1 /* $NetBSD: linux_file64.c,v 1.22 2003/08/10 20:16:27 jdolecek 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 1.22 2003/08/10 20:16:27 jdolecek 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 #if LINUX_STAT64_HAS_BROKEN_ST_INO 109 lsp->__lst_ino = (linux_ino_t) bsp->st_ino; 110 #endif 111 } 112 113 /* 114 * The stat functions below are plain sailing. stat and lstat are handled 115 * by one function to avoid code duplication. 116 */ 117 int 118 linux_sys_fstat64(l, v, retval) 119 struct lwp *l; 120 void *v; 121 register_t *retval; 122 { 123 struct linux_sys_fstat64_args /* { 124 syscallarg(int) fd; 125 syscallarg(struct linux_stat64 *) sp; 126 } */ *uap = v; 127 struct proc *p = l->l_proc; 128 struct sys___fstat13_args fsa; 129 struct linux_stat64 tmplst; 130 struct stat *st,tmpst; 131 caddr_t sg; 132 int error; 133 134 sg = stackgap_init(p, 0); 135 136 st = stackgap_alloc(p, &sg, sizeof (struct stat)); 137 138 SCARG(&fsa, fd) = SCARG(uap, fd); 139 SCARG(&fsa, sb) = st; 140 141 if ((error = sys___fstat13(l, &fsa, retval))) 142 return error; 143 144 if ((error = copyin(st, &tmpst, sizeof tmpst))) 145 return error; 146 147 bsd_to_linux_stat(&tmpst, &tmplst); 148 149 if ((error = copyout(&tmplst, SCARG(uap, sp), sizeof tmplst))) 150 return error; 151 152 return 0; 153 } 154 155 static int 156 linux_do_stat64(l, v, retval, dolstat) 157 struct lwp *l; 158 void *v; 159 register_t *retval; 160 int dolstat; 161 { 162 struct proc *p = l->l_proc; 163 struct sys___stat13_args sa; 164 struct linux_stat64 tmplst; 165 struct stat *st, tmpst; 166 caddr_t sg; 167 int error; 168 struct linux_sys_stat64_args *uap = v; 169 170 sg = stackgap_init(p, 0); 171 st = stackgap_alloc(p, &sg, sizeof (struct stat)); 172 CHECK_ALT_EXIST(p, &sg, SCARG(uap, path)); 173 174 SCARG(&sa, ub) = st; 175 SCARG(&sa, path) = SCARG(uap, path); 176 177 if ((error = (dolstat ? sys___lstat13(l, &sa, retval) : 178 sys___stat13(l, &sa, retval)))) 179 return error; 180 181 if ((error = copyin(st, &tmpst, sizeof tmpst))) 182 return error; 183 184 bsd_to_linux_stat(&tmpst, &tmplst); 185 186 if ((error = copyout(&tmplst, SCARG(uap, sp), sizeof tmplst))) 187 return error; 188 189 return 0; 190 } 191 192 int 193 linux_sys_stat64(l, v, retval) 194 struct lwp *l; 195 void *v; 196 register_t *retval; 197 { 198 struct linux_sys_stat64_args /* { 199 syscallarg(const char *) path; 200 syscallarg(struct linux_stat64 *) sp; 201 } */ *uap = v; 202 203 return linux_do_stat64(l, uap, retval, 0); 204 } 205 206 int 207 linux_sys_lstat64(l, v, retval) 208 struct lwp *l; 209 void *v; 210 register_t *retval; 211 { 212 struct linux_sys_lstat64_args /* { 213 syscallarg(const char *) path; 214 syscallarg(struct linux_stat64 *) sp; 215 } */ *uap = v; 216 217 return linux_do_stat64(l, uap, retval, 1); 218 } 219 220 int 221 linux_sys_truncate64(l, v, retval) 222 struct lwp *l; 223 void *v; 224 register_t *retval; 225 { 226 struct linux_sys_truncate64_args /* { 227 syscallarg(const char *) path; 228 syscallarg(off_t) length; 229 } */ *uap = v; 230 struct sys_truncate_args ta; 231 struct proc *p = l->l_proc; 232 caddr_t sg = stackgap_init(p, 0); 233 234 CHECK_ALT_EXIST(p, &sg, SCARG(uap, path)); 235 236 /* Linux doesn't have the 'pad' pseudo-parameter */ 237 SCARG(&ta, path) = SCARG(uap, path); 238 SCARG(&ta, pad) = 0; 239 SCARG(&ta, length) = SCARG(uap, length); 240 241 return sys_truncate(l, &ta, retval); 242 } 243 244 int 245 linux_sys_ftruncate64(l, v, retval) 246 struct lwp *l; 247 void *v; 248 register_t *retval; 249 { 250 struct linux_sys_ftruncate64_args /* { 251 syscallarg(unsigned int) fd; 252 syscallarg(off_t) length; 253 } */ *uap = v; 254 struct sys_ftruncate_args ta; 255 256 /* Linux doesn't have the 'pad' pseudo-parameter */ 257 SCARG(&ta, fd) = SCARG(uap, fd); 258 SCARG(&ta, pad) = 0; 259 SCARG(&ta, length) = SCARG(uap, length); 260 261 return sys_ftruncate(l, &ta, retval); 262 } 263 264 #if !defined(__m68k__) 265 static void bsd_to_linux_flock64 __P((struct linux_flock64 *, 266 const struct flock *)); 267 static void linux_to_bsd_flock64 __P((struct flock *, 268 const struct linux_flock64 *)); 269 270 static void 271 bsd_to_linux_flock64(lfp, bfp) 272 struct linux_flock64 *lfp; 273 const struct flock *bfp; 274 { 275 276 lfp->l_start = bfp->l_start; 277 lfp->l_len = bfp->l_len; 278 lfp->l_pid = bfp->l_pid; 279 lfp->l_whence = bfp->l_whence; 280 switch (bfp->l_type) { 281 case F_RDLCK: 282 lfp->l_type = LINUX_F_RDLCK; 283 break; 284 case F_UNLCK: 285 lfp->l_type = LINUX_F_UNLCK; 286 break; 287 case F_WRLCK: 288 lfp->l_type = LINUX_F_WRLCK; 289 break; 290 } 291 } 292 293 static void 294 linux_to_bsd_flock64(bfp, lfp) 295 struct flock *bfp; 296 const struct linux_flock64 *lfp; 297 { 298 299 bfp->l_start = lfp->l_start; 300 bfp->l_len = lfp->l_len; 301 bfp->l_pid = lfp->l_pid; 302 bfp->l_whence = lfp->l_whence; 303 switch (lfp->l_type) { 304 case LINUX_F_RDLCK: 305 bfp->l_type = F_RDLCK; 306 break; 307 case LINUX_F_UNLCK: 308 bfp->l_type = F_UNLCK; 309 break; 310 case LINUX_F_WRLCK: 311 bfp->l_type = F_WRLCK; 312 break; 313 } 314 } 315 316 int 317 linux_sys_fcntl64(l, v, retval) 318 struct lwp *l; 319 void *v; 320 register_t *retval; 321 { 322 struct linux_sys_fcntl64_args /* { 323 syscallarg(int) fd; 324 syscallarg(int) cmd; 325 syscallarg(void *) arg; 326 } */ *uap = v; 327 struct proc *p = l->l_proc; 328 struct sys_fcntl_args fca; 329 struct linux_flock64 lfl; 330 struct flock bfl, *bfp; 331 int error; 332 caddr_t sg; 333 void *arg = SCARG(uap, arg); 334 int cmd = SCARG(uap, cmd); 335 int fd = SCARG(uap, fd); 336 337 switch (cmd) { 338 case LINUX_F_GETLK64: 339 sg = stackgap_init(p, 0); 340 bfp = (struct flock *) stackgap_alloc(p, &sg, sizeof *bfp); 341 if ((error = copyin(arg, &lfl, sizeof lfl)) != 0) 342 return error; 343 linux_to_bsd_flock64(&bfl, &lfl); 344 if ((error = copyout(&bfl, bfp, sizeof bfl)) != 0) 345 return error; 346 SCARG(&fca, fd) = fd; 347 SCARG(&fca, cmd) = F_GETLK; 348 SCARG(&fca, arg) = bfp; 349 if ((error = sys_fcntl(l, &fca, retval)) != 0) 350 return error; 351 if ((error = copyin(bfp, &bfl, sizeof bfl)) != 0) 352 return error; 353 bsd_to_linux_flock64(&lfl, &bfl); 354 return copyout(&lfl, arg, sizeof lfl); 355 case LINUX_F_SETLK64: 356 case LINUX_F_SETLKW64: 357 cmd = (cmd == LINUX_F_SETLK64 ? F_SETLK : F_SETLKW); 358 if ((error = copyin(arg, &lfl, sizeof lfl)) != 0) 359 return error; 360 linux_to_bsd_flock64(&bfl, &lfl); 361 sg = stackgap_init(p, 0); 362 bfp = (struct flock *) stackgap_alloc(p, &sg, sizeof *bfp); 363 if ((error = copyout(&bfl, bfp, sizeof bfl)) != 0) 364 return error; 365 SCARG(&fca, fd) = fd; 366 SCARG(&fca, cmd) = cmd; 367 SCARG(&fca, arg) = bfp; 368 return sys_fcntl(l, &fca, retval); 369 default: 370 return linux_sys_fcntl(l, v, retval); 371 } 372 } 373 #endif /* !m68k */ 374 375 #endif /* !alpha */ 376 377 /* 378 * Linux 'readdir' call. This code is mostly taken from the 379 * SunOS getdents call (see compat/sunos/sunos_misc.c), though 380 * an attempt has been made to keep it a little cleaner. 381 * 382 * The d_off field contains the offset of the next valid entry, 383 * unless the older Linux getdents(2), which used to have it set 384 * to the offset of the entry itself. This function also doesn't 385 * need to deal with the old count == 1 glibc problem. 386 * 387 * Read in BSD-style entries, convert them, and copy them out. 388 * 389 * Note that this doesn't handle union-mounted filesystems. 390 */ 391 int 392 linux_sys_getdents64(l, v, retval) 393 struct lwp *l; 394 void *v; 395 register_t *retval; 396 { 397 struct linux_sys_getdents_args /* { 398 syscallarg(int) fd; 399 syscallarg(struct linux_dirent64 *) dent; 400 syscallarg(unsigned int) count; 401 } */ *uap = v; 402 struct proc *p = l->l_proc; 403 struct dirent *bdp; 404 struct vnode *vp; 405 caddr_t inp, buf; /* BSD-format */ 406 int len, reclen; /* BSD-format */ 407 caddr_t outp; /* Linux-format */ 408 int resid, linux_reclen = 0; /* Linux-format */ 409 struct file *fp; 410 struct uio auio; 411 struct iovec aiov; 412 struct linux_dirent64 idb; 413 off_t off; /* true file offset */ 414 int buflen, error, eofflag, nbytes; 415 struct vattr va; 416 off_t *cookiebuf = NULL, *cookie; 417 int ncookies; 418 419 /* getvnode() will use the descriptor for us */ 420 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 421 return (error); 422 423 if ((fp->f_flag & FREAD) == 0) { 424 error = EBADF; 425 goto out1; 426 } 427 428 vp = (struct vnode *)fp->f_data; 429 if (vp->v_type != VDIR) { 430 error = EINVAL; 431 goto out1; 432 } 433 434 if ((error = VOP_GETATTR(vp, &va, p->p_ucred, p))) 435 goto out1; 436 437 nbytes = SCARG(uap, count); 438 buflen = min(MAXBSIZE, nbytes); 439 if (buflen < va.va_blocksize) 440 buflen = va.va_blocksize; 441 buf = malloc(buflen, M_TEMP, M_WAITOK); 442 443 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 444 off = fp->f_offset; 445 again: 446 aiov.iov_base = buf; 447 aiov.iov_len = buflen; 448 auio.uio_iov = &aiov; 449 auio.uio_iovcnt = 1; 450 auio.uio_rw = UIO_READ; 451 auio.uio_segflg = UIO_SYSSPACE; 452 auio.uio_procp = p; 453 auio.uio_resid = buflen; 454 auio.uio_offset = off; 455 /* 456 * First we read into the malloc'ed buffer, then 457 * we massage it into user space, one record at a time. 458 */ 459 error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, &cookiebuf, 460 &ncookies); 461 if (error) 462 goto out; 463 464 inp = buf; 465 outp = (caddr_t)SCARG(uap, dent); 466 resid = nbytes; 467 if ((len = buflen - auio.uio_resid) == 0) 468 goto eof; 469 470 for (cookie = cookiebuf; len > 0; len -= reclen) { 471 bdp = (struct dirent *)inp; 472 reclen = bdp->d_reclen; 473 if (reclen & 3) 474 panic("linux_readdir"); 475 if (bdp->d_fileno == 0) { 476 inp += reclen; /* it is a hole; squish it out */ 477 off = *cookie++; 478 continue; 479 } 480 linux_reclen = LINUX_RECLEN(&idb, bdp->d_namlen); 481 if (reclen > len || resid < linux_reclen) { 482 /* entry too big for buffer, so just stop */ 483 outp++; 484 break; 485 } 486 off = *cookie++; /* each entry points to next */ 487 /* 488 * Massage in place to make a Linux-shaped dirent (otherwise 489 * we have to worry about touching user memory outside of 490 * the copyout() call). 491 */ 492 idb.d_ino = bdp->d_fileno; 493 idb.d_type = bdp->d_type; 494 idb.d_off = off; 495 idb.d_reclen = (u_short)linux_reclen; 496 strcpy(idb.d_name, bdp->d_name); 497 if ((error = copyout((caddr_t)&idb, outp, linux_reclen))) 498 goto out; 499 /* advance past this real entry */ 500 inp += reclen; 501 /* advance output past Linux-shaped entry */ 502 outp += linux_reclen; 503 resid -= linux_reclen; 504 } 505 506 /* if we squished out the whole block, try again */ 507 if (outp == (caddr_t)SCARG(uap, dent)) 508 goto again; 509 fp->f_offset = off; /* update the vnode offset */ 510 511 eof: 512 *retval = nbytes - resid; 513 out: 514 VOP_UNLOCK(vp, 0); 515 if (cookiebuf) 516 free(cookiebuf, M_TEMP); 517 free(buf, M_TEMP); 518 out1: 519 FILE_UNUSE(fp, p); 520 return error; 521 } 522