xref: /freebsd/sys/compat/linux/linux_file.c (revision 39ee7a7a6bdd1557b1c3532abf60d139798ac88b)
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 
77 #ifdef DEBUG
78 	if (ldebug(creat))
79 		printf(ARGS(creat, "%s, %d"), path, args->mode);
80 #endif
81     error = kern_openat(td, AT_FDCWD, path, UIO_SYSSPACE,
82 	O_WRONLY | O_CREAT | O_TRUNC, args->mode);
83     LFREEPATH(path);
84     return (error);
85 }
86 
87 
88 static int
89 linux_common_open(struct thread *td, int dirfd, char *path, int l_flags, int mode)
90 {
91     cap_rights_t rights;
92     struct proc *p = td->td_proc;
93     struct file *fp;
94     int fd;
95     int bsd_flags, error;
96 
97     bsd_flags = 0;
98     switch (l_flags & LINUX_O_ACCMODE) {
99     case LINUX_O_WRONLY:
100 	bsd_flags |= O_WRONLY;
101 	break;
102     case LINUX_O_RDWR:
103 	bsd_flags |= O_RDWR;
104 	break;
105     default:
106 	bsd_flags |= O_RDONLY;
107     }
108     if (l_flags & LINUX_O_NDELAY)
109 	bsd_flags |= O_NONBLOCK;
110     if (l_flags & LINUX_O_APPEND)
111 	bsd_flags |= O_APPEND;
112     if (l_flags & LINUX_O_SYNC)
113 	bsd_flags |= O_FSYNC;
114     if (l_flags & LINUX_O_NONBLOCK)
115 	bsd_flags |= O_NONBLOCK;
116     if (l_flags & LINUX_FASYNC)
117 	bsd_flags |= O_ASYNC;
118     if (l_flags & LINUX_O_CREAT)
119 	bsd_flags |= O_CREAT;
120     if (l_flags & LINUX_O_TRUNC)
121 	bsd_flags |= O_TRUNC;
122     if (l_flags & LINUX_O_EXCL)
123 	bsd_flags |= O_EXCL;
124     if (l_flags & LINUX_O_NOCTTY)
125 	bsd_flags |= O_NOCTTY;
126     if (l_flags & LINUX_O_DIRECT)
127 	bsd_flags |= O_DIRECT;
128     if (l_flags & LINUX_O_NOFOLLOW)
129 	bsd_flags |= O_NOFOLLOW;
130     if (l_flags & LINUX_O_DIRECTORY)
131 	bsd_flags |= O_DIRECTORY;
132     /* XXX LINUX_O_NOATIME: unable to be easily implemented. */
133 
134     error = kern_openat(td, dirfd, path, UIO_SYSSPACE, bsd_flags, mode);
135     if (error != 0)
136 	    goto done;
137 
138     if (bsd_flags & O_NOCTTY)
139 	    goto done;
140 
141     /*
142      * XXX In between kern_open() 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_rights_init(&rights, CAP_IOCTL), &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     if (ldebug(open))
170 	    printf(LMSG("open returns error %d"), error);
171 #endif
172     LFREEPATH(path);
173     return (error);
174 }
175 
176 int
177 linux_openat(struct thread *td, struct linux_openat_args *args)
178 {
179 	char *path;
180 	int dfd;
181 
182 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
183 	if (args->flags & LINUX_O_CREAT)
184 		LCONVPATH_AT(td, args->filename, &path, 1, dfd);
185 	else
186 		LCONVPATH_AT(td, args->filename, &path, 0, dfd);
187 #ifdef DEBUG
188 	if (ldebug(openat))
189 		printf(ARGS(openat, "%i, %s, 0x%x, 0x%x"), args->dfd,
190 		    path, args->flags, args->mode);
191 #endif
192 	return (linux_common_open(td, dfd, path, args->flags, args->mode));
193 }
194 
195 int
196 linux_open(struct thread *td, struct linux_open_args *args)
197 {
198     char *path;
199 
200     if (args->flags & LINUX_O_CREAT)
201 	LCONVPATHCREAT(td, args->path, &path);
202     else
203 	LCONVPATHEXIST(td, args->path, &path);
204 
205 #ifdef DEBUG
206 	if (ldebug(open))
207 		printf(ARGS(open, "%s, 0x%x, 0x%x"),
208 		    path, args->flags, args->mode);
209 #endif
210 
211 	return (linux_common_open(td, AT_FDCWD, path, args->flags, args->mode));
212 }
213 
214 int
215 linux_lseek(struct thread *td, struct linux_lseek_args *args)
216 {
217 
218     struct lseek_args /* {
219 	int fd;
220 	int pad;
221 	off_t offset;
222 	int whence;
223     } */ tmp_args;
224     int error;
225 
226 #ifdef DEBUG
227 	if (ldebug(lseek))
228 		printf(ARGS(lseek, "%d, %ld, %d"),
229 		    args->fdes, (long)args->off, args->whence);
230 #endif
231     tmp_args.fd = args->fdes;
232     tmp_args.offset = (off_t)args->off;
233     tmp_args.whence = args->whence;
234     error = sys_lseek(td, &tmp_args);
235     return error;
236 }
237 
238 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
239 int
240 linux_llseek(struct thread *td, struct linux_llseek_args *args)
241 {
242 	struct lseek_args bsd_args;
243 	int error;
244 	off_t off;
245 
246 #ifdef DEBUG
247 	if (ldebug(llseek))
248 		printf(ARGS(llseek, "%d, %d:%d, %d"),
249 		    args->fd, args->ohigh, args->olow, args->whence);
250 #endif
251 	off = (args->olow) | (((off_t) args->ohigh) << 32);
252 
253 	bsd_args.fd = args->fd;
254 	bsd_args.offset = off;
255 	bsd_args.whence = args->whence;
256 
257 	if ((error = sys_lseek(td, &bsd_args)))
258 		return error;
259 
260 	if ((error = copyout(td->td_retval, args->res, sizeof (off_t))))
261 		return error;
262 
263 	td->td_retval[0] = 0;
264 	return 0;
265 }
266 
267 int
268 linux_readdir(struct thread *td, struct linux_readdir_args *args)
269 {
270 	struct linux_getdents_args lda;
271 
272 	lda.fd = args->fd;
273 	lda.dent = args->dent;
274 	lda.count = 1;
275 	return linux_getdents(td, &lda);
276 }
277 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
278 
279 /*
280  * Note that linux_getdents(2) and linux_getdents64(2) have the same
281  * arguments. They only differ in the definition of struct dirent they
282  * operate on. We use this to common the code, with the exception of
283  * accessing struct dirent. Note that linux_readdir(2) is implemented
284  * by means of linux_getdents(2). In this case we never operate on
285  * struct dirent64 and thus don't need to handle it...
286  */
287 
288 struct l_dirent {
289 	l_ulong		d_ino;
290 	l_off_t		d_off;
291 	l_ushort	d_reclen;
292 	char		d_name[LINUX_NAME_MAX + 1];
293 };
294 
295 struct l_dirent64 {
296 	uint64_t	d_ino;
297 	int64_t		d_off;
298 	l_ushort	d_reclen;
299 	u_char		d_type;
300 	char		d_name[LINUX_NAME_MAX + 1];
301 };
302 
303 /*
304  * Linux uses the last byte in the dirent buffer to store d_type,
305  * at least glibc-2.7 requires it. That is why l_dirent is padded with 2 bytes.
306  */
307 #define LINUX_RECLEN(namlen)						\
308     roundup((offsetof(struct l_dirent, d_name) + (namlen) + 2),		\
309     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 #define LINUX_MAXRECLEN		max(LINUX_RECLEN(LINUX_NAME_MAX),	\
316 				    LINUX_RECLEN64(LINUX_NAME_MAX))
317 #define	LINUX_DIRBLKSIZ		512
318 
319 static int
320 getdents_common(struct thread *td, struct linux_getdents64_args *args,
321     int is64bit)
322 {
323 	struct dirent *bdp;
324 	struct vnode *vp;
325 	caddr_t inp, buf;		/* BSD-format */
326 	int len, reclen;		/* BSD-format */
327 	caddr_t outp;			/* Linux-format */
328 	int resid, linuxreclen=0;	/* Linux-format */
329 	caddr_t lbuf;			/* Linux-format */
330 	cap_rights_t rights;
331 	struct file *fp;
332 	struct uio auio;
333 	struct iovec aiov;
334 	off_t off;
335 	struct l_dirent *linux_dirent;
336 	struct l_dirent64 *linux_dirent64;
337 	int buflen, error, eofflag, nbytes, justone;
338 	u_long *cookies = NULL, *cookiep;
339 	int ncookies;
340 
341 	nbytes = args->count;
342 	if (nbytes == 1) {
343 		/* readdir(2) case. Always struct dirent. */
344 		if (is64bit)
345 			return (EINVAL);
346 		nbytes = sizeof(*linux_dirent);
347 		justone = 1;
348 	} else
349 		justone = 0;
350 
351 	error = getvnode(td, args->fd, cap_rights_init(&rights, CAP_READ), &fp);
352 	if (error != 0)
353 		return (error);
354 
355 	if ((fp->f_flag & FREAD) == 0) {
356 		fdrop(fp, td);
357 		return (EBADF);
358 	}
359 
360 	off = foffset_lock(fp, 0);
361 	vp = fp->f_vnode;
362 	if (vp->v_type != VDIR) {
363 		foffset_unlock(fp, off, 0);
364 		fdrop(fp, td);
365 		return (EINVAL);
366 	}
367 
368 
369 	buflen = max(LINUX_DIRBLKSIZ, nbytes);
370 	buflen = min(buflen, MAXBSIZE);
371 	buf = malloc(buflen, M_LINUX, M_WAITOK);
372 	lbuf = malloc(LINUX_MAXRECLEN, M_LINUX, M_WAITOK | M_ZERO);
373 	vn_lock(vp, LK_SHARED | LK_RETRY);
374 
375 	aiov.iov_base = buf;
376 	aiov.iov_len = buflen;
377 	auio.uio_iov = &aiov;
378 	auio.uio_iovcnt = 1;
379 	auio.uio_rw = UIO_READ;
380 	auio.uio_segflg = UIO_SYSSPACE;
381 	auio.uio_td = td;
382 	auio.uio_resid = buflen;
383 	auio.uio_offset = off;
384 
385 #ifdef MAC
386 	/*
387 	 * Do directory search MAC check using non-cached credentials.
388 	 */
389 	if ((error = mac_vnode_check_readdir(td->td_ucred, vp)))
390 		goto out;
391 #endif /* MAC */
392 	if ((error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, &ncookies,
393 		 &cookies)))
394 		goto out;
395 
396 	inp = buf;
397 	outp = (caddr_t)args->dirent;
398 	resid = nbytes;
399 	if ((len = buflen - auio.uio_resid) <= 0)
400 		goto eof;
401 
402 	cookiep = cookies;
403 
404 	if (cookies) {
405 		/*
406 		 * When using cookies, the vfs has the option of reading from
407 		 * a different offset than that supplied (UFS truncates the
408 		 * offset to a block boundary to make sure that it never reads
409 		 * partway through a directory entry, even if the directory
410 		 * has been compacted).
411 		 */
412 		while (len > 0 && ncookies > 0 && *cookiep <= off) {
413 			bdp = (struct dirent *) inp;
414 			len -= bdp->d_reclen;
415 			inp += bdp->d_reclen;
416 			cookiep++;
417 			ncookies--;
418 		}
419 	}
420 
421 	while (len > 0) {
422 		if (cookiep && ncookies == 0)
423 			break;
424 		bdp = (struct dirent *) inp;
425 		reclen = bdp->d_reclen;
426 		if (reclen & 3) {
427 			error = EFAULT;
428 			goto out;
429 		}
430 
431 		if (bdp->d_fileno == 0) {
432 			inp += reclen;
433 			if (cookiep) {
434 				off = *cookiep++;
435 				ncookies--;
436 			} else
437 				off += reclen;
438 
439 			len -= reclen;
440 			continue;
441 		}
442 
443 		linuxreclen = (is64bit)
444 		    ? LINUX_RECLEN64(bdp->d_namlen)
445 		    : LINUX_RECLEN(bdp->d_namlen);
446 
447 		if (reclen > len || resid < linuxreclen) {
448 			outp++;
449 			break;
450 		}
451 
452 		if (justone) {
453 			/* readdir(2) case. */
454 			linux_dirent = (struct l_dirent*)lbuf;
455 			linux_dirent->d_ino = bdp->d_fileno;
456 			linux_dirent->d_off = (l_off_t)linuxreclen;
457 			linux_dirent->d_reclen = (l_ushort)bdp->d_namlen;
458 			strlcpy(linux_dirent->d_name, bdp->d_name,
459 			    linuxreclen - offsetof(struct l_dirent, d_name));
460 			error = copyout(linux_dirent, outp, linuxreclen);
461 		}
462 		if (is64bit) {
463 			linux_dirent64 = (struct l_dirent64*)lbuf;
464 			linux_dirent64->d_ino = bdp->d_fileno;
465 			linux_dirent64->d_off = (cookiep)
466 			    ? (l_off_t)*cookiep
467 			    : (l_off_t)(off + reclen);
468 			linux_dirent64->d_reclen = (l_ushort)linuxreclen;
469 			linux_dirent64->d_type = bdp->d_type;
470 			strlcpy(linux_dirent64->d_name, bdp->d_name,
471 			    linuxreclen - offsetof(struct l_dirent64, d_name));
472 			error = copyout(linux_dirent64, outp, linuxreclen);
473 		} else if (!justone) {
474 			linux_dirent = (struct l_dirent*)lbuf;
475 			linux_dirent->d_ino = bdp->d_fileno;
476 			linux_dirent->d_off = (cookiep)
477 			    ? (l_off_t)*cookiep
478 			    : (l_off_t)(off + reclen);
479 			linux_dirent->d_reclen = (l_ushort)linuxreclen;
480 			/*
481 			 * Copy d_type to last byte of l_dirent buffer
482 			 */
483 			lbuf[linuxreclen-1] = bdp->d_type;
484 			strlcpy(linux_dirent->d_name, bdp->d_name,
485 			    linuxreclen - offsetof(struct l_dirent, d_name)-1);
486 			error = copyout(linux_dirent, outp, linuxreclen);
487 		}
488 
489 		if (error)
490 			goto out;
491 
492 		inp += reclen;
493 		if (cookiep) {
494 			off = *cookiep++;
495 			ncookies--;
496 		} else
497 			off += reclen;
498 
499 		outp += linuxreclen;
500 		resid -= linuxreclen;
501 		len -= reclen;
502 		if (justone)
503 			break;
504 	}
505 
506 	if (outp == (caddr_t)args->dirent) {
507 		nbytes = resid;
508 		goto eof;
509 	}
510 
511 	if (justone)
512 		nbytes = resid + linuxreclen;
513 
514 eof:
515 	td->td_retval[0] = nbytes - resid;
516 
517 out:
518 	free(cookies, M_TEMP);
519 
520 	VOP_UNLOCK(vp, 0);
521 	foffset_unlock(fp, off, 0);
522 	fdrop(fp, td);
523 	free(buf, M_LINUX);
524 	free(lbuf, M_LINUX);
525 	return (error);
526 }
527 
528 int
529 linux_getdents(struct thread *td, struct linux_getdents_args *args)
530 {
531 
532 #ifdef DEBUG
533 	if (ldebug(getdents))
534 		printf(ARGS(getdents, "%d, *, %d"), args->fd, args->count);
535 #endif
536 
537 	return (getdents_common(td, (struct linux_getdents64_args*)args, 0));
538 }
539 
540 int
541 linux_getdents64(struct thread *td, struct linux_getdents64_args *args)
542 {
543 
544 #ifdef DEBUG
545 	if (ldebug(getdents64))
546 		printf(ARGS(getdents64, "%d, *, %d"), args->fd, args->count);
547 #endif
548 
549 	return (getdents_common(td, args, 1));
550 }
551 
552 /*
553  * These exist mainly for hooks for doing /compat/linux translation.
554  */
555 
556 int
557 linux_access(struct thread *td, struct linux_access_args *args)
558 {
559 	char *path;
560 	int error;
561 
562 	/* linux convention */
563 	if (args->amode & ~(F_OK | X_OK | W_OK | R_OK))
564 		return (EINVAL);
565 
566 	LCONVPATHEXIST(td, args->path, &path);
567 
568 #ifdef DEBUG
569 	if (ldebug(access))
570 		printf(ARGS(access, "%s, %d"), path, args->amode);
571 #endif
572 	error = kern_accessat(td, AT_FDCWD, path, UIO_SYSSPACE, 0,
573 	    args->amode);
574 	LFREEPATH(path);
575 
576 	return (error);
577 }
578 
579 int
580 linux_faccessat(struct thread *td, struct linux_faccessat_args *args)
581 {
582 	char *path;
583 	int error, dfd;
584 
585 	/* linux convention */
586 	if (args->amode & ~(F_OK | X_OK | W_OK | R_OK))
587 		return (EINVAL);
588 
589 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
590 	LCONVPATHEXIST_AT(td, args->filename, &path, dfd);
591 
592 #ifdef DEBUG
593 	if (ldebug(access))
594 		printf(ARGS(access, "%s, %d"), path, args->amode);
595 #endif
596 
597 	error = kern_accessat(td, dfd, path, UIO_SYSSPACE, 0, args->amode);
598 	LFREEPATH(path);
599 
600 	return (error);
601 }
602 
603 int
604 linux_unlink(struct thread *td, struct linux_unlink_args *args)
605 {
606 	char *path;
607 	int error;
608 	struct stat st;
609 
610 	LCONVPATHEXIST(td, args->path, &path);
611 
612 #ifdef DEBUG
613 	if (ldebug(unlink))
614 		printf(ARGS(unlink, "%s"), path);
615 #endif
616 
617 	error = kern_unlinkat(td, AT_FDCWD, path, UIO_SYSSPACE, 0);
618 	if (error == EPERM) {
619 		/* Introduce POSIX noncompliant behaviour of Linux */
620 		if (kern_statat(td, 0, AT_FDCWD, path, UIO_SYSSPACE, &st,
621 		    NULL) == 0) {
622 			if (S_ISDIR(st.st_mode))
623 				error = EISDIR;
624 		}
625 	}
626 	LFREEPATH(path);
627 	return (error);
628 }
629 
630 int
631 linux_unlinkat(struct thread *td, struct linux_unlinkat_args *args)
632 {
633 	char *path;
634 	int error, dfd;
635 	struct stat st;
636 
637 	if (args->flag & ~LINUX_AT_REMOVEDIR)
638 		return (EINVAL);
639 
640 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
641 	LCONVPATHEXIST_AT(td, args->pathname, &path, dfd);
642 
643 #ifdef DEBUG
644 	if (ldebug(unlinkat))
645 		printf(ARGS(unlinkat, "%s"), path);
646 #endif
647 
648 	if (args->flag & LINUX_AT_REMOVEDIR)
649 		error = kern_rmdirat(td, dfd, path, UIO_SYSSPACE);
650 	else
651 		error = kern_unlinkat(td, dfd, path, UIO_SYSSPACE, 0);
652 	if (error == EPERM && !(args->flag & LINUX_AT_REMOVEDIR)) {
653 		/* Introduce POSIX noncompliant behaviour of Linux */
654 		if (kern_statat(td, AT_SYMLINK_NOFOLLOW, dfd, path,
655 		    UIO_SYSSPACE, &st, NULL) == 0 && S_ISDIR(st.st_mode))
656 			error = EISDIR;
657 	}
658 	LFREEPATH(path);
659 	return (error);
660 }
661 int
662 linux_chdir(struct thread *td, struct linux_chdir_args *args)
663 {
664 	char *path;
665 	int error;
666 
667 	LCONVPATHEXIST(td, args->path, &path);
668 
669 #ifdef DEBUG
670 	if (ldebug(chdir))
671 		printf(ARGS(chdir, "%s"), path);
672 #endif
673 	error = kern_chdir(td, path, UIO_SYSSPACE);
674 	LFREEPATH(path);
675 	return (error);
676 }
677 
678 int
679 linux_chmod(struct thread *td, struct linux_chmod_args *args)
680 {
681 	char *path;
682 	int error;
683 
684 	LCONVPATHEXIST(td, args->path, &path);
685 
686 #ifdef DEBUG
687 	if (ldebug(chmod))
688 		printf(ARGS(chmod, "%s, %d"), path, args->mode);
689 #endif
690 	error = kern_fchmodat(td, AT_FDCWD, path, UIO_SYSSPACE,
691 	    args->mode, 0);
692 	LFREEPATH(path);
693 	return (error);
694 }
695 
696 int
697 linux_fchmodat(struct thread *td, struct linux_fchmodat_args *args)
698 {
699 	char *path;
700 	int error, dfd;
701 
702 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
703 	LCONVPATHEXIST_AT(td, args->filename, &path, dfd);
704 
705 #ifdef DEBUG
706 	if (ldebug(fchmodat))
707 		printf(ARGS(fchmodat, "%s, %d"), path, args->mode);
708 #endif
709 
710 	error = kern_fchmodat(td, dfd, path, UIO_SYSSPACE, args->mode, 0);
711 	LFREEPATH(path);
712 	return (error);
713 }
714 
715 int
716 linux_mkdir(struct thread *td, struct linux_mkdir_args *args)
717 {
718 	char *path;
719 	int error;
720 
721 	LCONVPATHCREAT(td, args->path, &path);
722 
723 #ifdef DEBUG
724 	if (ldebug(mkdir))
725 		printf(ARGS(mkdir, "%s, %d"), path, args->mode);
726 #endif
727 	error = kern_mkdirat(td, AT_FDCWD, path, UIO_SYSSPACE, args->mode);
728 	LFREEPATH(path);
729 	return (error);
730 }
731 
732 int
733 linux_mkdirat(struct thread *td, struct linux_mkdirat_args *args)
734 {
735 	char *path;
736 	int error, dfd;
737 
738 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
739 	LCONVPATHCREAT_AT(td, args->pathname, &path, dfd);
740 
741 #ifdef DEBUG
742 	if (ldebug(mkdirat))
743 		printf(ARGS(mkdirat, "%s, %d"), path, args->mode);
744 #endif
745 	error = kern_mkdirat(td, dfd, path, UIO_SYSSPACE, args->mode);
746 	LFREEPATH(path);
747 	return (error);
748 }
749 
750 int
751 linux_rmdir(struct thread *td, struct linux_rmdir_args *args)
752 {
753 	char *path;
754 	int error;
755 
756 	LCONVPATHEXIST(td, args->path, &path);
757 
758 #ifdef DEBUG
759 	if (ldebug(rmdir))
760 		printf(ARGS(rmdir, "%s"), path);
761 #endif
762 	error = kern_rmdirat(td, AT_FDCWD, path, UIO_SYSSPACE);
763 	LFREEPATH(path);
764 	return (error);
765 }
766 
767 int
768 linux_rename(struct thread *td, struct linux_rename_args *args)
769 {
770 	char *from, *to;
771 	int error;
772 
773 	LCONVPATHEXIST(td, args->from, &from);
774 	/* Expand LCONVPATHCREATE so that `from' can be freed on errors */
775 	error = linux_emul_convpath(td, args->to, UIO_USERSPACE, &to, 1, AT_FDCWD);
776 	if (to == NULL) {
777 		LFREEPATH(from);
778 		return (error);
779 	}
780 
781 #ifdef DEBUG
782 	if (ldebug(rename))
783 		printf(ARGS(rename, "%s, %s"), from, to);
784 #endif
785 	error = kern_renameat(td, AT_FDCWD, from, AT_FDCWD, to, UIO_SYSSPACE);
786 	LFREEPATH(from);
787 	LFREEPATH(to);
788 	return (error);
789 }
790 
791 int
792 linux_renameat(struct thread *td, struct linux_renameat_args *args)
793 {
794 	char *from, *to;
795 	int error, olddfd, newdfd;
796 
797 	olddfd = (args->olddfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->olddfd;
798 	newdfd = (args->newdfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->newdfd;
799 	LCONVPATHEXIST_AT(td, args->oldname, &from, olddfd);
800 	/* Expand LCONVPATHCREATE so that `from' can be freed on errors */
801 	error = linux_emul_convpath(td, args->newname, UIO_USERSPACE, &to, 1, newdfd);
802 	if (to == NULL) {
803 		LFREEPATH(from);
804 		return (error);
805 	}
806 
807 #ifdef DEBUG
808 	if (ldebug(renameat))
809 		printf(ARGS(renameat, "%s, %s"), from, to);
810 #endif
811 	error = kern_renameat(td, olddfd, from, newdfd, to, UIO_SYSSPACE);
812 	LFREEPATH(from);
813 	LFREEPATH(to);
814 	return (error);
815 }
816 
817 int
818 linux_symlink(struct thread *td, struct linux_symlink_args *args)
819 {
820 	char *path, *to;
821 	int error;
822 
823 	LCONVPATHEXIST(td, args->path, &path);
824 	/* Expand LCONVPATHCREATE so that `path' can be freed on errors */
825 	error = linux_emul_convpath(td, args->to, UIO_USERSPACE, &to, 1, AT_FDCWD);
826 	if (to == NULL) {
827 		LFREEPATH(path);
828 		return (error);
829 	}
830 
831 #ifdef DEBUG
832 	if (ldebug(symlink))
833 		printf(ARGS(symlink, "%s, %s"), path, to);
834 #endif
835 	error = kern_symlinkat(td, path, AT_FDCWD, to, UIO_SYSSPACE);
836 	LFREEPATH(path);
837 	LFREEPATH(to);
838 	return (error);
839 }
840 
841 int
842 linux_symlinkat(struct thread *td, struct linux_symlinkat_args *args)
843 {
844 	char *path, *to;
845 	int error, dfd;
846 
847 	dfd = (args->newdfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->newdfd;
848 	LCONVPATHEXIST_AT(td, args->oldname, &path, dfd);
849 	/* Expand LCONVPATHCREATE so that `path' can be freed on errors */
850 	error = linux_emul_convpath(td, args->newname, UIO_USERSPACE, &to, 1, dfd);
851 	if (to == NULL) {
852 		LFREEPATH(path);
853 		return (error);
854 	}
855 
856 #ifdef DEBUG
857 	if (ldebug(symlinkat))
858 		printf(ARGS(symlinkat, "%s, %s"), path, to);
859 #endif
860 
861 	error = kern_symlinkat(td, path, dfd, to, UIO_SYSSPACE);
862 	LFREEPATH(path);
863 	LFREEPATH(to);
864 	return (error);
865 }
866 
867 int
868 linux_readlink(struct thread *td, struct linux_readlink_args *args)
869 {
870 	char *name;
871 	int error;
872 
873 	LCONVPATHEXIST(td, args->name, &name);
874 
875 #ifdef DEBUG
876 	if (ldebug(readlink))
877 		printf(ARGS(readlink, "%s, %p, %d"), name, (void *)args->buf,
878 		    args->count);
879 #endif
880 	error = kern_readlinkat(td, AT_FDCWD, name, UIO_SYSSPACE,
881 	    args->buf, UIO_USERSPACE, args->count);
882 	LFREEPATH(name);
883 	return (error);
884 }
885 
886 int
887 linux_readlinkat(struct thread *td, struct linux_readlinkat_args *args)
888 {
889 	char *name;
890 	int error, dfd;
891 
892 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
893 	LCONVPATHEXIST_AT(td, args->path, &name, dfd);
894 
895 #ifdef DEBUG
896 	if (ldebug(readlinkat))
897 		printf(ARGS(readlinkat, "%s, %p, %d"), name, (void *)args->buf,
898 		    args->bufsiz);
899 #endif
900 
901 	error = kern_readlinkat(td, dfd, name, UIO_SYSSPACE, args->buf,
902 	    UIO_USERSPACE, args->bufsiz);
903 	LFREEPATH(name);
904 	return (error);
905 }
906 
907 int
908 linux_truncate(struct thread *td, struct linux_truncate_args *args)
909 {
910 	char *path;
911 	int error;
912 
913 	LCONVPATHEXIST(td, args->path, &path);
914 
915 #ifdef DEBUG
916 	if (ldebug(truncate))
917 		printf(ARGS(truncate, "%s, %ld"), path, (long)args->length);
918 #endif
919 
920 	error = kern_truncate(td, path, UIO_SYSSPACE, args->length);
921 	LFREEPATH(path);
922 	return (error);
923 }
924 
925 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
926 int
927 linux_truncate64(struct thread *td, struct linux_truncate64_args *args)
928 {
929 	char *path;
930 	int error;
931 
932 	LCONVPATHEXIST(td, args->path, &path);
933 
934 #ifdef DEBUG
935 	if (ldebug(truncate64))
936 		printf(ARGS(truncate64, "%s, %jd"), path, args->length);
937 #endif
938 
939 	error = kern_truncate(td, path, UIO_SYSSPACE, args->length);
940 	LFREEPATH(path);
941 	return (error);
942 }
943 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
944 
945 int
946 linux_ftruncate(struct thread *td, struct linux_ftruncate_args *args)
947 {
948 	struct ftruncate_args /* {
949 		int fd;
950 		int pad;
951 		off_t length;
952 		} */ nuap;
953 
954 	nuap.fd = args->fd;
955 	nuap.length = args->length;
956 	return (sys_ftruncate(td, &nuap));
957 }
958 
959 int
960 linux_link(struct thread *td, struct linux_link_args *args)
961 {
962 	char *path, *to;
963 	int error;
964 
965 	LCONVPATHEXIST(td, args->path, &path);
966 	/* Expand LCONVPATHCREATE so that `path' can be freed on errors */
967 	error = linux_emul_convpath(td, args->to, UIO_USERSPACE, &to, 1, AT_FDCWD);
968 	if (to == NULL) {
969 		LFREEPATH(path);
970 		return (error);
971 	}
972 
973 #ifdef DEBUG
974 	if (ldebug(link))
975 		printf(ARGS(link, "%s, %s"), path, to);
976 #endif
977 	error = kern_linkat(td, AT_FDCWD, AT_FDCWD, path, to, UIO_SYSSPACE,
978 	    FOLLOW);
979 	LFREEPATH(path);
980 	LFREEPATH(to);
981 	return (error);
982 }
983 
984 int
985 linux_linkat(struct thread *td, struct linux_linkat_args *args)
986 {
987 	char *path, *to;
988 	int error, olddfd, newdfd, follow;
989 
990 	if (args->flag & ~LINUX_AT_SYMLINK_FOLLOW)
991 		return (EINVAL);
992 
993 	olddfd = (args->olddfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->olddfd;
994 	newdfd = (args->newdfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->newdfd;
995 	LCONVPATHEXIST_AT(td, args->oldname, &path, olddfd);
996 	/* Expand LCONVPATHCREATE so that `path' can be freed on errors */
997 	error = linux_emul_convpath(td, args->newname, UIO_USERSPACE, &to, 1, newdfd);
998 	if (to == NULL) {
999 		LFREEPATH(path);
1000 		return (error);
1001 	}
1002 
1003 #ifdef DEBUG
1004 	if (ldebug(linkat))
1005 		printf(ARGS(linkat, "%i, %s, %i, %s, %i"), args->olddfd, path,
1006 			args->newdfd, to, args->flag);
1007 #endif
1008 
1009 	follow = (args->flag & LINUX_AT_SYMLINK_FOLLOW) == 0 ? NOFOLLOW :
1010 	    FOLLOW;
1011 	error = kern_linkat(td, olddfd, newdfd, path, to, UIO_SYSSPACE, follow);
1012 	LFREEPATH(path);
1013 	LFREEPATH(to);
1014 	return (error);
1015 }
1016 
1017 int
1018 linux_fdatasync(td, uap)
1019 	struct thread *td;
1020 	struct linux_fdatasync_args *uap;
1021 {
1022 	struct fsync_args bsd;
1023 
1024 	bsd.fd = uap->fd;
1025 	return sys_fsync(td, &bsd);
1026 }
1027 
1028 int
1029 linux_pread(td, uap)
1030 	struct thread *td;
1031 	struct linux_pread_args *uap;
1032 {
1033 	struct pread_args bsd;
1034 	cap_rights_t rights;
1035 	struct vnode *vp;
1036 	int error;
1037 
1038 	bsd.fd = uap->fd;
1039 	bsd.buf = uap->buf;
1040 	bsd.nbyte = uap->nbyte;
1041 	bsd.offset = uap->offset;
1042 
1043 	error = sys_pread(td, &bsd);
1044 
1045 	if (error == 0) {
1046 		/* This seems to violate POSIX but linux does it */
1047 		error = fgetvp(td, uap->fd,
1048 		    cap_rights_init(&rights, CAP_PREAD), &vp);
1049 		if (error != 0)
1050 			return (error);
1051 		if (vp->v_type == VDIR) {
1052 			vrele(vp);
1053 			return (EISDIR);
1054 		}
1055 		vrele(vp);
1056 	}
1057 
1058 	return (error);
1059 }
1060 
1061 int
1062 linux_pwrite(td, uap)
1063 	struct thread *td;
1064 	struct linux_pwrite_args *uap;
1065 {
1066 	struct pwrite_args bsd;
1067 
1068 	bsd.fd = uap->fd;
1069 	bsd.buf = uap->buf;
1070 	bsd.nbyte = uap->nbyte;
1071 	bsd.offset = uap->offset;
1072 	return sys_pwrite(td, &bsd);
1073 }
1074 
1075 int
1076 linux_mount(struct thread *td, struct linux_mount_args *args)
1077 {
1078 	char fstypename[MFSNAMELEN];
1079 	char mntonname[MNAMELEN], mntfromname[MNAMELEN];
1080 	int error;
1081 	int fsflags;
1082 
1083 	error = copyinstr(args->filesystemtype, fstypename, MFSNAMELEN - 1,
1084 	    NULL);
1085 	if (error)
1086 		return (error);
1087 	error = copyinstr(args->specialfile, mntfromname, MNAMELEN - 1, NULL);
1088 	if (error)
1089 		return (error);
1090 	error = copyinstr(args->dir, mntonname, MNAMELEN - 1, NULL);
1091 	if (error)
1092 		return (error);
1093 
1094 #ifdef DEBUG
1095 	if (ldebug(mount))
1096 		printf(ARGS(mount, "%s, %s, %s"),
1097 		    fstypename, mntfromname, mntonname);
1098 #endif
1099 
1100 	if (strcmp(fstypename, "ext2") == 0) {
1101 		strcpy(fstypename, "ext2fs");
1102 	} else if (strcmp(fstypename, "proc") == 0) {
1103 		strcpy(fstypename, "linprocfs");
1104 	} else if (strcmp(fstypename, "vfat") == 0) {
1105 		strcpy(fstypename, "msdosfs");
1106 	}
1107 
1108 	fsflags = 0;
1109 
1110 	if ((args->rwflag & 0xffff0000) == 0xc0ed0000) {
1111 		/*
1112 		 * Linux SYNC flag is not included; the closest equivalent
1113 		 * FreeBSD has is !ASYNC, which is our default.
1114 		 */
1115 		if (args->rwflag & LINUX_MS_RDONLY)
1116 			fsflags |= MNT_RDONLY;
1117 		if (args->rwflag & LINUX_MS_NOSUID)
1118 			fsflags |= MNT_NOSUID;
1119 		if (args->rwflag & LINUX_MS_NOEXEC)
1120 			fsflags |= MNT_NOEXEC;
1121 		if (args->rwflag & LINUX_MS_REMOUNT)
1122 			fsflags |= MNT_UPDATE;
1123 	}
1124 
1125 	error = kernel_vmount(fsflags,
1126 	    "fstype", fstypename,
1127 	    "fspath", mntonname,
1128 	    "from", mntfromname,
1129 	    NULL);
1130 	return (error);
1131 }
1132 
1133 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1134 int
1135 linux_oldumount(struct thread *td, struct linux_oldumount_args *args)
1136 {
1137 	struct linux_umount_args args2;
1138 
1139 	args2.path = args->path;
1140 	args2.flags = 0;
1141 	return (linux_umount(td, &args2));
1142 }
1143 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1144 
1145 int
1146 linux_umount(struct thread *td, struct linux_umount_args *args)
1147 {
1148 	struct unmount_args bsd;
1149 
1150 	bsd.path = args->path;
1151 	bsd.flags = args->flags;	/* XXX correct? */
1152 	return (sys_unmount(td, &bsd));
1153 }
1154 
1155 /*
1156  * fcntl family of syscalls
1157  */
1158 
1159 struct l_flock {
1160 	l_short		l_type;
1161 	l_short		l_whence;
1162 	l_off_t		l_start;
1163 	l_off_t		l_len;
1164 	l_pid_t		l_pid;
1165 }
1166 #if defined(__amd64__) && defined(COMPAT_LINUX32)
1167 __packed
1168 #endif
1169 ;
1170 
1171 static void
1172 linux_to_bsd_flock(struct l_flock *linux_flock, struct flock *bsd_flock)
1173 {
1174 	switch (linux_flock->l_type) {
1175 	case LINUX_F_RDLCK:
1176 		bsd_flock->l_type = F_RDLCK;
1177 		break;
1178 	case LINUX_F_WRLCK:
1179 		bsd_flock->l_type = F_WRLCK;
1180 		break;
1181 	case LINUX_F_UNLCK:
1182 		bsd_flock->l_type = F_UNLCK;
1183 		break;
1184 	default:
1185 		bsd_flock->l_type = -1;
1186 		break;
1187 	}
1188 	bsd_flock->l_whence = linux_flock->l_whence;
1189 	bsd_flock->l_start = (off_t)linux_flock->l_start;
1190 	bsd_flock->l_len = (off_t)linux_flock->l_len;
1191 	bsd_flock->l_pid = (pid_t)linux_flock->l_pid;
1192 	bsd_flock->l_sysid = 0;
1193 }
1194 
1195 static void
1196 bsd_to_linux_flock(struct flock *bsd_flock, struct l_flock *linux_flock)
1197 {
1198 	switch (bsd_flock->l_type) {
1199 	case F_RDLCK:
1200 		linux_flock->l_type = LINUX_F_RDLCK;
1201 		break;
1202 	case F_WRLCK:
1203 		linux_flock->l_type = LINUX_F_WRLCK;
1204 		break;
1205 	case F_UNLCK:
1206 		linux_flock->l_type = LINUX_F_UNLCK;
1207 		break;
1208 	}
1209 	linux_flock->l_whence = bsd_flock->l_whence;
1210 	linux_flock->l_start = (l_off_t)bsd_flock->l_start;
1211 	linux_flock->l_len = (l_off_t)bsd_flock->l_len;
1212 	linux_flock->l_pid = (l_pid_t)bsd_flock->l_pid;
1213 }
1214 
1215 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1216 struct l_flock64 {
1217 	l_short		l_type;
1218 	l_short		l_whence;
1219 	l_loff_t	l_start;
1220 	l_loff_t	l_len;
1221 	l_pid_t		l_pid;
1222 }
1223 #if defined(__amd64__) && defined(COMPAT_LINUX32)
1224 __packed
1225 #endif
1226 ;
1227 
1228 static void
1229 linux_to_bsd_flock64(struct l_flock64 *linux_flock, struct flock *bsd_flock)
1230 {
1231 	switch (linux_flock->l_type) {
1232 	case LINUX_F_RDLCK:
1233 		bsd_flock->l_type = F_RDLCK;
1234 		break;
1235 	case LINUX_F_WRLCK:
1236 		bsd_flock->l_type = F_WRLCK;
1237 		break;
1238 	case LINUX_F_UNLCK:
1239 		bsd_flock->l_type = F_UNLCK;
1240 		break;
1241 	default:
1242 		bsd_flock->l_type = -1;
1243 		break;
1244 	}
1245 	bsd_flock->l_whence = linux_flock->l_whence;
1246 	bsd_flock->l_start = (off_t)linux_flock->l_start;
1247 	bsd_flock->l_len = (off_t)linux_flock->l_len;
1248 	bsd_flock->l_pid = (pid_t)linux_flock->l_pid;
1249 	bsd_flock->l_sysid = 0;
1250 }
1251 
1252 static void
1253 bsd_to_linux_flock64(struct flock *bsd_flock, struct l_flock64 *linux_flock)
1254 {
1255 	switch (bsd_flock->l_type) {
1256 	case F_RDLCK:
1257 		linux_flock->l_type = LINUX_F_RDLCK;
1258 		break;
1259 	case F_WRLCK:
1260 		linux_flock->l_type = LINUX_F_WRLCK;
1261 		break;
1262 	case F_UNLCK:
1263 		linux_flock->l_type = LINUX_F_UNLCK;
1264 		break;
1265 	}
1266 	linux_flock->l_whence = bsd_flock->l_whence;
1267 	linux_flock->l_start = (l_loff_t)bsd_flock->l_start;
1268 	linux_flock->l_len = (l_loff_t)bsd_flock->l_len;
1269 	linux_flock->l_pid = (l_pid_t)bsd_flock->l_pid;
1270 }
1271 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1272 
1273 static int
1274 fcntl_common(struct thread *td, struct linux_fcntl_args *args)
1275 {
1276 	struct l_flock linux_flock;
1277 	struct flock bsd_flock;
1278 	cap_rights_t rights;
1279 	struct file *fp;
1280 	long arg;
1281 	int error, result;
1282 
1283 	switch (args->cmd) {
1284 	case LINUX_F_DUPFD:
1285 		return (kern_fcntl(td, args->fd, F_DUPFD, args->arg));
1286 
1287 	case LINUX_F_GETFD:
1288 		return (kern_fcntl(td, args->fd, F_GETFD, 0));
1289 
1290 	case LINUX_F_SETFD:
1291 		return (kern_fcntl(td, args->fd, F_SETFD, args->arg));
1292 
1293 	case LINUX_F_GETFL:
1294 		error = kern_fcntl(td, args->fd, F_GETFL, 0);
1295 		result = td->td_retval[0];
1296 		td->td_retval[0] = 0;
1297 		if (result & O_RDONLY)
1298 			td->td_retval[0] |= LINUX_O_RDONLY;
1299 		if (result & O_WRONLY)
1300 			td->td_retval[0] |= LINUX_O_WRONLY;
1301 		if (result & O_RDWR)
1302 			td->td_retval[0] |= LINUX_O_RDWR;
1303 		if (result & O_NDELAY)
1304 			td->td_retval[0] |= LINUX_O_NONBLOCK;
1305 		if (result & O_APPEND)
1306 			td->td_retval[0] |= LINUX_O_APPEND;
1307 		if (result & O_FSYNC)
1308 			td->td_retval[0] |= LINUX_O_SYNC;
1309 		if (result & O_ASYNC)
1310 			td->td_retval[0] |= LINUX_FASYNC;
1311 #ifdef LINUX_O_NOFOLLOW
1312 		if (result & O_NOFOLLOW)
1313 			td->td_retval[0] |= LINUX_O_NOFOLLOW;
1314 #endif
1315 #ifdef LINUX_O_DIRECT
1316 		if (result & O_DIRECT)
1317 			td->td_retval[0] |= LINUX_O_DIRECT;
1318 #endif
1319 		return (error);
1320 
1321 	case LINUX_F_SETFL:
1322 		arg = 0;
1323 		if (args->arg & LINUX_O_NDELAY)
1324 			arg |= O_NONBLOCK;
1325 		if (args->arg & LINUX_O_APPEND)
1326 			arg |= O_APPEND;
1327 		if (args->arg & LINUX_O_SYNC)
1328 			arg |= O_FSYNC;
1329 		if (args->arg & LINUX_FASYNC)
1330 			arg |= O_ASYNC;
1331 #ifdef LINUX_O_NOFOLLOW
1332 		if (args->arg & LINUX_O_NOFOLLOW)
1333 			arg |= O_NOFOLLOW;
1334 #endif
1335 #ifdef LINUX_O_DIRECT
1336 		if (args->arg & LINUX_O_DIRECT)
1337 			arg |= O_DIRECT;
1338 #endif
1339 		return (kern_fcntl(td, args->fd, F_SETFL, arg));
1340 
1341 	case LINUX_F_GETLK:
1342 		error = copyin((void *)args->arg, &linux_flock,
1343 		    sizeof(linux_flock));
1344 		if (error)
1345 			return (error);
1346 		linux_to_bsd_flock(&linux_flock, &bsd_flock);
1347 		error = kern_fcntl(td, args->fd, F_GETLK, (intptr_t)&bsd_flock);
1348 		if (error)
1349 			return (error);
1350 		bsd_to_linux_flock(&bsd_flock, &linux_flock);
1351 		return (copyout(&linux_flock, (void *)args->arg,
1352 		    sizeof(linux_flock)));
1353 
1354 	case LINUX_F_SETLK:
1355 		error = copyin((void *)args->arg, &linux_flock,
1356 		    sizeof(linux_flock));
1357 		if (error)
1358 			return (error);
1359 		linux_to_bsd_flock(&linux_flock, &bsd_flock);
1360 		return (kern_fcntl(td, args->fd, F_SETLK,
1361 		    (intptr_t)&bsd_flock));
1362 
1363 	case LINUX_F_SETLKW:
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 		return (kern_fcntl(td, args->fd, F_SETLKW,
1370 		     (intptr_t)&bsd_flock));
1371 
1372 	case LINUX_F_GETOWN:
1373 		return (kern_fcntl(td, args->fd, F_GETOWN, 0));
1374 
1375 	case LINUX_F_SETOWN:
1376 		/*
1377 		 * XXX some Linux applications depend on F_SETOWN having no
1378 		 * significant effect for pipes (SIGIO is not delivered for
1379 		 * pipes under Linux-2.2.35 at least).
1380 		 */
1381 		error = fget(td, args->fd,
1382 		    cap_rights_init(&rights, CAP_FCNTL), &fp);
1383 		if (error)
1384 			return (error);
1385 		if (fp->f_type == DTYPE_PIPE) {
1386 			fdrop(fp, td);
1387 			return (EINVAL);
1388 		}
1389 		fdrop(fp, td);
1390 
1391 		return (kern_fcntl(td, args->fd, F_SETOWN, args->arg));
1392 
1393 	case LINUX_F_DUPFD_CLOEXEC:
1394 		return (kern_fcntl(td, args->fd, F_DUPFD_CLOEXEC, args->arg));
1395 	}
1396 
1397 	return (EINVAL);
1398 }
1399 
1400 int
1401 linux_fcntl(struct thread *td, struct linux_fcntl_args *args)
1402 {
1403 
1404 #ifdef DEBUG
1405 	if (ldebug(fcntl))
1406 		printf(ARGS(fcntl, "%d, %08x, *"), args->fd, args->cmd);
1407 #endif
1408 
1409 	return (fcntl_common(td, args));
1410 }
1411 
1412 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1413 int
1414 linux_fcntl64(struct thread *td, struct linux_fcntl64_args *args)
1415 {
1416 	struct l_flock64 linux_flock;
1417 	struct flock bsd_flock;
1418 	struct linux_fcntl_args fcntl_args;
1419 	int error;
1420 
1421 #ifdef DEBUG
1422 	if (ldebug(fcntl64))
1423 		printf(ARGS(fcntl64, "%d, %08x, *"), args->fd, args->cmd);
1424 #endif
1425 
1426 	switch (args->cmd) {
1427 	case LINUX_F_GETLK64:
1428 		error = copyin((void *)args->arg, &linux_flock,
1429 		    sizeof(linux_flock));
1430 		if (error)
1431 			return (error);
1432 		linux_to_bsd_flock64(&linux_flock, &bsd_flock);
1433 		error = kern_fcntl(td, args->fd, F_GETLK, (intptr_t)&bsd_flock);
1434 		if (error)
1435 			return (error);
1436 		bsd_to_linux_flock64(&bsd_flock, &linux_flock);
1437 		return (copyout(&linux_flock, (void *)args->arg,
1438 			    sizeof(linux_flock)));
1439 
1440 	case LINUX_F_SETLK64:
1441 		error = copyin((void *)args->arg, &linux_flock,
1442 		    sizeof(linux_flock));
1443 		if (error)
1444 			return (error);
1445 		linux_to_bsd_flock64(&linux_flock, &bsd_flock);
1446 		return (kern_fcntl(td, args->fd, F_SETLK,
1447 		    (intptr_t)&bsd_flock));
1448 
1449 	case LINUX_F_SETLKW64:
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 		return (kern_fcntl(td, args->fd, F_SETLKW,
1456 		    (intptr_t)&bsd_flock));
1457 	}
1458 
1459 	fcntl_args.fd = args->fd;
1460 	fcntl_args.cmd = args->cmd;
1461 	fcntl_args.arg = args->arg;
1462 	return (fcntl_common(td, &fcntl_args));
1463 }
1464 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1465 
1466 int
1467 linux_chown(struct thread *td, struct linux_chown_args *args)
1468 {
1469 	char *path;
1470 	int error;
1471 
1472 	LCONVPATHEXIST(td, args->path, &path);
1473 
1474 #ifdef DEBUG
1475 	if (ldebug(chown))
1476 		printf(ARGS(chown, "%s, %d, %d"), path, args->uid, args->gid);
1477 #endif
1478 	error = kern_fchownat(td, AT_FDCWD, path, UIO_SYSSPACE, args->uid,
1479 	    args->gid, 0);
1480 	LFREEPATH(path);
1481 	return (error);
1482 }
1483 
1484 int
1485 linux_fchownat(struct thread *td, struct linux_fchownat_args *args)
1486 {
1487 	char *path;
1488 	int error, dfd, flag;
1489 
1490 	if (args->flag & ~LINUX_AT_SYMLINK_NOFOLLOW)
1491 		return (EINVAL);
1492 
1493 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD :  args->dfd;
1494 	LCONVPATHEXIST_AT(td, args->filename, &path, dfd);
1495 
1496 #ifdef DEBUG
1497 	if (ldebug(fchownat))
1498 		printf(ARGS(fchownat, "%s, %d, %d"), path, args->uid, args->gid);
1499 #endif
1500 
1501 	flag = (args->flag & LINUX_AT_SYMLINK_NOFOLLOW) == 0 ? 0 :
1502 	    AT_SYMLINK_NOFOLLOW;
1503 	error = kern_fchownat(td, dfd, path, UIO_SYSSPACE, args->uid, args->gid,
1504 	    flag);
1505 	LFREEPATH(path);
1506 	return (error);
1507 }
1508 
1509 int
1510 linux_lchown(struct thread *td, struct linux_lchown_args *args)
1511 {
1512 	char *path;
1513 	int error;
1514 
1515 	LCONVPATHEXIST(td, args->path, &path);
1516 
1517 #ifdef DEBUG
1518 	if (ldebug(lchown))
1519 		printf(ARGS(lchown, "%s, %d, %d"), path, args->uid, args->gid);
1520 #endif
1521 	error = kern_fchownat(td, AT_FDCWD, path, UIO_SYSSPACE, args->uid,
1522 	    args->gid, AT_SYMLINK_NOFOLLOW);
1523 	LFREEPATH(path);
1524 	return (error);
1525 }
1526 
1527 static int
1528 convert_fadvice(int advice)
1529 {
1530 	switch (advice) {
1531 	case LINUX_POSIX_FADV_NORMAL:
1532 		return (POSIX_FADV_NORMAL);
1533 	case LINUX_POSIX_FADV_RANDOM:
1534 		return (POSIX_FADV_RANDOM);
1535 	case LINUX_POSIX_FADV_SEQUENTIAL:
1536 		return (POSIX_FADV_SEQUENTIAL);
1537 	case LINUX_POSIX_FADV_WILLNEED:
1538 		return (POSIX_FADV_WILLNEED);
1539 	case LINUX_POSIX_FADV_DONTNEED:
1540 		return (POSIX_FADV_DONTNEED);
1541 	case LINUX_POSIX_FADV_NOREUSE:
1542 		return (POSIX_FADV_NOREUSE);
1543 	default:
1544 		return (-1);
1545 	}
1546 }
1547 
1548 int
1549 linux_fadvise64(struct thread *td, struct linux_fadvise64_args *args)
1550 {
1551 	int advice;
1552 
1553 	advice = convert_fadvice(args->advice);
1554 	if (advice == -1)
1555 		return (EINVAL);
1556 	return (kern_posix_fadvise(td, args->fd, args->offset, args->len,
1557 	    advice));
1558 }
1559 
1560 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1561 int
1562 linux_fadvise64_64(struct thread *td, struct linux_fadvise64_64_args *args)
1563 {
1564 	int advice;
1565 
1566 	advice = convert_fadvice(args->advice);
1567 	if (advice == -1)
1568 		return (EINVAL);
1569 	return (kern_posix_fadvise(td, args->fd, args->offset, args->len,
1570 	    advice));
1571 }
1572 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1573 
1574 int
1575 linux_pipe(struct thread *td, struct linux_pipe_args *args)
1576 {
1577 	int fildes[2];
1578 	int error;
1579 
1580 #ifdef DEBUG
1581 	if (ldebug(pipe))
1582 		printf(ARGS(pipe, "*"));
1583 #endif
1584 
1585 	error = kern_pipe(td, fildes, 0, NULL, NULL);
1586 	if (error)
1587 		return (error);
1588 
1589 	/* XXX: Close descriptors on error. */
1590 	return (copyout(fildes, args->pipefds, sizeof(fildes)));
1591 }
1592 
1593 int
1594 linux_pipe2(struct thread *td, struct linux_pipe2_args *args)
1595 {
1596 	int fildes[2];
1597 	int error, flags;
1598 
1599 #ifdef DEBUG
1600 	if (ldebug(pipe2))
1601 		printf(ARGS(pipe2, "*, %d"), args->flags);
1602 #endif
1603 
1604 	if ((args->flags & ~(LINUX_O_NONBLOCK | LINUX_O_CLOEXEC)) != 0)
1605 		return (EINVAL);
1606 
1607 	flags = 0;
1608 	if ((args->flags & LINUX_O_NONBLOCK) != 0)
1609 		flags |= O_NONBLOCK;
1610 	if ((args->flags & LINUX_O_CLOEXEC) != 0)
1611 		flags |= O_CLOEXEC;
1612 	error = kern_pipe(td, fildes, flags, NULL, NULL);
1613 	if (error)
1614 		return (error);
1615 
1616 	/* XXX: Close descriptors on error. */
1617 	return (copyout(fildes, args->pipefds, sizeof(fildes)));
1618 }
1619 
1620 int
1621 linux_dup3(struct thread *td, struct linux_dup3_args *args)
1622 {
1623 	int cmd;
1624 	intptr_t newfd;
1625 
1626 	if (args->oldfd == args->newfd)
1627 		return (EINVAL);
1628 	if ((args->flags & ~LINUX_O_CLOEXEC) != 0)
1629 		return (EINVAL);
1630 	if (args->flags & LINUX_O_CLOEXEC)
1631 		cmd = F_DUP2FD_CLOEXEC;
1632 	else
1633 		cmd = F_DUP2FD;
1634 
1635 	newfd = args->newfd;
1636 	return (kern_fcntl(td, args->oldfd, cmd, newfd));
1637 }
1638 
1639 int
1640 linux_fallocate(struct thread *td, struct linux_fallocate_args *args)
1641 {
1642 
1643 	/*
1644 	 * We emulate only posix_fallocate system call for which
1645 	 * mode should be 0.
1646 	 */
1647 	if (args->mode != 0)
1648 		return (ENOSYS);
1649 
1650 	return (kern_posix_fallocate(td, args->fd, args->offset,
1651 	    args->len));
1652 }
1653