gen/FreeBSD/opendir.c

   1 /*-
   2  * Copyright (c) 1983, 1993
   3  *      The Regents of the University of California.  All rights reserved.
   4  *
   5  * Redistribution and use in source and binary forms, with or without
   6  * modification, are permitted provided that the following conditions
   7  * are met:
   8  * 1. Redistributions of source code must retain the above copyright
   9  *    notice, this list of conditions and the following disclaimer.
  10  * 2. Redistributions in binary form must reproduce the above copyright
  11  *    notice, this list of conditions and the following disclaimer in the
  12  *    documentation and/or other materials provided with the distribution.
  13  * 4. Neither the name of the University nor the names of its contributors
  14  *    may be used to endorse or promote products derived from this software
  15  *    without specific prior written permission.
  16  *
  17  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
  18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  27  * SUCH DAMAGE.
  28  */
  29
  30 #if defined(LIBC_SCCS) && !defined(lint)
  31 static char sccsid[] = "@(#)opendir.c   8.8 (Berkeley) 5/1/95";
  32 #endif /* LIBC_SCCS and not lint */
  33 #include <sys/cdefs.h>
  34 __FBSDID("$FreeBSD$");
  35
  36 #include "namespace.h"
  37 #include <sys/param.h>
  38 #include <sys/mount.h>
  39 #include <sys/stat.h>
  40 #include <sys/sysctl.h>
  41
  42 #include <dirent.h>
  43 #include <errno.h>
  44 #include <fcntl.h>
  45 #include <stdlib.h>
  46 #include <string.h>
  47 #include <unistd.h>
  48 #include <pthread.h>
  49 #include "un-namespace.h"
  50
  51 #include "telldir.h"
  52
  53 static bool
  54 __kernel_supports_unionfs(void)
  55 {
  56         static int8_t kernel_supports_unionfs = -1;
  57         if (kernel_supports_unionfs == -1) {
  58                 int value = 0;
  59                 size_t len = sizeof(value);
  60                 sysctlbyname("kern.secure_kernel", &value, &len, NULL, 0);
  61                 kernel_supports_unionfs = !value;
  62         }
  63         return kernel_supports_unionfs;
  64 }
  65
  66 static int
  67 __fd_is_on_union_mount(int fd)
  68 {
  69         struct statfs stbuf;
  70         int rc;
  71
  72         rc = fstatfs(fd, &stbuf);
  73         if (rc < 0) {
  74                 return rc;
  75         }
  76         return (stbuf.f_flags & MNT_UNION) != 0;
  77 }
  78
  79 static DIR * __opendir_common(int, int, bool);
  80
  81 /*
  82  * Open a directory.
  83  */
  84 DIR *
  85 opendir(const char *name)
  86 {
  87
  88         return (__opendir2(name, DTF_HIDEW|DTF_NODUP));
  89 }
  90
  91 /*
  92  * Open a directory with existing file descriptor.
  93  */
  94 DIR *
  95 fdopendir(int fd)
  96 {
  97         struct stat statb;
  98
  99         /* Check that fd is associated with a directory. */
 100         if (_fstat(fd, &statb) != 0)
 101                 return (NULL);
 102         if (!S_ISDIR(statb.st_mode)) {
 103                 errno = ENOTDIR;
 104                 return (NULL);
 105         }
 106         /* Make sure CLOEXEC is set on the fd */
 107         if (_fcntl(fd, F_SETFD, FD_CLOEXEC) == -1)
 108                 return (NULL);
 109         return (__opendir_common(fd, DTF_HIDEW|DTF_NODUP, true));
 110 }
 111
 112 DIR *
 113 __opendir2(const char *name, int flags)
 114 {
 115         int fd;
 116         DIR *dir;
 117         int saved_errno;
 118
 119         if ((flags & (__DTF_READALL | __DTF_SKIPREAD)) != 0)
 120                 return (NULL);
 121         if ((fd = _open(name,
 122             O_RDONLY | O_NONBLOCK | O_DIRECTORY | O_CLOEXEC)) == -1)
 123                 return (NULL);
 124
 125         dir = __opendir_common(fd, flags, false);
 126         if (dir == NULL) {
 127                 saved_errno = errno;
 128                 _close(fd);
 129                 errno = saved_errno;
 130         }
 131         return (dir);
 132 }
 133
 134 static int
 135 opendir_compar(const void *p1, const void *p2)
 136 {
 137
 138         return (strcmp((*(const struct dirent **)p1)->d_name,
 139             (*(const struct dirent **)p2)->d_name));
 140 }
 141
 142 /*
 143  * For a directory at the top of a unionfs stack, the entire directory's
 144  * contents are read and cached locally until the next call to rewinddir().
 145  * For the fdopendir() case, the initial seek position must be preserved.
 146  * For rewinddir(), the full directory should always be re-read from the
 147  * beginning.
 148  *
 149  * If an error occurs, the existing buffer and state of 'dirp' is left
 150  * unchanged.
 151  */
 152 bool
 153 _filldir(DIR *dirp, bool use_current_pos)
 154 {
 155         struct dirent **dpv;
 156         char *buf, *ddptr, *ddeptr;
 157         off_t pos;
 158         int fd2, incr, len, n, saved_errno, space;
 159
 160         len = 0;
 161         space = 0;
 162         buf = NULL;
 163         ddptr = NULL;
 164
 165         /*
 166          * Use the system page size if that is a multiple of DIRBLKSIZ.
 167          * Hopefully this can be a big win someday by allowing page
 168          * trades to user space to be done by _getdirentries().
 169          */
 170         incr = getpagesize();
 171         if ((incr % DIRBLKSIZ) != 0)
 172                 incr = DIRBLKSIZ;
 173
 174         /*
 175          * The strategy here is to read all the directory
 176          * entries into a buffer, sort the buffer, and
 177          * remove duplicate entries by setting the inode
 178          * number to zero.
 179          *
 180          * We reopen the directory because _getdirentries()
 181          * on a MNT_UNION mount modifies the open directory,
 182          * making it refer to the lower directory after the
 183          * upper directory's entries are exhausted.
 184          * This would otherwise break software that uses
 185          * the directory descriptor for fchdir or *at
 186          * functions, such as fts.c.
 187          */
 188         if ((fd2 = openat(dirp->dd_fd, ".", O_RDONLY | O_CLOEXEC)) == -1)
 189                 return (false);
 190
 191         if (use_current_pos) {
 192                 pos = lseek(dirp->dd_fd, 0, SEEK_CUR);
 193                 if (pos == -1 || lseek(fd2, pos, SEEK_SET) == -1) {
 194                         saved_errno = errno;
 195                         _close(fd2);
 196                         errno = saved_errno;
 197                         return (false);
 198                 }
 199         }
 200
 201         do {
 202                 /*
 203                  * Always make at least DIRBLKSIZ bytes
 204                  * available to _getdirentries
 205                  */
 206                 if (space < DIRBLKSIZ) {
 207                         space += incr;
 208                         len += incr;
 209                         buf = reallocf(buf, len);
 210                         if (buf == NULL) {
 211                                 saved_errno = errno;
 212                                 _close(fd2);
 213                                 errno = saved_errno;
 214                                 return (false);
 215                         }
 216                         ddptr = buf + (len - space);
 217                 }
 218
 219 #if __DARWIN_64_BIT_INO_T
 220                 n = (int)__getdirentries64(fd2, ddptr, space, &dirp->dd_td->seekoff);
 221 #else /* !__DARWIN_64_BIT_INO_T */
 222                 n = _getdirentries(fd2, ddptr, space, &dirp->dd_seek);
 223 #endif /* __DARWIN_64_BIT_INO_T */
 224                 if (n > 0) {
 225                         ddptr += n;
 226                         space -= n;
 227                 }
 228                 if (n < 0) {
 229                         saved_errno = errno;
 230                         _close(fd2);
 231                         errno = saved_errno;
 232                         return (false);
 233                 }
 234         } while (n > 0);
 235         _close(fd2);
 236
 237         ddeptr = ddptr;
 238
 239         /*
 240          * There is now a buffer full of (possibly) duplicate
 241          * names.
 242          */
 243         dirp->dd_buf = buf;
 244
 245         /*
 246          * Go round this loop twice...
 247          *
 248          * Scan through the buffer, counting entries.
 249          * On the second pass, save pointers to each one.
 250          * Then sort the pointers and remove duplicate names.
 251          */
 252         for (dpv = NULL;;) {
 253                 n = 0;
 254                 ddptr = buf;
 255                 while (ddptr < ddeptr) {
 256                         struct dirent *dp;
 257
 258                         dp = (struct dirent *) ddptr;
 259                         if ((long)dp & 03L)
 260                                 break;
 261                         if ((dp->d_reclen <= 0) ||
 262                             (dp->d_reclen > (ddeptr + 1 - ddptr)))
 263                                 break;
 264                         ddptr += dp->d_reclen;
 265                         if (dp->d_fileno) {
 266                                 if (dpv)
 267                                         dpv[n] = dp;
 268                                 n++;
 269                         }
 270                 }
 271
 272                 if (dpv) {
 273                         struct dirent *xp;
 274
 275                         /*
 276                          * This sort must be stable.
 277                          */
 278                         mergesort(dpv, n, sizeof(*dpv), opendir_compar);
 279
 280                         dpv[n] = NULL;
 281                         xp = NULL;
 282
 283                         /*
 284                          * Scan through the buffer in sort order,
 285                          * zapping the inode number of any
 286                          * duplicate names.
 287                          */
 288                         for (n = 0; dpv[n]; n++) {
 289                                 struct dirent *dp = dpv[n];
 290
 291                                 if ((xp == NULL) ||
 292                                     strcmp(dp->d_name, xp->d_name)) {
 293                                         xp = dp;
 294                                 } else {
 295                                         dp->d_fileno = 0;
 296                                 }
 297                                 if (dp->d_type == DT_WHT &&
 298                                     (dirp->dd_flags & DTF_HIDEW))
 299                                         dp->d_fileno = 0;
 300                         }
 301
 302                         free(dpv);
 303                         break;
 304                 } else {
 305                         dpv = malloc((n+1) * sizeof(struct dirent *));
 306                         if (dpv == NULL)
 307                                 break;
 308                 }
 309         }
 310
 311         dirp->dd_len = len;
 312         dirp->dd_size = ddptr - dirp->dd_buf;
 313         return (true);
 314 }
 315
 316
 317 /*
 318  * Common routine for opendir(3), __opendir2(3) and fdopendir(3).
 319  */
 320 static DIR *
 321 __opendir_common(int fd, int flags, bool use_current_pos)
 322 {
 323         DIR *dirp;
 324         int saved_errno;
 325         int unionstack;
 326
 327         if ((dirp = malloc(sizeof(DIR) + sizeof(struct _telldir))) == NULL)
 328                 return (NULL);
 329
 330         dirp->dd_buf = NULL;
 331         dirp->dd_fd = fd;
 332         dirp->dd_flags = flags;
 333         dirp->dd_loc = 0;
 334         dirp->dd_lock = (pthread_mutex_t)PTHREAD_MUTEX_INITIALIZER;
 335         dirp->dd_td = (struct _telldir *)((char *)dirp + sizeof(DIR));
 336         LIST_INIT(&dirp->dd_td->td_locq);
 337         dirp->dd_td->td_loccnt = 0;
 338
 339         /*
 340          * Determine whether this directory is the top of a union stack.
 341          */
 342         if ((flags & DTF_NODUP) && __kernel_supports_unionfs()) {
 343                 unionstack = __fd_is_on_union_mount(fd);
 344                 if (unionstack < 0)
 345                         goto fail;
 346         } else {
 347                 unionstack = 0;
 348         }
 349
 350         if (unionstack) {
 351                 if (!_filldir(dirp, use_current_pos))
 352                         goto fail;
 353                 dirp->dd_flags |= __DTF_READALL;
 354         } else {
 355                 /*
 356                  * Start with a small-ish size to avoid allocating full pages.
 357                  * readdir() will allocate a larger buffer if it didn't fit
 358                  * to stay fast for large directories.
 359                  */
 360                 _Static_assert(GETDIRENTRIES64_EXTENDED_BUFSIZE <= READDIR_INITIAL_SIZE,
 361                     "Make sure we'll get extended metadata");
 362                 dirp->dd_len = READDIR_INITIAL_SIZE;
 363                 dirp->dd_buf = malloc(dirp->dd_len);
 364                 if (dirp->dd_buf == NULL)
 365                         goto fail;
 366                 if (use_current_pos) {
 367                         /*
 368                          * Read the first batch of directory entries
 369                          * to prime dd_seek.  This also checks if the
 370                          * fd passed to fdopendir() is a directory.
 371                          */
 372 #if __DARWIN_64_BIT_INO_T
 373                         /*
 374                          * sufficiently recent kernels when the buffer is large enough,
 375                          * will use the last bytes of the buffer to return status.
 376                          *
 377                          * To support older kernels:
 378                          * - make sure it's 0 initialized
 379                          * - make sure it's past `dd_size` before reading it
 380                          */
 381                         getdirentries64_flags_t *gdeflags =
 382                             (getdirentries64_flags_t *)(dirp->dd_buf + dirp->dd_len -
 383                             sizeof(getdirentries64_flags_t));
 384                         *gdeflags = 0;
 385                         dirp->dd_size = (long)__getdirentries64(dirp->dd_fd,
 386                             dirp->dd_buf, dirp->dd_len, &dirp->dd_td->seekoff);
 387                         if (dirp->dd_size >= 0 &&
 388                             dirp->dd_size <= dirp->dd_len - sizeof(getdirentries64_flags_t)) {
 389                                 if (*gdeflags & GETDIRENTRIES64_EOF) {
 390                                         dirp->dd_flags |= __DTF_ATEND;
 391                                 }
 392                         }
 393 #else /* !__DARWIN_64_BIT_INO_T */
 394                         dirp->dd_size = _getdirentries(dirp->dd_fd,
 395                             dirp->dd_buf, dirp->dd_len, &dirp->dd_seek);
 396 #endif /* __DARWIN_64_BIT_INO_T */
 397                         if (dirp->dd_size < 0) {
 398                                 if (errno == EINVAL)
 399                                         errno = ENOTDIR;
 400                                 goto fail;
 401                         }
 402                         dirp->dd_flags |= __DTF_SKIPREAD;
 403                 } else {
 404                         dirp->dd_size = 0;
 405 #if __DARWIN_64_BIT_INO_T
 406                         dirp->dd_td->seekoff = 0;
 407 #else /* !__DARWIN_64_BIT_INO_T */
 408                         dirp->dd_seek = 0;
 409 #endif /* __DARWIN_64_BIT_INO_T */
 410                 }
 411         }
 412
 413         return (dirp);
 414
 415 fail:
 416         saved_errno = errno;
 417         free(dirp->dd_buf);
 418         free(dirp);
 419         errno = saved_errno;
 420         return (NULL);
 421 }