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