898         start = NULL;
 899         length = fill_array((char *) page, pid, type,
 900                             &start, file->f_pos, count);
 901         if (length < 0) {
 902                 free_page(page);
 903                 return length;
 904         }
 905         if (start != NULL) {
 906                 /* We have had block-adjusting processing! */
 907                 memcpy_tofs(buf, start, length);
 908                 file->f_pos += length;
 909                 count = length;
 910         } else {
 911                 /* Static 4kB (or whatever) block capacity */
 912                 if (file->f_pos >= length) {
 913                         free_page(page);
 914                         return 0;
 915                 }
 916                 if (count + file->f_pos > length)
 917                         count = length - file->f_pos;
 918                 end = count + file->f_pos;
 919                 memcpy_tofs(buf, (char *) page + file->f_pos, count);
 920                 file->f_pos = end;
 921         }
 922         free_page(page);
 923         return count;
 924 }
 925 
 926 static struct file_operations proc_array_operations = {
 927         NULL,           /* array_lseek */
 928         array_read,
 929         NULL,           /* array_write */
 930         NULL,           /* array_readdir */
 931         NULL,           /* array_select */
 932         NULL,           /* array_ioctl */
 933         NULL,           /* mmap */
 934         NULL,           /* no special open code */
 935         NULL,           /* no special release code */
 936         NULL            /* can't fsync */
 937 };
 938 
 939 struct inode_operations proc_array_inode_operations = {
 940         &proc_array_operations, /* default base directory file-ops */
 941         NULL,                   /* create */
 942         NULL,                   /* lookup */
 943         NULL,                   /* link */
 944         NULL,                   /* unlink */
 945         NULL,                   /* symlink */
 946         NULL,                   /* mkdir */
 947         NULL,                   /* rmdir */
 948         NULL,                   /* mknod */
 949         NULL,                   /* rename */
 950         NULL,                   /* readlink */
 951         NULL,                   /* follow_link */
 952         NULL,                   /* readpage */
 953         NULL,                   /* writepage */
 954         NULL,                   /* bmap */
 955         NULL,                   /* truncate */
 956         NULL                    /* permission */
 957 };
 958 
 959 static int arraylong_read (struct inode * inode, struct file * file, char * buf, int count)
     /*  */
 960 {
 961         unsigned int pid = inode->i_ino >> 16;
 962         unsigned int type = inode->i_ino & 0x0000ffff;
 963 
 964         if (count < 0)
 965                 return -EINVAL;
 966 
 967         switch (type) {
 968                 case PROC_PID_MAPS:
 969                         return read_maps(pid, file, buf, count);
 970         }
 971         return -EINVAL;
 972 }
 973 
 974 static struct file_operations proc_arraylong_operations = {
 975         NULL,           /* array_lseek */
 976         arraylong_read,
 977         NULL,           /* array_write */
 978         NULL,           /* array_readdir */
 979         NULL,           /* array_select */
 980         NULL,           /* array_ioctl */
 981         NULL,           /* mmap */
 982         NULL,           /* no special open code */
 983         NULL,           /* no special release code */
 984         NULL            /* can't fsync */
 985 };
 986 
 987 struct inode_operations proc_arraylong_inode_operations = {
 988         &proc_arraylong_operations,     /* default base directory file-ops */
 989         NULL,                   /* create */
 990         NULL,                   /* lookup */
 991         NULL,                   /* link */
 992         NULL,                   /* unlink */
 993         NULL,                   /* symlink */
 994         NULL,                   /* mkdir */
 995         NULL,                   /* rmdir */
 996         NULL,                   /* mknod */
 997         NULL,                   /* rename */
 998         NULL,                   /* readlink */
 999         NULL,                   /* follow_link */
1000         NULL,                   /* readpage */
1001         NULL,                   /* writepage */
1002         NULL,                   /* bmap */
1003         NULL,                   /* truncate */
1004         NULL                    /* permission */
1005 };
/* */
root/fs/proc/array.c

DEFINITIONS
