xref: /freebsd/sys/compat/linux/linux_file.c (revision 93a065e7496dfbfbd0a5b0208ef763f37ea975c7)
1 /*-
2  * Copyright (c) 1994-1995 Søren Schmidt
3  * All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer
10  *    in this position and unchanged.
11  * 2. Redistributions in binary form must reproduce the above copyright
12  *    notice, this list of conditions and the following disclaimer in the
13  *    documentation and/or other materials provided with the distribution.
14  * 3. The name of the author may not be used to endorse or promote products
15  *    derived from this software without specific prior written permission
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
18  * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
19  * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
20  * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
21  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
22  * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
23  * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
24  * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
25  * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
26  * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 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 #include <security/mac/mac_framework.h>
57 
58 #ifdef COMPAT_LINUX32
59 #include <machine/../linux32/linux.h>
60 #include <machine/../linux32/linux32_proto.h>
61 #else
62 #include <machine/../linux/linux.h>
63 #include <machine/../linux/linux_proto.h>
64 #endif
65 #include <compat/linux/linux_misc.h>
66 #include <compat/linux/linux_util.h>
67 #include <compat/linux/linux_file.h>
68 
69 int
70 linux_creat(struct thread *td, struct linux_creat_args *args)
71 {
72 	char *path;
73 	int error;
74 
75 	LCONVPATHEXIST(td, args->path, &path);
76 #ifdef DEBUG
77 	if (ldebug(creat))
78 		printf(ARGS(creat, "%s, %d"), path, args->mode);
79 #endif
80 	error = kern_openat(td, AT_FDCWD, path, UIO_SYSSPACE,
81 	    O_WRONLY | O_CREAT | O_TRUNC, args->mode);
82 	LFREEPATH(path);
83 	return (error);
84 }
85 
86 
87 static int
88 linux_common_open(struct thread *td, int dirfd, char *path, int l_flags, int mode)
89 {
90 	cap_rights_t rights;
91 	struct proc *p = td->td_proc;
92 	struct file *fp;
93 	int fd;
94 	int bsd_flags, error;
95 
96 	bsd_flags = 0;
97 	switch (l_flags & LINUX_O_ACCMODE) {
98 	case LINUX_O_WRONLY:
99 		bsd_flags |= O_WRONLY;
100 		break;
101 	case LINUX_O_RDWR:
102 		bsd_flags |= O_RDWR;
103 		break;
104 	default:
105 		bsd_flags |= O_RDONLY;
106 	}
107 	if (l_flags & LINUX_O_NDELAY)
108 		bsd_flags |= O_NONBLOCK;
109 	if (l_flags & LINUX_O_APPEND)
110 		bsd_flags |= O_APPEND;
111 	if (l_flags & LINUX_O_SYNC)
112 		bsd_flags |= O_FSYNC;
113 	if (l_flags & LINUX_O_NONBLOCK)
114 		bsd_flags |= O_NONBLOCK;
115 	if (l_flags & LINUX_FASYNC)
116 		bsd_flags |= O_ASYNC;
117 	if (l_flags & LINUX_O_CREAT)
118 		bsd_flags |= O_CREAT;
119 	if (l_flags & LINUX_O_TRUNC)
120 		bsd_flags |= O_TRUNC;
121 	if (l_flags & LINUX_O_EXCL)
122 		bsd_flags |= O_EXCL;
123 	if (l_flags & LINUX_O_NOCTTY)
124 		bsd_flags |= O_NOCTTY;
125 	if (l_flags & LINUX_O_DIRECT)
126 		bsd_flags |= O_DIRECT;
127 	if (l_flags & LINUX_O_NOFOLLOW)
128 		bsd_flags |= O_NOFOLLOW;
129 	if (l_flags & LINUX_O_DIRECTORY)
130 		bsd_flags |= O_DIRECTORY;
131 	/* XXX LINUX_O_NOATIME: unable to be easily implemented. */
132 
133 	error = kern_openat(td, dirfd, path, UIO_SYSSPACE, bsd_flags, mode);
134 	if (error != 0)
135 		goto done;
136 	if (bsd_flags & O_NOCTTY)
137 		goto done;
138 
139 	/*
140 	 * XXX In between kern_open() and fget(), another process
141 	 * having the same filedesc could use that fd without
142 	 * checking below.
143 	*/
144 	fd = td->td_retval[0];
145 	if (fget(td, fd, cap_rights_init(&rights, CAP_IOCTL), &fp) == 0) {
146 		if (fp->f_type != DTYPE_VNODE) {
147 			fdrop(fp, td);
148 			goto done;
149 		}
150 		sx_slock(&proctree_lock);
151 		PROC_LOCK(p);
152 		if (SESS_LEADER(p) && !(p->p_flag & P_CONTROLT)) {
153 			PROC_UNLOCK(p);
154 			sx_sunlock(&proctree_lock);
155 			/* XXXPJD: Verify if TIOCSCTTY is allowed. */
156 			(void) fo_ioctl(fp, TIOCSCTTY, (caddr_t) 0,
157 			    td->td_ucred, td);
158 		} else {
159 			PROC_UNLOCK(p);
160 			sx_sunlock(&proctree_lock);
161 		}
162 		fdrop(fp, td);
163 	}
164 
165 done:
166 #ifdef DEBUG
167 	if (ldebug(open))
168 		printf(LMSG("open returns error %d"), error);
169 #endif
170 	LFREEPATH(path);
171 	return (error);
172 }
173 
174 int
175 linux_openat(struct thread *td, struct linux_openat_args *args)
176 {
177 	char *path;
178 	int dfd;
179 
180 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
181 	if (args->flags & LINUX_O_CREAT)
182 		LCONVPATH_AT(td, args->filename, &path, 1, dfd);
183 	else
184 		LCONVPATH_AT(td, args->filename, &path, 0, dfd);
185 #ifdef DEBUG
186 	if (ldebug(openat))
187 		printf(ARGS(openat, "%i, %s, 0x%x, 0x%x"), args->dfd,
188 		    path, args->flags, args->mode);
189 #endif
190 	return (linux_common_open(td, dfd, path, args->flags, args->mode));
191 }
192 
193 int
194 linux_open(struct thread *td, struct linux_open_args *args)
195 {
196 	char *path;
197 
198 	if (args->flags & LINUX_O_CREAT)
199 		LCONVPATHCREAT(td, args->path, &path);
200 	else
201 		LCONVPATHEXIST(td, args->path, &path);
202 #ifdef DEBUG
203 	if (ldebug(open))
204 		printf(ARGS(open, "%s, 0x%x, 0x%x"),
205 		    path, args->flags, args->mode);
206 #endif
207 	return (linux_common_open(td, AT_FDCWD, path, args->flags, args->mode));
208 }
209 
210 int
211 linux_lseek(struct thread *td, struct linux_lseek_args *args)
212 {
213 
214 #ifdef DEBUG
215 	if (ldebug(lseek))
216 		printf(ARGS(lseek, "%d, %ld, %d"),
217 		    args->fdes, (long)args->off, args->whence);
218 #endif
219 	return (kern_lseek(td, args->fdes, args->off, args->whence));
220 }
221 
222 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
223 int
224 linux_llseek(struct thread *td, struct linux_llseek_args *args)
225 {
226 	int error;
227 	off_t off;
228 
229 #ifdef DEBUG
230 	if (ldebug(llseek))
231 		printf(ARGS(llseek, "%d, %d:%d, %d"),
232 		    args->fd, args->ohigh, args->olow, args->whence);
233 #endif
234 	off = (args->olow) | (((off_t) args->ohigh) << 32);
235 
236 	error = kern_lseek(td, args->fd, off, args->whence);
237 	if (error != 0)
238 		return (error);
239 
240 	error = copyout(td->td_retval, args->res, sizeof(off_t));
241 	if (error != 0)
242 		return (error);
243 
244 	td->td_retval[0] = 0;
245 	return (0);
246 }
247 
248 int
249 linux_readdir(struct thread *td, struct linux_readdir_args *args)
250 {
251 	struct linux_getdents_args lda;
252 
253 	lda.fd = args->fd;
254 	lda.dent = args->dent;
255 	lda.count = 1;
256 	return (linux_getdents(td, &lda));
257 }
258 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
259 
260 /*
261  * Note that linux_getdents(2) and linux_getdents64(2) have the same
262  * arguments. They only differ in the definition of struct dirent they
263  * operate on. We use this to common the code, with the exception of
264  * accessing struct dirent. Note that linux_readdir(2) is implemented
265  * by means of linux_getdents(2). In this case we never operate on
266  * struct dirent64 and thus don't need to handle it...
267  */
268 
269 struct l_dirent {
270 	l_ulong		d_ino;
271 	l_off_t		d_off;
272 	l_ushort	d_reclen;
273 	char		d_name[LINUX_NAME_MAX + 1];
274 };
275 
276 struct l_dirent64 {
277 	uint64_t	d_ino;
278 	int64_t		d_off;
279 	l_ushort	d_reclen;
280 	u_char		d_type;
281 	char		d_name[LINUX_NAME_MAX + 1];
282 };
283 
284 /*
285  * Linux uses the last byte in the dirent buffer to store d_type,
286  * at least glibc-2.7 requires it. That is why l_dirent is padded with 2 bytes.
287  */
288 #define LINUX_RECLEN(namlen)						\
289     roundup(offsetof(struct l_dirent, d_name) + (namlen) + 2, sizeof(l_ulong))
290 
291 #define LINUX_RECLEN64(namlen)						\
292     roundup(offsetof(struct l_dirent64, d_name) + (namlen) + 1,		\
293     sizeof(uint64_t))
294 
295 #define LINUX_MAXRECLEN		max(LINUX_RECLEN(LINUX_NAME_MAX),	\
296 				    LINUX_RECLEN64(LINUX_NAME_MAX))
297 #define	LINUX_DIRBLKSIZ		512
298 
299 static int
300 getdents_common(struct thread *td, struct linux_getdents64_args *args,
301     int is64bit)
302 {
303 	struct dirent *bdp;
304 	struct vnode *vp;
305 	caddr_t inp, buf;		/* BSD-format */
306 	int len, reclen;		/* BSD-format */
307 	caddr_t outp;			/* Linux-format */
308 	int resid, linuxreclen=0;	/* Linux-format */
309 	caddr_t lbuf;			/* Linux-format */
310 	cap_rights_t rights;
311 	struct file *fp;
312 	struct uio auio;
313 	struct iovec aiov;
314 	off_t off;
315 	struct l_dirent *linux_dirent;
316 	struct l_dirent64 *linux_dirent64;
317 	int buflen, error, eofflag, nbytes, justone;
318 	u_long *cookies = NULL, *cookiep;
319 	int ncookies;
320 
321 	nbytes = args->count;
322 	if (nbytes == 1) {
323 		/* readdir(2) case. Always struct dirent. */
324 		if (is64bit)
325 			return (EINVAL);
326 		nbytes = sizeof(*linux_dirent);
327 		justone = 1;
328 	} else
329 		justone = 0;
330 
331 	error = getvnode(td, args->fd, cap_rights_init(&rights, CAP_READ), &fp);
332 	if (error != 0)
333 		return (error);
334 
335 	if ((fp->f_flag & FREAD) == 0) {
336 		fdrop(fp, td);
337 		return (EBADF);
338 	}
339 
340 	off = foffset_lock(fp, 0);
341 	vp = fp->f_vnode;
342 	if (vp->v_type != VDIR) {
343 		foffset_unlock(fp, off, 0);
344 		fdrop(fp, td);
345 		return (EINVAL);
346 	}
347 
348 
349 	buflen = max(LINUX_DIRBLKSIZ, nbytes);
350 	buflen = min(buflen, MAXBSIZE);
351 	buf = malloc(buflen, M_LINUX, M_WAITOK);
352 	lbuf = malloc(LINUX_MAXRECLEN, M_LINUX, M_WAITOK | M_ZERO);
353 	vn_lock(vp, LK_SHARED | LK_RETRY);
354 
355 	aiov.iov_base = buf;
356 	aiov.iov_len = buflen;
357 	auio.uio_iov = &aiov;
358 	auio.uio_iovcnt = 1;
359 	auio.uio_rw = UIO_READ;
360 	auio.uio_segflg = UIO_SYSSPACE;
361 	auio.uio_td = td;
362 	auio.uio_resid = buflen;
363 	auio.uio_offset = off;
364 
365 #ifdef MAC
366 	/*
367 	 * Do directory search MAC check using non-cached credentials.
368 	 */
369 	if ((error = mac_vnode_check_readdir(td->td_ucred, vp)))
370 		goto out;
371 #endif /* MAC */
372 	if ((error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, &ncookies,
373 		 &cookies)))
374 		goto out;
375 
376 	inp = buf;
377 	outp = (caddr_t)args->dirent;
378 	resid = nbytes;
379 	if ((len = buflen - auio.uio_resid) <= 0)
380 		goto eof;
381 
382 	cookiep = cookies;
383 
384 	if (cookies) {
385 		/*
386 		 * When using cookies, the vfs has the option of reading from
387 		 * a different offset than that supplied (UFS truncates the
388 		 * offset to a block boundary to make sure that it never reads
389 		 * partway through a directory entry, even if the directory
390 		 * has been compacted).
391 		 */
392 		while (len > 0 && ncookies > 0 && *cookiep <= off) {
393 			bdp = (struct dirent *) inp;
394 			len -= bdp->d_reclen;
395 			inp += bdp->d_reclen;
396 			cookiep++;
397 			ncookies--;
398 		}
399 	}
400 
401 	while (len > 0) {
402 		if (cookiep && ncookies == 0)
403 			break;
404 		bdp = (struct dirent *) inp;
405 		reclen = bdp->d_reclen;
406 		if (reclen & 3) {
407 			error = EFAULT;
408 			goto out;
409 		}
410 
411 		if (bdp->d_fileno == 0) {
412 			inp += reclen;
413 			if (cookiep) {
414 				off = *cookiep++;
415 				ncookies--;
416 			} else
417 				off += reclen;
418 
419 			len -= reclen;
420 			continue;
421 		}
422 
423 		linuxreclen = (is64bit)
424 		    ? LINUX_RECLEN64(bdp->d_namlen)
425 		    : LINUX_RECLEN(bdp->d_namlen);
426 
427 		if (reclen > len || resid < linuxreclen) {
428 			outp++;
429 			break;
430 		}
431 
432 		if (justone) {
433 			/* readdir(2) case. */
434 			linux_dirent = (struct l_dirent*)lbuf;
435 			linux_dirent->d_ino = bdp->d_fileno;
436 			linux_dirent->d_off = (l_off_t)linuxreclen;
437 			linux_dirent->d_reclen = (l_ushort)bdp->d_namlen;
438 			strlcpy(linux_dirent->d_name, bdp->d_name,
439 			    linuxreclen - offsetof(struct l_dirent, d_name));
440 			error = copyout(linux_dirent, outp, linuxreclen);
441 		}
442 		if (is64bit) {
443 			linux_dirent64 = (struct l_dirent64*)lbuf;
444 			linux_dirent64->d_ino = bdp->d_fileno;
445 			linux_dirent64->d_off = (cookiep)
446 			    ? (l_off_t)*cookiep
447 			    : (l_off_t)(off + reclen);
448 			linux_dirent64->d_reclen = (l_ushort)linuxreclen;
449 			linux_dirent64->d_type = bdp->d_type;
450 			strlcpy(linux_dirent64->d_name, bdp->d_name,
451 			    linuxreclen - offsetof(struct l_dirent64, d_name));
452 			error = copyout(linux_dirent64, outp, linuxreclen);
453 		} else if (!justone) {
454 			linux_dirent = (struct l_dirent*)lbuf;
455 			linux_dirent->d_ino = bdp->d_fileno;
456 			linux_dirent->d_off = (cookiep)
457 			    ? (l_off_t)*cookiep
458 			    : (l_off_t)(off + reclen);
459 			linux_dirent->d_reclen = (l_ushort)linuxreclen;
460 			/*
461 			 * Copy d_type to last byte of l_dirent buffer
462 			 */
463 			lbuf[linuxreclen-1] = bdp->d_type;
464 			strlcpy(linux_dirent->d_name, bdp->d_name,
465 			    linuxreclen - offsetof(struct l_dirent, d_name)-1);
466 			error = copyout(linux_dirent, outp, linuxreclen);
467 		}
468 
469 		if (error)
470 			goto out;
471 
472 		inp += reclen;
473 		if (cookiep) {
474 			off = *cookiep++;
475 			ncookies--;
476 		} else
477 			off += reclen;
478 
479 		outp += linuxreclen;
480 		resid -= linuxreclen;
481 		len -= reclen;
482 		if (justone)
483 			break;
484 	}
485 
486 	if (outp == (caddr_t)args->dirent) {
487 		nbytes = resid;
488 		goto eof;
489 	}
490 
491 	if (justone)
492 		nbytes = resid + linuxreclen;
493 
494 eof:
495 	td->td_retval[0] = nbytes - resid;
496 
497 out:
498 	free(cookies, M_TEMP);
499 
500 	VOP_UNLOCK(vp, 0);
501 	foffset_unlock(fp, off, 0);
502 	fdrop(fp, td);
503 	free(buf, M_LINUX);
504 	free(lbuf, M_LINUX);
505 	return (error);
506 }
507 
508 int
509 linux_getdents(struct thread *td, struct linux_getdents_args *args)
510 {
511 
512 #ifdef DEBUG
513 	if (ldebug(getdents))
514 		printf(ARGS(getdents, "%d, *, %d"), args->fd, args->count);
515 #endif
516 
517 	return (getdents_common(td, (struct linux_getdents64_args*)args, 0));
518 }
519 
520 int
521 linux_getdents64(struct thread *td, struct linux_getdents64_args *args)
522 {
523 
524 #ifdef DEBUG
525 	if (ldebug(getdents64))
526 		printf(ARGS(getdents64, "%d, *, %d"), args->fd, args->count);
527 #endif
528 
529 	return (getdents_common(td, args, 1));
530 }
531 
532 /*
533  * These exist mainly for hooks for doing /compat/linux translation.
534  */
535 
536 int
537 linux_access(struct thread *td, struct linux_access_args *args)
538 {
539 	char *path;
540 	int error;
541 
542 	/* linux convention */
543 	if (args->amode & ~(F_OK | X_OK | W_OK | R_OK))
544 		return (EINVAL);
545 
546 	LCONVPATHEXIST(td, args->path, &path);
547 
548 #ifdef DEBUG
549 	if (ldebug(access))
550 		printf(ARGS(access, "%s, %d"), path, args->amode);
551 #endif
552 	error = kern_accessat(td, AT_FDCWD, path, UIO_SYSSPACE, 0,
553 	    args->amode);
554 	LFREEPATH(path);
555 
556 	return (error);
557 }
558 
559 int
560 linux_faccessat(struct thread *td, struct linux_faccessat_args *args)
561 {
562 	char *path;
563 	int error, dfd;
564 
565 	/* linux convention */
566 	if (args->amode & ~(F_OK | X_OK | W_OK | R_OK))
567 		return (EINVAL);
568 
569 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
570 	LCONVPATHEXIST_AT(td, args->filename, &path, dfd);
571 
572 #ifdef DEBUG
573 	if (ldebug(access))
574 		printf(ARGS(access, "%s, %d"), path, args->amode);
575 #endif
576 
577 	error = kern_accessat(td, dfd, path, UIO_SYSSPACE, 0, args->amode);
578 	LFREEPATH(path);
579 
580 	return (error);
581 }
582 
583 int
584 linux_unlink(struct thread *td, struct linux_unlink_args *args)
585 {
586 	char *path;
587 	int error;
588 	struct stat st;
589 
590 	LCONVPATHEXIST(td, args->path, &path);
591 
592 #ifdef DEBUG
593 	if (ldebug(unlink))
594 		printf(ARGS(unlink, "%s"), path);
595 #endif
596 
597 	error = kern_unlinkat(td, AT_FDCWD, path, UIO_SYSSPACE, 0);
598 	if (error == EPERM) {
599 		/* Introduce POSIX noncompliant behaviour of Linux */
600 		if (kern_statat(td, 0, AT_FDCWD, path, UIO_SYSSPACE, &st,
601 		    NULL) == 0) {
602 			if (S_ISDIR(st.st_mode))
603 				error = EISDIR;
604 		}
605 	}
606 	LFREEPATH(path);
607 	return (error);
608 }
609 
610 int
611 linux_unlinkat(struct thread *td, struct linux_unlinkat_args *args)
612 {
613 	char *path;
614 	int error, dfd;
615 	struct stat st;
616 
617 	if (args->flag & ~LINUX_AT_REMOVEDIR)
618 		return (EINVAL);
619 
620 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
621 	LCONVPATHEXIST_AT(td, args->pathname, &path, dfd);
622 
623 #ifdef DEBUG
624 	if (ldebug(unlinkat))
625 		printf(ARGS(unlinkat, "%s"), path);
626 #endif
627 
628 	if (args->flag & LINUX_AT_REMOVEDIR)
629 		error = kern_rmdirat(td, dfd, path, UIO_SYSSPACE);
630 	else
631 		error = kern_unlinkat(td, dfd, path, UIO_SYSSPACE, 0);
632 	if (error == EPERM && !(args->flag & LINUX_AT_REMOVEDIR)) {
633 		/* Introduce POSIX noncompliant behaviour of Linux */
634 		if (kern_statat(td, AT_SYMLINK_NOFOLLOW, dfd, path,
635 		    UIO_SYSSPACE, &st, NULL) == 0 && S_ISDIR(st.st_mode))
636 			error = EISDIR;
637 	}
638 	LFREEPATH(path);
639 	return (error);
640 }
641 int
642 linux_chdir(struct thread *td, struct linux_chdir_args *args)
643 {
644 	char *path;
645 	int error;
646 
647 	LCONVPATHEXIST(td, args->path, &path);
648 
649 #ifdef DEBUG
650 	if (ldebug(chdir))
651 		printf(ARGS(chdir, "%s"), path);
652 #endif
653 	error = kern_chdir(td, path, UIO_SYSSPACE);
654 	LFREEPATH(path);
655 	return (error);
656 }
657 
658 int
659 linux_chmod(struct thread *td, struct linux_chmod_args *args)
660 {
661 	char *path;
662 	int error;
663 
664 	LCONVPATHEXIST(td, args->path, &path);
665 
666 #ifdef DEBUG
667 	if (ldebug(chmod))
668 		printf(ARGS(chmod, "%s, %d"), path, args->mode);
669 #endif
670 	error = kern_fchmodat(td, AT_FDCWD, path, UIO_SYSSPACE,
671 	    args->mode, 0);
672 	LFREEPATH(path);
673 	return (error);
674 }
675 
676 int
677 linux_fchmodat(struct thread *td, struct linux_fchmodat_args *args)
678 {
679 	char *path;
680 	int error, dfd;
681 
682 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
683 	LCONVPATHEXIST_AT(td, args->filename, &path, dfd);
684 
685 #ifdef DEBUG
686 	if (ldebug(fchmodat))
687 		printf(ARGS(fchmodat, "%s, %d"), path, args->mode);
688 #endif
689 
690 	error = kern_fchmodat(td, dfd, path, UIO_SYSSPACE, args->mode, 0);
691 	LFREEPATH(path);
692 	return (error);
693 }
694 
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 
712 int
713 linux_mkdirat(struct thread *td, struct linux_mkdirat_args *args)
714 {
715 	char *path;
716 	int error, dfd;
717 
718 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
719 	LCONVPATHCREAT_AT(td, args->pathname, &path, dfd);
720 
721 #ifdef DEBUG
722 	if (ldebug(mkdirat))
723 		printf(ARGS(mkdirat, "%s, %d"), path, args->mode);
724 #endif
725 	error = kern_mkdirat(td, dfd, path, UIO_SYSSPACE, args->mode);
726 	LFREEPATH(path);
727 	return (error);
728 }
729 
730 int
731 linux_rmdir(struct thread *td, struct linux_rmdir_args *args)
732 {
733 	char *path;
734 	int error;
735 
736 	LCONVPATHEXIST(td, args->path, &path);
737 
738 #ifdef DEBUG
739 	if (ldebug(rmdir))
740 		printf(ARGS(rmdir, "%s"), path);
741 #endif
742 	error = kern_rmdirat(td, AT_FDCWD, path, UIO_SYSSPACE);
743 	LFREEPATH(path);
744 	return (error);
745 }
746 
747 int
748 linux_rename(struct thread *td, struct linux_rename_args *args)
749 {
750 	char *from, *to;
751 	int error;
752 
753 	LCONVPATHEXIST(td, args->from, &from);
754 	/* Expand LCONVPATHCREATE so that `from' can be freed on errors */
755 	error = linux_emul_convpath(td, args->to, UIO_USERSPACE, &to, 1, AT_FDCWD);
756 	if (to == NULL) {
757 		LFREEPATH(from);
758 		return (error);
759 	}
760 
761 #ifdef DEBUG
762 	if (ldebug(rename))
763 		printf(ARGS(rename, "%s, %s"), from, to);
764 #endif
765 	error = kern_renameat(td, AT_FDCWD, from, AT_FDCWD, to, UIO_SYSSPACE);
766 	LFREEPATH(from);
767 	LFREEPATH(to);
768 	return (error);
769 }
770 
771 int
772 linux_renameat(struct thread *td, struct linux_renameat_args *args)
773 {
774 	char *from, *to;
775 	int error, olddfd, newdfd;
776 
777 	olddfd = (args->olddfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->olddfd;
778 	newdfd = (args->newdfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->newdfd;
779 	LCONVPATHEXIST_AT(td, args->oldname, &from, olddfd);
780 	/* Expand LCONVPATHCREATE so that `from' can be freed on errors */
781 	error = linux_emul_convpath(td, args->newname, UIO_USERSPACE, &to, 1, newdfd);
782 	if (to == NULL) {
783 		LFREEPATH(from);
784 		return (error);
785 	}
786 
787 #ifdef DEBUG
788 	if (ldebug(renameat))
789 		printf(ARGS(renameat, "%s, %s"), from, to);
790 #endif
791 	error = kern_renameat(td, olddfd, from, newdfd, to, UIO_SYSSPACE);
792 	LFREEPATH(from);
793 	LFREEPATH(to);
794 	return (error);
795 }
796 
797 int
798 linux_symlink(struct thread *td, struct linux_symlink_args *args)
799 {
800 	char *path, *to;
801 	int error;
802 
803 	LCONVPATHEXIST(td, args->path, &path);
804 	/* Expand LCONVPATHCREATE so that `path' can be freed on errors */
805 	error = linux_emul_convpath(td, args->to, UIO_USERSPACE, &to, 1, AT_FDCWD);
806 	if (to == NULL) {
807 		LFREEPATH(path);
808 		return (error);
809 	}
810 
811 #ifdef DEBUG
812 	if (ldebug(symlink))
813 		printf(ARGS(symlink, "%s, %s"), path, to);
814 #endif
815 	error = kern_symlinkat(td, path, AT_FDCWD, to, UIO_SYSSPACE);
816 	LFREEPATH(path);
817 	LFREEPATH(to);
818 	return (error);
819 }
820 
821 int
822 linux_symlinkat(struct thread *td, struct linux_symlinkat_args *args)
823 {
824 	char *path, *to;
825 	int error, dfd;
826 
827 	dfd = (args->newdfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->newdfd;
828 	LCONVPATHEXIST_AT(td, args->oldname, &path, dfd);
829 	/* Expand LCONVPATHCREATE so that `path' can be freed on errors */
830 	error = linux_emul_convpath(td, args->newname, UIO_USERSPACE, &to, 1, dfd);
831 	if (to == NULL) {
832 		LFREEPATH(path);
833 		return (error);
834 	}
835 
836 #ifdef DEBUG
837 	if (ldebug(symlinkat))
838 		printf(ARGS(symlinkat, "%s, %s"), path, to);
839 #endif
840 
841 	error = kern_symlinkat(td, path, dfd, to, UIO_SYSSPACE);
842 	LFREEPATH(path);
843 	LFREEPATH(to);
844 	return (error);
845 }
846 
847 int
848 linux_readlink(struct thread *td, struct linux_readlink_args *args)
849 {
850 	char *name;
851 	int error;
852 
853 	LCONVPATHEXIST(td, args->name, &name);
854 
855 #ifdef DEBUG
856 	if (ldebug(readlink))
857 		printf(ARGS(readlink, "%s, %p, %d"), name, (void *)args->buf,
858 		    args->count);
859 #endif
860 	error = kern_readlinkat(td, AT_FDCWD, name, UIO_SYSSPACE,
861 	    args->buf, UIO_USERSPACE, args->count);
862 	LFREEPATH(name);
863 	return (error);
864 }
865 
866 int
867 linux_readlinkat(struct thread *td, struct linux_readlinkat_args *args)
868 {
869 	char *name;
870 	int error, dfd;
871 
872 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
873 	LCONVPATHEXIST_AT(td, args->path, &name, dfd);
874 
875 #ifdef DEBUG
876 	if (ldebug(readlinkat))
877 		printf(ARGS(readlinkat, "%s, %p, %d"), name, (void *)args->buf,
878 		    args->bufsiz);
879 #endif
880 
881 	error = kern_readlinkat(td, dfd, name, UIO_SYSSPACE, args->buf,
882 	    UIO_USERSPACE, args->bufsiz);
883 	LFREEPATH(name);
884 	return (error);
885 }
886 
887 int
888 linux_truncate(struct thread *td, struct linux_truncate_args *args)
889 {
890 	char *path;
891 	int error;
892 
893 	LCONVPATHEXIST(td, args->path, &path);
894 
895 #ifdef DEBUG
896 	if (ldebug(truncate))
897 		printf(ARGS(truncate, "%s, %ld"), path, (long)args->length);
898 #endif
899 
900 	error = kern_truncate(td, path, UIO_SYSSPACE, args->length);
901 	LFREEPATH(path);
902 	return (error);
903 }
904 
905 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
906 int
907 linux_truncate64(struct thread *td, struct linux_truncate64_args *args)
908 {
909 	char *path;
910 	int error;
911 
912 	LCONVPATHEXIST(td, args->path, &path);
913 
914 #ifdef DEBUG
915 	if (ldebug(truncate64))
916 		printf(ARGS(truncate64, "%s, %jd"), path, args->length);
917 #endif
918 
919 	error = kern_truncate(td, path, UIO_SYSSPACE, args->length);
920 	LFREEPATH(path);
921 	return (error);
922 }
923 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
924 
925 int
926 linux_ftruncate(struct thread *td, struct linux_ftruncate_args *args)
927 {
928 
929 	return (kern_ftruncate(td, args->fd, args->length));
930 }
931 
932 int
933 linux_link(struct thread *td, struct linux_link_args *args)
934 {
935 	char *path, *to;
936 	int error;
937 
938 	LCONVPATHEXIST(td, args->path, &path);
939 	/* Expand LCONVPATHCREATE so that `path' can be freed on errors */
940 	error = linux_emul_convpath(td, args->to, UIO_USERSPACE, &to, 1, AT_FDCWD);
941 	if (to == NULL) {
942 		LFREEPATH(path);
943 		return (error);
944 	}
945 
946 #ifdef DEBUG
947 	if (ldebug(link))
948 		printf(ARGS(link, "%s, %s"), path, to);
949 #endif
950 	error = kern_linkat(td, AT_FDCWD, AT_FDCWD, path, to, UIO_SYSSPACE,
951 	    FOLLOW);
952 	LFREEPATH(path);
953 	LFREEPATH(to);
954 	return (error);
955 }
956 
957 int
958 linux_linkat(struct thread *td, struct linux_linkat_args *args)
959 {
960 	char *path, *to;
961 	int error, olddfd, newdfd, follow;
962 
963 	if (args->flag & ~LINUX_AT_SYMLINK_FOLLOW)
964 		return (EINVAL);
965 
966 	olddfd = (args->olddfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->olddfd;
967 	newdfd = (args->newdfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->newdfd;
968 	LCONVPATHEXIST_AT(td, args->oldname, &path, olddfd);
969 	/* Expand LCONVPATHCREATE so that `path' can be freed on errors */
970 	error = linux_emul_convpath(td, args->newname, UIO_USERSPACE, &to, 1, newdfd);
971 	if (to == NULL) {
972 		LFREEPATH(path);
973 		return (error);
974 	}
975 
976 #ifdef DEBUG
977 	if (ldebug(linkat))
978 		printf(ARGS(linkat, "%i, %s, %i, %s, %i"), args->olddfd, path,
979 			args->newdfd, to, args->flag);
980 #endif
981 
982 	follow = (args->flag & LINUX_AT_SYMLINK_FOLLOW) == 0 ? NOFOLLOW :
983 	    FOLLOW;
984 	error = kern_linkat(td, olddfd, newdfd, path, to, UIO_SYSSPACE, follow);
985 	LFREEPATH(path);
986 	LFREEPATH(to);
987 	return (error);
988 }
989 
990 int
991 linux_fdatasync(td, uap)
992 	struct thread *td;
993 	struct linux_fdatasync_args *uap;
994 {
995 
996 	return (kern_fsync(td, uap->fd, false));
997 }
998 
999 int
1000 linux_pread(struct thread *td, struct linux_pread_args *uap)
1001 {
1002 	cap_rights_t rights;
1003 	struct vnode *vp;
1004 	int error;
1005 
1006 	error = kern_pread(td, uap->fd, uap->buf, uap->nbyte, uap->offset);
1007 	if (error == 0) {
1008 		/* This seems to violate POSIX but linux does it */
1009 		error = fgetvp(td, uap->fd,
1010 		    cap_rights_init(&rights, CAP_PREAD), &vp);
1011 		if (error != 0)
1012 			return (error);
1013 		if (vp->v_type == VDIR) {
1014 			vrele(vp);
1015 			return (EISDIR);
1016 		}
1017 		vrele(vp);
1018 	}
1019 	return (error);
1020 }
1021 
1022 int
1023 linux_pwrite(struct thread *td, struct linux_pwrite_args *uap)
1024 {
1025 
1026 	return (kern_pwrite(td, uap->fd, uap->buf, uap->nbyte, uap->offset));
1027 }
1028 
1029 int
1030 linux_mount(struct thread *td, struct linux_mount_args *args)
1031 {
1032 	char fstypename[MFSNAMELEN];
1033 	char mntonname[MNAMELEN], mntfromname[MNAMELEN];
1034 	int error;
1035 	int fsflags;
1036 
1037 	error = copyinstr(args->filesystemtype, fstypename, MFSNAMELEN - 1,
1038 	    NULL);
1039 	if (error)
1040 		return (error);
1041 	error = copyinstr(args->specialfile, mntfromname, MNAMELEN - 1, NULL);
1042 	if (error)
1043 		return (error);
1044 	error = copyinstr(args->dir, mntonname, MNAMELEN - 1, NULL);
1045 	if (error)
1046 		return (error);
1047 
1048 #ifdef DEBUG
1049 	if (ldebug(mount))
1050 		printf(ARGS(mount, "%s, %s, %s"),
1051 		    fstypename, mntfromname, mntonname);
1052 #endif
1053 
1054 	if (strcmp(fstypename, "ext2") == 0) {
1055 		strcpy(fstypename, "ext2fs");
1056 	} else if (strcmp(fstypename, "proc") == 0) {
1057 		strcpy(fstypename, "linprocfs");
1058 	} else if (strcmp(fstypename, "vfat") == 0) {
1059 		strcpy(fstypename, "msdosfs");
1060 	}
1061 
1062 	fsflags = 0;
1063 
1064 	if ((args->rwflag & 0xffff0000) == 0xc0ed0000) {
1065 		/*
1066 		 * Linux SYNC flag is not included; the closest equivalent
1067 		 * FreeBSD has is !ASYNC, which is our default.
1068 		 */
1069 		if (args->rwflag & LINUX_MS_RDONLY)
1070 			fsflags |= MNT_RDONLY;
1071 		if (args->rwflag & LINUX_MS_NOSUID)
1072 			fsflags |= MNT_NOSUID;
1073 		if (args->rwflag & LINUX_MS_NOEXEC)
1074 			fsflags |= MNT_NOEXEC;
1075 		if (args->rwflag & LINUX_MS_REMOUNT)
1076 			fsflags |= MNT_UPDATE;
1077 	}
1078 
1079 	error = kernel_vmount(fsflags,
1080 	    "fstype", fstypename,
1081 	    "fspath", mntonname,
1082 	    "from", mntfromname,
1083 	    NULL);
1084 	return (error);
1085 }
1086 
1087 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1088 int
1089 linux_oldumount(struct thread *td, struct linux_oldumount_args *args)
1090 {
1091 	struct linux_umount_args args2;
1092 
1093 	args2.path = args->path;
1094 	args2.flags = 0;
1095 	return (linux_umount(td, &args2));
1096 }
1097 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1098 
1099 int
1100 linux_umount(struct thread *td, struct linux_umount_args *args)
1101 {
1102 	struct unmount_args bsd;
1103 
1104 	bsd.path = args->path;
1105 	bsd.flags = args->flags;	/* XXX correct? */
1106 	return (sys_unmount(td, &bsd));
1107 }
1108 
1109 /*
1110  * fcntl family of syscalls
1111  */
1112 
1113 struct l_flock {
1114 	l_short		l_type;
1115 	l_short		l_whence;
1116 	l_off_t		l_start;
1117 	l_off_t		l_len;
1118 	l_pid_t		l_pid;
1119 }
1120 #if defined(__amd64__) && defined(COMPAT_LINUX32)
1121 __packed
1122 #endif
1123 ;
1124 
1125 static void
1126 linux_to_bsd_flock(struct l_flock *linux_flock, struct flock *bsd_flock)
1127 {
1128 	switch (linux_flock->l_type) {
1129 	case LINUX_F_RDLCK:
1130 		bsd_flock->l_type = F_RDLCK;
1131 		break;
1132 	case LINUX_F_WRLCK:
1133 		bsd_flock->l_type = F_WRLCK;
1134 		break;
1135 	case LINUX_F_UNLCK:
1136 		bsd_flock->l_type = F_UNLCK;
1137 		break;
1138 	default:
1139 		bsd_flock->l_type = -1;
1140 		break;
1141 	}
1142 	bsd_flock->l_whence = linux_flock->l_whence;
1143 	bsd_flock->l_start = (off_t)linux_flock->l_start;
1144 	bsd_flock->l_len = (off_t)linux_flock->l_len;
1145 	bsd_flock->l_pid = (pid_t)linux_flock->l_pid;
1146 	bsd_flock->l_sysid = 0;
1147 }
1148 
1149 static void
1150 bsd_to_linux_flock(struct flock *bsd_flock, struct l_flock *linux_flock)
1151 {
1152 	switch (bsd_flock->l_type) {
1153 	case F_RDLCK:
1154 		linux_flock->l_type = LINUX_F_RDLCK;
1155 		break;
1156 	case F_WRLCK:
1157 		linux_flock->l_type = LINUX_F_WRLCK;
1158 		break;
1159 	case F_UNLCK:
1160 		linux_flock->l_type = LINUX_F_UNLCK;
1161 		break;
1162 	}
1163 	linux_flock->l_whence = bsd_flock->l_whence;
1164 	linux_flock->l_start = (l_off_t)bsd_flock->l_start;
1165 	linux_flock->l_len = (l_off_t)bsd_flock->l_len;
1166 	linux_flock->l_pid = (l_pid_t)bsd_flock->l_pid;
1167 }
1168 
1169 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1170 struct l_flock64 {
1171 	l_short		l_type;
1172 	l_short		l_whence;
1173 	l_loff_t	l_start;
1174 	l_loff_t	l_len;
1175 	l_pid_t		l_pid;
1176 }
1177 #if defined(__amd64__) && defined(COMPAT_LINUX32)
1178 __packed
1179 #endif
1180 ;
1181 
1182 static void
1183 linux_to_bsd_flock64(struct l_flock64 *linux_flock, struct flock *bsd_flock)
1184 {
1185 	switch (linux_flock->l_type) {
1186 	case LINUX_F_RDLCK:
1187 		bsd_flock->l_type = F_RDLCK;
1188 		break;
1189 	case LINUX_F_WRLCK:
1190 		bsd_flock->l_type = F_WRLCK;
1191 		break;
1192 	case LINUX_F_UNLCK:
1193 		bsd_flock->l_type = F_UNLCK;
1194 		break;
1195 	default:
1196 		bsd_flock->l_type = -1;
1197 		break;
1198 	}
1199 	bsd_flock->l_whence = linux_flock->l_whence;
1200 	bsd_flock->l_start = (off_t)linux_flock->l_start;
1201 	bsd_flock->l_len = (off_t)linux_flock->l_len;
1202 	bsd_flock->l_pid = (pid_t)linux_flock->l_pid;
1203 	bsd_flock->l_sysid = 0;
1204 }
1205 
1206 static void
1207 bsd_to_linux_flock64(struct flock *bsd_flock, struct l_flock64 *linux_flock)
1208 {
1209 	switch (bsd_flock->l_type) {
1210 	case F_RDLCK:
1211 		linux_flock->l_type = LINUX_F_RDLCK;
1212 		break;
1213 	case F_WRLCK:
1214 		linux_flock->l_type = LINUX_F_WRLCK;
1215 		break;
1216 	case F_UNLCK:
1217 		linux_flock->l_type = LINUX_F_UNLCK;
1218 		break;
1219 	}
1220 	linux_flock->l_whence = bsd_flock->l_whence;
1221 	linux_flock->l_start = (l_loff_t)bsd_flock->l_start;
1222 	linux_flock->l_len = (l_loff_t)bsd_flock->l_len;
1223 	linux_flock->l_pid = (l_pid_t)bsd_flock->l_pid;
1224 }
1225 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1226 
1227 static int
1228 fcntl_common(struct thread *td, struct linux_fcntl_args *args)
1229 {
1230 	struct l_flock linux_flock;
1231 	struct flock bsd_flock;
1232 	cap_rights_t rights;
1233 	struct file *fp;
1234 	long arg;
1235 	int error, result;
1236 
1237 	switch (args->cmd) {
1238 	case LINUX_F_DUPFD:
1239 		return (kern_fcntl(td, args->fd, F_DUPFD, args->arg));
1240 
1241 	case LINUX_F_GETFD:
1242 		return (kern_fcntl(td, args->fd, F_GETFD, 0));
1243 
1244 	case LINUX_F_SETFD:
1245 		return (kern_fcntl(td, args->fd, F_SETFD, args->arg));
1246 
1247 	case LINUX_F_GETFL:
1248 		error = kern_fcntl(td, args->fd, F_GETFL, 0);
1249 		result = td->td_retval[0];
1250 		td->td_retval[0] = 0;
1251 		if (result & O_RDONLY)
1252 			td->td_retval[0] |= LINUX_O_RDONLY;
1253 		if (result & O_WRONLY)
1254 			td->td_retval[0] |= LINUX_O_WRONLY;
1255 		if (result & O_RDWR)
1256 			td->td_retval[0] |= LINUX_O_RDWR;
1257 		if (result & O_NDELAY)
1258 			td->td_retval[0] |= LINUX_O_NONBLOCK;
1259 		if (result & O_APPEND)
1260 			td->td_retval[0] |= LINUX_O_APPEND;
1261 		if (result & O_FSYNC)
1262 			td->td_retval[0] |= LINUX_O_SYNC;
1263 		if (result & O_ASYNC)
1264 			td->td_retval[0] |= LINUX_FASYNC;
1265 #ifdef LINUX_O_NOFOLLOW
1266 		if (result & O_NOFOLLOW)
1267 			td->td_retval[0] |= LINUX_O_NOFOLLOW;
1268 #endif
1269 #ifdef LINUX_O_DIRECT
1270 		if (result & O_DIRECT)
1271 			td->td_retval[0] |= LINUX_O_DIRECT;
1272 #endif
1273 		return (error);
1274 
1275 	case LINUX_F_SETFL:
1276 		arg = 0;
1277 		if (args->arg & LINUX_O_NDELAY)
1278 			arg |= O_NONBLOCK;
1279 		if (args->arg & LINUX_O_APPEND)
1280 			arg |= O_APPEND;
1281 		if (args->arg & LINUX_O_SYNC)
1282 			arg |= O_FSYNC;
1283 		if (args->arg & LINUX_FASYNC)
1284 			arg |= O_ASYNC;
1285 #ifdef LINUX_O_NOFOLLOW
1286 		if (args->arg & LINUX_O_NOFOLLOW)
1287 			arg |= O_NOFOLLOW;
1288 #endif
1289 #ifdef LINUX_O_DIRECT
1290 		if (args->arg & LINUX_O_DIRECT)
1291 			arg |= O_DIRECT;
1292 #endif
1293 		return (kern_fcntl(td, args->fd, F_SETFL, arg));
1294 
1295 	case LINUX_F_GETLK:
1296 		error = copyin((void *)args->arg, &linux_flock,
1297 		    sizeof(linux_flock));
1298 		if (error)
1299 			return (error);
1300 		linux_to_bsd_flock(&linux_flock, &bsd_flock);
1301 		error = kern_fcntl(td, args->fd, F_GETLK, (intptr_t)&bsd_flock);
1302 		if (error)
1303 			return (error);
1304 		bsd_to_linux_flock(&bsd_flock, &linux_flock);
1305 		return (copyout(&linux_flock, (void *)args->arg,
1306 		    sizeof(linux_flock)));
1307 
1308 	case LINUX_F_SETLK:
1309 		error = copyin((void *)args->arg, &linux_flock,
1310 		    sizeof(linux_flock));
1311 		if (error)
1312 			return (error);
1313 		linux_to_bsd_flock(&linux_flock, &bsd_flock);
1314 		return (kern_fcntl(td, args->fd, F_SETLK,
1315 		    (intptr_t)&bsd_flock));
1316 
1317 	case LINUX_F_SETLKW:
1318 		error = copyin((void *)args->arg, &linux_flock,
1319 		    sizeof(linux_flock));
1320 		if (error)
1321 			return (error);
1322 		linux_to_bsd_flock(&linux_flock, &bsd_flock);
1323 		return (kern_fcntl(td, args->fd, F_SETLKW,
1324 		     (intptr_t)&bsd_flock));
1325 
1326 	case LINUX_F_GETOWN:
1327 		return (kern_fcntl(td, args->fd, F_GETOWN, 0));
1328 
1329 	case LINUX_F_SETOWN:
1330 		/*
1331 		 * XXX some Linux applications depend on F_SETOWN having no
1332 		 * significant effect for pipes (SIGIO is not delivered for
1333 		 * pipes under Linux-2.2.35 at least).
1334 		 */
1335 		error = fget(td, args->fd,
1336 		    cap_rights_init(&rights, CAP_FCNTL), &fp);
1337 		if (error)
1338 			return (error);
1339 		if (fp->f_type == DTYPE_PIPE) {
1340 			fdrop(fp, td);
1341 			return (EINVAL);
1342 		}
1343 		fdrop(fp, td);
1344 
1345 		return (kern_fcntl(td, args->fd, F_SETOWN, args->arg));
1346 
1347 	case LINUX_F_DUPFD_CLOEXEC:
1348 		return (kern_fcntl(td, args->fd, F_DUPFD_CLOEXEC, args->arg));
1349 	}
1350 
1351 	return (EINVAL);
1352 }
1353 
1354 int
1355 linux_fcntl(struct thread *td, struct linux_fcntl_args *args)
1356 {
1357 
1358 #ifdef DEBUG
1359 	if (ldebug(fcntl))
1360 		printf(ARGS(fcntl, "%d, %08x, *"), args->fd, args->cmd);
1361 #endif
1362 
1363 	return (fcntl_common(td, args));
1364 }
1365 
1366 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1367 int
1368 linux_fcntl64(struct thread *td, struct linux_fcntl64_args *args)
1369 {
1370 	struct l_flock64 linux_flock;
1371 	struct flock bsd_flock;
1372 	struct linux_fcntl_args fcntl_args;
1373 	int error;
1374 
1375 #ifdef DEBUG
1376 	if (ldebug(fcntl64))
1377 		printf(ARGS(fcntl64, "%d, %08x, *"), args->fd, args->cmd);
1378 #endif
1379 
1380 	switch (args->cmd) {
1381 	case LINUX_F_GETLK64:
1382 		error = copyin((void *)args->arg, &linux_flock,
1383 		    sizeof(linux_flock));
1384 		if (error)
1385 			return (error);
1386 		linux_to_bsd_flock64(&linux_flock, &bsd_flock);
1387 		error = kern_fcntl(td, args->fd, F_GETLK, (intptr_t)&bsd_flock);
1388 		if (error)
1389 			return (error);
1390 		bsd_to_linux_flock64(&bsd_flock, &linux_flock);
1391 		return (copyout(&linux_flock, (void *)args->arg,
1392 			    sizeof(linux_flock)));
1393 
1394 	case LINUX_F_SETLK64:
1395 		error = copyin((void *)args->arg, &linux_flock,
1396 		    sizeof(linux_flock));
1397 		if (error)
1398 			return (error);
1399 		linux_to_bsd_flock64(&linux_flock, &bsd_flock);
1400 		return (kern_fcntl(td, args->fd, F_SETLK,
1401 		    (intptr_t)&bsd_flock));
1402 
1403 	case LINUX_F_SETLKW64:
1404 		error = copyin((void *)args->arg, &linux_flock,
1405 		    sizeof(linux_flock));
1406 		if (error)
1407 			return (error);
1408 		linux_to_bsd_flock64(&linux_flock, &bsd_flock);
1409 		return (kern_fcntl(td, args->fd, F_SETLKW,
1410 		    (intptr_t)&bsd_flock));
1411 	}
1412 
1413 	fcntl_args.fd = args->fd;
1414 	fcntl_args.cmd = args->cmd;
1415 	fcntl_args.arg = args->arg;
1416 	return (fcntl_common(td, &fcntl_args));
1417 }
1418 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1419 
1420 int
1421 linux_chown(struct thread *td, struct linux_chown_args *args)
1422 {
1423 	char *path;
1424 	int error;
1425 
1426 	LCONVPATHEXIST(td, args->path, &path);
1427 
1428 #ifdef DEBUG
1429 	if (ldebug(chown))
1430 		printf(ARGS(chown, "%s, %d, %d"), path, args->uid, args->gid);
1431 #endif
1432 	error = kern_fchownat(td, AT_FDCWD, path, UIO_SYSSPACE, args->uid,
1433 	    args->gid, 0);
1434 	LFREEPATH(path);
1435 	return (error);
1436 }
1437 
1438 int
1439 linux_fchownat(struct thread *td, struct linux_fchownat_args *args)
1440 {
1441 	char *path;
1442 	int error, dfd, flag;
1443 
1444 	if (args->flag & ~LINUX_AT_SYMLINK_NOFOLLOW)
1445 		return (EINVAL);
1446 
1447 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD :  args->dfd;
1448 	LCONVPATHEXIST_AT(td, args->filename, &path, dfd);
1449 
1450 #ifdef DEBUG
1451 	if (ldebug(fchownat))
1452 		printf(ARGS(fchownat, "%s, %d, %d"), path, args->uid, args->gid);
1453 #endif
1454 
1455 	flag = (args->flag & LINUX_AT_SYMLINK_NOFOLLOW) == 0 ? 0 :
1456 	    AT_SYMLINK_NOFOLLOW;
1457 	error = kern_fchownat(td, dfd, path, UIO_SYSSPACE, args->uid, args->gid,
1458 	    flag);
1459 	LFREEPATH(path);
1460 	return (error);
1461 }
1462 
1463 int
1464 linux_lchown(struct thread *td, struct linux_lchown_args *args)
1465 {
1466 	char *path;
1467 	int error;
1468 
1469 	LCONVPATHEXIST(td, args->path, &path);
1470 
1471 #ifdef DEBUG
1472 	if (ldebug(lchown))
1473 		printf(ARGS(lchown, "%s, %d, %d"), path, args->uid, args->gid);
1474 #endif
1475 	error = kern_fchownat(td, AT_FDCWD, path, UIO_SYSSPACE, args->uid,
1476 	    args->gid, AT_SYMLINK_NOFOLLOW);
1477 	LFREEPATH(path);
1478 	return (error);
1479 }
1480 
1481 static int
1482 convert_fadvice(int advice)
1483 {
1484 	switch (advice) {
1485 	case LINUX_POSIX_FADV_NORMAL:
1486 		return (POSIX_FADV_NORMAL);
1487 	case LINUX_POSIX_FADV_RANDOM:
1488 		return (POSIX_FADV_RANDOM);
1489 	case LINUX_POSIX_FADV_SEQUENTIAL:
1490 		return (POSIX_FADV_SEQUENTIAL);
1491 	case LINUX_POSIX_FADV_WILLNEED:
1492 		return (POSIX_FADV_WILLNEED);
1493 	case LINUX_POSIX_FADV_DONTNEED:
1494 		return (POSIX_FADV_DONTNEED);
1495 	case LINUX_POSIX_FADV_NOREUSE:
1496 		return (POSIX_FADV_NOREUSE);
1497 	default:
1498 		return (-1);
1499 	}
1500 }
1501 
1502 int
1503 linux_fadvise64(struct thread *td, struct linux_fadvise64_args *args)
1504 {
1505 	int advice;
1506 
1507 	advice = convert_fadvice(args->advice);
1508 	if (advice == -1)
1509 		return (EINVAL);
1510 	return (kern_posix_fadvise(td, args->fd, args->offset, args->len,
1511 	    advice));
1512 }
1513 
1514 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1515 int
1516 linux_fadvise64_64(struct thread *td, struct linux_fadvise64_64_args *args)
1517 {
1518 	int advice;
1519 
1520 	advice = convert_fadvice(args->advice);
1521 	if (advice == -1)
1522 		return (EINVAL);
1523 	return (kern_posix_fadvise(td, args->fd, args->offset, args->len,
1524 	    advice));
1525 }
1526 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1527 
1528 int
1529 linux_pipe(struct thread *td, struct linux_pipe_args *args)
1530 {
1531 	int fildes[2];
1532 	int error;
1533 
1534 #ifdef DEBUG
1535 	if (ldebug(pipe))
1536 		printf(ARGS(pipe, "*"));
1537 #endif
1538 
1539 	error = kern_pipe(td, fildes, 0, NULL, NULL);
1540 	if (error)
1541 		return (error);
1542 
1543 	/* XXX: Close descriptors on error. */
1544 	return (copyout(fildes, args->pipefds, sizeof(fildes)));
1545 }
1546 
1547 int
1548 linux_pipe2(struct thread *td, struct linux_pipe2_args *args)
1549 {
1550 	int fildes[2];
1551 	int error, flags;
1552 
1553 #ifdef DEBUG
1554 	if (ldebug(pipe2))
1555 		printf(ARGS(pipe2, "*, %d"), args->flags);
1556 #endif
1557 
1558 	if ((args->flags & ~(LINUX_O_NONBLOCK | LINUX_O_CLOEXEC)) != 0)
1559 		return (EINVAL);
1560 
1561 	flags = 0;
1562 	if ((args->flags & LINUX_O_NONBLOCK) != 0)
1563 		flags |= O_NONBLOCK;
1564 	if ((args->flags & LINUX_O_CLOEXEC) != 0)
1565 		flags |= O_CLOEXEC;
1566 	error = kern_pipe(td, fildes, flags, NULL, NULL);
1567 	if (error)
1568 		return (error);
1569 
1570 	/* XXX: Close descriptors on error. */
1571 	return (copyout(fildes, args->pipefds, sizeof(fildes)));
1572 }
1573 
1574 int
1575 linux_dup3(struct thread *td, struct linux_dup3_args *args)
1576 {
1577 	int cmd;
1578 	intptr_t newfd;
1579 
1580 	if (args->oldfd == args->newfd)
1581 		return (EINVAL);
1582 	if ((args->flags & ~LINUX_O_CLOEXEC) != 0)
1583 		return (EINVAL);
1584 	if (args->flags & LINUX_O_CLOEXEC)
1585 		cmd = F_DUP2FD_CLOEXEC;
1586 	else
1587 		cmd = F_DUP2FD;
1588 
1589 	newfd = args->newfd;
1590 	return (kern_fcntl(td, args->oldfd, cmd, newfd));
1591 }
1592 
1593 int
1594 linux_fallocate(struct thread *td, struct linux_fallocate_args *args)
1595 {
1596 
1597 	/*
1598 	 * We emulate only posix_fallocate system call for which
1599 	 * mode should be 0.
1600 	 */
1601 	if (args->mode != 0)
1602 		return (ENOSYS);
1603 
1604 	return (kern_posix_fallocate(td, args->fd, args->offset,
1605 	    args->len));
1606 }
1607