xref: /freebsd/sys/compat/linux/linux_file.c (revision f05cddf9)
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 int
74 linux_creat(struct thread *td, struct linux_creat_args *args)
75 {
76     char *path;
77     int error;
78 
79     LCONVPATHEXIST(td, args->path, &path);
80 
81 #ifdef DEBUG
82 	if (ldebug(creat))
83 		printf(ARGS(creat, "%s, %d"), path, args->mode);
84 #endif
85     error = kern_open(td, path, UIO_SYSSPACE, O_WRONLY | O_CREAT | O_TRUNC,
86 	args->mode);
87     LFREEPATH(path);
88     return (error);
89 }
90 
91 
92 static int
93 linux_common_open(struct thread *td, int dirfd, char *path, int l_flags, int mode)
94 {
95     struct proc *p = td->td_proc;
96     struct file *fp;
97     int fd;
98     int bsd_flags, error;
99 
100     bsd_flags = 0;
101     switch (l_flags & LINUX_O_ACCMODE) {
102     case LINUX_O_WRONLY:
103 	bsd_flags |= O_WRONLY;
104 	break;
105     case LINUX_O_RDWR:
106 	bsd_flags |= O_RDWR;
107 	break;
108     default:
109 	bsd_flags |= O_RDONLY;
110     }
111     if (l_flags & LINUX_O_NDELAY)
112 	bsd_flags |= O_NONBLOCK;
113     if (l_flags & LINUX_O_APPEND)
114 	bsd_flags |= O_APPEND;
115     if (l_flags & LINUX_O_SYNC)
116 	bsd_flags |= O_FSYNC;
117     if (l_flags & LINUX_O_NONBLOCK)
118 	bsd_flags |= O_NONBLOCK;
119     if (l_flags & LINUX_FASYNC)
120 	bsd_flags |= O_ASYNC;
121     if (l_flags & LINUX_O_CREAT)
122 	bsd_flags |= O_CREAT;
123     if (l_flags & LINUX_O_TRUNC)
124 	bsd_flags |= O_TRUNC;
125     if (l_flags & LINUX_O_EXCL)
126 	bsd_flags |= O_EXCL;
127     if (l_flags & LINUX_O_NOCTTY)
128 	bsd_flags |= O_NOCTTY;
129     if (l_flags & LINUX_O_DIRECT)
130 	bsd_flags |= O_DIRECT;
131     if (l_flags & LINUX_O_NOFOLLOW)
132 	bsd_flags |= O_NOFOLLOW;
133     if (l_flags & LINUX_O_DIRECTORY)
134 	bsd_flags |= O_DIRECTORY;
135     /* XXX LINUX_O_NOATIME: unable to be easily implemented. */
136 
137     error = kern_openat(td, dirfd, path, UIO_SYSSPACE, bsd_flags, mode);
138 
139     if (!error) {
140 	    fd = td->td_retval[0];
141 	    /*
142 	     * XXX In between kern_open() and fget(), another process
143 	     * having the same filedesc could use that fd without
144 	     * checking below.
145 	     */
146 	    error = fget(td, fd, CAP_IOCTL, &fp);
147 	    if (!error) {
148 		    sx_slock(&proctree_lock);
149 		    PROC_LOCK(p);
150 		    if (!(bsd_flags & O_NOCTTY) &&
151 			SESS_LEADER(p) && !(p->p_flag & P_CONTROLT)) {
152 			    PROC_UNLOCK(p);
153 			    sx_unlock(&proctree_lock);
154 			    /* XXXPJD: Verify if TIOCSCTTY is allowed. */
155 			    if (fp->f_type == DTYPE_VNODE)
156 				    (void) fo_ioctl(fp, TIOCSCTTY, (caddr_t) 0,
157 					     td->td_ucred, td);
158 		    } else {
159 			    PROC_UNLOCK(p);
160 			    sx_sunlock(&proctree_lock);
161 		    }
162 		    fdrop(fp, td);
163 		    /*
164 		     * XXX as above, fdrop()/kern_close() pair is racy.
165 		     */
166 		    if (error)
167 			    kern_close(td, fd);
168 	    }
169     }
170 
171 #ifdef DEBUG
172     if (ldebug(open))
173 	    printf(LMSG("open returns error %d"), error);
174 #endif
175     LFREEPATH(path);
176     return (error);
177 }
178 
179 int
180 linux_openat(struct thread *td, struct linux_openat_args *args)
181 {
182 	char *path;
183 	int dfd;
184 
185 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
186 	if (args->flags & LINUX_O_CREAT)
187 		LCONVPATH_AT(td, args->filename, &path, 1, dfd);
188 	else
189 		LCONVPATH_AT(td, args->filename, &path, 0, dfd);
190 #ifdef DEBUG
191 	if (ldebug(openat))
192 		printf(ARGS(openat, "%i, %s, 0x%x, 0x%x"), args->dfd,
193 		    path, args->flags, args->mode);
194 #endif
195 	return (linux_common_open(td, dfd, path, args->flags, args->mode));
196 }
197 
198 int
199 linux_open(struct thread *td, struct linux_open_args *args)
200 {
201     char *path;
202 
203     if (args->flags & LINUX_O_CREAT)
204 	LCONVPATHCREAT(td, args->path, &path);
205     else
206 	LCONVPATHEXIST(td, args->path, &path);
207 
208 #ifdef DEBUG
209 	if (ldebug(open))
210 		printf(ARGS(open, "%s, 0x%x, 0x%x"),
211 		    path, args->flags, args->mode);
212 #endif
213 
214 	return (linux_common_open(td, AT_FDCWD, path, args->flags, args->mode));
215 }
216 
217 int
218 linux_lseek(struct thread *td, struct linux_lseek_args *args)
219 {
220 
221     struct lseek_args /* {
222 	int fd;
223 	int pad;
224 	off_t offset;
225 	int whence;
226     } */ tmp_args;
227     int error;
228 
229 #ifdef DEBUG
230 	if (ldebug(lseek))
231 		printf(ARGS(lseek, "%d, %ld, %d"),
232 		    args->fdes, (long)args->off, args->whence);
233 #endif
234     tmp_args.fd = args->fdes;
235     tmp_args.offset = (off_t)args->off;
236     tmp_args.whence = args->whence;
237     error = sys_lseek(td, &tmp_args);
238     return error;
239 }
240 
241 int
242 linux_llseek(struct thread *td, struct linux_llseek_args *args)
243 {
244 	struct lseek_args bsd_args;
245 	int error;
246 	off_t off;
247 
248 #ifdef DEBUG
249 	if (ldebug(llseek))
250 		printf(ARGS(llseek, "%d, %d:%d, %d"),
251 		    args->fd, args->ohigh, args->olow, args->whence);
252 #endif
253 	off = (args->olow) | (((off_t) args->ohigh) << 32);
254 
255 	bsd_args.fd = args->fd;
256 	bsd_args.offset = off;
257 	bsd_args.whence = args->whence;
258 
259 	if ((error = sys_lseek(td, &bsd_args)))
260 		return error;
261 
262 	if ((error = copyout(td->td_retval, args->res, sizeof (off_t))))
263 		return error;
264 
265 	td->td_retval[0] = 0;
266 	return 0;
267 }
268 
269 int
270 linux_readdir(struct thread *td, struct linux_readdir_args *args)
271 {
272 	struct linux_getdents_args lda;
273 
274 	lda.fd = args->fd;
275 	lda.dent = args->dent;
276 	lda.count = 1;
277 	return linux_getdents(td, &lda);
278 }
279 
280 /*
281  * Note that linux_getdents(2) and linux_getdents64(2) have the same
282  * arguments. They only differ in the definition of struct dirent they
283  * operate on. We use this to common the code, with the exception of
284  * accessing struct dirent. Note that linux_readdir(2) is implemented
285  * by means of linux_getdents(2). In this case we never operate on
286  * struct dirent64 and thus don't need to handle it...
287  */
288 
289 struct l_dirent {
290 	l_ulong		d_ino;
291 	l_off_t		d_off;
292 	l_ushort	d_reclen;
293 	char		d_name[LINUX_NAME_MAX + 1];
294 };
295 
296 struct l_dirent64 {
297 	uint64_t	d_ino;
298 	int64_t		d_off;
299 	l_ushort	d_reclen;
300 	u_char		d_type;
301 	char		d_name[LINUX_NAME_MAX + 1];
302 };
303 
304 /*
305  * Linux uses the last byte in the dirent buffer to store d_type,
306  * at least glibc-2.7 requires it. That is why l_dirent is padded with 2 bytes.
307  */
308 #define LINUX_RECLEN(namlen)						\
309     roundup((offsetof(struct l_dirent, d_name) + (namlen) + 2),		\
310     sizeof(l_ulong))
311 
312 #define LINUX_RECLEN64(namlen)						\
313     roundup((offsetof(struct l_dirent64, d_name) + (namlen) + 1),	\
314     sizeof(uint64_t))
315 
316 #define LINUX_MAXRECLEN		max(LINUX_RECLEN(LINUX_NAME_MAX),	\
317 				    LINUX_RECLEN64(LINUX_NAME_MAX))
318 #define	LINUX_DIRBLKSIZ		512
319 
320 static int
321 getdents_common(struct thread *td, struct linux_getdents64_args *args,
322     int is64bit)
323 {
324 	struct dirent *bdp;
325 	struct vnode *vp;
326 	caddr_t inp, buf;		/* BSD-format */
327 	int len, reclen;		/* BSD-format */
328 	caddr_t outp;			/* Linux-format */
329 	int resid, linuxreclen=0;	/* Linux-format */
330 	caddr_t lbuf;			/* Linux-format */
331 	struct file *fp;
332 	struct uio auio;
333 	struct iovec aiov;
334 	off_t off;
335 	struct l_dirent *linux_dirent;
336 	struct l_dirent64 *linux_dirent64;
337 	int buflen, error, eofflag, nbytes, justone;
338 	u_long *cookies = NULL, *cookiep;
339 	int ncookies;
340 
341 	nbytes = args->count;
342 	if (nbytes == 1) {
343 		/* readdir(2) case. Always struct dirent. */
344 		if (is64bit)
345 			return (EINVAL);
346 		nbytes = sizeof(*linux_dirent);
347 		justone = 1;
348 	} else
349 		justone = 0;
350 
351 	if ((error = getvnode(td->td_proc->p_fd, args->fd, CAP_READ, &fp)) != 0)
352 		return (error);
353 
354 	if ((fp->f_flag & FREAD) == 0) {
355 		fdrop(fp, td);
356 		return (EBADF);
357 	}
358 
359 	off = foffset_lock(fp, 0);
360 	vp = fp->f_vnode;
361 	if (vp->v_type != VDIR) {
362 		foffset_unlock(fp, off, 0);
363 		fdrop(fp, td);
364 		return (EINVAL);
365 	}
366 
367 
368 	buflen = max(LINUX_DIRBLKSIZ, nbytes);
369 	buflen = min(buflen, MAXBSIZE);
370 	buf = malloc(buflen, M_TEMP, M_WAITOK);
371 	lbuf = malloc(LINUX_MAXRECLEN, M_TEMP, M_WAITOK | M_ZERO);
372 	vn_lock(vp, LK_SHARED | LK_RETRY);
373 
374 	aiov.iov_base = buf;
375 	aiov.iov_len = buflen;
376 	auio.uio_iov = &aiov;
377 	auio.uio_iovcnt = 1;
378 	auio.uio_rw = UIO_READ;
379 	auio.uio_segflg = UIO_SYSSPACE;
380 	auio.uio_td = td;
381 	auio.uio_resid = buflen;
382 	auio.uio_offset = off;
383 
384 #ifdef MAC
385 	/*
386 	 * Do directory search MAC check using non-cached credentials.
387 	 */
388 	if ((error = mac_vnode_check_readdir(td->td_ucred, vp)))
389 		goto out;
390 #endif /* MAC */
391 	if ((error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, &ncookies,
392 		 &cookies)))
393 		goto out;
394 
395 	inp = buf;
396 	outp = (caddr_t)args->dirent;
397 	resid = nbytes;
398 	if ((len = buflen - auio.uio_resid) <= 0)
399 		goto eof;
400 
401 	cookiep = cookies;
402 
403 	if (cookies) {
404 		/*
405 		 * When using cookies, the vfs has the option of reading from
406 		 * a different offset than that supplied (UFS truncates the
407 		 * offset to a block boundary to make sure that it never reads
408 		 * partway through a directory entry, even if the directory
409 		 * has been compacted).
410 		 */
411 		while (len > 0 && ncookies > 0 && *cookiep <= off) {
412 			bdp = (struct dirent *) inp;
413 			len -= bdp->d_reclen;
414 			inp += bdp->d_reclen;
415 			cookiep++;
416 			ncookies--;
417 		}
418 	}
419 
420 	while (len > 0) {
421 		if (cookiep && ncookies == 0)
422 			break;
423 		bdp = (struct dirent *) inp;
424 		reclen = bdp->d_reclen;
425 		if (reclen & 3) {
426 			error = EFAULT;
427 			goto out;
428 		}
429 
430 		if (bdp->d_fileno == 0) {
431 			inp += reclen;
432 			if (cookiep) {
433 				off = *cookiep++;
434 				ncookies--;
435 			} else
436 				off += reclen;
437 
438 			len -= reclen;
439 			continue;
440 		}
441 
442 		linuxreclen = (is64bit)
443 		    ? LINUX_RECLEN64(bdp->d_namlen)
444 		    : LINUX_RECLEN(bdp->d_namlen);
445 
446 		if (reclen > len || resid < linuxreclen) {
447 			outp++;
448 			break;
449 		}
450 
451 		if (justone) {
452 			/* readdir(2) case. */
453 			linux_dirent = (struct l_dirent*)lbuf;
454 			linux_dirent->d_ino = bdp->d_fileno;
455 			linux_dirent->d_off = (l_off_t)linuxreclen;
456 			linux_dirent->d_reclen = (l_ushort)bdp->d_namlen;
457 			strlcpy(linux_dirent->d_name, bdp->d_name,
458 			    linuxreclen - offsetof(struct l_dirent, d_name));
459 			error = copyout(linux_dirent, outp, linuxreclen);
460 		}
461 		if (is64bit) {
462 			linux_dirent64 = (struct l_dirent64*)lbuf;
463 			linux_dirent64->d_ino = bdp->d_fileno;
464 			linux_dirent64->d_off = (cookiep)
465 			    ? (l_off_t)*cookiep
466 			    : (l_off_t)(off + reclen);
467 			linux_dirent64->d_reclen = (l_ushort)linuxreclen;
468 			linux_dirent64->d_type = bdp->d_type;
469 			strlcpy(linux_dirent64->d_name, bdp->d_name,
470 			    linuxreclen - offsetof(struct l_dirent64, d_name));
471 			error = copyout(linux_dirent64, outp, linuxreclen);
472 		} else if (!justone) {
473 			linux_dirent = (struct l_dirent*)lbuf;
474 			linux_dirent->d_ino = bdp->d_fileno;
475 			linux_dirent->d_off = (cookiep)
476 			    ? (l_off_t)*cookiep
477 			    : (l_off_t)(off + reclen);
478 			linux_dirent->d_reclen = (l_ushort)linuxreclen;
479 			/*
480 			 * Copy d_type to last byte of l_dirent buffer
481 			 */
482 			lbuf[linuxreclen-1] = bdp->d_type;
483 			strlcpy(linux_dirent->d_name, bdp->d_name,
484 			    linuxreclen - offsetof(struct l_dirent, d_name)-1);
485 			error = copyout(linux_dirent, outp, linuxreclen);
486 		}
487 
488 		if (error)
489 			goto out;
490 
491 		inp += reclen;
492 		if (cookiep) {
493 			off = *cookiep++;
494 			ncookies--;
495 		} else
496 			off += reclen;
497 
498 		outp += linuxreclen;
499 		resid -= linuxreclen;
500 		len -= reclen;
501 		if (justone)
502 			break;
503 	}
504 
505 	if (outp == (caddr_t)args->dirent) {
506 		nbytes = resid;
507 		goto eof;
508 	}
509 
510 	if (justone)
511 		nbytes = resid + linuxreclen;
512 
513 eof:
514 	td->td_retval[0] = nbytes - resid;
515 
516 out:
517 	free(cookies, M_TEMP);
518 
519 	VOP_UNLOCK(vp, 0);
520 	foffset_unlock(fp, off, 0);
521 	fdrop(fp, td);
522 	free(buf, M_TEMP);
523 	free(lbuf, M_TEMP);
524 	return (error);
525 }
526 
527 int
528 linux_getdents(struct thread *td, struct linux_getdents_args *args)
529 {
530 
531 #ifdef DEBUG
532 	if (ldebug(getdents))
533 		printf(ARGS(getdents, "%d, *, %d"), args->fd, args->count);
534 #endif
535 
536 	return (getdents_common(td, (struct linux_getdents64_args*)args, 0));
537 }
538 
539 int
540 linux_getdents64(struct thread *td, struct linux_getdents64_args *args)
541 {
542 
543 #ifdef DEBUG
544 	if (ldebug(getdents64))
545 		printf(ARGS(getdents64, "%d, *, %d"), args->fd, args->count);
546 #endif
547 
548 	return (getdents_common(td, args, 1));
549 }
550 
551 /*
552  * These exist mainly for hooks for doing /compat/linux translation.
553  */
554 
555 int
556 linux_access(struct thread *td, struct linux_access_args *args)
557 {
558 	char *path;
559 	int error;
560 
561 	/* linux convention */
562 	if (args->amode & ~(F_OK | X_OK | W_OK | R_OK))
563 		return (EINVAL);
564 
565 	LCONVPATHEXIST(td, args->path, &path);
566 
567 #ifdef DEBUG
568 	if (ldebug(access))
569 		printf(ARGS(access, "%s, %d"), path, args->amode);
570 #endif
571 	error = kern_access(td, path, UIO_SYSSPACE, args->amode);
572 	LFREEPATH(path);
573 
574 	return (error);
575 }
576 
577 int
578 linux_faccessat(struct thread *td, struct linux_faccessat_args *args)
579 {
580 	char *path;
581 	int error, dfd, flag;
582 
583 	if (args->flag & ~LINUX_AT_EACCESS)
584 		return (EINVAL);
585 	/* linux convention */
586 	if (args->amode & ~(F_OK | X_OK | W_OK | R_OK))
587 		return (EINVAL);
588 
589 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
590 	LCONVPATHEXIST_AT(td, args->filename, &path, dfd);
591 
592 #ifdef DEBUG
593 	if (ldebug(access))
594 		printf(ARGS(access, "%s, %d"), path, args->amode);
595 #endif
596 
597 	flag = (args->flag & LINUX_AT_EACCESS) == 0 ? 0 : AT_EACCESS;
598 	error = kern_accessat(td, dfd, path, UIO_SYSSPACE, flag, args->amode);
599 	LFREEPATH(path);
600 
601 	return (error);
602 }
603 
604 int
605 linux_unlink(struct thread *td, struct linux_unlink_args *args)
606 {
607 	char *path;
608 	int error;
609 	struct stat st;
610 
611 	LCONVPATHEXIST(td, args->path, &path);
612 
613 #ifdef DEBUG
614 	if (ldebug(unlink))
615 		printf(ARGS(unlink, "%s"), path);
616 #endif
617 
618 	error = kern_unlink(td, path, UIO_SYSSPACE);
619 	if (error == EPERM)
620 		/* Introduce POSIX noncompliant behaviour of Linux */
621 		if (kern_stat(td, path, UIO_SYSSPACE, &st) == 0)
622 			if (S_ISDIR(st.st_mode))
623 				error = EISDIR;
624 	LFREEPATH(path);
625 	return (error);
626 }
627 
628 int
629 linux_unlinkat(struct thread *td, struct linux_unlinkat_args *args)
630 {
631 	char *path;
632 	int error, dfd;
633 	struct stat st;
634 
635 	if (args->flag & ~LINUX_AT_REMOVEDIR)
636 		return (EINVAL);
637 
638 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
639 	LCONVPATHEXIST_AT(td, args->pathname, &path, dfd);
640 
641 #ifdef DEBUG
642 	if (ldebug(unlinkat))
643 		printf(ARGS(unlinkat, "%s"), path);
644 #endif
645 
646 	if (args->flag & LINUX_AT_REMOVEDIR)
647 		error = kern_rmdirat(td, dfd, path, UIO_SYSSPACE);
648 	else
649 		error = kern_unlinkat(td, dfd, path, UIO_SYSSPACE, 0);
650 	if (error == EPERM && !(args->flag & LINUX_AT_REMOVEDIR)) {
651 		/* Introduce POSIX noncompliant behaviour of Linux */
652 		if (kern_statat(td, AT_SYMLINK_NOFOLLOW, dfd, path,
653 		    UIO_SYSSPACE, &st) == 0 && S_ISDIR(st.st_mode))
654 			error = EISDIR;
655 	}
656 	LFREEPATH(path);
657 	return (error);
658 }
659 int
660 linux_chdir(struct thread *td, struct linux_chdir_args *args)
661 {
662 	char *path;
663 	int error;
664 
665 	LCONVPATHEXIST(td, args->path, &path);
666 
667 #ifdef DEBUG
668 	if (ldebug(chdir))
669 		printf(ARGS(chdir, "%s"), path);
670 #endif
671 	error = kern_chdir(td, path, UIO_SYSSPACE);
672 	LFREEPATH(path);
673 	return (error);
674 }
675 
676 int
677 linux_chmod(struct thread *td, struct linux_chmod_args *args)
678 {
679 	char *path;
680 	int error;
681 
682 	LCONVPATHEXIST(td, args->path, &path);
683 
684 #ifdef DEBUG
685 	if (ldebug(chmod))
686 		printf(ARGS(chmod, "%s, %d"), path, args->mode);
687 #endif
688 	error = kern_chmod(td, path, UIO_SYSSPACE, args->mode);
689 	LFREEPATH(path);
690 	return (error);
691 }
692 
693 int
694 linux_fchmodat(struct thread *td, struct linux_fchmodat_args *args)
695 {
696 	char *path;
697 	int error, dfd;
698 
699 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
700 	LCONVPATHEXIST_AT(td, args->filename, &path, dfd);
701 
702 #ifdef DEBUG
703 	if (ldebug(fchmodat))
704 		printf(ARGS(fchmodat, "%s, %d"), path, args->mode);
705 #endif
706 
707 	error = kern_fchmodat(td, dfd, path, UIO_SYSSPACE, args->mode, 0);
708 	LFREEPATH(path);
709 	return (error);
710 }
711 
712 int
713 linux_mkdir(struct thread *td, struct linux_mkdir_args *args)
714 {
715 	char *path;
716 	int error;
717 
718 	LCONVPATHCREAT(td, args->path, &path);
719 
720 #ifdef DEBUG
721 	if (ldebug(mkdir))
722 		printf(ARGS(mkdir, "%s, %d"), path, args->mode);
723 #endif
724 	error = kern_mkdir(td, path, UIO_SYSSPACE, args->mode);
725 	LFREEPATH(path);
726 	return (error);
727 }
728 
729 int
730 linux_mkdirat(struct thread *td, struct linux_mkdirat_args *args)
731 {
732 	char *path;
733 	int error, dfd;
734 
735 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
736 	LCONVPATHCREAT_AT(td, args->pathname, &path, dfd);
737 
738 #ifdef DEBUG
739 	if (ldebug(mkdirat))
740 		printf(ARGS(mkdirat, "%s, %d"), path, args->mode);
741 #endif
742 	error = kern_mkdirat(td, dfd, path, UIO_SYSSPACE, args->mode);
743 	LFREEPATH(path);
744 	return (error);
745 }
746 
747 int
748 linux_rmdir(struct thread *td, struct linux_rmdir_args *args)
749 {
750 	char *path;
751 	int error;
752 
753 	LCONVPATHEXIST(td, args->path, &path);
754 
755 #ifdef DEBUG
756 	if (ldebug(rmdir))
757 		printf(ARGS(rmdir, "%s"), path);
758 #endif
759 	error = kern_rmdir(td, path, UIO_SYSSPACE);
760 	LFREEPATH(path);
761 	return (error);
762 }
763 
764 int
765 linux_rename(struct thread *td, struct linux_rename_args *args)
766 {
767 	char *from, *to;
768 	int error;
769 
770 	LCONVPATHEXIST(td, args->from, &from);
771 	/* Expand LCONVPATHCREATE so that `from' can be freed on errors */
772 	error = linux_emul_convpath(td, args->to, UIO_USERSPACE, &to, 1, AT_FDCWD);
773 	if (to == NULL) {
774 		LFREEPATH(from);
775 		return (error);
776 	}
777 
778 #ifdef DEBUG
779 	if (ldebug(rename))
780 		printf(ARGS(rename, "%s, %s"), from, to);
781 #endif
782 	error = kern_rename(td, from, to, UIO_SYSSPACE);
783 	LFREEPATH(from);
784 	LFREEPATH(to);
785 	return (error);
786 }
787 
788 int
789 linux_renameat(struct thread *td, struct linux_renameat_args *args)
790 {
791 	char *from, *to;
792 	int error, olddfd, newdfd;
793 
794 	olddfd = (args->olddfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->olddfd;
795 	newdfd = (args->newdfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->newdfd;
796 	LCONVPATHEXIST_AT(td, args->oldname, &from, olddfd);
797 	/* Expand LCONVPATHCREATE so that `from' can be freed on errors */
798 	error = linux_emul_convpath(td, args->newname, UIO_USERSPACE, &to, 1, newdfd);
799 	if (to == NULL) {
800 		LFREEPATH(from);
801 		return (error);
802 	}
803 
804 #ifdef DEBUG
805 	if (ldebug(renameat))
806 		printf(ARGS(renameat, "%s, %s"), from, to);
807 #endif
808 	error = kern_renameat(td, olddfd, from, newdfd, to, UIO_SYSSPACE);
809 	LFREEPATH(from);
810 	LFREEPATH(to);
811 	return (error);
812 }
813 
814 int
815 linux_symlink(struct thread *td, struct linux_symlink_args *args)
816 {
817 	char *path, *to;
818 	int error;
819 
820 	LCONVPATHEXIST(td, args->path, &path);
821 	/* Expand LCONVPATHCREATE so that `path' can be freed on errors */
822 	error = linux_emul_convpath(td, args->to, UIO_USERSPACE, &to, 1, AT_FDCWD);
823 	if (to == NULL) {
824 		LFREEPATH(path);
825 		return (error);
826 	}
827 
828 #ifdef DEBUG
829 	if (ldebug(symlink))
830 		printf(ARGS(symlink, "%s, %s"), path, to);
831 #endif
832 	error = kern_symlink(td, path, to, UIO_SYSSPACE);
833 	LFREEPATH(path);
834 	LFREEPATH(to);
835 	return (error);
836 }
837 
838 int
839 linux_symlinkat(struct thread *td, struct linux_symlinkat_args *args)
840 {
841 	char *path, *to;
842 	int error, dfd;
843 
844 	dfd = (args->newdfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->newdfd;
845 	LCONVPATHEXIST_AT(td, args->oldname, &path, dfd);
846 	/* Expand LCONVPATHCREATE so that `path' can be freed on errors */
847 	error = linux_emul_convpath(td, args->newname, UIO_USERSPACE, &to, 1, dfd);
848 	if (to == NULL) {
849 		LFREEPATH(path);
850 		return (error);
851 	}
852 
853 #ifdef DEBUG
854 	if (ldebug(symlinkat))
855 		printf(ARGS(symlinkat, "%s, %s"), path, to);
856 #endif
857 
858 	error = kern_symlinkat(td, path, dfd, to, UIO_SYSSPACE);
859 	LFREEPATH(path);
860 	LFREEPATH(to);
861 	return (error);
862 }
863 
864 int
865 linux_readlink(struct thread *td, struct linux_readlink_args *args)
866 {
867 	char *name;
868 	int error;
869 
870 	LCONVPATHEXIST(td, args->name, &name);
871 
872 #ifdef DEBUG
873 	if (ldebug(readlink))
874 		printf(ARGS(readlink, "%s, %p, %d"), name, (void *)args->buf,
875 		    args->count);
876 #endif
877 	error = kern_readlink(td, name, UIO_SYSSPACE, args->buf, UIO_USERSPACE,
878 	    args->count);
879 	LFREEPATH(name);
880 	return (error);
881 }
882 
883 int
884 linux_readlinkat(struct thread *td, struct linux_readlinkat_args *args)
885 {
886 	char *name;
887 	int error, dfd;
888 
889 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
890 	LCONVPATHEXIST_AT(td, args->path, &name, dfd);
891 
892 #ifdef DEBUG
893 	if (ldebug(readlinkat))
894 		printf(ARGS(readlinkat, "%s, %p, %d"), name, (void *)args->buf,
895 		    args->bufsiz);
896 #endif
897 
898 	error = kern_readlinkat(td, dfd, name, UIO_SYSSPACE, args->buf,
899 	    UIO_USERSPACE, args->bufsiz);
900 	LFREEPATH(name);
901 	return (error);
902 }
903 
904 int
905 linux_truncate(struct thread *td, struct linux_truncate_args *args)
906 {
907 	char *path;
908 	int error;
909 
910 	LCONVPATHEXIST(td, args->path, &path);
911 
912 #ifdef DEBUG
913 	if (ldebug(truncate))
914 		printf(ARGS(truncate, "%s, %ld"), path, (long)args->length);
915 #endif
916 
917 	error = kern_truncate(td, path, UIO_SYSSPACE, args->length);
918 	LFREEPATH(path);
919 	return (error);
920 }
921 
922 int
923 linux_truncate64(struct thread *td, struct linux_truncate64_args *args)
924 {
925 	char *path;
926 	int error;
927 
928 	LCONVPATHEXIST(td, args->path, &path);
929 
930 #ifdef DEBUG
931 	if (ldebug(truncate64))
932 		printf(ARGS(truncate64, "%s, %jd"), path, args->length);
933 #endif
934 
935 	error = kern_truncate(td, path, UIO_SYSSPACE, args->length);
936 	LFREEPATH(path);
937 	return (error);
938 }
939 int
940 linux_ftruncate(struct thread *td, struct linux_ftruncate_args *args)
941 {
942 	struct ftruncate_args /* {
943 		int fd;
944 		int pad;
945 		off_t length;
946 		} */ nuap;
947 
948 	nuap.fd = args->fd;
949 	nuap.length = args->length;
950 	return (sys_ftruncate(td, &nuap));
951 }
952 
953 int
954 linux_link(struct thread *td, struct linux_link_args *args)
955 {
956 	char *path, *to;
957 	int error;
958 
959 	LCONVPATHEXIST(td, args->path, &path);
960 	/* Expand LCONVPATHCREATE so that `path' can be freed on errors */
961 	error = linux_emul_convpath(td, args->to, UIO_USERSPACE, &to, 1, AT_FDCWD);
962 	if (to == NULL) {
963 		LFREEPATH(path);
964 		return (error);
965 	}
966 
967 #ifdef DEBUG
968 	if (ldebug(link))
969 		printf(ARGS(link, "%s, %s"), path, to);
970 #endif
971 	error = kern_link(td, path, to, UIO_SYSSPACE);
972 	LFREEPATH(path);
973 	LFREEPATH(to);
974 	return (error);
975 }
976 
977 int
978 linux_linkat(struct thread *td, struct linux_linkat_args *args)
979 {
980 	char *path, *to;
981 	int error, olddfd, newdfd, follow;
982 
983 	if (args->flag & ~LINUX_AT_SYMLINK_FOLLOW)
984 		return (EINVAL);
985 
986 	olddfd = (args->olddfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->olddfd;
987 	newdfd = (args->newdfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->newdfd;
988 	LCONVPATHEXIST_AT(td, args->oldname, &path, olddfd);
989 	/* Expand LCONVPATHCREATE so that `path' can be freed on errors */
990 	error = linux_emul_convpath(td, args->newname, UIO_USERSPACE, &to, 1, newdfd);
991 	if (to == NULL) {
992 		LFREEPATH(path);
993 		return (error);
994 	}
995 
996 #ifdef DEBUG
997 	if (ldebug(linkat))
998 		printf(ARGS(linkat, "%i, %s, %i, %s, %i"), args->olddfd, path,
999 			args->newdfd, to, args->flag);
1000 #endif
1001 
1002 	follow = (args->flag & LINUX_AT_SYMLINK_FOLLOW) == 0 ? NOFOLLOW :
1003 	    FOLLOW;
1004 	error = kern_linkat(td, olddfd, newdfd, path, to, UIO_SYSSPACE, follow);
1005 	LFREEPATH(path);
1006 	LFREEPATH(to);
1007 	return (error);
1008 }
1009 
1010 int
1011 linux_fdatasync(td, uap)
1012 	struct thread *td;
1013 	struct linux_fdatasync_args *uap;
1014 {
1015 	struct fsync_args bsd;
1016 
1017 	bsd.fd = uap->fd;
1018 	return sys_fsync(td, &bsd);
1019 }
1020 
1021 int
1022 linux_pread(td, uap)
1023 	struct thread *td;
1024 	struct linux_pread_args *uap;
1025 {
1026 	struct pread_args bsd;
1027 	struct vnode *vp;
1028 	int error;
1029 
1030 	bsd.fd = uap->fd;
1031 	bsd.buf = uap->buf;
1032 	bsd.nbyte = uap->nbyte;
1033 	bsd.offset = uap->offset;
1034 
1035 	error = sys_pread(td, &bsd);
1036 
1037 	if (error == 0) {
1038 		/* This seems to violate POSIX but linux does it */
1039 		if ((error = fgetvp(td, uap->fd, CAP_PREAD, &vp)) != 0)
1040 			return (error);
1041 		if (vp->v_type == VDIR) {
1042 			vrele(vp);
1043 			return (EISDIR);
1044 		}
1045 		vrele(vp);
1046 	}
1047 
1048 	return (error);
1049 }
1050 
1051 int
1052 linux_pwrite(td, uap)
1053 	struct thread *td;
1054 	struct linux_pwrite_args *uap;
1055 {
1056 	struct pwrite_args bsd;
1057 
1058 	bsd.fd = uap->fd;
1059 	bsd.buf = uap->buf;
1060 	bsd.nbyte = uap->nbyte;
1061 	bsd.offset = uap->offset;
1062 	return sys_pwrite(td, &bsd);
1063 }
1064 
1065 int
1066 linux_mount(struct thread *td, struct linux_mount_args *args)
1067 {
1068 	struct ufs_args ufs;
1069 	char fstypename[MFSNAMELEN];
1070 	char mntonname[MNAMELEN], mntfromname[MNAMELEN];
1071 	int error;
1072 	int fsflags;
1073 	void *fsdata;
1074 
1075 	error = copyinstr(args->filesystemtype, fstypename, MFSNAMELEN - 1,
1076 	    NULL);
1077 	if (error)
1078 		return (error);
1079 	error = copyinstr(args->specialfile, mntfromname, MNAMELEN - 1, NULL);
1080 	if (error)
1081 		return (error);
1082 	error = copyinstr(args->dir, mntonname, MNAMELEN - 1, NULL);
1083 	if (error)
1084 		return (error);
1085 
1086 #ifdef DEBUG
1087 	if (ldebug(mount))
1088 		printf(ARGS(mount, "%s, %s, %s"),
1089 		    fstypename, mntfromname, mntonname);
1090 #endif
1091 
1092 	if (strcmp(fstypename, "ext2") == 0) {
1093 		strcpy(fstypename, "ext2fs");
1094 		fsdata = &ufs;
1095 		ufs.fspec = mntfromname;
1096 #define DEFAULT_ROOTID		-2
1097 		ufs.export.ex_root = DEFAULT_ROOTID;
1098 		ufs.export.ex_flags =
1099 		    args->rwflag & LINUX_MS_RDONLY ? MNT_EXRDONLY : 0;
1100 	} else if (strcmp(fstypename, "proc") == 0) {
1101 		strcpy(fstypename, "linprocfs");
1102 		fsdata = NULL;
1103 	} else if (strcmp(fstypename, "vfat") == 0) {
1104 		strcpy(fstypename, "msdosfs");
1105 		fsdata = NULL;
1106 	} else {
1107 		return (ENODEV);
1108 	}
1109 
1110 	fsflags = 0;
1111 
1112 	if ((args->rwflag & 0xffff0000) == 0xc0ed0000) {
1113 		/*
1114 		 * Linux SYNC flag is not included; the closest equivalent
1115 		 * FreeBSD has is !ASYNC, which is our default.
1116 		 */
1117 		if (args->rwflag & LINUX_MS_RDONLY)
1118 			fsflags |= MNT_RDONLY;
1119 		if (args->rwflag & LINUX_MS_NOSUID)
1120 			fsflags |= MNT_NOSUID;
1121 		if (args->rwflag & LINUX_MS_NOEXEC)
1122 			fsflags |= MNT_NOEXEC;
1123 		if (args->rwflag & LINUX_MS_REMOUNT)
1124 			fsflags |= MNT_UPDATE;
1125 	}
1126 
1127 	if (strcmp(fstypename, "linprocfs") == 0) {
1128 		error = kernel_vmount(fsflags,
1129 			"fstype", fstypename,
1130 			"fspath", mntonname,
1131 			NULL);
1132 	} else if (strcmp(fstypename, "msdosfs") == 0) {
1133 		error = kernel_vmount(fsflags,
1134 			"fstype", fstypename,
1135 			"fspath", mntonname,
1136 			"from", mntfromname,
1137 			NULL);
1138 	} else
1139 		error = EOPNOTSUPP;
1140 	return (error);
1141 }
1142 
1143 int
1144 linux_oldumount(struct thread *td, struct linux_oldumount_args *args)
1145 {
1146 	struct linux_umount_args args2;
1147 
1148 	args2.path = args->path;
1149 	args2.flags = 0;
1150 	return (linux_umount(td, &args2));
1151 }
1152 
1153 int
1154 linux_umount(struct thread *td, struct linux_umount_args *args)
1155 {
1156 	struct unmount_args bsd;
1157 
1158 	bsd.path = args->path;
1159 	bsd.flags = args->flags;	/* XXX correct? */
1160 	return (sys_unmount(td, &bsd));
1161 }
1162 
1163 /*
1164  * fcntl family of syscalls
1165  */
1166 
1167 struct l_flock {
1168 	l_short		l_type;
1169 	l_short		l_whence;
1170 	l_off_t		l_start;
1171 	l_off_t		l_len;
1172 	l_pid_t		l_pid;
1173 }
1174 #if defined(__amd64__) && defined(COMPAT_LINUX32)
1175 __packed
1176 #endif
1177 ;
1178 
1179 static void
1180 linux_to_bsd_flock(struct l_flock *linux_flock, struct flock *bsd_flock)
1181 {
1182 	switch (linux_flock->l_type) {
1183 	case LINUX_F_RDLCK:
1184 		bsd_flock->l_type = F_RDLCK;
1185 		break;
1186 	case LINUX_F_WRLCK:
1187 		bsd_flock->l_type = F_WRLCK;
1188 		break;
1189 	case LINUX_F_UNLCK:
1190 		bsd_flock->l_type = F_UNLCK;
1191 		break;
1192 	default:
1193 		bsd_flock->l_type = -1;
1194 		break;
1195 	}
1196 	bsd_flock->l_whence = linux_flock->l_whence;
1197 	bsd_flock->l_start = (off_t)linux_flock->l_start;
1198 	bsd_flock->l_len = (off_t)linux_flock->l_len;
1199 	bsd_flock->l_pid = (pid_t)linux_flock->l_pid;
1200 	bsd_flock->l_sysid = 0;
1201 }
1202 
1203 static void
1204 bsd_to_linux_flock(struct flock *bsd_flock, struct l_flock *linux_flock)
1205 {
1206 	switch (bsd_flock->l_type) {
1207 	case F_RDLCK:
1208 		linux_flock->l_type = LINUX_F_RDLCK;
1209 		break;
1210 	case F_WRLCK:
1211 		linux_flock->l_type = LINUX_F_WRLCK;
1212 		break;
1213 	case F_UNLCK:
1214 		linux_flock->l_type = LINUX_F_UNLCK;
1215 		break;
1216 	}
1217 	linux_flock->l_whence = bsd_flock->l_whence;
1218 	linux_flock->l_start = (l_off_t)bsd_flock->l_start;
1219 	linux_flock->l_len = (l_off_t)bsd_flock->l_len;
1220 	linux_flock->l_pid = (l_pid_t)bsd_flock->l_pid;
1221 }
1222 
1223 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1224 struct l_flock64 {
1225 	l_short		l_type;
1226 	l_short		l_whence;
1227 	l_loff_t	l_start;
1228 	l_loff_t	l_len;
1229 	l_pid_t		l_pid;
1230 }
1231 #if defined(__amd64__) && defined(COMPAT_LINUX32)
1232 __packed
1233 #endif
1234 ;
1235 
1236 static void
1237 linux_to_bsd_flock64(struct l_flock64 *linux_flock, struct flock *bsd_flock)
1238 {
1239 	switch (linux_flock->l_type) {
1240 	case LINUX_F_RDLCK:
1241 		bsd_flock->l_type = F_RDLCK;
1242 		break;
1243 	case LINUX_F_WRLCK:
1244 		bsd_flock->l_type = F_WRLCK;
1245 		break;
1246 	case LINUX_F_UNLCK:
1247 		bsd_flock->l_type = F_UNLCK;
1248 		break;
1249 	default:
1250 		bsd_flock->l_type = -1;
1251 		break;
1252 	}
1253 	bsd_flock->l_whence = linux_flock->l_whence;
1254 	bsd_flock->l_start = (off_t)linux_flock->l_start;
1255 	bsd_flock->l_len = (off_t)linux_flock->l_len;
1256 	bsd_flock->l_pid = (pid_t)linux_flock->l_pid;
1257 	bsd_flock->l_sysid = 0;
1258 }
1259 
1260 static void
1261 bsd_to_linux_flock64(struct flock *bsd_flock, struct l_flock64 *linux_flock)
1262 {
1263 	switch (bsd_flock->l_type) {
1264 	case F_RDLCK:
1265 		linux_flock->l_type = LINUX_F_RDLCK;
1266 		break;
1267 	case F_WRLCK:
1268 		linux_flock->l_type = LINUX_F_WRLCK;
1269 		break;
1270 	case F_UNLCK:
1271 		linux_flock->l_type = LINUX_F_UNLCK;
1272 		break;
1273 	}
1274 	linux_flock->l_whence = bsd_flock->l_whence;
1275 	linux_flock->l_start = (l_loff_t)bsd_flock->l_start;
1276 	linux_flock->l_len = (l_loff_t)bsd_flock->l_len;
1277 	linux_flock->l_pid = (l_pid_t)bsd_flock->l_pid;
1278 }
1279 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1280 
1281 static int
1282 fcntl_common(struct thread *td, struct linux_fcntl64_args *args)
1283 {
1284 	struct l_flock linux_flock;
1285 	struct flock bsd_flock;
1286 	struct file *fp;
1287 	long arg;
1288 	int error, result;
1289 
1290 	switch (args->cmd) {
1291 	case LINUX_F_DUPFD:
1292 		return (kern_fcntl(td, args->fd, F_DUPFD, args->arg));
1293 
1294 	case LINUX_F_GETFD:
1295 		return (kern_fcntl(td, args->fd, F_GETFD, 0));
1296 
1297 	case LINUX_F_SETFD:
1298 		return (kern_fcntl(td, args->fd, F_SETFD, args->arg));
1299 
1300 	case LINUX_F_GETFL:
1301 		error = kern_fcntl(td, args->fd, F_GETFL, 0);
1302 		result = td->td_retval[0];
1303 		td->td_retval[0] = 0;
1304 		if (result & O_RDONLY)
1305 			td->td_retval[0] |= LINUX_O_RDONLY;
1306 		if (result & O_WRONLY)
1307 			td->td_retval[0] |= LINUX_O_WRONLY;
1308 		if (result & O_RDWR)
1309 			td->td_retval[0] |= LINUX_O_RDWR;
1310 		if (result & O_NDELAY)
1311 			td->td_retval[0] |= LINUX_O_NONBLOCK;
1312 		if (result & O_APPEND)
1313 			td->td_retval[0] |= LINUX_O_APPEND;
1314 		if (result & O_FSYNC)
1315 			td->td_retval[0] |= LINUX_O_SYNC;
1316 		if (result & O_ASYNC)
1317 			td->td_retval[0] |= LINUX_FASYNC;
1318 #ifdef LINUX_O_NOFOLLOW
1319 		if (result & O_NOFOLLOW)
1320 			td->td_retval[0] |= LINUX_O_NOFOLLOW;
1321 #endif
1322 #ifdef LINUX_O_DIRECT
1323 		if (result & O_DIRECT)
1324 			td->td_retval[0] |= LINUX_O_DIRECT;
1325 #endif
1326 		return (error);
1327 
1328 	case LINUX_F_SETFL:
1329 		arg = 0;
1330 		if (args->arg & LINUX_O_NDELAY)
1331 			arg |= O_NONBLOCK;
1332 		if (args->arg & LINUX_O_APPEND)
1333 			arg |= O_APPEND;
1334 		if (args->arg & LINUX_O_SYNC)
1335 			arg |= O_FSYNC;
1336 		if (args->arg & LINUX_FASYNC)
1337 			arg |= O_ASYNC;
1338 #ifdef LINUX_O_NOFOLLOW
1339 		if (args->arg & LINUX_O_NOFOLLOW)
1340 			arg |= O_NOFOLLOW;
1341 #endif
1342 #ifdef LINUX_O_DIRECT
1343 		if (args->arg & LINUX_O_DIRECT)
1344 			arg |= O_DIRECT;
1345 #endif
1346 		return (kern_fcntl(td, args->fd, F_SETFL, arg));
1347 
1348 	case LINUX_F_GETLK:
1349 		error = copyin((void *)args->arg, &linux_flock,
1350 		    sizeof(linux_flock));
1351 		if (error)
1352 			return (error);
1353 		linux_to_bsd_flock(&linux_flock, &bsd_flock);
1354 		error = kern_fcntl(td, args->fd, F_GETLK, (intptr_t)&bsd_flock);
1355 		if (error)
1356 			return (error);
1357 		bsd_to_linux_flock(&bsd_flock, &linux_flock);
1358 		return (copyout(&linux_flock, (void *)args->arg,
1359 		    sizeof(linux_flock)));
1360 
1361 	case LINUX_F_SETLK:
1362 		error = copyin((void *)args->arg, &linux_flock,
1363 		    sizeof(linux_flock));
1364 		if (error)
1365 			return (error);
1366 		linux_to_bsd_flock(&linux_flock, &bsd_flock);
1367 		return (kern_fcntl(td, args->fd, F_SETLK,
1368 		    (intptr_t)&bsd_flock));
1369 
1370 	case LINUX_F_SETLKW:
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_SETLKW,
1377 		     (intptr_t)&bsd_flock));
1378 
1379 	case LINUX_F_GETOWN:
1380 		return (kern_fcntl(td, args->fd, F_GETOWN, 0));
1381 
1382 	case LINUX_F_SETOWN:
1383 		/*
1384 		 * XXX some Linux applications depend on F_SETOWN having no
1385 		 * significant effect for pipes (SIGIO is not delivered for
1386 		 * pipes under Linux-2.2.35 at least).
1387 		 */
1388 		error = fget(td, args->fd, CAP_FCNTL, &fp);
1389 		if (error)
1390 			return (error);
1391 		if (fp->f_type == DTYPE_PIPE) {
1392 			fdrop(fp, td);
1393 			return (EINVAL);
1394 		}
1395 		fdrop(fp, td);
1396 
1397 		return (kern_fcntl(td, args->fd, F_SETOWN, args->arg));
1398 	}
1399 
1400 	return (EINVAL);
1401 }
1402 
1403 int
1404 linux_fcntl(struct thread *td, struct linux_fcntl_args *args)
1405 {
1406 	struct linux_fcntl64_args args64;
1407 
1408 #ifdef DEBUG
1409 	if (ldebug(fcntl))
1410 		printf(ARGS(fcntl, "%d, %08x, *"), args->fd, args->cmd);
1411 #endif
1412 
1413 	args64.fd = args->fd;
1414 	args64.cmd = args->cmd;
1415 	args64.arg = args->arg;
1416 	return (fcntl_common(td, &args64));
1417 }
1418 
1419 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1420 int
1421 linux_fcntl64(struct thread *td, struct linux_fcntl64_args *args)
1422 {
1423 	struct l_flock64 linux_flock;
1424 	struct flock bsd_flock;
1425 	int error;
1426 
1427 #ifdef DEBUG
1428 	if (ldebug(fcntl64))
1429 		printf(ARGS(fcntl64, "%d, %08x, *"), args->fd, args->cmd);
1430 #endif
1431 
1432 	switch (args->cmd) {
1433 	case LINUX_F_GETLK64:
1434 		error = copyin((void *)args->arg, &linux_flock,
1435 		    sizeof(linux_flock));
1436 		if (error)
1437 			return (error);
1438 		linux_to_bsd_flock64(&linux_flock, &bsd_flock);
1439 		error = kern_fcntl(td, args->fd, F_GETLK, (intptr_t)&bsd_flock);
1440 		if (error)
1441 			return (error);
1442 		bsd_to_linux_flock64(&bsd_flock, &linux_flock);
1443 		return (copyout(&linux_flock, (void *)args->arg,
1444 			    sizeof(linux_flock)));
1445 
1446 	case LINUX_F_SETLK64:
1447 		error = copyin((void *)args->arg, &linux_flock,
1448 		    sizeof(linux_flock));
1449 		if (error)
1450 			return (error);
1451 		linux_to_bsd_flock64(&linux_flock, &bsd_flock);
1452 		return (kern_fcntl(td, args->fd, F_SETLK,
1453 		    (intptr_t)&bsd_flock));
1454 
1455 	case LINUX_F_SETLKW64:
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_SETLKW,
1462 		    (intptr_t)&bsd_flock));
1463 	}
1464 
1465 	return (fcntl_common(td, args));
1466 }
1467 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1468 
1469 int
1470 linux_chown(struct thread *td, struct linux_chown_args *args)
1471 {
1472 	char *path;
1473 	int error;
1474 
1475 	LCONVPATHEXIST(td, args->path, &path);
1476 
1477 #ifdef DEBUG
1478 	if (ldebug(chown))
1479 		printf(ARGS(chown, "%s, %d, %d"), path, args->uid, args->gid);
1480 #endif
1481 	error = kern_chown(td, path, UIO_SYSSPACE, args->uid, args->gid);
1482 	LFREEPATH(path);
1483 	return (error);
1484 }
1485 
1486 int
1487 linux_fchownat(struct thread *td, struct linux_fchownat_args *args)
1488 {
1489 	char *path;
1490 	int error, dfd, flag;
1491 
1492 	if (args->flag & ~LINUX_AT_SYMLINK_NOFOLLOW)
1493 		return (EINVAL);
1494 
1495 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD :  args->dfd;
1496 	LCONVPATHEXIST_AT(td, args->filename, &path, dfd);
1497 
1498 #ifdef DEBUG
1499 	if (ldebug(fchownat))
1500 		printf(ARGS(fchownat, "%s, %d, %d"), path, args->uid, args->gid);
1501 #endif
1502 
1503 	flag = (args->flag & LINUX_AT_SYMLINK_NOFOLLOW) == 0 ? 0 :
1504 	    AT_SYMLINK_NOFOLLOW;
1505 	error = kern_fchownat(td, dfd, path, UIO_SYSSPACE, args->uid, args->gid,
1506 	    flag);
1507 	LFREEPATH(path);
1508 	return (error);
1509 }
1510 
1511 int
1512 linux_lchown(struct thread *td, struct linux_lchown_args *args)
1513 {
1514 	char *path;
1515 	int error;
1516 
1517 	LCONVPATHEXIST(td, args->path, &path);
1518 
1519 #ifdef DEBUG
1520 	if (ldebug(lchown))
1521 		printf(ARGS(lchown, "%s, %d, %d"), path, args->uid, args->gid);
1522 #endif
1523 	error = kern_lchown(td, path, UIO_SYSSPACE, args->uid, args->gid);
1524 	LFREEPATH(path);
1525 	return (error);
1526 }
1527 
1528 static int
1529 convert_fadvice(int advice)
1530 {
1531 	switch (advice) {
1532 	case LINUX_POSIX_FADV_NORMAL:
1533 		return (POSIX_FADV_NORMAL);
1534 	case LINUX_POSIX_FADV_RANDOM:
1535 		return (POSIX_FADV_RANDOM);
1536 	case LINUX_POSIX_FADV_SEQUENTIAL:
1537 		return (POSIX_FADV_SEQUENTIAL);
1538 	case LINUX_POSIX_FADV_WILLNEED:
1539 		return (POSIX_FADV_WILLNEED);
1540 	case LINUX_POSIX_FADV_DONTNEED:
1541 		return (POSIX_FADV_DONTNEED);
1542 	case LINUX_POSIX_FADV_NOREUSE:
1543 		return (POSIX_FADV_NOREUSE);
1544 	default:
1545 		return (-1);
1546 	}
1547 }
1548 
1549 int
1550 linux_fadvise64(struct thread *td, struct linux_fadvise64_args *args)
1551 {
1552 	int advice;
1553 
1554 	advice = convert_fadvice(args->advice);
1555 	if (advice == -1)
1556 		return (EINVAL);
1557 	return (kern_posix_fadvise(td, args->fd, args->offset, args->len,
1558 	    advice));
1559 }
1560 
1561 int
1562 linux_fadvise64_64(struct thread *td, struct linux_fadvise64_64_args *args)
1563 {
1564 	int advice;
1565 
1566 	advice = convert_fadvice(args->advice);
1567 	if (advice == -1)
1568 		return (EINVAL);
1569 	return (kern_posix_fadvise(td, args->fd, args->offset, args->len,
1570 	    advice));
1571 }
1572 
1573 int
1574 linux_pipe(struct thread *td, struct linux_pipe_args *args)
1575 {
1576 	int fildes[2];
1577 	int error;
1578 
1579 #ifdef DEBUG
1580 	if (ldebug(pipe))
1581 		printf(ARGS(pipe, "*"));
1582 #endif
1583 
1584 	error = kern_pipe2(td, fildes, 0);
1585 	if (error)
1586 		return (error);
1587 
1588 	/* XXX: Close descriptors on error. */
1589 	return (copyout(fildes, args->pipefds, sizeof(fildes)));
1590 }
1591 
1592 int
1593 linux_pipe2(struct thread *td, struct linux_pipe2_args *args)
1594 {
1595 	int fildes[2];
1596 	int error, flags;
1597 
1598 #ifdef DEBUG
1599 	if (ldebug(pipe2))
1600 		printf(ARGS(pipe2, "*, %d"), args->flags);
1601 #endif
1602 
1603 	if ((args->flags & ~(LINUX_O_NONBLOCK | LINUX_O_CLOEXEC)) != 0)
1604 		return (EINVAL);
1605 
1606 	flags = 0;
1607 	if ((args->flags & LINUX_O_NONBLOCK) != 0)
1608 		flags |= O_NONBLOCK;
1609 	if ((args->flags & LINUX_O_CLOEXEC) != 0)
1610 		flags |= O_CLOEXEC;
1611 	error = kern_pipe2(td, fildes, flags);
1612 	if (error)
1613 		return (error);
1614 
1615 	/* XXX: Close descriptors on error. */
1616 	return (copyout(fildes, args->pipefds, sizeof(fildes)));
1617 }
1618