xref: /freebsd/sys/compat/linux/linux_file.c (revision 6486b015fc84e96725fef22b0e3363351399ae83)
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 	vp = fp->f_vnode;
361 	vfslocked = VFS_LOCK_GIANT(vp->v_mount);
362 	if (vp->v_type != VDIR) {
363 		VFS_UNLOCK_GIANT(vfslocked);
364 		fdrop(fp, td);
365 		return (EINVAL);
366 	}
367 
368 	off = fp->f_offset;
369 
370 	buflen = max(LINUX_DIRBLKSIZ, nbytes);
371 	buflen = min(buflen, MAXBSIZE);
372 	buf = malloc(buflen, M_TEMP, M_WAITOK);
373 	lbuf = malloc(LINUX_MAXRECLEN, M_TEMP, M_WAITOK | M_ZERO);
374 	vn_lock(vp, LK_SHARED | LK_RETRY);
375 
376 	aiov.iov_base = buf;
377 	aiov.iov_len = buflen;
378 	auio.uio_iov = &aiov;
379 	auio.uio_iovcnt = 1;
380 	auio.uio_rw = UIO_READ;
381 	auio.uio_segflg = UIO_SYSSPACE;
382 	auio.uio_td = td;
383 	auio.uio_resid = buflen;
384 	auio.uio_offset = off;
385 
386 	if (cookies) {
387 		free(cookies, M_TEMP);
388 		cookies = NULL;
389 	}
390 
391 #ifdef MAC
392 	/*
393 	 * Do directory search MAC check using non-cached credentials.
394 	 */
395 	if ((error = mac_vnode_check_readdir(td->td_ucred, vp)))
396 		goto out;
397 #endif /* MAC */
398 	if ((error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, &ncookies,
399 		 &cookies)))
400 		goto out;
401 
402 	inp = buf;
403 	outp = (caddr_t)args->dirent;
404 	resid = nbytes;
405 	if ((len = buflen - auio.uio_resid) <= 0)
406 		goto eof;
407 
408 	cookiep = cookies;
409 
410 	if (cookies) {
411 		/*
412 		 * When using cookies, the vfs has the option of reading from
413 		 * a different offset than that supplied (UFS truncates the
414 		 * offset to a block boundary to make sure that it never reads
415 		 * partway through a directory entry, even if the directory
416 		 * has been compacted).
417 		 */
418 		while (len > 0 && ncookies > 0 && *cookiep <= off) {
419 			bdp = (struct dirent *) inp;
420 			len -= bdp->d_reclen;
421 			inp += bdp->d_reclen;
422 			cookiep++;
423 			ncookies--;
424 		}
425 	}
426 
427 	while (len > 0) {
428 		if (cookiep && ncookies == 0)
429 			break;
430 		bdp = (struct dirent *) inp;
431 		reclen = bdp->d_reclen;
432 		if (reclen & 3) {
433 			error = EFAULT;
434 			goto out;
435 		}
436 
437 		if (bdp->d_fileno == 0) {
438 			inp += reclen;
439 			if (cookiep) {
440 				off = *cookiep++;
441 				ncookies--;
442 			} else
443 				off += reclen;
444 
445 			len -= reclen;
446 			continue;
447 		}
448 
449 		linuxreclen = (is64bit)
450 		    ? LINUX_RECLEN64(bdp->d_namlen)
451 		    : LINUX_RECLEN(bdp->d_namlen);
452 
453 		if (reclen > len || resid < linuxreclen) {
454 			outp++;
455 			break;
456 		}
457 
458 		if (justone) {
459 			/* readdir(2) case. */
460 			linux_dirent = (struct l_dirent*)lbuf;
461 			linux_dirent->d_ino = bdp->d_fileno;
462 			linux_dirent->d_off = (l_off_t)linuxreclen;
463 			linux_dirent->d_reclen = (l_ushort)bdp->d_namlen;
464 			strlcpy(linux_dirent->d_name, bdp->d_name,
465 			    linuxreclen - offsetof(struct l_dirent, d_name));
466 			error = copyout(linux_dirent, outp, linuxreclen);
467 		}
468 		if (is64bit) {
469 			linux_dirent64 = (struct l_dirent64*)lbuf;
470 			linux_dirent64->d_ino = bdp->d_fileno;
471 			linux_dirent64->d_off = (cookiep)
472 			    ? (l_off_t)*cookiep
473 			    : (l_off_t)(off + reclen);
474 			linux_dirent64->d_reclen = (l_ushort)linuxreclen;
475 			linux_dirent64->d_type = bdp->d_type;
476 			strlcpy(linux_dirent64->d_name, bdp->d_name,
477 			    linuxreclen - offsetof(struct l_dirent64, d_name));
478 			error = copyout(linux_dirent64, outp, linuxreclen);
479 		} else if (!justone) {
480 			linux_dirent = (struct l_dirent*)lbuf;
481 			linux_dirent->d_ino = bdp->d_fileno;
482 			linux_dirent->d_off = (cookiep)
483 			    ? (l_off_t)*cookiep
484 			    : (l_off_t)(off + reclen);
485 			linux_dirent->d_reclen = (l_ushort)linuxreclen;
486 			/*
487 			 * Copy d_type to last byte of l_dirent buffer
488 			 */
489 			lbuf[linuxreclen-1] = bdp->d_type;
490 			strlcpy(linux_dirent->d_name, bdp->d_name,
491 			    linuxreclen - offsetof(struct l_dirent, d_name)-1);
492 			error = copyout(linux_dirent, outp, linuxreclen);
493 		}
494 
495 		if (error)
496 			goto out;
497 
498 		inp += reclen;
499 		if (cookiep) {
500 			off = *cookiep++;
501 			ncookies--;
502 		} else
503 			off += reclen;
504 
505 		outp += linuxreclen;
506 		resid -= linuxreclen;
507 		len -= reclen;
508 		if (justone)
509 			break;
510 	}
511 
512 	if (outp == (caddr_t)args->dirent) {
513 		nbytes = resid;
514 		goto eof;
515 	}
516 
517 	fp->f_offset = off;
518 	if (justone)
519 		nbytes = resid + linuxreclen;
520 
521 eof:
522 	td->td_retval[0] = nbytes - resid;
523 
524 out:
525 	if (cookies)
526 		free(cookies, M_TEMP);
527 
528 	VOP_UNLOCK(vp, 0);
529 	VFS_UNLOCK_GIANT(vfslocked);
530 	fdrop(fp, td);
531 	free(buf, M_TEMP);
532 	free(lbuf, M_TEMP);
533 	return (error);
534 }
535 
536 int
537 linux_getdents(struct thread *td, struct linux_getdents_args *args)
538 {
539 
540 #ifdef DEBUG
541 	if (ldebug(getdents))
542 		printf(ARGS(getdents, "%d, *, %d"), args->fd, args->count);
543 #endif
544 
545 	return (getdents_common(td, (struct linux_getdents64_args*)args, 0));
546 }
547 
548 int
549 linux_getdents64(struct thread *td, struct linux_getdents64_args *args)
550 {
551 
552 #ifdef DEBUG
553 	if (ldebug(getdents64))
554 		printf(ARGS(getdents64, "%d, *, %d"), args->fd, args->count);
555 #endif
556 
557 	return (getdents_common(td, args, 1));
558 }
559 
560 /*
561  * These exist mainly for hooks for doing /compat/linux translation.
562  */
563 
564 int
565 linux_access(struct thread *td, struct linux_access_args *args)
566 {
567 	char *path;
568 	int error;
569 
570 	/* linux convention */
571 	if (args->amode & ~(F_OK | X_OK | W_OK | R_OK))
572 		return (EINVAL);
573 
574 	LCONVPATHEXIST(td, args->path, &path);
575 
576 #ifdef DEBUG
577 	if (ldebug(access))
578 		printf(ARGS(access, "%s, %d"), path, args->amode);
579 #endif
580 	error = kern_access(td, path, UIO_SYSSPACE, args->amode);
581 	LFREEPATH(path);
582 
583 	return (error);
584 }
585 
586 int
587 linux_faccessat(struct thread *td, struct linux_faccessat_args *args)
588 {
589 	char *path;
590 	int error, dfd, flag;
591 
592 	if (args->flag & ~LINUX_AT_EACCESS)
593 		return (EINVAL);
594 	/* linux convention */
595 	if (args->amode & ~(F_OK | X_OK | W_OK | R_OK))
596 		return (EINVAL);
597 
598 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
599 	LCONVPATHEXIST_AT(td, args->filename, &path, dfd);
600 
601 #ifdef DEBUG
602 	if (ldebug(access))
603 		printf(ARGS(access, "%s, %d"), path, args->amode);
604 #endif
605 
606 	flag = (args->flag & LINUX_AT_EACCESS) == 0 ? 0 : AT_EACCESS;
607 	error = kern_accessat(td, dfd, path, UIO_SYSSPACE, flag, args->amode);
608 	LFREEPATH(path);
609 
610 	return (error);
611 }
612 
613 int
614 linux_unlink(struct thread *td, struct linux_unlink_args *args)
615 {
616 	char *path;
617 	int error;
618 	struct stat st;
619 
620 	LCONVPATHEXIST(td, args->path, &path);
621 
622 #ifdef DEBUG
623 	if (ldebug(unlink))
624 		printf(ARGS(unlink, "%s"), path);
625 #endif
626 
627 	error = kern_unlink(td, path, UIO_SYSSPACE);
628 	if (error == EPERM)
629 		/* Introduce POSIX noncompliant behaviour of Linux */
630 		if (kern_stat(td, path, UIO_SYSSPACE, &st) == 0)
631 			if (S_ISDIR(st.st_mode))
632 				error = EISDIR;
633 	LFREEPATH(path);
634 	return (error);
635 }
636 
637 int
638 linux_unlinkat(struct thread *td, struct linux_unlinkat_args *args)
639 {
640 	char *path;
641 	int error, dfd;
642 	struct stat st;
643 
644 	if (args->flag & ~LINUX_AT_REMOVEDIR)
645 		return (EINVAL);
646 
647 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
648 	LCONVPATHEXIST_AT(td, args->pathname, &path, dfd);
649 
650 #ifdef DEBUG
651 	if (ldebug(unlinkat))
652 		printf(ARGS(unlinkat, "%s"), path);
653 #endif
654 
655 	if (args->flag & LINUX_AT_REMOVEDIR)
656 		error = kern_rmdirat(td, dfd, path, UIO_SYSSPACE);
657 	else
658 		error = kern_unlinkat(td, dfd, path, UIO_SYSSPACE, 0);
659 	if (error == EPERM && !(args->flag & LINUX_AT_REMOVEDIR)) {
660 		/* Introduce POSIX noncompliant behaviour of Linux */
661 		if (kern_statat(td, AT_SYMLINK_NOFOLLOW, dfd, path,
662 		    UIO_SYSSPACE, &st) == 0 && S_ISDIR(st.st_mode))
663 			error = EISDIR;
664 	}
665 	LFREEPATH(path);
666 	return (error);
667 }
668 int
669 linux_chdir(struct thread *td, struct linux_chdir_args *args)
670 {
671 	char *path;
672 	int error;
673 
674 	LCONVPATHEXIST(td, args->path, &path);
675 
676 #ifdef DEBUG
677 	if (ldebug(chdir))
678 		printf(ARGS(chdir, "%s"), path);
679 #endif
680 	error = kern_chdir(td, path, UIO_SYSSPACE);
681 	LFREEPATH(path);
682 	return (error);
683 }
684 
685 int
686 linux_chmod(struct thread *td, struct linux_chmod_args *args)
687 {
688 	char *path;
689 	int error;
690 
691 	LCONVPATHEXIST(td, args->path, &path);
692 
693 #ifdef DEBUG
694 	if (ldebug(chmod))
695 		printf(ARGS(chmod, "%s, %d"), path, args->mode);
696 #endif
697 	error = kern_chmod(td, path, UIO_SYSSPACE, args->mode);
698 	LFREEPATH(path);
699 	return (error);
700 }
701 
702 int
703 linux_fchmodat(struct thread *td, struct linux_fchmodat_args *args)
704 {
705 	char *path;
706 	int error, dfd;
707 
708 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
709 	LCONVPATHEXIST_AT(td, args->filename, &path, dfd);
710 
711 #ifdef DEBUG
712 	if (ldebug(fchmodat))
713 		printf(ARGS(fchmodat, "%s, %d"), path, args->mode);
714 #endif
715 
716 	error = kern_fchmodat(td, dfd, path, UIO_SYSSPACE, args->mode, 0);
717 	LFREEPATH(path);
718 	return (error);
719 }
720 
721 int
722 linux_mkdir(struct thread *td, struct linux_mkdir_args *args)
723 {
724 	char *path;
725 	int error;
726 
727 	LCONVPATHCREAT(td, args->path, &path);
728 
729 #ifdef DEBUG
730 	if (ldebug(mkdir))
731 		printf(ARGS(mkdir, "%s, %d"), path, args->mode);
732 #endif
733 	error = kern_mkdir(td, path, UIO_SYSSPACE, args->mode);
734 	LFREEPATH(path);
735 	return (error);
736 }
737 
738 int
739 linux_mkdirat(struct thread *td, struct linux_mkdirat_args *args)
740 {
741 	char *path;
742 	int error, dfd;
743 
744 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
745 	LCONVPATHCREAT_AT(td, args->pathname, &path, dfd);
746 
747 #ifdef DEBUG
748 	if (ldebug(mkdirat))
749 		printf(ARGS(mkdirat, "%s, %d"), path, args->mode);
750 #endif
751 	error = kern_mkdirat(td, dfd, path, UIO_SYSSPACE, args->mode);
752 	LFREEPATH(path);
753 	return (error);
754 }
755 
756 int
757 linux_rmdir(struct thread *td, struct linux_rmdir_args *args)
758 {
759 	char *path;
760 	int error;
761 
762 	LCONVPATHEXIST(td, args->path, &path);
763 
764 #ifdef DEBUG
765 	if (ldebug(rmdir))
766 		printf(ARGS(rmdir, "%s"), path);
767 #endif
768 	error = kern_rmdir(td, path, UIO_SYSSPACE);
769 	LFREEPATH(path);
770 	return (error);
771 }
772 
773 int
774 linux_rename(struct thread *td, struct linux_rename_args *args)
775 {
776 	char *from, *to;
777 	int error;
778 
779 	LCONVPATHEXIST(td, args->from, &from);
780 	/* Expand LCONVPATHCREATE so that `from' can be freed on errors */
781 	error = linux_emul_convpath(td, args->to, UIO_USERSPACE, &to, 1, AT_FDCWD);
782 	if (to == NULL) {
783 		LFREEPATH(from);
784 		return (error);
785 	}
786 
787 #ifdef DEBUG
788 	if (ldebug(rename))
789 		printf(ARGS(rename, "%s, %s"), from, to);
790 #endif
791 	error = kern_rename(td, from, to, UIO_SYSSPACE);
792 	LFREEPATH(from);
793 	LFREEPATH(to);
794 	return (error);
795 }
796 
797 int
798 linux_renameat(struct thread *td, struct linux_renameat_args *args)
799 {
800 	char *from, *to;
801 	int error, olddfd, newdfd;
802 
803 	olddfd = (args->olddfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->olddfd;
804 	newdfd = (args->newdfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->newdfd;
805 	LCONVPATHEXIST_AT(td, args->oldname, &from, olddfd);
806 	/* Expand LCONVPATHCREATE so that `from' can be freed on errors */
807 	error = linux_emul_convpath(td, args->newname, UIO_USERSPACE, &to, 1, newdfd);
808 	if (to == NULL) {
809 		LFREEPATH(from);
810 		return (error);
811 	}
812 
813 #ifdef DEBUG
814 	if (ldebug(renameat))
815 		printf(ARGS(renameat, "%s, %s"), from, to);
816 #endif
817 	error = kern_renameat(td, olddfd, from, newdfd, to, UIO_SYSSPACE);
818 	LFREEPATH(from);
819 	LFREEPATH(to);
820 	return (error);
821 }
822 
823 int
824 linux_symlink(struct thread *td, struct linux_symlink_args *args)
825 {
826 	char *path, *to;
827 	int error;
828 
829 	LCONVPATHEXIST(td, args->path, &path);
830 	/* Expand LCONVPATHCREATE so that `path' can be freed on errors */
831 	error = linux_emul_convpath(td, args->to, UIO_USERSPACE, &to, 1, AT_FDCWD);
832 	if (to == NULL) {
833 		LFREEPATH(path);
834 		return (error);
835 	}
836 
837 #ifdef DEBUG
838 	if (ldebug(symlink))
839 		printf(ARGS(symlink, "%s, %s"), path, to);
840 #endif
841 	error = kern_symlink(td, path, to, UIO_SYSSPACE);
842 	LFREEPATH(path);
843 	LFREEPATH(to);
844 	return (error);
845 }
846 
847 int
848 linux_symlinkat(struct thread *td, struct linux_symlinkat_args *args)
849 {
850 	char *path, *to;
851 	int error, dfd;
852 
853 	dfd = (args->newdfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->newdfd;
854 	LCONVPATHEXIST_AT(td, args->oldname, &path, dfd);
855 	/* Expand LCONVPATHCREATE so that `path' can be freed on errors */
856 	error = linux_emul_convpath(td, args->newname, UIO_USERSPACE, &to, 1, dfd);
857 	if (to == NULL) {
858 		LFREEPATH(path);
859 		return (error);
860 	}
861 
862 #ifdef DEBUG
863 	if (ldebug(symlinkat))
864 		printf(ARGS(symlinkat, "%s, %s"), path, to);
865 #endif
866 
867 	error = kern_symlinkat(td, path, dfd, to, UIO_SYSSPACE);
868 	LFREEPATH(path);
869 	LFREEPATH(to);
870 	return (error);
871 }
872 
873 int
874 linux_readlink(struct thread *td, struct linux_readlink_args *args)
875 {
876 	char *name;
877 	int error;
878 
879 	LCONVPATHEXIST(td, args->name, &name);
880 
881 #ifdef DEBUG
882 	if (ldebug(readlink))
883 		printf(ARGS(readlink, "%s, %p, %d"), name, (void *)args->buf,
884 		    args->count);
885 #endif
886 	error = kern_readlink(td, name, UIO_SYSSPACE, args->buf, UIO_USERSPACE,
887 	    args->count);
888 	LFREEPATH(name);
889 	return (error);
890 }
891 
892 int
893 linux_readlinkat(struct thread *td, struct linux_readlinkat_args *args)
894 {
895 	char *name;
896 	int error, dfd;
897 
898 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
899 	LCONVPATHEXIST_AT(td, args->path, &name, dfd);
900 
901 #ifdef DEBUG
902 	if (ldebug(readlinkat))
903 		printf(ARGS(readlinkat, "%s, %p, %d"), name, (void *)args->buf,
904 		    args->bufsiz);
905 #endif
906 
907 	error = kern_readlinkat(td, dfd, name, UIO_SYSSPACE, args->buf,
908 	    UIO_USERSPACE, args->bufsiz);
909 	LFREEPATH(name);
910 	return (error);
911 }
912 
913 int
914 linux_truncate(struct thread *td, struct linux_truncate_args *args)
915 {
916 	char *path;
917 	int error;
918 
919 	LCONVPATHEXIST(td, args->path, &path);
920 
921 #ifdef DEBUG
922 	if (ldebug(truncate))
923 		printf(ARGS(truncate, "%s, %ld"), path, (long)args->length);
924 #endif
925 
926 	error = kern_truncate(td, path, UIO_SYSSPACE, args->length);
927 	LFREEPATH(path);
928 	return (error);
929 }
930 
931 int
932 linux_truncate64(struct thread *td, struct linux_truncate64_args *args)
933 {
934 	char *path;
935 	int error;
936 
937 	LCONVPATHEXIST(td, args->path, &path);
938 
939 #ifdef DEBUG
940 	if (ldebug(truncate64))
941 		printf(ARGS(truncate64, "%s, %jd"), path, args->length);
942 #endif
943 
944 	error = kern_truncate(td, path, UIO_SYSSPACE, args->length);
945 	LFREEPATH(path);
946 	return (error);
947 }
948 int
949 linux_ftruncate(struct thread *td, struct linux_ftruncate_args *args)
950 {
951 	struct ftruncate_args /* {
952 		int fd;
953 		int pad;
954 		off_t length;
955 		} */ nuap;
956 
957 	nuap.fd = args->fd;
958 	nuap.length = args->length;
959 	return (sys_ftruncate(td, &nuap));
960 }
961 
962 int
963 linux_link(struct thread *td, struct linux_link_args *args)
964 {
965 	char *path, *to;
966 	int error;
967 
968 	LCONVPATHEXIST(td, args->path, &path);
969 	/* Expand LCONVPATHCREATE so that `path' can be freed on errors */
970 	error = linux_emul_convpath(td, args->to, UIO_USERSPACE, &to, 1, AT_FDCWD);
971 	if (to == NULL) {
972 		LFREEPATH(path);
973 		return (error);
974 	}
975 
976 #ifdef DEBUG
977 	if (ldebug(link))
978 		printf(ARGS(link, "%s, %s"), path, to);
979 #endif
980 	error = kern_link(td, path, to, UIO_SYSSPACE);
981 	LFREEPATH(path);
982 	LFREEPATH(to);
983 	return (error);
984 }
985 
986 int
987 linux_linkat(struct thread *td, struct linux_linkat_args *args)
988 {
989 	char *path, *to;
990 	int error, olddfd, newdfd, follow;
991 
992 	if (args->flag & ~LINUX_AT_SYMLINK_FOLLOW)
993 		return (EINVAL);
994 
995 	olddfd = (args->olddfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->olddfd;
996 	newdfd = (args->newdfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->newdfd;
997 	LCONVPATHEXIST_AT(td, args->oldname, &path, olddfd);
998 	/* Expand LCONVPATHCREATE so that `path' can be freed on errors */
999 	error = linux_emul_convpath(td, args->newname, UIO_USERSPACE, &to, 1, newdfd);
1000 	if (to == NULL) {
1001 		LFREEPATH(path);
1002 		return (error);
1003 	}
1004 
1005 #ifdef DEBUG
1006 	if (ldebug(linkat))
1007 		printf(ARGS(linkat, "%i, %s, %i, %s, %i"), args->olddfd, path,
1008 			args->newdfd, to, args->flag);
1009 #endif
1010 
1011 	follow = (args->flag & LINUX_AT_SYMLINK_FOLLOW) == 0 ? NOFOLLOW :
1012 	    FOLLOW;
1013 	error = kern_linkat(td, olddfd, newdfd, path, to, UIO_SYSSPACE, follow);
1014 	LFREEPATH(path);
1015 	LFREEPATH(to);
1016 	return (error);
1017 }
1018 
1019 int
1020 linux_fdatasync(td, uap)
1021 	struct thread *td;
1022 	struct linux_fdatasync_args *uap;
1023 {
1024 	struct fsync_args bsd;
1025 
1026 	bsd.fd = uap->fd;
1027 	return sys_fsync(td, &bsd);
1028 }
1029 
1030 int
1031 linux_pread(td, uap)
1032 	struct thread *td;
1033 	struct linux_pread_args *uap;
1034 {
1035 	struct pread_args bsd;
1036 	struct vnode *vp;
1037 	int error;
1038 
1039 	bsd.fd = uap->fd;
1040 	bsd.buf = uap->buf;
1041 	bsd.nbyte = uap->nbyte;
1042 	bsd.offset = uap->offset;
1043 
1044 	error = sys_pread(td, &bsd);
1045 
1046 	if (error == 0) {
1047    	   	/* This seems to violate POSIX but linux does it */
1048 		if ((error = fgetvp(td, uap->fd, CAP_READ, &vp)) != 0)
1049    		   	return (error);
1050 		if (vp->v_type == VDIR) {
1051    		   	vrele(vp);
1052 			return (EISDIR);
1053 		}
1054 		vrele(vp);
1055 	}
1056 
1057 	return (error);
1058 }
1059 
1060 int
1061 linux_pwrite(td, uap)
1062 	struct thread *td;
1063 	struct linux_pwrite_args *uap;
1064 {
1065 	struct pwrite_args bsd;
1066 
1067 	bsd.fd = uap->fd;
1068 	bsd.buf = uap->buf;
1069 	bsd.nbyte = uap->nbyte;
1070 	bsd.offset = uap->offset;
1071 	return sys_pwrite(td, &bsd);
1072 }
1073 
1074 int
1075 linux_mount(struct thread *td, struct linux_mount_args *args)
1076 {
1077 	struct ufs_args ufs;
1078 	char fstypename[MFSNAMELEN];
1079 	char mntonname[MNAMELEN], mntfromname[MNAMELEN];
1080 	int error;
1081 	int fsflags;
1082 	void *fsdata;
1083 
1084 	error = copyinstr(args->filesystemtype, fstypename, MFSNAMELEN - 1,
1085 	    NULL);
1086 	if (error)
1087 		return (error);
1088 	error = copyinstr(args->specialfile, mntfromname, MNAMELEN - 1, NULL);
1089 	if (error)
1090 		return (error);
1091 	error = copyinstr(args->dir, mntonname, MNAMELEN - 1, NULL);
1092 	if (error)
1093 		return (error);
1094 
1095 #ifdef DEBUG
1096 	if (ldebug(mount))
1097 		printf(ARGS(mount, "%s, %s, %s"),
1098 		    fstypename, mntfromname, mntonname);
1099 #endif
1100 
1101 	if (strcmp(fstypename, "ext2") == 0) {
1102 		strcpy(fstypename, "ext2fs");
1103 		fsdata = &ufs;
1104 		ufs.fspec = mntfromname;
1105 #define DEFAULT_ROOTID		-2
1106 		ufs.export.ex_root = DEFAULT_ROOTID;
1107 		ufs.export.ex_flags =
1108 		    args->rwflag & LINUX_MS_RDONLY ? MNT_EXRDONLY : 0;
1109 	} else if (strcmp(fstypename, "proc") == 0) {
1110 		strcpy(fstypename, "linprocfs");
1111 		fsdata = NULL;
1112 	} else if (strcmp(fstypename, "vfat") == 0) {
1113 		strcpy(fstypename, "msdosfs");
1114 		fsdata = NULL;
1115 	} else {
1116 		return (ENODEV);
1117 	}
1118 
1119 	fsflags = 0;
1120 
1121 	if ((args->rwflag & 0xffff0000) == 0xc0ed0000) {
1122 		/*
1123 		 * Linux SYNC flag is not included; the closest equivalent
1124 		 * FreeBSD has is !ASYNC, which is our default.
1125 		 */
1126 		if (args->rwflag & LINUX_MS_RDONLY)
1127 			fsflags |= MNT_RDONLY;
1128 		if (args->rwflag & LINUX_MS_NOSUID)
1129 			fsflags |= MNT_NOSUID;
1130 		if (args->rwflag & LINUX_MS_NOEXEC)
1131 			fsflags |= MNT_NOEXEC;
1132 		if (args->rwflag & LINUX_MS_REMOUNT)
1133 			fsflags |= MNT_UPDATE;
1134 	}
1135 
1136 	if (strcmp(fstypename, "linprocfs") == 0) {
1137 		error = kernel_vmount(fsflags,
1138 			"fstype", fstypename,
1139 			"fspath", mntonname,
1140 			NULL);
1141 	} else if (strcmp(fstypename, "msdosfs") == 0) {
1142 		error = kernel_vmount(fsflags,
1143 			"fstype", fstypename,
1144 			"fspath", mntonname,
1145 			"from", mntfromname,
1146 			NULL);
1147 	} else
1148 		error = EOPNOTSUPP;
1149 	return (error);
1150 }
1151 
1152 int
1153 linux_oldumount(struct thread *td, struct linux_oldumount_args *args)
1154 {
1155 	struct linux_umount_args args2;
1156 
1157 	args2.path = args->path;
1158 	args2.flags = 0;
1159 	return (linux_umount(td, &args2));
1160 }
1161 
1162 int
1163 linux_umount(struct thread *td, struct linux_umount_args *args)
1164 {
1165 	struct unmount_args bsd;
1166 
1167 	bsd.path = args->path;
1168 	bsd.flags = args->flags;	/* XXX correct? */
1169 	return (sys_unmount(td, &bsd));
1170 }
1171 
1172 /*
1173  * fcntl family of syscalls
1174  */
1175 
1176 struct l_flock {
1177 	l_short		l_type;
1178 	l_short		l_whence;
1179 	l_off_t		l_start;
1180 	l_off_t		l_len;
1181 	l_pid_t		l_pid;
1182 }
1183 #if defined(__amd64__) && defined(COMPAT_LINUX32)
1184 __packed
1185 #endif
1186 ;
1187 
1188 static void
1189 linux_to_bsd_flock(struct l_flock *linux_flock, struct flock *bsd_flock)
1190 {
1191 	switch (linux_flock->l_type) {
1192 	case LINUX_F_RDLCK:
1193 		bsd_flock->l_type = F_RDLCK;
1194 		break;
1195 	case LINUX_F_WRLCK:
1196 		bsd_flock->l_type = F_WRLCK;
1197 		break;
1198 	case LINUX_F_UNLCK:
1199 		bsd_flock->l_type = F_UNLCK;
1200 		break;
1201 	default:
1202 		bsd_flock->l_type = -1;
1203 		break;
1204 	}
1205 	bsd_flock->l_whence = linux_flock->l_whence;
1206 	bsd_flock->l_start = (off_t)linux_flock->l_start;
1207 	bsd_flock->l_len = (off_t)linux_flock->l_len;
1208 	bsd_flock->l_pid = (pid_t)linux_flock->l_pid;
1209 	bsd_flock->l_sysid = 0;
1210 }
1211 
1212 static void
1213 bsd_to_linux_flock(struct flock *bsd_flock, struct l_flock *linux_flock)
1214 {
1215 	switch (bsd_flock->l_type) {
1216 	case F_RDLCK:
1217 		linux_flock->l_type = LINUX_F_RDLCK;
1218 		break;
1219 	case F_WRLCK:
1220 		linux_flock->l_type = LINUX_F_WRLCK;
1221 		break;
1222 	case F_UNLCK:
1223 		linux_flock->l_type = LINUX_F_UNLCK;
1224 		break;
1225 	}
1226 	linux_flock->l_whence = bsd_flock->l_whence;
1227 	linux_flock->l_start = (l_off_t)bsd_flock->l_start;
1228 	linux_flock->l_len = (l_off_t)bsd_flock->l_len;
1229 	linux_flock->l_pid = (l_pid_t)bsd_flock->l_pid;
1230 }
1231 
1232 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1233 struct l_flock64 {
1234 	l_short		l_type;
1235 	l_short		l_whence;
1236 	l_loff_t	l_start;
1237 	l_loff_t	l_len;
1238 	l_pid_t		l_pid;
1239 }
1240 #if defined(__amd64__) && defined(COMPAT_LINUX32)
1241 __packed
1242 #endif
1243 ;
1244 
1245 static void
1246 linux_to_bsd_flock64(struct l_flock64 *linux_flock, struct flock *bsd_flock)
1247 {
1248 	switch (linux_flock->l_type) {
1249 	case LINUX_F_RDLCK:
1250 		bsd_flock->l_type = F_RDLCK;
1251 		break;
1252 	case LINUX_F_WRLCK:
1253 		bsd_flock->l_type = F_WRLCK;
1254 		break;
1255 	case LINUX_F_UNLCK:
1256 		bsd_flock->l_type = F_UNLCK;
1257 		break;
1258 	default:
1259 		bsd_flock->l_type = -1;
1260 		break;
1261 	}
1262 	bsd_flock->l_whence = linux_flock->l_whence;
1263 	bsd_flock->l_start = (off_t)linux_flock->l_start;
1264 	bsd_flock->l_len = (off_t)linux_flock->l_len;
1265 	bsd_flock->l_pid = (pid_t)linux_flock->l_pid;
1266 	bsd_flock->l_sysid = 0;
1267 }
1268 
1269 static void
1270 bsd_to_linux_flock64(struct flock *bsd_flock, struct l_flock64 *linux_flock)
1271 {
1272 	switch (bsd_flock->l_type) {
1273 	case F_RDLCK:
1274 		linux_flock->l_type = LINUX_F_RDLCK;
1275 		break;
1276 	case F_WRLCK:
1277 		linux_flock->l_type = LINUX_F_WRLCK;
1278 		break;
1279 	case F_UNLCK:
1280 		linux_flock->l_type = LINUX_F_UNLCK;
1281 		break;
1282 	}
1283 	linux_flock->l_whence = bsd_flock->l_whence;
1284 	linux_flock->l_start = (l_loff_t)bsd_flock->l_start;
1285 	linux_flock->l_len = (l_loff_t)bsd_flock->l_len;
1286 	linux_flock->l_pid = (l_pid_t)bsd_flock->l_pid;
1287 }
1288 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1289 
1290 static int
1291 fcntl_common(struct thread *td, struct linux_fcntl64_args *args)
1292 {
1293 	struct l_flock linux_flock;
1294 	struct flock bsd_flock;
1295 	struct file *fp;
1296 	long arg;
1297 	int error, result;
1298 
1299 	switch (args->cmd) {
1300 	case LINUX_F_DUPFD:
1301 		return (kern_fcntl(td, args->fd, F_DUPFD, args->arg));
1302 
1303 	case LINUX_F_GETFD:
1304 		return (kern_fcntl(td, args->fd, F_GETFD, 0));
1305 
1306 	case LINUX_F_SETFD:
1307 		return (kern_fcntl(td, args->fd, F_SETFD, args->arg));
1308 
1309 	case LINUX_F_GETFL:
1310 		error = kern_fcntl(td, args->fd, F_GETFL, 0);
1311 		result = td->td_retval[0];
1312 		td->td_retval[0] = 0;
1313 		if (result & O_RDONLY)
1314 			td->td_retval[0] |= LINUX_O_RDONLY;
1315 		if (result & O_WRONLY)
1316 			td->td_retval[0] |= LINUX_O_WRONLY;
1317 		if (result & O_RDWR)
1318 			td->td_retval[0] |= LINUX_O_RDWR;
1319 		if (result & O_NDELAY)
1320 			td->td_retval[0] |= LINUX_O_NONBLOCK;
1321 		if (result & O_APPEND)
1322 			td->td_retval[0] |= LINUX_O_APPEND;
1323 		if (result & O_FSYNC)
1324 			td->td_retval[0] |= LINUX_O_SYNC;
1325 		if (result & O_ASYNC)
1326 			td->td_retval[0] |= LINUX_FASYNC;
1327 #ifdef LINUX_O_NOFOLLOW
1328 		if (result & O_NOFOLLOW)
1329 			td->td_retval[0] |= LINUX_O_NOFOLLOW;
1330 #endif
1331 #ifdef LINUX_O_DIRECT
1332 		if (result & O_DIRECT)
1333 			td->td_retval[0] |= LINUX_O_DIRECT;
1334 #endif
1335 		return (error);
1336 
1337 	case LINUX_F_SETFL:
1338 		arg = 0;
1339 		if (args->arg & LINUX_O_NDELAY)
1340 			arg |= O_NONBLOCK;
1341 		if (args->arg & LINUX_O_APPEND)
1342 			arg |= O_APPEND;
1343 		if (args->arg & LINUX_O_SYNC)
1344 			arg |= O_FSYNC;
1345 		if (args->arg & LINUX_FASYNC)
1346 			arg |= O_ASYNC;
1347 #ifdef LINUX_O_NOFOLLOW
1348 		if (args->arg & LINUX_O_NOFOLLOW)
1349 			arg |= O_NOFOLLOW;
1350 #endif
1351 #ifdef LINUX_O_DIRECT
1352 		if (args->arg & LINUX_O_DIRECT)
1353 			arg |= O_DIRECT;
1354 #endif
1355 		return (kern_fcntl(td, args->fd, F_SETFL, arg));
1356 
1357 	case LINUX_F_GETLK:
1358 		error = copyin((void *)args->arg, &linux_flock,
1359 		    sizeof(linux_flock));
1360 		if (error)
1361 			return (error);
1362 		linux_to_bsd_flock(&linux_flock, &bsd_flock);
1363 		error = kern_fcntl(td, args->fd, F_GETLK, (intptr_t)&bsd_flock);
1364 		if (error)
1365 			return (error);
1366 		bsd_to_linux_flock(&bsd_flock, &linux_flock);
1367 		return (copyout(&linux_flock, (void *)args->arg,
1368 		    sizeof(linux_flock)));
1369 
1370 	case LINUX_F_SETLK:
1371 		error = copyin((void *)args->arg, &linux_flock,
1372 		    sizeof(linux_flock));
1373 		if (error)
1374 			return (error);
1375 		linux_to_bsd_flock(&linux_flock, &bsd_flock);
1376 		return (kern_fcntl(td, args->fd, F_SETLK,
1377 		    (intptr_t)&bsd_flock));
1378 
1379 	case LINUX_F_SETLKW:
1380 		error = copyin((void *)args->arg, &linux_flock,
1381 		    sizeof(linux_flock));
1382 		if (error)
1383 			return (error);
1384 		linux_to_bsd_flock(&linux_flock, &bsd_flock);
1385 		return (kern_fcntl(td, args->fd, F_SETLKW,
1386 		     (intptr_t)&bsd_flock));
1387 
1388 	case LINUX_F_GETOWN:
1389 		return (kern_fcntl(td, args->fd, F_GETOWN, 0));
1390 
1391 	case LINUX_F_SETOWN:
1392 		/*
1393 		 * XXX some Linux applications depend on F_SETOWN having no
1394 		 * significant effect for pipes (SIGIO is not delivered for
1395 		 * pipes under Linux-2.2.35 at least).
1396 		 */
1397 		error = fget(td, args->fd, CAP_FCNTL, &fp);
1398 		if (error)
1399 			return (error);
1400 		if (fp->f_type == DTYPE_PIPE) {
1401 			fdrop(fp, td);
1402 			return (EINVAL);
1403 		}
1404 		fdrop(fp, td);
1405 
1406 		return (kern_fcntl(td, args->fd, F_SETOWN, args->arg));
1407 	}
1408 
1409 	return (EINVAL);
1410 }
1411 
1412 int
1413 linux_fcntl(struct thread *td, struct linux_fcntl_args *args)
1414 {
1415 	struct linux_fcntl64_args args64;
1416 
1417 #ifdef DEBUG
1418 	if (ldebug(fcntl))
1419 		printf(ARGS(fcntl, "%d, %08x, *"), args->fd, args->cmd);
1420 #endif
1421 
1422 	args64.fd = args->fd;
1423 	args64.cmd = args->cmd;
1424 	args64.arg = args->arg;
1425 	return (fcntl_common(td, &args64));
1426 }
1427 
1428 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1429 int
1430 linux_fcntl64(struct thread *td, struct linux_fcntl64_args *args)
1431 {
1432 	struct l_flock64 linux_flock;
1433 	struct flock bsd_flock;
1434 	int error;
1435 
1436 #ifdef DEBUG
1437 	if (ldebug(fcntl64))
1438 		printf(ARGS(fcntl64, "%d, %08x, *"), args->fd, args->cmd);
1439 #endif
1440 
1441 	switch (args->cmd) {
1442 	case LINUX_F_GETLK64:
1443 		error = copyin((void *)args->arg, &linux_flock,
1444 		    sizeof(linux_flock));
1445 		if (error)
1446 			return (error);
1447 		linux_to_bsd_flock64(&linux_flock, &bsd_flock);
1448 		error = kern_fcntl(td, args->fd, F_GETLK, (intptr_t)&bsd_flock);
1449 		if (error)
1450 			return (error);
1451 		bsd_to_linux_flock64(&bsd_flock, &linux_flock);
1452 		return (copyout(&linux_flock, (void *)args->arg,
1453 			    sizeof(linux_flock)));
1454 
1455 	case LINUX_F_SETLK64:
1456 		error = copyin((void *)args->arg, &linux_flock,
1457 		    sizeof(linux_flock));
1458 		if (error)
1459 			return (error);
1460 		linux_to_bsd_flock64(&linux_flock, &bsd_flock);
1461 		return (kern_fcntl(td, args->fd, F_SETLK,
1462 		    (intptr_t)&bsd_flock));
1463 
1464 	case LINUX_F_SETLKW64:
1465 		error = copyin((void *)args->arg, &linux_flock,
1466 		    sizeof(linux_flock));
1467 		if (error)
1468 			return (error);
1469 		linux_to_bsd_flock64(&linux_flock, &bsd_flock);
1470 		return (kern_fcntl(td, args->fd, F_SETLKW,
1471 		    (intptr_t)&bsd_flock));
1472 	}
1473 
1474 	return (fcntl_common(td, args));
1475 }
1476 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1477 
1478 int
1479 linux_chown(struct thread *td, struct linux_chown_args *args)
1480 {
1481 	char *path;
1482 	int error;
1483 
1484 	LCONVPATHEXIST(td, args->path, &path);
1485 
1486 #ifdef DEBUG
1487 	if (ldebug(chown))
1488 		printf(ARGS(chown, "%s, %d, %d"), path, args->uid, args->gid);
1489 #endif
1490 	error = kern_chown(td, path, UIO_SYSSPACE, args->uid, args->gid);
1491 	LFREEPATH(path);
1492 	return (error);
1493 }
1494 
1495 int
1496 linux_fchownat(struct thread *td, struct linux_fchownat_args *args)
1497 {
1498 	char *path;
1499 	int error, dfd, flag;
1500 
1501 	if (args->flag & ~LINUX_AT_SYMLINK_NOFOLLOW)
1502 		return (EINVAL);
1503 
1504 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD :  args->dfd;
1505 	LCONVPATHEXIST_AT(td, args->filename, &path, dfd);
1506 
1507 #ifdef DEBUG
1508 	if (ldebug(fchownat))
1509 		printf(ARGS(fchownat, "%s, %d, %d"), path, args->uid, args->gid);
1510 #endif
1511 
1512 	flag = (args->flag & LINUX_AT_SYMLINK_NOFOLLOW) == 0 ? 0 :
1513 	    AT_SYMLINK_NOFOLLOW;
1514 	error = kern_fchownat(td, dfd, path, UIO_SYSSPACE, args->uid, args->gid,
1515 	    flag);
1516 	LFREEPATH(path);
1517 	return (error);
1518 }
1519 
1520 int
1521 linux_lchown(struct thread *td, struct linux_lchown_args *args)
1522 {
1523 	char *path;
1524 	int error;
1525 
1526 	LCONVPATHEXIST(td, args->path, &path);
1527 
1528 #ifdef DEBUG
1529 	if (ldebug(lchown))
1530 		printf(ARGS(lchown, "%s, %d, %d"), path, args->uid, args->gid);
1531 #endif
1532 	error = kern_lchown(td, path, UIO_SYSSPACE, args->uid, args->gid);
1533 	LFREEPATH(path);
1534 	return (error);
1535 }
1536 
1537 static int
1538 convert_fadvice(int advice)
1539 {
1540 	switch (advice) {
1541 	case LINUX_POSIX_FADV_NORMAL:
1542 		return (POSIX_FADV_NORMAL);
1543 	case LINUX_POSIX_FADV_RANDOM:
1544 		return (POSIX_FADV_RANDOM);
1545 	case LINUX_POSIX_FADV_SEQUENTIAL:
1546 		return (POSIX_FADV_SEQUENTIAL);
1547 	case LINUX_POSIX_FADV_WILLNEED:
1548 		return (POSIX_FADV_WILLNEED);
1549 	case LINUX_POSIX_FADV_DONTNEED:
1550 		return (POSIX_FADV_DONTNEED);
1551 	case LINUX_POSIX_FADV_NOREUSE:
1552 		return (POSIX_FADV_NOREUSE);
1553 	default:
1554 		return (-1);
1555 	}
1556 }
1557 
1558 int
1559 linux_fadvise64(struct thread *td, struct linux_fadvise64_args *args)
1560 {
1561 	int advice;
1562 
1563 	advice = convert_fadvice(args->advice);
1564 	if (advice == -1)
1565 		return (EINVAL);
1566 	return (kern_posix_fadvise(td, args->fd, args->offset, args->len,
1567 	    advice));
1568 }
1569 
1570 int
1571 linux_fadvise64_64(struct thread *td, struct linux_fadvise64_64_args *args)
1572 {
1573 	int advice;
1574 
1575 	advice = convert_fadvice(args->advice);
1576 	if (advice == -1)
1577 		return (EINVAL);
1578 	return (kern_posix_fadvise(td, args->fd, args->offset, args->len,
1579 	    advice));
1580 }
1581 
1582 int
1583 linux_pipe(struct thread *td, struct linux_pipe_args *args)
1584 {
1585 	int fildes[2];
1586 	int error;
1587 
1588 #ifdef DEBUG
1589 	if (ldebug(pipe))
1590 		printf(ARGS(pipe, "*"));
1591 #endif
1592 
1593 	error = do_pipe(td, fildes, 0);
1594 	if (error)
1595 		return (error);
1596 
1597 	/* XXX: Close descriptors on error. */
1598 	return (copyout(fildes, args->pipefds, sizeof(fildes)));
1599 }
1600 
1601 int
1602 linux_pipe2(struct thread *td, struct linux_pipe2_args *args)
1603 {
1604 	int fildes[2];
1605 	int error, flags;
1606 
1607 #ifdef DEBUG
1608 	if (ldebug(pipe2))
1609 		printf(ARGS(pipe2, "*, %d"), args->flags);
1610 #endif
1611 
1612 	if ((args->flags & ~(LINUX_O_NONBLOCK | LINUX_O_CLOEXEC)) != 0)
1613 		return (EINVAL);
1614 
1615 	flags = 0;
1616 	if ((args->flags & LINUX_O_NONBLOCK) != 0)
1617 		flags |= O_NONBLOCK;
1618 	if ((args->flags & LINUX_O_CLOEXEC) != 0)
1619 		flags |= O_CLOEXEC;
1620 	error = do_pipe(td, fildes, flags);
1621 	if (error)
1622 		return (error);
1623 
1624 	/* XXX: Close descriptors on error. */
1625 	return (copyout(fildes, args->pipefds, sizeof(fildes)));
1626 }
1627