1 /*- 2 * SPDX-License-Identifier: BSD-2-Clause 3 * 4 * Copyright (c) 1994-1995 Søren Schmidt 5 * All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 16 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND 17 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 18 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 19 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE 20 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 21 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 22 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 23 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 24 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 25 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 26 * SUCH DAMAGE. 27 */ 28 29 #include <sys/cdefs.h> 30 #include <sys/param.h> 31 #include <sys/systm.h> 32 #include <sys/dirent.h> 33 #include <sys/fcntl.h> 34 #include <sys/file.h> 35 #include <sys/filedesc.h> 36 #include <sys/lock.h> 37 #include <sys/mman.h> 38 #include <sys/selinfo.h> 39 #include <sys/pipe.h> 40 #include <sys/proc.h> 41 #include <sys/stat.h> 42 #include <sys/sx.h> 43 #include <sys/syscallsubr.h> 44 #include <sys/tty.h> 45 #include <sys/unistd.h> 46 #include <sys/vnode.h> 47 48 #ifdef COMPAT_LINUX32 49 #include <compat/freebsd32/freebsd32_misc.h> 50 #include <machine/../linux32/linux.h> 51 #include <machine/../linux32/linux32_proto.h> 52 #else 53 #include <machine/../linux/linux.h> 54 #include <machine/../linux/linux_proto.h> 55 #endif 56 #include <compat/linux/linux_misc.h> 57 #include <compat/linux/linux_util.h> 58 #include <compat/linux/linux_file.h> 59 60 static int linux_common_open(struct thread *, int, const char *, int, int, 61 enum uio_seg); 62 static int linux_do_accessat(struct thread *t, int, const char *, int, int); 63 static int linux_getdents_error(struct thread *, int, int); 64 65 static struct bsd_to_linux_bitmap seal_bitmap[] = { 66 BITMAP_1t1_LINUX(F_SEAL_SEAL), 67 BITMAP_1t1_LINUX(F_SEAL_SHRINK), 68 BITMAP_1t1_LINUX(F_SEAL_GROW), 69 BITMAP_1t1_LINUX(F_SEAL_WRITE), 70 }; 71 72 #define MFD_HUGETLB_ENTRY(_size) \ 73 { \ 74 .bsd_value = MFD_HUGE_##_size, \ 75 .linux_value = LINUX_HUGETLB_FLAG_ENCODE_##_size \ 76 } 77 static struct bsd_to_linux_bitmap mfd_bitmap[] = { 78 BITMAP_1t1_LINUX(MFD_CLOEXEC), 79 BITMAP_1t1_LINUX(MFD_ALLOW_SEALING), 80 BITMAP_1t1_LINUX(MFD_HUGETLB), 81 MFD_HUGETLB_ENTRY(64KB), 82 MFD_HUGETLB_ENTRY(512KB), 83 MFD_HUGETLB_ENTRY(1MB), 84 MFD_HUGETLB_ENTRY(2MB), 85 MFD_HUGETLB_ENTRY(8MB), 86 MFD_HUGETLB_ENTRY(16MB), 87 MFD_HUGETLB_ENTRY(32MB), 88 MFD_HUGETLB_ENTRY(256MB), 89 MFD_HUGETLB_ENTRY(512MB), 90 MFD_HUGETLB_ENTRY(1GB), 91 MFD_HUGETLB_ENTRY(2GB), 92 MFD_HUGETLB_ENTRY(16GB), 93 }; 94 #undef MFD_HUGETLB_ENTRY 95 96 #ifdef LINUX_LEGACY_SYSCALLS 97 int 98 linux_creat(struct thread *td, struct linux_creat_args *args) 99 { 100 101 return (kern_openat(td, AT_FDCWD, args->path, UIO_USERSPACE, 102 O_WRONLY | O_CREAT | O_TRUNC, args->mode)); 103 } 104 #endif 105 106 static int 107 linux_common_openflags(int l_flags) 108 { 109 int bsd_flags; 110 111 bsd_flags = 0; 112 switch (l_flags & LINUX_O_ACCMODE) { 113 case LINUX_O_WRONLY: 114 bsd_flags |= O_WRONLY; 115 break; 116 case LINUX_O_RDWR: 117 bsd_flags |= O_RDWR; 118 break; 119 default: 120 bsd_flags |= O_RDONLY; 121 } 122 if (l_flags & LINUX_O_NDELAY) 123 bsd_flags |= O_NONBLOCK; 124 if (l_flags & LINUX_O_APPEND) 125 bsd_flags |= O_APPEND; 126 if (l_flags & LINUX_O_SYNC) 127 bsd_flags |= O_FSYNC; 128 if (l_flags & LINUX_O_CLOEXEC) 129 bsd_flags |= O_CLOEXEC; 130 if (l_flags & LINUX_O_NONBLOCK) 131 bsd_flags |= O_NONBLOCK; 132 if (l_flags & LINUX_O_ASYNC) 133 bsd_flags |= O_ASYNC; 134 if (l_flags & LINUX_O_CREAT) 135 bsd_flags |= O_CREAT; 136 if (l_flags & LINUX_O_TRUNC) 137 bsd_flags |= O_TRUNC; 138 if (l_flags & LINUX_O_EXCL) 139 bsd_flags |= O_EXCL; 140 if (l_flags & LINUX_O_NOCTTY) 141 bsd_flags |= O_NOCTTY; 142 if (l_flags & LINUX_O_DIRECT) 143 bsd_flags |= O_DIRECT; 144 if (l_flags & LINUX_O_NOFOLLOW) 145 bsd_flags |= O_NOFOLLOW; 146 if (l_flags & LINUX_O_DIRECTORY) 147 bsd_flags |= O_DIRECTORY; 148 if (l_flags & LINUX_O_PATH) 149 bsd_flags |= O_PATH; 150 /* XXX LINUX_O_NOATIME: unable to be easily implemented. */ 151 return (bsd_flags); 152 } 153 154 static int 155 linux_common_open(struct thread *td, int dirfd, const char *path, int l_flags, 156 int mode, enum uio_seg seg) 157 { 158 struct proc *p = td->td_proc; 159 struct file *fp; 160 int fd; 161 int bsd_flags, error; 162 163 bsd_flags = linux_common_openflags(l_flags); 164 error = kern_openat(td, dirfd, path, seg, bsd_flags, mode); 165 if (error != 0) { 166 if (error == EMLINK) 167 error = ELOOP; 168 goto done; 169 } 170 if (p->p_flag & P_CONTROLT) 171 goto done; 172 if (bsd_flags & O_NOCTTY) 173 goto done; 174 175 /* 176 * XXX In between kern_openat() and fget(), another process 177 * having the same filedesc could use that fd without 178 * checking below. 179 */ 180 fd = td->td_retval[0]; 181 if (fget(td, fd, &cap_ioctl_rights, &fp) == 0) { 182 if (fp->f_type != DTYPE_VNODE) { 183 fdrop(fp, td); 184 goto done; 185 } 186 sx_slock(&proctree_lock); 187 PROC_LOCK(p); 188 if (SESS_LEADER(p) && !(p->p_flag & P_CONTROLT)) { 189 PROC_UNLOCK(p); 190 sx_sunlock(&proctree_lock); 191 /* XXXPJD: Verify if TIOCSCTTY is allowed. */ 192 (void) fo_ioctl(fp, TIOCSCTTY, (caddr_t) 0, 193 td->td_ucred, td); 194 } else { 195 PROC_UNLOCK(p); 196 sx_sunlock(&proctree_lock); 197 } 198 fdrop(fp, td); 199 } 200 201 done: 202 return (error); 203 } 204 205 int 206 linux_openat(struct thread *td, struct linux_openat_args *args) 207 { 208 int dfd; 209 210 dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd; 211 return (linux_common_open(td, dfd, args->filename, args->flags, 212 args->mode, UIO_USERSPACE)); 213 } 214 215 #ifdef LINUX_LEGACY_SYSCALLS 216 int 217 linux_open(struct thread *td, struct linux_open_args *args) 218 { 219 220 return (linux_common_open(td, AT_FDCWD, args->path, args->flags, 221 args->mode, UIO_USERSPACE)); 222 } 223 #endif 224 225 int 226 linux_name_to_handle_at(struct thread *td, 227 struct linux_name_to_handle_at_args *args) 228 { 229 static const l_int valid_flags = (LINUX_AT_SYMLINK_FOLLOW | 230 LINUX_AT_EMPTY_PATH); 231 static const l_uint fh_size = sizeof(fhandle_t); 232 233 fhandle_t fh; 234 l_uint fh_bytes; 235 l_int mount_id; 236 int error, fd, bsd_flags; 237 238 if (args->flags & ~valid_flags) 239 return (EINVAL); 240 241 fd = args->dirfd; 242 if (fd == LINUX_AT_FDCWD) 243 fd = AT_FDCWD; 244 245 bsd_flags = 0; 246 if (!(args->flags & LINUX_AT_SYMLINK_FOLLOW)) 247 bsd_flags |= AT_SYMLINK_NOFOLLOW; 248 if ((args->flags & LINUX_AT_EMPTY_PATH) != 0) 249 bsd_flags |= AT_EMPTY_PATH; 250 251 error = kern_getfhat(td, bsd_flags, fd, args->name, 252 UIO_USERSPACE, &fh, UIO_SYSSPACE); 253 if (error != 0) 254 return (error); 255 256 /* Emit mount_id -- required before EOVERFLOW case. */ 257 mount_id = (fh.fh_fsid.val[0] ^ fh.fh_fsid.val[1]); 258 error = copyout(&mount_id, args->mnt_id, sizeof(mount_id)); 259 if (error != 0) 260 return (error); 261 262 /* Check if there is room for handle. */ 263 error = copyin(&args->handle->handle_bytes, &fh_bytes, 264 sizeof(fh_bytes)); 265 if (error != 0) 266 return (error); 267 268 if (fh_bytes < fh_size) { 269 error = copyout(&fh_size, &args->handle->handle_bytes, 270 sizeof(fh_size)); 271 if (error == 0) 272 error = EOVERFLOW; 273 return (error); 274 } 275 276 /* Emit handle. */ 277 mount_id = 0; 278 /* 279 * We don't use handle_type for anything yet, but initialize a known 280 * value. 281 */ 282 error = copyout(&mount_id, &args->handle->handle_type, 283 sizeof(mount_id)); 284 if (error != 0) 285 return (error); 286 287 error = copyout(&fh, &args->handle->f_handle, 288 sizeof(fh)); 289 return (error); 290 } 291 292 int 293 linux_open_by_handle_at(struct thread *td, 294 struct linux_open_by_handle_at_args *args) 295 { 296 l_uint fh_bytes; 297 int bsd_flags, error; 298 299 error = copyin(&args->handle->handle_bytes, &fh_bytes, 300 sizeof(fh_bytes)); 301 if (error != 0) 302 return (error); 303 304 if (fh_bytes < sizeof(fhandle_t)) 305 return (EINVAL); 306 307 bsd_flags = linux_common_openflags(args->flags); 308 return (kern_fhopen(td, (void *)&args->handle->f_handle, bsd_flags)); 309 } 310 311 int 312 linux_lseek(struct thread *td, struct linux_lseek_args *args) 313 { 314 315 return (kern_lseek(td, args->fdes, args->off, args->whence)); 316 } 317 318 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32)) 319 int 320 linux_llseek(struct thread *td, struct linux_llseek_args *args) 321 { 322 int error; 323 off_t off; 324 325 off = (args->olow) | (((off_t) args->ohigh) << 32); 326 327 error = kern_lseek(td, args->fd, off, args->whence); 328 if (error != 0) 329 return (error); 330 331 error = copyout(td->td_retval, args->res, sizeof(off_t)); 332 if (error != 0) 333 return (error); 334 335 td->td_retval[0] = 0; 336 return (0); 337 } 338 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */ 339 340 /* 341 * Note that linux_getdents(2) and linux_getdents64(2) have the same 342 * arguments. They only differ in the definition of struct dirent they 343 * operate on. 344 * Note that linux_readdir(2) is a special case of linux_getdents(2) 345 * where count is always equals 1, meaning that the buffer is one 346 * dirent-structure in size and that the code can't handle more anyway. 347 * Note that linux_readdir(2) can't be implemented by means of linux_getdents(2) 348 * as in case when the *dent buffer size is equal to 1 linux_getdents(2) will 349 * trash user stack. 350 */ 351 352 static int 353 linux_getdents_error(struct thread *td, int fd, int err) 354 { 355 struct vnode *vp; 356 struct file *fp; 357 int error; 358 359 /* Linux return ENOTDIR in case when fd is not a directory. */ 360 error = getvnode(td, fd, &cap_read_rights, &fp); 361 if (error != 0) 362 return (error); 363 vp = fp->f_vnode; 364 if (vp->v_type != VDIR) { 365 fdrop(fp, td); 366 return (ENOTDIR); 367 } 368 fdrop(fp, td); 369 return (err); 370 } 371 372 struct l_dirent { 373 l_ulong d_ino; 374 l_off_t d_off; 375 l_ushort d_reclen; 376 char d_name[LINUX_NAME_MAX + 1]; 377 }; 378 379 struct l_dirent64 { 380 uint64_t d_ino; 381 int64_t d_off; 382 l_ushort d_reclen; 383 u_char d_type; 384 char d_name[LINUX_NAME_MAX + 1]; 385 }; 386 387 /* 388 * Linux uses the last byte in the dirent buffer to store d_type, 389 * at least glibc-2.7 requires it. That is why l_dirent is padded with 2 bytes. 390 */ 391 #define LINUX_RECLEN(namlen) \ 392 roundup(offsetof(struct l_dirent, d_name) + (namlen) + 2, sizeof(l_ulong)) 393 394 #define LINUX_RECLEN64(namlen) \ 395 roundup(offsetof(struct l_dirent64, d_name) + (namlen) + 1, \ 396 sizeof(uint64_t)) 397 398 #ifdef LINUX_LEGACY_SYSCALLS 399 int 400 linux_getdents(struct thread *td, struct linux_getdents_args *args) 401 { 402 struct dirent *bdp; 403 caddr_t inp, buf; /* BSD-format */ 404 int len, reclen; /* BSD-format */ 405 caddr_t outp; /* Linux-format */ 406 int resid, linuxreclen; /* Linux-format */ 407 caddr_t lbuf; /* Linux-format */ 408 off_t base; 409 struct l_dirent *linux_dirent; 410 int buflen, error; 411 size_t retval; 412 413 buflen = min(args->count, MAXBSIZE); 414 buf = malloc(buflen, M_LINUX, M_WAITOK); 415 416 error = kern_getdirentries(td, args->fd, buf, buflen, 417 &base, NULL, UIO_SYSSPACE); 418 if (error != 0) { 419 error = linux_getdents_error(td, args->fd, error); 420 goto out1; 421 } 422 423 lbuf = malloc(LINUX_RECLEN(LINUX_NAME_MAX), M_LINUX, M_WAITOK | M_ZERO); 424 425 len = td->td_retval[0]; 426 inp = buf; 427 outp = (caddr_t)args->dent; 428 resid = args->count; 429 retval = 0; 430 431 while (len > 0) { 432 bdp = (struct dirent *) inp; 433 reclen = bdp->d_reclen; 434 linuxreclen = LINUX_RECLEN(bdp->d_namlen); 435 /* 436 * No more space in the user supplied dirent buffer. 437 * Return EINVAL. 438 */ 439 if (resid < linuxreclen) { 440 error = EINVAL; 441 goto out; 442 } 443 444 linux_dirent = (struct l_dirent*)lbuf; 445 linux_dirent->d_ino = bdp->d_fileno; 446 linux_dirent->d_off = bdp->d_off; 447 linux_dirent->d_reclen = linuxreclen; 448 /* 449 * Copy d_type to last byte of l_dirent buffer 450 */ 451 lbuf[linuxreclen - 1] = bdp->d_type; 452 strlcpy(linux_dirent->d_name, bdp->d_name, 453 linuxreclen - offsetof(struct l_dirent, d_name)-1); 454 error = copyout(linux_dirent, outp, linuxreclen); 455 if (error != 0) 456 goto out; 457 458 inp += reclen; 459 base += reclen; 460 len -= reclen; 461 462 retval += linuxreclen; 463 outp += linuxreclen; 464 resid -= linuxreclen; 465 } 466 td->td_retval[0] = retval; 467 468 out: 469 free(lbuf, M_LINUX); 470 out1: 471 free(buf, M_LINUX); 472 return (error); 473 } 474 #endif 475 476 int 477 linux_getdents64(struct thread *td, struct linux_getdents64_args *args) 478 { 479 struct dirent *bdp; 480 caddr_t inp, buf; /* BSD-format */ 481 int len, reclen; /* BSD-format */ 482 caddr_t outp; /* Linux-format */ 483 int resid, linuxreclen; /* Linux-format */ 484 off_t base; 485 struct l_dirent64 *linux_dirent64; 486 int buflen, error; 487 size_t retval; 488 489 buflen = min(args->count, MAXBSIZE); 490 buf = malloc(buflen, M_LINUX, M_WAITOK); 491 492 error = kern_getdirentries(td, args->fd, buf, buflen, 493 &base, NULL, UIO_SYSSPACE); 494 if (error != 0) { 495 error = linux_getdents_error(td, args->fd, error); 496 goto out1; 497 } 498 499 linux_dirent64 = malloc(LINUX_RECLEN64(LINUX_NAME_MAX), M_LINUX, 500 M_WAITOK | M_ZERO); 501 502 len = td->td_retval[0]; 503 inp = buf; 504 outp = (caddr_t)args->dirent; 505 resid = args->count; 506 retval = 0; 507 508 while (len > 0) { 509 bdp = (struct dirent *) inp; 510 reclen = bdp->d_reclen; 511 linuxreclen = LINUX_RECLEN64(bdp->d_namlen); 512 /* 513 * No more space in the user supplied dirent buffer. 514 * Return EINVAL. 515 */ 516 if (resid < linuxreclen) { 517 error = EINVAL; 518 goto out; 519 } 520 521 linux_dirent64->d_ino = bdp->d_fileno; 522 linux_dirent64->d_off = bdp->d_off; 523 linux_dirent64->d_reclen = linuxreclen; 524 linux_dirent64->d_type = bdp->d_type; 525 strlcpy(linux_dirent64->d_name, bdp->d_name, 526 linuxreclen - offsetof(struct l_dirent64, d_name)); 527 error = copyout(linux_dirent64, outp, linuxreclen); 528 if (error != 0) 529 goto out; 530 531 inp += reclen; 532 base += reclen; 533 len -= reclen; 534 535 retval += linuxreclen; 536 outp += linuxreclen; 537 resid -= linuxreclen; 538 } 539 td->td_retval[0] = retval; 540 541 out: 542 free(linux_dirent64, M_LINUX); 543 out1: 544 free(buf, M_LINUX); 545 return (error); 546 } 547 548 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32)) 549 int 550 linux_readdir(struct thread *td, struct linux_readdir_args *args) 551 { 552 struct dirent *bdp; 553 caddr_t buf; /* BSD-format */ 554 int linuxreclen; /* Linux-format */ 555 off_t base; 556 struct l_dirent *linux_dirent; /* Linux-format */ 557 int buflen, error; 558 559 buflen = sizeof(*bdp); 560 buf = malloc(buflen, M_LINUX, M_WAITOK); 561 562 error = kern_getdirentries(td, args->fd, buf, buflen, 563 &base, NULL, UIO_SYSSPACE); 564 if (error != 0) { 565 error = linux_getdents_error(td, args->fd, error); 566 goto out; 567 } 568 if (td->td_retval[0] == 0) 569 goto out; 570 571 linux_dirent = malloc(LINUX_RECLEN(LINUX_NAME_MAX), M_LINUX, 572 M_WAITOK | M_ZERO); 573 574 bdp = (struct dirent *) buf; 575 linuxreclen = LINUX_RECLEN(bdp->d_namlen); 576 577 linux_dirent->d_ino = bdp->d_fileno; 578 linux_dirent->d_off = bdp->d_off; 579 linux_dirent->d_reclen = bdp->d_namlen; 580 strlcpy(linux_dirent->d_name, bdp->d_name, 581 linuxreclen - offsetof(struct l_dirent, d_name)); 582 error = copyout(linux_dirent, args->dent, linuxreclen); 583 if (error == 0) 584 td->td_retval[0] = linuxreclen; 585 586 free(linux_dirent, M_LINUX); 587 out: 588 free(buf, M_LINUX); 589 return (error); 590 } 591 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */ 592 593 /* 594 * These exist mainly for hooks for doing /compat/linux translation. 595 */ 596 597 #ifdef LINUX_LEGACY_SYSCALLS 598 int 599 linux_access(struct thread *td, struct linux_access_args *args) 600 { 601 602 /* Linux convention. */ 603 if (args->amode & ~(F_OK | X_OK | W_OK | R_OK)) 604 return (EINVAL); 605 606 return (kern_accessat(td, AT_FDCWD, args->path, UIO_USERSPACE, 0, 607 args->amode)); 608 } 609 #endif 610 611 static int 612 linux_do_accessat(struct thread *td, int ldfd, const char *filename, 613 int amode, int flags) 614 { 615 int dfd; 616 617 /* Linux convention. */ 618 if (amode & ~(F_OK | X_OK | W_OK | R_OK)) 619 return (EINVAL); 620 621 dfd = (ldfd == LINUX_AT_FDCWD) ? AT_FDCWD : ldfd; 622 return (kern_accessat(td, dfd, filename, UIO_USERSPACE, flags, amode)); 623 } 624 625 int 626 linux_faccessat(struct thread *td, struct linux_faccessat_args *args) 627 { 628 629 return (linux_do_accessat(td, args->dfd, args->filename, args->amode, 630 0)); 631 } 632 633 int 634 linux_faccessat2(struct thread *td, struct linux_faccessat2_args *args) 635 { 636 int flags, unsupported; 637 638 /* XXX. AT_SYMLINK_NOFOLLOW is not supported by kern_accessat */ 639 unsupported = args->flags & ~(LINUX_AT_EACCESS | LINUX_AT_EMPTY_PATH); 640 if (unsupported != 0) { 641 linux_msg(td, "faccessat2 unsupported flag 0x%x", unsupported); 642 return (EINVAL); 643 } 644 645 flags = (args->flags & LINUX_AT_EACCESS) == 0 ? 0 : 646 AT_EACCESS; 647 flags |= (args->flags & LINUX_AT_EMPTY_PATH) == 0 ? 0 : 648 AT_EMPTY_PATH; 649 return (linux_do_accessat(td, args->dfd, args->filename, args->amode, 650 flags)); 651 } 652 653 654 #ifdef LINUX_LEGACY_SYSCALLS 655 int 656 linux_unlink(struct thread *td, struct linux_unlink_args *args) 657 { 658 int error; 659 struct stat st; 660 661 error = kern_funlinkat(td, AT_FDCWD, args->path, FD_NONE, 662 UIO_USERSPACE, 0, 0); 663 if (error == EPERM) { 664 /* Introduce POSIX noncompliant behaviour of Linux */ 665 if (kern_statat(td, 0, AT_FDCWD, args->path, 666 UIO_USERSPACE, &st) == 0) { 667 if (S_ISDIR(st.st_mode)) 668 error = EISDIR; 669 } 670 } 671 672 return (error); 673 } 674 #endif 675 676 static int 677 linux_unlinkat_impl(struct thread *td, enum uio_seg pathseg, const char *path, 678 int dfd, struct linux_unlinkat_args *args) 679 { 680 struct stat st; 681 int error; 682 683 if (args->flag & LINUX_AT_REMOVEDIR) 684 error = kern_frmdirat(td, dfd, path, FD_NONE, pathseg, 0); 685 else 686 error = kern_funlinkat(td, dfd, path, FD_NONE, pathseg, 0, 0); 687 if (error == EPERM && !(args->flag & LINUX_AT_REMOVEDIR)) { 688 /* Introduce POSIX noncompliant behaviour of Linux */ 689 if (kern_statat(td, AT_SYMLINK_NOFOLLOW, dfd, path, 690 pathseg, &st) == 0 && S_ISDIR(st.st_mode)) 691 error = EISDIR; 692 } 693 return (error); 694 } 695 696 int 697 linux_unlinkat(struct thread *td, struct linux_unlinkat_args *args) 698 { 699 int dfd; 700 701 if (args->flag & ~LINUX_AT_REMOVEDIR) 702 return (EINVAL); 703 dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd; 704 return (linux_unlinkat_impl(td, UIO_USERSPACE, args->pathname, 705 dfd, args)); 706 } 707 708 int 709 linux_chdir(struct thread *td, struct linux_chdir_args *args) 710 { 711 712 return (kern_chdir(td, args->path, UIO_USERSPACE)); 713 } 714 715 #ifdef LINUX_LEGACY_SYSCALLS 716 int 717 linux_chmod(struct thread *td, struct linux_chmod_args *args) 718 { 719 720 return (kern_fchmodat(td, AT_FDCWD, args->path, UIO_USERSPACE, 721 args->mode, 0)); 722 } 723 #endif 724 725 int 726 linux_fchmodat(struct thread *td, struct linux_fchmodat_args *args) 727 { 728 int dfd; 729 730 dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd; 731 return (kern_fchmodat(td, dfd, args->filename, UIO_USERSPACE, 732 args->mode, 0)); 733 } 734 735 #ifdef LINUX_LEGACY_SYSCALLS 736 int 737 linux_mkdir(struct thread *td, struct linux_mkdir_args *args) 738 { 739 740 return (kern_mkdirat(td, AT_FDCWD, args->path, UIO_USERSPACE, args->mode)); 741 } 742 #endif 743 744 int 745 linux_mkdirat(struct thread *td, struct linux_mkdirat_args *args) 746 { 747 int dfd; 748 749 dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd; 750 return (kern_mkdirat(td, dfd, args->pathname, UIO_USERSPACE, args->mode)); 751 } 752 753 #ifdef LINUX_LEGACY_SYSCALLS 754 int 755 linux_rmdir(struct thread *td, struct linux_rmdir_args *args) 756 { 757 758 return (kern_frmdirat(td, AT_FDCWD, args->path, FD_NONE, 759 UIO_USERSPACE, 0)); 760 } 761 762 int 763 linux_rename(struct thread *td, struct linux_rename_args *args) 764 { 765 766 return (kern_renameat(td, AT_FDCWD, args->from, AT_FDCWD, 767 args->to, UIO_USERSPACE)); 768 } 769 #endif 770 771 int 772 linux_renameat(struct thread *td, struct linux_renameat_args *args) 773 { 774 struct linux_renameat2_args renameat2_args = { 775 .olddfd = args->olddfd, 776 .oldname = args->oldname, 777 .newdfd = args->newdfd, 778 .newname = args->newname, 779 .flags = 0 780 }; 781 782 return (linux_renameat2(td, &renameat2_args)); 783 } 784 785 int 786 linux_renameat2(struct thread *td, struct linux_renameat2_args *args) 787 { 788 int olddfd, newdfd; 789 790 if (args->flags != 0) { 791 if (args->flags & ~(LINUX_RENAME_EXCHANGE | 792 LINUX_RENAME_NOREPLACE | LINUX_RENAME_WHITEOUT)) 793 return (EINVAL); 794 if (args->flags & LINUX_RENAME_EXCHANGE && 795 args->flags & (LINUX_RENAME_NOREPLACE | 796 LINUX_RENAME_WHITEOUT)) 797 return (EINVAL); 798 #if 0 799 /* 800 * This spams the console on Ubuntu Focal. 801 * 802 * What's needed here is a general mechanism to let users know 803 * about missing features without hogging the system. 804 */ 805 linux_msg(td, "renameat2 unsupported flags 0x%x", 806 args->flags); 807 #endif 808 return (EINVAL); 809 } 810 811 olddfd = (args->olddfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->olddfd; 812 newdfd = (args->newdfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->newdfd; 813 return (kern_renameat(td, olddfd, args->oldname, newdfd, 814 args->newname, UIO_USERSPACE)); 815 } 816 817 #ifdef LINUX_LEGACY_SYSCALLS 818 int 819 linux_symlink(struct thread *td, struct linux_symlink_args *args) 820 { 821 822 return (kern_symlinkat(td, args->path, AT_FDCWD, args->to, 823 UIO_USERSPACE)); 824 } 825 #endif 826 827 int 828 linux_symlinkat(struct thread *td, struct linux_symlinkat_args *args) 829 { 830 int dfd; 831 832 dfd = (args->newdfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->newdfd; 833 return (kern_symlinkat(td, args->oldname, dfd, args->newname, 834 UIO_USERSPACE)); 835 } 836 837 #ifdef LINUX_LEGACY_SYSCALLS 838 int 839 linux_readlink(struct thread *td, struct linux_readlink_args *args) 840 { 841 842 if (args->count <= 0) 843 return (EINVAL); 844 845 return (kern_readlinkat(td, AT_FDCWD, args->name, UIO_USERSPACE, 846 args->buf, UIO_USERSPACE, args->count)); 847 } 848 #endif 849 850 int 851 linux_readlinkat(struct thread *td, struct linux_readlinkat_args *args) 852 { 853 int dfd; 854 855 if (args->bufsiz <= 0) 856 return (EINVAL); 857 858 dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd; 859 return (kern_readlinkat(td, dfd, args->path, UIO_USERSPACE, 860 args->buf, UIO_USERSPACE, args->bufsiz)); 861 } 862 863 int 864 linux_truncate(struct thread *td, struct linux_truncate_args *args) 865 { 866 867 return (kern_truncate(td, args->path, UIO_USERSPACE, args->length)); 868 } 869 870 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32)) 871 int 872 linux_truncate64(struct thread *td, struct linux_truncate64_args *args) 873 { 874 off_t length; 875 876 #if defined(__amd64__) && defined(COMPAT_LINUX32) 877 length = PAIR32TO64(off_t, args->length); 878 #else 879 length = args->length; 880 #endif 881 882 return (kern_truncate(td, args->path, UIO_USERSPACE, length)); 883 } 884 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */ 885 886 int 887 linux_ftruncate(struct thread *td, struct linux_ftruncate_args *args) 888 { 889 890 return (kern_ftruncate(td, args->fd, args->length)); 891 } 892 893 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32)) 894 int 895 linux_ftruncate64(struct thread *td, struct linux_ftruncate64_args *args) 896 { 897 off_t length; 898 899 #if defined(__amd64__) && defined(COMPAT_LINUX32) 900 length = PAIR32TO64(off_t, args->length); 901 #else 902 length = args->length; 903 #endif 904 905 return (kern_ftruncate(td, args->fd, length)); 906 } 907 #endif 908 909 #ifdef LINUX_LEGACY_SYSCALLS 910 int 911 linux_link(struct thread *td, struct linux_link_args *args) 912 { 913 914 return (kern_linkat(td, AT_FDCWD, AT_FDCWD, args->path, args->to, 915 UIO_USERSPACE, AT_SYMLINK_FOLLOW)); 916 } 917 #endif 918 919 int 920 linux_linkat(struct thread *td, struct linux_linkat_args *args) 921 { 922 int olddfd, newdfd, flag; 923 924 if (args->flag & ~(LINUX_AT_SYMLINK_FOLLOW | LINUX_AT_EMPTY_PATH)) 925 return (EINVAL); 926 927 flag = (args->flag & LINUX_AT_SYMLINK_FOLLOW) != 0 ? AT_SYMLINK_FOLLOW : 928 0; 929 flag |= (args->flag & LINUX_AT_EMPTY_PATH) != 0 ? AT_EMPTY_PATH : 0; 930 931 olddfd = (args->olddfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->olddfd; 932 newdfd = (args->newdfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->newdfd; 933 return (kern_linkat(td, olddfd, newdfd, args->oldname, 934 args->newname, UIO_USERSPACE, flag)); 935 } 936 937 int 938 linux_fdatasync(struct thread *td, struct linux_fdatasync_args *uap) 939 { 940 941 return (kern_fsync(td, uap->fd, false)); 942 } 943 944 int 945 linux_sync_file_range(struct thread *td, struct linux_sync_file_range_args *uap) 946 { 947 off_t nbytes, offset; 948 949 #if defined(__amd64__) && defined(COMPAT_LINUX32) 950 nbytes = PAIR32TO64(off_t, uap->nbytes); 951 offset = PAIR32TO64(off_t, uap->offset); 952 #else 953 nbytes = uap->nbytes; 954 offset = uap->offset; 955 #endif 956 957 if (offset < 0 || nbytes < 0 || 958 (uap->flags & ~(LINUX_SYNC_FILE_RANGE_WAIT_BEFORE | 959 LINUX_SYNC_FILE_RANGE_WRITE | 960 LINUX_SYNC_FILE_RANGE_WAIT_AFTER)) != 0) { 961 return (EINVAL); 962 } 963 964 return (kern_fsync(td, uap->fd, false)); 965 } 966 967 int 968 linux_pread(struct thread *td, struct linux_pread_args *uap) 969 { 970 struct vnode *vp; 971 off_t offset; 972 int error; 973 974 #if defined(__amd64__) && defined(COMPAT_LINUX32) 975 offset = PAIR32TO64(off_t, uap->offset); 976 #else 977 offset = uap->offset; 978 #endif 979 980 error = kern_pread(td, uap->fd, uap->buf, uap->nbyte, offset); 981 if (error == 0) { 982 /* This seems to violate POSIX but Linux does it. */ 983 error = fgetvp(td, uap->fd, &cap_pread_rights, &vp); 984 if (error != 0) 985 return (error); 986 if (vp->v_type == VDIR) 987 error = EISDIR; 988 vrele(vp); 989 } 990 return (error); 991 } 992 993 int 994 linux_pwrite(struct thread *td, struct linux_pwrite_args *uap) 995 { 996 off_t offset; 997 998 #if defined(__amd64__) && defined(COMPAT_LINUX32) 999 offset = PAIR32TO64(off_t, uap->offset); 1000 #else 1001 offset = uap->offset; 1002 #endif 1003 1004 return (kern_pwrite(td, uap->fd, uap->buf, uap->nbyte, offset)); 1005 } 1006 1007 #define HALF_LONG_BITS ((sizeof(l_long) * NBBY / 2)) 1008 1009 static inline off_t 1010 pos_from_hilo(unsigned long high, unsigned long low) 1011 { 1012 1013 return (((off_t)high << HALF_LONG_BITS) << HALF_LONG_BITS) | low; 1014 } 1015 1016 int 1017 linux_preadv(struct thread *td, struct linux_preadv_args *uap) 1018 { 1019 struct uio *auio; 1020 int error; 1021 off_t offset; 1022 1023 /* 1024 * According http://man7.org/linux/man-pages/man2/preadv.2.html#NOTES 1025 * pos_l and pos_h, respectively, contain the 1026 * low order and high order 32 bits of offset. 1027 */ 1028 offset = pos_from_hilo(uap->pos_h, uap->pos_l); 1029 if (offset < 0) 1030 return (EINVAL); 1031 #ifdef COMPAT_LINUX32 1032 error = linux32_copyinuio(PTRIN(uap->vec), uap->vlen, &auio); 1033 #else 1034 error = copyinuio(uap->vec, uap->vlen, &auio); 1035 #endif 1036 if (error != 0) 1037 return (error); 1038 error = kern_preadv(td, uap->fd, auio, offset); 1039 free(auio, M_IOV); 1040 return (error); 1041 } 1042 1043 int 1044 linux_pwritev(struct thread *td, struct linux_pwritev_args *uap) 1045 { 1046 struct uio *auio; 1047 int error; 1048 off_t offset; 1049 1050 /* 1051 * According http://man7.org/linux/man-pages/man2/pwritev.2.html#NOTES 1052 * pos_l and pos_h, respectively, contain the 1053 * low order and high order 32 bits of offset. 1054 */ 1055 offset = pos_from_hilo(uap->pos_h, uap->pos_l); 1056 if (offset < 0) 1057 return (EINVAL); 1058 #ifdef COMPAT_LINUX32 1059 error = linux32_copyinuio(PTRIN(uap->vec), uap->vlen, &auio); 1060 #else 1061 error = copyinuio(uap->vec, uap->vlen, &auio); 1062 #endif 1063 if (error != 0) 1064 return (error); 1065 error = kern_pwritev(td, uap->fd, auio, offset); 1066 free(auio, M_IOV); 1067 return (error); 1068 } 1069 1070 int 1071 linux_mount(struct thread *td, struct linux_mount_args *args) 1072 { 1073 struct mntarg *ma = NULL; 1074 char *fstypename, *mntonname, *mntfromname, *data; 1075 int error, fsflags; 1076 1077 fstypename = malloc(MNAMELEN, M_TEMP, M_WAITOK); 1078 mntonname = malloc(MNAMELEN, M_TEMP, M_WAITOK); 1079 mntfromname = malloc(MNAMELEN, M_TEMP, M_WAITOK); 1080 data = NULL; 1081 error = copyinstr(args->filesystemtype, fstypename, MNAMELEN - 1, 1082 NULL); 1083 if (error != 0) 1084 goto out; 1085 if (args->specialfile != NULL) { 1086 error = copyinstr(args->specialfile, mntfromname, MNAMELEN - 1, NULL); 1087 if (error != 0) 1088 goto out; 1089 } else { 1090 mntfromname[0] = '\0'; 1091 } 1092 error = copyinstr(args->dir, mntonname, MNAMELEN - 1, NULL); 1093 if (error != 0) 1094 goto out; 1095 1096 if (strcmp(fstypename, "ext2") == 0) { 1097 strcpy(fstypename, "ext2fs"); 1098 } else if (strcmp(fstypename, "proc") == 0) { 1099 strcpy(fstypename, "linprocfs"); 1100 } else if (strcmp(fstypename, "vfat") == 0) { 1101 strcpy(fstypename, "msdosfs"); 1102 } else if (strcmp(fstypename, "fuse") == 0 || 1103 strncmp(fstypename, "fuse.", 5) == 0) { 1104 char *fuse_options, *fuse_option, *fuse_name; 1105 1106 strcpy(mntfromname, "/dev/fuse"); 1107 strcpy(fstypename, "fusefs"); 1108 data = malloc(MNAMELEN, M_TEMP, M_WAITOK); 1109 error = copyinstr(args->data, data, MNAMELEN - 1, NULL); 1110 if (error != 0) 1111 goto out; 1112 1113 fuse_options = data; 1114 while ((fuse_option = strsep(&fuse_options, ",")) != NULL) { 1115 fuse_name = strsep(&fuse_option, "="); 1116 if (fuse_name == NULL || fuse_option == NULL) 1117 goto out; 1118 ma = mount_arg(ma, fuse_name, fuse_option, -1); 1119 } 1120 1121 /* 1122 * The FUSE server uses Linux errno values instead of FreeBSD 1123 * ones; add a flag to tell fuse(4) to do errno translation. 1124 */ 1125 ma = mount_arg(ma, "linux_errnos", "1", -1); 1126 } 1127 1128 fsflags = 0; 1129 1130 /* 1131 * Linux SYNC flag is not included; the closest equivalent 1132 * FreeBSD has is !ASYNC, which is our default. 1133 */ 1134 if (args->rwflag & LINUX_MS_RDONLY) 1135 fsflags |= MNT_RDONLY; 1136 if (args->rwflag & LINUX_MS_NOSUID) 1137 fsflags |= MNT_NOSUID; 1138 if (args->rwflag & LINUX_MS_NOEXEC) 1139 fsflags |= MNT_NOEXEC; 1140 if (args->rwflag & LINUX_MS_REMOUNT) 1141 fsflags |= MNT_UPDATE; 1142 1143 ma = mount_arg(ma, "fstype", fstypename, -1); 1144 ma = mount_arg(ma, "fspath", mntonname, -1); 1145 ma = mount_arg(ma, "from", mntfromname, -1); 1146 error = kernel_mount(ma, fsflags); 1147 out: 1148 free(fstypename, M_TEMP); 1149 free(mntonname, M_TEMP); 1150 free(mntfromname, M_TEMP); 1151 free(data, M_TEMP); 1152 return (error); 1153 } 1154 1155 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32)) 1156 int 1157 linux_oldumount(struct thread *td, struct linux_oldumount_args *args) 1158 { 1159 1160 return (kern_unmount(td, args->path, 0)); 1161 } 1162 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */ 1163 1164 #ifdef LINUX_LEGACY_SYSCALLS 1165 int 1166 linux_umount(struct thread *td, struct linux_umount_args *args) 1167 { 1168 int flags; 1169 1170 flags = 0; 1171 if ((args->flags & LINUX_MNT_FORCE) != 0) { 1172 args->flags &= ~LINUX_MNT_FORCE; 1173 flags |= MNT_FORCE; 1174 } 1175 if (args->flags != 0) { 1176 linux_msg(td, "unsupported umount2 flags %#x", args->flags); 1177 return (EINVAL); 1178 } 1179 1180 return (kern_unmount(td, args->path, flags)); 1181 } 1182 #endif 1183 1184 /* 1185 * fcntl family of syscalls 1186 */ 1187 1188 struct l_flock { 1189 l_short l_type; 1190 l_short l_whence; 1191 l_off_t l_start; 1192 l_off_t l_len; 1193 l_pid_t l_pid; 1194 } 1195 #if defined(__amd64__) && defined(COMPAT_LINUX32) 1196 __packed 1197 #endif 1198 ; 1199 1200 static void 1201 linux_to_bsd_flock(struct l_flock *linux_flock, struct flock *bsd_flock) 1202 { 1203 switch (linux_flock->l_type) { 1204 case LINUX_F_RDLCK: 1205 bsd_flock->l_type = F_RDLCK; 1206 break; 1207 case LINUX_F_WRLCK: 1208 bsd_flock->l_type = F_WRLCK; 1209 break; 1210 case LINUX_F_UNLCK: 1211 bsd_flock->l_type = F_UNLCK; 1212 break; 1213 default: 1214 bsd_flock->l_type = -1; 1215 break; 1216 } 1217 bsd_flock->l_whence = linux_flock->l_whence; 1218 bsd_flock->l_start = (off_t)linux_flock->l_start; 1219 bsd_flock->l_len = (off_t)linux_flock->l_len; 1220 bsd_flock->l_pid = (pid_t)linux_flock->l_pid; 1221 bsd_flock->l_sysid = 0; 1222 } 1223 1224 static void 1225 bsd_to_linux_flock(struct flock *bsd_flock, struct l_flock *linux_flock) 1226 { 1227 switch (bsd_flock->l_type) { 1228 case F_RDLCK: 1229 linux_flock->l_type = LINUX_F_RDLCK; 1230 break; 1231 case F_WRLCK: 1232 linux_flock->l_type = LINUX_F_WRLCK; 1233 break; 1234 case F_UNLCK: 1235 linux_flock->l_type = LINUX_F_UNLCK; 1236 break; 1237 } 1238 linux_flock->l_whence = bsd_flock->l_whence; 1239 linux_flock->l_start = (l_off_t)bsd_flock->l_start; 1240 linux_flock->l_len = (l_off_t)bsd_flock->l_len; 1241 linux_flock->l_pid = (l_pid_t)bsd_flock->l_pid; 1242 } 1243 1244 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32)) 1245 struct l_flock64 { 1246 l_short l_type; 1247 l_short l_whence; 1248 l_loff_t l_start; 1249 l_loff_t l_len; 1250 l_pid_t l_pid; 1251 } 1252 #if defined(__amd64__) && defined(COMPAT_LINUX32) 1253 __packed 1254 #endif 1255 ; 1256 1257 static void 1258 linux_to_bsd_flock64(struct l_flock64 *linux_flock, struct flock *bsd_flock) 1259 { 1260 switch (linux_flock->l_type) { 1261 case LINUX_F_RDLCK: 1262 bsd_flock->l_type = F_RDLCK; 1263 break; 1264 case LINUX_F_WRLCK: 1265 bsd_flock->l_type = F_WRLCK; 1266 break; 1267 case LINUX_F_UNLCK: 1268 bsd_flock->l_type = F_UNLCK; 1269 break; 1270 default: 1271 bsd_flock->l_type = -1; 1272 break; 1273 } 1274 bsd_flock->l_whence = linux_flock->l_whence; 1275 bsd_flock->l_start = (off_t)linux_flock->l_start; 1276 bsd_flock->l_len = (off_t)linux_flock->l_len; 1277 bsd_flock->l_pid = (pid_t)linux_flock->l_pid; 1278 bsd_flock->l_sysid = 0; 1279 } 1280 1281 static void 1282 bsd_to_linux_flock64(struct flock *bsd_flock, struct l_flock64 *linux_flock) 1283 { 1284 switch (bsd_flock->l_type) { 1285 case F_RDLCK: 1286 linux_flock->l_type = LINUX_F_RDLCK; 1287 break; 1288 case F_WRLCK: 1289 linux_flock->l_type = LINUX_F_WRLCK; 1290 break; 1291 case F_UNLCK: 1292 linux_flock->l_type = LINUX_F_UNLCK; 1293 break; 1294 } 1295 linux_flock->l_whence = bsd_flock->l_whence; 1296 linux_flock->l_start = (l_loff_t)bsd_flock->l_start; 1297 linux_flock->l_len = (l_loff_t)bsd_flock->l_len; 1298 linux_flock->l_pid = (l_pid_t)bsd_flock->l_pid; 1299 } 1300 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */ 1301 1302 static int 1303 fcntl_common(struct thread *td, struct linux_fcntl_args *args) 1304 { 1305 struct l_flock linux_flock; 1306 struct flock bsd_flock; 1307 struct pipe *fpipe; 1308 struct file *fp; 1309 long arg; 1310 int error, result; 1311 1312 switch (args->cmd) { 1313 case LINUX_F_DUPFD: 1314 return (kern_fcntl(td, args->fd, F_DUPFD, args->arg)); 1315 1316 case LINUX_F_GETFD: 1317 return (kern_fcntl(td, args->fd, F_GETFD, 0)); 1318 1319 case LINUX_F_SETFD: 1320 return (kern_fcntl(td, args->fd, F_SETFD, args->arg)); 1321 1322 case LINUX_F_GETFL: 1323 error = kern_fcntl(td, args->fd, F_GETFL, 0); 1324 result = td->td_retval[0]; 1325 td->td_retval[0] = 0; 1326 if (result & O_RDONLY) 1327 td->td_retval[0] |= LINUX_O_RDONLY; 1328 if (result & O_WRONLY) 1329 td->td_retval[0] |= LINUX_O_WRONLY; 1330 if (result & O_RDWR) 1331 td->td_retval[0] |= LINUX_O_RDWR; 1332 if (result & O_NDELAY) 1333 td->td_retval[0] |= LINUX_O_NONBLOCK; 1334 if (result & O_APPEND) 1335 td->td_retval[0] |= LINUX_O_APPEND; 1336 if (result & O_FSYNC) 1337 td->td_retval[0] |= LINUX_O_SYNC; 1338 if (result & O_ASYNC) 1339 td->td_retval[0] |= LINUX_O_ASYNC; 1340 #ifdef LINUX_O_NOFOLLOW 1341 if (result & O_NOFOLLOW) 1342 td->td_retval[0] |= LINUX_O_NOFOLLOW; 1343 #endif 1344 #ifdef LINUX_O_DIRECT 1345 if (result & O_DIRECT) 1346 td->td_retval[0] |= LINUX_O_DIRECT; 1347 #endif 1348 return (error); 1349 1350 case LINUX_F_SETFL: 1351 arg = 0; 1352 if (args->arg & LINUX_O_NDELAY) 1353 arg |= O_NONBLOCK; 1354 if (args->arg & LINUX_O_APPEND) 1355 arg |= O_APPEND; 1356 if (args->arg & LINUX_O_SYNC) 1357 arg |= O_FSYNC; 1358 if (args->arg & LINUX_O_ASYNC) 1359 arg |= O_ASYNC; 1360 #ifdef LINUX_O_NOFOLLOW 1361 if (args->arg & LINUX_O_NOFOLLOW) 1362 arg |= O_NOFOLLOW; 1363 #endif 1364 #ifdef LINUX_O_DIRECT 1365 if (args->arg & LINUX_O_DIRECT) 1366 arg |= O_DIRECT; 1367 #endif 1368 return (kern_fcntl(td, args->fd, F_SETFL, arg)); 1369 1370 case LINUX_F_GETLK: 1371 error = copyin((void *)args->arg, &linux_flock, 1372 sizeof(linux_flock)); 1373 if (error) 1374 return (error); 1375 linux_to_bsd_flock(&linux_flock, &bsd_flock); 1376 error = kern_fcntl(td, args->fd, F_GETLK, (intptr_t)&bsd_flock); 1377 if (error) 1378 return (error); 1379 bsd_to_linux_flock(&bsd_flock, &linux_flock); 1380 return (copyout(&linux_flock, (void *)args->arg, 1381 sizeof(linux_flock))); 1382 1383 case LINUX_F_SETLK: 1384 error = copyin((void *)args->arg, &linux_flock, 1385 sizeof(linux_flock)); 1386 if (error) 1387 return (error); 1388 linux_to_bsd_flock(&linux_flock, &bsd_flock); 1389 return (kern_fcntl(td, args->fd, F_SETLK, 1390 (intptr_t)&bsd_flock)); 1391 1392 case LINUX_F_SETLKW: 1393 error = copyin((void *)args->arg, &linux_flock, 1394 sizeof(linux_flock)); 1395 if (error) 1396 return (error); 1397 linux_to_bsd_flock(&linux_flock, &bsd_flock); 1398 return (kern_fcntl(td, args->fd, F_SETLKW, 1399 (intptr_t)&bsd_flock)); 1400 1401 case LINUX_F_GETOWN: 1402 return (kern_fcntl(td, args->fd, F_GETOWN, 0)); 1403 1404 case LINUX_F_SETOWN: 1405 /* 1406 * XXX some Linux applications depend on F_SETOWN having no 1407 * significant effect for pipes (SIGIO is not delivered for 1408 * pipes under Linux-2.2.35 at least). 1409 */ 1410 error = fget(td, args->fd, 1411 &cap_fcntl_rights, &fp); 1412 if (error) 1413 return (error); 1414 if (fp->f_type == DTYPE_PIPE) { 1415 fdrop(fp, td); 1416 return (EINVAL); 1417 } 1418 fdrop(fp, td); 1419 1420 return (kern_fcntl(td, args->fd, F_SETOWN, args->arg)); 1421 1422 case LINUX_F_DUPFD_CLOEXEC: 1423 return (kern_fcntl(td, args->fd, F_DUPFD_CLOEXEC, args->arg)); 1424 /* 1425 * Our F_SEAL_* values match Linux one for maximum compatibility. So we 1426 * only needed to account for different values for fcntl(2) commands. 1427 */ 1428 case LINUX_F_GET_SEALS: 1429 error = kern_fcntl(td, args->fd, F_GET_SEALS, 0); 1430 if (error != 0) 1431 return (error); 1432 td->td_retval[0] = bsd_to_linux_bits(td->td_retval[0], 1433 seal_bitmap, 0); 1434 return (0); 1435 1436 case LINUX_F_ADD_SEALS: 1437 return (kern_fcntl(td, args->fd, F_ADD_SEALS, 1438 linux_to_bsd_bits(args->arg, seal_bitmap, 0))); 1439 1440 case LINUX_F_GETPIPE_SZ: 1441 error = fget(td, args->fd, 1442 &cap_fcntl_rights, &fp); 1443 if (error != 0) 1444 return (error); 1445 if (fp->f_type != DTYPE_PIPE) { 1446 fdrop(fp, td); 1447 return (EINVAL); 1448 } 1449 fpipe = fp->f_data; 1450 td->td_retval[0] = fpipe->pipe_buffer.size; 1451 fdrop(fp, td); 1452 return (0); 1453 1454 default: 1455 linux_msg(td, "unsupported fcntl cmd %d", args->cmd); 1456 return (EINVAL); 1457 } 1458 } 1459 1460 int 1461 linux_fcntl(struct thread *td, struct linux_fcntl_args *args) 1462 { 1463 1464 return (fcntl_common(td, args)); 1465 } 1466 1467 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32)) 1468 int 1469 linux_fcntl64(struct thread *td, struct linux_fcntl64_args *args) 1470 { 1471 struct l_flock64 linux_flock; 1472 struct flock bsd_flock; 1473 struct linux_fcntl_args fcntl_args; 1474 int error; 1475 1476 switch (args->cmd) { 1477 case LINUX_F_GETLK64: 1478 error = copyin((void *)args->arg, &linux_flock, 1479 sizeof(linux_flock)); 1480 if (error) 1481 return (error); 1482 linux_to_bsd_flock64(&linux_flock, &bsd_flock); 1483 error = kern_fcntl(td, args->fd, F_GETLK, (intptr_t)&bsd_flock); 1484 if (error) 1485 return (error); 1486 bsd_to_linux_flock64(&bsd_flock, &linux_flock); 1487 return (copyout(&linux_flock, (void *)args->arg, 1488 sizeof(linux_flock))); 1489 1490 case LINUX_F_SETLK64: 1491 error = copyin((void *)args->arg, &linux_flock, 1492 sizeof(linux_flock)); 1493 if (error) 1494 return (error); 1495 linux_to_bsd_flock64(&linux_flock, &bsd_flock); 1496 return (kern_fcntl(td, args->fd, F_SETLK, 1497 (intptr_t)&bsd_flock)); 1498 1499 case LINUX_F_SETLKW64: 1500 error = copyin((void *)args->arg, &linux_flock, 1501 sizeof(linux_flock)); 1502 if (error) 1503 return (error); 1504 linux_to_bsd_flock64(&linux_flock, &bsd_flock); 1505 return (kern_fcntl(td, args->fd, F_SETLKW, 1506 (intptr_t)&bsd_flock)); 1507 } 1508 1509 fcntl_args.fd = args->fd; 1510 fcntl_args.cmd = args->cmd; 1511 fcntl_args.arg = args->arg; 1512 return (fcntl_common(td, &fcntl_args)); 1513 } 1514 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */ 1515 1516 #ifdef LINUX_LEGACY_SYSCALLS 1517 int 1518 linux_chown(struct thread *td, struct linux_chown_args *args) 1519 { 1520 1521 return (kern_fchownat(td, AT_FDCWD, args->path, UIO_USERSPACE, 1522 args->uid, args->gid, 0)); 1523 } 1524 #endif 1525 1526 int 1527 linux_fchownat(struct thread *td, struct linux_fchownat_args *args) 1528 { 1529 int dfd, flag, unsupported; 1530 1531 unsupported = args->flag & ~(LINUX_AT_SYMLINK_NOFOLLOW | LINUX_AT_EMPTY_PATH); 1532 if (unsupported != 0) { 1533 linux_msg(td, "fchownat unsupported flag 0x%x", unsupported); 1534 return (EINVAL); 1535 } 1536 1537 flag = (args->flag & LINUX_AT_SYMLINK_NOFOLLOW) == 0 ? 0 : 1538 AT_SYMLINK_NOFOLLOW; 1539 flag |= (args->flag & LINUX_AT_EMPTY_PATH) == 0 ? 0 : 1540 AT_EMPTY_PATH; 1541 1542 dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd; 1543 return (kern_fchownat(td, dfd, args->filename, UIO_USERSPACE, 1544 args->uid, args->gid, flag)); 1545 } 1546 1547 #ifdef LINUX_LEGACY_SYSCALLS 1548 int 1549 linux_lchown(struct thread *td, struct linux_lchown_args *args) 1550 { 1551 1552 return (kern_fchownat(td, AT_FDCWD, args->path, UIO_USERSPACE, args->uid, 1553 args->gid, AT_SYMLINK_NOFOLLOW)); 1554 } 1555 #endif 1556 1557 static int 1558 convert_fadvice(int advice) 1559 { 1560 switch (advice) { 1561 case LINUX_POSIX_FADV_NORMAL: 1562 return (POSIX_FADV_NORMAL); 1563 case LINUX_POSIX_FADV_RANDOM: 1564 return (POSIX_FADV_RANDOM); 1565 case LINUX_POSIX_FADV_SEQUENTIAL: 1566 return (POSIX_FADV_SEQUENTIAL); 1567 case LINUX_POSIX_FADV_WILLNEED: 1568 return (POSIX_FADV_WILLNEED); 1569 case LINUX_POSIX_FADV_DONTNEED: 1570 return (POSIX_FADV_DONTNEED); 1571 case LINUX_POSIX_FADV_NOREUSE: 1572 return (POSIX_FADV_NOREUSE); 1573 default: 1574 return (-1); 1575 } 1576 } 1577 1578 int 1579 linux_fadvise64(struct thread *td, struct linux_fadvise64_args *args) 1580 { 1581 off_t offset; 1582 int advice; 1583 1584 #if defined(__amd64__) && defined(COMPAT_LINUX32) 1585 offset = PAIR32TO64(off_t, args->offset); 1586 #else 1587 offset = args->offset; 1588 #endif 1589 1590 advice = convert_fadvice(args->advice); 1591 if (advice == -1) 1592 return (EINVAL); 1593 return (kern_posix_fadvise(td, args->fd, offset, args->len, advice)); 1594 } 1595 1596 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32)) 1597 int 1598 linux_fadvise64_64(struct thread *td, struct linux_fadvise64_64_args *args) 1599 { 1600 off_t len, offset; 1601 int advice; 1602 1603 #if defined(__amd64__) && defined(COMPAT_LINUX32) 1604 len = PAIR32TO64(off_t, args->len); 1605 offset = PAIR32TO64(off_t, args->offset); 1606 #else 1607 len = args->len; 1608 offset = args->offset; 1609 #endif 1610 1611 advice = convert_fadvice(args->advice); 1612 if (advice == -1) 1613 return (EINVAL); 1614 return (kern_posix_fadvise(td, args->fd, offset, len, advice)); 1615 } 1616 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */ 1617 1618 #ifdef LINUX_LEGACY_SYSCALLS 1619 int 1620 linux_pipe(struct thread *td, struct linux_pipe_args *args) 1621 { 1622 int fildes[2]; 1623 int error; 1624 1625 error = kern_pipe(td, fildes, 0, NULL, NULL); 1626 if (error != 0) 1627 return (error); 1628 1629 error = copyout(fildes, args->pipefds, sizeof(fildes)); 1630 if (error != 0) { 1631 (void)kern_close(td, fildes[0]); 1632 (void)kern_close(td, fildes[1]); 1633 } 1634 1635 return (error); 1636 } 1637 #endif 1638 1639 int 1640 linux_pipe2(struct thread *td, struct linux_pipe2_args *args) 1641 { 1642 int fildes[2]; 1643 int error, flags; 1644 1645 if ((args->flags & ~(LINUX_O_NONBLOCK | LINUX_O_CLOEXEC)) != 0) 1646 return (EINVAL); 1647 1648 flags = 0; 1649 if ((args->flags & LINUX_O_NONBLOCK) != 0) 1650 flags |= O_NONBLOCK; 1651 if ((args->flags & LINUX_O_CLOEXEC) != 0) 1652 flags |= O_CLOEXEC; 1653 error = kern_pipe(td, fildes, flags, NULL, NULL); 1654 if (error != 0) 1655 return (error); 1656 1657 error = copyout(fildes, args->pipefds, sizeof(fildes)); 1658 if (error != 0) { 1659 (void)kern_close(td, fildes[0]); 1660 (void)kern_close(td, fildes[1]); 1661 } 1662 1663 return (error); 1664 } 1665 1666 int 1667 linux_dup3(struct thread *td, struct linux_dup3_args *args) 1668 { 1669 int cmd; 1670 intptr_t newfd; 1671 1672 if (args->oldfd == args->newfd) 1673 return (EINVAL); 1674 if ((args->flags & ~LINUX_O_CLOEXEC) != 0) 1675 return (EINVAL); 1676 if (args->flags & LINUX_O_CLOEXEC) 1677 cmd = F_DUP2FD_CLOEXEC; 1678 else 1679 cmd = F_DUP2FD; 1680 1681 newfd = args->newfd; 1682 return (kern_fcntl(td, args->oldfd, cmd, newfd)); 1683 } 1684 1685 int 1686 linux_fallocate(struct thread *td, struct linux_fallocate_args *args) 1687 { 1688 off_t len, offset; 1689 1690 /* 1691 * We emulate only posix_fallocate system call for which 1692 * mode should be 0. 1693 */ 1694 if (args->mode != 0) 1695 return (EOPNOTSUPP); 1696 1697 #if defined(__amd64__) && defined(COMPAT_LINUX32) 1698 len = PAIR32TO64(off_t, args->len); 1699 offset = PAIR32TO64(off_t, args->offset); 1700 #else 1701 len = args->len; 1702 offset = args->offset; 1703 #endif 1704 1705 return (kern_posix_fallocate(td, args->fd, offset, len)); 1706 } 1707 1708 int 1709 linux_copy_file_range(struct thread *td, struct linux_copy_file_range_args 1710 *args) 1711 { 1712 l_loff_t inoff, outoff, *inoffp, *outoffp; 1713 int error, flags; 1714 1715 /* 1716 * copy_file_range(2) on Linux doesn't define any flags (yet), so is 1717 * the native implementation. Enforce it. 1718 */ 1719 if (args->flags != 0) { 1720 linux_msg(td, "copy_file_range unsupported flags 0x%x", 1721 args->flags); 1722 return (EINVAL); 1723 } 1724 flags = 0; 1725 inoffp = outoffp = NULL; 1726 if (args->off_in != NULL) { 1727 error = copyin(args->off_in, &inoff, sizeof(l_loff_t)); 1728 if (error != 0) 1729 return (error); 1730 inoffp = &inoff; 1731 } 1732 if (args->off_out != NULL) { 1733 error = copyin(args->off_out, &outoff, sizeof(l_loff_t)); 1734 if (error != 0) 1735 return (error); 1736 outoffp = &outoff; 1737 } 1738 1739 error = kern_copy_file_range(td, args->fd_in, inoffp, args->fd_out, 1740 outoffp, args->len, flags); 1741 if (error == 0 && args->off_in != NULL) 1742 error = copyout(inoffp, args->off_in, sizeof(l_loff_t)); 1743 if (error == 0 && args->off_out != NULL) 1744 error = copyout(outoffp, args->off_out, sizeof(l_loff_t)); 1745 return (error); 1746 } 1747 1748 #define LINUX_MEMFD_PREFIX "memfd:" 1749 1750 int 1751 linux_memfd_create(struct thread *td, struct linux_memfd_create_args *args) 1752 { 1753 char memfd_name[LINUX_NAME_MAX + 1]; 1754 int error, flags, shmflags, oflags; 1755 1756 /* 1757 * This is our clever trick to avoid the heap allocation to copy in the 1758 * uname. We don't really need to go this far out of our way, but it 1759 * does keep the rest of this function fairly clean as they don't have 1760 * to worry about cleanup on the way out. 1761 */ 1762 error = copyinstr(args->uname_ptr, 1763 memfd_name + sizeof(LINUX_MEMFD_PREFIX) - 1, 1764 LINUX_NAME_MAX - sizeof(LINUX_MEMFD_PREFIX) - 1, NULL); 1765 if (error != 0) { 1766 if (error == ENAMETOOLONG) 1767 error = EINVAL; 1768 return (error); 1769 } 1770 1771 memcpy(memfd_name, LINUX_MEMFD_PREFIX, sizeof(LINUX_MEMFD_PREFIX) - 1); 1772 flags = linux_to_bsd_bits(args->flags, mfd_bitmap, 0); 1773 if ((flags & ~(MFD_CLOEXEC | MFD_ALLOW_SEALING | MFD_HUGETLB | 1774 MFD_HUGE_MASK)) != 0) 1775 return (EINVAL); 1776 /* Size specified but no HUGETLB. */ 1777 if ((flags & MFD_HUGE_MASK) != 0 && (flags & MFD_HUGETLB) == 0) 1778 return (EINVAL); 1779 /* We don't actually support HUGETLB. */ 1780 if ((flags & MFD_HUGETLB) != 0) 1781 return (ENOSYS); 1782 oflags = O_RDWR; 1783 shmflags = SHM_GROW_ON_WRITE; 1784 if ((flags & MFD_CLOEXEC) != 0) 1785 oflags |= O_CLOEXEC; 1786 if ((flags & MFD_ALLOW_SEALING) != 0) 1787 shmflags |= SHM_ALLOW_SEALING; 1788 return (kern_shm_open2(td, SHM_ANON, oflags, 0, shmflags, NULL, 1789 memfd_name)); 1790 } 1791 1792 int 1793 linux_splice(struct thread *td, struct linux_splice_args *args) 1794 { 1795 1796 linux_msg(td, "syscall splice not really implemented"); 1797 1798 /* 1799 * splice(2) is documented to return EINVAL in various circumstances; 1800 * returning it instead of ENOSYS should hint the caller to use fallback 1801 * instead. 1802 */ 1803 return (EINVAL); 1804 } 1805 1806 int 1807 linux_close_range(struct thread *td, struct linux_close_range_args *args) 1808 { 1809 u_int flags = 0; 1810 1811 /* 1812 * Implementing close_range(CLOSE_RANGE_UNSHARE) allows Linux to 1813 * unshare filedesc table of the calling thread from others threads 1814 * in a thread group (i.e., process in the FreeBSD) or others processes, 1815 * which shares the same table, before closing the files. FreeBSD does 1816 * not have compatible unsharing mechanism due to the fact that sharing 1817 * process resources, including filedesc table, is at thread level in the 1818 * Linux, while in the FreeBSD it is at the process level. 1819 * Return EINVAL for now if the CLOSE_RANGE_UNSHARE flag is specified 1820 * until this new Linux API stabilizes. 1821 */ 1822 1823 if ((args->flags & ~(LINUX_CLOSE_RANGE_CLOEXEC)) != 0) 1824 return (EINVAL); 1825 if (args->first > args->last) 1826 return (EINVAL); 1827 if ((args->flags & LINUX_CLOSE_RANGE_CLOEXEC) != 0) 1828 flags |= CLOSE_RANGE_CLOEXEC; 1829 return (kern_close_range(td, flags, args->first, args->last)); 1830 } 1831