1 /* $NetBSD: linux_file64.c,v 1.30 2005/12/11 12:20:19 christos Exp $ */ 2 3 /*- 4 * Copyright (c) 1995, 1998, 2000 The NetBSD Foundation, Inc. 5 * All rights reserved. 6 * 7 * This code is derived from software contributed to The NetBSD Foundation 8 * by Frank van der Linden and Eric Haszlakiewicz. 9 * 10 * Redistribution and use in source and binary forms, with or without 11 * modification, are permitted provided that the following conditions 12 * are met: 13 * 1. Redistributions of source code must retain the above copyright 14 * notice, this list of conditions and the following disclaimer. 15 * 2. Redistributions in binary form must reproduce the above copyright 16 * notice, this list of conditions and the following disclaimer in the 17 * documentation and/or other materials provided with the distribution. 18 * 3. All advertising materials mentioning features or use of this software 19 * must display the following acknowledgement: 20 * This product includes software developed by the NetBSD 21 * Foundation, Inc. and its contributors. 22 * 4. Neither the name of The NetBSD Foundation nor the names of its 23 * contributors may be used to endorse or promote products derived 24 * from this software without specific prior written permission. 25 * 26 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 27 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 28 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 29 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 30 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 31 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 32 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 33 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 34 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 35 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 36 * POSSIBILITY OF SUCH DAMAGE. 37 */ 38 39 /* 40 * Linux 64bit filesystem calls. Used on 32bit archs, not used on 64bit ones. 41 */ 42 43 #include <sys/cdefs.h> 44 __KERNEL_RCSID(0, "$NetBSD: linux_file64.c,v 1.30 2005/12/11 12:20:19 christos Exp $"); 45 46 #include <sys/param.h> 47 #include <sys/systm.h> 48 #include <sys/namei.h> 49 #include <sys/proc.h> 50 #include <sys/dirent.h> 51 #include <sys/file.h> 52 #include <sys/stat.h> 53 #include <sys/filedesc.h> 54 #include <sys/ioctl.h> 55 #include <sys/kernel.h> 56 #include <sys/mount.h> 57 #include <sys/malloc.h> 58 #include <sys/vnode.h> 59 #include <sys/tty.h> 60 #include <sys/conf.h> 61 62 #include <sys/sa.h> 63 #include <sys/syscallargs.h> 64 65 #include <compat/linux/common/linux_types.h> 66 #include <compat/linux/common/linux_signal.h> 67 #include <compat/linux/common/linux_fcntl.h> 68 #include <compat/linux/common/linux_util.h> 69 #include <compat/linux/common/linux_machdep.h> 70 #include <compat/linux/common/linux_dirent.h> 71 72 #include <compat/linux/linux_syscallargs.h> 73 74 #ifndef alpha 75 76 static void bsd_to_linux_stat __P((struct stat *, struct linux_stat64 *)); 77 static int linux_do_stat64 __P((struct lwp *, void *, register_t *, int)); 78 79 /* 80 * Convert a NetBSD stat structure to a Linux stat structure. 81 * Only the order of the fields and the padding in the structure 82 * is different. linux_fakedev is a machine-dependent function 83 * which optionally converts device driver major/minor numbers 84 * (XXX horrible, but what can you do against code that compares 85 * things against constant major device numbers? sigh) 86 */ 87 static void 88 bsd_to_linux_stat(bsp, lsp) 89 struct stat *bsp; 90 struct linux_stat64 *lsp; 91 { 92 lsp->lst_dev = linux_fakedev(bsp->st_dev, 0); 93 lsp->lst_ino = bsp->st_ino; 94 lsp->lst_mode = (linux_mode_t)bsp->st_mode; 95 if (bsp->st_nlink >= (1 << 15)) 96 lsp->lst_nlink = (1 << 15) - 1; 97 else 98 lsp->lst_nlink = (linux_nlink_t)bsp->st_nlink; 99 lsp->lst_uid = bsp->st_uid; 100 lsp->lst_gid = bsp->st_gid; 101 lsp->lst_rdev = linux_fakedev(bsp->st_rdev, 1); 102 lsp->lst_size = bsp->st_size; 103 lsp->lst_blksize = bsp->st_blksize; 104 lsp->lst_blocks = bsp->st_blocks; 105 lsp->lst_atime = bsp->st_atime; 106 lsp->lst_mtime = bsp->st_mtime; 107 lsp->lst_ctime = bsp->st_ctime; 108 #ifdef LINUX_STAT64_HAS_NSEC 109 lsp->lst_atime_nsec = bsp->st_atimensec; 110 lsp->lst_mtime_nsec = bsp->st_mtimensec; 111 lsp->lst_ctime_nsec = bsp->st_ctimensec; 112 #endif 113 #if LINUX_STAT64_HAS_BROKEN_ST_INO 114 lsp->__lst_ino = (linux_ino_t) bsp->st_ino; 115 #endif 116 } 117 118 /* 119 * The stat functions below are plain sailing. stat and lstat are handled 120 * by one function to avoid code duplication. 121 */ 122 int 123 linux_sys_fstat64(l, v, retval) 124 struct lwp *l; 125 void *v; 126 register_t *retval; 127 { 128 struct linux_sys_fstat64_args /* { 129 syscallarg(int) fd; 130 syscallarg(struct linux_stat64 *) sp; 131 } */ *uap = v; 132 struct proc *p = l->l_proc; 133 struct sys___fstat30_args fsa; 134 struct linux_stat64 tmplst; 135 struct stat *st,tmpst; 136 caddr_t sg; 137 int error; 138 139 sg = stackgap_init(p, 0); 140 141 st = stackgap_alloc(p, &sg, sizeof (struct stat)); 142 143 SCARG(&fsa, fd) = SCARG(uap, fd); 144 SCARG(&fsa, sb) = st; 145 146 if ((error = sys___fstat30(l, &fsa, retval))) 147 return error; 148 149 if ((error = copyin(st, &tmpst, sizeof tmpst))) 150 return error; 151 152 bsd_to_linux_stat(&tmpst, &tmplst); 153 154 if ((error = copyout(&tmplst, SCARG(uap, sp), sizeof tmplst))) 155 return error; 156 157 return 0; 158 } 159 160 static int 161 linux_do_stat64(l, v, retval, dolstat) 162 struct lwp *l; 163 void *v; 164 register_t *retval; 165 int dolstat; 166 { 167 struct proc *p = l->l_proc; 168 struct sys___stat30_args sa; 169 struct linux_stat64 tmplst; 170 struct stat *st, tmpst; 171 caddr_t sg; 172 int error; 173 struct linux_sys_stat64_args *uap = v; 174 175 sg = stackgap_init(p, 0); 176 st = stackgap_alloc(p, &sg, sizeof (struct stat)); 177 CHECK_ALT_EXIST(l, &sg, SCARG(uap, path)); 178 179 SCARG(&sa, ub) = st; 180 SCARG(&sa, path) = SCARG(uap, path); 181 182 if ((error = (dolstat ? sys___lstat30(l, &sa, retval) : 183 sys___stat30(l, &sa, retval)))) 184 return error; 185 186 if ((error = copyin(st, &tmpst, sizeof tmpst))) 187 return error; 188 189 bsd_to_linux_stat(&tmpst, &tmplst); 190 191 if ((error = copyout(&tmplst, SCARG(uap, sp), sizeof tmplst))) 192 return error; 193 194 return 0; 195 } 196 197 int 198 linux_sys_stat64(l, v, retval) 199 struct lwp *l; 200 void *v; 201 register_t *retval; 202 { 203 struct linux_sys_stat64_args /* { 204 syscallarg(const char *) path; 205 syscallarg(struct linux_stat64 *) sp; 206 } */ *uap = v; 207 208 return linux_do_stat64(l, uap, retval, 0); 209 } 210 211 int 212 linux_sys_lstat64(l, v, retval) 213 struct lwp *l; 214 void *v; 215 register_t *retval; 216 { 217 struct linux_sys_lstat64_args /* { 218 syscallarg(const char *) path; 219 syscallarg(struct linux_stat64 *) sp; 220 } */ *uap = v; 221 222 return linux_do_stat64(l, uap, retval, 1); 223 } 224 225 int 226 linux_sys_truncate64(l, v, retval) 227 struct lwp *l; 228 void *v; 229 register_t *retval; 230 { 231 struct linux_sys_truncate64_args /* { 232 syscallarg(const char *) path; 233 syscallarg(off_t) length; 234 } */ *uap = v; 235 struct sys_truncate_args ta; 236 struct proc *p = l->l_proc; 237 caddr_t sg = stackgap_init(p, 0); 238 239 CHECK_ALT_EXIST(l, &sg, SCARG(uap, path)); 240 241 /* Linux doesn't have the 'pad' pseudo-parameter */ 242 SCARG(&ta, path) = SCARG(uap, path); 243 SCARG(&ta, pad) = 0; 244 SCARG(&ta, length) = SCARG(uap, length); 245 246 return sys_truncate(l, &ta, retval); 247 } 248 249 int 250 linux_sys_ftruncate64(l, v, retval) 251 struct lwp *l; 252 void *v; 253 register_t *retval; 254 { 255 struct linux_sys_ftruncate64_args /* { 256 syscallarg(unsigned int) fd; 257 syscallarg(off_t) length; 258 } */ *uap = v; 259 struct sys_ftruncate_args ta; 260 261 /* Linux doesn't have the 'pad' pseudo-parameter */ 262 SCARG(&ta, fd) = SCARG(uap, fd); 263 SCARG(&ta, pad) = 0; 264 SCARG(&ta, length) = SCARG(uap, length); 265 266 return sys_ftruncate(l, &ta, retval); 267 } 268 269 #if !defined(__m68k__) && !defined(__amd64__) 270 static void bsd_to_linux_flock64 __P((struct linux_flock64 *, 271 const struct flock *)); 272 static void linux_to_bsd_flock64 __P((struct flock *, 273 const struct linux_flock64 *)); 274 275 static void 276 bsd_to_linux_flock64(lfp, bfp) 277 struct linux_flock64 *lfp; 278 const struct flock *bfp; 279 { 280 281 lfp->l_start = bfp->l_start; 282 lfp->l_len = bfp->l_len; 283 lfp->l_pid = bfp->l_pid; 284 lfp->l_whence = bfp->l_whence; 285 switch (bfp->l_type) { 286 case F_RDLCK: 287 lfp->l_type = LINUX_F_RDLCK; 288 break; 289 case F_UNLCK: 290 lfp->l_type = LINUX_F_UNLCK; 291 break; 292 case F_WRLCK: 293 lfp->l_type = LINUX_F_WRLCK; 294 break; 295 } 296 } 297 298 static void 299 linux_to_bsd_flock64(bfp, lfp) 300 struct flock *bfp; 301 const struct linux_flock64 *lfp; 302 { 303 304 bfp->l_start = lfp->l_start; 305 bfp->l_len = lfp->l_len; 306 bfp->l_pid = lfp->l_pid; 307 bfp->l_whence = lfp->l_whence; 308 switch (lfp->l_type) { 309 case LINUX_F_RDLCK: 310 bfp->l_type = F_RDLCK; 311 break; 312 case LINUX_F_UNLCK: 313 bfp->l_type = F_UNLCK; 314 break; 315 case LINUX_F_WRLCK: 316 bfp->l_type = F_WRLCK; 317 break; 318 } 319 } 320 321 int 322 linux_sys_fcntl64(l, v, retval) 323 struct lwp *l; 324 void *v; 325 register_t *retval; 326 { 327 struct linux_sys_fcntl64_args /* { 328 syscallarg(int) fd; 329 syscallarg(int) cmd; 330 syscallarg(void *) arg; 331 } */ *uap = v; 332 struct proc *p = l->l_proc; 333 struct sys_fcntl_args fca; 334 struct linux_flock64 lfl; 335 struct flock bfl, *bfp; 336 int error; 337 caddr_t sg; 338 void *arg = SCARG(uap, arg); 339 int cmd = SCARG(uap, cmd); 340 int fd = SCARG(uap, fd); 341 342 switch (cmd) { 343 case LINUX_F_GETLK64: 344 sg = stackgap_init(p, 0); 345 bfp = (struct flock *) stackgap_alloc(p, &sg, sizeof *bfp); 346 if ((error = copyin(arg, &lfl, sizeof lfl)) != 0) 347 return error; 348 linux_to_bsd_flock64(&bfl, &lfl); 349 if ((error = copyout(&bfl, bfp, sizeof bfl)) != 0) 350 return error; 351 SCARG(&fca, fd) = fd; 352 SCARG(&fca, cmd) = F_GETLK; 353 SCARG(&fca, arg) = bfp; 354 if ((error = sys_fcntl(l, &fca, retval)) != 0) 355 return error; 356 if ((error = copyin(bfp, &bfl, sizeof bfl)) != 0) 357 return error; 358 bsd_to_linux_flock64(&lfl, &bfl); 359 return copyout(&lfl, arg, sizeof lfl); 360 case LINUX_F_SETLK64: 361 case LINUX_F_SETLKW64: 362 cmd = (cmd == LINUX_F_SETLK64 ? F_SETLK : F_SETLKW); 363 if ((error = copyin(arg, &lfl, sizeof lfl)) != 0) 364 return error; 365 linux_to_bsd_flock64(&bfl, &lfl); 366 sg = stackgap_init(p, 0); 367 bfp = (struct flock *) stackgap_alloc(p, &sg, sizeof *bfp); 368 if ((error = copyout(&bfl, bfp, sizeof bfl)) != 0) 369 return error; 370 SCARG(&fca, fd) = fd; 371 SCARG(&fca, cmd) = cmd; 372 SCARG(&fca, arg) = bfp; 373 return sys_fcntl(l, &fca, retval); 374 default: 375 return linux_sys_fcntl(l, v, retval); 376 } 377 } 378 #endif /* !m68k && !amd64 */ 379 380 #endif /* !alpha */ 381 382 /* 383 * Linux 'readdir' call. This code is mostly taken from the 384 * SunOS getdents call (see compat/sunos/sunos_misc.c), though 385 * an attempt has been made to keep it a little cleaner. 386 * 387 * The d_off field contains the offset of the next valid entry, 388 * unless the older Linux getdents(2), which used to have it set 389 * to the offset of the entry itself. This function also doesn't 390 * need to deal with the old count == 1 glibc problem. 391 * 392 * Read in BSD-style entries, convert them, and copy them out. 393 * 394 * Note that this doesn't handle union-mounted filesystems. 395 */ 396 int 397 linux_sys_getdents64(l, v, retval) 398 struct lwp *l; 399 void *v; 400 register_t *retval; 401 { 402 struct linux_sys_getdents_args /* { 403 syscallarg(int) fd; 404 syscallarg(struct linux_dirent64 *) dent; 405 syscallarg(unsigned int) count; 406 } */ *uap = v; 407 struct proc *p = l->l_proc; 408 struct dirent *bdp; 409 struct vnode *vp; 410 caddr_t inp, tbuf; /* BSD-format */ 411 int len, reclen; /* BSD-format */ 412 caddr_t outp; /* Linux-format */ 413 int resid, linux_reclen = 0; /* Linux-format */ 414 struct file *fp; 415 struct uio auio; 416 struct iovec aiov; 417 struct linux_dirent64 idb; 418 off_t off; /* true file offset */ 419 int buflen, error, eofflag, nbytes; 420 struct vattr va; 421 off_t *cookiebuf = NULL, *cookie; 422 int ncookies; 423 424 /* getvnode() will use the descriptor for us */ 425 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 426 return (error); 427 428 if ((fp->f_flag & FREAD) == 0) { 429 error = EBADF; 430 goto out1; 431 } 432 433 vp = (struct vnode *)fp->f_data; 434 if (vp->v_type != VDIR) { 435 error = EINVAL; 436 goto out1; 437 } 438 439 if ((error = VOP_GETATTR(vp, &va, p->p_ucred, l))) 440 goto out1; 441 442 nbytes = SCARG(uap, count); 443 buflen = min(MAXBSIZE, nbytes); 444 if (buflen < va.va_blocksize) 445 buflen = va.va_blocksize; 446 tbuf = malloc(buflen, M_TEMP, M_WAITOK); 447 448 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 449 off = fp->f_offset; 450 again: 451 aiov.iov_base = tbuf; 452 aiov.iov_len = buflen; 453 auio.uio_iov = &aiov; 454 auio.uio_iovcnt = 1; 455 auio.uio_rw = UIO_READ; 456 auio.uio_segflg = UIO_SYSSPACE; 457 auio.uio_lwp = NULL; 458 auio.uio_resid = buflen; 459 auio.uio_offset = off; 460 /* 461 * First we read into the malloc'ed buffer, then 462 * we massage it into user space, one record at a time. 463 */ 464 error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, &cookiebuf, 465 &ncookies); 466 if (error) 467 goto out; 468 469 inp = tbuf; 470 outp = (caddr_t)SCARG(uap, dent); 471 resid = nbytes; 472 if ((len = buflen - auio.uio_resid) == 0) 473 goto eof; 474 475 for (cookie = cookiebuf; len > 0; len -= reclen) { 476 bdp = (struct dirent *)inp; 477 reclen = bdp->d_reclen; 478 if (reclen & 3) 479 panic("linux_readdir"); 480 if (bdp->d_fileno == 0) { 481 inp += reclen; /* it is a hole; squish it out */ 482 if (cookie) 483 off = *cookie++; 484 else 485 off += reclen; 486 continue; 487 } 488 linux_reclen = LINUX_RECLEN(&idb, bdp->d_namlen); 489 if (reclen > len || resid < linux_reclen) { 490 /* entry too big for buffer, so just stop */ 491 outp++; 492 break; 493 } 494 if (cookie) 495 off = *cookie++; /* each entry points to next */ 496 else 497 off += reclen; 498 /* 499 * Massage in place to make a Linux-shaped dirent (otherwise 500 * we have to worry about touching user memory outside of 501 * the copyout() call). 502 */ 503 idb.d_ino = bdp->d_fileno; 504 idb.d_type = bdp->d_type; 505 idb.d_off = off; 506 idb.d_reclen = (u_short)linux_reclen; 507 strcpy(idb.d_name, bdp->d_name); 508 if ((error = copyout((caddr_t)&idb, outp, linux_reclen))) 509 goto out; 510 /* advance past this real entry */ 511 inp += reclen; 512 /* advance output past Linux-shaped entry */ 513 outp += linux_reclen; 514 resid -= linux_reclen; 515 } 516 517 /* if we squished out the whole block, try again */ 518 if (outp == (caddr_t)SCARG(uap, dent)) 519 goto again; 520 fp->f_offset = off; /* update the vnode offset */ 521 522 eof: 523 *retval = nbytes - resid; 524 out: 525 VOP_UNLOCK(vp, 0); 526 if (cookiebuf) 527 free(cookiebuf, M_TEMP); 528 free(tbuf, M_TEMP); 529 out1: 530 FILE_UNUSE(fp, l); 531 return error; 532 } 533