fs/proc/base.c

   1 /*
   2  *  linux/fs/proc/base.c
   3  *
   4  *  Copyright (C) 1991, 1992 Linus Torvalds
   5  *
   6  *  proc base directory handling functions
   7  *
   8  *  1999, Al Viro. Rewritten. Now it covers the whole per-process part.
   9  *  Instead of using magical inumbers to determine the kind of object
  10  *  we allocate and fill in-core inodes upon lookup. They don't even
  11  *  go into icache. We cache the reference to task_struct upon lookup too.
  12  *  Eventually it should become a filesystem in its own. We don't use the
  13  *  rest of procfs anymore.
  14  *
  15  *
  16  *  Changelog:
  17  *  17-Jan-2005
  18  *  Allan Bezerra
  19  *  Bruna Moreira <bruna.moreira@indt.org.br>
  20  *  Edjard Mota <edjard.mota@indt.org.br>
  21  *  Ilias Biris <ilias.biris@indt.org.br>
  22  *  Mauricio Lin <mauricio.lin@indt.org.br>
  23  *
  24  *  Embedded Linux Lab - 10LE Instituto Nokia de Tecnologia - INdT
  25  *
  26  *  A new process specific entry (smaps) included in /proc. It shows the
  27  *  size of rss for each memory area. The maps entry lacks information
  28  *  about physical memory size (rss) for each mapped file, i.e.,
  29  *  rss information for executables and library files.
  30  *  This additional information is useful for any tools that need to know
  31  *  about physical memory consumption for a process specific library.
  32  *
  33  *  Changelog:
  34  *  21-Feb-2005
  35  *  Embedded Linux Lab - 10LE Instituto Nokia de Tecnologia - INdT
  36  *  Pud inclusion in the page table walking.
  37  *
  38  *  ChangeLog:
  39  *  10-Mar-2005
  40  *  10LE Instituto Nokia de Tecnologia - INdT:
  41  *  A better way to walks through the page table as suggested by Hugh Dickins.
  42  *
  43  *  Simo Piiroinen <simo.piiroinen@nokia.com>:
  44  *  Smaps information related to shared, private, clean and dirty pages.
  45  *
  46  *  Paul Mundt <paul.mundt@nokia.com>:
  47  *  Overall revision about smaps.
  48  */
  49
  50 #include <asm/uaccess.h>
  51
  52 #include <linux/config.h>
  53 #include <linux/errno.h>
  54 #include <linux/time.h>
  55 #include <linux/proc_fs.h>
  56 #include <linux/stat.h>
  57 #include <linux/init.h>
  58 #include <linux/file.h>
  59 #include <linux/string.h>
  60 #include <linux/seq_file.h>
  61 #include <linux/namei.h>
  62 #include <linux/namespace.h>
  63 #include <linux/mm.h>
  64 #include <linux/smp_lock.h>
  65 #include <linux/rcupdate.h>
  66 #include <linux/kallsyms.h>
  67 #include <linux/mount.h>
  68 #include <linux/security.h>
  69 #include <linux/ptrace.h>
  70 #include <linux/seccomp.h>
  71 #include <linux/cpuset.h>
  72 #include <linux/audit.h>
  73 #include "internal.h"
  74
  75 /*
  76  * For hysterical raisins we keep the same inumbers as in the old procfs.
  77  * Feel free to change the macro below - just keep the range distinct from
  78  * inumbers of the rest of procfs (currently those are in 0x0000--0xffff).
  79  * As soon as we'll get a separate superblock we will be able to forget
  80  * about magical ranges too.
  81  */
  82
  83 #define fake_ino(pid,ino) (((pid)<<16)|(ino))
  84
  85 enum pid_directory_inos {
  86         PROC_TGID_INO = 2,
  87         PROC_TGID_TASK,
  88         PROC_TGID_STATUS,
  89         PROC_TGID_MEM,
  90 #ifdef CONFIG_SECCOMP
  91         PROC_TGID_SECCOMP,
  92 #endif
  93         PROC_TGID_CWD,
  94         PROC_TGID_ROOT,
  95         PROC_TGID_EXE,
  96         PROC_TGID_FD,
  97         PROC_TGID_ENVIRON,
  98         PROC_TGID_AUXV,
  99         PROC_TGID_CMDLINE,
 100         PROC_TGID_STAT,
 101         PROC_TGID_STATM,
 102         PROC_TGID_MAPS,
 103         PROC_TGID_NUMA_MAPS,
 104         PROC_TGID_MOUNTS,
 105         PROC_TGID_WCHAN,
 106         PROC_TGID_SMAPS,
 107 #ifdef CONFIG_SCHEDSTATS
 108         PROC_TGID_SCHEDSTAT,
 109 #endif
 110 #ifdef CONFIG_CPUSETS
 111         PROC_TGID_CPUSET,
 112 #endif
 113 #ifdef CONFIG_SECURITY
 114         PROC_TGID_ATTR,
 115         PROC_TGID_ATTR_CURRENT,
 116         PROC_TGID_ATTR_PREV,
 117         PROC_TGID_ATTR_EXEC,
 118         PROC_TGID_ATTR_FSCREATE,
 119 #endif
 120 #ifdef CONFIG_AUDITSYSCALL
 121         PROC_TGID_LOGINUID,
 122 #endif
 123         PROC_TGID_OOM_SCORE,
 124         PROC_TGID_OOM_ADJUST,
 125         PROC_TID_INO,
 126         PROC_TID_STATUS,
 127         PROC_TID_MEM,
 128 #ifdef CONFIG_SECCOMP
 129         PROC_TID_SECCOMP,
 130 #endif
 131         PROC_TID_CWD,
 132         PROC_TID_ROOT,
 133         PROC_TID_EXE,
 134         PROC_TID_FD,
 135         PROC_TID_ENVIRON,
 136         PROC_TID_AUXV,
 137         PROC_TID_CMDLINE,
 138         PROC_TID_STAT,
 139         PROC_TID_STATM,
 140         PROC_TID_MAPS,
 141         PROC_TID_NUMA_MAPS,
 142         PROC_TID_MOUNTS,
 143         PROC_TID_WCHAN,
 144         PROC_TID_SMAPS,
 145 #ifdef CONFIG_SCHEDSTATS
 146         PROC_TID_SCHEDSTAT,
 147 #endif
 148 #ifdef CONFIG_CPUSETS
 149         PROC_TID_CPUSET,
 150 #endif
 151 #ifdef CONFIG_SECURITY
 152         PROC_TID_ATTR,
 153         PROC_TID_ATTR_CURRENT,
 154         PROC_TID_ATTR_PREV,
 155         PROC_TID_ATTR_EXEC,
 156         PROC_TID_ATTR_FSCREATE,
 157 #endif
 158 #ifdef CONFIG_AUDITSYSCALL
 159         PROC_TID_LOGINUID,
 160 #endif
 161         PROC_TID_OOM_SCORE,
 162         PROC_TID_OOM_ADJUST,
 163
 164         /* Add new entries before this */
 165         PROC_TID_FD_DIR = 0x8000,       /* 0x8000-0xffff */
 166 };
 167
 168 struct pid_entry {
 169         int type;
 170         int len;
 171         char *name;
 172         mode_t mode;
 173 };
 174
 175 #define E(type,name,mode) {(type),sizeof(name)-1,(name),(mode)}
 176
 177 static struct pid_entry tgid_base_stuff[] = {
 178         E(PROC_TGID_TASK,      "task",    S_IFDIR|S_IRUGO|S_IXUGO),
 179         E(PROC_TGID_FD,        "fd",      S_IFDIR|S_IRUSR|S_IXUSR),
 180         E(PROC_TGID_ENVIRON,   "environ", S_IFREG|S_IRUSR),
 181         E(PROC_TGID_AUXV,      "auxv",    S_IFREG|S_IRUSR),
 182         E(PROC_TGID_STATUS,    "status",  S_IFREG|S_IRUGO),
 183         E(PROC_TGID_CMDLINE,   "cmdline", S_IFREG|S_IRUGO),
 184         E(PROC_TGID_STAT,      "stat",    S_IFREG|S_IRUGO),
 185         E(PROC_TGID_STATM,     "statm",   S_IFREG|S_IRUGO),
 186         E(PROC_TGID_MAPS,      "maps",    S_IFREG|S_IRUGO),
 187 #ifdef CONFIG_NUMA
 188         E(PROC_TGID_NUMA_MAPS, "numa_maps", S_IFREG|S_IRUGO),
 189 #endif
 190         E(PROC_TGID_MEM,       "mem",     S_IFREG|S_IRUSR|S_IWUSR),
 191 #ifdef CONFIG_SECCOMP
 192         E(PROC_TGID_SECCOMP,   "seccomp", S_IFREG|S_IRUSR|S_IWUSR),
 193 #endif
 194         E(PROC_TGID_CWD,       "cwd",     S_IFLNK|S_IRWXUGO),
 195         E(PROC_TGID_ROOT,      "root",    S_IFLNK|S_IRWXUGO),
 196         E(PROC_TGID_EXE,       "exe",     S_IFLNK|S_IRWXUGO),
 197         E(PROC_TGID_MOUNTS,    "mounts",  S_IFREG|S_IRUGO),
 198         E(PROC_TGID_SMAPS,     "smaps",   S_IFREG|S_IRUGO),
 199 #ifdef CONFIG_SECURITY
 200         E(PROC_TGID_ATTR,      "attr",    S_IFDIR|S_IRUGO|S_IXUGO),
 201 #endif
 202 #ifdef CONFIG_KALLSYMS
 203         E(PROC_TGID_WCHAN,     "wchan",   S_IFREG|S_IRUGO),
 204 #endif
 205 #ifdef CONFIG_SCHEDSTATS
 206         E(PROC_TGID_SCHEDSTAT, "schedstat", S_IFREG|S_IRUGO),
 207 #endif
 208 #ifdef CONFIG_CPUSETS
 209         E(PROC_TGID_CPUSET,    "cpuset",  S_IFREG|S_IRUGO),
 210 #endif
 211         E(PROC_TGID_OOM_SCORE, "oom_score",S_IFREG|S_IRUGO),
 212         E(PROC_TGID_OOM_ADJUST,"oom_adj", S_IFREG|S_IRUGO|S_IWUSR),
 213 #ifdef CONFIG_AUDITSYSCALL
 214         E(PROC_TGID_LOGINUID, "loginuid", S_IFREG|S_IWUSR|S_IRUGO),
 215 #endif
 216         {0,0,NULL,0}
 217 };
 218 static struct pid_entry tid_base_stuff[] = {
 219         E(PROC_TID_FD,         "fd",      S_IFDIR|S_IRUSR|S_IXUSR),
 220         E(PROC_TID_ENVIRON,    "environ", S_IFREG|S_IRUSR),
 221         E(PROC_TID_AUXV,       "auxv",    S_IFREG|S_IRUSR),
 222         E(PROC_TID_STATUS,     "status",  S_IFREG|S_IRUGO),
 223         E(PROC_TID_CMDLINE,    "cmdline", S_IFREG|S_IRUGO),
 224         E(PROC_TID_STAT,       "stat",    S_IFREG|S_IRUGO),
 225         E(PROC_TID_STATM,      "statm",   S_IFREG|S_IRUGO),
 226         E(PROC_TID_MAPS,       "maps",    S_IFREG|S_IRUGO),
 227 #ifdef CONFIG_NUMA
 228         E(PROC_TID_NUMA_MAPS,  "numa_maps",    S_IFREG|S_IRUGO),
 229 #endif
 230         E(PROC_TID_MEM,        "mem",     S_IFREG|S_IRUSR|S_IWUSR),
 231 #ifdef CONFIG_SECCOMP
 232         E(PROC_TID_SECCOMP,    "seccomp", S_IFREG|S_IRUSR|S_IWUSR),
 233 #endif
 234         E(PROC_TID_CWD,        "cwd",     S_IFLNK|S_IRWXUGO),
 235         E(PROC_TID_ROOT,       "root",    S_IFLNK|S_IRWXUGO),
 236         E(PROC_TID_EXE,        "exe",     S_IFLNK|S_IRWXUGO),
 237         E(PROC_TID_MOUNTS,     "mounts",  S_IFREG|S_IRUGO),
 238         E(PROC_TID_SMAPS,      "smaps",   S_IFREG|S_IRUGO),
 239 #ifdef CONFIG_SECURITY
 240         E(PROC_TID_ATTR,       "attr",    S_IFDIR|S_IRUGO|S_IXUGO),
 241 #endif
 242 #ifdef CONFIG_KALLSYMS
 243         E(PROC_TID_WCHAN,      "wchan",   S_IFREG|S_IRUGO),
 244 #endif
 245 #ifdef CONFIG_SCHEDSTATS
 246         E(PROC_TID_SCHEDSTAT, "schedstat",S_IFREG|S_IRUGO),
 247 #endif
 248 #ifdef CONFIG_CPUSETS
 249         E(PROC_TID_CPUSET,     "cpuset",  S_IFREG|S_IRUGO),
 250 #endif
 251         E(PROC_TID_OOM_SCORE,  "oom_score",S_IFREG|S_IRUGO),
 252         E(PROC_TID_OOM_ADJUST, "oom_adj", S_IFREG|S_IRUGO|S_IWUSR),
 253 #ifdef CONFIG_AUDITSYSCALL
 254         E(PROC_TID_LOGINUID, "loginuid", S_IFREG|S_IWUSR|S_IRUGO),
 255 #endif
 256         {0,0,NULL,0}
 257 };
 258
 259 #ifdef CONFIG_SECURITY
 260 static struct pid_entry tgid_attr_stuff[] = {
 261         E(PROC_TGID_ATTR_CURRENT,  "current",  S_IFREG|S_IRUGO|S_IWUGO),
 262         E(PROC_TGID_ATTR_PREV,     "prev",     S_IFREG|S_IRUGO),
 263         E(PROC_TGID_ATTR_EXEC,     "exec",     S_IFREG|S_IRUGO|S_IWUGO),
 264         E(PROC_TGID_ATTR_FSCREATE, "fscreate", S_IFREG|S_IRUGO|S_IWUGO),
 265         {0,0,NULL,0}
 266 };
 267 static struct pid_entry tid_attr_stuff[] = {
 268         E(PROC_TID_ATTR_CURRENT,   "current",  S_IFREG|S_IRUGO|S_IWUGO),
 269         E(PROC_TID_ATTR_PREV,      "prev",     S_IFREG|S_IRUGO),
 270         E(PROC_TID_ATTR_EXEC,      "exec",     S_IFREG|S_IRUGO|S_IWUGO),
 271         E(PROC_TID_ATTR_FSCREATE,  "fscreate", S_IFREG|S_IRUGO|S_IWUGO),
 272         {0,0,NULL,0}
 273 };
 274 #endif
 275
 276 #undef E
 277
 278 static int proc_fd_link(struct inode *inode, struct dentry **dentry, struct vfsmount **mnt)
 279 {
 280         struct task_struct *task = proc_task(inode);
 281         struct files_struct *files;
 282         struct file *file;
 283         int fd = proc_type(inode) - PROC_TID_FD_DIR;
 284
 285         files = get_files_struct(task);
 286         if (files) {
 287                 rcu_read_lock();
 288                 file = fcheck_files(files, fd);
 289                 if (file) {
 290                         *mnt = mntget(file->f_vfsmnt);
 291                         *dentry = dget(file->f_dentry);
 292                         rcu_read_unlock();
 293                         put_files_struct(files);
 294                         return 0;
 295                 }
 296                 rcu_read_unlock();
 297                 put_files_struct(files);
 298         }
 299         return -ENOENT;
 300 }
 301
 302 static struct fs_struct *get_fs_struct(struct task_struct *task)
 303 {
 304         struct fs_struct *fs;
 305         task_lock(task);
 306         fs = task->fs;
 307         if(fs)
 308                 atomic_inc(&fs->count);
 309         task_unlock(task);
 310         return fs;
 311 }
 312
 313 static int proc_cwd_link(struct inode *inode, struct dentry **dentry, struct vfsmount **mnt)
 314 {
 315         struct fs_struct *fs = get_fs_struct(proc_task(inode));
 316         int result = -ENOENT;
 317         if (fs) {
 318                 read_lock(&fs->lock);
 319                 *mnt = mntget(fs->pwdmnt);
 320                 *dentry = dget(fs->pwd);
 321                 read_unlock(&fs->lock);
 322                 result = 0;
 323                 put_fs_struct(fs);
 324         }
 325         return result;
 326 }
 327
 328 static int proc_root_link(struct inode *inode, struct dentry **dentry, struct vfsmount **mnt)
 329 {
 330         struct fs_struct *fs = get_fs_struct(proc_task(inode));
 331         int result = -ENOENT;
 332         if (fs) {
 333                 read_lock(&fs->lock);
 334                 *mnt = mntget(fs->rootmnt);
 335                 *dentry = dget(fs->root);
 336                 read_unlock(&fs->lock);
 337                 result = 0;
 338                 put_fs_struct(fs);
 339         }
 340         return result;
 341 }
 342
 343
 344 /* Same as proc_root_link, but this addionally tries to get fs from other
 345  * threads in the group */
 346 static int proc_task_root_link(struct inode *inode, struct dentry **dentry,
 347                                 struct vfsmount **mnt)
 348 {
 349         struct fs_struct *fs;
 350         int result = -ENOENT;
 351         struct task_struct *leader = proc_task(inode);
 352
 353         task_lock(leader);
 354         fs = leader->fs;
 355         if (fs) {
 356                 atomic_inc(&fs->count);
 357                 task_unlock(leader);
 358         } else {
 359                 /* Try to get fs from other threads */
 360                 task_unlock(leader);
 361                 read_lock(&tasklist_lock);
 362                 if (pid_alive(leader)) {
 363                         struct task_struct *task = leader;
 364
 365                         while ((task = next_thread(task)) != leader) {
 366                                 task_lock(task);
 367                                 fs = task->fs;
 368                                 if (fs) {
 369                                         atomic_inc(&fs->count);
 370                                         task_unlock(task);
 371                                         break;
 372                                 }
 373                                 task_unlock(task);
 374                         }
 375                 }
 376                 read_unlock(&tasklist_lock);
 377         }
 378
 379         if (fs) {
 380                 read_lock(&fs->lock);
 381                 *mnt = mntget(fs->rootmnt);
 382                 *dentry = dget(fs->root);
 383                 read_unlock(&fs->lock);
 384                 result = 0;
 385                 put_fs_struct(fs);
 386         }
 387         return result;
 388 }
 389
 390
 391 #define MAY_PTRACE(task) \
 392         (task == current || \
 393         (task->parent == current && \
 394         (task->ptrace & PT_PTRACED) && \
 395          (task->state == TASK_STOPPED || task->state == TASK_TRACED) && \
 396          security_ptrace(current,task) == 0))
 397
 398 static int proc_pid_environ(struct task_struct *task, char * buffer)
 399 {
 400         int res = 0;
 401         struct mm_struct *mm = get_task_mm(task);
 402         if (mm) {
 403                 unsigned int len = mm->env_end - mm->env_start;
 404                 if (len > PAGE_SIZE)
 405                         len = PAGE_SIZE;
 406                 res = access_process_vm(task, mm->env_start, buffer, len, 0);
 407                 if (!ptrace_may_attach(task))
 408                         res = -ESRCH;
 409                 mmput(mm);
 410         }
 411         return res;
 412 }
 413
 414 static int proc_pid_cmdline(struct task_struct *task, char * buffer)
 415 {
 416         int res = 0;
 417         unsigned int len;
 418         struct mm_struct *mm = get_task_mm(task);
 419         if (!mm)
 420                 goto out;
 421         if (!mm->arg_end)
 422                 goto out_mm;    /* Shh! No looking before we're done */
 423
 424         len = mm->arg_end - mm->arg_start;
 425
 426         if (len > PAGE_SIZE)
 427                 len = PAGE_SIZE;
 428
 429         res = access_process_vm(task, mm->arg_start, buffer, len, 0);
 430
 431         // If the nul at the end of args has been overwritten, then
 432         // assume application is using setproctitle(3).
 433         if (res > 0 && buffer[res-1] != '\0' && len < PAGE_SIZE) {
 434                 len = strnlen(buffer, res);
 435                 if (len < res) {
 436                     res = len;
 437                 } else {
 438                         len = mm->env_end - mm->env_start;
 439                         if (len > PAGE_SIZE - res)
 440                                 len = PAGE_SIZE - res;
 441                         res += access_process_vm(task, mm->env_start, buffer+res, len, 0);
 442                         res = strnlen(buffer, res);
 443                 }
 444         }
 445 out_mm:
 446         mmput(mm);
 447 out:
 448         return res;
 449 }
 450
 451 static int proc_pid_auxv(struct task_struct *task, char *buffer)
 452 {
 453         int res = 0;
 454         struct mm_struct *mm = get_task_mm(task);
 455         if (mm) {
 456                 unsigned int nwords = 0;
 457                 do
 458                         nwords += 2;
 459                 while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */
 460                 res = nwords * sizeof(mm->saved_auxv[0]);
 461                 if (res > PAGE_SIZE)
 462                         res = PAGE_SIZE;
 463                 memcpy(buffer, mm->saved_auxv, res);
 464                 mmput(mm);
 465         }
 466         return res;
 467 }
 468
 469
 470 #ifdef CONFIG_KALLSYMS
 471 /*
 472  * Provides a wchan file via kallsyms in a proper one-value-per-file format.
 473  * Returns the resolved symbol.  If that fails, simply return the address.
 474  */
 475 static int proc_pid_wchan(struct task_struct *task, char *buffer)
 476 {
 477         char *modname;
 478         const char *sym_name;
 479         unsigned long wchan, size, offset;
 480         char namebuf[KSYM_NAME_LEN+1];
 481
 482         wchan = get_wchan(task);
 483
 484         sym_name = kallsyms_lookup(wchan, &size, &offset, &modname, namebuf);
 485         if (sym_name)
 486                 return sprintf(buffer, "%s", sym_name);
 487         return sprintf(buffer, "%lu", wchan);
 488 }
 489 #endif /* CONFIG_KALLSYMS */
 490
 491 #ifdef CONFIG_SCHEDSTATS
 492 /*
 493  * Provides /proc/PID/schedstat
 494  */
 495 static int proc_pid_schedstat(struct task_struct *task, char *buffer)
 496 {
 497         return sprintf(buffer, "%lu %lu %lu\n",
 498                         task->sched_info.cpu_time,
 499                         task->sched_info.run_delay,
 500                         task->sched_info.pcnt);
 501 }
 502 #endif
 503
 504 /* The badness from the OOM killer */
 505 unsigned long badness(struct task_struct *p, unsigned long uptime);
 506 static int proc_oom_score(struct task_struct *task, char *buffer)
 507 {
 508         unsigned long points;
 509         struct timespec uptime;
 510
 511         do_posix_clock_monotonic_gettime(&uptime);
 512         points = badness(task, uptime.tv_sec);
 513         return sprintf(buffer, "%lu\n", points);
 514 }
 515
 516 /************************************************************************/
 517 /*                       Here the fs part begins                        */
 518 /************************************************************************/
 519
 520 /* permission checks */
 521
 522 /* If the process being read is separated by chroot from the reading process,
 523  * don't let the reader access the threads.
 524  */
 525 static int proc_check_chroot(struct dentry *root, struct vfsmount *vfsmnt)
 526 {
 527         struct dentry *de, *base;
 528         struct vfsmount *our_vfsmnt, *mnt;
 529         int res = 0;
 530         read_lock(&current->fs->lock);
 531         our_vfsmnt = mntget(current->fs->rootmnt);
 532         base = dget(current->fs->root);
 533         read_unlock(&current->fs->lock);
 534
 535         spin_lock(&vfsmount_lock);
 536         de = root;
 537         mnt = vfsmnt;
 538
 539         while (vfsmnt != our_vfsmnt) {
 540                 if (vfsmnt == vfsmnt->mnt_parent)
 541                         goto out;
 542                 de = vfsmnt->mnt_mountpoint;
 543                 vfsmnt = vfsmnt->mnt_parent;
 544         }
 545
 546         if (!is_subdir(de, base))
 547                 goto out;
 548         spin_unlock(&vfsmount_lock);
 549
 550 exit:
 551         dput(base);
 552         mntput(our_vfsmnt);
 553         dput(root);
 554         mntput(mnt);
 555         return res;
 556 out:
 557         spin_unlock(&vfsmount_lock);
 558         res = -EACCES;
 559         goto exit;
 560 }
 561
 562 static int proc_check_root(struct inode *inode)
 563 {
 564         struct dentry *root;
 565         struct vfsmount *vfsmnt;
 566
 567         if (proc_root_link(inode, &root, &vfsmnt)) /* Ewww... */
 568                 return -ENOENT;
 569         return proc_check_chroot(root, vfsmnt);
 570 }
 571
 572 static int proc_permission(struct inode *inode, int mask, struct nameidata *nd)
 573 {
 574         if (generic_permission(inode, mask, NULL) != 0)
 575                 return -EACCES;
 576         return proc_check_root(inode);
 577 }
 578
 579 static int proc_task_permission(struct inode *inode, int mask, struct nameidata *nd)
 580 {
 581         struct dentry *root;
 582         struct vfsmount *vfsmnt;
 583
 584         if (generic_permission(inode, mask, NULL) != 0)
 585                 return -EACCES;
 586
 587         if (proc_task_root_link(inode, &root, &vfsmnt))
 588                 return -ENOENT;
 589
 590         return proc_check_chroot(root, vfsmnt);
 591 }
 592
 593 extern struct seq_operations proc_pid_maps_op;
 594 static int maps_open(struct inode *inode, struct file *file)
 595 {
 596         struct task_struct *task = proc_task(inode);
 597         int ret = seq_open(file, &proc_pid_maps_op);
 598         if (!ret) {
 599                 struct seq_file *m = file->private_data;
 600                 m->private = task;
 601         }
 602         return ret;
 603 }
 604
 605 static struct file_operations proc_maps_operations = {
 606         .open           = maps_open,
 607         .read           = seq_read,
 608         .llseek         = seq_lseek,
 609         .release        = seq_release,
 610 };
 611
 612 #ifdef CONFIG_NUMA
 613 extern struct seq_operations proc_pid_numa_maps_op;
 614 static int numa_maps_open(struct inode *inode, struct file *file)
 615 {
 616         struct task_struct *task = proc_task(inode);
 617         int ret = seq_open(file, &proc_pid_numa_maps_op);
 618         if (!ret) {
 619                 struct seq_file *m = file->private_data;
 620                 m->private = task;
 621         }
 622         return ret;
 623 }
 624
 625 static struct file_operations proc_numa_maps_operations = {
 626         .open           = numa_maps_open,
 627         .read           = seq_read,
 628         .llseek         = seq_lseek,
 629         .release        = seq_release,
 630 };
 631 #endif
 632
 633 extern struct seq_operations proc_pid_smaps_op;
 634 static int smaps_open(struct inode *inode, struct file *file)
 635 {
 636         struct task_struct *task = proc_task(inode);
 637         int ret = seq_open(file, &proc_pid_smaps_op);
 638         if (!ret) {
 639                 struct seq_file *m = file->private_data;
 640                 m->private = task;
 641         }
 642         return ret;
 643 }
 644
 645 static struct file_operations proc_smaps_operations = {
 646         .open           = smaps_open,
 647         .read           = seq_read,
 648         .llseek         = seq_lseek,
 649         .release        = seq_release,
 650 };
 651
 652 extern struct seq_operations mounts_op;
 653 static int mounts_open(struct inode *inode, struct file *file)
 654 {
 655         struct task_struct *task = proc_task(inode);
 656         int ret = seq_open(file, &mounts_op);
 657
 658         if (!ret) {
 659                 struct seq_file *m = file->private_data;
 660                 struct namespace *namespace;
 661                 task_lock(task);
 662                 namespace = task->namespace;
 663                 if (namespace)
 664                         get_namespace(namespace);
 665                 task_unlock(task);
 666
 667                 if (namespace)
 668                         m->private = namespace;
 669                 else {
 670                         seq_release(inode, file);
 671                         ret = -EINVAL;
 672                 }
 673         }
 674         return ret;
 675 }
 676
 677 static int mounts_release(struct inode *inode, struct file *file)
 678 {
 679         struct seq_file *m = file->private_data;
 680         struct namespace *namespace = m->private;
 681         put_namespace(namespace);
 682         return seq_release(inode, file);
 683 }
 684
 685 static struct file_operations proc_mounts_operations = {
 686         .open           = mounts_open,
 687         .read           = seq_read,
 688         .llseek         = seq_lseek,
 689         .release        = mounts_release,
 690 };
 691
 692 #define PROC_BLOCK_SIZE (3*1024)                /* 4K page size but our output routines use some slack for overruns */
 693
 694 static ssize_t proc_info_read(struct file * file, char __user * buf,
 695                           size_t count, loff_t *ppos)
 696 {
 697         struct inode * inode = file->f_dentry->d_inode;
 698         unsigned long page;
 699         ssize_t length;
 700         struct task_struct *task = proc_task(inode);
 701
 702         if (count > PROC_BLOCK_SIZE)
 703                 count = PROC_BLOCK_SIZE;
 704         if (!(page = __get_free_page(GFP_KERNEL)))
 705                 return -ENOMEM;
 706
 707         length = PROC_I(inode)->op.proc_read(task, (char*)page);
 708
 709         if (length >= 0)
 710                 length = simple_read_from_buffer(buf, count, ppos, (char *)page, length);
 711         free_page(page);
 712         return length;
 713 }
 714
 715 static struct file_operations proc_info_file_operations = {
 716         .read           = proc_info_read,
 717 };
 718
 719 static int mem_open(struct inode* inode, struct file* file)
 720 {
 721         file->private_data = (void*)((long)current->self_exec_id);
 722         return 0;
 723 }
 724
 725 static ssize_t mem_read(struct file * file, char __user * buf,
 726                         size_t count, loff_t *ppos)
 727 {
 728         struct task_struct *task = proc_task(file->f_dentry->d_inode);
 729         char *page;
 730         unsigned long src = *ppos;
 731         int ret = -ESRCH;
 732         struct mm_struct *mm;
 733
 734         if (!MAY_PTRACE(task) || !ptrace_may_attach(task))
 735                 goto out;
 736
 737         ret = -ENOMEM;
 738         page = (char *)__get_free_page(GFP_USER);
 739         if (!page)
 740                 goto out;
 741
 742         ret = 0;
 743
 744         mm = get_task_mm(task);
 745         if (!mm)
 746                 goto out_free;
 747
 748         ret = -EIO;
 749
 750         if (file->private_data != (void*)((long)current->self_exec_id))
 751                 goto out_put;
 752
 753         ret = 0;
 754
 755         while (count > 0) {
 756                 int this_len, retval;
 757
 758                 this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
 759                 retval = access_process_vm(task, src, page, this_len, 0);
 760                 if (!retval || !MAY_PTRACE(task) || !ptrace_may_attach(task)) {
 761                         if (!ret)
 762                                 ret = -EIO;
 763                         break;
 764                 }
 765
 766                 if (copy_to_user(buf, page, retval)) {
 767                         ret = -EFAULT;
 768                         break;
 769                 }
 770
 771                 ret += retval;
 772                 src += retval;
 773                 buf += retval;
 774                 count -= retval;
 775         }
 776         *ppos = src;
 777
 778 out_put:
 779         mmput(mm);
 780 out_free:
 781         free_page((unsigned long) page);
 782 out:
 783         return ret;
 784 }
 785
 786 #define mem_write NULL
 787
 788 #ifndef mem_write
 789 /* This is a security hazard */
 790 static ssize_t mem_write(struct file * file, const char * buf,
 791                          size_t count, loff_t *ppos)
 792 {
 793         int copied = 0;
 794         char *page;
 795         struct task_struct *task = proc_task(file->f_dentry->d_inode);
 796         unsigned long dst = *ppos;
 797
 798         if (!MAY_PTRACE(task) || !ptrace_may_attach(task))
 799                 return -ESRCH;
 800
 801         page = (char *)__get_free_page(GFP_USER);
 802         if (!page)
 803                 return -ENOMEM;
 804
 805         while (count > 0) {
 806                 int this_len, retval;
 807
 808                 this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
 809                 if (copy_from_user(page, buf, this_len)) {
 810                         copied = -EFAULT;
 811                         break;
 812                 }
 813                 retval = access_process_vm(task, dst, page, this_len, 1);
 814                 if (!retval) {
 815                         if (!copied)
 816                                 copied = -EIO;
 817                         break;
 818                 }
 819                 copied += retval;
 820                 buf += retval;
 821                 dst += retval;
 822                 count -= retval;
 823         }
 824         *ppos = dst;
 825         free_page((unsigned long) page);
 826         return copied;
 827 }
 828 #endif
 829
 830 static loff_t mem_lseek(struct file * file, loff_t offset, int orig)
 831 {
 832         switch (orig) {
 833         case 0:
 834                 file->f_pos = offset;
 835                 break;
 836         case 1:
 837                 file->f_pos += offset;
 838                 break;
 839         default:
 840                 return -EINVAL;
 841         }
 842         force_successful_syscall_return();
 843         return file->f_pos;
 844 }
 845
 846 static struct file_operations proc_mem_operations = {
 847         .llseek         = mem_lseek,
 848         .read           = mem_read,
 849         .write          = mem_write,
 850         .open           = mem_open,
 851 };
 852
 853 static ssize_t oom_adjust_read(struct file *file, char __user *buf,
 854                                 size_t count, loff_t *ppos)
 855 {
 856         struct task_struct *task = proc_task(file->f_dentry->d_inode);
 857         char buffer[8];
 858         size_t len;
 859         int oom_adjust = task->oomkilladj;
 860         loff_t __ppos = *ppos;
 861
 862         len = sprintf(buffer, "%i\n", oom_adjust);
 863         if (__ppos >= len)
 864                 return 0;
 865         if (count > len-__ppos)
 866                 count = len-__ppos;
 867         if (copy_to_user(buf, buffer + __ppos, count))
 868                 return -EFAULT;
 869         *ppos = __ppos + count;
 870         return count;
 871 }
 872
 873 static ssize_t oom_adjust_write(struct file *file, const char __user *buf,
 874                                 size_t count, loff_t *ppos)
 875 {
 876         struct task_struct *task = proc_task(file->f_dentry->d_inode);
 877         char buffer[8], *end;
 878         int oom_adjust;
 879
 880         if (!capable(CAP_SYS_RESOURCE))
 881                 return -EPERM;
 882         memset(buffer, 0, 8);
 883         if (count > 6)
 884                 count = 6;
 885         if (copy_from_user(buffer, buf, count))
 886                 return -EFAULT;
 887         oom_adjust = simple_strtol(buffer, &end, 0);
 888         if ((oom_adjust < -16 || oom_adjust > 15) && oom_adjust != OOM_DISABLE)
 889                 return -EINVAL;
 890         if (*end == '\n')
 891                 end++;
 892         task->oomkilladj = oom_adjust;
 893         if (end - buffer == 0)
 894                 return -EIO;
 895         return end - buffer;
 896 }
 897
 898 static struct file_operations proc_oom_adjust_operations = {
 899         .read           = oom_adjust_read,
 900         .write          = oom_adjust_write,
 901 };
 902
 903 static struct inode_operations proc_mem_inode_operations = {
 904         .permission     = proc_permission,
 905 };
 906
 907 #ifdef CONFIG_AUDITSYSCALL
 908 #define TMPBUFLEN 21
 909 static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
 910                                   size_t count, loff_t *ppos)
 911 {
 912         struct inode * inode = file->f_dentry->d_inode;
 913         struct task_struct *task = proc_task(inode);
 914         ssize_t length;
 915         char tmpbuf[TMPBUFLEN];
 916
 917         length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
 918                                 audit_get_loginuid(task->audit_context));
 919         return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
 920 }
 921
 922 static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
 923                                    size_t count, loff_t *ppos)
 924 {
 925         struct inode * inode = file->f_dentry->d_inode;
 926         char *page, *tmp;
 927         ssize_t length;
 928         struct task_struct *task = proc_task(inode);
 929         uid_t loginuid;
 930
 931         if (!capable(CAP_AUDIT_CONTROL))
 932                 return -EPERM;
 933
 934         if (current != task)
 935                 return -EPERM;
 936
 937         if (count > PAGE_SIZE)
 938                 count = PAGE_SIZE;
 939
 940         if (*ppos != 0) {
 941                 /* No partial writes. */
 942                 return -EINVAL;
 943         }
 944         page = (char*)__get_free_page(GFP_USER);
 945         if (!page)
 946                 return -ENOMEM;
 947         length = -EFAULT;
 948         if (copy_from_user(page, buf, count))
 949                 goto out_free_page;
 950
 951         loginuid = simple_strtoul(page, &tmp, 10);
 952         if (tmp == page) {
 953                 length = -EINVAL;
 954                 goto out_free_page;
 955
 956         }
 957         length = audit_set_loginuid(task, loginuid);
 958         if (likely(length == 0))
 959                 length = count;
 960
 961 out_free_page:
 962         free_page((unsigned long) page);
 963         return length;
 964 }
 965
 966 static struct file_operations proc_loginuid_operations = {
 967         .read           = proc_loginuid_read,
 968         .write          = proc_loginuid_write,
 969 };
 970 #endif
 971
 972 #ifdef CONFIG_SECCOMP
 973 static ssize_t seccomp_read(struct file *file, char __user *buf,
 974                             size_t count, loff_t *ppos)
 975 {
 976         struct task_struct *tsk = proc_task(file->f_dentry->d_inode);
 977         char __buf[20];
 978         loff_t __ppos = *ppos;
 979         size_t len;
 980
 981         /* no need to print the trailing zero, so use only len */
 982         len = sprintf(__buf, "%u\n", tsk->seccomp.mode);
 983         if (__ppos >= len)
 984                 return 0;
 985         if (count > len - __ppos)
 986                 count = len - __ppos;
 987         if (copy_to_user(buf, __buf + __ppos, count))
 988                 return -EFAULT;
 989         *ppos = __ppos + count;
 990         return count;
 991 }
 992
 993 static ssize_t seccomp_write(struct file *file, const char __user *buf,
 994                              size_t count, loff_t *ppos)
 995 {
 996         struct task_struct *tsk = proc_task(file->f_dentry->d_inode);
 997         char __buf[20], *end;
 998         unsigned int seccomp_mode;
 999
1000         /* can set it only once to be even more secure */
1001         if (unlikely(tsk->seccomp.mode))
1002                 return -EPERM;
1003
1004         memset(__buf, 0, sizeof(__buf));
1005         count = min(count, sizeof(__buf) - 1);
1006         if (copy_from_user(__buf, buf, count))
1007                 return -EFAULT;
1008         seccomp_mode = simple_strtoul(__buf, &end, 0);
1009         if (*end == '\n')
1010                 end++;
1011         if (seccomp_mode && seccomp_mode <= NR_SECCOMP_MODES) {
1012                 tsk->seccomp.mode = seccomp_mode;
1013                 set_tsk_thread_flag(tsk, TIF_SECCOMP);
1014         } else
1015                 return -EINVAL;
1016         if (unlikely(!(end - __buf)))
1017                 return -EIO;
1018         return end - __buf;
1019 }
1020
1021 static struct file_operations proc_seccomp_operations = {
1022         .read           = seccomp_read,
1023         .write          = seccomp_write,
1024 };
1025 #endif /* CONFIG_SECCOMP */
1026
1027 static void *proc_pid_follow_link(struct dentry *dentry, struct nameidata *nd)
1028 {
1029         struct inode *inode = dentry->d_inode;
1030         int error = -EACCES;
1031
1032         /* We don't need a base pointer in the /proc filesystem */
1033         path_release(nd);
1034
1035         if (current->fsuid != inode->i_uid && !capable(CAP_DAC_OVERRIDE))
1036                 goto out;
1037         error = proc_check_root(inode);
1038         if (error)
1039                 goto out;
1040
1041         error = PROC_I(inode)->op.proc_get_link(inode, &nd->dentry, &nd->mnt);
1042         nd->last_type = LAST_BIND;
1043 out:
1044         return ERR_PTR(error);
1045 }
1046
1047 static int do_proc_readlink(struct dentry *dentry, struct vfsmount *mnt,
1048                             char __user *buffer, int buflen)
1049 {
1050         struct inode * inode;
1051         char *tmp = (char*)__get_free_page(GFP_KERNEL), *path;
1052         int len;
1053
1054         if (!tmp)
1055                 return -ENOMEM;
1056
1057         inode = dentry->d_inode;
1058         path = d_path(dentry, mnt, tmp, PAGE_SIZE);
1059         len = PTR_ERR(path);
1060         if (IS_ERR(path))
1061                 goto out;
1062         len = tmp + PAGE_SIZE - 1 - path;
1063
1064         if (len > buflen)
1065                 len = buflen;
1066         if (copy_to_user(buffer, path, len))
1067                 len = -EFAULT;
1068  out:
1069         free_page((unsigned long)tmp);
1070         return len;
1071 }
1072
1073 static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1074 {
1075         int error = -EACCES;
1076         struct inode *inode = dentry->d_inode;
1077         struct dentry *de;
1078         struct vfsmount *mnt = NULL;
1079
1080         lock_kernel();
1081
1082         if (current->fsuid != inode->i_uid && !capable(CAP_DAC_OVERRIDE))
1083                 goto out;
1084         error = proc_check_root(inode);
1085         if (error)
1086                 goto out;
1087
1088         error = PROC_I(inode)->op.proc_get_link(inode, &de, &mnt);
1089         if (error)
1090                 goto out;
1091
1092         error = do_proc_readlink(de, mnt, buffer, buflen);
1093         dput(de);
1094         mntput(mnt);
1095 out:
1096         unlock_kernel();
1097         return error;
1098 }
1099
1100 static struct inode_operations proc_pid_link_inode_operations = {
1101         .readlink       = proc_pid_readlink,
1102         .follow_link    = proc_pid_follow_link
1103 };
1104
1105 #define NUMBUF 10
1106
1107 static int proc_readfd(struct file * filp, void * dirent, filldir_t filldir)
1108 {
1109         struct inode *inode = filp->f_dentry->d_inode;
1110         struct task_struct *p = proc_task(inode);
1111         unsigned int fd, tid, ino;
1112         int retval;
1113         char buf[NUMBUF];
1114         struct files_struct * files;
1115         struct fdtable *fdt;
1116
1117         retval = -ENOENT;
1118         if (!pid_alive(p))
1119                 goto out;
1120         retval = 0;
1121         tid = p->pid;
1122
1123         fd = filp->f_pos;
1124         switch (fd) {
1125                 case 0:
1126                         if (filldir(dirent, ".", 1, 0, inode->i_ino, DT_DIR) < 0)
1127                                 goto out;
1128                         filp->f_pos++;
1129                 case 1:
1130                         ino = fake_ino(tid, PROC_TID_INO);
1131                         if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0)
1132                                 goto out;
1133                         filp->f_pos++;
1134                 default:
1135                         files = get_files_struct(p);
1136                         if (!files)
1137                                 goto out;
1138                         rcu_read_lock();
1139                         fdt = files_fdtable(files);
1140                         for (fd = filp->f_pos-2;
1141                              fd < fdt->max_fds;
1142                              fd++, filp->f_pos++) {
1143                                 unsigned int i,j;
1144
1145                                 if (!fcheck_files(files, fd))
1146                                         continue;
1147                                 rcu_read_unlock();
1148
1149                                 j = NUMBUF;
1150                                 i = fd;
1151                                 do {
1152                                         j--;
1153                                         buf[j] = '0' + (i % 10);
1154                                         i /= 10;
1155                                 } while (i);
1156
1157                                 ino = fake_ino(tid, PROC_TID_FD_DIR + fd);
1158                                 if (filldir(dirent, buf+j, NUMBUF-j, fd+2, ino, DT_LNK) < 0) {
1159                                         rcu_read_lock();
1160                                         break;
1161                                 }
1162                                 rcu_read_lock();
1163                         }
1164                         rcu_read_unlock();
1165                         put_files_struct(files);
1166         }
1167 out:
1168         return retval;
1169 }
1170
1171 static int proc_pident_readdir(struct file *filp,
1172                 void *dirent, filldir_t filldir,
1173                 struct pid_entry *ents, unsigned int nents)
1174 {
1175         int i;
1176         int pid;
1177         struct dentry *dentry = filp->f_dentry;
1178         struct inode *inode = dentry->d_inode;
1179         struct pid_entry *p;
1180         ino_t ino;
1181         int ret;
1182
1183         ret = -ENOENT;
1184         if (!pid_alive(proc_task(inode)))
1185                 goto out;
1186
1187         ret = 0;
1188         pid = proc_task(inode)->pid;
1189         i = filp->f_pos;
1190         switch (i) {
1191         case 0:
1192                 ino = inode->i_ino;
1193                 if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0)
1194                         goto out;
1195                 i++;
1196                 filp->f_pos++;
1197                 /* fall through */
1198         case 1:
1199                 ino = parent_ino(dentry);
1200                 if (filldir(dirent, "..", 2, i, ino, DT_DIR) < 0)
1201                         goto out;
1202                 i++;
1203                 filp->f_pos++;
1204                 /* fall through */
1205         default:
1206                 i -= 2;
1207                 if (i >= nents) {
1208                         ret = 1;
1209                         goto out;
1210                 }
1211                 p = ents + i;
1212                 while (p->name) {
1213                         if (filldir(dirent, p->name, p->len, filp->f_pos,
1214                                     fake_ino(pid, p->type), p->mode >> 12) < 0)
1215                                 goto out;
1216                         filp->f_pos++;
1217                         p++;
1218                 }
1219         }
1220
1221         ret = 1;
1222 out:
1223         return ret;
1224 }
1225
1226 static int proc_tgid_base_readdir(struct file * filp,
1227                              void * dirent, filldir_t filldir)
1228 {
1229         return proc_pident_readdir(filp,dirent,filldir,
1230                                    tgid_base_stuff,ARRAY_SIZE(tgid_base_stuff));
1231 }
1232
1233 static int proc_tid_base_readdir(struct file * filp,
1234                              void * dirent, filldir_t filldir)
1235 {
1236         return proc_pident_readdir(filp,dirent,filldir,
1237                                    tid_base_stuff,ARRAY_SIZE(tid_base_stuff));
1238 }
1239
1240 /* building an inode */
1241
1242 static int task_dumpable(struct task_struct *task)
1243 {
1244         int dumpable = 0;
1245         struct mm_struct *mm;
1246
1247         task_lock(task);
1248         mm = task->mm;
1249         if (mm)
1250                 dumpable = mm->dumpable;
1251         task_unlock(task);
1252         if(dumpable == 1)
1253                 return 1;
1254         return 0;
1255 }
1256
1257
1258 static struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task, int ino)
1259 {
1260         struct inode * inode;
1261         struct proc_inode *ei;
1262
1263         /* We need a new inode */
1264
1265         inode = new_inode(sb);
1266         if (!inode)
1267                 goto out;
1268
1269         /* Common stuff */
1270         ei = PROC_I(inode);
1271         ei->task = NULL;
1272         inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1273         inode->i_ino = fake_ino(task->pid, ino);
1274
1275         if (!pid_alive(task))
1276                 goto out_unlock;
1277
1278         /*
1279          * grab the reference to task.
1280          */
1281         get_task_struct(task);
1282         ei->task = task;
1283         ei->type = ino;
1284         inode->i_uid = 0;
1285         inode->i_gid = 0;
1286         if (ino == PROC_TGID_INO || ino == PROC_TID_INO || task_dumpable(task)) {
1287                 inode->i_uid = task->euid;
1288                 inode->i_gid = task->egid;
1289         }
1290         security_task_to_inode(task, inode);
1291
1292 out:
1293         return inode;
1294
1295 out_unlock:
1296         ei->pde = NULL;
1297         iput(inode);
1298         return NULL;
1299 }
1300
1301 /* dentry stuff */
1302
1303 /*
1304  *      Exceptional case: normally we are not allowed to unhash a busy
1305  * directory. In this case, however, we can do it - no aliasing problems
1306  * due to the way we treat inodes.
1307  *
1308  * Rewrite the inode's ownerships here because the owning task may have
1309  * performed a setuid(), etc.
1310  */
1311 static int pid_revalidate(struct dentry *dentry, struct nameidata *nd)
1312 {
1313         struct inode *inode = dentry->d_inode;
1314         struct task_struct *task = proc_task(inode);
1315         if (pid_alive(task)) {
1316                 if (proc_type(inode) == PROC_TGID_INO || proc_type(inode) == PROC_TID_INO || task_dumpable(task)) {
1317                         inode->i_uid = task->euid;
1318                         inode->i_gid = task->egid;
1319                 } else {
1320                         inode->i_uid = 0;
1321                         inode->i_gid = 0;
1322                 }
1323                 security_task_to_inode(task, inode);
1324                 return 1;
1325         }
1326         d_drop(dentry);
1327         return 0;
1328 }
1329
1330 static int tid_fd_revalidate(struct dentry *dentry, struct nameidata *nd)
1331 {
1332         struct inode *inode = dentry->d_inode;
1333         struct task_struct *task = proc_task(inode);
1334         int fd = proc_type(inode) - PROC_TID_FD_DIR;
1335         struct files_struct *files;
1336
1337         files = get_files_struct(task);
1338         if (files) {
1339                 rcu_read_lock();
1340                 if (fcheck_files(files, fd)) {
1341                         rcu_read_unlock();
1342                         put_files_struct(files);
1343                         if (task_dumpable(task)) {
1344                                 inode->i_uid = task->euid;
1345                                 inode->i_gid = task->egid;
1346                         } else {
1347                                 inode->i_uid = 0;
1348                                 inode->i_gid = 0;
1349                         }
1350                         security_task_to_inode(task, inode);
1351                         return 1;
1352                 }
1353                 rcu_read_unlock();
1354                 put_files_struct(files);
1355         }
1356         d_drop(dentry);
1357         return 0;
1358 }
1359
1360 static void pid_base_iput(struct dentry *dentry, struct inode *inode)
1361 {
1362         struct task_struct *task = proc_task(inode);
1363         spin_lock(&task->proc_lock);
1364         if (task->proc_dentry == dentry)
1365                 task->proc_dentry = NULL;
1366         spin_unlock(&task->proc_lock);
1367         iput(inode);
1368 }
1369
1370 static int pid_delete_dentry(struct dentry * dentry)
1371 {
1372         /* Is the task we represent dead?
1373          * If so, then don't put the dentry on the lru list,
1374          * kill it immediately.
1375          */
1376         return !pid_alive(proc_task(dentry->d_inode));
1377 }
1378
1379 static struct dentry_operations tid_fd_dentry_operations =
1380 {
1381         .d_revalidate   = tid_fd_revalidate,
1382         .d_delete       = pid_delete_dentry,
1383 };
1384
1385 static struct dentry_operations pid_dentry_operations =
1386 {
1387         .d_revalidate   = pid_revalidate,
1388         .d_delete       = pid_delete_dentry,
1389 };
1390
1391 static struct dentry_operations pid_base_dentry_operations =
1392 {
1393         .d_revalidate   = pid_revalidate,
1394         .d_iput         = pid_base_iput,
1395         .d_delete       = pid_delete_dentry,
1396 };
1397
1398 /* Lookups */
1399
1400 static unsigned name_to_int(struct dentry *dentry)
1401 {
1402         const char *name = dentry->d_name.name;
1403         int len = dentry->d_name.len;
1404         unsigned n = 0;
1405
1406         if (len > 1 && *name == '0')
1407                 goto out;
1408         while (len-- > 0) {
1409                 unsigned c = *name++ - '0';
1410                 if (c > 9)
1411                         goto out;
1412                 if (n >= (~0U-9)/10)
1413                         goto out;
1414                 n *= 10;
1415                 n += c;
1416         }
1417         return n;
1418 out:
1419         return ~0U;
1420 }
1421
1422 /* SMP-safe */
1423 static struct dentry *proc_lookupfd(struct inode * dir, struct dentry * dentry, struct nameidata *nd)
1424 {
1425         struct task_struct *task = proc_task(dir);
1426         unsigned fd = name_to_int(dentry);
1427         struct file * file;
1428         struct files_struct * files;
1429         struct inode *inode;
1430         struct proc_inode *ei;
1431
1432         if (fd == ~0U)
1433                 goto out;
1434         if (!pid_alive(task))
1435                 goto out;
1436
1437         inode = proc_pid_make_inode(dir->i_sb, task, PROC_TID_FD_DIR+fd);
1438         if (!inode)
1439                 goto out;
1440         ei = PROC_I(inode);
1441         files = get_files_struct(task);
1442         if (!files)
1443                 goto out_unlock;
1444         inode->i_mode = S_IFLNK;
1445         rcu_read_lock();
1446         file = fcheck_files(files, fd);
1447         if (!file)
1448                 goto out_unlock2;
1449         if (file->f_mode & 1)
1450                 inode->i_mode |= S_IRUSR | S_IXUSR;
1451         if (file->f_mode & 2)
1452                 inode->i_mode |= S_IWUSR | S_IXUSR;
1453         rcu_read_unlock();
1454         put_files_struct(files);
1455         inode->i_op = &proc_pid_link_inode_operations;
1456         inode->i_size = 64;
1457         ei->op.proc_get_link = proc_fd_link;
1458         dentry->d_op = &tid_fd_dentry_operations;
1459         d_add(dentry, inode);
1460         return NULL;
1461
1462 out_unlock2:
1463         rcu_read_unlock();
1464         put_files_struct(files);
1465 out_unlock:
1466         iput(inode);
1467 out:
1468         return ERR_PTR(-ENOENT);
1469 }
1470
1471 static int proc_task_readdir(struct file * filp, void * dirent, filldir_t filldir);
1472 static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd);
1473
1474 static struct file_operations proc_fd_operations = {
1475         .read           = generic_read_dir,
1476         .readdir        = proc_readfd,
1477 };
1478
1479 static struct file_operations proc_task_operations = {
1480         .read           = generic_read_dir,
1481         .readdir        = proc_task_readdir,
1482 };
1483
1484 /*
1485  * proc directories can do almost nothing..
1486  */
1487 static struct inode_operations proc_fd_inode_operations = {
1488         .lookup         = proc_lookupfd,
1489         .permission     = proc_permission,
1490 };
1491
1492 static struct inode_operations proc_task_inode_operations = {
1493         .lookup         = proc_task_lookup,
1494         .permission     = proc_task_permission,
1495 };
1496
1497 #ifdef CONFIG_SECURITY
1498 static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
1499                                   size_t count, loff_t *ppos)
1500 {
1501         struct inode * inode = file->f_dentry->d_inode;
1502         unsigned long page;
1503         ssize_t length;
1504         struct task_struct *task = proc_task(inode);
1505
1506         if (count > PAGE_SIZE)
1507                 count = PAGE_SIZE;
1508         if (!(page = __get_free_page(GFP_KERNEL)))
1509                 return -ENOMEM;
1510
1511         length = security_getprocattr(task,
1512                                       (char*)file->f_dentry->d_name.name,
1513                                       (void*)page, count);
1514         if (length >= 0)
1515                 length = simple_read_from_buffer(buf, count, ppos, (char *)page, length);
1516         free_page(page);
1517         return length;
1518 }
1519
1520 static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
1521                                    size_t count, loff_t *ppos)
1522 {
1523         struct inode * inode = file->f_dentry->d_inode;
1524         char *page;
1525         ssize_t length;
1526         struct task_struct *task = proc_task(inode);
1527
1528         if (count > PAGE_SIZE)
1529                 count = PAGE_SIZE;
1530         if (*ppos != 0) {
1531                 /* No partial writes. */
1532                 return -EINVAL;
1533         }
1534         page = (char*)__get_free_page(GFP_USER);
1535         if (!page)
1536                 return -ENOMEM;
1537         length = -EFAULT;
1538         if (copy_from_user(page, buf, count))
1539                 goto out;
1540
1541         length = security_setprocattr(task,
1542                                       (char*)file->f_dentry->d_name.name,
1543                                       (void*)page, count);
1544 out:
1545         free_page((unsigned long) page);
1546         return length;
1547 }
1548
1549 static struct file_operations proc_pid_attr_operations = {
1550         .read           = proc_pid_attr_read,
1551         .write          = proc_pid_attr_write,
1552 };
1553
1554 static struct file_operations proc_tid_attr_operations;
1555 static struct inode_operations proc_tid_attr_inode_operations;
1556 static struct file_operations proc_tgid_attr_operations;
1557 static struct inode_operations proc_tgid_attr_inode_operations;
1558 #endif
1559
1560 static int get_tid_list(int index, unsigned int *tids, struct inode *dir);
1561
1562 /* SMP-safe */
1563 static struct dentry *proc_pident_lookup(struct inode *dir,
1564                                          struct dentry *dentry,
1565                                          struct pid_entry *ents)
1566 {
1567         struct inode *inode;
1568         int error;
1569         struct task_struct *task = proc_task(dir);
1570         struct pid_entry *p;
1571         struct proc_inode *ei;
1572
1573         error = -ENOENT;
1574         inode = NULL;
1575
1576         if (!pid_alive(task))
1577                 goto out;
1578
1579         for (p = ents; p->name; p++) {
1580                 if (p->len != dentry->d_name.len)
1581                         continue;
1582                 if (!memcmp(dentry->d_name.name, p->name, p->len))
1583                         break;
1584         }
1585         if (!p->name)
1586                 goto out;
1587
1588         error = -EINVAL;
1589         inode = proc_pid_make_inode(dir->i_sb, task, p->type);
1590         if (!inode)
1591                 goto out;
1592
1593         ei = PROC_I(inode);
1594         inode->i_mode = p->mode;
1595         /*
1596          * Yes, it does not scale. And it should not. Don't add
1597          * new entries into /proc/<tgid>/ without very good reasons.
1598          */
1599         switch(p->type) {
1600                 case PROC_TGID_TASK:
1601                         inode->i_nlink = 2 + get_tid_list(2, NULL, dir);
1602                         inode->i_op = &proc_task_inode_operations;
1603                         inode->i_fop = &proc_task_operations;
1604                         break;
1605                 case PROC_TID_FD:
1606                 case PROC_TGID_FD:
1607                         inode->i_nlink = 2;
1608                         inode->i_op = &proc_fd_inode_operations;
1609                         inode->i_fop = &proc_fd_operations;
1610                         break;
1611                 case PROC_TID_EXE:
1612                 case PROC_TGID_EXE:
1613                         inode->i_op = &proc_pid_link_inode_operations;
1614                         ei->op.proc_get_link = proc_exe_link;
1615                         break;
1616                 case PROC_TID_CWD:
1617                 case PROC_TGID_CWD:
1618                         inode->i_op = &proc_pid_link_inode_operations;
1619                         ei->op.proc_get_link = proc_cwd_link;
1620                         break;
1621                 case PROC_TID_ROOT:
1622                 case PROC_TGID_ROOT:
1623                         inode->i_op = &proc_pid_link_inode_operations;
1624                         ei->op.proc_get_link = proc_root_link;
1625                         break;
1626                 case PROC_TID_ENVIRON:
1627                 case PROC_TGID_ENVIRON:
1628                         inode->i_fop = &proc_info_file_operations;
1629                         ei->op.proc_read = proc_pid_environ;
1630                         break;
1631                 case PROC_TID_AUXV:
1632                 case PROC_TGID_AUXV:
1633                         inode->i_fop = &proc_info_file_operations;
1634                         ei->op.proc_read = proc_pid_auxv;
1635                         break;
1636                 case PROC_TID_STATUS:
1637                 case PROC_TGID_STATUS:
1638                         inode->i_fop = &proc_info_file_operations;
1639                         ei->op.proc_read = proc_pid_status;
1640                         break;
1641                 case PROC_TID_STAT:
1642                         inode->i_fop = &proc_info_file_operations;
1643                         ei->op.proc_read = proc_tid_stat;
1644                         break;
1645                 case PROC_TGID_STAT:
1646                         inode->i_fop = &proc_info_file_operations;
1647                         ei->op.proc_read = proc_tgid_stat;
1648                         break;
1649                 case PROC_TID_CMDLINE:
1650                 case PROC_TGID_CMDLINE:
1651                         inode->i_fop = &proc_info_file_operations;
1652                         ei->op.proc_read = proc_pid_cmdline;
1653                         break;
1654                 case PROC_TID_STATM:
1655                 case PROC_TGID_STATM:
1656                         inode->i_fop = &proc_info_file_operations;
1657                         ei->op.proc_read = proc_pid_statm;
1658                         break;
1659                 case PROC_TID_MAPS:
1660                 case PROC_TGID_MAPS:
1661                         inode->i_fop = &proc_maps_operations;
1662                         break;
1663 #ifdef CONFIG_NUMA
1664                 case PROC_TID_NUMA_MAPS:
1665                 case PROC_TGID_NUMA_MAPS:
1666                         inode->i_fop = &proc_numa_maps_operations;
1667                         break;
1668 #endif
1669                 case PROC_TID_MEM:
1670                 case PROC_TGID_MEM:
1671                         inode->i_op = &proc_mem_inode_operations;
1672                         inode->i_fop = &proc_mem_operations;
1673                         break;
1674 #ifdef CONFIG_SECCOMP
1675                 case PROC_TID_SECCOMP:
1676                 case PROC_TGID_SECCOMP:
1677                         inode->i_fop = &proc_seccomp_operations;
1678                         break;
1679 #endif /* CONFIG_SECCOMP */
1680                 case PROC_TID_MOUNTS:
1681                 case PROC_TGID_MOUNTS:
1682                         inode->i_fop = &proc_mounts_operations;
1683                         break;
1684                 case PROC_TID_SMAPS:
1685                 case PROC_TGID_SMAPS:
1686                         inode->i_fop = &proc_smaps_operations;
1687                         break;
1688 #ifdef CONFIG_SECURITY
1689                 case PROC_TID_ATTR:
1690                         inode->i_nlink = 2;
1691                         inode->i_op = &proc_tid_attr_inode_operations;
1692                         inode->i_fop = &proc_tid_attr_operations;
1693                         break;
1694                 case PROC_TGID_ATTR:
1695                         inode->i_nlink = 2;
1696                         inode->i_op = &proc_tgid_attr_inode_operations;
1697                         inode->i_fop = &proc_tgid_attr_operations;
1698                         break;
1699                 case PROC_TID_ATTR_CURRENT:
1700                 case PROC_TGID_ATTR_CURRENT:
1701                 case PROC_TID_ATTR_PREV:
1702                 case PROC_TGID_ATTR_PREV:
1703                 case PROC_TID_ATTR_EXEC:
1704                 case PROC_TGID_ATTR_EXEC:
1705                 case PROC_TID_ATTR_FSCREATE:
1706                 case PROC_TGID_ATTR_FSCREATE:
1707                         inode->i_fop = &proc_pid_attr_operations;
1708                         break;
1709 #endif
1710 #ifdef CONFIG_KALLSYMS
1711                 case PROC_TID_WCHAN:
1712                 case PROC_TGID_WCHAN:
1713                         inode->i_fop = &proc_info_file_operations;
1714                         ei->op.proc_read = proc_pid_wchan;
1715                         break;
1716 #endif
1717 #ifdef CONFIG_SCHEDSTATS
1718                 case PROC_TID_SCHEDSTAT:
1719                 case PROC_TGID_SCHEDSTAT:
1720                         inode->i_fop = &proc_info_file_operations;
1721                         ei->op.proc_read = proc_pid_schedstat;
1722                         break;
1723 #endif
1724 #ifdef CONFIG_CPUSETS
1725                 case PROC_TID_CPUSET:
1726                 case PROC_TGID_CPUSET:
1727                         inode->i_fop = &proc_cpuset_operations;
1728                         break;
1729 #endif
1730                 case PROC_TID_OOM_SCORE:
1731                 case PROC_TGID_OOM_SCORE:
1732                         inode->i_fop = &proc_info_file_operations;
1733                         ei->op.proc_read = proc_oom_score;
1734                         break;
1735                 case PROC_TID_OOM_ADJUST:
1736                 case PROC_TGID_OOM_ADJUST:
1737                         inode->i_fop = &proc_oom_adjust_operations;
1738                         break;
1739 #ifdef CONFIG_AUDITSYSCALL
1740                 case PROC_TID_LOGINUID:
1741                 case PROC_TGID_LOGINUID:
1742                         inode->i_fop = &proc_loginuid_operations;
1743                         break;
1744 #endif
1745                 default:
1746                         printk("procfs: impossible type (%d)",p->type);
1747                         iput(inode);
1748                         return ERR_PTR(-EINVAL);
1749         }
1750         dentry->d_op = &pid_dentry_operations;
1751         d_add(dentry, inode);
1752         return NULL;
1753
1754 out:
1755         return ERR_PTR(error);
1756 }
1757
1758 static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
1759         return proc_pident_lookup(dir, dentry, tgid_base_stuff);
1760 }
1761
1762 static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
1763         return proc_pident_lookup(dir, dentry, tid_base_stuff);
1764 }
1765
1766 static struct file_operations proc_tgid_base_operations = {
1767         .read           = generic_read_dir,
1768         .readdir        = proc_tgid_base_readdir,
1769 };
1770
1771 static struct file_operations proc_tid_base_operations = {
1772         .read           = generic_read_dir,
1773         .readdir        = proc_tid_base_readdir,
1774 };
1775
1776 static struct inode_operations proc_tgid_base_inode_operations = {
1777         .lookup         = proc_tgid_base_lookup,
1778 };
1779
1780 static struct inode_operations proc_tid_base_inode_operations = {
1781         .lookup         = proc_tid_base_lookup,
1782 };
1783
1784 #ifdef CONFIG_SECURITY
1785 static int proc_tgid_attr_readdir(struct file * filp,
1786                              void * dirent, filldir_t filldir)
1787 {
1788         return proc_pident_readdir(filp,dirent,filldir,
1789                                    tgid_attr_stuff,ARRAY_SIZE(tgid_attr_stuff));
1790 }
1791
1792 static int proc_tid_attr_readdir(struct file * filp,
1793                              void * dirent, filldir_t filldir)
1794 {
1795         return proc_pident_readdir(filp,dirent,filldir,
1796                                    tid_attr_stuff,ARRAY_SIZE(tid_attr_stuff));
1797 }
1798
1799 static struct file_operations proc_tgid_attr_operations = {
1800         .read           = generic_read_dir,
1801         .readdir        = proc_tgid_attr_readdir,
1802 };
1803
1804 static struct file_operations proc_tid_attr_operations = {
1805         .read           = generic_read_dir,
1806         .readdir        = proc_tid_attr_readdir,
1807 };
1808
1809 static struct dentry *proc_tgid_attr_lookup(struct inode *dir,
1810                                 struct dentry *dentry, struct nameidata *nd)
1811 {
1812         return proc_pident_lookup(dir, dentry, tgid_attr_stuff);
1813 }
1814
1815 static struct dentry *proc_tid_attr_lookup(struct inode *dir,
1816                                 struct dentry *dentry, struct nameidata *nd)
1817 {
1818         return proc_pident_lookup(dir, dentry, tid_attr_stuff);
1819 }
1820
1821 static struct inode_operations proc_tgid_attr_inode_operations = {
1822         .lookup         = proc_tgid_attr_lookup,
1823 };
1824
1825 static struct inode_operations proc_tid_attr_inode_operations = {
1826         .lookup         = proc_tid_attr_lookup,
1827 };
1828 #endif
1829
1830 /*
1831  * /proc/self:
1832  */
1833 static int proc_self_readlink(struct dentry *dentry, char __user *buffer,
1834                               int buflen)
1835 {
1836         char tmp[30];
1837         sprintf(tmp, "%d", current->tgid);
1838         return vfs_readlink(dentry,buffer,buflen,tmp);
1839 }
1840
1841 static void *proc_self_follow_link(struct dentry *dentry, struct nameidata *nd)
1842 {
1843         char tmp[30];
1844         sprintf(tmp, "%d", current->tgid);
1845         return ERR_PTR(vfs_follow_link(nd,tmp));
1846 }
1847
1848 static struct inode_operations proc_self_inode_operations = {
1849         .readlink       = proc_self_readlink,
1850         .follow_link    = proc_self_follow_link,
1851 };
1852
1853 /**
1854  * proc_pid_unhash -  Unhash /proc/@pid entry from the dcache.
1855  * @p: task that should be flushed.
1856  *
1857  * Drops the /proc/@pid dcache entry from the hash chains.
1858  *
1859  * Dropping /proc/@pid entries and detach_pid must be synchroneous,
1860  * otherwise e.g. /proc/@pid/exe might point to the wrong executable,
1861  * if the pid value is immediately reused. This is enforced by
1862  * - caller must acquire spin_lock(p->proc_lock)
1863  * - must be called before detach_pid()
1864  * - proc_pid_lookup acquires proc_lock, and checks that
1865  *   the target is not dead by looking at the attach count
1866  *   of PIDTYPE_PID.
1867  */
1868
1869 struct dentry *proc_pid_unhash(struct task_struct *p)
1870 {
1871         struct dentry *proc_dentry;
1872
1873         proc_dentry = p->proc_dentry;
1874         if (proc_dentry != NULL) {
1875
1876                 spin_lock(&dcache_lock);
1877                 spin_lock(&proc_dentry->d_lock);
1878                 if (!d_unhashed(proc_dentry)) {
1879                         dget_locked(proc_dentry);
1880                         __d_drop(proc_dentry);
1881                         spin_unlock(&proc_dentry->d_lock);
1882                 } else {
1883                         spin_unlock(&proc_dentry->d_lock);
1884                         proc_dentry = NULL;
1885                 }
1886                 spin_unlock(&dcache_lock);
1887         }
1888         return proc_dentry;
1889 }
1890
1891 /**
1892  * proc_pid_flush - recover memory used by stale /proc/@pid/x entries
1893  * @proc_dentry: directoy to prune.
1894  *
1895  * Shrink the /proc directory that was used by the just killed thread.
1896  */
1897
1898 void proc_pid_flush(struct dentry *proc_dentry)
1899 {
1900         might_sleep();
1901         if(proc_dentry != NULL) {
1902                 shrink_dcache_parent(proc_dentry);
1903                 dput(proc_dentry);
1904         }
1905 }
1906
1907 /* SMP-safe */
1908 struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
1909 {
1910         struct task_struct *task;
1911         struct inode *inode;
1912         struct proc_inode *ei;
1913         unsigned tgid;
1914         int died;
1915
1916         if (dentry->d_name.len == 4 && !memcmp(dentry->d_name.name,"self",4)) {
1917                 inode = new_inode(dir->i_sb);
1918                 if (!inode)
1919                         return ERR_PTR(-ENOMEM);
1920                 ei = PROC_I(inode);
1921                 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1922                 inode->i_ino = fake_ino(0, PROC_TGID_INO);
1923                 ei->pde = NULL;
1924                 inode->i_mode = S_IFLNK|S_IRWXUGO;
1925                 inode->i_uid = inode->i_gid = 0;
1926                 inode->i_size = 64;
1927                 inode->i_op = &proc_self_inode_operations;
1928                 d_add(dentry, inode);
1929                 return NULL;
1930         }
1931         tgid = name_to_int(dentry);
1932         if (tgid == ~0U)
1933                 goto out;
1934
1935         read_lock(&tasklist_lock);
1936         task = find_task_by_pid(tgid);
1937         if (task)
1938                 get_task_struct(task);
1939         read_unlock(&tasklist_lock);
1940         if (!task)
1941                 goto out;
1942
1943         inode = proc_pid_make_inode(dir->i_sb, task, PROC_TGID_INO);
1944
1945
1946         if (!inode) {
1947                 put_task_struct(task);
1948                 goto out;
1949         }
1950         inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
1951         inode->i_op = &proc_tgid_base_inode_operations;
1952         inode->i_fop = &proc_tgid_base_operations;
1953         inode->i_flags|=S_IMMUTABLE;
1954 #ifdef CONFIG_SECURITY
1955         inode->i_nlink = 5;
1956 #else
1957         inode->i_nlink = 4;
1958 #endif
1959
1960         dentry->d_op = &pid_base_dentry_operations;
1961
1962         died = 0;
1963         d_add(dentry, inode);
1964         spin_lock(&task->proc_lock);
1965         task->proc_dentry = dentry;
1966         if (!pid_alive(task)) {
1967                 dentry = proc_pid_unhash(task);
1968                 died = 1;
1969         }
1970         spin_unlock(&task->proc_lock);
1971
1972         put_task_struct(task);
1973         if (died) {
1974                 proc_pid_flush(dentry);
1975                 goto out;
1976         }
1977         return NULL;
1978 out:
1979         return ERR_PTR(-ENOENT);
1980 }
1981
1982 /* SMP-safe */
1983 static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
1984 {
1985         struct task_struct *task;
1986         struct task_struct *leader = proc_task(dir);
1987         struct inode *inode;
1988         unsigned tid;
1989
1990         tid = name_to_int(dentry);
1991         if (tid == ~0U)
1992                 goto out;
1993
1994         read_lock(&tasklist_lock);
1995         task = find_task_by_pid(tid);
1996         if (task)
1997                 get_task_struct(task);
1998         read_unlock(&tasklist_lock);
1999         if (!task)
2000                 goto out;
2001         if (leader->tgid != task->tgid)
2002                 goto out_drop_task;
2003
2004         inode = proc_pid_make_inode(dir->i_sb, task, PROC_TID_INO);
2005
2006
2007         if (!inode)
2008                 goto out_drop_task;
2009         inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
2010         inode->i_op = &proc_tid_base_inode_operations;
2011         inode->i_fop = &proc_tid_base_operations;
2012         inode->i_flags|=S_IMMUTABLE;
2013 #ifdef CONFIG_SECURITY
2014         inode->i_nlink = 4;
2015 #else
2016         inode->i_nlink = 3;
2017 #endif
2018
2019         dentry->d_op = &pid_base_dentry_operations;
2020
2021         d_add(dentry, inode);
2022
2023         put_task_struct(task);
2024         return NULL;
2025 out_drop_task:
2026         put_task_struct(task);
2027 out:
2028         return ERR_PTR(-ENOENT);
2029 }
2030
2031 #define PROC_NUMBUF 10
2032 #define PROC_MAXPIDS 20
2033
2034 /*
2035  * Get a few tgid's to return for filldir - we need to hold the
2036  * tasklist lock while doing this, and we must release it before
2037  * we actually do the filldir itself, so we use a temp buffer..
2038  */
2039 static int get_tgid_list(int index, unsigned long version, unsigned int *tgids)
2040 {
2041         struct task_struct *p;
2042         int nr_tgids = 0;
2043
2044         index--;
2045         read_lock(&tasklist_lock);
2046         p = NULL;
2047         if (version) {
2048                 p = find_task_by_pid(version);
2049                 if (p && !thread_group_leader(p))
2050                         p = NULL;
2051         }
2052
2053         if (p)
2054                 index = 0;
2055         else
2056                 p = next_task(&init_task);
2057
2058         for ( ; p != &init_task; p = next_task(p)) {
2059                 int tgid = p->pid;
2060                 if (!pid_alive(p))
2061                         continue;
2062                 if (--index >= 0)
2063                         continue;
2064                 tgids[nr_tgids] = tgid;
2065                 nr_tgids++;
2066                 if (nr_tgids >= PROC_MAXPIDS)
2067                         break;
2068         }
2069         read_unlock(&tasklist_lock);
2070         return nr_tgids;
2071 }
2072
2073 /*
2074  * Get a few tid's to return for filldir - we need to hold the
2075  * tasklist lock while doing this, and we must release it before
2076  * we actually do the filldir itself, so we use a temp buffer..
2077  */
2078 static int get_tid_list(int index, unsigned int *tids, struct inode *dir)
2079 {
2080         struct task_struct *leader_task = proc_task(dir);
2081         struct task_struct *task = leader_task;
2082         int nr_tids = 0;
2083
2084         index -= 2;
2085         read_lock(&tasklist_lock);
2086         /*
2087          * The starting point task (leader_task) might be an already
2088          * unlinked task, which cannot be used to access the task-list
2089          * via next_thread().
2090          */
2091         if (pid_alive(task)) do {
2092                 int tid = task->pid;
2093
2094                 if (--index >= 0)
2095                         continue;
2096                 if (tids != NULL)
2097                         tids[nr_tids] = tid;
2098                 nr_tids++;
2099                 if (nr_tids >= PROC_MAXPIDS)
2100                         break;
2101         } while ((task = next_thread(task)) != leader_task);
2102         read_unlock(&tasklist_lock);
2103         return nr_tids;
2104 }
2105
2106 /* for the /proc/ directory itself, after non-process stuff has been done */
2107 int proc_pid_readdir(struct file * filp, void * dirent, filldir_t filldir)
2108 {
2109         unsigned int tgid_array[PROC_MAXPIDS];
2110         char buf[PROC_NUMBUF];
2111         unsigned int nr = filp->f_pos - FIRST_PROCESS_ENTRY;
2112         unsigned int nr_tgids, i;
2113         int next_tgid;
2114
2115         if (!nr) {
2116                 ino_t ino = fake_ino(0,PROC_TGID_INO);
2117                 if (filldir(dirent, "self", 4, filp->f_pos, ino, DT_LNK) < 0)
2118                         return 0;
2119                 filp->f_pos++;
2120                 nr++;
2121         }
2122
2123         /* f_version caches the tgid value that the last readdir call couldn't
2124          * return. lseek aka telldir automagically resets f_version to 0.
2125          */
2126         next_tgid = filp->f_version;
2127         filp->f_version = 0;
2128         for (;;) {
2129                 nr_tgids = get_tgid_list(nr, next_tgid, tgid_array);
2130                 if (!nr_tgids) {
2131                         /* no more entries ! */
2132                         break;
2133                 }
2134                 next_tgid = 0;
2135
2136                 /* do not use the last found pid, reserve it for next_tgid */
2137                 if (nr_tgids == PROC_MAXPIDS) {
2138                         nr_tgids--;
2139                         next_tgid = tgid_array[nr_tgids];
2140                 }
2141
2142                 for (i=0;i<nr_tgids;i++) {
2143                         int tgid = tgid_array[i];
2144                         ino_t ino = fake_ino(tgid,PROC_TGID_INO);
2145                         unsigned long j = PROC_NUMBUF;
2146
2147                         do
2148                                 buf[--j] = '0' + (tgid % 10);
2149                         while ((tgid /= 10) != 0);
2150
2151                         if (filldir(dirent, buf+j, PROC_NUMBUF-j, filp->f_pos, ino, DT_DIR) < 0) {
2152                                 /* returning this tgid failed, save it as the first
2153                                  * pid for the next readir call */
2154                                 filp->f_version = tgid_array[i];
2155                                 goto out;
2156                         }
2157                         filp->f_pos++;
2158                         nr++;
2159                 }
2160         }
2161 out:
2162         return 0;
2163 }
2164
2165 /* for the /proc/TGID/task/ directories */
2166 static int proc_task_readdir(struct file * filp, void * dirent, filldir_t filldir)
2167 {
2168         unsigned int tid_array[PROC_MAXPIDS];
2169         char buf[PROC_NUMBUF];
2170         unsigned int nr_tids, i;
2171         struct dentry *dentry = filp->f_dentry;
2172         struct inode *inode = dentry->d_inode;
2173         int retval = -ENOENT;
2174         ino_t ino;
2175         unsigned long pos = filp->f_pos;  /* avoiding "long long" filp->f_pos */
2176
2177         if (!pid_alive(proc_task(inode)))
2178                 goto out;
2179         retval = 0;
2180
2181         switch (pos) {
2182         case 0:
2183                 ino = inode->i_ino;
2184                 if (filldir(dirent, ".", 1, pos, ino, DT_DIR) < 0)
2185                         goto out;
2186                 pos++;
2187                 /* fall through */
2188         case 1:
2189                 ino = parent_ino(dentry);
2190                 if (filldir(dirent, "..", 2, pos, ino, DT_DIR) < 0)
2191                         goto out;
2192                 pos++;
2193                 /* fall through */
2194         }
2195
2196         nr_tids = get_tid_list(pos, tid_array, inode);
2197         inode->i_nlink = pos + nr_tids;
2198
2199         for (i = 0; i < nr_tids; i++) {
2200                 unsigned long j = PROC_NUMBUF;
2201                 int tid = tid_array[i];
2202
2203                 ino = fake_ino(tid,PROC_TID_INO);
2204
2205                 do
2206                         buf[--j] = '0' + (tid % 10);
2207                 while ((tid /= 10) != 0);
2208
2209                 if (filldir(dirent, buf+j, PROC_NUMBUF-j, pos, ino, DT_DIR) < 0)
2210                         break;
2211                 pos++;
2212         }
2213 out:
2214         filp->f_pos = pos;
2215         return retval;
2216 }