xref: /freebsd/sys/compat/linux/linux_file.c (revision c243e4902be8df1e643c76b5f18b68bb77cc5268)
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 	if (cookies) {
388 		free(cookies, M_TEMP);
389 		cookies = NULL;
390 	}
391 
392 #ifdef MAC
393 	/*
394 	 * Do directory search MAC check using non-cached credentials.
395 	 */
396 	if ((error = mac_vnode_check_readdir(td->td_ucred, vp)))
397 		goto out;
398 #endif /* MAC */
399 	if ((error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, &ncookies,
400 		 &cookies)))
401 		goto out;
402 
403 	inp = buf;
404 	outp = (caddr_t)args->dirent;
405 	resid = nbytes;
406 	if ((len = buflen - auio.uio_resid) <= 0)
407 		goto eof;
408 
409 	cookiep = cookies;
410 
411 	if (cookies) {
412 		/*
413 		 * When using cookies, the vfs has the option of reading from
414 		 * a different offset than that supplied (UFS truncates the
415 		 * offset to a block boundary to make sure that it never reads
416 		 * partway through a directory entry, even if the directory
417 		 * has been compacted).
418 		 */
419 		while (len > 0 && ncookies > 0 && *cookiep <= off) {
420 			bdp = (struct dirent *) inp;
421 			len -= bdp->d_reclen;
422 			inp += bdp->d_reclen;
423 			cookiep++;
424 			ncookies--;
425 		}
426 	}
427 
428 	while (len > 0) {
429 		if (cookiep && ncookies == 0)
430 			break;
431 		bdp = (struct dirent *) inp;
432 		reclen = bdp->d_reclen;
433 		if (reclen & 3) {
434 			error = EFAULT;
435 			goto out;
436 		}
437 
438 		if (bdp->d_fileno == 0) {
439 			inp += reclen;
440 			if (cookiep) {
441 				off = *cookiep++;
442 				ncookies--;
443 			} else
444 				off += reclen;
445 
446 			len -= reclen;
447 			continue;
448 		}
449 
450 		linuxreclen = (is64bit)
451 		    ? LINUX_RECLEN64(bdp->d_namlen)
452 		    : LINUX_RECLEN(bdp->d_namlen);
453 
454 		if (reclen > len || resid < linuxreclen) {
455 			outp++;
456 			break;
457 		}
458 
459 		if (justone) {
460 			/* readdir(2) case. */
461 			linux_dirent = (struct l_dirent*)lbuf;
462 			linux_dirent->d_ino = bdp->d_fileno;
463 			linux_dirent->d_off = (l_off_t)linuxreclen;
464 			linux_dirent->d_reclen = (l_ushort)bdp->d_namlen;
465 			strlcpy(linux_dirent->d_name, bdp->d_name,
466 			    linuxreclen - offsetof(struct l_dirent, d_name));
467 			error = copyout(linux_dirent, outp, linuxreclen);
468 		}
469 		if (is64bit) {
470 			linux_dirent64 = (struct l_dirent64*)lbuf;
471 			linux_dirent64->d_ino = bdp->d_fileno;
472 			linux_dirent64->d_off = (cookiep)
473 			    ? (l_off_t)*cookiep
474 			    : (l_off_t)(off + reclen);
475 			linux_dirent64->d_reclen = (l_ushort)linuxreclen;
476 			linux_dirent64->d_type = bdp->d_type;
477 			strlcpy(linux_dirent64->d_name, bdp->d_name,
478 			    linuxreclen - offsetof(struct l_dirent64, d_name));
479 			error = copyout(linux_dirent64, outp, linuxreclen);
480 		} else if (!justone) {
481 			linux_dirent = (struct l_dirent*)lbuf;
482 			linux_dirent->d_ino = bdp->d_fileno;
483 			linux_dirent->d_off = (cookiep)
484 			    ? (l_off_t)*cookiep
485 			    : (l_off_t)(off + reclen);
486 			linux_dirent->d_reclen = (l_ushort)linuxreclen;
487 			/*
488 			 * Copy d_type to last byte of l_dirent buffer
489 			 */
490 			lbuf[linuxreclen-1] = bdp->d_type;
491 			strlcpy(linux_dirent->d_name, bdp->d_name,
492 			    linuxreclen - offsetof(struct l_dirent, d_name)-1);
493 			error = copyout(linux_dirent, outp, linuxreclen);
494 		}
495 
496 		if (error)
497 			goto out;
498 
499 		inp += reclen;
500 		if (cookiep) {
501 			off = *cookiep++;
502 			ncookies--;
503 		} else
504 			off += reclen;
505 
506 		outp += linuxreclen;
507 		resid -= linuxreclen;
508 		len -= reclen;
509 		if (justone)
510 			break;
511 	}
512 
513 	if (outp == (caddr_t)args->dirent) {
514 		nbytes = resid;
515 		goto eof;
516 	}
517 
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 	foffset_unlock(fp, off, 0);
531 	fdrop(fp, td);
532 	free(buf, M_TEMP);
533 	free(lbuf, M_TEMP);
534 	return (error);
535 }
536 
537 int
538 linux_getdents(struct thread *td, struct linux_getdents_args *args)
539 {
540 
541 #ifdef DEBUG
542 	if (ldebug(getdents))
543 		printf(ARGS(getdents, "%d, *, %d"), args->fd, args->count);
544 #endif
545 
546 	return (getdents_common(td, (struct linux_getdents64_args*)args, 0));
547 }
548 
549 int
550 linux_getdents64(struct thread *td, struct linux_getdents64_args *args)
551 {
552 
553 #ifdef DEBUG
554 	if (ldebug(getdents64))
555 		printf(ARGS(getdents64, "%d, *, %d"), args->fd, args->count);
556 #endif
557 
558 	return (getdents_common(td, args, 1));
559 }
560 
561 /*
562  * These exist mainly for hooks for doing /compat/linux translation.
563  */
564 
565 int
566 linux_access(struct thread *td, struct linux_access_args *args)
567 {
568 	char *path;
569 	int error;
570 
571 	/* linux convention */
572 	if (args->amode & ~(F_OK | X_OK | W_OK | R_OK))
573 		return (EINVAL);
574 
575 	LCONVPATHEXIST(td, args->path, &path);
576 
577 #ifdef DEBUG
578 	if (ldebug(access))
579 		printf(ARGS(access, "%s, %d"), path, args->amode);
580 #endif
581 	error = kern_access(td, path, UIO_SYSSPACE, args->amode);
582 	LFREEPATH(path);
583 
584 	return (error);
585 }
586 
587 int
588 linux_faccessat(struct thread *td, struct linux_faccessat_args *args)
589 {
590 	char *path;
591 	int error, dfd, flag;
592 
593 	if (args->flag & ~LINUX_AT_EACCESS)
594 		return (EINVAL);
595 	/* linux convention */
596 	if (args->amode & ~(F_OK | X_OK | W_OK | R_OK))
597 		return (EINVAL);
598 
599 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
600 	LCONVPATHEXIST_AT(td, args->filename, &path, dfd);
601 
602 #ifdef DEBUG
603 	if (ldebug(access))
604 		printf(ARGS(access, "%s, %d"), path, args->amode);
605 #endif
606 
607 	flag = (args->flag & LINUX_AT_EACCESS) == 0 ? 0 : AT_EACCESS;
608 	error = kern_accessat(td, dfd, path, UIO_SYSSPACE, flag, args->amode);
609 	LFREEPATH(path);
610 
611 	return (error);
612 }
613 
614 int
615 linux_unlink(struct thread *td, struct linux_unlink_args *args)
616 {
617 	char *path;
618 	int error;
619 	struct stat st;
620 
621 	LCONVPATHEXIST(td, args->path, &path);
622 
623 #ifdef DEBUG
624 	if (ldebug(unlink))
625 		printf(ARGS(unlink, "%s"), path);
626 #endif
627 
628 	error = kern_unlink(td, path, UIO_SYSSPACE);
629 	if (error == EPERM)
630 		/* Introduce POSIX noncompliant behaviour of Linux */
631 		if (kern_stat(td, path, UIO_SYSSPACE, &st) == 0)
632 			if (S_ISDIR(st.st_mode))
633 				error = EISDIR;
634 	LFREEPATH(path);
635 	return (error);
636 }
637 
638 int
639 linux_unlinkat(struct thread *td, struct linux_unlinkat_args *args)
640 {
641 	char *path;
642 	int error, dfd;
643 	struct stat st;
644 
645 	if (args->flag & ~LINUX_AT_REMOVEDIR)
646 		return (EINVAL);
647 
648 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
649 	LCONVPATHEXIST_AT(td, args->pathname, &path, dfd);
650 
651 #ifdef DEBUG
652 	if (ldebug(unlinkat))
653 		printf(ARGS(unlinkat, "%s"), path);
654 #endif
655 
656 	if (args->flag & LINUX_AT_REMOVEDIR)
657 		error = kern_rmdirat(td, dfd, path, UIO_SYSSPACE);
658 	else
659 		error = kern_unlinkat(td, dfd, path, UIO_SYSSPACE, 0);
660 	if (error == EPERM && !(args->flag & LINUX_AT_REMOVEDIR)) {
661 		/* Introduce POSIX noncompliant behaviour of Linux */
662 		if (kern_statat(td, AT_SYMLINK_NOFOLLOW, dfd, path,
663 		    UIO_SYSSPACE, &st) == 0 && S_ISDIR(st.st_mode))
664 			error = EISDIR;
665 	}
666 	LFREEPATH(path);
667 	return (error);
668 }
669 int
670 linux_chdir(struct thread *td, struct linux_chdir_args *args)
671 {
672 	char *path;
673 	int error;
674 
675 	LCONVPATHEXIST(td, args->path, &path);
676 
677 #ifdef DEBUG
678 	if (ldebug(chdir))
679 		printf(ARGS(chdir, "%s"), path);
680 #endif
681 	error = kern_chdir(td, path, UIO_SYSSPACE);
682 	LFREEPATH(path);
683 	return (error);
684 }
685 
686 int
687 linux_chmod(struct thread *td, struct linux_chmod_args *args)
688 {
689 	char *path;
690 	int error;
691 
692 	LCONVPATHEXIST(td, args->path, &path);
693 
694 #ifdef DEBUG
695 	if (ldebug(chmod))
696 		printf(ARGS(chmod, "%s, %d"), path, args->mode);
697 #endif
698 	error = kern_chmod(td, path, UIO_SYSSPACE, args->mode);
699 	LFREEPATH(path);
700 	return (error);
701 }
702 
703 int
704 linux_fchmodat(struct thread *td, struct linux_fchmodat_args *args)
705 {
706 	char *path;
707 	int error, dfd;
708 
709 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
710 	LCONVPATHEXIST_AT(td, args->filename, &path, dfd);
711 
712 #ifdef DEBUG
713 	if (ldebug(fchmodat))
714 		printf(ARGS(fchmodat, "%s, %d"), path, args->mode);
715 #endif
716 
717 	error = kern_fchmodat(td, dfd, path, UIO_SYSSPACE, args->mode, 0);
718 	LFREEPATH(path);
719 	return (error);
720 }
721 
722 int
723 linux_mkdir(struct thread *td, struct linux_mkdir_args *args)
724 {
725 	char *path;
726 	int error;
727 
728 	LCONVPATHCREAT(td, args->path, &path);
729 
730 #ifdef DEBUG
731 	if (ldebug(mkdir))
732 		printf(ARGS(mkdir, "%s, %d"), path, args->mode);
733 #endif
734 	error = kern_mkdir(td, path, UIO_SYSSPACE, args->mode);
735 	LFREEPATH(path);
736 	return (error);
737 }
738 
739 int
740 linux_mkdirat(struct thread *td, struct linux_mkdirat_args *args)
741 {
742 	char *path;
743 	int error, dfd;
744 
745 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
746 	LCONVPATHCREAT_AT(td, args->pathname, &path, dfd);
747 
748 #ifdef DEBUG
749 	if (ldebug(mkdirat))
750 		printf(ARGS(mkdirat, "%s, %d"), path, args->mode);
751 #endif
752 	error = kern_mkdirat(td, dfd, path, UIO_SYSSPACE, args->mode);
753 	LFREEPATH(path);
754 	return (error);
755 }
756 
757 int
758 linux_rmdir(struct thread *td, struct linux_rmdir_args *args)
759 {
760 	char *path;
761 	int error;
762 
763 	LCONVPATHEXIST(td, args->path, &path);
764 
765 #ifdef DEBUG
766 	if (ldebug(rmdir))
767 		printf(ARGS(rmdir, "%s"), path);
768 #endif
769 	error = kern_rmdir(td, path, UIO_SYSSPACE);
770 	LFREEPATH(path);
771 	return (error);
772 }
773 
774 int
775 linux_rename(struct thread *td, struct linux_rename_args *args)
776 {
777 	char *from, *to;
778 	int error;
779 
780 	LCONVPATHEXIST(td, args->from, &from);
781 	/* Expand LCONVPATHCREATE so that `from' can be freed on errors */
782 	error = linux_emul_convpath(td, args->to, UIO_USERSPACE, &to, 1, AT_FDCWD);
783 	if (to == NULL) {
784 		LFREEPATH(from);
785 		return (error);
786 	}
787 
788 #ifdef DEBUG
789 	if (ldebug(rename))
790 		printf(ARGS(rename, "%s, %s"), from, to);
791 #endif
792 	error = kern_rename(td, from, to, UIO_SYSSPACE);
793 	LFREEPATH(from);
794 	LFREEPATH(to);
795 	return (error);
796 }
797 
798 int
799 linux_renameat(struct thread *td, struct linux_renameat_args *args)
800 {
801 	char *from, *to;
802 	int error, olddfd, newdfd;
803 
804 	olddfd = (args->olddfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->olddfd;
805 	newdfd = (args->newdfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->newdfd;
806 	LCONVPATHEXIST_AT(td, args->oldname, &from, olddfd);
807 	/* Expand LCONVPATHCREATE so that `from' can be freed on errors */
808 	error = linux_emul_convpath(td, args->newname, UIO_USERSPACE, &to, 1, newdfd);
809 	if (to == NULL) {
810 		LFREEPATH(from);
811 		return (error);
812 	}
813 
814 #ifdef DEBUG
815 	if (ldebug(renameat))
816 		printf(ARGS(renameat, "%s, %s"), from, to);
817 #endif
818 	error = kern_renameat(td, olddfd, from, newdfd, to, UIO_SYSSPACE);
819 	LFREEPATH(from);
820 	LFREEPATH(to);
821 	return (error);
822 }
823 
824 int
825 linux_symlink(struct thread *td, struct linux_symlink_args *args)
826 {
827 	char *path, *to;
828 	int error;
829 
830 	LCONVPATHEXIST(td, args->path, &path);
831 	/* Expand LCONVPATHCREATE so that `path' can be freed on errors */
832 	error = linux_emul_convpath(td, args->to, UIO_USERSPACE, &to, 1, AT_FDCWD);
833 	if (to == NULL) {
834 		LFREEPATH(path);
835 		return (error);
836 	}
837 
838 #ifdef DEBUG
839 	if (ldebug(symlink))
840 		printf(ARGS(symlink, "%s, %s"), path, to);
841 #endif
842 	error = kern_symlink(td, path, to, UIO_SYSSPACE);
843 	LFREEPATH(path);
844 	LFREEPATH(to);
845 	return (error);
846 }
847 
848 int
849 linux_symlinkat(struct thread *td, struct linux_symlinkat_args *args)
850 {
851 	char *path, *to;
852 	int error, dfd;
853 
854 	dfd = (args->newdfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->newdfd;
855 	LCONVPATHEXIST_AT(td, args->oldname, &path, dfd);
856 	/* Expand LCONVPATHCREATE so that `path' can be freed on errors */
857 	error = linux_emul_convpath(td, args->newname, UIO_USERSPACE, &to, 1, dfd);
858 	if (to == NULL) {
859 		LFREEPATH(path);
860 		return (error);
861 	}
862 
863 #ifdef DEBUG
864 	if (ldebug(symlinkat))
865 		printf(ARGS(symlinkat, "%s, %s"), path, to);
866 #endif
867 
868 	error = kern_symlinkat(td, path, dfd, to, UIO_SYSSPACE);
869 	LFREEPATH(path);
870 	LFREEPATH(to);
871 	return (error);
872 }
873 
874 int
875 linux_readlink(struct thread *td, struct linux_readlink_args *args)
876 {
877 	char *name;
878 	int error;
879 
880 	LCONVPATHEXIST(td, args->name, &name);
881 
882 #ifdef DEBUG
883 	if (ldebug(readlink))
884 		printf(ARGS(readlink, "%s, %p, %d"), name, (void *)args->buf,
885 		    args->count);
886 #endif
887 	error = kern_readlink(td, name, UIO_SYSSPACE, args->buf, UIO_USERSPACE,
888 	    args->count);
889 	LFREEPATH(name);
890 	return (error);
891 }
892 
893 int
894 linux_readlinkat(struct thread *td, struct linux_readlinkat_args *args)
895 {
896 	char *name;
897 	int error, dfd;
898 
899 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
900 	LCONVPATHEXIST_AT(td, args->path, &name, dfd);
901 
902 #ifdef DEBUG
903 	if (ldebug(readlinkat))
904 		printf(ARGS(readlinkat, "%s, %p, %d"), name, (void *)args->buf,
905 		    args->bufsiz);
906 #endif
907 
908 	error = kern_readlinkat(td, dfd, name, UIO_SYSSPACE, args->buf,
909 	    UIO_USERSPACE, args->bufsiz);
910 	LFREEPATH(name);
911 	return (error);
912 }
913 
914 int
915 linux_truncate(struct thread *td, struct linux_truncate_args *args)
916 {
917 	char *path;
918 	int error;
919 
920 	LCONVPATHEXIST(td, args->path, &path);
921 
922 #ifdef DEBUG
923 	if (ldebug(truncate))
924 		printf(ARGS(truncate, "%s, %ld"), path, (long)args->length);
925 #endif
926 
927 	error = kern_truncate(td, path, UIO_SYSSPACE, args->length);
928 	LFREEPATH(path);
929 	return (error);
930 }
931 
932 int
933 linux_truncate64(struct thread *td, struct linux_truncate64_args *args)
934 {
935 	char *path;
936 	int error;
937 
938 	LCONVPATHEXIST(td, args->path, &path);
939 
940 #ifdef DEBUG
941 	if (ldebug(truncate64))
942 		printf(ARGS(truncate64, "%s, %jd"), path, args->length);
943 #endif
944 
945 	error = kern_truncate(td, path, UIO_SYSSPACE, args->length);
946 	LFREEPATH(path);
947 	return (error);
948 }
949 int
950 linux_ftruncate(struct thread *td, struct linux_ftruncate_args *args)
951 {
952 	struct ftruncate_args /* {
953 		int fd;
954 		int pad;
955 		off_t length;
956 		} */ nuap;
957 
958 	nuap.fd = args->fd;
959 	nuap.length = args->length;
960 	return (sys_ftruncate(td, &nuap));
961 }
962 
963 int
964 linux_link(struct thread *td, struct linux_link_args *args)
965 {
966 	char *path, *to;
967 	int error;
968 
969 	LCONVPATHEXIST(td, args->path, &path);
970 	/* Expand LCONVPATHCREATE so that `path' can be freed on errors */
971 	error = linux_emul_convpath(td, args->to, UIO_USERSPACE, &to, 1, AT_FDCWD);
972 	if (to == NULL) {
973 		LFREEPATH(path);
974 		return (error);
975 	}
976 
977 #ifdef DEBUG
978 	if (ldebug(link))
979 		printf(ARGS(link, "%s, %s"), path, to);
980 #endif
981 	error = kern_link(td, path, to, UIO_SYSSPACE);
982 	LFREEPATH(path);
983 	LFREEPATH(to);
984 	return (error);
985 }
986 
987 int
988 linux_linkat(struct thread *td, struct linux_linkat_args *args)
989 {
990 	char *path, *to;
991 	int error, olddfd, newdfd, follow;
992 
993 	if (args->flag & ~LINUX_AT_SYMLINK_FOLLOW)
994 		return (EINVAL);
995 
996 	olddfd = (args->olddfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->olddfd;
997 	newdfd = (args->newdfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->newdfd;
998 	LCONVPATHEXIST_AT(td, args->oldname, &path, olddfd);
999 	/* Expand LCONVPATHCREATE so that `path' can be freed on errors */
1000 	error = linux_emul_convpath(td, args->newname, UIO_USERSPACE, &to, 1, newdfd);
1001 	if (to == NULL) {
1002 		LFREEPATH(path);
1003 		return (error);
1004 	}
1005 
1006 #ifdef DEBUG
1007 	if (ldebug(linkat))
1008 		printf(ARGS(linkat, "%i, %s, %i, %s, %i"), args->olddfd, path,
1009 			args->newdfd, to, args->flag);
1010 #endif
1011 
1012 	follow = (args->flag & LINUX_AT_SYMLINK_FOLLOW) == 0 ? NOFOLLOW :
1013 	    FOLLOW;
1014 	error = kern_linkat(td, olddfd, newdfd, path, to, UIO_SYSSPACE, follow);
1015 	LFREEPATH(path);
1016 	LFREEPATH(to);
1017 	return (error);
1018 }
1019 
1020 int
1021 linux_fdatasync(td, uap)
1022 	struct thread *td;
1023 	struct linux_fdatasync_args *uap;
1024 {
1025 	struct fsync_args bsd;
1026 
1027 	bsd.fd = uap->fd;
1028 	return sys_fsync(td, &bsd);
1029 }
1030 
1031 int
1032 linux_pread(td, uap)
1033 	struct thread *td;
1034 	struct linux_pread_args *uap;
1035 {
1036 	struct pread_args bsd;
1037 	struct vnode *vp;
1038 	int error;
1039 
1040 	bsd.fd = uap->fd;
1041 	bsd.buf = uap->buf;
1042 	bsd.nbyte = uap->nbyte;
1043 	bsd.offset = uap->offset;
1044 
1045 	error = sys_pread(td, &bsd);
1046 
1047 	if (error == 0) {
1048    	   	/* This seems to violate POSIX but linux does it */
1049 		if ((error = fgetvp(td, uap->fd, CAP_READ, &vp)) != 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 	struct ufs_args ufs;
1079 	char fstypename[MFSNAMELEN];
1080 	char mntonname[MNAMELEN], mntfromname[MNAMELEN];
1081 	int error;
1082 	int fsflags;
1083 	void *fsdata;
1084 
1085 	error = copyinstr(args->filesystemtype, fstypename, MFSNAMELEN - 1,
1086 	    NULL);
1087 	if (error)
1088 		return (error);
1089 	error = copyinstr(args->specialfile, mntfromname, MNAMELEN - 1, NULL);
1090 	if (error)
1091 		return (error);
1092 	error = copyinstr(args->dir, mntonname, MNAMELEN - 1, NULL);
1093 	if (error)
1094 		return (error);
1095 
1096 #ifdef DEBUG
1097 	if (ldebug(mount))
1098 		printf(ARGS(mount, "%s, %s, %s"),
1099 		    fstypename, mntfromname, mntonname);
1100 #endif
1101 
1102 	if (strcmp(fstypename, "ext2") == 0) {
1103 		strcpy(fstypename, "ext2fs");
1104 		fsdata = &ufs;
1105 		ufs.fspec = mntfromname;
1106 #define DEFAULT_ROOTID		-2
1107 		ufs.export.ex_root = DEFAULT_ROOTID;
1108 		ufs.export.ex_flags =
1109 		    args->rwflag & LINUX_MS_RDONLY ? MNT_EXRDONLY : 0;
1110 	} else if (strcmp(fstypename, "proc") == 0) {
1111 		strcpy(fstypename, "linprocfs");
1112 		fsdata = NULL;
1113 	} else if (strcmp(fstypename, "vfat") == 0) {
1114 		strcpy(fstypename, "msdosfs");
1115 		fsdata = NULL;
1116 	} else {
1117 		return (ENODEV);
1118 	}
1119 
1120 	fsflags = 0;
1121 
1122 	if ((args->rwflag & 0xffff0000) == 0xc0ed0000) {
1123 		/*
1124 		 * Linux SYNC flag is not included; the closest equivalent
1125 		 * FreeBSD has is !ASYNC, which is our default.
1126 		 */
1127 		if (args->rwflag & LINUX_MS_RDONLY)
1128 			fsflags |= MNT_RDONLY;
1129 		if (args->rwflag & LINUX_MS_NOSUID)
1130 			fsflags |= MNT_NOSUID;
1131 		if (args->rwflag & LINUX_MS_NOEXEC)
1132 			fsflags |= MNT_NOEXEC;
1133 		if (args->rwflag & LINUX_MS_REMOUNT)
1134 			fsflags |= MNT_UPDATE;
1135 	}
1136 
1137 	if (strcmp(fstypename, "linprocfs") == 0) {
1138 		error = kernel_vmount(fsflags,
1139 			"fstype", fstypename,
1140 			"fspath", mntonname,
1141 			NULL);
1142 	} else if (strcmp(fstypename, "msdosfs") == 0) {
1143 		error = kernel_vmount(fsflags,
1144 			"fstype", fstypename,
1145 			"fspath", mntonname,
1146 			"from", mntfromname,
1147 			NULL);
1148 	} else
1149 		error = EOPNOTSUPP;
1150 	return (error);
1151 }
1152 
1153 int
1154 linux_oldumount(struct thread *td, struct linux_oldumount_args *args)
1155 {
1156 	struct linux_umount_args args2;
1157 
1158 	args2.path = args->path;
1159 	args2.flags = 0;
1160 	return (linux_umount(td, &args2));
1161 }
1162 
1163 int
1164 linux_umount(struct thread *td, struct linux_umount_args *args)
1165 {
1166 	struct unmount_args bsd;
1167 
1168 	bsd.path = args->path;
1169 	bsd.flags = args->flags;	/* XXX correct? */
1170 	return (sys_unmount(td, &bsd));
1171 }
1172 
1173 /*
1174  * fcntl family of syscalls
1175  */
1176 
1177 struct l_flock {
1178 	l_short		l_type;
1179 	l_short		l_whence;
1180 	l_off_t		l_start;
1181 	l_off_t		l_len;
1182 	l_pid_t		l_pid;
1183 }
1184 #if defined(__amd64__) && defined(COMPAT_LINUX32)
1185 __packed
1186 #endif
1187 ;
1188 
1189 static void
1190 linux_to_bsd_flock(struct l_flock *linux_flock, struct flock *bsd_flock)
1191 {
1192 	switch (linux_flock->l_type) {
1193 	case LINUX_F_RDLCK:
1194 		bsd_flock->l_type = F_RDLCK;
1195 		break;
1196 	case LINUX_F_WRLCK:
1197 		bsd_flock->l_type = F_WRLCK;
1198 		break;
1199 	case LINUX_F_UNLCK:
1200 		bsd_flock->l_type = F_UNLCK;
1201 		break;
1202 	default:
1203 		bsd_flock->l_type = -1;
1204 		break;
1205 	}
1206 	bsd_flock->l_whence = linux_flock->l_whence;
1207 	bsd_flock->l_start = (off_t)linux_flock->l_start;
1208 	bsd_flock->l_len = (off_t)linux_flock->l_len;
1209 	bsd_flock->l_pid = (pid_t)linux_flock->l_pid;
1210 	bsd_flock->l_sysid = 0;
1211 }
1212 
1213 static void
1214 bsd_to_linux_flock(struct flock *bsd_flock, struct l_flock *linux_flock)
1215 {
1216 	switch (bsd_flock->l_type) {
1217 	case F_RDLCK:
1218 		linux_flock->l_type = LINUX_F_RDLCK;
1219 		break;
1220 	case F_WRLCK:
1221 		linux_flock->l_type = LINUX_F_WRLCK;
1222 		break;
1223 	case F_UNLCK:
1224 		linux_flock->l_type = LINUX_F_UNLCK;
1225 		break;
1226 	}
1227 	linux_flock->l_whence = bsd_flock->l_whence;
1228 	linux_flock->l_start = (l_off_t)bsd_flock->l_start;
1229 	linux_flock->l_len = (l_off_t)bsd_flock->l_len;
1230 	linux_flock->l_pid = (l_pid_t)bsd_flock->l_pid;
1231 }
1232 
1233 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1234 struct l_flock64 {
1235 	l_short		l_type;
1236 	l_short		l_whence;
1237 	l_loff_t	l_start;
1238 	l_loff_t	l_len;
1239 	l_pid_t		l_pid;
1240 }
1241 #if defined(__amd64__) && defined(COMPAT_LINUX32)
1242 __packed
1243 #endif
1244 ;
1245 
1246 static void
1247 linux_to_bsd_flock64(struct l_flock64 *linux_flock, struct flock *bsd_flock)
1248 {
1249 	switch (linux_flock->l_type) {
1250 	case LINUX_F_RDLCK:
1251 		bsd_flock->l_type = F_RDLCK;
1252 		break;
1253 	case LINUX_F_WRLCK:
1254 		bsd_flock->l_type = F_WRLCK;
1255 		break;
1256 	case LINUX_F_UNLCK:
1257 		bsd_flock->l_type = F_UNLCK;
1258 		break;
1259 	default:
1260 		bsd_flock->l_type = -1;
1261 		break;
1262 	}
1263 	bsd_flock->l_whence = linux_flock->l_whence;
1264 	bsd_flock->l_start = (off_t)linux_flock->l_start;
1265 	bsd_flock->l_len = (off_t)linux_flock->l_len;
1266 	bsd_flock->l_pid = (pid_t)linux_flock->l_pid;
1267 	bsd_flock->l_sysid = 0;
1268 }
1269 
1270 static void
1271 bsd_to_linux_flock64(struct flock *bsd_flock, struct l_flock64 *linux_flock)
1272 {
1273 	switch (bsd_flock->l_type) {
1274 	case F_RDLCK:
1275 		linux_flock->l_type = LINUX_F_RDLCK;
1276 		break;
1277 	case F_WRLCK:
1278 		linux_flock->l_type = LINUX_F_WRLCK;
1279 		break;
1280 	case F_UNLCK:
1281 		linux_flock->l_type = LINUX_F_UNLCK;
1282 		break;
1283 	}
1284 	linux_flock->l_whence = bsd_flock->l_whence;
1285 	linux_flock->l_start = (l_loff_t)bsd_flock->l_start;
1286 	linux_flock->l_len = (l_loff_t)bsd_flock->l_len;
1287 	linux_flock->l_pid = (l_pid_t)bsd_flock->l_pid;
1288 }
1289 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1290 
1291 static int
1292 fcntl_common(struct thread *td, struct linux_fcntl64_args *args)
1293 {
1294 	struct l_flock linux_flock;
1295 	struct flock bsd_flock;
1296 	struct file *fp;
1297 	long arg;
1298 	int error, result;
1299 
1300 	switch (args->cmd) {
1301 	case LINUX_F_DUPFD:
1302 		return (kern_fcntl(td, args->fd, F_DUPFD, args->arg));
1303 
1304 	case LINUX_F_GETFD:
1305 		return (kern_fcntl(td, args->fd, F_GETFD, 0));
1306 
1307 	case LINUX_F_SETFD:
1308 		return (kern_fcntl(td, args->fd, F_SETFD, args->arg));
1309 
1310 	case LINUX_F_GETFL:
1311 		error = kern_fcntl(td, args->fd, F_GETFL, 0);
1312 		result = td->td_retval[0];
1313 		td->td_retval[0] = 0;
1314 		if (result & O_RDONLY)
1315 			td->td_retval[0] |= LINUX_O_RDONLY;
1316 		if (result & O_WRONLY)
1317 			td->td_retval[0] |= LINUX_O_WRONLY;
1318 		if (result & O_RDWR)
1319 			td->td_retval[0] |= LINUX_O_RDWR;
1320 		if (result & O_NDELAY)
1321 			td->td_retval[0] |= LINUX_O_NONBLOCK;
1322 		if (result & O_APPEND)
1323 			td->td_retval[0] |= LINUX_O_APPEND;
1324 		if (result & O_FSYNC)
1325 			td->td_retval[0] |= LINUX_O_SYNC;
1326 		if (result & O_ASYNC)
1327 			td->td_retval[0] |= LINUX_FASYNC;
1328 #ifdef LINUX_O_NOFOLLOW
1329 		if (result & O_NOFOLLOW)
1330 			td->td_retval[0] |= LINUX_O_NOFOLLOW;
1331 #endif
1332 #ifdef LINUX_O_DIRECT
1333 		if (result & O_DIRECT)
1334 			td->td_retval[0] |= LINUX_O_DIRECT;
1335 #endif
1336 		return (error);
1337 
1338 	case LINUX_F_SETFL:
1339 		arg = 0;
1340 		if (args->arg & LINUX_O_NDELAY)
1341 			arg |= O_NONBLOCK;
1342 		if (args->arg & LINUX_O_APPEND)
1343 			arg |= O_APPEND;
1344 		if (args->arg & LINUX_O_SYNC)
1345 			arg |= O_FSYNC;
1346 		if (args->arg & LINUX_FASYNC)
1347 			arg |= O_ASYNC;
1348 #ifdef LINUX_O_NOFOLLOW
1349 		if (args->arg & LINUX_O_NOFOLLOW)
1350 			arg |= O_NOFOLLOW;
1351 #endif
1352 #ifdef LINUX_O_DIRECT
1353 		if (args->arg & LINUX_O_DIRECT)
1354 			arg |= O_DIRECT;
1355 #endif
1356 		return (kern_fcntl(td, args->fd, F_SETFL, arg));
1357 
1358 	case LINUX_F_GETLK:
1359 		error = copyin((void *)args->arg, &linux_flock,
1360 		    sizeof(linux_flock));
1361 		if (error)
1362 			return (error);
1363 		linux_to_bsd_flock(&linux_flock, &bsd_flock);
1364 		error = kern_fcntl(td, args->fd, F_GETLK, (intptr_t)&bsd_flock);
1365 		if (error)
1366 			return (error);
1367 		bsd_to_linux_flock(&bsd_flock, &linux_flock);
1368 		return (copyout(&linux_flock, (void *)args->arg,
1369 		    sizeof(linux_flock)));
1370 
1371 	case LINUX_F_SETLK:
1372 		error = copyin((void *)args->arg, &linux_flock,
1373 		    sizeof(linux_flock));
1374 		if (error)
1375 			return (error);
1376 		linux_to_bsd_flock(&linux_flock, &bsd_flock);
1377 		return (kern_fcntl(td, args->fd, F_SETLK,
1378 		    (intptr_t)&bsd_flock));
1379 
1380 	case LINUX_F_SETLKW:
1381 		error = copyin((void *)args->arg, &linux_flock,
1382 		    sizeof(linux_flock));
1383 		if (error)
1384 			return (error);
1385 		linux_to_bsd_flock(&linux_flock, &bsd_flock);
1386 		return (kern_fcntl(td, args->fd, F_SETLKW,
1387 		     (intptr_t)&bsd_flock));
1388 
1389 	case LINUX_F_GETOWN:
1390 		return (kern_fcntl(td, args->fd, F_GETOWN, 0));
1391 
1392 	case LINUX_F_SETOWN:
1393 		/*
1394 		 * XXX some Linux applications depend on F_SETOWN having no
1395 		 * significant effect for pipes (SIGIO is not delivered for
1396 		 * pipes under Linux-2.2.35 at least).
1397 		 */
1398 		error = fget(td, args->fd, CAP_FCNTL, &fp);
1399 		if (error)
1400 			return (error);
1401 		if (fp->f_type == DTYPE_PIPE) {
1402 			fdrop(fp, td);
1403 			return (EINVAL);
1404 		}
1405 		fdrop(fp, td);
1406 
1407 		return (kern_fcntl(td, args->fd, F_SETOWN, args->arg));
1408 	}
1409 
1410 	return (EINVAL);
1411 }
1412 
1413 int
1414 linux_fcntl(struct thread *td, struct linux_fcntl_args *args)
1415 {
1416 	struct linux_fcntl64_args args64;
1417 
1418 #ifdef DEBUG
1419 	if (ldebug(fcntl))
1420 		printf(ARGS(fcntl, "%d, %08x, *"), args->fd, args->cmd);
1421 #endif
1422 
1423 	args64.fd = args->fd;
1424 	args64.cmd = args->cmd;
1425 	args64.arg = args->arg;
1426 	return (fcntl_common(td, &args64));
1427 }
1428 
1429 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1430 int
1431 linux_fcntl64(struct thread *td, struct linux_fcntl64_args *args)
1432 {
1433 	struct l_flock64 linux_flock;
1434 	struct flock bsd_flock;
1435 	int error;
1436 
1437 #ifdef DEBUG
1438 	if (ldebug(fcntl64))
1439 		printf(ARGS(fcntl64, "%d, %08x, *"), args->fd, args->cmd);
1440 #endif
1441 
1442 	switch (args->cmd) {
1443 	case LINUX_F_GETLK64:
1444 		error = copyin((void *)args->arg, &linux_flock,
1445 		    sizeof(linux_flock));
1446 		if (error)
1447 			return (error);
1448 		linux_to_bsd_flock64(&linux_flock, &bsd_flock);
1449 		error = kern_fcntl(td, args->fd, F_GETLK, (intptr_t)&bsd_flock);
1450 		if (error)
1451 			return (error);
1452 		bsd_to_linux_flock64(&bsd_flock, &linux_flock);
1453 		return (copyout(&linux_flock, (void *)args->arg,
1454 			    sizeof(linux_flock)));
1455 
1456 	case LINUX_F_SETLK64:
1457 		error = copyin((void *)args->arg, &linux_flock,
1458 		    sizeof(linux_flock));
1459 		if (error)
1460 			return (error);
1461 		linux_to_bsd_flock64(&linux_flock, &bsd_flock);
1462 		return (kern_fcntl(td, args->fd, F_SETLK,
1463 		    (intptr_t)&bsd_flock));
1464 
1465 	case LINUX_F_SETLKW64:
1466 		error = copyin((void *)args->arg, &linux_flock,
1467 		    sizeof(linux_flock));
1468 		if (error)
1469 			return (error);
1470 		linux_to_bsd_flock64(&linux_flock, &bsd_flock);
1471 		return (kern_fcntl(td, args->fd, F_SETLKW,
1472 		    (intptr_t)&bsd_flock));
1473 	}
1474 
1475 	return (fcntl_common(td, args));
1476 }
1477 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1478 
1479 int
1480 linux_chown(struct thread *td, struct linux_chown_args *args)
1481 {
1482 	char *path;
1483 	int error;
1484 
1485 	LCONVPATHEXIST(td, args->path, &path);
1486 
1487 #ifdef DEBUG
1488 	if (ldebug(chown))
1489 		printf(ARGS(chown, "%s, %d, %d"), path, args->uid, args->gid);
1490 #endif
1491 	error = kern_chown(td, path, UIO_SYSSPACE, args->uid, args->gid);
1492 	LFREEPATH(path);
1493 	return (error);
1494 }
1495 
1496 int
1497 linux_fchownat(struct thread *td, struct linux_fchownat_args *args)
1498 {
1499 	char *path;
1500 	int error, dfd, flag;
1501 
1502 	if (args->flag & ~LINUX_AT_SYMLINK_NOFOLLOW)
1503 		return (EINVAL);
1504 
1505 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD :  args->dfd;
1506 	LCONVPATHEXIST_AT(td, args->filename, &path, dfd);
1507 
1508 #ifdef DEBUG
1509 	if (ldebug(fchownat))
1510 		printf(ARGS(fchownat, "%s, %d, %d"), path, args->uid, args->gid);
1511 #endif
1512 
1513 	flag = (args->flag & LINUX_AT_SYMLINK_NOFOLLOW) == 0 ? 0 :
1514 	    AT_SYMLINK_NOFOLLOW;
1515 	error = kern_fchownat(td, dfd, path, UIO_SYSSPACE, args->uid, args->gid,
1516 	    flag);
1517 	LFREEPATH(path);
1518 	return (error);
1519 }
1520 
1521 int
1522 linux_lchown(struct thread *td, struct linux_lchown_args *args)
1523 {
1524 	char *path;
1525 	int error;
1526 
1527 	LCONVPATHEXIST(td, args->path, &path);
1528 
1529 #ifdef DEBUG
1530 	if (ldebug(lchown))
1531 		printf(ARGS(lchown, "%s, %d, %d"), path, args->uid, args->gid);
1532 #endif
1533 	error = kern_lchown(td, path, UIO_SYSSPACE, args->uid, args->gid);
1534 	LFREEPATH(path);
1535 	return (error);
1536 }
1537 
1538 static int
1539 convert_fadvice(int advice)
1540 {
1541 	switch (advice) {
1542 	case LINUX_POSIX_FADV_NORMAL:
1543 		return (POSIX_FADV_NORMAL);
1544 	case LINUX_POSIX_FADV_RANDOM:
1545 		return (POSIX_FADV_RANDOM);
1546 	case LINUX_POSIX_FADV_SEQUENTIAL:
1547 		return (POSIX_FADV_SEQUENTIAL);
1548 	case LINUX_POSIX_FADV_WILLNEED:
1549 		return (POSIX_FADV_WILLNEED);
1550 	case LINUX_POSIX_FADV_DONTNEED:
1551 		return (POSIX_FADV_DONTNEED);
1552 	case LINUX_POSIX_FADV_NOREUSE:
1553 		return (POSIX_FADV_NOREUSE);
1554 	default:
1555 		return (-1);
1556 	}
1557 }
1558 
1559 int
1560 linux_fadvise64(struct thread *td, struct linux_fadvise64_args *args)
1561 {
1562 	int advice;
1563 
1564 	advice = convert_fadvice(args->advice);
1565 	if (advice == -1)
1566 		return (EINVAL);
1567 	return (kern_posix_fadvise(td, args->fd, args->offset, args->len,
1568 	    advice));
1569 }
1570 
1571 int
1572 linux_fadvise64_64(struct thread *td, struct linux_fadvise64_64_args *args)
1573 {
1574 	int advice;
1575 
1576 	advice = convert_fadvice(args->advice);
1577 	if (advice == -1)
1578 		return (EINVAL);
1579 	return (kern_posix_fadvise(td, args->fd, args->offset, args->len,
1580 	    advice));
1581 }
1582 
1583 int
1584 linux_pipe(struct thread *td, struct linux_pipe_args *args)
1585 {
1586 	int fildes[2];
1587 	int error;
1588 
1589 #ifdef DEBUG
1590 	if (ldebug(pipe))
1591 		printf(ARGS(pipe, "*"));
1592 #endif
1593 
1594 	error = do_pipe(td, fildes, 0);
1595 	if (error)
1596 		return (error);
1597 
1598 	/* XXX: Close descriptors on error. */
1599 	return (copyout(fildes, args->pipefds, sizeof(fildes)));
1600 }
1601 
1602 int
1603 linux_pipe2(struct thread *td, struct linux_pipe2_args *args)
1604 {
1605 	int fildes[2];
1606 	int error, flags;
1607 
1608 #ifdef DEBUG
1609 	if (ldebug(pipe2))
1610 		printf(ARGS(pipe2, "*, %d"), args->flags);
1611 #endif
1612 
1613 	if ((args->flags & ~(LINUX_O_NONBLOCK | LINUX_O_CLOEXEC)) != 0)
1614 		return (EINVAL);
1615 
1616 	flags = 0;
1617 	if ((args->flags & LINUX_O_NONBLOCK) != 0)
1618 		flags |= O_NONBLOCK;
1619 	if ((args->flags & LINUX_O_CLOEXEC) != 0)
1620 		flags |= O_CLOEXEC;
1621 	error = do_pipe(td, fildes, flags);
1622 	if (error)
1623 		return (error);
1624 
1625 	/* XXX: Close descriptors on error. */
1626 	return (copyout(fildes, args->pipefds, sizeof(fildes)));
1627 }
1628