xref: /freebsd/sys/compat/linux/linux_file.c (revision 9a41df2a0e6408e9b329bbd8b9e37c2b44461a1b)
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/capability.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 #include <ufs/ufs/extattr.h>
59 #include <ufs/ufs/quota.h>
60 #include <ufs/ufs/ufsmount.h>
61 
62 #ifdef COMPAT_LINUX32
63 #include <machine/../linux32/linux.h>
64 #include <machine/../linux32/linux32_proto.h>
65 #else
66 #include <machine/../linux/linux.h>
67 #include <machine/../linux/linux_proto.h>
68 #endif
69 #include <compat/linux/linux_util.h>
70 #include <compat/linux/linux_file.h>
71 
72 /* XXX */
73 int	do_pipe(struct thread *td, int fildes[2], int flags);
74 
75 int
76 linux_creat(struct thread *td, struct linux_creat_args *args)
77 {
78     char *path;
79     int error;
80 
81     LCONVPATHEXIST(td, args->path, &path);
82 
83 #ifdef DEBUG
84 	if (ldebug(creat))
85 		printf(ARGS(creat, "%s, %d"), path, args->mode);
86 #endif
87     error = kern_open(td, path, UIO_SYSSPACE, O_WRONLY | O_CREAT | O_TRUNC,
88 	args->mode);
89     LFREEPATH(path);
90     return (error);
91 }
92 
93 
94 static int
95 linux_common_open(struct thread *td, int dirfd, char *path, int l_flags, int mode)
96 {
97     struct proc *p = td->td_proc;
98     struct file *fp;
99     int fd;
100     int bsd_flags, error;
101 
102     bsd_flags = 0;
103     switch (l_flags & LINUX_O_ACCMODE) {
104     case LINUX_O_WRONLY:
105 	bsd_flags |= O_WRONLY;
106 	break;
107     case LINUX_O_RDWR:
108 	bsd_flags |= O_RDWR;
109 	break;
110     default:
111 	bsd_flags |= O_RDONLY;
112     }
113     if (l_flags & LINUX_O_NDELAY)
114 	bsd_flags |= O_NONBLOCK;
115     if (l_flags & LINUX_O_APPEND)
116 	bsd_flags |= O_APPEND;
117     if (l_flags & LINUX_O_SYNC)
118 	bsd_flags |= O_FSYNC;
119     if (l_flags & LINUX_O_NONBLOCK)
120 	bsd_flags |= O_NONBLOCK;
121     if (l_flags & LINUX_FASYNC)
122 	bsd_flags |= O_ASYNC;
123     if (l_flags & LINUX_O_CREAT)
124 	bsd_flags |= O_CREAT;
125     if (l_flags & LINUX_O_TRUNC)
126 	bsd_flags |= O_TRUNC;
127     if (l_flags & LINUX_O_EXCL)
128 	bsd_flags |= O_EXCL;
129     if (l_flags & LINUX_O_NOCTTY)
130 	bsd_flags |= O_NOCTTY;
131     if (l_flags & LINUX_O_DIRECT)
132 	bsd_flags |= O_DIRECT;
133     if (l_flags & LINUX_O_NOFOLLOW)
134 	bsd_flags |= O_NOFOLLOW;
135     if (l_flags & LINUX_O_DIRECTORY)
136 	bsd_flags |= O_DIRECTORY;
137     /* XXX LINUX_O_NOATIME: unable to be easily implemented. */
138 
139     error = kern_openat(td, dirfd, path, UIO_SYSSPACE, bsd_flags, mode);
140 
141     if (!error) {
142 	    fd = td->td_retval[0];
143 	    /*
144 	     * XXX In between kern_open() and fget(), another process
145 	     * having the same filedesc could use that fd without
146 	     * checking below.
147 	     */
148 	    error = fget(td, fd, CAP_IOCTL, &fp);
149 	    if (!error) {
150 		    sx_slock(&proctree_lock);
151 		    PROC_LOCK(p);
152 		    if (!(bsd_flags & O_NOCTTY) &&
153 			SESS_LEADER(p) && !(p->p_flag & P_CONTROLT)) {
154 			    PROC_UNLOCK(p);
155 			    sx_unlock(&proctree_lock);
156 			    if (fp->f_type == DTYPE_VNODE)
157 				    (void) fo_ioctl(fp, TIOCSCTTY, (caddr_t) 0,
158 					     td->td_ucred, td);
159 		    } else {
160 			    PROC_UNLOCK(p);
161 			    sx_sunlock(&proctree_lock);
162 		    }
163 		    fdrop(fp, td);
164 		    /*
165 		     * XXX as above, fdrop()/kern_close() pair is racy.
166 		     */
167 		    if (error)
168 			    kern_close(td, fd);
169 	    }
170     }
171 
172 #ifdef DEBUG
173     if (ldebug(open))
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 int
200 linux_open(struct thread *td, struct linux_open_args *args)
201 {
202     char *path;
203 
204     if (args->flags & LINUX_O_CREAT)
205 	LCONVPATHCREAT(td, args->path, &path);
206     else
207 	LCONVPATHEXIST(td, args->path, &path);
208 
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 
215 	return (linux_common_open(td, AT_FDCWD, path, args->flags, args->mode));
216 }
217 
218 int
219 linux_lseek(struct thread *td, struct linux_lseek_args *args)
220 {
221 
222     struct lseek_args /* {
223 	int fd;
224 	int pad;
225 	off_t offset;
226 	int whence;
227     } */ tmp_args;
228     int error;
229 
230 #ifdef DEBUG
231 	if (ldebug(lseek))
232 		printf(ARGS(lseek, "%d, %ld, %d"),
233 		    args->fdes, (long)args->off, args->whence);
234 #endif
235     tmp_args.fd = args->fdes;
236     tmp_args.offset = (off_t)args->off;
237     tmp_args.whence = args->whence;
238     error = sys_lseek(td, &tmp_args);
239     return error;
240 }
241 
242 int
243 linux_llseek(struct thread *td, struct linux_llseek_args *args)
244 {
245 	struct lseek_args bsd_args;
246 	int error;
247 	off_t off;
248 
249 #ifdef DEBUG
250 	if (ldebug(llseek))
251 		printf(ARGS(llseek, "%d, %d:%d, %d"),
252 		    args->fd, args->ohigh, args->olow, args->whence);
253 #endif
254 	off = (args->olow) | (((off_t) args->ohigh) << 32);
255 
256 	bsd_args.fd = args->fd;
257 	bsd_args.offset = off;
258 	bsd_args.whence = args->whence;
259 
260 	if ((error = sys_lseek(td, &bsd_args)))
261 		return error;
262 
263 	if ((error = copyout(td->td_retval, args->res, sizeof (off_t))))
264 		return error;
265 
266 	td->td_retval[0] = 0;
267 	return 0;
268 }
269 
270 int
271 linux_readdir(struct thread *td, struct linux_readdir_args *args)
272 {
273 	struct linux_getdents_args lda;
274 
275 	lda.fd = args->fd;
276 	lda.dent = args->dent;
277 	lda.count = 1;
278 	return linux_getdents(td, &lda);
279 }
280 
281 /*
282  * Note that linux_getdents(2) and linux_getdents64(2) have the same
283  * arguments. They only differ in the definition of struct dirent they
284  * operate on. We use this to common the code, with the exception of
285  * accessing struct dirent. Note that linux_readdir(2) is implemented
286  * by means of linux_getdents(2). In this case we never operate on
287  * struct dirent64 and thus don't need to handle it...
288  */
289 
290 struct l_dirent {
291 	l_ulong		d_ino;
292 	l_off_t		d_off;
293 	l_ushort	d_reclen;
294 	char		d_name[LINUX_NAME_MAX + 1];
295 };
296 
297 struct l_dirent64 {
298 	uint64_t	d_ino;
299 	int64_t		d_off;
300 	l_ushort	d_reclen;
301 	u_char		d_type;
302 	char		d_name[LINUX_NAME_MAX + 1];
303 };
304 
305 /*
306  * Linux uses the last byte in the dirent buffer to store d_type,
307  * at least glibc-2.7 requires it. That is why l_dirent is padded with 2 bytes.
308  */
309 #define LINUX_RECLEN(namlen)						\
310     roundup((offsetof(struct l_dirent, d_name) + (namlen) + 2),		\
311     sizeof(l_ulong))
312 
313 #define LINUX_RECLEN64(namlen)						\
314     roundup((offsetof(struct l_dirent64, d_name) + (namlen) + 1),	\
315     sizeof(uint64_t))
316 
317 #define LINUX_MAXRECLEN		max(LINUX_RECLEN(LINUX_NAME_MAX),	\
318 				    LINUX_RECLEN64(LINUX_NAME_MAX))
319 #define	LINUX_DIRBLKSIZ		512
320 
321 static int
322 getdents_common(struct thread *td, struct linux_getdents64_args *args,
323     int is64bit)
324 {
325 	struct dirent *bdp;
326 	struct vnode *vp;
327 	caddr_t inp, buf;		/* BSD-format */
328 	int len, reclen;		/* BSD-format */
329 	caddr_t outp;			/* Linux-format */
330 	int resid, linuxreclen=0;	/* Linux-format */
331 	caddr_t lbuf;			/* Linux-format */
332 	struct file *fp;
333 	struct uio auio;
334 	struct iovec aiov;
335 	off_t off;
336 	struct l_dirent *linux_dirent;
337 	struct l_dirent64 *linux_dirent64;
338 	int buflen, error, eofflag, nbytes, justone;
339 	u_long *cookies = NULL, *cookiep;
340 	int ncookies, vfslocked;
341 
342 	nbytes = args->count;
343 	if (nbytes == 1) {
344 		/* readdir(2) case. Always struct dirent. */
345 		if (is64bit)
346 			return (EINVAL);
347 		nbytes = sizeof(*linux_dirent);
348 		justone = 1;
349 	} else
350 		justone = 0;
351 
352 	if ((error = getvnode(td->td_proc->p_fd, args->fd, CAP_READ, &fp)) != 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 	vfslocked = VFS_LOCK_GIANT(vp->v_mount);
363 	if (vp->v_type != VDIR) {
364 		VFS_UNLOCK_GIANT(vfslocked);
365 		foffset_unlock(fp, off, 0);
366 		fdrop(fp, td);
367 		return (EINVAL);
368 	}
369 
370 
371 	buflen = max(LINUX_DIRBLKSIZ, nbytes);
372 	buflen = min(buflen, MAXBSIZE);
373 	buf = malloc(buflen, M_TEMP, M_WAITOK);
374 	lbuf = malloc(LINUX_MAXRECLEN, M_TEMP, M_WAITOK | M_ZERO);
375 	vn_lock(vp, LK_SHARED | LK_RETRY);
376 
377 	aiov.iov_base = buf;
378 	aiov.iov_len = buflen;
379 	auio.uio_iov = &aiov;
380 	auio.uio_iovcnt = 1;
381 	auio.uio_rw = UIO_READ;
382 	auio.uio_segflg = UIO_SYSSPACE;
383 	auio.uio_td = td;
384 	auio.uio_resid = buflen;
385 	auio.uio_offset = off;
386 
387 #ifdef MAC
388 	/*
389 	 * Do directory search MAC check using non-cached credentials.
390 	 */
391 	if ((error = mac_vnode_check_readdir(td->td_ucred, vp)))
392 		goto out;
393 #endif /* MAC */
394 	if ((error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, &ncookies,
395 		 &cookies)))
396 		goto out;
397 
398 	inp = buf;
399 	outp = (caddr_t)args->dirent;
400 	resid = nbytes;
401 	if ((len = buflen - auio.uio_resid) <= 0)
402 		goto eof;
403 
404 	cookiep = cookies;
405 
406 	if (cookies) {
407 		/*
408 		 * When using cookies, the vfs has the option of reading from
409 		 * a different offset than that supplied (UFS truncates the
410 		 * offset to a block boundary to make sure that it never reads
411 		 * partway through a directory entry, even if the directory
412 		 * has been compacted).
413 		 */
414 		while (len > 0 && ncookies > 0 && *cookiep <= off) {
415 			bdp = (struct dirent *) inp;
416 			len -= bdp->d_reclen;
417 			inp += bdp->d_reclen;
418 			cookiep++;
419 			ncookies--;
420 		}
421 	}
422 
423 	while (len > 0) {
424 		if (cookiep && ncookies == 0)
425 			break;
426 		bdp = (struct dirent *) inp;
427 		reclen = bdp->d_reclen;
428 		if (reclen & 3) {
429 			error = EFAULT;
430 			goto out;
431 		}
432 
433 		if (bdp->d_fileno == 0) {
434 			inp += reclen;
435 			if (cookiep) {
436 				off = *cookiep++;
437 				ncookies--;
438 			} else
439 				off += reclen;
440 
441 			len -= reclen;
442 			continue;
443 		}
444 
445 		linuxreclen = (is64bit)
446 		    ? LINUX_RECLEN64(bdp->d_namlen)
447 		    : LINUX_RECLEN(bdp->d_namlen);
448 
449 		if (reclen > len || resid < linuxreclen) {
450 			outp++;
451 			break;
452 		}
453 
454 		if (justone) {
455 			/* readdir(2) case. */
456 			linux_dirent = (struct l_dirent*)lbuf;
457 			linux_dirent->d_ino = bdp->d_fileno;
458 			linux_dirent->d_off = (l_off_t)linuxreclen;
459 			linux_dirent->d_reclen = (l_ushort)bdp->d_namlen;
460 			strlcpy(linux_dirent->d_name, bdp->d_name,
461 			    linuxreclen - offsetof(struct l_dirent, d_name));
462 			error = copyout(linux_dirent, outp, linuxreclen);
463 		}
464 		if (is64bit) {
465 			linux_dirent64 = (struct l_dirent64*)lbuf;
466 			linux_dirent64->d_ino = bdp->d_fileno;
467 			linux_dirent64->d_off = (cookiep)
468 			    ? (l_off_t)*cookiep
469 			    : (l_off_t)(off + reclen);
470 			linux_dirent64->d_reclen = (l_ushort)linuxreclen;
471 			linux_dirent64->d_type = bdp->d_type;
472 			strlcpy(linux_dirent64->d_name, bdp->d_name,
473 			    linuxreclen - offsetof(struct l_dirent64, d_name));
474 			error = copyout(linux_dirent64, outp, linuxreclen);
475 		} else if (!justone) {
476 			linux_dirent = (struct l_dirent*)lbuf;
477 			linux_dirent->d_ino = bdp->d_fileno;
478 			linux_dirent->d_off = (cookiep)
479 			    ? (l_off_t)*cookiep
480 			    : (l_off_t)(off + reclen);
481 			linux_dirent->d_reclen = (l_ushort)linuxreclen;
482 			/*
483 			 * Copy d_type to last byte of l_dirent buffer
484 			 */
485 			lbuf[linuxreclen-1] = bdp->d_type;
486 			strlcpy(linux_dirent->d_name, bdp->d_name,
487 			    linuxreclen - offsetof(struct l_dirent, d_name)-1);
488 			error = copyout(linux_dirent, outp, linuxreclen);
489 		}
490 
491 		if (error)
492 			goto out;
493 
494 		inp += reclen;
495 		if (cookiep) {
496 			off = *cookiep++;
497 			ncookies--;
498 		} else
499 			off += reclen;
500 
501 		outp += linuxreclen;
502 		resid -= linuxreclen;
503 		len -= reclen;
504 		if (justone)
505 			break;
506 	}
507 
508 	if (outp == (caddr_t)args->dirent) {
509 		nbytes = resid;
510 		goto eof;
511 	}
512 
513 	if (justone)
514 		nbytes = resid + linuxreclen;
515 
516 eof:
517 	td->td_retval[0] = nbytes - resid;
518 
519 out:
520 	if (cookies)
521 		free(cookies, M_TEMP);
522 
523 	VOP_UNLOCK(vp, 0);
524 	VFS_UNLOCK_GIANT(vfslocked);
525 	foffset_unlock(fp, off, 0);
526 	fdrop(fp, td);
527 	free(buf, M_TEMP);
528 	free(lbuf, M_TEMP);
529 	return (error);
530 }
531 
532 int
533 linux_getdents(struct thread *td, struct linux_getdents_args *args)
534 {
535 
536 #ifdef DEBUG
537 	if (ldebug(getdents))
538 		printf(ARGS(getdents, "%d, *, %d"), args->fd, args->count);
539 #endif
540 
541 	return (getdents_common(td, (struct linux_getdents64_args*)args, 0));
542 }
543 
544 int
545 linux_getdents64(struct thread *td, struct linux_getdents64_args *args)
546 {
547 
548 #ifdef DEBUG
549 	if (ldebug(getdents64))
550 		printf(ARGS(getdents64, "%d, *, %d"), args->fd, args->count);
551 #endif
552 
553 	return (getdents_common(td, args, 1));
554 }
555 
556 /*
557  * These exist mainly for hooks for doing /compat/linux translation.
558  */
559 
560 int
561 linux_access(struct thread *td, struct linux_access_args *args)
562 {
563 	char *path;
564 	int error;
565 
566 	/* linux convention */
567 	if (args->amode & ~(F_OK | X_OK | W_OK | R_OK))
568 		return (EINVAL);
569 
570 	LCONVPATHEXIST(td, args->path, &path);
571 
572 #ifdef DEBUG
573 	if (ldebug(access))
574 		printf(ARGS(access, "%s, %d"), path, args->amode);
575 #endif
576 	error = kern_access(td, path, UIO_SYSSPACE, args->amode);
577 	LFREEPATH(path);
578 
579 	return (error);
580 }
581 
582 int
583 linux_faccessat(struct thread *td, struct linux_faccessat_args *args)
584 {
585 	char *path;
586 	int error, dfd, flag;
587 
588 	if (args->flag & ~LINUX_AT_EACCESS)
589 		return (EINVAL);
590 	/* linux convention */
591 	if (args->amode & ~(F_OK | X_OK | W_OK | R_OK))
592 		return (EINVAL);
593 
594 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
595 	LCONVPATHEXIST_AT(td, args->filename, &path, dfd);
596 
597 #ifdef DEBUG
598 	if (ldebug(access))
599 		printf(ARGS(access, "%s, %d"), path, args->amode);
600 #endif
601 
602 	flag = (args->flag & LINUX_AT_EACCESS) == 0 ? 0 : AT_EACCESS;
603 	error = kern_accessat(td, dfd, path, UIO_SYSSPACE, flag, args->amode);
604 	LFREEPATH(path);
605 
606 	return (error);
607 }
608 
609 int
610 linux_unlink(struct thread *td, struct linux_unlink_args *args)
611 {
612 	char *path;
613 	int error;
614 	struct stat st;
615 
616 	LCONVPATHEXIST(td, args->path, &path);
617 
618 #ifdef DEBUG
619 	if (ldebug(unlink))
620 		printf(ARGS(unlink, "%s"), path);
621 #endif
622 
623 	error = kern_unlink(td, path, UIO_SYSSPACE);
624 	if (error == EPERM)
625 		/* Introduce POSIX noncompliant behaviour of Linux */
626 		if (kern_stat(td, path, UIO_SYSSPACE, &st) == 0)
627 			if (S_ISDIR(st.st_mode))
628 				error = EISDIR;
629 	LFREEPATH(path);
630 	return (error);
631 }
632 
633 int
634 linux_unlinkat(struct thread *td, struct linux_unlinkat_args *args)
635 {
636 	char *path;
637 	int error, dfd;
638 	struct stat st;
639 
640 	if (args->flag & ~LINUX_AT_REMOVEDIR)
641 		return (EINVAL);
642 
643 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
644 	LCONVPATHEXIST_AT(td, args->pathname, &path, dfd);
645 
646 #ifdef DEBUG
647 	if (ldebug(unlinkat))
648 		printf(ARGS(unlinkat, "%s"), path);
649 #endif
650 
651 	if (args->flag & LINUX_AT_REMOVEDIR)
652 		error = kern_rmdirat(td, dfd, path, UIO_SYSSPACE);
653 	else
654 		error = kern_unlinkat(td, dfd, path, UIO_SYSSPACE, 0);
655 	if (error == EPERM && !(args->flag & LINUX_AT_REMOVEDIR)) {
656 		/* Introduce POSIX noncompliant behaviour of Linux */
657 		if (kern_statat(td, AT_SYMLINK_NOFOLLOW, dfd, path,
658 		    UIO_SYSSPACE, &st) == 0 && S_ISDIR(st.st_mode))
659 			error = EISDIR;
660 	}
661 	LFREEPATH(path);
662 	return (error);
663 }
664 int
665 linux_chdir(struct thread *td, struct linux_chdir_args *args)
666 {
667 	char *path;
668 	int error;
669 
670 	LCONVPATHEXIST(td, args->path, &path);
671 
672 #ifdef DEBUG
673 	if (ldebug(chdir))
674 		printf(ARGS(chdir, "%s"), path);
675 #endif
676 	error = kern_chdir(td, path, UIO_SYSSPACE);
677 	LFREEPATH(path);
678 	return (error);
679 }
680 
681 int
682 linux_chmod(struct thread *td, struct linux_chmod_args *args)
683 {
684 	char *path;
685 	int error;
686 
687 	LCONVPATHEXIST(td, args->path, &path);
688 
689 #ifdef DEBUG
690 	if (ldebug(chmod))
691 		printf(ARGS(chmod, "%s, %d"), path, args->mode);
692 #endif
693 	error = kern_chmod(td, path, UIO_SYSSPACE, args->mode);
694 	LFREEPATH(path);
695 	return (error);
696 }
697 
698 int
699 linux_fchmodat(struct thread *td, struct linux_fchmodat_args *args)
700 {
701 	char *path;
702 	int error, dfd;
703 
704 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
705 	LCONVPATHEXIST_AT(td, args->filename, &path, dfd);
706 
707 #ifdef DEBUG
708 	if (ldebug(fchmodat))
709 		printf(ARGS(fchmodat, "%s, %d"), path, args->mode);
710 #endif
711 
712 	error = kern_fchmodat(td, dfd, path, UIO_SYSSPACE, args->mode, 0);
713 	LFREEPATH(path);
714 	return (error);
715 }
716 
717 int
718 linux_mkdir(struct thread *td, struct linux_mkdir_args *args)
719 {
720 	char *path;
721 	int error;
722 
723 	LCONVPATHCREAT(td, args->path, &path);
724 
725 #ifdef DEBUG
726 	if (ldebug(mkdir))
727 		printf(ARGS(mkdir, "%s, %d"), path, args->mode);
728 #endif
729 	error = kern_mkdir(td, path, UIO_SYSSPACE, args->mode);
730 	LFREEPATH(path);
731 	return (error);
732 }
733 
734 int
735 linux_mkdirat(struct thread *td, struct linux_mkdirat_args *args)
736 {
737 	char *path;
738 	int error, dfd;
739 
740 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
741 	LCONVPATHCREAT_AT(td, args->pathname, &path, dfd);
742 
743 #ifdef DEBUG
744 	if (ldebug(mkdirat))
745 		printf(ARGS(mkdirat, "%s, %d"), path, args->mode);
746 #endif
747 	error = kern_mkdirat(td, dfd, path, UIO_SYSSPACE, args->mode);
748 	LFREEPATH(path);
749 	return (error);
750 }
751 
752 int
753 linux_rmdir(struct thread *td, struct linux_rmdir_args *args)
754 {
755 	char *path;
756 	int error;
757 
758 	LCONVPATHEXIST(td, args->path, &path);
759 
760 #ifdef DEBUG
761 	if (ldebug(rmdir))
762 		printf(ARGS(rmdir, "%s"), path);
763 #endif
764 	error = kern_rmdir(td, path, UIO_SYSSPACE);
765 	LFREEPATH(path);
766 	return (error);
767 }
768 
769 int
770 linux_rename(struct thread *td, struct linux_rename_args *args)
771 {
772 	char *from, *to;
773 	int error;
774 
775 	LCONVPATHEXIST(td, args->from, &from);
776 	/* Expand LCONVPATHCREATE so that `from' can be freed on errors */
777 	error = linux_emul_convpath(td, args->to, UIO_USERSPACE, &to, 1, AT_FDCWD);
778 	if (to == NULL) {
779 		LFREEPATH(from);
780 		return (error);
781 	}
782 
783 #ifdef DEBUG
784 	if (ldebug(rename))
785 		printf(ARGS(rename, "%s, %s"), from, to);
786 #endif
787 	error = kern_rename(td, from, to, UIO_SYSSPACE);
788 	LFREEPATH(from);
789 	LFREEPATH(to);
790 	return (error);
791 }
792 
793 int
794 linux_renameat(struct thread *td, struct linux_renameat_args *args)
795 {
796 	char *from, *to;
797 	int error, olddfd, newdfd;
798 
799 	olddfd = (args->olddfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->olddfd;
800 	newdfd = (args->newdfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->newdfd;
801 	LCONVPATHEXIST_AT(td, args->oldname, &from, olddfd);
802 	/* Expand LCONVPATHCREATE so that `from' can be freed on errors */
803 	error = linux_emul_convpath(td, args->newname, UIO_USERSPACE, &to, 1, newdfd);
804 	if (to == NULL) {
805 		LFREEPATH(from);
806 		return (error);
807 	}
808 
809 #ifdef DEBUG
810 	if (ldebug(renameat))
811 		printf(ARGS(renameat, "%s, %s"), from, to);
812 #endif
813 	error = kern_renameat(td, olddfd, from, newdfd, to, UIO_SYSSPACE);
814 	LFREEPATH(from);
815 	LFREEPATH(to);
816 	return (error);
817 }
818 
819 int
820 linux_symlink(struct thread *td, struct linux_symlink_args *args)
821 {
822 	char *path, *to;
823 	int error;
824 
825 	LCONVPATHEXIST(td, args->path, &path);
826 	/* Expand LCONVPATHCREATE so that `path' can be freed on errors */
827 	error = linux_emul_convpath(td, args->to, UIO_USERSPACE, &to, 1, AT_FDCWD);
828 	if (to == NULL) {
829 		LFREEPATH(path);
830 		return (error);
831 	}
832 
833 #ifdef DEBUG
834 	if (ldebug(symlink))
835 		printf(ARGS(symlink, "%s, %s"), path, to);
836 #endif
837 	error = kern_symlink(td, path, to, UIO_SYSSPACE);
838 	LFREEPATH(path);
839 	LFREEPATH(to);
840 	return (error);
841 }
842 
843 int
844 linux_symlinkat(struct thread *td, struct linux_symlinkat_args *args)
845 {
846 	char *path, *to;
847 	int error, dfd;
848 
849 	dfd = (args->newdfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->newdfd;
850 	LCONVPATHEXIST_AT(td, args->oldname, &path, dfd);
851 	/* Expand LCONVPATHCREATE so that `path' can be freed on errors */
852 	error = linux_emul_convpath(td, args->newname, UIO_USERSPACE, &to, 1, dfd);
853 	if (to == NULL) {
854 		LFREEPATH(path);
855 		return (error);
856 	}
857 
858 #ifdef DEBUG
859 	if (ldebug(symlinkat))
860 		printf(ARGS(symlinkat, "%s, %s"), path, to);
861 #endif
862 
863 	error = kern_symlinkat(td, path, dfd, to, UIO_SYSSPACE);
864 	LFREEPATH(path);
865 	LFREEPATH(to);
866 	return (error);
867 }
868 
869 int
870 linux_readlink(struct thread *td, struct linux_readlink_args *args)
871 {
872 	char *name;
873 	int error;
874 
875 	LCONVPATHEXIST(td, args->name, &name);
876 
877 #ifdef DEBUG
878 	if (ldebug(readlink))
879 		printf(ARGS(readlink, "%s, %p, %d"), name, (void *)args->buf,
880 		    args->count);
881 #endif
882 	error = kern_readlink(td, name, UIO_SYSSPACE, args->buf, UIO_USERSPACE,
883 	    args->count);
884 	LFREEPATH(name);
885 	return (error);
886 }
887 
888 int
889 linux_readlinkat(struct thread *td, struct linux_readlinkat_args *args)
890 {
891 	char *name;
892 	int error, dfd;
893 
894 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
895 	LCONVPATHEXIST_AT(td, args->path, &name, dfd);
896 
897 #ifdef DEBUG
898 	if (ldebug(readlinkat))
899 		printf(ARGS(readlinkat, "%s, %p, %d"), name, (void *)args->buf,
900 		    args->bufsiz);
901 #endif
902 
903 	error = kern_readlinkat(td, dfd, name, UIO_SYSSPACE, args->buf,
904 	    UIO_USERSPACE, args->bufsiz);
905 	LFREEPATH(name);
906 	return (error);
907 }
908 
909 int
910 linux_truncate(struct thread *td, struct linux_truncate_args *args)
911 {
912 	char *path;
913 	int error;
914 
915 	LCONVPATHEXIST(td, args->path, &path);
916 
917 #ifdef DEBUG
918 	if (ldebug(truncate))
919 		printf(ARGS(truncate, "%s, %ld"), path, (long)args->length);
920 #endif
921 
922 	error = kern_truncate(td, path, UIO_SYSSPACE, args->length);
923 	LFREEPATH(path);
924 	return (error);
925 }
926 
927 int
928 linux_truncate64(struct thread *td, struct linux_truncate64_args *args)
929 {
930 	char *path;
931 	int error;
932 
933 	LCONVPATHEXIST(td, args->path, &path);
934 
935 #ifdef DEBUG
936 	if (ldebug(truncate64))
937 		printf(ARGS(truncate64, "%s, %jd"), path, args->length);
938 #endif
939 
940 	error = kern_truncate(td, path, UIO_SYSSPACE, args->length);
941 	LFREEPATH(path);
942 	return (error);
943 }
944 int
945 linux_ftruncate(struct thread *td, struct linux_ftruncate_args *args)
946 {
947 	struct ftruncate_args /* {
948 		int fd;
949 		int pad;
950 		off_t length;
951 		} */ nuap;
952 
953 	nuap.fd = args->fd;
954 	nuap.length = args->length;
955 	return (sys_ftruncate(td, &nuap));
956 }
957 
958 int
959 linux_link(struct thread *td, struct linux_link_args *args)
960 {
961 	char *path, *to;
962 	int error;
963 
964 	LCONVPATHEXIST(td, args->path, &path);
965 	/* Expand LCONVPATHCREATE so that `path' can be freed on errors */
966 	error = linux_emul_convpath(td, args->to, UIO_USERSPACE, &to, 1, AT_FDCWD);
967 	if (to == NULL) {
968 		LFREEPATH(path);
969 		return (error);
970 	}
971 
972 #ifdef DEBUG
973 	if (ldebug(link))
974 		printf(ARGS(link, "%s, %s"), path, to);
975 #endif
976 	error = kern_link(td, path, to, UIO_SYSSPACE);
977 	LFREEPATH(path);
978 	LFREEPATH(to);
979 	return (error);
980 }
981 
982 int
983 linux_linkat(struct thread *td, struct linux_linkat_args *args)
984 {
985 	char *path, *to;
986 	int error, olddfd, newdfd, follow;
987 
988 	if (args->flag & ~LINUX_AT_SYMLINK_FOLLOW)
989 		return (EINVAL);
990 
991 	olddfd = (args->olddfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->olddfd;
992 	newdfd = (args->newdfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->newdfd;
993 	LCONVPATHEXIST_AT(td, args->oldname, &path, olddfd);
994 	/* Expand LCONVPATHCREATE so that `path' can be freed on errors */
995 	error = linux_emul_convpath(td, args->newname, UIO_USERSPACE, &to, 1, newdfd);
996 	if (to == NULL) {
997 		LFREEPATH(path);
998 		return (error);
999 	}
1000 
1001 #ifdef DEBUG
1002 	if (ldebug(linkat))
1003 		printf(ARGS(linkat, "%i, %s, %i, %s, %i"), args->olddfd, path,
1004 			args->newdfd, to, args->flag);
1005 #endif
1006 
1007 	follow = (args->flag & LINUX_AT_SYMLINK_FOLLOW) == 0 ? NOFOLLOW :
1008 	    FOLLOW;
1009 	error = kern_linkat(td, olddfd, newdfd, path, to, UIO_SYSSPACE, follow);
1010 	LFREEPATH(path);
1011 	LFREEPATH(to);
1012 	return (error);
1013 }
1014 
1015 int
1016 linux_fdatasync(td, uap)
1017 	struct thread *td;
1018 	struct linux_fdatasync_args *uap;
1019 {
1020 	struct fsync_args bsd;
1021 
1022 	bsd.fd = uap->fd;
1023 	return sys_fsync(td, &bsd);
1024 }
1025 
1026 int
1027 linux_pread(td, uap)
1028 	struct thread *td;
1029 	struct linux_pread_args *uap;
1030 {
1031 	struct pread_args bsd;
1032 	struct vnode *vp;
1033 	int error;
1034 
1035 	bsd.fd = uap->fd;
1036 	bsd.buf = uap->buf;
1037 	bsd.nbyte = uap->nbyte;
1038 	bsd.offset = uap->offset;
1039 
1040 	error = sys_pread(td, &bsd);
1041 
1042 	if (error == 0) {
1043    	   	/* This seems to violate POSIX but linux does it */
1044 		if ((error = fgetvp(td, uap->fd, CAP_READ, &vp)) != 0)
1045    		   	return (error);
1046 		if (vp->v_type == VDIR) {
1047    		   	vrele(vp);
1048 			return (EISDIR);
1049 		}
1050 		vrele(vp);
1051 	}
1052 
1053 	return (error);
1054 }
1055 
1056 int
1057 linux_pwrite(td, uap)
1058 	struct thread *td;
1059 	struct linux_pwrite_args *uap;
1060 {
1061 	struct pwrite_args bsd;
1062 
1063 	bsd.fd = uap->fd;
1064 	bsd.buf = uap->buf;
1065 	bsd.nbyte = uap->nbyte;
1066 	bsd.offset = uap->offset;
1067 	return sys_pwrite(td, &bsd);
1068 }
1069 
1070 int
1071 linux_mount(struct thread *td, struct linux_mount_args *args)
1072 {
1073 	struct ufs_args ufs;
1074 	char fstypename[MFSNAMELEN];
1075 	char mntonname[MNAMELEN], mntfromname[MNAMELEN];
1076 	int error;
1077 	int fsflags;
1078 	void *fsdata;
1079 
1080 	error = copyinstr(args->filesystemtype, fstypename, MFSNAMELEN - 1,
1081 	    NULL);
1082 	if (error)
1083 		return (error);
1084 	error = copyinstr(args->specialfile, mntfromname, MNAMELEN - 1, NULL);
1085 	if (error)
1086 		return (error);
1087 	error = copyinstr(args->dir, mntonname, MNAMELEN - 1, NULL);
1088 	if (error)
1089 		return (error);
1090 
1091 #ifdef DEBUG
1092 	if (ldebug(mount))
1093 		printf(ARGS(mount, "%s, %s, %s"),
1094 		    fstypename, mntfromname, mntonname);
1095 #endif
1096 
1097 	if (strcmp(fstypename, "ext2") == 0) {
1098 		strcpy(fstypename, "ext2fs");
1099 		fsdata = &ufs;
1100 		ufs.fspec = mntfromname;
1101 #define DEFAULT_ROOTID		-2
1102 		ufs.export.ex_root = DEFAULT_ROOTID;
1103 		ufs.export.ex_flags =
1104 		    args->rwflag & LINUX_MS_RDONLY ? MNT_EXRDONLY : 0;
1105 	} else if (strcmp(fstypename, "proc") == 0) {
1106 		strcpy(fstypename, "linprocfs");
1107 		fsdata = NULL;
1108 	} else if (strcmp(fstypename, "vfat") == 0) {
1109 		strcpy(fstypename, "msdosfs");
1110 		fsdata = NULL;
1111 	} else {
1112 		return (ENODEV);
1113 	}
1114 
1115 	fsflags = 0;
1116 
1117 	if ((args->rwflag & 0xffff0000) == 0xc0ed0000) {
1118 		/*
1119 		 * Linux SYNC flag is not included; the closest equivalent
1120 		 * FreeBSD has is !ASYNC, which is our default.
1121 		 */
1122 		if (args->rwflag & LINUX_MS_RDONLY)
1123 			fsflags |= MNT_RDONLY;
1124 		if (args->rwflag & LINUX_MS_NOSUID)
1125 			fsflags |= MNT_NOSUID;
1126 		if (args->rwflag & LINUX_MS_NOEXEC)
1127 			fsflags |= MNT_NOEXEC;
1128 		if (args->rwflag & LINUX_MS_REMOUNT)
1129 			fsflags |= MNT_UPDATE;
1130 	}
1131 
1132 	if (strcmp(fstypename, "linprocfs") == 0) {
1133 		error = kernel_vmount(fsflags,
1134 			"fstype", fstypename,
1135 			"fspath", mntonname,
1136 			NULL);
1137 	} else if (strcmp(fstypename, "msdosfs") == 0) {
1138 		error = kernel_vmount(fsflags,
1139 			"fstype", fstypename,
1140 			"fspath", mntonname,
1141 			"from", mntfromname,
1142 			NULL);
1143 	} else
1144 		error = EOPNOTSUPP;
1145 	return (error);
1146 }
1147 
1148 int
1149 linux_oldumount(struct thread *td, struct linux_oldumount_args *args)
1150 {
1151 	struct linux_umount_args args2;
1152 
1153 	args2.path = args->path;
1154 	args2.flags = 0;
1155 	return (linux_umount(td, &args2));
1156 }
1157 
1158 int
1159 linux_umount(struct thread *td, struct linux_umount_args *args)
1160 {
1161 	struct unmount_args bsd;
1162 
1163 	bsd.path = args->path;
1164 	bsd.flags = args->flags;	/* XXX correct? */
1165 	return (sys_unmount(td, &bsd));
1166 }
1167 
1168 /*
1169  * fcntl family of syscalls
1170  */
1171 
1172 struct l_flock {
1173 	l_short		l_type;
1174 	l_short		l_whence;
1175 	l_off_t		l_start;
1176 	l_off_t		l_len;
1177 	l_pid_t		l_pid;
1178 }
1179 #if defined(__amd64__) && defined(COMPAT_LINUX32)
1180 __packed
1181 #endif
1182 ;
1183 
1184 static void
1185 linux_to_bsd_flock(struct l_flock *linux_flock, struct flock *bsd_flock)
1186 {
1187 	switch (linux_flock->l_type) {
1188 	case LINUX_F_RDLCK:
1189 		bsd_flock->l_type = F_RDLCK;
1190 		break;
1191 	case LINUX_F_WRLCK:
1192 		bsd_flock->l_type = F_WRLCK;
1193 		break;
1194 	case LINUX_F_UNLCK:
1195 		bsd_flock->l_type = F_UNLCK;
1196 		break;
1197 	default:
1198 		bsd_flock->l_type = -1;
1199 		break;
1200 	}
1201 	bsd_flock->l_whence = linux_flock->l_whence;
1202 	bsd_flock->l_start = (off_t)linux_flock->l_start;
1203 	bsd_flock->l_len = (off_t)linux_flock->l_len;
1204 	bsd_flock->l_pid = (pid_t)linux_flock->l_pid;
1205 	bsd_flock->l_sysid = 0;
1206 }
1207 
1208 static void
1209 bsd_to_linux_flock(struct flock *bsd_flock, struct l_flock *linux_flock)
1210 {
1211 	switch (bsd_flock->l_type) {
1212 	case F_RDLCK:
1213 		linux_flock->l_type = LINUX_F_RDLCK;
1214 		break;
1215 	case F_WRLCK:
1216 		linux_flock->l_type = LINUX_F_WRLCK;
1217 		break;
1218 	case F_UNLCK:
1219 		linux_flock->l_type = LINUX_F_UNLCK;
1220 		break;
1221 	}
1222 	linux_flock->l_whence = bsd_flock->l_whence;
1223 	linux_flock->l_start = (l_off_t)bsd_flock->l_start;
1224 	linux_flock->l_len = (l_off_t)bsd_flock->l_len;
1225 	linux_flock->l_pid = (l_pid_t)bsd_flock->l_pid;
1226 }
1227 
1228 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1229 struct l_flock64 {
1230 	l_short		l_type;
1231 	l_short		l_whence;
1232 	l_loff_t	l_start;
1233 	l_loff_t	l_len;
1234 	l_pid_t		l_pid;
1235 }
1236 #if defined(__amd64__) && defined(COMPAT_LINUX32)
1237 __packed
1238 #endif
1239 ;
1240 
1241 static void
1242 linux_to_bsd_flock64(struct l_flock64 *linux_flock, struct flock *bsd_flock)
1243 {
1244 	switch (linux_flock->l_type) {
1245 	case LINUX_F_RDLCK:
1246 		bsd_flock->l_type = F_RDLCK;
1247 		break;
1248 	case LINUX_F_WRLCK:
1249 		bsd_flock->l_type = F_WRLCK;
1250 		break;
1251 	case LINUX_F_UNLCK:
1252 		bsd_flock->l_type = F_UNLCK;
1253 		break;
1254 	default:
1255 		bsd_flock->l_type = -1;
1256 		break;
1257 	}
1258 	bsd_flock->l_whence = linux_flock->l_whence;
1259 	bsd_flock->l_start = (off_t)linux_flock->l_start;
1260 	bsd_flock->l_len = (off_t)linux_flock->l_len;
1261 	bsd_flock->l_pid = (pid_t)linux_flock->l_pid;
1262 	bsd_flock->l_sysid = 0;
1263 }
1264 
1265 static void
1266 bsd_to_linux_flock64(struct flock *bsd_flock, struct l_flock64 *linux_flock)
1267 {
1268 	switch (bsd_flock->l_type) {
1269 	case F_RDLCK:
1270 		linux_flock->l_type = LINUX_F_RDLCK;
1271 		break;
1272 	case F_WRLCK:
1273 		linux_flock->l_type = LINUX_F_WRLCK;
1274 		break;
1275 	case F_UNLCK:
1276 		linux_flock->l_type = LINUX_F_UNLCK;
1277 		break;
1278 	}
1279 	linux_flock->l_whence = bsd_flock->l_whence;
1280 	linux_flock->l_start = (l_loff_t)bsd_flock->l_start;
1281 	linux_flock->l_len = (l_loff_t)bsd_flock->l_len;
1282 	linux_flock->l_pid = (l_pid_t)bsd_flock->l_pid;
1283 }
1284 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1285 
1286 static int
1287 fcntl_common(struct thread *td, struct linux_fcntl64_args *args)
1288 {
1289 	struct l_flock linux_flock;
1290 	struct flock bsd_flock;
1291 	struct file *fp;
1292 	long arg;
1293 	int error, result;
1294 
1295 	switch (args->cmd) {
1296 	case LINUX_F_DUPFD:
1297 		return (kern_fcntl(td, args->fd, F_DUPFD, args->arg));
1298 
1299 	case LINUX_F_GETFD:
1300 		return (kern_fcntl(td, args->fd, F_GETFD, 0));
1301 
1302 	case LINUX_F_SETFD:
1303 		return (kern_fcntl(td, args->fd, F_SETFD, args->arg));
1304 
1305 	case LINUX_F_GETFL:
1306 		error = kern_fcntl(td, args->fd, F_GETFL, 0);
1307 		result = td->td_retval[0];
1308 		td->td_retval[0] = 0;
1309 		if (result & O_RDONLY)
1310 			td->td_retval[0] |= LINUX_O_RDONLY;
1311 		if (result & O_WRONLY)
1312 			td->td_retval[0] |= LINUX_O_WRONLY;
1313 		if (result & O_RDWR)
1314 			td->td_retval[0] |= LINUX_O_RDWR;
1315 		if (result & O_NDELAY)
1316 			td->td_retval[0] |= LINUX_O_NONBLOCK;
1317 		if (result & O_APPEND)
1318 			td->td_retval[0] |= LINUX_O_APPEND;
1319 		if (result & O_FSYNC)
1320 			td->td_retval[0] |= LINUX_O_SYNC;
1321 		if (result & O_ASYNC)
1322 			td->td_retval[0] |= LINUX_FASYNC;
1323 #ifdef LINUX_O_NOFOLLOW
1324 		if (result & O_NOFOLLOW)
1325 			td->td_retval[0] |= LINUX_O_NOFOLLOW;
1326 #endif
1327 #ifdef LINUX_O_DIRECT
1328 		if (result & O_DIRECT)
1329 			td->td_retval[0] |= LINUX_O_DIRECT;
1330 #endif
1331 		return (error);
1332 
1333 	case LINUX_F_SETFL:
1334 		arg = 0;
1335 		if (args->arg & LINUX_O_NDELAY)
1336 			arg |= O_NONBLOCK;
1337 		if (args->arg & LINUX_O_APPEND)
1338 			arg |= O_APPEND;
1339 		if (args->arg & LINUX_O_SYNC)
1340 			arg |= O_FSYNC;
1341 		if (args->arg & LINUX_FASYNC)
1342 			arg |= O_ASYNC;
1343 #ifdef LINUX_O_NOFOLLOW
1344 		if (args->arg & LINUX_O_NOFOLLOW)
1345 			arg |= O_NOFOLLOW;
1346 #endif
1347 #ifdef LINUX_O_DIRECT
1348 		if (args->arg & LINUX_O_DIRECT)
1349 			arg |= O_DIRECT;
1350 #endif
1351 		return (kern_fcntl(td, args->fd, F_SETFL, arg));
1352 
1353 	case LINUX_F_GETLK:
1354 		error = copyin((void *)args->arg, &linux_flock,
1355 		    sizeof(linux_flock));
1356 		if (error)
1357 			return (error);
1358 		linux_to_bsd_flock(&linux_flock, &bsd_flock);
1359 		error = kern_fcntl(td, args->fd, F_GETLK, (intptr_t)&bsd_flock);
1360 		if (error)
1361 			return (error);
1362 		bsd_to_linux_flock(&bsd_flock, &linux_flock);
1363 		return (copyout(&linux_flock, (void *)args->arg,
1364 		    sizeof(linux_flock)));
1365 
1366 	case LINUX_F_SETLK:
1367 		error = copyin((void *)args->arg, &linux_flock,
1368 		    sizeof(linux_flock));
1369 		if (error)
1370 			return (error);
1371 		linux_to_bsd_flock(&linux_flock, &bsd_flock);
1372 		return (kern_fcntl(td, args->fd, F_SETLK,
1373 		    (intptr_t)&bsd_flock));
1374 
1375 	case LINUX_F_SETLKW:
1376 		error = copyin((void *)args->arg, &linux_flock,
1377 		    sizeof(linux_flock));
1378 		if (error)
1379 			return (error);
1380 		linux_to_bsd_flock(&linux_flock, &bsd_flock);
1381 		return (kern_fcntl(td, args->fd, F_SETLKW,
1382 		     (intptr_t)&bsd_flock));
1383 
1384 	case LINUX_F_GETOWN:
1385 		return (kern_fcntl(td, args->fd, F_GETOWN, 0));
1386 
1387 	case LINUX_F_SETOWN:
1388 		/*
1389 		 * XXX some Linux applications depend on F_SETOWN having no
1390 		 * significant effect for pipes (SIGIO is not delivered for
1391 		 * pipes under Linux-2.2.35 at least).
1392 		 */
1393 		error = fget(td, args->fd, CAP_FCNTL, &fp);
1394 		if (error)
1395 			return (error);
1396 		if (fp->f_type == DTYPE_PIPE) {
1397 			fdrop(fp, td);
1398 			return (EINVAL);
1399 		}
1400 		fdrop(fp, td);
1401 
1402 		return (kern_fcntl(td, args->fd, F_SETOWN, args->arg));
1403 	}
1404 
1405 	return (EINVAL);
1406 }
1407 
1408 int
1409 linux_fcntl(struct thread *td, struct linux_fcntl_args *args)
1410 {
1411 	struct linux_fcntl64_args args64;
1412 
1413 #ifdef DEBUG
1414 	if (ldebug(fcntl))
1415 		printf(ARGS(fcntl, "%d, %08x, *"), args->fd, args->cmd);
1416 #endif
1417 
1418 	args64.fd = args->fd;
1419 	args64.cmd = args->cmd;
1420 	args64.arg = args->arg;
1421 	return (fcntl_common(td, &args64));
1422 }
1423 
1424 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1425 int
1426 linux_fcntl64(struct thread *td, struct linux_fcntl64_args *args)
1427 {
1428 	struct l_flock64 linux_flock;
1429 	struct flock bsd_flock;
1430 	int error;
1431 
1432 #ifdef DEBUG
1433 	if (ldebug(fcntl64))
1434 		printf(ARGS(fcntl64, "%d, %08x, *"), args->fd, args->cmd);
1435 #endif
1436 
1437 	switch (args->cmd) {
1438 	case LINUX_F_GETLK64:
1439 		error = copyin((void *)args->arg, &linux_flock,
1440 		    sizeof(linux_flock));
1441 		if (error)
1442 			return (error);
1443 		linux_to_bsd_flock64(&linux_flock, &bsd_flock);
1444 		error = kern_fcntl(td, args->fd, F_GETLK, (intptr_t)&bsd_flock);
1445 		if (error)
1446 			return (error);
1447 		bsd_to_linux_flock64(&bsd_flock, &linux_flock);
1448 		return (copyout(&linux_flock, (void *)args->arg,
1449 			    sizeof(linux_flock)));
1450 
1451 	case LINUX_F_SETLK64:
1452 		error = copyin((void *)args->arg, &linux_flock,
1453 		    sizeof(linux_flock));
1454 		if (error)
1455 			return (error);
1456 		linux_to_bsd_flock64(&linux_flock, &bsd_flock);
1457 		return (kern_fcntl(td, args->fd, F_SETLK,
1458 		    (intptr_t)&bsd_flock));
1459 
1460 	case LINUX_F_SETLKW64:
1461 		error = copyin((void *)args->arg, &linux_flock,
1462 		    sizeof(linux_flock));
1463 		if (error)
1464 			return (error);
1465 		linux_to_bsd_flock64(&linux_flock, &bsd_flock);
1466 		return (kern_fcntl(td, args->fd, F_SETLKW,
1467 		    (intptr_t)&bsd_flock));
1468 	}
1469 
1470 	return (fcntl_common(td, args));
1471 }
1472 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1473 
1474 int
1475 linux_chown(struct thread *td, struct linux_chown_args *args)
1476 {
1477 	char *path;
1478 	int error;
1479 
1480 	LCONVPATHEXIST(td, args->path, &path);
1481 
1482 #ifdef DEBUG
1483 	if (ldebug(chown))
1484 		printf(ARGS(chown, "%s, %d, %d"), path, args->uid, args->gid);
1485 #endif
1486 	error = kern_chown(td, path, UIO_SYSSPACE, args->uid, args->gid);
1487 	LFREEPATH(path);
1488 	return (error);
1489 }
1490 
1491 int
1492 linux_fchownat(struct thread *td, struct linux_fchownat_args *args)
1493 {
1494 	char *path;
1495 	int error, dfd, flag;
1496 
1497 	if (args->flag & ~LINUX_AT_SYMLINK_NOFOLLOW)
1498 		return (EINVAL);
1499 
1500 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD :  args->dfd;
1501 	LCONVPATHEXIST_AT(td, args->filename, &path, dfd);
1502 
1503 #ifdef DEBUG
1504 	if (ldebug(fchownat))
1505 		printf(ARGS(fchownat, "%s, %d, %d"), path, args->uid, args->gid);
1506 #endif
1507 
1508 	flag = (args->flag & LINUX_AT_SYMLINK_NOFOLLOW) == 0 ? 0 :
1509 	    AT_SYMLINK_NOFOLLOW;
1510 	error = kern_fchownat(td, dfd, path, UIO_SYSSPACE, args->uid, args->gid,
1511 	    flag);
1512 	LFREEPATH(path);
1513 	return (error);
1514 }
1515 
1516 int
1517 linux_lchown(struct thread *td, struct linux_lchown_args *args)
1518 {
1519 	char *path;
1520 	int error;
1521 
1522 	LCONVPATHEXIST(td, args->path, &path);
1523 
1524 #ifdef DEBUG
1525 	if (ldebug(lchown))
1526 		printf(ARGS(lchown, "%s, %d, %d"), path, args->uid, args->gid);
1527 #endif
1528 	error = kern_lchown(td, path, UIO_SYSSPACE, args->uid, args->gid);
1529 	LFREEPATH(path);
1530 	return (error);
1531 }
1532 
1533 static int
1534 convert_fadvice(int advice)
1535 {
1536 	switch (advice) {
1537 	case LINUX_POSIX_FADV_NORMAL:
1538 		return (POSIX_FADV_NORMAL);
1539 	case LINUX_POSIX_FADV_RANDOM:
1540 		return (POSIX_FADV_RANDOM);
1541 	case LINUX_POSIX_FADV_SEQUENTIAL:
1542 		return (POSIX_FADV_SEQUENTIAL);
1543 	case LINUX_POSIX_FADV_WILLNEED:
1544 		return (POSIX_FADV_WILLNEED);
1545 	case LINUX_POSIX_FADV_DONTNEED:
1546 		return (POSIX_FADV_DONTNEED);
1547 	case LINUX_POSIX_FADV_NOREUSE:
1548 		return (POSIX_FADV_NOREUSE);
1549 	default:
1550 		return (-1);
1551 	}
1552 }
1553 
1554 int
1555 linux_fadvise64(struct thread *td, struct linux_fadvise64_args *args)
1556 {
1557 	int advice;
1558 
1559 	advice = convert_fadvice(args->advice);
1560 	if (advice == -1)
1561 		return (EINVAL);
1562 	return (kern_posix_fadvise(td, args->fd, args->offset, args->len,
1563 	    advice));
1564 }
1565 
1566 int
1567 linux_fadvise64_64(struct thread *td, struct linux_fadvise64_64_args *args)
1568 {
1569 	int advice;
1570 
1571 	advice = convert_fadvice(args->advice);
1572 	if (advice == -1)
1573 		return (EINVAL);
1574 	return (kern_posix_fadvise(td, args->fd, args->offset, args->len,
1575 	    advice));
1576 }
1577 
1578 int
1579 linux_pipe(struct thread *td, struct linux_pipe_args *args)
1580 {
1581 	int fildes[2];
1582 	int error;
1583 
1584 #ifdef DEBUG
1585 	if (ldebug(pipe))
1586 		printf(ARGS(pipe, "*"));
1587 #endif
1588 
1589 	error = do_pipe(td, fildes, 0);
1590 	if (error)
1591 		return (error);
1592 
1593 	/* XXX: Close descriptors on error. */
1594 	return (copyout(fildes, args->pipefds, sizeof(fildes)));
1595 }
1596 
1597 int
1598 linux_pipe2(struct thread *td, struct linux_pipe2_args *args)
1599 {
1600 	int fildes[2];
1601 	int error, flags;
1602 
1603 #ifdef DEBUG
1604 	if (ldebug(pipe2))
1605 		printf(ARGS(pipe2, "*, %d"), args->flags);
1606 #endif
1607 
1608 	if ((args->flags & ~(LINUX_O_NONBLOCK | LINUX_O_CLOEXEC)) != 0)
1609 		return (EINVAL);
1610 
1611 	flags = 0;
1612 	if ((args->flags & LINUX_O_NONBLOCK) != 0)
1613 		flags |= O_NONBLOCK;
1614 	if ((args->flags & LINUX_O_CLOEXEC) != 0)
1615 		flags |= O_CLOEXEC;
1616 	error = do_pipe(td, fildes, flags);
1617 	if (error)
1618 		return (error);
1619 
1620 	/* XXX: Close descriptors on error. */
1621 	return (copyout(fildes, args->pipefds, sizeof(fildes)));
1622 }
1623