xref: /freebsd/sys/compat/linux/linux_file.c (revision 0957b409)
1 /*-
2  * SPDX-License-Identifier: BSD-2-Clause-FreeBSD
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 __FBSDID("$FreeBSD$");
31 
32 #include "opt_compat.h"
33 
34 #include <sys/param.h>
35 #include <sys/systm.h>
36 #include <sys/capsicum.h>
37 #include <sys/conf.h>
38 #include <sys/dirent.h>
39 #include <sys/fcntl.h>
40 #include <sys/file.h>
41 #include <sys/filedesc.h>
42 #include <sys/lock.h>
43 #include <sys/malloc.h>
44 #include <sys/mount.h>
45 #include <sys/mutex.h>
46 #include <sys/namei.h>
47 #include <sys/proc.h>
48 #include <sys/stat.h>
49 #include <sys/sx.h>
50 #include <sys/syscallsubr.h>
51 #include <sys/sysproto.h>
52 #include <sys/tty.h>
53 #include <sys/unistd.h>
54 #include <sys/vnode.h>
55 
56 #ifdef COMPAT_LINUX32
57 #include <machine/../linux32/linux.h>
58 #include <machine/../linux32/linux32_proto.h>
59 #else
60 #include <machine/../linux/linux.h>
61 #include <machine/../linux/linux_proto.h>
62 #endif
63 #include <compat/linux/linux_misc.h>
64 #include <compat/linux/linux_util.h>
65 #include <compat/linux/linux_file.h>
66 
67 static int	linux_common_open(struct thread *, int, char *, int, int);
68 static int	linux_getdents_error(struct thread *, int, int);
69 
70 
71 #ifdef LINUX_LEGACY_SYSCALLS
72 int
73 linux_creat(struct thread *td, struct linux_creat_args *args)
74 {
75 	char *path;
76 	int error;
77 
78 	LCONVPATHEXIST(td, args->path, &path);
79 #ifdef DEBUG
80 	if (ldebug(creat))
81 		printf(ARGS(creat, "%s, %d"), path, args->mode);
82 #endif
83 	error = kern_openat(td, AT_FDCWD, path, UIO_SYSSPACE,
84 	    O_WRONLY | O_CREAT | O_TRUNC, args->mode);
85 	LFREEPATH(path);
86 	return (error);
87 }
88 #endif
89 
90 static int
91 linux_common_open(struct thread *td, int dirfd, char *path, int l_flags, int mode)
92 {
93 	struct proc *p = td->td_proc;
94 	struct file *fp;
95 	int fd;
96 	int bsd_flags, error;
97 
98 	bsd_flags = 0;
99 	switch (l_flags & LINUX_O_ACCMODE) {
100 	case LINUX_O_WRONLY:
101 		bsd_flags |= O_WRONLY;
102 		break;
103 	case LINUX_O_RDWR:
104 		bsd_flags |= O_RDWR;
105 		break;
106 	default:
107 		bsd_flags |= O_RDONLY;
108 	}
109 	if (l_flags & LINUX_O_NDELAY)
110 		bsd_flags |= O_NONBLOCK;
111 	if (l_flags & LINUX_O_APPEND)
112 		bsd_flags |= O_APPEND;
113 	if (l_flags & LINUX_O_SYNC)
114 		bsd_flags |= O_FSYNC;
115 	if (l_flags & LINUX_O_NONBLOCK)
116 		bsd_flags |= O_NONBLOCK;
117 	if (l_flags & LINUX_FASYNC)
118 		bsd_flags |= O_ASYNC;
119 	if (l_flags & LINUX_O_CREAT)
120 		bsd_flags |= O_CREAT;
121 	if (l_flags & LINUX_O_TRUNC)
122 		bsd_flags |= O_TRUNC;
123 	if (l_flags & LINUX_O_EXCL)
124 		bsd_flags |= O_EXCL;
125 	if (l_flags & LINUX_O_NOCTTY)
126 		bsd_flags |= O_NOCTTY;
127 	if (l_flags & LINUX_O_DIRECT)
128 		bsd_flags |= O_DIRECT;
129 	if (l_flags & LINUX_O_NOFOLLOW)
130 		bsd_flags |= O_NOFOLLOW;
131 	if (l_flags & LINUX_O_DIRECTORY)
132 		bsd_flags |= O_DIRECTORY;
133 	/* XXX LINUX_O_NOATIME: unable to be easily implemented. */
134 
135 	error = kern_openat(td, dirfd, path, UIO_SYSSPACE, bsd_flags, mode);
136 	if (error != 0)
137 		goto done;
138 	if (bsd_flags & O_NOCTTY)
139 		goto done;
140 
141 	/*
142 	 * XXX In between kern_openat() and fget(), another process
143 	 * having the same filedesc could use that fd without
144 	 * checking below.
145 	*/
146 	fd = td->td_retval[0];
147 	if (fget(td, fd, &cap_ioctl_rights, &fp) == 0) {
148 		if (fp->f_type != DTYPE_VNODE) {
149 			fdrop(fp, td);
150 			goto done;
151 		}
152 		sx_slock(&proctree_lock);
153 		PROC_LOCK(p);
154 		if (SESS_LEADER(p) && !(p->p_flag & P_CONTROLT)) {
155 			PROC_UNLOCK(p);
156 			sx_sunlock(&proctree_lock);
157 			/* XXXPJD: Verify if TIOCSCTTY is allowed. */
158 			(void) fo_ioctl(fp, TIOCSCTTY, (caddr_t) 0,
159 			    td->td_ucred, td);
160 		} else {
161 			PROC_UNLOCK(p);
162 			sx_sunlock(&proctree_lock);
163 		}
164 		fdrop(fp, td);
165 	}
166 
167 done:
168 #ifdef DEBUG
169 #ifdef LINUX_LEGACY_SYSCALLS
170 	if (ldebug(open))
171 #else
172 	if (ldebug(openat))
173 #endif
174 		printf(LMSG("open returns error %d"), error);
175 #endif
176 	LFREEPATH(path);
177 	return (error);
178 }
179 
180 int
181 linux_openat(struct thread *td, struct linux_openat_args *args)
182 {
183 	char *path;
184 	int dfd;
185 
186 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
187 	if (args->flags & LINUX_O_CREAT)
188 		LCONVPATH_AT(td, args->filename, &path, 1, dfd);
189 	else
190 		LCONVPATH_AT(td, args->filename, &path, 0, dfd);
191 #ifdef DEBUG
192 	if (ldebug(openat))
193 		printf(ARGS(openat, "%i, %s, 0x%x, 0x%x"), args->dfd,
194 		    path, args->flags, args->mode);
195 #endif
196 	return (linux_common_open(td, dfd, path, args->flags, args->mode));
197 }
198 
199 #ifdef LINUX_LEGACY_SYSCALLS
200 int
201 linux_open(struct thread *td, struct linux_open_args *args)
202 {
203 	char *path;
204 
205 	if (args->flags & LINUX_O_CREAT)
206 		LCONVPATHCREAT(td, args->path, &path);
207 	else
208 		LCONVPATHEXIST(td, args->path, &path);
209 #ifdef DEBUG
210 	if (ldebug(open))
211 		printf(ARGS(open, "%s, 0x%x, 0x%x"),
212 		    path, args->flags, args->mode);
213 #endif
214 	return (linux_common_open(td, AT_FDCWD, path, args->flags, args->mode));
215 }
216 #endif
217 
218 int
219 linux_lseek(struct thread *td, struct linux_lseek_args *args)
220 {
221 
222 #ifdef DEBUG
223 	if (ldebug(lseek))
224 		printf(ARGS(lseek, "%d, %ld, %d"),
225 		    args->fdes, (long)args->off, args->whence);
226 #endif
227 	return (kern_lseek(td, args->fdes, args->off, args->whence));
228 }
229 
230 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
231 int
232 linux_llseek(struct thread *td, struct linux_llseek_args *args)
233 {
234 	int error;
235 	off_t off;
236 
237 #ifdef DEBUG
238 	if (ldebug(llseek))
239 		printf(ARGS(llseek, "%d, %d:%d, %d"),
240 		    args->fd, args->ohigh, args->olow, args->whence);
241 #endif
242 	off = (args->olow) | (((off_t) args->ohigh) << 32);
243 
244 	error = kern_lseek(td, args->fd, off, args->whence);
245 	if (error != 0)
246 		return (error);
247 
248 	error = copyout(td->td_retval, args->res, sizeof(off_t));
249 	if (error != 0)
250 		return (error);
251 
252 	td->td_retval[0] = 0;
253 	return (0);
254 }
255 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
256 
257 /*
258  * Note that linux_getdents(2) and linux_getdents64(2) have the same
259  * arguments. They only differ in the definition of struct dirent they
260  * operate on.
261  * Note that linux_readdir(2) is a special case of linux_getdents(2)
262  * where count is always equals 1, meaning that the buffer is one
263  * dirent-structure in size and that the code can't handle more anyway.
264  * Note that linux_readdir(2) can't be implemented by means of linux_getdents(2)
265  * as in case when the *dent buffer size is equal to 1 linux_getdents(2) will
266  * trash user stack.
267  */
268 
269 static int
270 linux_getdents_error(struct thread *td, int fd, int err)
271 {
272 	struct vnode *vp;
273 	struct file *fp;
274 	int error;
275 
276 	/* Linux return ENOTDIR in case when fd is not a directory. */
277 	error = getvnode(td, fd, &cap_read_rights, &fp);
278 	if (error != 0)
279 		return (error);
280 	vp = fp->f_vnode;
281 	if (vp->v_type != VDIR) {
282 		fdrop(fp, td);
283 		return (ENOTDIR);
284 	}
285 	fdrop(fp, td);
286 	return (err);
287 }
288 
289 struct l_dirent {
290 	l_ulong		d_ino;
291 	l_off_t		d_off;
292 	l_ushort	d_reclen;
293 	char		d_name[LINUX_NAME_MAX + 1];
294 };
295 
296 struct l_dirent64 {
297 	uint64_t	d_ino;
298 	int64_t		d_off;
299 	l_ushort	d_reclen;
300 	u_char		d_type;
301 	char		d_name[LINUX_NAME_MAX + 1];
302 };
303 
304 /*
305  * Linux uses the last byte in the dirent buffer to store d_type,
306  * at least glibc-2.7 requires it. That is why l_dirent is padded with 2 bytes.
307  */
308 #define LINUX_RECLEN(namlen)						\
309     roundup(offsetof(struct l_dirent, d_name) + (namlen) + 2, sizeof(l_ulong))
310 
311 #define LINUX_RECLEN64(namlen)						\
312     roundup(offsetof(struct l_dirent64, d_name) + (namlen) + 1,		\
313     sizeof(uint64_t))
314 
315 #ifdef LINUX_LEGACY_SYSCALLS
316 int
317 linux_getdents(struct thread *td, struct linux_getdents_args *args)
318 {
319 	struct dirent *bdp;
320 	caddr_t inp, buf;		/* BSD-format */
321 	int len, reclen;		/* BSD-format */
322 	caddr_t outp;			/* Linux-format */
323 	int resid, linuxreclen;		/* Linux-format */
324 	caddr_t lbuf;			/* Linux-format */
325 	off_t base;
326 	struct l_dirent *linux_dirent;
327 	int buflen, error;
328 	size_t retval;
329 
330 #ifdef DEBUG
331 	if (ldebug(getdents))
332 		printf(ARGS(getdents, "%d, *, %d"), args->fd, args->count);
333 #endif
334 	buflen = min(args->count, MAXBSIZE);
335 	buf = malloc(buflen, M_TEMP, M_WAITOK);
336 
337 	error = kern_getdirentries(td, args->fd, buf, buflen,
338 	    &base, NULL, UIO_SYSSPACE);
339 	if (error != 0) {
340 		error = linux_getdents_error(td, args->fd, error);
341 		goto out1;
342 	}
343 
344 	lbuf = malloc(LINUX_RECLEN(LINUX_NAME_MAX), M_TEMP, M_WAITOK | M_ZERO);
345 
346 	len = td->td_retval[0];
347 	inp = buf;
348 	outp = (caddr_t)args->dent;
349 	resid = args->count;
350 	retval = 0;
351 
352 	while (len > 0) {
353 		bdp = (struct dirent *) inp;
354 		reclen = bdp->d_reclen;
355 		linuxreclen = LINUX_RECLEN(bdp->d_namlen);
356 		/*
357 		 * No more space in the user supplied dirent buffer.
358 		 * Return EINVAL.
359 		 */
360 		if (resid < linuxreclen) {
361 			error = EINVAL;
362 			goto out;
363 		}
364 
365 		linux_dirent = (struct l_dirent*)lbuf;
366 		linux_dirent->d_ino = bdp->d_fileno;
367 		linux_dirent->d_off = base + reclen;
368 		linux_dirent->d_reclen = linuxreclen;
369 		/*
370 		 * Copy d_type to last byte of l_dirent buffer
371 		 */
372 		lbuf[linuxreclen - 1] = bdp->d_type;
373 		strlcpy(linux_dirent->d_name, bdp->d_name,
374 		    linuxreclen - offsetof(struct l_dirent, d_name)-1);
375 		error = copyout(linux_dirent, outp, linuxreclen);
376 		if (error != 0)
377 			goto out;
378 
379 		inp += reclen;
380 		base += reclen;
381 		len -= reclen;
382 
383 		retval += linuxreclen;
384 		outp += linuxreclen;
385 		resid -= linuxreclen;
386 	}
387 	td->td_retval[0] = retval;
388 
389 out:
390 	free(lbuf, M_TEMP);
391 out1:
392 	free(buf, M_TEMP);
393 	return (error);
394 }
395 #endif
396 
397 int
398 linux_getdents64(struct thread *td, struct linux_getdents64_args *args)
399 {
400 	struct dirent *bdp;
401 	caddr_t inp, buf;		/* BSD-format */
402 	int len, reclen;		/* BSD-format */
403 	caddr_t outp;			/* Linux-format */
404 	int resid, linuxreclen;		/* Linux-format */
405 	caddr_t lbuf;			/* Linux-format */
406 	off_t base;
407 	struct l_dirent64 *linux_dirent64;
408 	int buflen, error;
409 	size_t retval;
410 
411 #ifdef DEBUG
412 	if (ldebug(getdents64))
413 		uprintf(ARGS(getdents64, "%d, *, %d"), args->fd, args->count);
414 #endif
415 	buflen = min(args->count, MAXBSIZE);
416 	buf = malloc(buflen, M_TEMP, M_WAITOK);
417 
418 	error = kern_getdirentries(td, args->fd, buf, buflen,
419 	    &base, NULL, UIO_SYSSPACE);
420 	if (error != 0) {
421 		error = linux_getdents_error(td, args->fd, error);
422 		goto out1;
423 	}
424 
425 	lbuf = malloc(LINUX_RECLEN64(LINUX_NAME_MAX), M_TEMP, M_WAITOK | M_ZERO);
426 
427 	len = td->td_retval[0];
428 	inp = buf;
429 	outp = (caddr_t)args->dirent;
430 	resid = args->count;
431 	retval = 0;
432 
433 	while (len > 0) {
434 		bdp = (struct dirent *) inp;
435 		reclen = bdp->d_reclen;
436 		linuxreclen = LINUX_RECLEN64(bdp->d_namlen);
437 		/*
438 		 * No more space in the user supplied dirent buffer.
439 		 * Return EINVAL.
440 		 */
441 		if (resid < linuxreclen) {
442 			error = EINVAL;
443 			goto out;
444 		}
445 
446 		linux_dirent64 = (struct l_dirent64*)lbuf;
447 		linux_dirent64->d_ino = bdp->d_fileno;
448 		linux_dirent64->d_off = base + reclen;
449 		linux_dirent64->d_reclen = linuxreclen;
450 		linux_dirent64->d_type = bdp->d_type;
451 		strlcpy(linux_dirent64->d_name, bdp->d_name,
452 		    linuxreclen - offsetof(struct l_dirent64, d_name));
453 		error = copyout(linux_dirent64, outp, linuxreclen);
454 		if (error != 0)
455 			goto out;
456 
457 		inp += reclen;
458 		base += reclen;
459 		len -= reclen;
460 
461 		retval += linuxreclen;
462 		outp += linuxreclen;
463 		resid -= linuxreclen;
464 	}
465 	td->td_retval[0] = retval;
466 
467 out:
468 	free(lbuf, M_TEMP);
469 out1:
470 	free(buf, M_TEMP);
471 	return (error);
472 }
473 
474 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
475 int
476 linux_readdir(struct thread *td, struct linux_readdir_args *args)
477 {
478 	struct dirent *bdp;
479 	caddr_t buf;			/* BSD-format */
480 	int linuxreclen;		/* Linux-format */
481 	caddr_t lbuf;			/* Linux-format */
482 	off_t base;
483 	struct l_dirent *linux_dirent;
484 	int buflen, error;
485 
486 #ifdef DEBUG
487 	if (ldebug(readdir))
488 		printf(ARGS(readdir, "%d, *"), args->fd);
489 #endif
490 	buflen = LINUX_RECLEN(LINUX_NAME_MAX);
491 	buf = malloc(buflen, M_TEMP, M_WAITOK);
492 
493 	error = kern_getdirentries(td, args->fd, buf, buflen,
494 	    &base, NULL, UIO_SYSSPACE);
495 	if (error != 0) {
496 		error = linux_getdents_error(td, args->fd, error);
497 		goto out;
498 	}
499 	if (td->td_retval[0] == 0)
500 		goto out;
501 
502 	lbuf = malloc(LINUX_RECLEN(LINUX_NAME_MAX), M_TEMP, M_WAITOK | M_ZERO);
503 
504 	bdp = (struct dirent *) buf;
505 	linuxreclen = LINUX_RECLEN(bdp->d_namlen);
506 
507 	linux_dirent = (struct l_dirent*)lbuf;
508 	linux_dirent->d_ino = bdp->d_fileno;
509 	linux_dirent->d_off = linuxreclen;
510 	linux_dirent->d_reclen = bdp->d_namlen;
511 	strlcpy(linux_dirent->d_name, bdp->d_name,
512 	    linuxreclen - offsetof(struct l_dirent, d_name));
513 	error = copyout(linux_dirent, args->dent, linuxreclen);
514 	if (error == 0)
515 		td->td_retval[0] = linuxreclen;
516 
517 	free(lbuf, M_TEMP);
518 out:
519 	free(buf, M_TEMP);
520 	return (error);
521 }
522 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
523 
524 
525 /*
526  * These exist mainly for hooks for doing /compat/linux translation.
527  */
528 
529 #ifdef LINUX_LEGACY_SYSCALLS
530 int
531 linux_access(struct thread *td, struct linux_access_args *args)
532 {
533 	char *path;
534 	int error;
535 
536 	/* Linux convention. */
537 	if (args->amode & ~(F_OK | X_OK | W_OK | R_OK))
538 		return (EINVAL);
539 
540 	LCONVPATHEXIST(td, args->path, &path);
541 
542 #ifdef DEBUG
543 	if (ldebug(access))
544 		printf(ARGS(access, "%s, %d"), path, args->amode);
545 #endif
546 	error = kern_accessat(td, AT_FDCWD, path, UIO_SYSSPACE, 0,
547 	    args->amode);
548 	LFREEPATH(path);
549 
550 	return (error);
551 }
552 #endif
553 
554 int
555 linux_faccessat(struct thread *td, struct linux_faccessat_args *args)
556 {
557 	char *path;
558 	int error, dfd;
559 
560 	/* Linux convention. */
561 	if (args->amode & ~(F_OK | X_OK | W_OK | R_OK))
562 		return (EINVAL);
563 
564 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
565 	LCONVPATHEXIST_AT(td, args->filename, &path, dfd);
566 
567 #ifdef DEBUG
568 	if (ldebug(faccessat))
569 		printf(ARGS(access, "%s, %d"), path, args->amode);
570 #endif
571 
572 	error = kern_accessat(td, dfd, path, UIO_SYSSPACE, 0, args->amode);
573 	LFREEPATH(path);
574 
575 	return (error);
576 }
577 
578 #ifdef LINUX_LEGACY_SYSCALLS
579 int
580 linux_unlink(struct thread *td, struct linux_unlink_args *args)
581 {
582 	char *path;
583 	int error;
584 	struct stat st;
585 
586 	LCONVPATHEXIST(td, args->path, &path);
587 
588 #ifdef DEBUG
589 	if (ldebug(unlink))
590 		printf(ARGS(unlink, "%s"), path);
591 #endif
592 
593 	error = kern_unlinkat(td, AT_FDCWD, path, UIO_SYSSPACE, 0, 0);
594 	if (error == EPERM) {
595 		/* Introduce POSIX noncompliant behaviour of Linux */
596 		if (kern_statat(td, 0, AT_FDCWD, path, UIO_SYSSPACE, &st,
597 		    NULL) == 0) {
598 			if (S_ISDIR(st.st_mode))
599 				error = EISDIR;
600 		}
601 	}
602 	LFREEPATH(path);
603 	return (error);
604 }
605 #endif
606 
607 int
608 linux_unlinkat(struct thread *td, struct linux_unlinkat_args *args)
609 {
610 	char *path;
611 	int error, dfd;
612 	struct stat st;
613 
614 	if (args->flag & ~LINUX_AT_REMOVEDIR)
615 		return (EINVAL);
616 
617 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
618 	LCONVPATHEXIST_AT(td, args->pathname, &path, dfd);
619 
620 #ifdef DEBUG
621 	if (ldebug(unlinkat))
622 		printf(ARGS(unlinkat, "%s"), path);
623 #endif
624 
625 	if (args->flag & LINUX_AT_REMOVEDIR)
626 		error = kern_rmdirat(td, dfd, path, UIO_SYSSPACE, 0);
627 	else
628 		error = kern_unlinkat(td, dfd, path, UIO_SYSSPACE, 0, 0);
629 	if (error == EPERM && !(args->flag & LINUX_AT_REMOVEDIR)) {
630 		/* Introduce POSIX noncompliant behaviour of Linux */
631 		if (kern_statat(td, AT_SYMLINK_NOFOLLOW, dfd, path,
632 		    UIO_SYSSPACE, &st, NULL) == 0 && S_ISDIR(st.st_mode))
633 			error = EISDIR;
634 	}
635 	LFREEPATH(path);
636 	return (error);
637 }
638 int
639 linux_chdir(struct thread *td, struct linux_chdir_args *args)
640 {
641 	char *path;
642 	int error;
643 
644 	LCONVPATHEXIST(td, args->path, &path);
645 
646 #ifdef DEBUG
647 	if (ldebug(chdir))
648 		printf(ARGS(chdir, "%s"), path);
649 #endif
650 	error = kern_chdir(td, path, UIO_SYSSPACE);
651 	LFREEPATH(path);
652 	return (error);
653 }
654 
655 #ifdef LINUX_LEGACY_SYSCALLS
656 int
657 linux_chmod(struct thread *td, struct linux_chmod_args *args)
658 {
659 	char *path;
660 	int error;
661 
662 	LCONVPATHEXIST(td, args->path, &path);
663 
664 #ifdef DEBUG
665 	if (ldebug(chmod))
666 		printf(ARGS(chmod, "%s, %d"), path, args->mode);
667 #endif
668 	error = kern_fchmodat(td, AT_FDCWD, path, UIO_SYSSPACE,
669 	    args->mode, 0);
670 	LFREEPATH(path);
671 	return (error);
672 }
673 #endif
674 
675 int
676 linux_fchmodat(struct thread *td, struct linux_fchmodat_args *args)
677 {
678 	char *path;
679 	int error, dfd;
680 
681 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
682 	LCONVPATHEXIST_AT(td, args->filename, &path, dfd);
683 
684 #ifdef DEBUG
685 	if (ldebug(fchmodat))
686 		printf(ARGS(fchmodat, "%s, %d"), path, args->mode);
687 #endif
688 
689 	error = kern_fchmodat(td, dfd, path, UIO_SYSSPACE, args->mode, 0);
690 	LFREEPATH(path);
691 	return (error);
692 }
693 
694 #ifdef LINUX_LEGACY_SYSCALLS
695 int
696 linux_mkdir(struct thread *td, struct linux_mkdir_args *args)
697 {
698 	char *path;
699 	int error;
700 
701 	LCONVPATHCREAT(td, args->path, &path);
702 
703 #ifdef DEBUG
704 	if (ldebug(mkdir))
705 		printf(ARGS(mkdir, "%s, %d"), path, args->mode);
706 #endif
707 	error = kern_mkdirat(td, AT_FDCWD, path, UIO_SYSSPACE, args->mode);
708 	LFREEPATH(path);
709 	return (error);
710 }
711 #endif
712 
713 int
714 linux_mkdirat(struct thread *td, struct linux_mkdirat_args *args)
715 {
716 	char *path;
717 	int error, dfd;
718 
719 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
720 	LCONVPATHCREAT_AT(td, args->pathname, &path, dfd);
721 
722 #ifdef DEBUG
723 	if (ldebug(mkdirat))
724 		printf(ARGS(mkdirat, "%s, %d"), path, args->mode);
725 #endif
726 	error = kern_mkdirat(td, dfd, path, UIO_SYSSPACE, args->mode);
727 	LFREEPATH(path);
728 	return (error);
729 }
730 
731 #ifdef LINUX_LEGACY_SYSCALLS
732 int
733 linux_rmdir(struct thread *td, struct linux_rmdir_args *args)
734 {
735 	char *path;
736 	int error;
737 
738 	LCONVPATHEXIST(td, args->path, &path);
739 
740 #ifdef DEBUG
741 	if (ldebug(rmdir))
742 		printf(ARGS(rmdir, "%s"), path);
743 #endif
744 	error = kern_rmdirat(td, AT_FDCWD, path, UIO_SYSSPACE, 0);
745 	LFREEPATH(path);
746 	return (error);
747 }
748 
749 int
750 linux_rename(struct thread *td, struct linux_rename_args *args)
751 {
752 	char *from, *to;
753 	int error;
754 
755 	LCONVPATHEXIST(td, args->from, &from);
756 	/* Expand LCONVPATHCREATE so that `from' can be freed on errors */
757 	error = linux_emul_convpath(td, args->to, UIO_USERSPACE, &to, 1, AT_FDCWD);
758 	if (to == NULL) {
759 		LFREEPATH(from);
760 		return (error);
761 	}
762 
763 #ifdef DEBUG
764 	if (ldebug(rename))
765 		printf(ARGS(rename, "%s, %s"), from, to);
766 #endif
767 	error = kern_renameat(td, AT_FDCWD, from, AT_FDCWD, to, UIO_SYSSPACE);
768 	LFREEPATH(from);
769 	LFREEPATH(to);
770 	return (error);
771 }
772 #endif
773 
774 int
775 linux_renameat(struct thread *td, struct linux_renameat_args *args)
776 {
777 	char *from, *to;
778 	int error, olddfd, newdfd;
779 
780 	olddfd = (args->olddfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->olddfd;
781 	newdfd = (args->newdfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->newdfd;
782 	LCONVPATHEXIST_AT(td, args->oldname, &from, olddfd);
783 	/* Expand LCONVPATHCREATE so that `from' can be freed on errors */
784 	error = linux_emul_convpath(td, args->newname, UIO_USERSPACE, &to, 1, newdfd);
785 	if (to == NULL) {
786 		LFREEPATH(from);
787 		return (error);
788 	}
789 
790 #ifdef DEBUG
791 	if (ldebug(renameat))
792 		printf(ARGS(renameat, "%s, %s"), from, to);
793 #endif
794 	error = kern_renameat(td, olddfd, from, newdfd, to, UIO_SYSSPACE);
795 	LFREEPATH(from);
796 	LFREEPATH(to);
797 	return (error);
798 }
799 
800 #ifdef LINUX_LEGACY_SYSCALLS
801 int
802 linux_symlink(struct thread *td, struct linux_symlink_args *args)
803 {
804 	char *path, *to;
805 	int error;
806 
807 	LCONVPATHEXIST(td, args->path, &path);
808 	/* Expand LCONVPATHCREATE so that `path' can be freed on errors */
809 	error = linux_emul_convpath(td, args->to, UIO_USERSPACE, &to, 1, AT_FDCWD);
810 	if (to == NULL) {
811 		LFREEPATH(path);
812 		return (error);
813 	}
814 
815 #ifdef DEBUG
816 	if (ldebug(symlink))
817 		printf(ARGS(symlink, "%s, %s"), path, to);
818 #endif
819 	error = kern_symlinkat(td, path, AT_FDCWD, to, UIO_SYSSPACE);
820 	LFREEPATH(path);
821 	LFREEPATH(to);
822 	return (error);
823 }
824 #endif
825 
826 int
827 linux_symlinkat(struct thread *td, struct linux_symlinkat_args *args)
828 {
829 	char *path, *to;
830 	int error, dfd;
831 
832 	dfd = (args->newdfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->newdfd;
833 	LCONVPATHEXIST(td, args->oldname, &path);
834 	/* Expand LCONVPATHCREATE so that `path' can be freed on errors */
835 	error = linux_emul_convpath(td, args->newname, UIO_USERSPACE, &to, 1, dfd);
836 	if (to == NULL) {
837 		LFREEPATH(path);
838 		return (error);
839 	}
840 
841 #ifdef DEBUG
842 	if (ldebug(symlinkat))
843 		printf(ARGS(symlinkat, "%s, %s"), path, to);
844 #endif
845 
846 	error = kern_symlinkat(td, path, dfd, to, UIO_SYSSPACE);
847 	LFREEPATH(path);
848 	LFREEPATH(to);
849 	return (error);
850 }
851 
852 #ifdef LINUX_LEGACY_SYSCALLS
853 int
854 linux_readlink(struct thread *td, struct linux_readlink_args *args)
855 {
856 	char *name;
857 	int error;
858 
859 	LCONVPATHEXIST(td, args->name, &name);
860 
861 #ifdef DEBUG
862 	if (ldebug(readlink))
863 		printf(ARGS(readlink, "%s, %p, %d"), name, (void *)args->buf,
864 		    args->count);
865 #endif
866 	error = kern_readlinkat(td, AT_FDCWD, name, UIO_SYSSPACE,
867 	    args->buf, UIO_USERSPACE, args->count);
868 	LFREEPATH(name);
869 	return (error);
870 }
871 #endif
872 
873 int
874 linux_readlinkat(struct thread *td, struct linux_readlinkat_args *args)
875 {
876 	char *name;
877 	int error, dfd;
878 
879 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
880 	LCONVPATHEXIST_AT(td, args->path, &name, dfd);
881 
882 #ifdef DEBUG
883 	if (ldebug(readlinkat))
884 		printf(ARGS(readlinkat, "%s, %p, %d"), name, (void *)args->buf,
885 		    args->bufsiz);
886 #endif
887 
888 	error = kern_readlinkat(td, dfd, name, UIO_SYSSPACE, args->buf,
889 	    UIO_USERSPACE, args->bufsiz);
890 	LFREEPATH(name);
891 	return (error);
892 }
893 
894 int
895 linux_truncate(struct thread *td, struct linux_truncate_args *args)
896 {
897 	char *path;
898 	int error;
899 
900 	LCONVPATHEXIST(td, args->path, &path);
901 
902 #ifdef DEBUG
903 	if (ldebug(truncate))
904 		printf(ARGS(truncate, "%s, %ld"), path, (long)args->length);
905 #endif
906 
907 	error = kern_truncate(td, path, UIO_SYSSPACE, args->length);
908 	LFREEPATH(path);
909 	return (error);
910 }
911 
912 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
913 int
914 linux_truncate64(struct thread *td, struct linux_truncate64_args *args)
915 {
916 	char *path;
917 	int error;
918 
919 	LCONVPATHEXIST(td, args->path, &path);
920 
921 #ifdef DEBUG
922 	if (ldebug(truncate64))
923 		printf(ARGS(truncate64, "%s, %jd"), path, args->length);
924 #endif
925 
926 	error = kern_truncate(td, path, UIO_SYSSPACE, args->length);
927 	LFREEPATH(path);
928 	return (error);
929 }
930 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
931 
932 int
933 linux_ftruncate(struct thread *td, struct linux_ftruncate_args *args)
934 {
935 
936 	return (kern_ftruncate(td, args->fd, args->length));
937 }
938 
939 #ifdef LINUX_LEGACY_SYSCALLS
940 int
941 linux_link(struct thread *td, struct linux_link_args *args)
942 {
943 	char *path, *to;
944 	int error;
945 
946 	LCONVPATHEXIST(td, args->path, &path);
947 	/* Expand LCONVPATHCREATE so that `path' can be freed on errors */
948 	error = linux_emul_convpath(td, args->to, UIO_USERSPACE, &to, 1, AT_FDCWD);
949 	if (to == NULL) {
950 		LFREEPATH(path);
951 		return (error);
952 	}
953 
954 #ifdef DEBUG
955 	if (ldebug(link))
956 		printf(ARGS(link, "%s, %s"), path, to);
957 #endif
958 	error = kern_linkat(td, AT_FDCWD, AT_FDCWD, path, to, UIO_SYSSPACE,
959 	    FOLLOW);
960 	LFREEPATH(path);
961 	LFREEPATH(to);
962 	return (error);
963 }
964 #endif
965 
966 int
967 linux_linkat(struct thread *td, struct linux_linkat_args *args)
968 {
969 	char *path, *to;
970 	int error, olddfd, newdfd, follow;
971 
972 	if (args->flag & ~LINUX_AT_SYMLINK_FOLLOW)
973 		return (EINVAL);
974 
975 	olddfd = (args->olddfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->olddfd;
976 	newdfd = (args->newdfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->newdfd;
977 	LCONVPATHEXIST_AT(td, args->oldname, &path, olddfd);
978 	/* Expand LCONVPATHCREATE so that `path' can be freed on errors */
979 	error = linux_emul_convpath(td, args->newname, UIO_USERSPACE, &to, 1, newdfd);
980 	if (to == NULL) {
981 		LFREEPATH(path);
982 		return (error);
983 	}
984 
985 #ifdef DEBUG
986 	if (ldebug(linkat))
987 		printf(ARGS(linkat, "%i, %s, %i, %s, %i"), args->olddfd, path,
988 			args->newdfd, to, args->flag);
989 #endif
990 
991 	follow = (args->flag & LINUX_AT_SYMLINK_FOLLOW) == 0 ? NOFOLLOW :
992 	    FOLLOW;
993 	error = kern_linkat(td, olddfd, newdfd, path, to, UIO_SYSSPACE, follow);
994 	LFREEPATH(path);
995 	LFREEPATH(to);
996 	return (error);
997 }
998 
999 int
1000 linux_fdatasync(td, uap)
1001 	struct thread *td;
1002 	struct linux_fdatasync_args *uap;
1003 {
1004 
1005 	return (kern_fsync(td, uap->fd, false));
1006 }
1007 
1008 int
1009 linux_pread(struct thread *td, struct linux_pread_args *uap)
1010 {
1011 	struct vnode *vp;
1012 	int error;
1013 
1014 	error = kern_pread(td, uap->fd, uap->buf, uap->nbyte, uap->offset);
1015 	if (error == 0) {
1016 		/* This seems to violate POSIX but Linux does it. */
1017 		error = fgetvp(td, uap->fd, &cap_pread_rights, &vp);
1018 		if (error != 0)
1019 			return (error);
1020 		if (vp->v_type == VDIR) {
1021 			vrele(vp);
1022 			return (EISDIR);
1023 		}
1024 		vrele(vp);
1025 	}
1026 	return (error);
1027 }
1028 
1029 int
1030 linux_pwrite(struct thread *td, struct linux_pwrite_args *uap)
1031 {
1032 
1033 	return (kern_pwrite(td, uap->fd, uap->buf, uap->nbyte, uap->offset));
1034 }
1035 
1036 int
1037 linux_preadv(struct thread *td, struct linux_preadv_args *uap)
1038 {
1039 	struct uio *auio;
1040 	int error;
1041 	off_t offset;
1042 
1043 	/*
1044 	 * According http://man7.org/linux/man-pages/man2/preadv.2.html#NOTES
1045 	 * pos_l and pos_h, respectively, contain the
1046 	 * low order and high order 32 bits of offset.
1047 	 */
1048 	offset = (((off_t)uap->pos_h << (sizeof(offset) * 4)) <<
1049 	    (sizeof(offset) * 4)) | uap->pos_l;
1050 	if (offset < 0)
1051 		return (EINVAL);
1052 #ifdef COMPAT_LINUX32
1053 	error = linux32_copyinuio(PTRIN(uap->vec), uap->vlen, &auio);
1054 #else
1055 	error = copyinuio(uap->vec, uap->vlen, &auio);
1056 #endif
1057 	if (error != 0)
1058 		return (error);
1059 	error = kern_preadv(td, uap->fd, auio, offset);
1060 	free(auio, M_IOV);
1061 	return (error);
1062 }
1063 
1064 int
1065 linux_pwritev(struct thread *td, struct linux_pwritev_args *uap)
1066 {
1067 	struct uio *auio;
1068 	int error;
1069 	off_t offset;
1070 
1071 	/*
1072 	 * According http://man7.org/linux/man-pages/man2/pwritev.2.html#NOTES
1073 	 * pos_l and pos_h, respectively, contain the
1074 	 * low order and high order 32 bits of offset.
1075 	 */
1076 	offset = (((off_t)uap->pos_h << (sizeof(offset) * 4)) <<
1077 	    (sizeof(offset) * 4)) | uap->pos_l;
1078 	if (offset < 0)
1079 		return (EINVAL);
1080 #ifdef COMPAT_LINUX32
1081 	error = linux32_copyinuio(PTRIN(uap->vec), uap->vlen, &auio);
1082 #else
1083 	error = copyinuio(uap->vec, uap->vlen, &auio);
1084 #endif
1085 	if (error != 0)
1086 		return (error);
1087 	error = kern_pwritev(td, uap->fd, auio, offset);
1088 	free(auio, M_IOV);
1089 	return (error);
1090 }
1091 
1092 int
1093 linux_mount(struct thread *td, struct linux_mount_args *args)
1094 {
1095 	char fstypename[MFSNAMELEN];
1096 	char *mntonname, *mntfromname;
1097 	int error, fsflags;
1098 
1099 	mntonname = malloc(MNAMELEN, M_TEMP, M_WAITOK);
1100 	mntfromname = malloc(MNAMELEN, M_TEMP, M_WAITOK);
1101 	error = copyinstr(args->filesystemtype, fstypename, MFSNAMELEN - 1,
1102 	    NULL);
1103 	if (error != 0)
1104 		goto out;
1105 	error = copyinstr(args->specialfile, mntfromname, MNAMELEN - 1, NULL);
1106 	if (error != 0)
1107 		goto out;
1108 	error = copyinstr(args->dir, mntonname, MNAMELEN - 1, NULL);
1109 	if (error != 0)
1110 		goto out;
1111 
1112 #ifdef DEBUG
1113 	if (ldebug(mount))
1114 		printf(ARGS(mount, "%s, %s, %s"),
1115 		    fstypename, mntfromname, mntonname);
1116 #endif
1117 
1118 	if (strcmp(fstypename, "ext2") == 0) {
1119 		strcpy(fstypename, "ext2fs");
1120 	} else if (strcmp(fstypename, "proc") == 0) {
1121 		strcpy(fstypename, "linprocfs");
1122 	} else if (strcmp(fstypename, "vfat") == 0) {
1123 		strcpy(fstypename, "msdosfs");
1124 	}
1125 
1126 	fsflags = 0;
1127 
1128 	if ((args->rwflag & 0xffff0000) == 0xc0ed0000) {
1129 		/*
1130 		 * Linux SYNC flag is not included; the closest equivalent
1131 		 * FreeBSD has is !ASYNC, which is our default.
1132 		 */
1133 		if (args->rwflag & LINUX_MS_RDONLY)
1134 			fsflags |= MNT_RDONLY;
1135 		if (args->rwflag & LINUX_MS_NOSUID)
1136 			fsflags |= MNT_NOSUID;
1137 		if (args->rwflag & LINUX_MS_NOEXEC)
1138 			fsflags |= MNT_NOEXEC;
1139 		if (args->rwflag & LINUX_MS_REMOUNT)
1140 			fsflags |= MNT_UPDATE;
1141 	}
1142 
1143 	error = kernel_vmount(fsflags,
1144 	    "fstype", fstypename,
1145 	    "fspath", mntonname,
1146 	    "from", mntfromname,
1147 	    NULL);
1148 out:
1149 	free(mntonname, M_TEMP);
1150 	free(mntfromname, M_TEMP);
1151 	return (error);
1152 }
1153 
1154 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1155 int
1156 linux_oldumount(struct thread *td, struct linux_oldumount_args *args)
1157 {
1158 	struct linux_umount_args args2;
1159 
1160 	args2.path = args->path;
1161 	args2.flags = 0;
1162 	return (linux_umount(td, &args2));
1163 }
1164 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1165 
1166 #ifdef LINUX_LEGACY_SYSCALLS
1167 int
1168 linux_umount(struct thread *td, struct linux_umount_args *args)
1169 {
1170 	struct unmount_args bsd;
1171 
1172 	bsd.path = args->path;
1173 	bsd.flags = args->flags;	/* XXX correct? */
1174 	return (sys_unmount(td, &bsd));
1175 }
1176 #endif
1177 
1178 /*
1179  * fcntl family of syscalls
1180  */
1181 
1182 struct l_flock {
1183 	l_short		l_type;
1184 	l_short		l_whence;
1185 	l_off_t		l_start;
1186 	l_off_t		l_len;
1187 	l_pid_t		l_pid;
1188 }
1189 #if defined(__amd64__) && defined(COMPAT_LINUX32)
1190 __packed
1191 #endif
1192 ;
1193 
1194 static void
1195 linux_to_bsd_flock(struct l_flock *linux_flock, struct flock *bsd_flock)
1196 {
1197 	switch (linux_flock->l_type) {
1198 	case LINUX_F_RDLCK:
1199 		bsd_flock->l_type = F_RDLCK;
1200 		break;
1201 	case LINUX_F_WRLCK:
1202 		bsd_flock->l_type = F_WRLCK;
1203 		break;
1204 	case LINUX_F_UNLCK:
1205 		bsd_flock->l_type = F_UNLCK;
1206 		break;
1207 	default:
1208 		bsd_flock->l_type = -1;
1209 		break;
1210 	}
1211 	bsd_flock->l_whence = linux_flock->l_whence;
1212 	bsd_flock->l_start = (off_t)linux_flock->l_start;
1213 	bsd_flock->l_len = (off_t)linux_flock->l_len;
1214 	bsd_flock->l_pid = (pid_t)linux_flock->l_pid;
1215 	bsd_flock->l_sysid = 0;
1216 }
1217 
1218 static void
1219 bsd_to_linux_flock(struct flock *bsd_flock, struct l_flock *linux_flock)
1220 {
1221 	switch (bsd_flock->l_type) {
1222 	case F_RDLCK:
1223 		linux_flock->l_type = LINUX_F_RDLCK;
1224 		break;
1225 	case F_WRLCK:
1226 		linux_flock->l_type = LINUX_F_WRLCK;
1227 		break;
1228 	case F_UNLCK:
1229 		linux_flock->l_type = LINUX_F_UNLCK;
1230 		break;
1231 	}
1232 	linux_flock->l_whence = bsd_flock->l_whence;
1233 	linux_flock->l_start = (l_off_t)bsd_flock->l_start;
1234 	linux_flock->l_len = (l_off_t)bsd_flock->l_len;
1235 	linux_flock->l_pid = (l_pid_t)bsd_flock->l_pid;
1236 }
1237 
1238 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1239 struct l_flock64 {
1240 	l_short		l_type;
1241 	l_short		l_whence;
1242 	l_loff_t	l_start;
1243 	l_loff_t	l_len;
1244 	l_pid_t		l_pid;
1245 }
1246 #if defined(__amd64__) && defined(COMPAT_LINUX32)
1247 __packed
1248 #endif
1249 ;
1250 
1251 static void
1252 linux_to_bsd_flock64(struct l_flock64 *linux_flock, struct flock *bsd_flock)
1253 {
1254 	switch (linux_flock->l_type) {
1255 	case LINUX_F_RDLCK:
1256 		bsd_flock->l_type = F_RDLCK;
1257 		break;
1258 	case LINUX_F_WRLCK:
1259 		bsd_flock->l_type = F_WRLCK;
1260 		break;
1261 	case LINUX_F_UNLCK:
1262 		bsd_flock->l_type = F_UNLCK;
1263 		break;
1264 	default:
1265 		bsd_flock->l_type = -1;
1266 		break;
1267 	}
1268 	bsd_flock->l_whence = linux_flock->l_whence;
1269 	bsd_flock->l_start = (off_t)linux_flock->l_start;
1270 	bsd_flock->l_len = (off_t)linux_flock->l_len;
1271 	bsd_flock->l_pid = (pid_t)linux_flock->l_pid;
1272 	bsd_flock->l_sysid = 0;
1273 }
1274 
1275 static void
1276 bsd_to_linux_flock64(struct flock *bsd_flock, struct l_flock64 *linux_flock)
1277 {
1278 	switch (bsd_flock->l_type) {
1279 	case F_RDLCK:
1280 		linux_flock->l_type = LINUX_F_RDLCK;
1281 		break;
1282 	case F_WRLCK:
1283 		linux_flock->l_type = LINUX_F_WRLCK;
1284 		break;
1285 	case F_UNLCK:
1286 		linux_flock->l_type = LINUX_F_UNLCK;
1287 		break;
1288 	}
1289 	linux_flock->l_whence = bsd_flock->l_whence;
1290 	linux_flock->l_start = (l_loff_t)bsd_flock->l_start;
1291 	linux_flock->l_len = (l_loff_t)bsd_flock->l_len;
1292 	linux_flock->l_pid = (l_pid_t)bsd_flock->l_pid;
1293 }
1294 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1295 
1296 static int
1297 fcntl_common(struct thread *td, struct linux_fcntl_args *args)
1298 {
1299 	struct l_flock linux_flock;
1300 	struct flock bsd_flock;
1301 	struct file *fp;
1302 	long arg;
1303 	int error, result;
1304 
1305 	switch (args->cmd) {
1306 	case LINUX_F_DUPFD:
1307 		return (kern_fcntl(td, args->fd, F_DUPFD, args->arg));
1308 
1309 	case LINUX_F_GETFD:
1310 		return (kern_fcntl(td, args->fd, F_GETFD, 0));
1311 
1312 	case LINUX_F_SETFD:
1313 		return (kern_fcntl(td, args->fd, F_SETFD, args->arg));
1314 
1315 	case LINUX_F_GETFL:
1316 		error = kern_fcntl(td, args->fd, F_GETFL, 0);
1317 		result = td->td_retval[0];
1318 		td->td_retval[0] = 0;
1319 		if (result & O_RDONLY)
1320 			td->td_retval[0] |= LINUX_O_RDONLY;
1321 		if (result & O_WRONLY)
1322 			td->td_retval[0] |= LINUX_O_WRONLY;
1323 		if (result & O_RDWR)
1324 			td->td_retval[0] |= LINUX_O_RDWR;
1325 		if (result & O_NDELAY)
1326 			td->td_retval[0] |= LINUX_O_NONBLOCK;
1327 		if (result & O_APPEND)
1328 			td->td_retval[0] |= LINUX_O_APPEND;
1329 		if (result & O_FSYNC)
1330 			td->td_retval[0] |= LINUX_O_SYNC;
1331 		if (result & O_ASYNC)
1332 			td->td_retval[0] |= LINUX_FASYNC;
1333 #ifdef LINUX_O_NOFOLLOW
1334 		if (result & O_NOFOLLOW)
1335 			td->td_retval[0] |= LINUX_O_NOFOLLOW;
1336 #endif
1337 #ifdef LINUX_O_DIRECT
1338 		if (result & O_DIRECT)
1339 			td->td_retval[0] |= LINUX_O_DIRECT;
1340 #endif
1341 		return (error);
1342 
1343 	case LINUX_F_SETFL:
1344 		arg = 0;
1345 		if (args->arg & LINUX_O_NDELAY)
1346 			arg |= O_NONBLOCK;
1347 		if (args->arg & LINUX_O_APPEND)
1348 			arg |= O_APPEND;
1349 		if (args->arg & LINUX_O_SYNC)
1350 			arg |= O_FSYNC;
1351 		if (args->arg & LINUX_FASYNC)
1352 			arg |= O_ASYNC;
1353 #ifdef LINUX_O_NOFOLLOW
1354 		if (args->arg & LINUX_O_NOFOLLOW)
1355 			arg |= O_NOFOLLOW;
1356 #endif
1357 #ifdef LINUX_O_DIRECT
1358 		if (args->arg & LINUX_O_DIRECT)
1359 			arg |= O_DIRECT;
1360 #endif
1361 		return (kern_fcntl(td, args->fd, F_SETFL, arg));
1362 
1363 	case LINUX_F_GETLK:
1364 		error = copyin((void *)args->arg, &linux_flock,
1365 		    sizeof(linux_flock));
1366 		if (error)
1367 			return (error);
1368 		linux_to_bsd_flock(&linux_flock, &bsd_flock);
1369 		error = kern_fcntl(td, args->fd, F_GETLK, (intptr_t)&bsd_flock);
1370 		if (error)
1371 			return (error);
1372 		bsd_to_linux_flock(&bsd_flock, &linux_flock);
1373 		return (copyout(&linux_flock, (void *)args->arg,
1374 		    sizeof(linux_flock)));
1375 
1376 	case LINUX_F_SETLK:
1377 		error = copyin((void *)args->arg, &linux_flock,
1378 		    sizeof(linux_flock));
1379 		if (error)
1380 			return (error);
1381 		linux_to_bsd_flock(&linux_flock, &bsd_flock);
1382 		return (kern_fcntl(td, args->fd, F_SETLK,
1383 		    (intptr_t)&bsd_flock));
1384 
1385 	case LINUX_F_SETLKW:
1386 		error = copyin((void *)args->arg, &linux_flock,
1387 		    sizeof(linux_flock));
1388 		if (error)
1389 			return (error);
1390 		linux_to_bsd_flock(&linux_flock, &bsd_flock);
1391 		return (kern_fcntl(td, args->fd, F_SETLKW,
1392 		     (intptr_t)&bsd_flock));
1393 
1394 	case LINUX_F_GETOWN:
1395 		return (kern_fcntl(td, args->fd, F_GETOWN, 0));
1396 
1397 	case LINUX_F_SETOWN:
1398 		/*
1399 		 * XXX some Linux applications depend on F_SETOWN having no
1400 		 * significant effect for pipes (SIGIO is not delivered for
1401 		 * pipes under Linux-2.2.35 at least).
1402 		 */
1403 		error = fget(td, args->fd,
1404 		    &cap_fcntl_rights, &fp);
1405 		if (error)
1406 			return (error);
1407 		if (fp->f_type == DTYPE_PIPE) {
1408 			fdrop(fp, td);
1409 			return (EINVAL);
1410 		}
1411 		fdrop(fp, td);
1412 
1413 		return (kern_fcntl(td, args->fd, F_SETOWN, args->arg));
1414 
1415 	case LINUX_F_DUPFD_CLOEXEC:
1416 		return (kern_fcntl(td, args->fd, F_DUPFD_CLOEXEC, args->arg));
1417 	}
1418 
1419 	return (EINVAL);
1420 }
1421 
1422 int
1423 linux_fcntl(struct thread *td, struct linux_fcntl_args *args)
1424 {
1425 
1426 #ifdef DEBUG
1427 	if (ldebug(fcntl))
1428 		printf(ARGS(fcntl, "%d, %08x, *"), args->fd, args->cmd);
1429 #endif
1430 
1431 	return (fcntl_common(td, args));
1432 }
1433 
1434 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1435 int
1436 linux_fcntl64(struct thread *td, struct linux_fcntl64_args *args)
1437 {
1438 	struct l_flock64 linux_flock;
1439 	struct flock bsd_flock;
1440 	struct linux_fcntl_args fcntl_args;
1441 	int error;
1442 
1443 #ifdef DEBUG
1444 	if (ldebug(fcntl64))
1445 		printf(ARGS(fcntl64, "%d, %08x, *"), args->fd, args->cmd);
1446 #endif
1447 
1448 	switch (args->cmd) {
1449 	case LINUX_F_GETLK64:
1450 		error = copyin((void *)args->arg, &linux_flock,
1451 		    sizeof(linux_flock));
1452 		if (error)
1453 			return (error);
1454 		linux_to_bsd_flock64(&linux_flock, &bsd_flock);
1455 		error = kern_fcntl(td, args->fd, F_GETLK, (intptr_t)&bsd_flock);
1456 		if (error)
1457 			return (error);
1458 		bsd_to_linux_flock64(&bsd_flock, &linux_flock);
1459 		return (copyout(&linux_flock, (void *)args->arg,
1460 			    sizeof(linux_flock)));
1461 
1462 	case LINUX_F_SETLK64:
1463 		error = copyin((void *)args->arg, &linux_flock,
1464 		    sizeof(linux_flock));
1465 		if (error)
1466 			return (error);
1467 		linux_to_bsd_flock64(&linux_flock, &bsd_flock);
1468 		return (kern_fcntl(td, args->fd, F_SETLK,
1469 		    (intptr_t)&bsd_flock));
1470 
1471 	case LINUX_F_SETLKW64:
1472 		error = copyin((void *)args->arg, &linux_flock,
1473 		    sizeof(linux_flock));
1474 		if (error)
1475 			return (error);
1476 		linux_to_bsd_flock64(&linux_flock, &bsd_flock);
1477 		return (kern_fcntl(td, args->fd, F_SETLKW,
1478 		    (intptr_t)&bsd_flock));
1479 	}
1480 
1481 	fcntl_args.fd = args->fd;
1482 	fcntl_args.cmd = args->cmd;
1483 	fcntl_args.arg = args->arg;
1484 	return (fcntl_common(td, &fcntl_args));
1485 }
1486 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1487 
1488 #ifdef LINUX_LEGACY_SYSCALLS
1489 int
1490 linux_chown(struct thread *td, struct linux_chown_args *args)
1491 {
1492 	char *path;
1493 	int error;
1494 
1495 	LCONVPATHEXIST(td, args->path, &path);
1496 
1497 #ifdef DEBUG
1498 	if (ldebug(chown))
1499 		printf(ARGS(chown, "%s, %d, %d"), path, args->uid, args->gid);
1500 #endif
1501 	error = kern_fchownat(td, AT_FDCWD, path, UIO_SYSSPACE, args->uid,
1502 	    args->gid, 0);
1503 	LFREEPATH(path);
1504 	return (error);
1505 }
1506 #endif
1507 
1508 int
1509 linux_fchownat(struct thread *td, struct linux_fchownat_args *args)
1510 {
1511 	char *path;
1512 	int error, dfd, flag;
1513 
1514 	if (args->flag & ~LINUX_AT_SYMLINK_NOFOLLOW)
1515 		return (EINVAL);
1516 
1517 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD :  args->dfd;
1518 	LCONVPATHEXIST_AT(td, args->filename, &path, dfd);
1519 
1520 #ifdef DEBUG
1521 	if (ldebug(fchownat))
1522 		printf(ARGS(fchownat, "%s, %d, %d"), path, args->uid, args->gid);
1523 #endif
1524 
1525 	flag = (args->flag & LINUX_AT_SYMLINK_NOFOLLOW) == 0 ? 0 :
1526 	    AT_SYMLINK_NOFOLLOW;
1527 	error = kern_fchownat(td, dfd, path, UIO_SYSSPACE, args->uid, args->gid,
1528 	    flag);
1529 	LFREEPATH(path);
1530 	return (error);
1531 }
1532 
1533 #ifdef LINUX_LEGACY_SYSCALLS
1534 int
1535 linux_lchown(struct thread *td, struct linux_lchown_args *args)
1536 {
1537 	char *path;
1538 	int error;
1539 
1540 	LCONVPATHEXIST(td, args->path, &path);
1541 
1542 #ifdef DEBUG
1543 	if (ldebug(lchown))
1544 		printf(ARGS(lchown, "%s, %d, %d"), path, args->uid, args->gid);
1545 #endif
1546 	error = kern_fchownat(td, AT_FDCWD, path, UIO_SYSSPACE, args->uid,
1547 	    args->gid, AT_SYMLINK_NOFOLLOW);
1548 	LFREEPATH(path);
1549 	return (error);
1550 }
1551 #endif
1552 
1553 static int
1554 convert_fadvice(int advice)
1555 {
1556 	switch (advice) {
1557 	case LINUX_POSIX_FADV_NORMAL:
1558 		return (POSIX_FADV_NORMAL);
1559 	case LINUX_POSIX_FADV_RANDOM:
1560 		return (POSIX_FADV_RANDOM);
1561 	case LINUX_POSIX_FADV_SEQUENTIAL:
1562 		return (POSIX_FADV_SEQUENTIAL);
1563 	case LINUX_POSIX_FADV_WILLNEED:
1564 		return (POSIX_FADV_WILLNEED);
1565 	case LINUX_POSIX_FADV_DONTNEED:
1566 		return (POSIX_FADV_DONTNEED);
1567 	case LINUX_POSIX_FADV_NOREUSE:
1568 		return (POSIX_FADV_NOREUSE);
1569 	default:
1570 		return (-1);
1571 	}
1572 }
1573 
1574 int
1575 linux_fadvise64(struct thread *td, struct linux_fadvise64_args *args)
1576 {
1577 	int advice;
1578 
1579 	advice = convert_fadvice(args->advice);
1580 	if (advice == -1)
1581 		return (EINVAL);
1582 	return (kern_posix_fadvise(td, args->fd, args->offset, args->len,
1583 	    advice));
1584 }
1585 
1586 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1587 int
1588 linux_fadvise64_64(struct thread *td, struct linux_fadvise64_64_args *args)
1589 {
1590 	int advice;
1591 
1592 	advice = convert_fadvice(args->advice);
1593 	if (advice == -1)
1594 		return (EINVAL);
1595 	return (kern_posix_fadvise(td, args->fd, args->offset, args->len,
1596 	    advice));
1597 }
1598 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1599 
1600 #ifdef LINUX_LEGACY_SYSCALLS
1601 int
1602 linux_pipe(struct thread *td, struct linux_pipe_args *args)
1603 {
1604 	int fildes[2];
1605 	int error;
1606 
1607 #ifdef DEBUG
1608 	if (ldebug(pipe))
1609 		printf(ARGS(pipe, "*"));
1610 #endif
1611 
1612 	error = kern_pipe(td, fildes, 0, NULL, NULL);
1613 	if (error != 0)
1614 		return (error);
1615 
1616 	error = copyout(fildes, args->pipefds, sizeof(fildes));
1617 	if (error != 0) {
1618 		(void)kern_close(td, fildes[0]);
1619 		(void)kern_close(td, fildes[1]);
1620 	}
1621 
1622 	return (error);
1623 }
1624 #endif
1625 
1626 int
1627 linux_pipe2(struct thread *td, struct linux_pipe2_args *args)
1628 {
1629 	int fildes[2];
1630 	int error, flags;
1631 
1632 #ifdef DEBUG
1633 	if (ldebug(pipe2))
1634 		printf(ARGS(pipe2, "*, %d"), args->flags);
1635 #endif
1636 
1637 	if ((args->flags & ~(LINUX_O_NONBLOCK | LINUX_O_CLOEXEC)) != 0)
1638 		return (EINVAL);
1639 
1640 	flags = 0;
1641 	if ((args->flags & LINUX_O_NONBLOCK) != 0)
1642 		flags |= O_NONBLOCK;
1643 	if ((args->flags & LINUX_O_CLOEXEC) != 0)
1644 		flags |= O_CLOEXEC;
1645 	error = kern_pipe(td, fildes, flags, NULL, NULL);
1646 	if (error != 0)
1647 		return (error);
1648 
1649 	error = copyout(fildes, args->pipefds, sizeof(fildes));
1650 	if (error != 0) {
1651 		(void)kern_close(td, fildes[0]);
1652 		(void)kern_close(td, fildes[1]);
1653 	}
1654 
1655 	return (error);
1656 }
1657 
1658 int
1659 linux_dup3(struct thread *td, struct linux_dup3_args *args)
1660 {
1661 	int cmd;
1662 	intptr_t newfd;
1663 
1664 	if (args->oldfd == args->newfd)
1665 		return (EINVAL);
1666 	if ((args->flags & ~LINUX_O_CLOEXEC) != 0)
1667 		return (EINVAL);
1668 	if (args->flags & LINUX_O_CLOEXEC)
1669 		cmd = F_DUP2FD_CLOEXEC;
1670 	else
1671 		cmd = F_DUP2FD;
1672 
1673 	newfd = args->newfd;
1674 	return (kern_fcntl(td, args->oldfd, cmd, newfd));
1675 }
1676 
1677 int
1678 linux_fallocate(struct thread *td, struct linux_fallocate_args *args)
1679 {
1680 
1681 	/*
1682 	 * We emulate only posix_fallocate system call for which
1683 	 * mode should be 0.
1684 	 */
1685 	if (args->mode != 0)
1686 		return (ENOSYS);
1687 
1688 	return (kern_posix_fallocate(td, args->fd, args->offset,
1689 	    args->len));
1690 }
1691