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