xref: /freebsd/sys/compat/linux/linux_file.c (revision 7e46ff5f4cb541537722f87bb20a57bc41493d54)
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_misc.h>
70 #include <compat/linux/linux_util.h>
71 #include <compat/linux/linux_file.h>
72 
73 /* XXX */
74 int	do_pipe(struct thread *td, int fildes[2], int flags);
75 
76 int
77 linux_creat(struct thread *td, struct linux_creat_args *args)
78 {
79     char *path;
80     int error;
81 
82     LCONVPATHEXIST(td, args->path, &path);
83 
84 #ifdef DEBUG
85 	if (ldebug(creat))
86 		printf(ARGS(creat, "%s, %d"), path, args->mode);
87 #endif
88     error = kern_open(td, path, UIO_SYSSPACE, O_WRONLY | O_CREAT | O_TRUNC,
89 	args->mode);
90     LFREEPATH(path);
91     return (error);
92 }
93 
94 
95 static int
96 linux_common_open(struct thread *td, int dirfd, char *path, int l_flags, int mode)
97 {
98     struct proc *p = td->td_proc;
99     struct file *fp;
100     int fd;
101     int bsd_flags, error;
102 
103     bsd_flags = 0;
104     switch (l_flags & LINUX_O_ACCMODE) {
105     case LINUX_O_WRONLY:
106 	bsd_flags |= O_WRONLY;
107 	break;
108     case LINUX_O_RDWR:
109 	bsd_flags |= O_RDWR;
110 	break;
111     default:
112 	bsd_flags |= O_RDONLY;
113     }
114     if (l_flags & LINUX_O_NDELAY)
115 	bsd_flags |= O_NONBLOCK;
116     if (l_flags & LINUX_O_APPEND)
117 	bsd_flags |= O_APPEND;
118     if (l_flags & LINUX_O_SYNC)
119 	bsd_flags |= O_FSYNC;
120     if (l_flags & LINUX_O_NONBLOCK)
121 	bsd_flags |= O_NONBLOCK;
122     if (l_flags & LINUX_FASYNC)
123 	bsd_flags |= O_ASYNC;
124     if (l_flags & LINUX_O_CREAT)
125 	bsd_flags |= O_CREAT;
126     if (l_flags & LINUX_O_TRUNC)
127 	bsd_flags |= O_TRUNC;
128     if (l_flags & LINUX_O_EXCL)
129 	bsd_flags |= O_EXCL;
130     if (l_flags & LINUX_O_NOCTTY)
131 	bsd_flags |= O_NOCTTY;
132     if (l_flags & LINUX_O_DIRECT)
133 	bsd_flags |= O_DIRECT;
134     if (l_flags & LINUX_O_NOFOLLOW)
135 	bsd_flags |= O_NOFOLLOW;
136     if (l_flags & LINUX_O_DIRECTORY)
137 	bsd_flags |= O_DIRECTORY;
138     /* XXX LINUX_O_NOATIME: unable to be easily implemented. */
139 
140     error = kern_openat(td, dirfd, path, UIO_SYSSPACE, bsd_flags, mode);
141 
142     if (!error) {
143 	    fd = td->td_retval[0];
144 	    /*
145 	     * XXX In between kern_open() and fget(), another process
146 	     * having the same filedesc could use that fd without
147 	     * checking below.
148 	     */
149 	    error = fget(td, fd, CAP_IOCTL, &fp);
150 	    if (!error) {
151 		    sx_slock(&proctree_lock);
152 		    PROC_LOCK(p);
153 		    if (!(bsd_flags & O_NOCTTY) &&
154 			SESS_LEADER(p) && !(p->p_flag & P_CONTROLT)) {
155 			    PROC_UNLOCK(p);
156 			    sx_unlock(&proctree_lock);
157 			    if (fp->f_type == DTYPE_VNODE)
158 				    (void) fo_ioctl(fp, TIOCSCTTY, (caddr_t) 0,
159 					     td->td_ucred, td);
160 		    } else {
161 			    PROC_UNLOCK(p);
162 			    sx_sunlock(&proctree_lock);
163 		    }
164 		    fdrop(fp, td);
165 		    /*
166 		     * XXX as above, fdrop()/kern_close() pair is racy.
167 		     */
168 		    if (error)
169 			    kern_close(td, fd);
170 	    }
171     }
172 
173 #ifdef DEBUG
174     if (ldebug(open))
175 	    printf(LMSG("open returns error %d"), error);
176 #endif
177     LFREEPATH(path);
178     return (error);
179 }
180 
181 int
182 linux_openat(struct thread *td, struct linux_openat_args *args)
183 {
184 	char *path;
185 	int dfd;
186 
187 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
188 	if (args->flags & LINUX_O_CREAT)
189 		LCONVPATH_AT(td, args->filename, &path, 1, dfd);
190 	else
191 		LCONVPATH_AT(td, args->filename, &path, 0, dfd);
192 #ifdef DEBUG
193 	if (ldebug(openat))
194 		printf(ARGS(openat, "%i, %s, 0x%x, 0x%x"), args->dfd,
195 		    path, args->flags, args->mode);
196 #endif
197 	return (linux_common_open(td, dfd, path, args->flags, args->mode));
198 }
199 
200 int
201 linux_open(struct thread *td, struct linux_open_args *args)
202 {
203     char *path;
204 
205     if (args->flags & LINUX_O_CREAT)
206 	LCONVPATHCREAT(td, args->path, &path);
207     else
208 	LCONVPATHEXIST(td, args->path, &path);
209 
210 #ifdef DEBUG
211 	if (ldebug(open))
212 		printf(ARGS(open, "%s, 0x%x, 0x%x"),
213 		    path, args->flags, args->mode);
214 #endif
215 
216 	return (linux_common_open(td, AT_FDCWD, path, args->flags, args->mode));
217 }
218 
219 int
220 linux_lseek(struct thread *td, struct linux_lseek_args *args)
221 {
222 
223     struct lseek_args /* {
224 	int fd;
225 	int pad;
226 	off_t offset;
227 	int whence;
228     } */ tmp_args;
229     int error;
230 
231 #ifdef DEBUG
232 	if (ldebug(lseek))
233 		printf(ARGS(lseek, "%d, %ld, %d"),
234 		    args->fdes, (long)args->off, args->whence);
235 #endif
236     tmp_args.fd = args->fdes;
237     tmp_args.offset = (off_t)args->off;
238     tmp_args.whence = args->whence;
239     error = sys_lseek(td, &tmp_args);
240     return error;
241 }
242 
243 int
244 linux_llseek(struct thread *td, struct linux_llseek_args *args)
245 {
246 	struct lseek_args bsd_args;
247 	int error;
248 	off_t off;
249 
250 #ifdef DEBUG
251 	if (ldebug(llseek))
252 		printf(ARGS(llseek, "%d, %d:%d, %d"),
253 		    args->fd, args->ohigh, args->olow, args->whence);
254 #endif
255 	off = (args->olow) | (((off_t) args->ohigh) << 32);
256 
257 	bsd_args.fd = args->fd;
258 	bsd_args.offset = off;
259 	bsd_args.whence = args->whence;
260 
261 	if ((error = sys_lseek(td, &bsd_args)))
262 		return error;
263 
264 	if ((error = copyout(td->td_retval, args->res, sizeof (off_t))))
265 		return error;
266 
267 	td->td_retval[0] = 0;
268 	return 0;
269 }
270 
271 int
272 linux_readdir(struct thread *td, struct linux_readdir_args *args)
273 {
274 	struct linux_getdents_args lda;
275 
276 	lda.fd = args->fd;
277 	lda.dent = args->dent;
278 	lda.count = 1;
279 	return linux_getdents(td, &lda);
280 }
281 
282 /*
283  * Note that linux_getdents(2) and linux_getdents64(2) have the same
284  * arguments. They only differ in the definition of struct dirent they
285  * operate on. We use this to common the code, with the exception of
286  * accessing struct dirent. Note that linux_readdir(2) is implemented
287  * by means of linux_getdents(2). In this case we never operate on
288  * struct dirent64 and thus don't need to handle it...
289  */
290 
291 struct l_dirent {
292 	l_ulong		d_ino;
293 	l_off_t		d_off;
294 	l_ushort	d_reclen;
295 	char		d_name[LINUX_NAME_MAX + 1];
296 };
297 
298 struct l_dirent64 {
299 	uint64_t	d_ino;
300 	int64_t		d_off;
301 	l_ushort	d_reclen;
302 	u_char		d_type;
303 	char		d_name[LINUX_NAME_MAX + 1];
304 };
305 
306 /*
307  * Linux uses the last byte in the dirent buffer to store d_type,
308  * at least glibc-2.7 requires it. That is why l_dirent is padded with 2 bytes.
309  */
310 #define LINUX_RECLEN(namlen)						\
311     roundup((offsetof(struct l_dirent, d_name) + (namlen) + 2),		\
312     sizeof(l_ulong))
313 
314 #define LINUX_RECLEN64(namlen)						\
315     roundup((offsetof(struct l_dirent64, d_name) + (namlen) + 1),	\
316     sizeof(uint64_t))
317 
318 #define LINUX_MAXRECLEN		max(LINUX_RECLEN(LINUX_NAME_MAX),	\
319 				    LINUX_RECLEN64(LINUX_NAME_MAX))
320 #define	LINUX_DIRBLKSIZ		512
321 
322 static int
323 getdents_common(struct thread *td, struct linux_getdents64_args *args,
324     int is64bit)
325 {
326 	struct dirent *bdp;
327 	struct vnode *vp;
328 	caddr_t inp, buf;		/* BSD-format */
329 	int len, reclen;		/* BSD-format */
330 	caddr_t outp;			/* Linux-format */
331 	int resid, linuxreclen=0;	/* Linux-format */
332 	caddr_t lbuf;			/* Linux-format */
333 	struct file *fp;
334 	struct uio auio;
335 	struct iovec aiov;
336 	off_t off;
337 	struct l_dirent *linux_dirent;
338 	struct l_dirent64 *linux_dirent64;
339 	int buflen, error, eofflag, nbytes, justone;
340 	u_long *cookies = NULL, *cookiep;
341 	int ncookies;
342 
343 	nbytes = args->count;
344 	if (nbytes == 1) {
345 		/* readdir(2) case. Always struct dirent. */
346 		if (is64bit)
347 			return (EINVAL);
348 		nbytes = sizeof(*linux_dirent);
349 		justone = 1;
350 	} else
351 		justone = 0;
352 
353 	if ((error = getvnode(td->td_proc->p_fd, args->fd, CAP_READ, &fp)) != 0)
354 		return (error);
355 
356 	if ((fp->f_flag & FREAD) == 0) {
357 		fdrop(fp, td);
358 		return (EBADF);
359 	}
360 
361 	off = foffset_lock(fp, 0);
362 	vp = fp->f_vnode;
363 	if (vp->v_type != VDIR) {
364 		foffset_unlock(fp, off, 0);
365 		fdrop(fp, td);
366 		return (EINVAL);
367 	}
368 
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 #ifdef MAC
387 	/*
388 	 * Do directory search MAC check using non-cached credentials.
389 	 */
390 	if ((error = mac_vnode_check_readdir(td->td_ucred, vp)))
391 		goto out;
392 #endif /* MAC */
393 	if ((error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, &ncookies,
394 		 &cookies)))
395 		goto out;
396 
397 	inp = buf;
398 	outp = (caddr_t)args->dirent;
399 	resid = nbytes;
400 	if ((len = buflen - auio.uio_resid) <= 0)
401 		goto eof;
402 
403 	cookiep = cookies;
404 
405 	if (cookies) {
406 		/*
407 		 * When using cookies, the vfs has the option of reading from
408 		 * a different offset than that supplied (UFS truncates the
409 		 * offset to a block boundary to make sure that it never reads
410 		 * partway through a directory entry, even if the directory
411 		 * has been compacted).
412 		 */
413 		while (len > 0 && ncookies > 0 && *cookiep <= off) {
414 			bdp = (struct dirent *) inp;
415 			len -= bdp->d_reclen;
416 			inp += bdp->d_reclen;
417 			cookiep++;
418 			ncookies--;
419 		}
420 	}
421 
422 	while (len > 0) {
423 		if (cookiep && ncookies == 0)
424 			break;
425 		bdp = (struct dirent *) inp;
426 		reclen = bdp->d_reclen;
427 		if (reclen & 3) {
428 			error = EFAULT;
429 			goto out;
430 		}
431 
432 		if (bdp->d_fileno == 0) {
433 			inp += reclen;
434 			if (cookiep) {
435 				off = *cookiep++;
436 				ncookies--;
437 			} else
438 				off += reclen;
439 
440 			len -= reclen;
441 			continue;
442 		}
443 
444 		linuxreclen = (is64bit)
445 		    ? LINUX_RECLEN64(bdp->d_namlen)
446 		    : LINUX_RECLEN(bdp->d_namlen);
447 
448 		if (reclen > len || resid < linuxreclen) {
449 			outp++;
450 			break;
451 		}
452 
453 		if (justone) {
454 			/* readdir(2) case. */
455 			linux_dirent = (struct l_dirent*)lbuf;
456 			linux_dirent->d_ino = bdp->d_fileno;
457 			linux_dirent->d_off = (l_off_t)linuxreclen;
458 			linux_dirent->d_reclen = (l_ushort)bdp->d_namlen;
459 			strlcpy(linux_dirent->d_name, bdp->d_name,
460 			    linuxreclen - offsetof(struct l_dirent, d_name));
461 			error = copyout(linux_dirent, outp, linuxreclen);
462 		}
463 		if (is64bit) {
464 			linux_dirent64 = (struct l_dirent64*)lbuf;
465 			linux_dirent64->d_ino = bdp->d_fileno;
466 			linux_dirent64->d_off = (cookiep)
467 			    ? (l_off_t)*cookiep
468 			    : (l_off_t)(off + reclen);
469 			linux_dirent64->d_reclen = (l_ushort)linuxreclen;
470 			linux_dirent64->d_type = bdp->d_type;
471 			strlcpy(linux_dirent64->d_name, bdp->d_name,
472 			    linuxreclen - offsetof(struct l_dirent64, d_name));
473 			error = copyout(linux_dirent64, outp, linuxreclen);
474 		} else if (!justone) {
475 			linux_dirent = (struct l_dirent*)lbuf;
476 			linux_dirent->d_ino = bdp->d_fileno;
477 			linux_dirent->d_off = (cookiep)
478 			    ? (l_off_t)*cookiep
479 			    : (l_off_t)(off + reclen);
480 			linux_dirent->d_reclen = (l_ushort)linuxreclen;
481 			/*
482 			 * Copy d_type to last byte of l_dirent buffer
483 			 */
484 			lbuf[linuxreclen-1] = bdp->d_type;
485 			strlcpy(linux_dirent->d_name, bdp->d_name,
486 			    linuxreclen - offsetof(struct l_dirent, d_name)-1);
487 			error = copyout(linux_dirent, outp, linuxreclen);
488 		}
489 
490 		if (error)
491 			goto out;
492 
493 		inp += reclen;
494 		if (cookiep) {
495 			off = *cookiep++;
496 			ncookies--;
497 		} else
498 			off += reclen;
499 
500 		outp += linuxreclen;
501 		resid -= linuxreclen;
502 		len -= reclen;
503 		if (justone)
504 			break;
505 	}
506 
507 	if (outp == (caddr_t)args->dirent) {
508 		nbytes = resid;
509 		goto eof;
510 	}
511 
512 	if (justone)
513 		nbytes = resid + linuxreclen;
514 
515 eof:
516 	td->td_retval[0] = nbytes - resid;
517 
518 out:
519 	if (cookies)
520 		free(cookies, M_TEMP);
521 
522 	VOP_UNLOCK(vp, 0);
523 	foffset_unlock(fp, off, 0);
524 	fdrop(fp, td);
525 	free(buf, M_TEMP);
526 	free(lbuf, M_TEMP);
527 	return (error);
528 }
529 
530 int
531 linux_getdents(struct thread *td, struct linux_getdents_args *args)
532 {
533 
534 #ifdef DEBUG
535 	if (ldebug(getdents))
536 		printf(ARGS(getdents, "%d, *, %d"), args->fd, args->count);
537 #endif
538 
539 	return (getdents_common(td, (struct linux_getdents64_args*)args, 0));
540 }
541 
542 int
543 linux_getdents64(struct thread *td, struct linux_getdents64_args *args)
544 {
545 
546 #ifdef DEBUG
547 	if (ldebug(getdents64))
548 		printf(ARGS(getdents64, "%d, *, %d"), args->fd, args->count);
549 #endif
550 
551 	return (getdents_common(td, args, 1));
552 }
553 
554 /*
555  * These exist mainly for hooks for doing /compat/linux translation.
556  */
557 
558 int
559 linux_access(struct thread *td, struct linux_access_args *args)
560 {
561 	char *path;
562 	int error;
563 
564 	/* linux convention */
565 	if (args->amode & ~(F_OK | X_OK | W_OK | R_OK))
566 		return (EINVAL);
567 
568 	LCONVPATHEXIST(td, args->path, &path);
569 
570 #ifdef DEBUG
571 	if (ldebug(access))
572 		printf(ARGS(access, "%s, %d"), path, args->amode);
573 #endif
574 	error = kern_access(td, path, UIO_SYSSPACE, args->amode);
575 	LFREEPATH(path);
576 
577 	return (error);
578 }
579 
580 int
581 linux_faccessat(struct thread *td, struct linux_faccessat_args *args)
582 {
583 	char *path;
584 	int error, dfd, flag;
585 
586 	if (args->flag & ~LINUX_AT_EACCESS)
587 		return (EINVAL);
588 	/* linux convention */
589 	if (args->amode & ~(F_OK | X_OK | W_OK | R_OK))
590 		return (EINVAL);
591 
592 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
593 	LCONVPATHEXIST_AT(td, args->filename, &path, dfd);
594 
595 #ifdef DEBUG
596 	if (ldebug(access))
597 		printf(ARGS(access, "%s, %d"), path, args->amode);
598 #endif
599 
600 	flag = (args->flag & LINUX_AT_EACCESS) == 0 ? 0 : AT_EACCESS;
601 	error = kern_accessat(td, dfd, path, UIO_SYSSPACE, flag, args->amode);
602 	LFREEPATH(path);
603 
604 	return (error);
605 }
606 
607 int
608 linux_unlink(struct thread *td, struct linux_unlink_args *args)
609 {
610 	char *path;
611 	int error;
612 	struct stat st;
613 
614 	LCONVPATHEXIST(td, args->path, &path);
615 
616 #ifdef DEBUG
617 	if (ldebug(unlink))
618 		printf(ARGS(unlink, "%s"), path);
619 #endif
620 
621 	error = kern_unlink(td, path, UIO_SYSSPACE);
622 	if (error == EPERM)
623 		/* Introduce POSIX noncompliant behaviour of Linux */
624 		if (kern_stat(td, path, UIO_SYSSPACE, &st) == 0)
625 			if (S_ISDIR(st.st_mode))
626 				error = EISDIR;
627 	LFREEPATH(path);
628 	return (error);
629 }
630 
631 int
632 linux_unlinkat(struct thread *td, struct linux_unlinkat_args *args)
633 {
634 	char *path;
635 	int error, dfd;
636 	struct stat st;
637 
638 	if (args->flag & ~LINUX_AT_REMOVEDIR)
639 		return (EINVAL);
640 
641 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
642 	LCONVPATHEXIST_AT(td, args->pathname, &path, dfd);
643 
644 #ifdef DEBUG
645 	if (ldebug(unlinkat))
646 		printf(ARGS(unlinkat, "%s"), path);
647 #endif
648 
649 	if (args->flag & LINUX_AT_REMOVEDIR)
650 		error = kern_rmdirat(td, dfd, path, UIO_SYSSPACE);
651 	else
652 		error = kern_unlinkat(td, dfd, path, UIO_SYSSPACE, 0);
653 	if (error == EPERM && !(args->flag & LINUX_AT_REMOVEDIR)) {
654 		/* Introduce POSIX noncompliant behaviour of Linux */
655 		if (kern_statat(td, AT_SYMLINK_NOFOLLOW, dfd, path,
656 		    UIO_SYSSPACE, &st) == 0 && S_ISDIR(st.st_mode))
657 			error = EISDIR;
658 	}
659 	LFREEPATH(path);
660 	return (error);
661 }
662 int
663 linux_chdir(struct thread *td, struct linux_chdir_args *args)
664 {
665 	char *path;
666 	int error;
667 
668 	LCONVPATHEXIST(td, args->path, &path);
669 
670 #ifdef DEBUG
671 	if (ldebug(chdir))
672 		printf(ARGS(chdir, "%s"), path);
673 #endif
674 	error = kern_chdir(td, path, UIO_SYSSPACE);
675 	LFREEPATH(path);
676 	return (error);
677 }
678 
679 int
680 linux_chmod(struct thread *td, struct linux_chmod_args *args)
681 {
682 	char *path;
683 	int error;
684 
685 	LCONVPATHEXIST(td, args->path, &path);
686 
687 #ifdef DEBUG
688 	if (ldebug(chmod))
689 		printf(ARGS(chmod, "%s, %d"), path, args->mode);
690 #endif
691 	error = kern_chmod(td, path, UIO_SYSSPACE, args->mode);
692 	LFREEPATH(path);
693 	return (error);
694 }
695 
696 int
697 linux_fchmodat(struct thread *td, struct linux_fchmodat_args *args)
698 {
699 	char *path;
700 	int error, dfd;
701 
702 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
703 	LCONVPATHEXIST_AT(td, args->filename, &path, dfd);
704 
705 #ifdef DEBUG
706 	if (ldebug(fchmodat))
707 		printf(ARGS(fchmodat, "%s, %d"), path, args->mode);
708 #endif
709 
710 	error = kern_fchmodat(td, dfd, path, UIO_SYSSPACE, args->mode, 0);
711 	LFREEPATH(path);
712 	return (error);
713 }
714 
715 int
716 linux_mkdir(struct thread *td, struct linux_mkdir_args *args)
717 {
718 	char *path;
719 	int error;
720 
721 	LCONVPATHCREAT(td, args->path, &path);
722 
723 #ifdef DEBUG
724 	if (ldebug(mkdir))
725 		printf(ARGS(mkdir, "%s, %d"), path, args->mode);
726 #endif
727 	error = kern_mkdir(td, path, UIO_SYSSPACE, args->mode);
728 	LFREEPATH(path);
729 	return (error);
730 }
731 
732 int
733 linux_mkdirat(struct thread *td, struct linux_mkdirat_args *args)
734 {
735 	char *path;
736 	int error, dfd;
737 
738 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
739 	LCONVPATHCREAT_AT(td, args->pathname, &path, dfd);
740 
741 #ifdef DEBUG
742 	if (ldebug(mkdirat))
743 		printf(ARGS(mkdirat, "%s, %d"), path, args->mode);
744 #endif
745 	error = kern_mkdirat(td, dfd, path, UIO_SYSSPACE, args->mode);
746 	LFREEPATH(path);
747 	return (error);
748 }
749 
750 int
751 linux_rmdir(struct thread *td, struct linux_rmdir_args *args)
752 {
753 	char *path;
754 	int error;
755 
756 	LCONVPATHEXIST(td, args->path, &path);
757 
758 #ifdef DEBUG
759 	if (ldebug(rmdir))
760 		printf(ARGS(rmdir, "%s"), path);
761 #endif
762 	error = kern_rmdir(td, path, UIO_SYSSPACE);
763 	LFREEPATH(path);
764 	return (error);
765 }
766 
767 int
768 linux_rename(struct thread *td, struct linux_rename_args *args)
769 {
770 	char *from, *to;
771 	int error;
772 
773 	LCONVPATHEXIST(td, args->from, &from);
774 	/* Expand LCONVPATHCREATE so that `from' can be freed on errors */
775 	error = linux_emul_convpath(td, args->to, UIO_USERSPACE, &to, 1, AT_FDCWD);
776 	if (to == NULL) {
777 		LFREEPATH(from);
778 		return (error);
779 	}
780 
781 #ifdef DEBUG
782 	if (ldebug(rename))
783 		printf(ARGS(rename, "%s, %s"), from, to);
784 #endif
785 	error = kern_rename(td, from, to, UIO_SYSSPACE);
786 	LFREEPATH(from);
787 	LFREEPATH(to);
788 	return (error);
789 }
790 
791 int
792 linux_renameat(struct thread *td, struct linux_renameat_args *args)
793 {
794 	char *from, *to;
795 	int error, olddfd, newdfd;
796 
797 	olddfd = (args->olddfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->olddfd;
798 	newdfd = (args->newdfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->newdfd;
799 	LCONVPATHEXIST_AT(td, args->oldname, &from, olddfd);
800 	/* Expand LCONVPATHCREATE so that `from' can be freed on errors */
801 	error = linux_emul_convpath(td, args->newname, UIO_USERSPACE, &to, 1, newdfd);
802 	if (to == NULL) {
803 		LFREEPATH(from);
804 		return (error);
805 	}
806 
807 #ifdef DEBUG
808 	if (ldebug(renameat))
809 		printf(ARGS(renameat, "%s, %s"), from, to);
810 #endif
811 	error = kern_renameat(td, olddfd, from, newdfd, to, UIO_SYSSPACE);
812 	LFREEPATH(from);
813 	LFREEPATH(to);
814 	return (error);
815 }
816 
817 int
818 linux_symlink(struct thread *td, struct linux_symlink_args *args)
819 {
820 	char *path, *to;
821 	int error;
822 
823 	LCONVPATHEXIST(td, args->path, &path);
824 	/* Expand LCONVPATHCREATE so that `path' can be freed on errors */
825 	error = linux_emul_convpath(td, args->to, UIO_USERSPACE, &to, 1, AT_FDCWD);
826 	if (to == NULL) {
827 		LFREEPATH(path);
828 		return (error);
829 	}
830 
831 #ifdef DEBUG
832 	if (ldebug(symlink))
833 		printf(ARGS(symlink, "%s, %s"), path, to);
834 #endif
835 	error = kern_symlink(td, path, to, UIO_SYSSPACE);
836 	LFREEPATH(path);
837 	LFREEPATH(to);
838 	return (error);
839 }
840 
841 int
842 linux_symlinkat(struct thread *td, struct linux_symlinkat_args *args)
843 {
844 	char *path, *to;
845 	int error, dfd;
846 
847 	dfd = (args->newdfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->newdfd;
848 	LCONVPATHEXIST_AT(td, args->oldname, &path, dfd);
849 	/* Expand LCONVPATHCREATE so that `path' can be freed on errors */
850 	error = linux_emul_convpath(td, args->newname, UIO_USERSPACE, &to, 1, dfd);
851 	if (to == NULL) {
852 		LFREEPATH(path);
853 		return (error);
854 	}
855 
856 #ifdef DEBUG
857 	if (ldebug(symlinkat))
858 		printf(ARGS(symlinkat, "%s, %s"), path, to);
859 #endif
860 
861 	error = kern_symlinkat(td, path, dfd, to, UIO_SYSSPACE);
862 	LFREEPATH(path);
863 	LFREEPATH(to);
864 	return (error);
865 }
866 
867 int
868 linux_readlink(struct thread *td, struct linux_readlink_args *args)
869 {
870 	char *name;
871 	int error;
872 
873 	LCONVPATHEXIST(td, args->name, &name);
874 
875 #ifdef DEBUG
876 	if (ldebug(readlink))
877 		printf(ARGS(readlink, "%s, %p, %d"), name, (void *)args->buf,
878 		    args->count);
879 #endif
880 	error = kern_readlink(td, name, UIO_SYSSPACE, args->buf, UIO_USERSPACE,
881 	    args->count);
882 	LFREEPATH(name);
883 	return (error);
884 }
885 
886 int
887 linux_readlinkat(struct thread *td, struct linux_readlinkat_args *args)
888 {
889 	char *name;
890 	int error, dfd;
891 
892 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
893 	LCONVPATHEXIST_AT(td, args->path, &name, dfd);
894 
895 #ifdef DEBUG
896 	if (ldebug(readlinkat))
897 		printf(ARGS(readlinkat, "%s, %p, %d"), name, (void *)args->buf,
898 		    args->bufsiz);
899 #endif
900 
901 	error = kern_readlinkat(td, dfd, name, UIO_SYSSPACE, args->buf,
902 	    UIO_USERSPACE, args->bufsiz);
903 	LFREEPATH(name);
904 	return (error);
905 }
906 
907 int
908 linux_truncate(struct thread *td, struct linux_truncate_args *args)
909 {
910 	char *path;
911 	int error;
912 
913 	LCONVPATHEXIST(td, args->path, &path);
914 
915 #ifdef DEBUG
916 	if (ldebug(truncate))
917 		printf(ARGS(truncate, "%s, %ld"), path, (long)args->length);
918 #endif
919 
920 	error = kern_truncate(td, path, UIO_SYSSPACE, args->length);
921 	LFREEPATH(path);
922 	return (error);
923 }
924 
925 int
926 linux_truncate64(struct thread *td, struct linux_truncate64_args *args)
927 {
928 	char *path;
929 	int error;
930 
931 	LCONVPATHEXIST(td, args->path, &path);
932 
933 #ifdef DEBUG
934 	if (ldebug(truncate64))
935 		printf(ARGS(truncate64, "%s, %jd"), path, args->length);
936 #endif
937 
938 	error = kern_truncate(td, path, UIO_SYSSPACE, args->length);
939 	LFREEPATH(path);
940 	return (error);
941 }
942 int
943 linux_ftruncate(struct thread *td, struct linux_ftruncate_args *args)
944 {
945 	struct ftruncate_args /* {
946 		int fd;
947 		int pad;
948 		off_t length;
949 		} */ nuap;
950 
951 	nuap.fd = args->fd;
952 	nuap.length = args->length;
953 	return (sys_ftruncate(td, &nuap));
954 }
955 
956 int
957 linux_link(struct thread *td, struct linux_link_args *args)
958 {
959 	char *path, *to;
960 	int error;
961 
962 	LCONVPATHEXIST(td, args->path, &path);
963 	/* Expand LCONVPATHCREATE so that `path' can be freed on errors */
964 	error = linux_emul_convpath(td, args->to, UIO_USERSPACE, &to, 1, AT_FDCWD);
965 	if (to == NULL) {
966 		LFREEPATH(path);
967 		return (error);
968 	}
969 
970 #ifdef DEBUG
971 	if (ldebug(link))
972 		printf(ARGS(link, "%s, %s"), path, to);
973 #endif
974 	error = kern_link(td, path, to, UIO_SYSSPACE);
975 	LFREEPATH(path);
976 	LFREEPATH(to);
977 	return (error);
978 }
979 
980 int
981 linux_linkat(struct thread *td, struct linux_linkat_args *args)
982 {
983 	char *path, *to;
984 	int error, olddfd, newdfd, follow;
985 
986 	if (args->flag & ~LINUX_AT_SYMLINK_FOLLOW)
987 		return (EINVAL);
988 
989 	olddfd = (args->olddfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->olddfd;
990 	newdfd = (args->newdfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->newdfd;
991 	LCONVPATHEXIST_AT(td, args->oldname, &path, olddfd);
992 	/* Expand LCONVPATHCREATE so that `path' can be freed on errors */
993 	error = linux_emul_convpath(td, args->newname, UIO_USERSPACE, &to, 1, newdfd);
994 	if (to == NULL) {
995 		LFREEPATH(path);
996 		return (error);
997 	}
998 
999 #ifdef DEBUG
1000 	if (ldebug(linkat))
1001 		printf(ARGS(linkat, "%i, %s, %i, %s, %i"), args->olddfd, path,
1002 			args->newdfd, to, args->flag);
1003 #endif
1004 
1005 	follow = (args->flag & LINUX_AT_SYMLINK_FOLLOW) == 0 ? NOFOLLOW :
1006 	    FOLLOW;
1007 	error = kern_linkat(td, olddfd, newdfd, path, to, UIO_SYSSPACE, follow);
1008 	LFREEPATH(path);
1009 	LFREEPATH(to);
1010 	return (error);
1011 }
1012 
1013 int
1014 linux_fdatasync(td, uap)
1015 	struct thread *td;
1016 	struct linux_fdatasync_args *uap;
1017 {
1018 	struct fsync_args bsd;
1019 
1020 	bsd.fd = uap->fd;
1021 	return sys_fsync(td, &bsd);
1022 }
1023 
1024 int
1025 linux_pread(td, uap)
1026 	struct thread *td;
1027 	struct linux_pread_args *uap;
1028 {
1029 	struct pread_args bsd;
1030 	struct vnode *vp;
1031 	int error;
1032 
1033 	bsd.fd = uap->fd;
1034 	bsd.buf = uap->buf;
1035 	bsd.nbyte = uap->nbyte;
1036 	bsd.offset = uap->offset;
1037 
1038 	error = sys_pread(td, &bsd);
1039 
1040 	if (error == 0) {
1041    	   	/* This seems to violate POSIX but linux does it */
1042 		if ((error = fgetvp(td, uap->fd, CAP_READ, &vp)) != 0)
1043    		   	return (error);
1044 		if (vp->v_type == VDIR) {
1045    		   	vrele(vp);
1046 			return (EISDIR);
1047 		}
1048 		vrele(vp);
1049 	}
1050 
1051 	return (error);
1052 }
1053 
1054 int
1055 linux_pwrite(td, uap)
1056 	struct thread *td;
1057 	struct linux_pwrite_args *uap;
1058 {
1059 	struct pwrite_args bsd;
1060 
1061 	bsd.fd = uap->fd;
1062 	bsd.buf = uap->buf;
1063 	bsd.nbyte = uap->nbyte;
1064 	bsd.offset = uap->offset;
1065 	return sys_pwrite(td, &bsd);
1066 }
1067 
1068 int
1069 linux_mount(struct thread *td, struct linux_mount_args *args)
1070 {
1071 	struct ufs_args ufs;
1072 	char fstypename[MFSNAMELEN];
1073 	char mntonname[MNAMELEN], mntfromname[MNAMELEN];
1074 	int error;
1075 	int fsflags;
1076 	void *fsdata;
1077 
1078 	error = copyinstr(args->filesystemtype, fstypename, MFSNAMELEN - 1,
1079 	    NULL);
1080 	if (error)
1081 		return (error);
1082 	error = copyinstr(args->specialfile, mntfromname, MNAMELEN - 1, NULL);
1083 	if (error)
1084 		return (error);
1085 	error = copyinstr(args->dir, mntonname, MNAMELEN - 1, NULL);
1086 	if (error)
1087 		return (error);
1088 
1089 #ifdef DEBUG
1090 	if (ldebug(mount))
1091 		printf(ARGS(mount, "%s, %s, %s"),
1092 		    fstypename, mntfromname, mntonname);
1093 #endif
1094 
1095 	if (strcmp(fstypename, "ext2") == 0) {
1096 		strcpy(fstypename, "ext2fs");
1097 		fsdata = &ufs;
1098 		ufs.fspec = mntfromname;
1099 #define DEFAULT_ROOTID		-2
1100 		ufs.export.ex_root = DEFAULT_ROOTID;
1101 		ufs.export.ex_flags =
1102 		    args->rwflag & LINUX_MS_RDONLY ? MNT_EXRDONLY : 0;
1103 	} else if (strcmp(fstypename, "proc") == 0) {
1104 		strcpy(fstypename, "linprocfs");
1105 		fsdata = NULL;
1106 	} else if (strcmp(fstypename, "vfat") == 0) {
1107 		strcpy(fstypename, "msdosfs");
1108 		fsdata = NULL;
1109 	} else {
1110 		return (ENODEV);
1111 	}
1112 
1113 	fsflags = 0;
1114 
1115 	if ((args->rwflag & 0xffff0000) == 0xc0ed0000) {
1116 		/*
1117 		 * Linux SYNC flag is not included; the closest equivalent
1118 		 * FreeBSD has is !ASYNC, which is our default.
1119 		 */
1120 		if (args->rwflag & LINUX_MS_RDONLY)
1121 			fsflags |= MNT_RDONLY;
1122 		if (args->rwflag & LINUX_MS_NOSUID)
1123 			fsflags |= MNT_NOSUID;
1124 		if (args->rwflag & LINUX_MS_NOEXEC)
1125 			fsflags |= MNT_NOEXEC;
1126 		if (args->rwflag & LINUX_MS_REMOUNT)
1127 			fsflags |= MNT_UPDATE;
1128 	}
1129 
1130 	if (strcmp(fstypename, "linprocfs") == 0) {
1131 		error = kernel_vmount(fsflags,
1132 			"fstype", fstypename,
1133 			"fspath", mntonname,
1134 			NULL);
1135 	} else if (strcmp(fstypename, "msdosfs") == 0) {
1136 		error = kernel_vmount(fsflags,
1137 			"fstype", fstypename,
1138 			"fspath", mntonname,
1139 			"from", mntfromname,
1140 			NULL);
1141 	} else
1142 		error = EOPNOTSUPP;
1143 	return (error);
1144 }
1145 
1146 int
1147 linux_oldumount(struct thread *td, struct linux_oldumount_args *args)
1148 {
1149 	struct linux_umount_args args2;
1150 
1151 	args2.path = args->path;
1152 	args2.flags = 0;
1153 	return (linux_umount(td, &args2));
1154 }
1155 
1156 int
1157 linux_umount(struct thread *td, struct linux_umount_args *args)
1158 {
1159 	struct unmount_args bsd;
1160 
1161 	bsd.path = args->path;
1162 	bsd.flags = args->flags;	/* XXX correct? */
1163 	return (sys_unmount(td, &bsd));
1164 }
1165 
1166 /*
1167  * fcntl family of syscalls
1168  */
1169 
1170 struct l_flock {
1171 	l_short		l_type;
1172 	l_short		l_whence;
1173 	l_off_t		l_start;
1174 	l_off_t		l_len;
1175 	l_pid_t		l_pid;
1176 }
1177 #if defined(__amd64__) && defined(COMPAT_LINUX32)
1178 __packed
1179 #endif
1180 ;
1181 
1182 static void
1183 linux_to_bsd_flock(struct l_flock *linux_flock, struct flock *bsd_flock)
1184 {
1185 	switch (linux_flock->l_type) {
1186 	case LINUX_F_RDLCK:
1187 		bsd_flock->l_type = F_RDLCK;
1188 		break;
1189 	case LINUX_F_WRLCK:
1190 		bsd_flock->l_type = F_WRLCK;
1191 		break;
1192 	case LINUX_F_UNLCK:
1193 		bsd_flock->l_type = F_UNLCK;
1194 		break;
1195 	default:
1196 		bsd_flock->l_type = -1;
1197 		break;
1198 	}
1199 	bsd_flock->l_whence = linux_flock->l_whence;
1200 	bsd_flock->l_start = (off_t)linux_flock->l_start;
1201 	bsd_flock->l_len = (off_t)linux_flock->l_len;
1202 	bsd_flock->l_pid = (pid_t)linux_flock->l_pid;
1203 	bsd_flock->l_sysid = 0;
1204 }
1205 
1206 static void
1207 bsd_to_linux_flock(struct flock *bsd_flock, struct l_flock *linux_flock)
1208 {
1209 	switch (bsd_flock->l_type) {
1210 	case F_RDLCK:
1211 		linux_flock->l_type = LINUX_F_RDLCK;
1212 		break;
1213 	case F_WRLCK:
1214 		linux_flock->l_type = LINUX_F_WRLCK;
1215 		break;
1216 	case F_UNLCK:
1217 		linux_flock->l_type = LINUX_F_UNLCK;
1218 		break;
1219 	}
1220 	linux_flock->l_whence = bsd_flock->l_whence;
1221 	linux_flock->l_start = (l_off_t)bsd_flock->l_start;
1222 	linux_flock->l_len = (l_off_t)bsd_flock->l_len;
1223 	linux_flock->l_pid = (l_pid_t)bsd_flock->l_pid;
1224 }
1225 
1226 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1227 struct l_flock64 {
1228 	l_short		l_type;
1229 	l_short		l_whence;
1230 	l_loff_t	l_start;
1231 	l_loff_t	l_len;
1232 	l_pid_t		l_pid;
1233 }
1234 #if defined(__amd64__) && defined(COMPAT_LINUX32)
1235 __packed
1236 #endif
1237 ;
1238 
1239 static void
1240 linux_to_bsd_flock64(struct l_flock64 *linux_flock, struct flock *bsd_flock)
1241 {
1242 	switch (linux_flock->l_type) {
1243 	case LINUX_F_RDLCK:
1244 		bsd_flock->l_type = F_RDLCK;
1245 		break;
1246 	case LINUX_F_WRLCK:
1247 		bsd_flock->l_type = F_WRLCK;
1248 		break;
1249 	case LINUX_F_UNLCK:
1250 		bsd_flock->l_type = F_UNLCK;
1251 		break;
1252 	default:
1253 		bsd_flock->l_type = -1;
1254 		break;
1255 	}
1256 	bsd_flock->l_whence = linux_flock->l_whence;
1257 	bsd_flock->l_start = (off_t)linux_flock->l_start;
1258 	bsd_flock->l_len = (off_t)linux_flock->l_len;
1259 	bsd_flock->l_pid = (pid_t)linux_flock->l_pid;
1260 	bsd_flock->l_sysid = 0;
1261 }
1262 
1263 static void
1264 bsd_to_linux_flock64(struct flock *bsd_flock, struct l_flock64 *linux_flock)
1265 {
1266 	switch (bsd_flock->l_type) {
1267 	case F_RDLCK:
1268 		linux_flock->l_type = LINUX_F_RDLCK;
1269 		break;
1270 	case F_WRLCK:
1271 		linux_flock->l_type = LINUX_F_WRLCK;
1272 		break;
1273 	case F_UNLCK:
1274 		linux_flock->l_type = LINUX_F_UNLCK;
1275 		break;
1276 	}
1277 	linux_flock->l_whence = bsd_flock->l_whence;
1278 	linux_flock->l_start = (l_loff_t)bsd_flock->l_start;
1279 	linux_flock->l_len = (l_loff_t)bsd_flock->l_len;
1280 	linux_flock->l_pid = (l_pid_t)bsd_flock->l_pid;
1281 }
1282 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1283 
1284 static int
1285 fcntl_common(struct thread *td, struct linux_fcntl64_args *args)
1286 {
1287 	struct l_flock linux_flock;
1288 	struct flock bsd_flock;
1289 	struct file *fp;
1290 	long arg;
1291 	int error, result;
1292 
1293 	switch (args->cmd) {
1294 	case LINUX_F_DUPFD:
1295 		return (kern_fcntl(td, args->fd, F_DUPFD, args->arg));
1296 
1297 	case LINUX_F_GETFD:
1298 		return (kern_fcntl(td, args->fd, F_GETFD, 0));
1299 
1300 	case LINUX_F_SETFD:
1301 		return (kern_fcntl(td, args->fd, F_SETFD, args->arg));
1302 
1303 	case LINUX_F_GETFL:
1304 		error = kern_fcntl(td, args->fd, F_GETFL, 0);
1305 		result = td->td_retval[0];
1306 		td->td_retval[0] = 0;
1307 		if (result & O_RDONLY)
1308 			td->td_retval[0] |= LINUX_O_RDONLY;
1309 		if (result & O_WRONLY)
1310 			td->td_retval[0] |= LINUX_O_WRONLY;
1311 		if (result & O_RDWR)
1312 			td->td_retval[0] |= LINUX_O_RDWR;
1313 		if (result & O_NDELAY)
1314 			td->td_retval[0] |= LINUX_O_NONBLOCK;
1315 		if (result & O_APPEND)
1316 			td->td_retval[0] |= LINUX_O_APPEND;
1317 		if (result & O_FSYNC)
1318 			td->td_retval[0] |= LINUX_O_SYNC;
1319 		if (result & O_ASYNC)
1320 			td->td_retval[0] |= LINUX_FASYNC;
1321 #ifdef LINUX_O_NOFOLLOW
1322 		if (result & O_NOFOLLOW)
1323 			td->td_retval[0] |= LINUX_O_NOFOLLOW;
1324 #endif
1325 #ifdef LINUX_O_DIRECT
1326 		if (result & O_DIRECT)
1327 			td->td_retval[0] |= LINUX_O_DIRECT;
1328 #endif
1329 		return (error);
1330 
1331 	case LINUX_F_SETFL:
1332 		arg = 0;
1333 		if (args->arg & LINUX_O_NDELAY)
1334 			arg |= O_NONBLOCK;
1335 		if (args->arg & LINUX_O_APPEND)
1336 			arg |= O_APPEND;
1337 		if (args->arg & LINUX_O_SYNC)
1338 			arg |= O_FSYNC;
1339 		if (args->arg & LINUX_FASYNC)
1340 			arg |= O_ASYNC;
1341 #ifdef LINUX_O_NOFOLLOW
1342 		if (args->arg & LINUX_O_NOFOLLOW)
1343 			arg |= O_NOFOLLOW;
1344 #endif
1345 #ifdef LINUX_O_DIRECT
1346 		if (args->arg & LINUX_O_DIRECT)
1347 			arg |= O_DIRECT;
1348 #endif
1349 		return (kern_fcntl(td, args->fd, F_SETFL, arg));
1350 
1351 	case LINUX_F_GETLK:
1352 		error = copyin((void *)args->arg, &linux_flock,
1353 		    sizeof(linux_flock));
1354 		if (error)
1355 			return (error);
1356 		linux_to_bsd_flock(&linux_flock, &bsd_flock);
1357 		error = kern_fcntl(td, args->fd, F_GETLK, (intptr_t)&bsd_flock);
1358 		if (error)
1359 			return (error);
1360 		bsd_to_linux_flock(&bsd_flock, &linux_flock);
1361 		return (copyout(&linux_flock, (void *)args->arg,
1362 		    sizeof(linux_flock)));
1363 
1364 	case LINUX_F_SETLK:
1365 		error = copyin((void *)args->arg, &linux_flock,
1366 		    sizeof(linux_flock));
1367 		if (error)
1368 			return (error);
1369 		linux_to_bsd_flock(&linux_flock, &bsd_flock);
1370 		return (kern_fcntl(td, args->fd, F_SETLK,
1371 		    (intptr_t)&bsd_flock));
1372 
1373 	case LINUX_F_SETLKW:
1374 		error = copyin((void *)args->arg, &linux_flock,
1375 		    sizeof(linux_flock));
1376 		if (error)
1377 			return (error);
1378 		linux_to_bsd_flock(&linux_flock, &bsd_flock);
1379 		return (kern_fcntl(td, args->fd, F_SETLKW,
1380 		     (intptr_t)&bsd_flock));
1381 
1382 	case LINUX_F_GETOWN:
1383 		return (kern_fcntl(td, args->fd, F_GETOWN, 0));
1384 
1385 	case LINUX_F_SETOWN:
1386 		/*
1387 		 * XXX some Linux applications depend on F_SETOWN having no
1388 		 * significant effect for pipes (SIGIO is not delivered for
1389 		 * pipes under Linux-2.2.35 at least).
1390 		 */
1391 		error = fget(td, args->fd, CAP_FCNTL, &fp);
1392 		if (error)
1393 			return (error);
1394 		if (fp->f_type == DTYPE_PIPE) {
1395 			fdrop(fp, td);
1396 			return (EINVAL);
1397 		}
1398 		fdrop(fp, td);
1399 
1400 		return (kern_fcntl(td, args->fd, F_SETOWN, args->arg));
1401 	}
1402 
1403 	return (EINVAL);
1404 }
1405 
1406 int
1407 linux_fcntl(struct thread *td, struct linux_fcntl_args *args)
1408 {
1409 	struct linux_fcntl64_args args64;
1410 
1411 #ifdef DEBUG
1412 	if (ldebug(fcntl))
1413 		printf(ARGS(fcntl, "%d, %08x, *"), args->fd, args->cmd);
1414 #endif
1415 
1416 	args64.fd = args->fd;
1417 	args64.cmd = args->cmd;
1418 	args64.arg = args->arg;
1419 	return (fcntl_common(td, &args64));
1420 }
1421 
1422 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1423 int
1424 linux_fcntl64(struct thread *td, struct linux_fcntl64_args *args)
1425 {
1426 	struct l_flock64 linux_flock;
1427 	struct flock bsd_flock;
1428 	int error;
1429 
1430 #ifdef DEBUG
1431 	if (ldebug(fcntl64))
1432 		printf(ARGS(fcntl64, "%d, %08x, *"), args->fd, args->cmd);
1433 #endif
1434 
1435 	switch (args->cmd) {
1436 	case LINUX_F_GETLK64:
1437 		error = copyin((void *)args->arg, &linux_flock,
1438 		    sizeof(linux_flock));
1439 		if (error)
1440 			return (error);
1441 		linux_to_bsd_flock64(&linux_flock, &bsd_flock);
1442 		error = kern_fcntl(td, args->fd, F_GETLK, (intptr_t)&bsd_flock);
1443 		if (error)
1444 			return (error);
1445 		bsd_to_linux_flock64(&bsd_flock, &linux_flock);
1446 		return (copyout(&linux_flock, (void *)args->arg,
1447 			    sizeof(linux_flock)));
1448 
1449 	case LINUX_F_SETLK64:
1450 		error = copyin((void *)args->arg, &linux_flock,
1451 		    sizeof(linux_flock));
1452 		if (error)
1453 			return (error);
1454 		linux_to_bsd_flock64(&linux_flock, &bsd_flock);
1455 		return (kern_fcntl(td, args->fd, F_SETLK,
1456 		    (intptr_t)&bsd_flock));
1457 
1458 	case LINUX_F_SETLKW64:
1459 		error = copyin((void *)args->arg, &linux_flock,
1460 		    sizeof(linux_flock));
1461 		if (error)
1462 			return (error);
1463 		linux_to_bsd_flock64(&linux_flock, &bsd_flock);
1464 		return (kern_fcntl(td, args->fd, F_SETLKW,
1465 		    (intptr_t)&bsd_flock));
1466 	}
1467 
1468 	return (fcntl_common(td, args));
1469 }
1470 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1471 
1472 int
1473 linux_chown(struct thread *td, struct linux_chown_args *args)
1474 {
1475 	char *path;
1476 	int error;
1477 
1478 	LCONVPATHEXIST(td, args->path, &path);
1479 
1480 #ifdef DEBUG
1481 	if (ldebug(chown))
1482 		printf(ARGS(chown, "%s, %d, %d"), path, args->uid, args->gid);
1483 #endif
1484 	error = kern_chown(td, path, UIO_SYSSPACE, args->uid, args->gid);
1485 	LFREEPATH(path);
1486 	return (error);
1487 }
1488 
1489 int
1490 linux_fchownat(struct thread *td, struct linux_fchownat_args *args)
1491 {
1492 	char *path;
1493 	int error, dfd, flag;
1494 
1495 	if (args->flag & ~LINUX_AT_SYMLINK_NOFOLLOW)
1496 		return (EINVAL);
1497 
1498 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD :  args->dfd;
1499 	LCONVPATHEXIST_AT(td, args->filename, &path, dfd);
1500 
1501 #ifdef DEBUG
1502 	if (ldebug(fchownat))
1503 		printf(ARGS(fchownat, "%s, %d, %d"), path, args->uid, args->gid);
1504 #endif
1505 
1506 	flag = (args->flag & LINUX_AT_SYMLINK_NOFOLLOW) == 0 ? 0 :
1507 	    AT_SYMLINK_NOFOLLOW;
1508 	error = kern_fchownat(td, dfd, path, UIO_SYSSPACE, args->uid, args->gid,
1509 	    flag);
1510 	LFREEPATH(path);
1511 	return (error);
1512 }
1513 
1514 int
1515 linux_lchown(struct thread *td, struct linux_lchown_args *args)
1516 {
1517 	char *path;
1518 	int error;
1519 
1520 	LCONVPATHEXIST(td, args->path, &path);
1521 
1522 #ifdef DEBUG
1523 	if (ldebug(lchown))
1524 		printf(ARGS(lchown, "%s, %d, %d"), path, args->uid, args->gid);
1525 #endif
1526 	error = kern_lchown(td, path, UIO_SYSSPACE, args->uid, args->gid);
1527 	LFREEPATH(path);
1528 	return (error);
1529 }
1530 
1531 static int
1532 convert_fadvice(int advice)
1533 {
1534 	switch (advice) {
1535 	case LINUX_POSIX_FADV_NORMAL:
1536 		return (POSIX_FADV_NORMAL);
1537 	case LINUX_POSIX_FADV_RANDOM:
1538 		return (POSIX_FADV_RANDOM);
1539 	case LINUX_POSIX_FADV_SEQUENTIAL:
1540 		return (POSIX_FADV_SEQUENTIAL);
1541 	case LINUX_POSIX_FADV_WILLNEED:
1542 		return (POSIX_FADV_WILLNEED);
1543 	case LINUX_POSIX_FADV_DONTNEED:
1544 		return (POSIX_FADV_DONTNEED);
1545 	case LINUX_POSIX_FADV_NOREUSE:
1546 		return (POSIX_FADV_NOREUSE);
1547 	default:
1548 		return (-1);
1549 	}
1550 }
1551 
1552 int
1553 linux_fadvise64(struct thread *td, struct linux_fadvise64_args *args)
1554 {
1555 	int advice;
1556 
1557 	advice = convert_fadvice(args->advice);
1558 	if (advice == -1)
1559 		return (EINVAL);
1560 	return (kern_posix_fadvise(td, args->fd, args->offset, args->len,
1561 	    advice));
1562 }
1563 
1564 int
1565 linux_fadvise64_64(struct thread *td, struct linux_fadvise64_64_args *args)
1566 {
1567 	int advice;
1568 
1569 	advice = convert_fadvice(args->advice);
1570 	if (advice == -1)
1571 		return (EINVAL);
1572 	return (kern_posix_fadvise(td, args->fd, args->offset, args->len,
1573 	    advice));
1574 }
1575 
1576 int
1577 linux_pipe(struct thread *td, struct linux_pipe_args *args)
1578 {
1579 	int fildes[2];
1580 	int error;
1581 
1582 #ifdef DEBUG
1583 	if (ldebug(pipe))
1584 		printf(ARGS(pipe, "*"));
1585 #endif
1586 
1587 	error = do_pipe(td, fildes, 0);
1588 	if (error)
1589 		return (error);
1590 
1591 	/* XXX: Close descriptors on error. */
1592 	return (copyout(fildes, args->pipefds, sizeof(fildes)));
1593 }
1594 
1595 int
1596 linux_pipe2(struct thread *td, struct linux_pipe2_args *args)
1597 {
1598 	int fildes[2];
1599 	int error, flags;
1600 
1601 #ifdef DEBUG
1602 	if (ldebug(pipe2))
1603 		printf(ARGS(pipe2, "*, %d"), args->flags);
1604 #endif
1605 
1606 	if ((args->flags & ~(LINUX_O_NONBLOCK | LINUX_O_CLOEXEC)) != 0)
1607 		return (EINVAL);
1608 
1609 	flags = 0;
1610 	if ((args->flags & LINUX_O_NONBLOCK) != 0)
1611 		flags |= O_NONBLOCK;
1612 	if ((args->flags & LINUX_O_CLOEXEC) != 0)
1613 		flags |= O_CLOEXEC;
1614 	error = do_pipe(td, fildes, flags);
1615 	if (error)
1616 		return (error);
1617 
1618 	/* XXX: Close descriptors on error. */
1619 	return (copyout(fildes, args->pipefds, sizeof(fildes)));
1620 }
1621