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/vs_network.h>
  36
  37 /*
  38  * For hysterical raisins we keep the same inumbers as in the old procfs.
  39  * Feel free to change the macro below - just keep the range distinct from
  40  * inumbers of the rest of procfs (currently those are in 0x0000--0xffff).
  41  * As soon as we'll get a separate superblock we will be able to forget
  42  * about magical ranges too.
  43  */
  44
  45 #define fake_ino(pid,ino) (((pid)<<16)|(ino))
  46
  47 enum pid_directory_inos {
  48         PROC_TGID_INO = 2,
  49         PROC_TGID_TASK,
  50         PROC_TGID_STATUS,
  51         PROC_TGID_MEM,
  52         PROC_TGID_CWD,
  53         PROC_TGID_ROOT,
  54         PROC_TGID_EXE,
  55         PROC_TGID_FD,
  56         PROC_TGID_ENVIRON,
  57         PROC_TGID_AUXV,
  58         PROC_TGID_CMDLINE,
  59         PROC_TGID_STAT,
  60         PROC_TGID_STATM,
  61         PROC_TGID_MAPS,
  62         PROC_TGID_MOUNTS,
  63         PROC_TGID_WCHAN,
  64 #ifdef CONFIG_SCHEDSTATS
  65         PROC_TGID_SCHEDSTAT,
  66 #endif
  67 #ifdef CONFIG_SECURITY
  68         PROC_TGID_ATTR,
  69         PROC_TGID_ATTR_CURRENT,
  70         PROC_TGID_ATTR_PREV,
  71         PROC_TGID_ATTR_EXEC,
  72         PROC_TGID_ATTR_FSCREATE,
  73 #endif
  74         PROC_TGID_VX_INFO,
  75         PROC_TGID_IP_INFO,
  76         PROC_TGID_FD_DIR,
  77         PROC_TID_INO,
  78         PROC_TID_STATUS,
  79         PROC_TID_MEM,
  80         PROC_TID_CWD,
  81         PROC_TID_ROOT,
  82         PROC_TID_EXE,
  83         PROC_TID_FD,
  84         PROC_TID_ENVIRON,
  85         PROC_TID_AUXV,
  86         PROC_TID_CMDLINE,
  87         PROC_TID_STAT,
  88         PROC_TID_STATM,
  89         PROC_TID_MAPS,
  90         PROC_TID_MOUNTS,
  91         PROC_TID_WCHAN,
  92 #ifdef CONFIG_SCHEDSTATS
  93         PROC_TID_SCHEDSTAT,
  94 #endif
  95 #ifdef CONFIG_SECURITY
  96         PROC_TID_ATTR,
  97         PROC_TID_ATTR_CURRENT,
  98         PROC_TID_ATTR_PREV,
  99         PROC_TID_ATTR_EXEC,
 100         PROC_TID_ATTR_FSCREATE,
 101 #endif
 102         PROC_TID_VX_INFO,
 103         PROC_TID_IP_INFO,
 104 #ifdef CONFIG_DELAY_ACCT
 105         PROC_TID_DELAY_ACCT,
 106         PROC_TGID_DELAY_ACCT,
 107 #endif
 108         PROC_TID_FD_DIR = 0x8000,       /* 0x8000-0xffff */
 109 };
 110
 111 struct pid_entry {
 112         int type;
 113         int len;
 114         char *name;
 115         mode_t mode;
 116 };
 117
 118 #define E(type,name,mode) {(type),sizeof(name)-1,(name),(mode)}
 119
 120 static struct pid_entry tgid_base_stuff[] = {
 121         E(PROC_TGID_TASK,      "task",    S_IFDIR|S_IRUGO|S_IXUGO),
 122         E(PROC_TGID_FD,        "fd",      S_IFDIR|S_IRUSR|S_IXUSR),
 123         E(PROC_TGID_ENVIRON,   "environ", S_IFREG|S_IRUSR),
 124         E(PROC_TGID_AUXV,      "auxv",    S_IFREG|S_IRUSR),
 125         E(PROC_TGID_STATUS,    "status",  S_IFREG|S_IRUGO),
 126         E(PROC_TGID_CMDLINE,   "cmdline", S_IFREG|S_IRUGO),
 127         E(PROC_TGID_STAT,      "stat",    S_IFREG|S_IRUGO),
 128         E(PROC_TGID_STATM,     "statm",   S_IFREG|S_IRUGO),
 129         E(PROC_TGID_MAPS,      "maps",    S_IFREG|S_IRUSR),
 130         E(PROC_TGID_MEM,       "mem",     S_IFREG|S_IRUSR|S_IWUSR),
 131         E(PROC_TGID_CWD,       "cwd",     S_IFLNK|S_IRWXUGO),
 132         E(PROC_TGID_ROOT,      "root",    S_IFLNK|S_IRWXUGO),
 133         E(PROC_TGID_EXE,       "exe",     S_IFLNK|S_IRWXUGO),
 134         E(PROC_TGID_MOUNTS,    "mounts",  S_IFREG|S_IRUGO),
 135 #ifdef CONFIG_SECURITY
 136         E(PROC_TGID_ATTR,      "attr",    S_IFDIR|S_IRUGO|S_IXUGO),
 137 #endif
 138 #ifdef CONFIG_DELAY_ACCT
 139         E(PROC_TGID_DELAY_ACCT,"delay",   S_IFREG|S_IRUGO),
 140 #endif
 141 #ifdef CONFIG_KALLSYMS
 142         E(PROC_TGID_WCHAN,     "wchan",   S_IFREG|S_IRUGO),
 143 #endif
 144 #ifdef CONFIG_SCHEDSTATS
 145         E(PROC_TGID_SCHEDSTAT, "schedstat", S_IFREG|S_IRUGO),
 146 #endif
 147         E(PROC_TGID_VX_INFO,   "vinfo",   S_IFREG|S_IRUGO),
 148         E(PROC_TGID_IP_INFO,   "ninfo",   S_IFREG|S_IRUGO),
 149 #ifdef CONFIG_SCHEDSTATS
 150         E(PROC_TGID_SCHEDSTAT, "schedstat", S_IFREG|S_IRUGO),
 151 #endif
 152         {0,0,NULL,0}
 153 };
 154 static struct pid_entry tid_base_stuff[] = {
 155         E(PROC_TID_FD,         "fd",      S_IFDIR|S_IRUSR|S_IXUSR),
 156         E(PROC_TID_ENVIRON,    "environ", S_IFREG|S_IRUSR),
 157         E(PROC_TID_AUXV,       "auxv",    S_IFREG|S_IRUSR),
 158         E(PROC_TID_STATUS,     "status",  S_IFREG|S_IRUGO),
 159         E(PROC_TID_CMDLINE,    "cmdline", S_IFREG|S_IRUGO),
 160         E(PROC_TID_STAT,       "stat",    S_IFREG|S_IRUGO),
 161         E(PROC_TID_STATM,      "statm",   S_IFREG|S_IRUGO),
 162         E(PROC_TID_MAPS,       "maps",    S_IFREG|S_IRUSR),
 163         E(PROC_TID_MEM,        "mem",     S_IFREG|S_IRUSR|S_IWUSR),
 164         E(PROC_TID_CWD,        "cwd",     S_IFLNK|S_IRWXUGO),
 165         E(PROC_TID_ROOT,       "root",    S_IFLNK|S_IRWXUGO),
 166         E(PROC_TID_EXE,        "exe",     S_IFLNK|S_IRWXUGO),
 167         E(PROC_TID_MOUNTS,     "mounts",  S_IFREG|S_IRUGO),
 168 #ifdef CONFIG_SECURITY
 169         E(PROC_TID_ATTR,       "attr",    S_IFDIR|S_IRUGO|S_IXUGO),
 170 #endif
 171 #ifdef CONFIG_DELAY_ACCT
 172         E(PROC_TGID_DELAY_ACCT,"delay",   S_IFREG|S_IRUGO),
 173 #endif
 174 #ifdef CONFIG_KALLSYMS
 175         E(PROC_TID_WCHAN,      "wchan",   S_IFREG|S_IRUGO),
 176 #endif
 177 #ifdef CONFIG_SCHEDSTATS
 178         E(PROC_TID_SCHEDSTAT, "schedstat",S_IFREG|S_IRUGO),
 179 #endif
 180         E(PROC_TID_VX_INFO,    "vinfo",   S_IFREG|S_IRUGO),
 181         E(PROC_TID_IP_INFO,    "ninfo",   S_IFREG|S_IRUGO),
 182 #ifdef CONFIG_SCHEDSTATS
 183         E(PROC_TID_SCHEDSTAT, "schedstat",S_IFREG|S_IRUGO),
 184 #endif
 185         {0,0,NULL,0}
 186 };
 187
 188 #ifdef CONFIG_SECURITY
 189 static struct pid_entry tgid_attr_stuff[] = {
 190         E(PROC_TGID_ATTR_CURRENT,  "current",  S_IFREG|S_IRUGO|S_IWUGO),
 191         E(PROC_TGID_ATTR_PREV,     "prev",     S_IFREG|S_IRUGO),
 192         E(PROC_TGID_ATTR_EXEC,     "exec",     S_IFREG|S_IRUGO|S_IWUGO),
 193         E(PROC_TGID_ATTR_FSCREATE, "fscreate", S_IFREG|S_IRUGO|S_IWUGO),
 194         {0,0,NULL,0}
 195 };
 196 static struct pid_entry tid_attr_stuff[] = {
 197         E(PROC_TID_ATTR_CURRENT,   "current",  S_IFREG|S_IRUGO|S_IWUGO),
 198         E(PROC_TID_ATTR_PREV,      "prev",     S_IFREG|S_IRUGO),
 199         E(PROC_TID_ATTR_EXEC,      "exec",     S_IFREG|S_IRUGO|S_IWUGO),
 200         E(PROC_TID_ATTR_FSCREATE,  "fscreate", S_IFREG|S_IRUGO|S_IWUGO),
 201         {0,0,NULL,0}
 202 };
 203 #endif
 204
 205 #undef E
 206
 207 static inline struct task_struct *proc_task(struct inode *inode)
 208 {
 209         return PROC_I(inode)->task;
 210 }
 211
 212 static inline int proc_type(struct inode *inode)
 213 {
 214         return PROC_I(inode)->type;
 215 }
 216
 217 int proc_pid_stat(struct task_struct*,char*);
 218 int proc_pid_status(struct task_struct*,char*);
 219 int proc_pid_statm(struct task_struct*,char*);
 220 #ifdef CONFIG_DELAY_ACCT
 221 int proc_pid_delay(struct task_struct*,char*);
 222 #endif
 223
 224 static int proc_fd_link(struct inode *inode, struct dentry **dentry, struct vfsmount **mnt)
 225 {
 226         struct task_struct *task = proc_task(inode);
 227         struct files_struct *files;
 228         struct file *file;
 229         int fd = proc_type(inode) - PROC_TID_FD_DIR;
 230
 231         files = get_files_struct(task);
 232         if (files) {
 233                 spin_lock(&files->file_lock);
 234                 file = fcheck_files(files, fd);
 235                 if (file) {
 236                         *mnt = mntget(file->f_vfsmnt);
 237                         *dentry = dget(file->f_dentry);
 238                         spin_unlock(&files->file_lock);
 239                         put_files_struct(files);
 240                         return 0;
 241                 }
 242                 spin_unlock(&files->file_lock);
 243                 put_files_struct(files);
 244         }
 245         return -ENOENT;
 246 }
 247
 248 static int proc_exe_link(struct inode *inode, struct dentry **dentry, struct vfsmount **mnt)
 249 {
 250         struct vm_area_struct * vma;
 251         int result = -ENOENT;
 252         struct task_struct *task = proc_task(inode);
 253         struct mm_struct * mm = get_task_mm(task);
 254
 255         if (!mm)
 256                 goto out;
 257         down_read(&mm->mmap_sem);
 258         vma = mm->mmap;
 259         while (vma) {
 260                 if ((vma->vm_flags & VM_EXECUTABLE) &&
 261                     vma->vm_file) {
 262                         *mnt = mntget(vma->vm_file->f_vfsmnt);
 263                         *dentry = dget(vma->vm_file->f_dentry);
 264                         result = 0;
 265                         break;
 266                 }
 267                 vma = vma->vm_next;
 268         }
 269         up_read(&mm->mmap_sem);
 270         mmput(mm);
 271 out:
 272         return result;
 273 }
 274
 275 static int proc_cwd_link(struct inode *inode, struct dentry **dentry, struct vfsmount **mnt)
 276 {
 277         struct fs_struct *fs;
 278         int result = -ENOENT;
 279         task_lock(proc_task(inode));
 280         fs = proc_task(inode)->fs;
 281         if(fs)
 282                 atomic_inc(&fs->count);
 283         task_unlock(proc_task(inode));
 284         if (fs) {
 285                 read_lock(&fs->lock);
 286                 *mnt = mntget(fs->pwdmnt);
 287                 *dentry = dget(fs->pwd);
 288                 read_unlock(&fs->lock);
 289                 result = 0;
 290                 put_fs_struct(fs);
 291         }
 292         return result;
 293 }
 294
 295 static int proc_root_link(struct inode *inode, struct dentry **dentry, struct vfsmount **mnt)
 296 {
 297         struct fs_struct *fs;
 298         int result = -ENOENT;
 299         task_lock(proc_task(inode));
 300         fs = proc_task(inode)->fs;
 301         if(fs)
 302                 atomic_inc(&fs->count);
 303         task_unlock(proc_task(inode));
 304         if (fs) {
 305                 read_lock(&fs->lock);
 306                 *mnt = mntget(fs->rootmnt);
 307                 *dentry = dget(fs->root);
 308                 read_unlock(&fs->lock);
 309                 result = 0;
 310                 put_fs_struct(fs);
 311         }
 312         return result;
 313 }
 314
 315 #define MAY_PTRACE(task) \
 316         (task == current || \
 317         (task->parent == current && \
 318         (task->ptrace & PT_PTRACED) && \
 319          (task->state == TASK_STOPPED || task->state == TASK_TRACED) && \
 320          security_ptrace(current,task) == 0))
 321
 322 static int may_ptrace_attach(struct task_struct *task)
 323 {
 324         int retval = 0;
 325
 326         task_lock(task);
 327
 328         if (!task->mm)
 329                 goto out;
 330         if (((current->uid != task->euid) ||
 331              (current->uid != task->suid) ||
 332              (current->uid != task->uid) ||
 333              (current->gid != task->egid) ||
 334              (current->gid != task->sgid) ||
 335              (current->gid != task->gid)) && !capable(CAP_SYS_PTRACE))
 336                 goto out;
 337         rmb();
 338         if (task->mm->dumpable != 1 && !capable(CAP_SYS_PTRACE))
 339                 goto out;
 340         if (security_ptrace(current, task))
 341                 goto out;
 342
 343         retval = 1;
 344 out:
 345         task_unlock(task);
 346         return retval;
 347 }
 348
 349 static int proc_pid_environ(struct task_struct *task, char * buffer)
 350 {
 351         int res = 0;
 352         struct mm_struct *mm = get_task_mm(task);
 353         if (mm) {
 354                 unsigned int len = mm->env_end - mm->env_start;
 355                 if (len > PAGE_SIZE)
 356                         len = PAGE_SIZE;
 357                 res = access_process_vm(task, mm->env_start, buffer, len, 0);
 358                 if (!may_ptrace_attach(task))
 359                         res = -ESRCH;
 360                 mmput(mm);
 361         }
 362         return res;
 363 }
 364
 365 static int proc_pid_cmdline(struct task_struct *task, char * buffer)
 366 {
 367         int res = 0;
 368         unsigned int len;
 369         struct mm_struct *mm = get_task_mm(task);
 370         if (!mm)
 371                 goto out;
 372         if (!mm->arg_end)
 373                 goto out_mm;    /* Shh! No looking before we're done */
 374
 375         len = mm->arg_end - mm->arg_start;
 376
 377         if (len > PAGE_SIZE)
 378                 len = PAGE_SIZE;
 379
 380         res = access_process_vm(task, mm->arg_start, buffer, len, 0);
 381
 382         // If the nul at the end of args has been overwritten, then
 383         // assume application is using setproctitle(3).
 384         if (res > 0 && buffer[res-1] != '\0' && len < PAGE_SIZE) {
 385                 len = strnlen(buffer, res);
 386                 if (len < res) {
 387                     res = len;
 388                 } else {
 389                         len = mm->env_end - mm->env_start;
 390                         if (len > PAGE_SIZE - res)
 391                                 len = PAGE_SIZE - res;
 392                         res += access_process_vm(task, mm->env_start, buffer+res, len, 0);
 393                         res = strnlen(buffer, res);
 394                 }
 395         }
 396 out_mm:
 397         mmput(mm);
 398 out:
 399         return res;
 400 }
 401
 402 static int proc_pid_auxv(struct task_struct *task, char *buffer)
 403 {
 404         int res = 0;
 405         struct mm_struct *mm = get_task_mm(task);
 406         if (mm) {
 407                 unsigned int nwords = 0;
 408                 do
 409                         nwords += 2;
 410                 while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */
 411                 res = nwords * sizeof(mm->saved_auxv[0]);
 412                 if (res > PAGE_SIZE)
 413                         res = PAGE_SIZE;
 414                 memcpy(buffer, mm->saved_auxv, res);
 415                 mmput(mm);
 416         }
 417         return res;
 418 }
 419
 420
 421 #ifdef CONFIG_KALLSYMS
 422 /*
 423  * Provides a wchan file via kallsyms in a proper one-value-per-file format.
 424  * Returns the resolved symbol.  If that fails, simply return the address.
 425  */
 426 static int proc_pid_wchan(struct task_struct *task, char *buffer)
 427 {
 428         char *modname;
 429         const char *sym_name;
 430         unsigned long wchan, size, offset;
 431         char namebuf[128];
 432
 433         wchan = get_wchan(task);
 434
 435         sym_name = kallsyms_lookup(wchan, &size, &offset, &modname, namebuf);
 436         if (sym_name)
 437                 return sprintf(buffer, "%s", sym_name);
 438         return sprintf(buffer, "%lu", wchan);
 439 }
 440 #endif /* CONFIG_KALLSYMS */
 441
 442 #ifdef CONFIG_SCHEDSTATS
 443 /*
 444  * Provides /proc/PID/schedstat
 445  */
 446 static int proc_pid_schedstat(struct task_struct *task, char *buffer)
 447 {
 448         return sprintf(buffer, "%lu %lu %lu\n",
 449                         task->sched_info.cpu_time,
 450                         task->sched_info.run_delay,
 451                         task->sched_info.pcnt);
 452 }
 453 #endif
 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 (vfs_permission(inode, mask) != 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 struct inode_operations proc_mem_inode_operations = {
 729         .permission     = proc_permission,
 730 };
 731
 732 static int proc_pid_follow_link(struct dentry *dentry, struct nameidata *nd)
 733 {
 734         struct inode *inode = dentry->d_inode;
 735         int error = -EACCES;
 736
 737         /* We don't need a base pointer in the /proc filesystem */
 738         path_release(nd);
 739
 740         if (current->fsuid != inode->i_uid && !capable(CAP_DAC_OVERRIDE))
 741                 goto out;
 742         error = proc_check_root(inode);
 743         if (error)
 744                 goto out;
 745
 746         error = PROC_I(inode)->op.proc_get_link(inode, &nd->dentry, &nd->mnt);
 747         nd->last_type = LAST_BIND;
 748 out:
 749         return error;
 750 }
 751
 752 static int do_proc_readlink(struct dentry *dentry, struct vfsmount *mnt,
 753                             char __user *buffer, int buflen)
 754 {
 755         struct inode * inode;
 756         char *tmp = (char*)__get_free_page(GFP_KERNEL), *path;
 757         int len;
 758
 759         if (!tmp)
 760                 return -ENOMEM;
 761
 762         inode = dentry->d_inode;
 763         path = d_path(dentry, mnt, tmp, PAGE_SIZE);
 764         len = PTR_ERR(path);
 765         if (IS_ERR(path))
 766                 goto out;
 767         len = tmp + PAGE_SIZE - 1 - path;
 768
 769         if (len > buflen)
 770                 len = buflen;
 771         if (copy_to_user(buffer, path, len))
 772                 len = -EFAULT;
 773  out:
 774         free_page((unsigned long)tmp);
 775         return len;
 776 }
 777
 778 static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
 779 {
 780         int error = -EACCES;
 781         struct inode *inode = dentry->d_inode;
 782         struct dentry *de;
 783         struct vfsmount *mnt = NULL;
 784
 785         lock_kernel();
 786
 787         if (current->fsuid != inode->i_uid && !capable(CAP_DAC_OVERRIDE))
 788                 goto out;
 789         error = proc_check_root(inode);
 790         if (error)
 791                 goto out;
 792
 793         error = PROC_I(inode)->op.proc_get_link(inode, &de, &mnt);
 794         if (error)
 795                 goto out;
 796
 797         error = do_proc_readlink(de, mnt, buffer, buflen);
 798         dput(de);
 799         mntput(mnt);
 800 out:
 801         unlock_kernel();
 802         return error;
 803 }
 804
 805 static struct inode_operations proc_pid_link_inode_operations = {
 806         .readlink       = proc_pid_readlink,
 807         .follow_link    = proc_pid_follow_link
 808 };
 809
 810 #define NUMBUF 10
 811
 812 static int proc_readfd(struct file * filp, void * dirent, filldir_t filldir)
 813 {
 814         struct inode *inode = filp->f_dentry->d_inode;
 815         struct task_struct *p = proc_task(inode);
 816         unsigned int fd, tid, ino;
 817         int retval;
 818         char buf[NUMBUF];
 819         struct files_struct * files;
 820
 821         retval = -ENOENT;
 822         if (!pid_alive(p))
 823                 goto out;
 824         retval = 0;
 825         tid = p->pid;
 826
 827         fd = filp->f_pos;
 828         switch (fd) {
 829                 case 0:
 830                         if (filldir(dirent, ".", 1, 0, inode->i_ino, DT_DIR) < 0)
 831                                 goto out;
 832                         filp->f_pos++;
 833                 case 1:
 834                         ino = fake_ino(tid, PROC_TID_INO);
 835                         if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0)
 836                                 goto out;
 837                         filp->f_pos++;
 838                 default:
 839                         files = get_files_struct(p);
 840                         if (!files)
 841                                 goto out;
 842                         spin_lock(&files->file_lock);
 843                         for (fd = filp->f_pos-2;
 844                              fd < files->max_fds;
 845                              fd++, filp->f_pos++) {
 846                                 unsigned int i,j;
 847
 848                                 if (!fcheck_files(files, fd))
 849                                         continue;
 850                                 spin_unlock(&files->file_lock);
 851
 852                                 j = NUMBUF;
 853                                 i = fd;
 854                                 do {
 855                                         j--;
 856                                         buf[j] = '0' + (i % 10);
 857                                         i /= 10;
 858                                 } while (i);
 859
 860                                 ino = fake_ino(tid, PROC_TID_FD_DIR + fd);
 861                                 if (filldir(dirent, buf+j, NUMBUF-j, fd+2, ino, DT_LNK) < 0) {
 862                                         spin_lock(&files->file_lock);
 863                                         break;
 864                                 }
 865                                 spin_lock(&files->file_lock);
 866                         }
 867                         spin_unlock(&files->file_lock);
 868                         put_files_struct(files);
 869         }
 870 out:
 871         return retval;
 872 }
 873
 874 static int proc_pident_readdir(struct file *filp,
 875                 void *dirent, filldir_t filldir,
 876                 struct pid_entry *ents, unsigned int nents)
 877 {
 878         int i;
 879         int pid;
 880         struct dentry *dentry = filp->f_dentry;
 881         struct inode *inode = dentry->d_inode;
 882         struct pid_entry *p;
 883         ino_t ino;
 884         int ret;
 885
 886         ret = -ENOENT;
 887         if (!pid_alive(proc_task(inode)))
 888                 goto out;
 889
 890         ret = 0;
 891         pid = proc_task(inode)->pid;
 892         i = filp->f_pos;
 893         switch (i) {
 894         case 0:
 895                 ino = inode->i_ino;
 896                 if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0)
 897                         goto out;
 898                 i++;
 899                 filp->f_pos++;
 900                 /* fall through */
 901         case 1:
 902                 ino = parent_ino(dentry);
 903                 if (filldir(dirent, "..", 2, i, ino, DT_DIR) < 0)
 904                         goto out;
 905                 i++;
 906                 filp->f_pos++;
 907                 /* fall through */
 908         default:
 909                 i -= 2;
 910                 if (i >= nents) {
 911                         ret = 1;
 912                         goto out;
 913                 }
 914                 p = ents + i;
 915                 while (p->name) {
 916                         if (filldir(dirent, p->name, p->len, filp->f_pos,
 917                                     fake_ino(pid, p->type), p->mode >> 12) < 0)
 918                                 goto out;
 919                         filp->f_pos++;
 920                         p++;
 921                 }
 922         }
 923
 924         ret = 1;
 925 out:
 926         return ret;
 927 }
 928
 929 static int proc_tgid_base_readdir(struct file * filp,
 930                              void * dirent, filldir_t filldir)
 931 {
 932         return proc_pident_readdir(filp,dirent,filldir,
 933                                    tgid_base_stuff,ARRAY_SIZE(tgid_base_stuff));
 934 }
 935
 936 static int proc_tid_base_readdir(struct file * filp,
 937                              void * dirent, filldir_t filldir)
 938 {
 939         return proc_pident_readdir(filp,dirent,filldir,
 940                                    tid_base_stuff,ARRAY_SIZE(tid_base_stuff));
 941 }
 942
 943 /* building an inode */
 944
 945 static int task_dumpable(struct task_struct *task)
 946 {
 947         int dumpable = 0;
 948         struct mm_struct *mm;
 949
 950         task_lock(task);
 951         mm = task->mm;
 952         if (mm)
 953                 dumpable = mm->dumpable;
 954         task_unlock(task);
 955         if(dumpable == 1)
 956                 return 1;
 957         return 0;
 958 }
 959
 960
 961 static struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task, int ino)
 962 {
 963         struct inode * inode;
 964         struct proc_inode *ei;
 965
 966         /* We need a new inode */
 967
 968         inode = new_inode(sb);
 969         if (!inode)
 970                 goto out;
 971
 972         /* Common stuff */
 973         ei = PROC_I(inode);
 974         ei->task = NULL;
 975         inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
 976         inode->i_ino = fake_ino(task->pid, ino);
 977
 978         if (!pid_alive(task))
 979                 goto out_unlock;
 980
 981         /*
 982          * grab the reference to task.
 983          */
 984         get_task_struct(task);
 985         ei->task = task;
 986         ei->type = ino;
 987         inode->i_uid = 0;
 988         inode->i_gid = 0;
 989         if (ino == PROC_TGID_INO || ino == PROC_TID_INO || task_dumpable(task)) {
 990                 inode->i_uid = task->euid;
 991                 inode->i_gid = task->egid;
 992         }
 993         inode->i_xid = vx_task_xid(task);
 994         security_task_to_inode(task, inode);
 995
 996 out:
 997         return inode;
 998
 999 out_unlock:
1000         ei->pde = NULL;
1001         iput(inode);
1002         return NULL;
1003 }
1004
1005 /* dentry stuff */
1006
1007 /*
1008  *      Exceptional case: normally we are not allowed to unhash a busy
1009  * directory. In this case, however, we can do it - no aliasing problems
1010  * due to the way we treat inodes.
1011  *
1012  * Rewrite the inode's ownerships here because the owning task may have
1013  * performed a setuid(), etc.
1014  */
1015 static int pid_revalidate(struct dentry *dentry, struct nameidata *nd)
1016 {
1017         struct inode *inode = dentry->d_inode;
1018         struct task_struct *task = proc_task(inode);
1019
1020         if (!vx_check(vx_task_xid(task), VX_WATCH|VX_IDENT))
1021                 goto out_drop;
1022         /* discard wrong fakeinit */
1023
1024         if (pid_alive(task)) {
1025                 if (proc_type(inode) == PROC_TGID_INO || proc_type(inode) == PROC_TID_INO || task_dumpable(task)) {
1026                         inode->i_uid = task->euid;
1027                         inode->i_gid = task->egid;
1028                 } else {
1029                         inode->i_uid = 0;
1030                         inode->i_gid = 0;
1031                 }
1032                 security_task_to_inode(task, inode);
1033                 return 1;
1034         }
1035 out_drop:
1036         d_drop(dentry);
1037         return 0;
1038 }
1039
1040 static int tid_fd_revalidate(struct dentry *dentry, struct nameidata *nd)
1041 {
1042         struct inode *inode = dentry->d_inode;
1043         struct task_struct *task = proc_task(inode);
1044         int fd = proc_type(inode) - PROC_TID_FD_DIR;
1045         struct files_struct *files;
1046
1047         files = get_files_struct(task);
1048         if (files) {
1049                 spin_lock(&files->file_lock);
1050                 if (fcheck_files(files, fd)) {
1051                         spin_unlock(&files->file_lock);
1052                         put_files_struct(files);
1053                         if (task_dumpable(task)) {
1054                                 inode->i_uid = task->euid;
1055                                 inode->i_gid = task->egid;
1056                         } else {
1057                                 inode->i_uid = 0;
1058                                 inode->i_gid = 0;
1059                         }
1060                         security_task_to_inode(task, inode);
1061                         return 1;
1062                 }
1063                 spin_unlock(&files->file_lock);
1064                 put_files_struct(files);
1065         }
1066         d_drop(dentry);
1067         return 0;
1068 }
1069
1070 static void pid_base_iput(struct dentry *dentry, struct inode *inode)
1071 {
1072         struct task_struct *task = proc_task(inode);
1073         spin_lock(&task->proc_lock);
1074         if (task->proc_dentry == dentry)
1075                 task->proc_dentry = NULL;
1076         spin_unlock(&task->proc_lock);
1077         iput(inode);
1078 }
1079
1080 static int pid_delete_dentry(struct dentry * dentry)
1081 {
1082         /* Is the task we represent dead?
1083          * If so, then don't put the dentry on the lru list,
1084          * kill it immediately.
1085          */
1086         return !pid_alive(proc_task(dentry->d_inode));
1087 }
1088
1089 static struct dentry_operations tid_fd_dentry_operations =
1090 {
1091         .d_revalidate   = tid_fd_revalidate,
1092         .d_delete       = pid_delete_dentry,
1093 };
1094
1095 static struct dentry_operations pid_dentry_operations =
1096 {
1097         .d_revalidate   = pid_revalidate,
1098         .d_delete       = pid_delete_dentry,
1099 };
1100
1101 static struct dentry_operations pid_base_dentry_operations =
1102 {
1103         .d_revalidate   = pid_revalidate,
1104         .d_iput         = pid_base_iput,
1105         .d_delete       = pid_delete_dentry,
1106 };
1107
1108 /* Lookups */
1109
1110 static unsigned name_to_int(struct dentry *dentry)
1111 {
1112         const char *name = dentry->d_name.name;
1113         int len = dentry->d_name.len;
1114         unsigned n = 0;
1115
1116         if (len > 1 && *name == '0')
1117                 goto out;
1118         while (len-- > 0) {
1119                 unsigned c = *name++ - '0';
1120                 if (c > 9)
1121                         goto out;
1122                 if (n >= (~0U-9)/10)
1123                         goto out;
1124                 n *= 10;
1125                 n += c;
1126         }
1127         return n;
1128 out:
1129         return ~0U;
1130 }
1131
1132 /* SMP-safe */
1133 static struct dentry *proc_lookupfd(struct inode * dir, struct dentry * dentry, struct nameidata *nd)
1134 {
1135         struct task_struct *task = proc_task(dir);
1136         unsigned fd = name_to_int(dentry);
1137         struct file * file;
1138         struct files_struct * files;
1139         struct inode *inode;
1140         struct proc_inode *ei;
1141
1142         if (fd == ~0U)
1143                 goto out;
1144         if (!pid_alive(task))
1145                 goto out;
1146
1147         inode = proc_pid_make_inode(dir->i_sb, task, PROC_TID_FD_DIR+fd);
1148         if (!inode)
1149                 goto out;
1150         ei = PROC_I(inode);
1151         files = get_files_struct(task);
1152         if (!files)
1153                 goto out_unlock;
1154         inode->i_mode = S_IFLNK;
1155         spin_lock(&files->file_lock);
1156         file = fcheck_files(files, fd);
1157         if (!file)
1158                 goto out_unlock2;
1159         if (file->f_mode & 1)
1160                 inode->i_mode |= S_IRUSR | S_IXUSR;
1161         if (file->f_mode & 2)
1162                 inode->i_mode |= S_IWUSR | S_IXUSR;
1163         spin_unlock(&files->file_lock);
1164         put_files_struct(files);
1165         inode->i_op = &proc_pid_link_inode_operations;
1166         inode->i_size = 64;
1167         ei->op.proc_get_link = proc_fd_link;
1168         dentry->d_op = &tid_fd_dentry_operations;
1169         d_add(dentry, inode);
1170         return NULL;
1171
1172 out_unlock2:
1173         spin_unlock(&files->file_lock);
1174         put_files_struct(files);
1175 out_unlock:
1176         iput(inode);
1177 out:
1178         return ERR_PTR(-ENOENT);
1179 }
1180
1181 static int proc_task_readdir(struct file * filp, void * dirent, filldir_t filldir);
1182 static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd);
1183
1184 static struct file_operations proc_fd_operations = {
1185         .read           = generic_read_dir,
1186         .readdir        = proc_readfd,
1187 };
1188
1189 static struct file_operations proc_task_operations = {
1190         .read           = generic_read_dir,
1191         .readdir        = proc_task_readdir,
1192 };
1193
1194 /*
1195  * proc directories can do almost nothing..
1196  */
1197 static struct inode_operations proc_fd_inode_operations = {
1198         .lookup         = proc_lookupfd,
1199         .permission     = proc_permission,
1200 };
1201
1202 static struct inode_operations proc_task_inode_operations = {
1203         .lookup         = proc_task_lookup,
1204         .permission     = proc_permission,
1205 };
1206
1207 #ifdef CONFIG_SECURITY
1208 static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
1209                                   size_t count, loff_t *ppos)
1210 {
1211         struct inode * inode = file->f_dentry->d_inode;
1212         unsigned long page;
1213         ssize_t length;
1214         struct task_struct *task = proc_task(inode);
1215
1216         if (count > PAGE_SIZE)
1217                 count = PAGE_SIZE;
1218         if (!(page = __get_free_page(GFP_KERNEL)))
1219                 return -ENOMEM;
1220
1221         length = security_getprocattr(task,
1222                                       (char*)file->f_dentry->d_name.name,
1223                                       (void*)page, count);
1224         if (length >= 0)
1225                 length = simple_read_from_buffer(buf, count, ppos, (char *)page, length);
1226         free_page(page);
1227         return length;
1228 }
1229
1230 static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
1231                                    size_t count, loff_t *ppos)
1232 {
1233         struct inode * inode = file->f_dentry->d_inode;
1234         char *page;
1235         ssize_t length;
1236         struct task_struct *task = proc_task(inode);
1237
1238         if (count > PAGE_SIZE)
1239                 count = PAGE_SIZE;
1240         if (*ppos != 0) {
1241                 /* No partial writes. */
1242                 return -EINVAL;
1243         }
1244         page = (char*)__get_free_page(GFP_USER);
1245         if (!page)
1246                 return -ENOMEM;
1247         length = -EFAULT;
1248         if (copy_from_user(page, buf, count))
1249                 goto out;
1250
1251         length = security_setprocattr(task,
1252                                       (char*)file->f_dentry->d_name.name,
1253                                       (void*)page, count);
1254 out:
1255         free_page((unsigned long) page);
1256         return length;
1257 }
1258
1259 static struct file_operations proc_pid_attr_operations = {
1260         .read           = proc_pid_attr_read,
1261         .write          = proc_pid_attr_write,
1262 };
1263
1264 static struct file_operations proc_tid_attr_operations;
1265 static struct inode_operations proc_tid_attr_inode_operations;
1266 static struct file_operations proc_tgid_attr_operations;
1267 static struct inode_operations proc_tgid_attr_inode_operations;
1268 #endif
1269
1270 /* SMP-safe */
1271 static struct dentry *proc_pident_lookup(struct inode *dir,
1272                                          struct dentry *dentry,
1273                                          struct pid_entry *ents)
1274 {
1275         struct inode *inode;
1276         int error;
1277         struct task_struct *task = proc_task(dir);
1278         struct pid_entry *p;
1279         struct proc_inode *ei;
1280
1281         error = -ENOENT;
1282         inode = NULL;
1283
1284         if (!pid_alive(task))
1285                 goto out;
1286
1287         for (p = ents; p->name; p++) {
1288                 if (p->len != dentry->d_name.len)
1289                         continue;
1290                 if (!memcmp(dentry->d_name.name, p->name, p->len))
1291                         break;
1292         }
1293         if (!p->name)
1294                 goto out;
1295
1296         error = -EINVAL;
1297         inode = proc_pid_make_inode(dir->i_sb, task, p->type);
1298         if (!inode)
1299                 goto out;
1300
1301         ei = PROC_I(inode);
1302         inode->i_mode = p->mode;
1303         /*
1304          * Yes, it does not scale. And it should not. Don't add
1305          * new entries into /proc/<tgid>/ without very good reasons.
1306          */
1307         switch(p->type) {
1308                 case PROC_TGID_TASK:
1309                         inode->i_nlink = 3;
1310                         inode->i_op = &proc_task_inode_operations;
1311                         inode->i_fop = &proc_task_operations;
1312                         break;
1313                 case PROC_TID_FD:
1314                 case PROC_TGID_FD:
1315                         inode->i_nlink = 2;
1316                         inode->i_op = &proc_fd_inode_operations;
1317                         inode->i_fop = &proc_fd_operations;
1318                         break;
1319                 case PROC_TID_EXE:
1320                 case PROC_TGID_EXE:
1321                         inode->i_op = &proc_pid_link_inode_operations;
1322                         ei->op.proc_get_link = proc_exe_link;
1323                         break;
1324                 case PROC_TID_CWD:
1325                 case PROC_TGID_CWD:
1326                         inode->i_op = &proc_pid_link_inode_operations;
1327                         ei->op.proc_get_link = proc_cwd_link;
1328                         break;
1329                 case PROC_TID_ROOT:
1330                 case PROC_TGID_ROOT:
1331                         inode->i_op = &proc_pid_link_inode_operations;
1332                         ei->op.proc_get_link = proc_root_link;
1333                         break;
1334                 case PROC_TID_ENVIRON:
1335                 case PROC_TGID_ENVIRON:
1336                         inode->i_fop = &proc_info_file_operations;
1337                         ei->op.proc_read = proc_pid_environ;
1338                         break;
1339                 case PROC_TID_AUXV:
1340                 case PROC_TGID_AUXV:
1341                         inode->i_fop = &proc_info_file_operations;
1342                         ei->op.proc_read = proc_pid_auxv;
1343                         break;
1344                 case PROC_TID_STATUS:
1345                 case PROC_TGID_STATUS:
1346                         inode->i_fop = &proc_info_file_operations;
1347                         ei->op.proc_read = proc_pid_status;
1348                         break;
1349                 case PROC_TID_STAT:
1350                 case PROC_TGID_STAT:
1351                         inode->i_fop = &proc_info_file_operations;
1352                         ei->op.proc_read = proc_pid_stat;
1353                         break;
1354                 case PROC_TID_CMDLINE:
1355                 case PROC_TGID_CMDLINE:
1356                         inode->i_fop = &proc_info_file_operations;
1357                         ei->op.proc_read = proc_pid_cmdline;
1358                         break;
1359                 case PROC_TID_STATM:
1360                 case PROC_TGID_STATM:
1361                         inode->i_fop = &proc_info_file_operations;
1362                         ei->op.proc_read = proc_pid_statm;
1363                         break;
1364                 case PROC_TID_MAPS:
1365                 case PROC_TGID_MAPS:
1366                         inode->i_fop = &proc_maps_operations;
1367                         break;
1368                 case PROC_TID_MEM:
1369                 case PROC_TGID_MEM:
1370                         inode->i_op = &proc_mem_inode_operations;
1371                         inode->i_fop = &proc_mem_operations;
1372                         break;
1373                 case PROC_TID_MOUNTS:
1374                 case PROC_TGID_MOUNTS:
1375                         inode->i_fop = &proc_mounts_operations;
1376                         break;
1377 #ifdef CONFIG_SECURITY
1378                 case PROC_TID_ATTR:
1379                         inode->i_nlink = 2;
1380                         inode->i_op = &proc_tid_attr_inode_operations;
1381                         inode->i_fop = &proc_tid_attr_operations;
1382                         break;
1383                 case PROC_TGID_ATTR:
1384                         inode->i_nlink = 2;
1385                         inode->i_op = &proc_tgid_attr_inode_operations;
1386                         inode->i_fop = &proc_tgid_attr_operations;
1387                         break;
1388                 case PROC_TID_ATTR_CURRENT:
1389                 case PROC_TGID_ATTR_CURRENT:
1390                 case PROC_TID_ATTR_PREV:
1391                 case PROC_TGID_ATTR_PREV:
1392                 case PROC_TID_ATTR_EXEC:
1393                 case PROC_TGID_ATTR_EXEC:
1394                 case PROC_TID_ATTR_FSCREATE:
1395                 case PROC_TGID_ATTR_FSCREATE:
1396                         inode->i_fop = &proc_pid_attr_operations;
1397                         break;
1398 #endif
1399 #ifdef CONFIG_KALLSYMS
1400                 case PROC_TID_WCHAN:
1401                 case PROC_TGID_WCHAN:
1402                         inode->i_fop = &proc_info_file_operations;
1403                         ei->op.proc_read = proc_pid_wchan;
1404                         break;
1405 #endif
1406 #ifdef CONFIG_SCHEDSTATS
1407                 case PROC_TID_SCHEDSTAT:
1408                 case PROC_TGID_SCHEDSTAT:
1409                         inode->i_fop = &proc_info_file_operations;
1410                         ei->op.proc_read = proc_pid_schedstat;
1411                         break;
1412 #endif
1413                 case PROC_TID_VX_INFO:
1414                 case PROC_TGID_VX_INFO:
1415                         inode->i_fop = &proc_info_file_operations;
1416                         ei->op.proc_read = proc_pid_vx_info;
1417                         break;
1418                 case PROC_TID_IP_INFO:
1419                 case PROC_TGID_IP_INFO:
1420                         inode->i_fop = &proc_info_file_operations;
1421                         ei->op.proc_read = proc_pid_nx_info;
1422                         break;
1423 #ifdef CONFIG_DELAY_ACCT
1424                 case PROC_TID_DELAY_ACCT:
1425                 case PROC_TGID_DELAY_ACCT:
1426                         inode->i_fop = &proc_info_file_operations;
1427                         ei->op.proc_read = proc_pid_delay;
1428                         break;
1429 #endif
1430 #ifdef CONFIG_SCHEDSTATS
1431                 case PROC_TID_SCHEDSTAT:
1432                 case PROC_TGID_SCHEDSTAT:
1433                         inode->i_fop = &proc_info_file_operations;
1434                         ei->op.proc_read = proc_pid_schedstat;
1435                         break;
1436 #endif
1437                 default:
1438                         printk("procfs: impossible type (%d)",p->type);
1439                         iput(inode);
1440                         return ERR_PTR(-EINVAL);
1441         }
1442         dentry->d_op = &pid_dentry_operations;
1443         d_add(dentry, inode);
1444         return NULL;
1445
1446 out:
1447         return ERR_PTR(error);
1448 }
1449
1450 static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
1451         return proc_pident_lookup(dir, dentry, tgid_base_stuff);
1452 }
1453
1454 static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
1455         return proc_pident_lookup(dir, dentry, tid_base_stuff);
1456 }
1457
1458 static struct file_operations proc_tgid_base_operations = {
1459         .read           = generic_read_dir,
1460         .readdir        = proc_tgid_base_readdir,
1461 };
1462
1463 static struct file_operations proc_tid_base_operations = {
1464         .read           = generic_read_dir,
1465         .readdir        = proc_tid_base_readdir,
1466 };
1467
1468 static struct inode_operations proc_tgid_base_inode_operations = {
1469         .lookup         = proc_tgid_base_lookup,
1470 };
1471
1472 static struct inode_operations proc_tid_base_inode_operations = {
1473         .lookup         = proc_tid_base_lookup,
1474 };
1475
1476 #ifdef CONFIG_SECURITY
1477 static int proc_tgid_attr_readdir(struct file * filp,
1478                              void * dirent, filldir_t filldir)
1479 {
1480         return proc_pident_readdir(filp,dirent,filldir,
1481                                    tgid_attr_stuff,ARRAY_SIZE(tgid_attr_stuff));
1482 }
1483
1484 static int proc_tid_attr_readdir(struct file * filp,
1485                              void * dirent, filldir_t filldir)
1486 {
1487         return proc_pident_readdir(filp,dirent,filldir,
1488                                    tid_attr_stuff,ARRAY_SIZE(tid_attr_stuff));
1489 }
1490
1491 static struct file_operations proc_tgid_attr_operations = {
1492         .read           = generic_read_dir,
1493         .readdir        = proc_tgid_attr_readdir,
1494 };
1495
1496 static struct file_operations proc_tid_attr_operations = {
1497         .read           = generic_read_dir,
1498         .readdir        = proc_tid_attr_readdir,
1499 };
1500
1501 static struct dentry *proc_tgid_attr_lookup(struct inode *dir,
1502                                 struct dentry *dentry, struct nameidata *nd)
1503 {
1504         return proc_pident_lookup(dir, dentry, tgid_attr_stuff);
1505 }
1506
1507 static struct dentry *proc_tid_attr_lookup(struct inode *dir,
1508                                 struct dentry *dentry, struct nameidata *nd)
1509 {
1510         return proc_pident_lookup(dir, dentry, tid_attr_stuff);
1511 }
1512
1513 static struct inode_operations proc_tgid_attr_inode_operations = {
1514         .lookup         = proc_tgid_attr_lookup,
1515 };
1516
1517 static struct inode_operations proc_tid_attr_inode_operations = {
1518         .lookup         = proc_tid_attr_lookup,
1519 };
1520 #endif
1521
1522 /*
1523  * /proc/self:
1524  */
1525 static int proc_self_readlink(struct dentry *dentry, char __user *buffer,
1526                               int buflen)
1527 {
1528         char tmp[30];
1529         sprintf(tmp, "%d", vx_map_pid(current->tgid));
1530         return vfs_readlink(dentry,buffer,buflen,tmp);
1531 }
1532
1533 static int proc_self_follow_link(struct dentry *dentry, struct nameidata *nd)
1534 {
1535         char tmp[30];
1536         sprintf(tmp, "%d", vx_map_pid(current->tgid));
1537         return vfs_follow_link(nd,tmp);
1538 }
1539
1540 static struct inode_operations proc_self_inode_operations = {
1541         .readlink       = proc_self_readlink,
1542         .follow_link    = proc_self_follow_link,
1543 };
1544
1545 /**
1546  * proc_pid_unhash -  Unhash /proc/<pid> entry from the dcache.
1547  * @p: task that should be flushed.
1548  *
1549  * Drops the /proc/<pid> dcache entry from the hash chains.
1550  *
1551  * Dropping /proc/<pid> entries and detach_pid must be synchroneous,
1552  * otherwise e.g. /proc/<pid>/exe might point to the wrong executable,
1553  * if the pid value is immediately reused. This is enforced by
1554  * - caller must acquire spin_lock(p->proc_lock)
1555  * - must be called before detach_pid()
1556  * - proc_pid_lookup acquires proc_lock, and checks that
1557  *   the target is not dead by looking at the attach count
1558  *   of PIDTYPE_PID.
1559  */
1560
1561 struct dentry *proc_pid_unhash(struct task_struct *p)
1562 {
1563         struct dentry *proc_dentry;
1564
1565         proc_dentry = p->proc_dentry;
1566         if (proc_dentry != NULL) {
1567
1568                 spin_lock(&dcache_lock);
1569                 if (!d_unhashed(proc_dentry)) {
1570                         dget_locked(proc_dentry);
1571                         __d_drop(proc_dentry);
1572                 } else
1573                         proc_dentry = NULL;
1574                 spin_unlock(&dcache_lock);
1575         }
1576         return proc_dentry;
1577 }
1578
1579 /**
1580  * proc_pid_flush - recover memory used by stale /proc/<pid>/x entries
1581  * @proc_entry: directoy to prune.
1582  *
1583  * Shrink the /proc directory that was used by the just killed thread.
1584  */
1585
1586 void proc_pid_flush(struct dentry *proc_dentry)
1587 {
1588         might_sleep();
1589         if(proc_dentry != NULL) {
1590                 shrink_dcache_parent(proc_dentry);
1591                 dput(proc_dentry);
1592         }
1593 }
1594
1595 /* SMP-safe */
1596 struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
1597 {
1598         struct task_struct *task;
1599         struct inode *inode;
1600         struct proc_inode *ei;
1601         unsigned tgid;
1602         int died;
1603
1604         if (dentry->d_name.len == 4 && !memcmp(dentry->d_name.name,"self",4)) {
1605                 inode = new_inode(dir->i_sb);
1606                 if (!inode)
1607                         return ERR_PTR(-ENOMEM);
1608                 ei = PROC_I(inode);
1609                 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1610                 inode->i_ino = fake_ino(0, PROC_TGID_INO);
1611                 ei->pde = NULL;
1612                 inode->i_mode = S_IFLNK|S_IRWXUGO;
1613                 inode->i_uid = inode->i_gid = 0;
1614                 inode->i_size = 64;
1615                 inode->i_op = &proc_self_inode_operations;
1616                 d_add(dentry, inode);
1617                 return NULL;
1618         }
1619         tgid = name_to_int(dentry);
1620         if (tgid == ~0U)
1621                 goto out;
1622
1623         read_lock(&tasklist_lock);
1624         task = find_task_by_pid(tgid);
1625         if (task)
1626                 get_task_struct(task);
1627         read_unlock(&tasklist_lock);
1628         if (!task)
1629                 goto out;
1630
1631         if (!vx_check(vx_task_xid(task), VX_WATCH|VX_IDENT))
1632                 goto out_drop_task;
1633
1634         inode = proc_pid_make_inode(dir->i_sb, task, PROC_TGID_INO);
1635         if (!inode)
1636                 goto out_drop_task;
1637
1638         inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
1639         inode->i_op = &proc_tgid_base_inode_operations;
1640         inode->i_fop = &proc_tgid_base_operations;
1641         inode->i_nlink = 3;
1642         inode->i_flags|=S_IMMUTABLE;
1643
1644         dentry->d_op = &pid_base_dentry_operations;
1645
1646         died = 0;
1647         d_add(dentry, inode);
1648         spin_lock(&task->proc_lock);
1649         task->proc_dentry = dentry;
1650         if (!pid_alive(task)) {
1651                 dentry = proc_pid_unhash(task);
1652                 died = 1;
1653         }
1654         spin_unlock(&task->proc_lock);
1655
1656         put_task_struct(task);
1657         if (died) {
1658                 proc_pid_flush(dentry);
1659                 goto out;
1660         }
1661         return NULL;
1662 out_drop_task:
1663         put_task_struct(task);
1664 out:
1665         return ERR_PTR(-ENOENT);
1666 }
1667
1668 /* SMP-safe */
1669 static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
1670 {
1671         struct task_struct *task;
1672         struct task_struct *leader = proc_task(dir);
1673         struct inode *inode;
1674         unsigned tid;
1675
1676         tid = name_to_int(dentry);
1677         if (tid == ~0U)
1678                 goto out;
1679         if (vx_current_initpid(tid))
1680                 goto out;
1681
1682         read_lock(&tasklist_lock);
1683         task = find_task_by_pid(tid);
1684         if (task)
1685                 get_task_struct(task);
1686         read_unlock(&tasklist_lock);
1687         if (!task)
1688                 goto out;
1689         if (leader->tgid != task->tgid)
1690                 goto out_drop_task;
1691
1692         if (!vx_check(vx_task_xid(task), VX_WATCH|VX_IDENT))
1693                 goto out_drop_task;
1694
1695         inode = proc_pid_make_inode(dir->i_sb, task, PROC_TID_INO);
1696         if (!inode)
1697                 goto out_drop_task;
1698
1699         inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
1700         inode->i_op = &proc_tid_base_inode_operations;
1701         inode->i_fop = &proc_tid_base_operations;
1702         inode->i_nlink = 3;
1703         inode->i_flags|=S_IMMUTABLE;
1704
1705         dentry->d_op = &pid_base_dentry_operations;
1706
1707         d_add(dentry, inode);
1708
1709         put_task_struct(task);
1710         return NULL;
1711 out_drop_task:
1712         put_task_struct(task);
1713 out:
1714         return ERR_PTR(-ENOENT);
1715 }
1716
1717 #define PROC_NUMBUF 10
1718 #define PROC_MAXPIDS 20
1719
1720 /*
1721  * Get a few tgid's to return for filldir - we need to hold the
1722  * tasklist lock while doing this, and we must release it before
1723  * we actually do the filldir itself, so we use a temp buffer..
1724  */
1725 static int get_tgid_list(int index, unsigned long version, unsigned int *tgids)
1726 {
1727         struct task_struct *p;
1728         int nr_tgids = 0;
1729
1730         index--;
1731         read_lock(&tasklist_lock);
1732         p = NULL;
1733         if (version) {
1734                 p = find_task_by_real_pid(version);
1735                 if (p && !thread_group_leader(p))
1736                         p = NULL;
1737         }
1738
1739         if (p)
1740                 index = 0;
1741         else
1742                 p = next_task(&init_task);
1743
1744         for ( ; p != &init_task; p = next_task(p)) {
1745                 int tgid = p->pid;
1746
1747                 if (!pid_alive(p))
1748                         continue;
1749                 if (!vx_check(vx_task_xid(p), VX_WATCH|VX_IDENT))
1750                         continue;
1751                 if (--index >= 0)
1752                         continue;
1753                 tgids[nr_tgids] = vx_map_tgid(tgid);
1754                 nr_tgids++;
1755                 if (nr_tgids >= PROC_MAXPIDS)
1756                         break;
1757         }
1758         read_unlock(&tasklist_lock);
1759         return nr_tgids;
1760 }
1761
1762 /*
1763  * Get a few tid's to return for filldir - we need to hold the
1764  * tasklist lock while doing this, and we must release it before
1765  * we actually do the filldir itself, so we use a temp buffer..
1766  */
1767 static int get_tid_list(int index, unsigned int *tids, struct inode *dir)
1768 {
1769         struct task_struct *leader_task = proc_task(dir);
1770         struct task_struct *task = leader_task;
1771         int nr_tids = 0;
1772
1773         index -= 2;
1774         read_lock(&tasklist_lock);
1775         /*
1776          * The starting point task (leader_task) might be an already
1777          * unlinked task, which cannot be used to access the task-list
1778          * via next_thread().
1779          */
1780         if (pid_alive(task)) do {
1781                 int tid = task->pid;
1782
1783                 if (!vx_check(vx_task_xid(task), VX_WATCH|VX_IDENT))
1784                         continue;
1785                 if (--index >= 0)
1786                         continue;
1787                 tids[nr_tids] = vx_map_pid(tid);
1788                 nr_tids++;
1789                 if (nr_tids >= PROC_MAXPIDS)
1790                         break;
1791         } while ((task = next_thread(task)) != leader_task);
1792         read_unlock(&tasklist_lock);
1793         return nr_tids;
1794 }
1795
1796 /* for the /proc/ directory itself, after non-process stuff has been done */
1797 int proc_pid_readdir(struct file * filp, void * dirent, filldir_t filldir)
1798 {
1799         unsigned int tgid_array[PROC_MAXPIDS];
1800         char buf[PROC_NUMBUF];
1801         unsigned int nr = filp->f_pos - FIRST_PROCESS_ENTRY;
1802         unsigned int nr_tgids, i;
1803         int next_tgid;
1804
1805         if (!nr) {
1806                 ino_t ino = fake_ino(0,PROC_TGID_INO);
1807                 if (filldir(dirent, "self", 4, filp->f_pos, ino, DT_LNK) < 0)
1808                         return 0;
1809                 filp->f_pos++;
1810                 nr++;
1811         }
1812
1813         /* f_version caches the tgid value that the last readdir call couldn't
1814          * return. lseek aka telldir automagically resets f_version to 0.
1815          */
1816         next_tgid = filp->f_version;
1817         filp->f_version = 0;
1818         for (;;) {
1819                 nr_tgids = get_tgid_list(nr, next_tgid, tgid_array);
1820                 if (!nr_tgids) {
1821                         /* no more entries ! */
1822                         break;
1823                 }
1824                 next_tgid = 0;
1825
1826                 /* do not use the last found pid, reserve it for next_tgid */
1827                 if (nr_tgids == PROC_MAXPIDS) {
1828                         nr_tgids--;
1829                         next_tgid = tgid_array[nr_tgids];
1830                 }
1831
1832                 for (i=0;i<nr_tgids;i++) {
1833                         int tgid = tgid_array[i];
1834                         ino_t ino = fake_ino(tgid,PROC_TGID_INO);
1835                         unsigned long j = PROC_NUMBUF;
1836
1837                         do
1838                                 buf[--j] = '0' + (tgid % 10);
1839                         while ((tgid /= 10) != 0);
1840
1841                         if (filldir(dirent, buf+j, PROC_NUMBUF-j, filp->f_pos, ino, DT_DIR) < 0) {
1842                                 /* returning this tgid failed, save it as the first
1843                                  * pid for the next readir call */
1844                                 filp->f_version = tgid_array[i];
1845                                 goto out;
1846                         }
1847                         filp->f_pos++;
1848                         nr++;
1849                 }
1850         }
1851 out:
1852         return 0;
1853 }
1854
1855 /* for the /proc/TGID/task/ directories */
1856 static int proc_task_readdir(struct file * filp, void * dirent, filldir_t filldir)
1857 {
1858         unsigned int tid_array[PROC_MAXPIDS];
1859         char buf[PROC_NUMBUF];
1860         unsigned int nr_tids, i;
1861         struct dentry *dentry = filp->f_dentry;
1862         struct inode *inode = dentry->d_inode;
1863         struct task_struct *task = proc_task(inode);
1864         int retval = -ENOENT;
1865         ino_t ino;
1866         unsigned long pos = filp->f_pos;  /* avoiding "long long" filp->f_pos */
1867
1868         if (!vx_check(vx_task_xid(task), VX_WATCH|VX_IDENT))
1869                 goto out;
1870         if (!pid_alive(task))
1871                 goto out;
1872         retval = 0;
1873
1874         switch (pos) {
1875         case 0:
1876                 ino = inode->i_ino;
1877                 if (filldir(dirent, ".", 1, pos, ino, DT_DIR) < 0)
1878                         goto out;
1879                 pos++;
1880                 /* fall through */
1881         case 1:
1882                 ino = parent_ino(dentry);
1883                 if (filldir(dirent, "..", 2, pos, ino, DT_DIR) < 0)
1884                         goto out;
1885                 pos++;
1886                 /* fall through */
1887         }
1888
1889         nr_tids = get_tid_list(pos, tid_array, inode);
1890
1891         for (i = 0; i < nr_tids; i++) {
1892                 unsigned long j = PROC_NUMBUF;
1893                 int tid = tid_array[i];
1894
1895                 ino = fake_ino(tid,PROC_TID_INO);
1896
1897                 do
1898                         buf[--j] = '0' + (tid % 10);
1899                 while ((tid /= 10) != 0);
1900
1901                 if (filldir(dirent, buf+j, PROC_NUMBUF-j, pos, ino, DT_DIR) < 0)
1902                         break;
1903                 pos++;
1904         }
1905 out:
1906         filp->f_pos = pos;
1907         return retval;
1908 }