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