xref: /freebsd/sys/compat/linux/linux_file.c (revision c6ec7d31830ab1c80edae95ad5e4b9dba10c47ac)
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;
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 	if (vp->v_type != VDIR) {
363 		foffset_unlock(fp, off, 0);
364 		fdrop(fp, td);
365 		return (EINVAL);
366 	}
367 
368 
369 	buflen = max(LINUX_DIRBLKSIZ, nbytes);
370 	buflen = min(buflen, MAXBSIZE);
371 	buf = malloc(buflen, M_TEMP, M_WAITOK);
372 	lbuf = malloc(LINUX_MAXRECLEN, M_TEMP, M_WAITOK | M_ZERO);
373 	vn_lock(vp, LK_SHARED | LK_RETRY);
374 
375 	aiov.iov_base = buf;
376 	aiov.iov_len = buflen;
377 	auio.uio_iov = &aiov;
378 	auio.uio_iovcnt = 1;
379 	auio.uio_rw = UIO_READ;
380 	auio.uio_segflg = UIO_SYSSPACE;
381 	auio.uio_td = td;
382 	auio.uio_resid = buflen;
383 	auio.uio_offset = off;
384 
385 #ifdef MAC
386 	/*
387 	 * Do directory search MAC check using non-cached credentials.
388 	 */
389 	if ((error = mac_vnode_check_readdir(td->td_ucred, vp)))
390 		goto out;
391 #endif /* MAC */
392 	if ((error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, &ncookies,
393 		 &cookies)))
394 		goto out;
395 
396 	inp = buf;
397 	outp = (caddr_t)args->dirent;
398 	resid = nbytes;
399 	if ((len = buflen - auio.uio_resid) <= 0)
400 		goto eof;
401 
402 	cookiep = cookies;
403 
404 	if (cookies) {
405 		/*
406 		 * When using cookies, the vfs has the option of reading from
407 		 * a different offset than that supplied (UFS truncates the
408 		 * offset to a block boundary to make sure that it never reads
409 		 * partway through a directory entry, even if the directory
410 		 * has been compacted).
411 		 */
412 		while (len > 0 && ncookies > 0 && *cookiep <= off) {
413 			bdp = (struct dirent *) inp;
414 			len -= bdp->d_reclen;
415 			inp += bdp->d_reclen;
416 			cookiep++;
417 			ncookies--;
418 		}
419 	}
420 
421 	while (len > 0) {
422 		if (cookiep && ncookies == 0)
423 			break;
424 		bdp = (struct dirent *) inp;
425 		reclen = bdp->d_reclen;
426 		if (reclen & 3) {
427 			error = EFAULT;
428 			goto out;
429 		}
430 
431 		if (bdp->d_fileno == 0) {
432 			inp += reclen;
433 			if (cookiep) {
434 				off = *cookiep++;
435 				ncookies--;
436 			} else
437 				off += reclen;
438 
439 			len -= reclen;
440 			continue;
441 		}
442 
443 		linuxreclen = (is64bit)
444 		    ? LINUX_RECLEN64(bdp->d_namlen)
445 		    : LINUX_RECLEN(bdp->d_namlen);
446 
447 		if (reclen > len || resid < linuxreclen) {
448 			outp++;
449 			break;
450 		}
451 
452 		if (justone) {
453 			/* readdir(2) case. */
454 			linux_dirent = (struct l_dirent*)lbuf;
455 			linux_dirent->d_ino = bdp->d_fileno;
456 			linux_dirent->d_off = (l_off_t)linuxreclen;
457 			linux_dirent->d_reclen = (l_ushort)bdp->d_namlen;
458 			strlcpy(linux_dirent->d_name, bdp->d_name,
459 			    linuxreclen - offsetof(struct l_dirent, d_name));
460 			error = copyout(linux_dirent, outp, linuxreclen);
461 		}
462 		if (is64bit) {
463 			linux_dirent64 = (struct l_dirent64*)lbuf;
464 			linux_dirent64->d_ino = bdp->d_fileno;
465 			linux_dirent64->d_off = (cookiep)
466 			    ? (l_off_t)*cookiep
467 			    : (l_off_t)(off + reclen);
468 			linux_dirent64->d_reclen = (l_ushort)linuxreclen;
469 			linux_dirent64->d_type = bdp->d_type;
470 			strlcpy(linux_dirent64->d_name, bdp->d_name,
471 			    linuxreclen - offsetof(struct l_dirent64, d_name));
472 			error = copyout(linux_dirent64, outp, linuxreclen);
473 		} else if (!justone) {
474 			linux_dirent = (struct l_dirent*)lbuf;
475 			linux_dirent->d_ino = bdp->d_fileno;
476 			linux_dirent->d_off = (cookiep)
477 			    ? (l_off_t)*cookiep
478 			    : (l_off_t)(off + reclen);
479 			linux_dirent->d_reclen = (l_ushort)linuxreclen;
480 			/*
481 			 * Copy d_type to last byte of l_dirent buffer
482 			 */
483 			lbuf[linuxreclen-1] = bdp->d_type;
484 			strlcpy(linux_dirent->d_name, bdp->d_name,
485 			    linuxreclen - offsetof(struct l_dirent, d_name)-1);
486 			error = copyout(linux_dirent, outp, linuxreclen);
487 		}
488 
489 		if (error)
490 			goto out;
491 
492 		inp += reclen;
493 		if (cookiep) {
494 			off = *cookiep++;
495 			ncookies--;
496 		} else
497 			off += reclen;
498 
499 		outp += linuxreclen;
500 		resid -= linuxreclen;
501 		len -= reclen;
502 		if (justone)
503 			break;
504 	}
505 
506 	if (outp == (caddr_t)args->dirent) {
507 		nbytes = resid;
508 		goto eof;
509 	}
510 
511 	if (justone)
512 		nbytes = resid + linuxreclen;
513 
514 eof:
515 	td->td_retval[0] = nbytes - resid;
516 
517 out:
518 	if (cookies)
519 		free(cookies, M_TEMP);
520 
521 	VOP_UNLOCK(vp, 0);
522 	foffset_unlock(fp, off, 0);
523 	fdrop(fp, td);
524 	free(buf, M_TEMP);
525 	free(lbuf, M_TEMP);
526 	return (error);
527 }
528 
529 int
530 linux_getdents(struct thread *td, struct linux_getdents_args *args)
531 {
532 
533 #ifdef DEBUG
534 	if (ldebug(getdents))
535 		printf(ARGS(getdents, "%d, *, %d"), args->fd, args->count);
536 #endif
537 
538 	return (getdents_common(td, (struct linux_getdents64_args*)args, 0));
539 }
540 
541 int
542 linux_getdents64(struct thread *td, struct linux_getdents64_args *args)
543 {
544 
545 #ifdef DEBUG
546 	if (ldebug(getdents64))
547 		printf(ARGS(getdents64, "%d, *, %d"), args->fd, args->count);
548 #endif
549 
550 	return (getdents_common(td, args, 1));
551 }
552 
553 /*
554  * These exist mainly for hooks for doing /compat/linux translation.
555  */
556 
557 int
558 linux_access(struct thread *td, struct linux_access_args *args)
559 {
560 	char *path;
561 	int error;
562 
563 	/* linux convention */
564 	if (args->amode & ~(F_OK | X_OK | W_OK | R_OK))
565 		return (EINVAL);
566 
567 	LCONVPATHEXIST(td, args->path, &path);
568 
569 #ifdef DEBUG
570 	if (ldebug(access))
571 		printf(ARGS(access, "%s, %d"), path, args->amode);
572 #endif
573 	error = kern_access(td, path, UIO_SYSSPACE, args->amode);
574 	LFREEPATH(path);
575 
576 	return (error);
577 }
578 
579 int
580 linux_faccessat(struct thread *td, struct linux_faccessat_args *args)
581 {
582 	char *path;
583 	int error, dfd, flag;
584 
585 	if (args->flag & ~LINUX_AT_EACCESS)
586 		return (EINVAL);
587 	/* linux convention */
588 	if (args->amode & ~(F_OK | X_OK | W_OK | R_OK))
589 		return (EINVAL);
590 
591 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
592 	LCONVPATHEXIST_AT(td, args->filename, &path, dfd);
593 
594 #ifdef DEBUG
595 	if (ldebug(access))
596 		printf(ARGS(access, "%s, %d"), path, args->amode);
597 #endif
598 
599 	flag = (args->flag & LINUX_AT_EACCESS) == 0 ? 0 : AT_EACCESS;
600 	error = kern_accessat(td, dfd, path, UIO_SYSSPACE, flag, args->amode);
601 	LFREEPATH(path);
602 
603 	return (error);
604 }
605 
606 int
607 linux_unlink(struct thread *td, struct linux_unlink_args *args)
608 {
609 	char *path;
610 	int error;
611 	struct stat st;
612 
613 	LCONVPATHEXIST(td, args->path, &path);
614 
615 #ifdef DEBUG
616 	if (ldebug(unlink))
617 		printf(ARGS(unlink, "%s"), path);
618 #endif
619 
620 	error = kern_unlink(td, path, UIO_SYSSPACE);
621 	if (error == EPERM)
622 		/* Introduce POSIX noncompliant behaviour of Linux */
623 		if (kern_stat(td, path, UIO_SYSSPACE, &st) == 0)
624 			if (S_ISDIR(st.st_mode))
625 				error = EISDIR;
626 	LFREEPATH(path);
627 	return (error);
628 }
629 
630 int
631 linux_unlinkat(struct thread *td, struct linux_unlinkat_args *args)
632 {
633 	char *path;
634 	int error, dfd;
635 	struct stat st;
636 
637 	if (args->flag & ~LINUX_AT_REMOVEDIR)
638 		return (EINVAL);
639 
640 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
641 	LCONVPATHEXIST_AT(td, args->pathname, &path, dfd);
642 
643 #ifdef DEBUG
644 	if (ldebug(unlinkat))
645 		printf(ARGS(unlinkat, "%s"), path);
646 #endif
647 
648 	if (args->flag & LINUX_AT_REMOVEDIR)
649 		error = kern_rmdirat(td, dfd, path, UIO_SYSSPACE);
650 	else
651 		error = kern_unlinkat(td, dfd, path, UIO_SYSSPACE, 0);
652 	if (error == EPERM && !(args->flag & LINUX_AT_REMOVEDIR)) {
653 		/* Introduce POSIX noncompliant behaviour of Linux */
654 		if (kern_statat(td, AT_SYMLINK_NOFOLLOW, dfd, path,
655 		    UIO_SYSSPACE, &st) == 0 && S_ISDIR(st.st_mode))
656 			error = EISDIR;
657 	}
658 	LFREEPATH(path);
659 	return (error);
660 }
661 int
662 linux_chdir(struct thread *td, struct linux_chdir_args *args)
663 {
664 	char *path;
665 	int error;
666 
667 	LCONVPATHEXIST(td, args->path, &path);
668 
669 #ifdef DEBUG
670 	if (ldebug(chdir))
671 		printf(ARGS(chdir, "%s"), path);
672 #endif
673 	error = kern_chdir(td, path, UIO_SYSSPACE);
674 	LFREEPATH(path);
675 	return (error);
676 }
677 
678 int
679 linux_chmod(struct thread *td, struct linux_chmod_args *args)
680 {
681 	char *path;
682 	int error;
683 
684 	LCONVPATHEXIST(td, args->path, &path);
685 
686 #ifdef DEBUG
687 	if (ldebug(chmod))
688 		printf(ARGS(chmod, "%s, %d"), path, args->mode);
689 #endif
690 	error = kern_chmod(td, path, UIO_SYSSPACE, args->mode);
691 	LFREEPATH(path);
692 	return (error);
693 }
694 
695 int
696 linux_fchmodat(struct thread *td, struct linux_fchmodat_args *args)
697 {
698 	char *path;
699 	int error, dfd;
700 
701 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
702 	LCONVPATHEXIST_AT(td, args->filename, &path, dfd);
703 
704 #ifdef DEBUG
705 	if (ldebug(fchmodat))
706 		printf(ARGS(fchmodat, "%s, %d"), path, args->mode);
707 #endif
708 
709 	error = kern_fchmodat(td, dfd, path, UIO_SYSSPACE, args->mode, 0);
710 	LFREEPATH(path);
711 	return (error);
712 }
713 
714 int
715 linux_mkdir(struct thread *td, struct linux_mkdir_args *args)
716 {
717 	char *path;
718 	int error;
719 
720 	LCONVPATHCREAT(td, args->path, &path);
721 
722 #ifdef DEBUG
723 	if (ldebug(mkdir))
724 		printf(ARGS(mkdir, "%s, %d"), path, args->mode);
725 #endif
726 	error = kern_mkdir(td, path, UIO_SYSSPACE, args->mode);
727 	LFREEPATH(path);
728 	return (error);
729 }
730 
731 int
732 linux_mkdirat(struct thread *td, struct linux_mkdirat_args *args)
733 {
734 	char *path;
735 	int error, dfd;
736 
737 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
738 	LCONVPATHCREAT_AT(td, args->pathname, &path, dfd);
739 
740 #ifdef DEBUG
741 	if (ldebug(mkdirat))
742 		printf(ARGS(mkdirat, "%s, %d"), path, args->mode);
743 #endif
744 	error = kern_mkdirat(td, dfd, path, UIO_SYSSPACE, args->mode);
745 	LFREEPATH(path);
746 	return (error);
747 }
748 
749 int
750 linux_rmdir(struct thread *td, struct linux_rmdir_args *args)
751 {
752 	char *path;
753 	int error;
754 
755 	LCONVPATHEXIST(td, args->path, &path);
756 
757 #ifdef DEBUG
758 	if (ldebug(rmdir))
759 		printf(ARGS(rmdir, "%s"), path);
760 #endif
761 	error = kern_rmdir(td, path, UIO_SYSSPACE);
762 	LFREEPATH(path);
763 	return (error);
764 }
765 
766 int
767 linux_rename(struct thread *td, struct linux_rename_args *args)
768 {
769 	char *from, *to;
770 	int error;
771 
772 	LCONVPATHEXIST(td, args->from, &from);
773 	/* Expand LCONVPATHCREATE so that `from' can be freed on errors */
774 	error = linux_emul_convpath(td, args->to, UIO_USERSPACE, &to, 1, AT_FDCWD);
775 	if (to == NULL) {
776 		LFREEPATH(from);
777 		return (error);
778 	}
779 
780 #ifdef DEBUG
781 	if (ldebug(rename))
782 		printf(ARGS(rename, "%s, %s"), from, to);
783 #endif
784 	error = kern_rename(td, from, to, UIO_SYSSPACE);
785 	LFREEPATH(from);
786 	LFREEPATH(to);
787 	return (error);
788 }
789 
790 int
791 linux_renameat(struct thread *td, struct linux_renameat_args *args)
792 {
793 	char *from, *to;
794 	int error, olddfd, newdfd;
795 
796 	olddfd = (args->olddfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->olddfd;
797 	newdfd = (args->newdfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->newdfd;
798 	LCONVPATHEXIST_AT(td, args->oldname, &from, olddfd);
799 	/* Expand LCONVPATHCREATE so that `from' can be freed on errors */
800 	error = linux_emul_convpath(td, args->newname, UIO_USERSPACE, &to, 1, newdfd);
801 	if (to == NULL) {
802 		LFREEPATH(from);
803 		return (error);
804 	}
805 
806 #ifdef DEBUG
807 	if (ldebug(renameat))
808 		printf(ARGS(renameat, "%s, %s"), from, to);
809 #endif
810 	error = kern_renameat(td, olddfd, from, newdfd, to, UIO_SYSSPACE);
811 	LFREEPATH(from);
812 	LFREEPATH(to);
813 	return (error);
814 }
815 
816 int
817 linux_symlink(struct thread *td, struct linux_symlink_args *args)
818 {
819 	char *path, *to;
820 	int error;
821 
822 	LCONVPATHEXIST(td, args->path, &path);
823 	/* Expand LCONVPATHCREATE so that `path' can be freed on errors */
824 	error = linux_emul_convpath(td, args->to, UIO_USERSPACE, &to, 1, AT_FDCWD);
825 	if (to == NULL) {
826 		LFREEPATH(path);
827 		return (error);
828 	}
829 
830 #ifdef DEBUG
831 	if (ldebug(symlink))
832 		printf(ARGS(symlink, "%s, %s"), path, to);
833 #endif
834 	error = kern_symlink(td, path, to, UIO_SYSSPACE);
835 	LFREEPATH(path);
836 	LFREEPATH(to);
837 	return (error);
838 }
839 
840 int
841 linux_symlinkat(struct thread *td, struct linux_symlinkat_args *args)
842 {
843 	char *path, *to;
844 	int error, dfd;
845 
846 	dfd = (args->newdfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->newdfd;
847 	LCONVPATHEXIST_AT(td, args->oldname, &path, dfd);
848 	/* Expand LCONVPATHCREATE so that `path' can be freed on errors */
849 	error = linux_emul_convpath(td, args->newname, UIO_USERSPACE, &to, 1, dfd);
850 	if (to == NULL) {
851 		LFREEPATH(path);
852 		return (error);
853 	}
854 
855 #ifdef DEBUG
856 	if (ldebug(symlinkat))
857 		printf(ARGS(symlinkat, "%s, %s"), path, to);
858 #endif
859 
860 	error = kern_symlinkat(td, path, dfd, to, UIO_SYSSPACE);
861 	LFREEPATH(path);
862 	LFREEPATH(to);
863 	return (error);
864 }
865 
866 int
867 linux_readlink(struct thread *td, struct linux_readlink_args *args)
868 {
869 	char *name;
870 	int error;
871 
872 	LCONVPATHEXIST(td, args->name, &name);
873 
874 #ifdef DEBUG
875 	if (ldebug(readlink))
876 		printf(ARGS(readlink, "%s, %p, %d"), name, (void *)args->buf,
877 		    args->count);
878 #endif
879 	error = kern_readlink(td, name, UIO_SYSSPACE, args->buf, UIO_USERSPACE,
880 	    args->count);
881 	LFREEPATH(name);
882 	return (error);
883 }
884 
885 int
886 linux_readlinkat(struct thread *td, struct linux_readlinkat_args *args)
887 {
888 	char *name;
889 	int error, dfd;
890 
891 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
892 	LCONVPATHEXIST_AT(td, args->path, &name, dfd);
893 
894 #ifdef DEBUG
895 	if (ldebug(readlinkat))
896 		printf(ARGS(readlinkat, "%s, %p, %d"), name, (void *)args->buf,
897 		    args->bufsiz);
898 #endif
899 
900 	error = kern_readlinkat(td, dfd, name, UIO_SYSSPACE, args->buf,
901 	    UIO_USERSPACE, args->bufsiz);
902 	LFREEPATH(name);
903 	return (error);
904 }
905 
906 int
907 linux_truncate(struct thread *td, struct linux_truncate_args *args)
908 {
909 	char *path;
910 	int error;
911 
912 	LCONVPATHEXIST(td, args->path, &path);
913 
914 #ifdef DEBUG
915 	if (ldebug(truncate))
916 		printf(ARGS(truncate, "%s, %ld"), path, (long)args->length);
917 #endif
918 
919 	error = kern_truncate(td, path, UIO_SYSSPACE, args->length);
920 	LFREEPATH(path);
921 	return (error);
922 }
923 
924 int
925 linux_truncate64(struct thread *td, struct linux_truncate64_args *args)
926 {
927 	char *path;
928 	int error;
929 
930 	LCONVPATHEXIST(td, args->path, &path);
931 
932 #ifdef DEBUG
933 	if (ldebug(truncate64))
934 		printf(ARGS(truncate64, "%s, %jd"), path, args->length);
935 #endif
936 
937 	error = kern_truncate(td, path, UIO_SYSSPACE, args->length);
938 	LFREEPATH(path);
939 	return (error);
940 }
941 int
942 linux_ftruncate(struct thread *td, struct linux_ftruncate_args *args)
943 {
944 	struct ftruncate_args /* {
945 		int fd;
946 		int pad;
947 		off_t length;
948 		} */ nuap;
949 
950 	nuap.fd = args->fd;
951 	nuap.length = args->length;
952 	return (sys_ftruncate(td, &nuap));
953 }
954 
955 int
956 linux_link(struct thread *td, struct linux_link_args *args)
957 {
958 	char *path, *to;
959 	int error;
960 
961 	LCONVPATHEXIST(td, args->path, &path);
962 	/* Expand LCONVPATHCREATE so that `path' can be freed on errors */
963 	error = linux_emul_convpath(td, args->to, UIO_USERSPACE, &to, 1, AT_FDCWD);
964 	if (to == NULL) {
965 		LFREEPATH(path);
966 		return (error);
967 	}
968 
969 #ifdef DEBUG
970 	if (ldebug(link))
971 		printf(ARGS(link, "%s, %s"), path, to);
972 #endif
973 	error = kern_link(td, path, to, UIO_SYSSPACE);
974 	LFREEPATH(path);
975 	LFREEPATH(to);
976 	return (error);
977 }
978 
979 int
980 linux_linkat(struct thread *td, struct linux_linkat_args *args)
981 {
982 	char *path, *to;
983 	int error, olddfd, newdfd, follow;
984 
985 	if (args->flag & ~LINUX_AT_SYMLINK_FOLLOW)
986 		return (EINVAL);
987 
988 	olddfd = (args->olddfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->olddfd;
989 	newdfd = (args->newdfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->newdfd;
990 	LCONVPATHEXIST_AT(td, args->oldname, &path, olddfd);
991 	/* Expand LCONVPATHCREATE so that `path' can be freed on errors */
992 	error = linux_emul_convpath(td, args->newname, UIO_USERSPACE, &to, 1, newdfd);
993 	if (to == NULL) {
994 		LFREEPATH(path);
995 		return (error);
996 	}
997 
998 #ifdef DEBUG
999 	if (ldebug(linkat))
1000 		printf(ARGS(linkat, "%i, %s, %i, %s, %i"), args->olddfd, path,
1001 			args->newdfd, to, args->flag);
1002 #endif
1003 
1004 	follow = (args->flag & LINUX_AT_SYMLINK_FOLLOW) == 0 ? NOFOLLOW :
1005 	    FOLLOW;
1006 	error = kern_linkat(td, olddfd, newdfd, path, to, UIO_SYSSPACE, follow);
1007 	LFREEPATH(path);
1008 	LFREEPATH(to);
1009 	return (error);
1010 }
1011 
1012 int
1013 linux_fdatasync(td, uap)
1014 	struct thread *td;
1015 	struct linux_fdatasync_args *uap;
1016 {
1017 	struct fsync_args bsd;
1018 
1019 	bsd.fd = uap->fd;
1020 	return sys_fsync(td, &bsd);
1021 }
1022 
1023 int
1024 linux_pread(td, uap)
1025 	struct thread *td;
1026 	struct linux_pread_args *uap;
1027 {
1028 	struct pread_args bsd;
1029 	struct vnode *vp;
1030 	int error;
1031 
1032 	bsd.fd = uap->fd;
1033 	bsd.buf = uap->buf;
1034 	bsd.nbyte = uap->nbyte;
1035 	bsd.offset = uap->offset;
1036 
1037 	error = sys_pread(td, &bsd);
1038 
1039 	if (error == 0) {
1040    	   	/* This seems to violate POSIX but linux does it */
1041 		if ((error = fgetvp(td, uap->fd, CAP_READ, &vp)) != 0)
1042    		   	return (error);
1043 		if (vp->v_type == VDIR) {
1044    		   	vrele(vp);
1045 			return (EISDIR);
1046 		}
1047 		vrele(vp);
1048 	}
1049 
1050 	return (error);
1051 }
1052 
1053 int
1054 linux_pwrite(td, uap)
1055 	struct thread *td;
1056 	struct linux_pwrite_args *uap;
1057 {
1058 	struct pwrite_args bsd;
1059 
1060 	bsd.fd = uap->fd;
1061 	bsd.buf = uap->buf;
1062 	bsd.nbyte = uap->nbyte;
1063 	bsd.offset = uap->offset;
1064 	return sys_pwrite(td, &bsd);
1065 }
1066 
1067 int
1068 linux_mount(struct thread *td, struct linux_mount_args *args)
1069 {
1070 	struct ufs_args ufs;
1071 	char fstypename[MFSNAMELEN];
1072 	char mntonname[MNAMELEN], mntfromname[MNAMELEN];
1073 	int error;
1074 	int fsflags;
1075 	void *fsdata;
1076 
1077 	error = copyinstr(args->filesystemtype, fstypename, MFSNAMELEN - 1,
1078 	    NULL);
1079 	if (error)
1080 		return (error);
1081 	error = copyinstr(args->specialfile, mntfromname, MNAMELEN - 1, NULL);
1082 	if (error)
1083 		return (error);
1084 	error = copyinstr(args->dir, mntonname, MNAMELEN - 1, NULL);
1085 	if (error)
1086 		return (error);
1087 
1088 #ifdef DEBUG
1089 	if (ldebug(mount))
1090 		printf(ARGS(mount, "%s, %s, %s"),
1091 		    fstypename, mntfromname, mntonname);
1092 #endif
1093 
1094 	if (strcmp(fstypename, "ext2") == 0) {
1095 		strcpy(fstypename, "ext2fs");
1096 		fsdata = &ufs;
1097 		ufs.fspec = mntfromname;
1098 #define DEFAULT_ROOTID		-2
1099 		ufs.export.ex_root = DEFAULT_ROOTID;
1100 		ufs.export.ex_flags =
1101 		    args->rwflag & LINUX_MS_RDONLY ? MNT_EXRDONLY : 0;
1102 	} else if (strcmp(fstypename, "proc") == 0) {
1103 		strcpy(fstypename, "linprocfs");
1104 		fsdata = NULL;
1105 	} else if (strcmp(fstypename, "vfat") == 0) {
1106 		strcpy(fstypename, "msdosfs");
1107 		fsdata = NULL;
1108 	} else {
1109 		return (ENODEV);
1110 	}
1111 
1112 	fsflags = 0;
1113 
1114 	if ((args->rwflag & 0xffff0000) == 0xc0ed0000) {
1115 		/*
1116 		 * Linux SYNC flag is not included; the closest equivalent
1117 		 * FreeBSD has is !ASYNC, which is our default.
1118 		 */
1119 		if (args->rwflag & LINUX_MS_RDONLY)
1120 			fsflags |= MNT_RDONLY;
1121 		if (args->rwflag & LINUX_MS_NOSUID)
1122 			fsflags |= MNT_NOSUID;
1123 		if (args->rwflag & LINUX_MS_NOEXEC)
1124 			fsflags |= MNT_NOEXEC;
1125 		if (args->rwflag & LINUX_MS_REMOUNT)
1126 			fsflags |= MNT_UPDATE;
1127 	}
1128 
1129 	if (strcmp(fstypename, "linprocfs") == 0) {
1130 		error = kernel_vmount(fsflags,
1131 			"fstype", fstypename,
1132 			"fspath", mntonname,
1133 			NULL);
1134 	} else if (strcmp(fstypename, "msdosfs") == 0) {
1135 		error = kernel_vmount(fsflags,
1136 			"fstype", fstypename,
1137 			"fspath", mntonname,
1138 			"from", mntfromname,
1139 			NULL);
1140 	} else
1141 		error = EOPNOTSUPP;
1142 	return (error);
1143 }
1144 
1145 int
1146 linux_oldumount(struct thread *td, struct linux_oldumount_args *args)
1147 {
1148 	struct linux_umount_args args2;
1149 
1150 	args2.path = args->path;
1151 	args2.flags = 0;
1152 	return (linux_umount(td, &args2));
1153 }
1154 
1155 int
1156 linux_umount(struct thread *td, struct linux_umount_args *args)
1157 {
1158 	struct unmount_args bsd;
1159 
1160 	bsd.path = args->path;
1161 	bsd.flags = args->flags;	/* XXX correct? */
1162 	return (sys_unmount(td, &bsd));
1163 }
1164 
1165 /*
1166  * fcntl family of syscalls
1167  */
1168 
1169 struct l_flock {
1170 	l_short		l_type;
1171 	l_short		l_whence;
1172 	l_off_t		l_start;
1173 	l_off_t		l_len;
1174 	l_pid_t		l_pid;
1175 }
1176 #if defined(__amd64__) && defined(COMPAT_LINUX32)
1177 __packed
1178 #endif
1179 ;
1180 
1181 static void
1182 linux_to_bsd_flock(struct l_flock *linux_flock, struct flock *bsd_flock)
1183 {
1184 	switch (linux_flock->l_type) {
1185 	case LINUX_F_RDLCK:
1186 		bsd_flock->l_type = F_RDLCK;
1187 		break;
1188 	case LINUX_F_WRLCK:
1189 		bsd_flock->l_type = F_WRLCK;
1190 		break;
1191 	case LINUX_F_UNLCK:
1192 		bsd_flock->l_type = F_UNLCK;
1193 		break;
1194 	default:
1195 		bsd_flock->l_type = -1;
1196 		break;
1197 	}
1198 	bsd_flock->l_whence = linux_flock->l_whence;
1199 	bsd_flock->l_start = (off_t)linux_flock->l_start;
1200 	bsd_flock->l_len = (off_t)linux_flock->l_len;
1201 	bsd_flock->l_pid = (pid_t)linux_flock->l_pid;
1202 	bsd_flock->l_sysid = 0;
1203 }
1204 
1205 static void
1206 bsd_to_linux_flock(struct flock *bsd_flock, struct l_flock *linux_flock)
1207 {
1208 	switch (bsd_flock->l_type) {
1209 	case F_RDLCK:
1210 		linux_flock->l_type = LINUX_F_RDLCK;
1211 		break;
1212 	case F_WRLCK:
1213 		linux_flock->l_type = LINUX_F_WRLCK;
1214 		break;
1215 	case F_UNLCK:
1216 		linux_flock->l_type = LINUX_F_UNLCK;
1217 		break;
1218 	}
1219 	linux_flock->l_whence = bsd_flock->l_whence;
1220 	linux_flock->l_start = (l_off_t)bsd_flock->l_start;
1221 	linux_flock->l_len = (l_off_t)bsd_flock->l_len;
1222 	linux_flock->l_pid = (l_pid_t)bsd_flock->l_pid;
1223 }
1224 
1225 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1226 struct l_flock64 {
1227 	l_short		l_type;
1228 	l_short		l_whence;
1229 	l_loff_t	l_start;
1230 	l_loff_t	l_len;
1231 	l_pid_t		l_pid;
1232 }
1233 #if defined(__amd64__) && defined(COMPAT_LINUX32)
1234 __packed
1235 #endif
1236 ;
1237 
1238 static void
1239 linux_to_bsd_flock64(struct l_flock64 *linux_flock, struct flock *bsd_flock)
1240 {
1241 	switch (linux_flock->l_type) {
1242 	case LINUX_F_RDLCK:
1243 		bsd_flock->l_type = F_RDLCK;
1244 		break;
1245 	case LINUX_F_WRLCK:
1246 		bsd_flock->l_type = F_WRLCK;
1247 		break;
1248 	case LINUX_F_UNLCK:
1249 		bsd_flock->l_type = F_UNLCK;
1250 		break;
1251 	default:
1252 		bsd_flock->l_type = -1;
1253 		break;
1254 	}
1255 	bsd_flock->l_whence = linux_flock->l_whence;
1256 	bsd_flock->l_start = (off_t)linux_flock->l_start;
1257 	bsd_flock->l_len = (off_t)linux_flock->l_len;
1258 	bsd_flock->l_pid = (pid_t)linux_flock->l_pid;
1259 	bsd_flock->l_sysid = 0;
1260 }
1261 
1262 static void
1263 bsd_to_linux_flock64(struct flock *bsd_flock, struct l_flock64 *linux_flock)
1264 {
1265 	switch (bsd_flock->l_type) {
1266 	case F_RDLCK:
1267 		linux_flock->l_type = LINUX_F_RDLCK;
1268 		break;
1269 	case F_WRLCK:
1270 		linux_flock->l_type = LINUX_F_WRLCK;
1271 		break;
1272 	case F_UNLCK:
1273 		linux_flock->l_type = LINUX_F_UNLCK;
1274 		break;
1275 	}
1276 	linux_flock->l_whence = bsd_flock->l_whence;
1277 	linux_flock->l_start = (l_loff_t)bsd_flock->l_start;
1278 	linux_flock->l_len = (l_loff_t)bsd_flock->l_len;
1279 	linux_flock->l_pid = (l_pid_t)bsd_flock->l_pid;
1280 }
1281 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1282 
1283 static int
1284 fcntl_common(struct thread *td, struct linux_fcntl64_args *args)
1285 {
1286 	struct l_flock linux_flock;
1287 	struct flock bsd_flock;
1288 	struct file *fp;
1289 	long arg;
1290 	int error, result;
1291 
1292 	switch (args->cmd) {
1293 	case LINUX_F_DUPFD:
1294 		return (kern_fcntl(td, args->fd, F_DUPFD, args->arg));
1295 
1296 	case LINUX_F_GETFD:
1297 		return (kern_fcntl(td, args->fd, F_GETFD, 0));
1298 
1299 	case LINUX_F_SETFD:
1300 		return (kern_fcntl(td, args->fd, F_SETFD, args->arg));
1301 
1302 	case LINUX_F_GETFL:
1303 		error = kern_fcntl(td, args->fd, F_GETFL, 0);
1304 		result = td->td_retval[0];
1305 		td->td_retval[0] = 0;
1306 		if (result & O_RDONLY)
1307 			td->td_retval[0] |= LINUX_O_RDONLY;
1308 		if (result & O_WRONLY)
1309 			td->td_retval[0] |= LINUX_O_WRONLY;
1310 		if (result & O_RDWR)
1311 			td->td_retval[0] |= LINUX_O_RDWR;
1312 		if (result & O_NDELAY)
1313 			td->td_retval[0] |= LINUX_O_NONBLOCK;
1314 		if (result & O_APPEND)
1315 			td->td_retval[0] |= LINUX_O_APPEND;
1316 		if (result & O_FSYNC)
1317 			td->td_retval[0] |= LINUX_O_SYNC;
1318 		if (result & O_ASYNC)
1319 			td->td_retval[0] |= LINUX_FASYNC;
1320 #ifdef LINUX_O_NOFOLLOW
1321 		if (result & O_NOFOLLOW)
1322 			td->td_retval[0] |= LINUX_O_NOFOLLOW;
1323 #endif
1324 #ifdef LINUX_O_DIRECT
1325 		if (result & O_DIRECT)
1326 			td->td_retval[0] |= LINUX_O_DIRECT;
1327 #endif
1328 		return (error);
1329 
1330 	case LINUX_F_SETFL:
1331 		arg = 0;
1332 		if (args->arg & LINUX_O_NDELAY)
1333 			arg |= O_NONBLOCK;
1334 		if (args->arg & LINUX_O_APPEND)
1335 			arg |= O_APPEND;
1336 		if (args->arg & LINUX_O_SYNC)
1337 			arg |= O_FSYNC;
1338 		if (args->arg & LINUX_FASYNC)
1339 			arg |= O_ASYNC;
1340 #ifdef LINUX_O_NOFOLLOW
1341 		if (args->arg & LINUX_O_NOFOLLOW)
1342 			arg |= O_NOFOLLOW;
1343 #endif
1344 #ifdef LINUX_O_DIRECT
1345 		if (args->arg & LINUX_O_DIRECT)
1346 			arg |= O_DIRECT;
1347 #endif
1348 		return (kern_fcntl(td, args->fd, F_SETFL, arg));
1349 
1350 	case LINUX_F_GETLK:
1351 		error = copyin((void *)args->arg, &linux_flock,
1352 		    sizeof(linux_flock));
1353 		if (error)
1354 			return (error);
1355 		linux_to_bsd_flock(&linux_flock, &bsd_flock);
1356 		error = kern_fcntl(td, args->fd, F_GETLK, (intptr_t)&bsd_flock);
1357 		if (error)
1358 			return (error);
1359 		bsd_to_linux_flock(&bsd_flock, &linux_flock);
1360 		return (copyout(&linux_flock, (void *)args->arg,
1361 		    sizeof(linux_flock)));
1362 
1363 	case LINUX_F_SETLK:
1364 		error = copyin((void *)args->arg, &linux_flock,
1365 		    sizeof(linux_flock));
1366 		if (error)
1367 			return (error);
1368 		linux_to_bsd_flock(&linux_flock, &bsd_flock);
1369 		return (kern_fcntl(td, args->fd, F_SETLK,
1370 		    (intptr_t)&bsd_flock));
1371 
1372 	case LINUX_F_SETLKW:
1373 		error = copyin((void *)args->arg, &linux_flock,
1374 		    sizeof(linux_flock));
1375 		if (error)
1376 			return (error);
1377 		linux_to_bsd_flock(&linux_flock, &bsd_flock);
1378 		return (kern_fcntl(td, args->fd, F_SETLKW,
1379 		     (intptr_t)&bsd_flock));
1380 
1381 	case LINUX_F_GETOWN:
1382 		return (kern_fcntl(td, args->fd, F_GETOWN, 0));
1383 
1384 	case LINUX_F_SETOWN:
1385 		/*
1386 		 * XXX some Linux applications depend on F_SETOWN having no
1387 		 * significant effect for pipes (SIGIO is not delivered for
1388 		 * pipes under Linux-2.2.35 at least).
1389 		 */
1390 		error = fget(td, args->fd, CAP_FCNTL, &fp);
1391 		if (error)
1392 			return (error);
1393 		if (fp->f_type == DTYPE_PIPE) {
1394 			fdrop(fp, td);
1395 			return (EINVAL);
1396 		}
1397 		fdrop(fp, td);
1398 
1399 		return (kern_fcntl(td, args->fd, F_SETOWN, args->arg));
1400 	}
1401 
1402 	return (EINVAL);
1403 }
1404 
1405 int
1406 linux_fcntl(struct thread *td, struct linux_fcntl_args *args)
1407 {
1408 	struct linux_fcntl64_args args64;
1409 
1410 #ifdef DEBUG
1411 	if (ldebug(fcntl))
1412 		printf(ARGS(fcntl, "%d, %08x, *"), args->fd, args->cmd);
1413 #endif
1414 
1415 	args64.fd = args->fd;
1416 	args64.cmd = args->cmd;
1417 	args64.arg = args->arg;
1418 	return (fcntl_common(td, &args64));
1419 }
1420 
1421 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1422 int
1423 linux_fcntl64(struct thread *td, struct linux_fcntl64_args *args)
1424 {
1425 	struct l_flock64 linux_flock;
1426 	struct flock bsd_flock;
1427 	int error;
1428 
1429 #ifdef DEBUG
1430 	if (ldebug(fcntl64))
1431 		printf(ARGS(fcntl64, "%d, %08x, *"), args->fd, args->cmd);
1432 #endif
1433 
1434 	switch (args->cmd) {
1435 	case LINUX_F_GETLK64:
1436 		error = copyin((void *)args->arg, &linux_flock,
1437 		    sizeof(linux_flock));
1438 		if (error)
1439 			return (error);
1440 		linux_to_bsd_flock64(&linux_flock, &bsd_flock);
1441 		error = kern_fcntl(td, args->fd, F_GETLK, (intptr_t)&bsd_flock);
1442 		if (error)
1443 			return (error);
1444 		bsd_to_linux_flock64(&bsd_flock, &linux_flock);
1445 		return (copyout(&linux_flock, (void *)args->arg,
1446 			    sizeof(linux_flock)));
1447 
1448 	case LINUX_F_SETLK64:
1449 		error = copyin((void *)args->arg, &linux_flock,
1450 		    sizeof(linux_flock));
1451 		if (error)
1452 			return (error);
1453 		linux_to_bsd_flock64(&linux_flock, &bsd_flock);
1454 		return (kern_fcntl(td, args->fd, F_SETLK,
1455 		    (intptr_t)&bsd_flock));
1456 
1457 	case LINUX_F_SETLKW64:
1458 		error = copyin((void *)args->arg, &linux_flock,
1459 		    sizeof(linux_flock));
1460 		if (error)
1461 			return (error);
1462 		linux_to_bsd_flock64(&linux_flock, &bsd_flock);
1463 		return (kern_fcntl(td, args->fd, F_SETLKW,
1464 		    (intptr_t)&bsd_flock));
1465 	}
1466 
1467 	return (fcntl_common(td, args));
1468 }
1469 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1470 
1471 int
1472 linux_chown(struct thread *td, struct linux_chown_args *args)
1473 {
1474 	char *path;
1475 	int error;
1476 
1477 	LCONVPATHEXIST(td, args->path, &path);
1478 
1479 #ifdef DEBUG
1480 	if (ldebug(chown))
1481 		printf(ARGS(chown, "%s, %d, %d"), path, args->uid, args->gid);
1482 #endif
1483 	error = kern_chown(td, path, UIO_SYSSPACE, args->uid, args->gid);
1484 	LFREEPATH(path);
1485 	return (error);
1486 }
1487 
1488 int
1489 linux_fchownat(struct thread *td, struct linux_fchownat_args *args)
1490 {
1491 	char *path;
1492 	int error, dfd, flag;
1493 
1494 	if (args->flag & ~LINUX_AT_SYMLINK_NOFOLLOW)
1495 		return (EINVAL);
1496 
1497 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD :  args->dfd;
1498 	LCONVPATHEXIST_AT(td, args->filename, &path, dfd);
1499 
1500 #ifdef DEBUG
1501 	if (ldebug(fchownat))
1502 		printf(ARGS(fchownat, "%s, %d, %d"), path, args->uid, args->gid);
1503 #endif
1504 
1505 	flag = (args->flag & LINUX_AT_SYMLINK_NOFOLLOW) == 0 ? 0 :
1506 	    AT_SYMLINK_NOFOLLOW;
1507 	error = kern_fchownat(td, dfd, path, UIO_SYSSPACE, args->uid, args->gid,
1508 	    flag);
1509 	LFREEPATH(path);
1510 	return (error);
1511 }
1512 
1513 int
1514 linux_lchown(struct thread *td, struct linux_lchown_args *args)
1515 {
1516 	char *path;
1517 	int error;
1518 
1519 	LCONVPATHEXIST(td, args->path, &path);
1520 
1521 #ifdef DEBUG
1522 	if (ldebug(lchown))
1523 		printf(ARGS(lchown, "%s, %d, %d"), path, args->uid, args->gid);
1524 #endif
1525 	error = kern_lchown(td, path, UIO_SYSSPACE, args->uid, args->gid);
1526 	LFREEPATH(path);
1527 	return (error);
1528 }
1529 
1530 static int
1531 convert_fadvice(int advice)
1532 {
1533 	switch (advice) {
1534 	case LINUX_POSIX_FADV_NORMAL:
1535 		return (POSIX_FADV_NORMAL);
1536 	case LINUX_POSIX_FADV_RANDOM:
1537 		return (POSIX_FADV_RANDOM);
1538 	case LINUX_POSIX_FADV_SEQUENTIAL:
1539 		return (POSIX_FADV_SEQUENTIAL);
1540 	case LINUX_POSIX_FADV_WILLNEED:
1541 		return (POSIX_FADV_WILLNEED);
1542 	case LINUX_POSIX_FADV_DONTNEED:
1543 		return (POSIX_FADV_DONTNEED);
1544 	case LINUX_POSIX_FADV_NOREUSE:
1545 		return (POSIX_FADV_NOREUSE);
1546 	default:
1547 		return (-1);
1548 	}
1549 }
1550 
1551 int
1552 linux_fadvise64(struct thread *td, struct linux_fadvise64_args *args)
1553 {
1554 	int advice;
1555 
1556 	advice = convert_fadvice(args->advice);
1557 	if (advice == -1)
1558 		return (EINVAL);
1559 	return (kern_posix_fadvise(td, args->fd, args->offset, args->len,
1560 	    advice));
1561 }
1562 
1563 int
1564 linux_fadvise64_64(struct thread *td, struct linux_fadvise64_64_args *args)
1565 {
1566 	int advice;
1567 
1568 	advice = convert_fadvice(args->advice);
1569 	if (advice == -1)
1570 		return (EINVAL);
1571 	return (kern_posix_fadvise(td, args->fd, args->offset, args->len,
1572 	    advice));
1573 }
1574 
1575 int
1576 linux_pipe(struct thread *td, struct linux_pipe_args *args)
1577 {
1578 	int fildes[2];
1579 	int error;
1580 
1581 #ifdef DEBUG
1582 	if (ldebug(pipe))
1583 		printf(ARGS(pipe, "*"));
1584 #endif
1585 
1586 	error = do_pipe(td, fildes, 0);
1587 	if (error)
1588 		return (error);
1589 
1590 	/* XXX: Close descriptors on error. */
1591 	return (copyout(fildes, args->pipefds, sizeof(fildes)));
1592 }
1593 
1594 int
1595 linux_pipe2(struct thread *td, struct linux_pipe2_args *args)
1596 {
1597 	int fildes[2];
1598 	int error, flags;
1599 
1600 #ifdef DEBUG
1601 	if (ldebug(pipe2))
1602 		printf(ARGS(pipe2, "*, %d"), args->flags);
1603 #endif
1604 
1605 	if ((args->flags & ~(LINUX_O_NONBLOCK | LINUX_O_CLOEXEC)) != 0)
1606 		return (EINVAL);
1607 
1608 	flags = 0;
1609 	if ((args->flags & LINUX_O_NONBLOCK) != 0)
1610 		flags |= O_NONBLOCK;
1611 	if ((args->flags & LINUX_O_CLOEXEC) != 0)
1612 		flags |= O_CLOEXEC;
1613 	error = do_pipe(td, fildes, flags);
1614 	if (error)
1615 		return (error);
1616 
1617 	/* XXX: Close descriptors on error. */
1618 	return (copyout(fildes, args->pipefds, sizeof(fildes)));
1619 }
1620