xref: /freebsd/sys/compat/linux/linux_file.c (revision 8ef24a0d4b28fe230e20637f56869cc4148cd2ca)
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/capsicum.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 #ifdef COMPAT_LINUX32
59 #include <machine/../linux32/linux.h>
60 #include <machine/../linux32/linux32_proto.h>
61 #else
62 #include <machine/../linux/linux.h>
63 #include <machine/../linux/linux_proto.h>
64 #endif
65 #include <compat/linux/linux_misc.h>
66 #include <compat/linux/linux_util.h>
67 #include <compat/linux/linux_file.h>
68 
69 int
70 linux_creat(struct thread *td, struct linux_creat_args *args)
71 {
72     char *path;
73     int error;
74 
75     LCONVPATHEXIST(td, args->path, &path);
76 
77 #ifdef DEBUG
78 	if (ldebug(creat))
79 		printf(ARGS(creat, "%s, %d"), path, args->mode);
80 #endif
81     error = kern_openat(td, AT_FDCWD, path, UIO_SYSSPACE,
82 	O_WRONLY | O_CREAT | O_TRUNC, args->mode);
83     LFREEPATH(path);
84     return (error);
85 }
86 
87 
88 static int
89 linux_common_open(struct thread *td, int dirfd, char *path, int l_flags, int mode)
90 {
91     cap_rights_t rights;
92     struct proc *p = td->td_proc;
93     struct file *fp;
94     int fd;
95     int bsd_flags, error;
96 
97     bsd_flags = 0;
98     switch (l_flags & LINUX_O_ACCMODE) {
99     case LINUX_O_WRONLY:
100 	bsd_flags |= O_WRONLY;
101 	break;
102     case LINUX_O_RDWR:
103 	bsd_flags |= O_RDWR;
104 	break;
105     default:
106 	bsd_flags |= O_RDONLY;
107     }
108     if (l_flags & LINUX_O_NDELAY)
109 	bsd_flags |= O_NONBLOCK;
110     if (l_flags & LINUX_O_APPEND)
111 	bsd_flags |= O_APPEND;
112     if (l_flags & LINUX_O_SYNC)
113 	bsd_flags |= O_FSYNC;
114     if (l_flags & LINUX_O_NONBLOCK)
115 	bsd_flags |= O_NONBLOCK;
116     if (l_flags & LINUX_FASYNC)
117 	bsd_flags |= O_ASYNC;
118     if (l_flags & LINUX_O_CREAT)
119 	bsd_flags |= O_CREAT;
120     if (l_flags & LINUX_O_TRUNC)
121 	bsd_flags |= O_TRUNC;
122     if (l_flags & LINUX_O_EXCL)
123 	bsd_flags |= O_EXCL;
124     if (l_flags & LINUX_O_NOCTTY)
125 	bsd_flags |= O_NOCTTY;
126     if (l_flags & LINUX_O_DIRECT)
127 	bsd_flags |= O_DIRECT;
128     if (l_flags & LINUX_O_NOFOLLOW)
129 	bsd_flags |= O_NOFOLLOW;
130     if (l_flags & LINUX_O_DIRECTORY)
131 	bsd_flags |= O_DIRECTORY;
132     /* XXX LINUX_O_NOATIME: unable to be easily implemented. */
133 
134     error = kern_openat(td, dirfd, path, UIO_SYSSPACE, bsd_flags, mode);
135     if (error != 0)
136 	    goto done;
137 
138     if (bsd_flags & O_NOCTTY)
139 	    goto done;
140 
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     fd = td->td_retval[0];
147     if (fget(td, fd, cap_rights_init(&rights, CAP_IOCTL), &fp) == 0) {
148 	    if (fp->f_type != DTYPE_VNODE) {
149 		    fdrop(fp, td);
150 		    goto done;
151 	    }
152 	    sx_slock(&proctree_lock);
153 	    PROC_LOCK(p);
154 	    if (SESS_LEADER(p) && !(p->p_flag & P_CONTROLT)) {
155 		    PROC_UNLOCK(p);
156 		    sx_sunlock(&proctree_lock);
157 		    /* XXXPJD: Verify if TIOCSCTTY is allowed. */
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 
167 done:
168 #ifdef DEBUG
169     if (ldebug(open))
170 	    printf(LMSG("open returns error %d"), error);
171 #endif
172     LFREEPATH(path);
173     return (error);
174 }
175 
176 int
177 linux_openat(struct thread *td, struct linux_openat_args *args)
178 {
179 	char *path;
180 	int dfd;
181 
182 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
183 	if (args->flags & LINUX_O_CREAT)
184 		LCONVPATH_AT(td, args->filename, &path, 1, dfd);
185 	else
186 		LCONVPATH_AT(td, args->filename, &path, 0, dfd);
187 #ifdef DEBUG
188 	if (ldebug(openat))
189 		printf(ARGS(openat, "%i, %s, 0x%x, 0x%x"), args->dfd,
190 		    path, args->flags, args->mode);
191 #endif
192 	return (linux_common_open(td, dfd, path, args->flags, args->mode));
193 }
194 
195 int
196 linux_open(struct thread *td, struct linux_open_args *args)
197 {
198     char *path;
199 
200     if (args->flags & LINUX_O_CREAT)
201 	LCONVPATHCREAT(td, args->path, &path);
202     else
203 	LCONVPATHEXIST(td, args->path, &path);
204 
205 #ifdef DEBUG
206 	if (ldebug(open))
207 		printf(ARGS(open, "%s, 0x%x, 0x%x"),
208 		    path, args->flags, args->mode);
209 #endif
210 
211 	return (linux_common_open(td, AT_FDCWD, path, args->flags, args->mode));
212 }
213 
214 int
215 linux_lseek(struct thread *td, struct linux_lseek_args *args)
216 {
217 
218     struct lseek_args /* {
219 	int fd;
220 	int pad;
221 	off_t offset;
222 	int whence;
223     } */ tmp_args;
224     int error;
225 
226 #ifdef DEBUG
227 	if (ldebug(lseek))
228 		printf(ARGS(lseek, "%d, %ld, %d"),
229 		    args->fdes, (long)args->off, args->whence);
230 #endif
231     tmp_args.fd = args->fdes;
232     tmp_args.offset = (off_t)args->off;
233     tmp_args.whence = args->whence;
234     error = sys_lseek(td, &tmp_args);
235     return error;
236 }
237 
238 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
239 int
240 linux_llseek(struct thread *td, struct linux_llseek_args *args)
241 {
242 	struct lseek_args bsd_args;
243 	int error;
244 	off_t off;
245 
246 #ifdef DEBUG
247 	if (ldebug(llseek))
248 		printf(ARGS(llseek, "%d, %d:%d, %d"),
249 		    args->fd, args->ohigh, args->olow, args->whence);
250 #endif
251 	off = (args->olow) | (((off_t) args->ohigh) << 32);
252 
253 	bsd_args.fd = args->fd;
254 	bsd_args.offset = off;
255 	bsd_args.whence = args->whence;
256 
257 	if ((error = sys_lseek(td, &bsd_args)))
258 		return error;
259 
260 	if ((error = copyout(td->td_retval, args->res, sizeof (off_t))))
261 		return error;
262 
263 	td->td_retval[0] = 0;
264 	return 0;
265 }
266 
267 int
268 linux_readdir(struct thread *td, struct linux_readdir_args *args)
269 {
270 	struct linux_getdents_args lda;
271 
272 	lda.fd = args->fd;
273 	lda.dent = args->dent;
274 	lda.count = 1;
275 	return linux_getdents(td, &lda);
276 }
277 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
278 
279 /*
280  * Note that linux_getdents(2) and linux_getdents64(2) have the same
281  * arguments. They only differ in the definition of struct dirent they
282  * operate on. We use this to common the code, with the exception of
283  * accessing struct dirent. Note that linux_readdir(2) is implemented
284  * by means of linux_getdents(2). In this case we never operate on
285  * struct dirent64 and thus don't need to handle it...
286  */
287 
288 struct l_dirent {
289 	l_ulong		d_ino;
290 	l_off_t		d_off;
291 	l_ushort	d_reclen;
292 	char		d_name[LINUX_NAME_MAX + 1];
293 };
294 
295 struct l_dirent64 {
296 	uint64_t	d_ino;
297 	int64_t		d_off;
298 	l_ushort	d_reclen;
299 	u_char		d_type;
300 	char		d_name[LINUX_NAME_MAX + 1];
301 };
302 
303 /*
304  * Linux uses the last byte in the dirent buffer to store d_type,
305  * at least glibc-2.7 requires it. That is why l_dirent is padded with 2 bytes.
306  */
307 #define LINUX_RECLEN(namlen)						\
308     roundup(offsetof(struct l_dirent, d_name) + (namlen) + 2, sizeof(l_ulong))
309 
310 #define LINUX_RECLEN64(namlen)						\
311     roundup(offsetof(struct l_dirent64, d_name) + (namlen) + 1,		\
312     sizeof(uint64_t))
313 
314 #define LINUX_MAXRECLEN		max(LINUX_RECLEN(LINUX_NAME_MAX),	\
315 				    LINUX_RECLEN64(LINUX_NAME_MAX))
316 #define	LINUX_DIRBLKSIZ		512
317 
318 static int
319 getdents_common(struct thread *td, struct linux_getdents64_args *args,
320     int is64bit)
321 {
322 	struct dirent *bdp;
323 	struct vnode *vp;
324 	caddr_t inp, buf;		/* BSD-format */
325 	int len, reclen;		/* BSD-format */
326 	caddr_t outp;			/* Linux-format */
327 	int resid, linuxreclen=0;	/* Linux-format */
328 	caddr_t lbuf;			/* Linux-format */
329 	cap_rights_t rights;
330 	struct file *fp;
331 	struct uio auio;
332 	struct iovec aiov;
333 	off_t off;
334 	struct l_dirent *linux_dirent;
335 	struct l_dirent64 *linux_dirent64;
336 	int buflen, error, eofflag, nbytes, justone;
337 	u_long *cookies = NULL, *cookiep;
338 	int ncookies;
339 
340 	nbytes = args->count;
341 	if (nbytes == 1) {
342 		/* readdir(2) case. Always struct dirent. */
343 		if (is64bit)
344 			return (EINVAL);
345 		nbytes = sizeof(*linux_dirent);
346 		justone = 1;
347 	} else
348 		justone = 0;
349 
350 	error = getvnode(td, args->fd, cap_rights_init(&rights, CAP_READ), &fp);
351 	if (error != 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_LINUX, M_WAITOK);
371 	lbuf = malloc(LINUX_MAXRECLEN, M_LINUX, 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_LINUX);
523 	free(lbuf, M_LINUX);
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_accessat(td, AT_FDCWD, path, UIO_SYSSPACE, 0,
572 	    args->amode);
573 	LFREEPATH(path);
574 
575 	return (error);
576 }
577 
578 int
579 linux_faccessat(struct thread *td, struct linux_faccessat_args *args)
580 {
581 	char *path;
582 	int error, dfd;
583 
584 	/* linux convention */
585 	if (args->amode & ~(F_OK | X_OK | W_OK | R_OK))
586 		return (EINVAL);
587 
588 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
589 	LCONVPATHEXIST_AT(td, args->filename, &path, dfd);
590 
591 #ifdef DEBUG
592 	if (ldebug(access))
593 		printf(ARGS(access, "%s, %d"), path, args->amode);
594 #endif
595 
596 	error = kern_accessat(td, dfd, path, UIO_SYSSPACE, 0, args->amode);
597 	LFREEPATH(path);
598 
599 	return (error);
600 }
601 
602 int
603 linux_unlink(struct thread *td, struct linux_unlink_args *args)
604 {
605 	char *path;
606 	int error;
607 	struct stat st;
608 
609 	LCONVPATHEXIST(td, args->path, &path);
610 
611 #ifdef DEBUG
612 	if (ldebug(unlink))
613 		printf(ARGS(unlink, "%s"), path);
614 #endif
615 
616 	error = kern_unlinkat(td, AT_FDCWD, path, UIO_SYSSPACE, 0);
617 	if (error == EPERM) {
618 		/* Introduce POSIX noncompliant behaviour of Linux */
619 		if (kern_statat(td, 0, AT_FDCWD, path, UIO_SYSSPACE, &st,
620 		    NULL) == 0) {
621 			if (S_ISDIR(st.st_mode))
622 				error = EISDIR;
623 		}
624 	}
625 	LFREEPATH(path);
626 	return (error);
627 }
628 
629 int
630 linux_unlinkat(struct thread *td, struct linux_unlinkat_args *args)
631 {
632 	char *path;
633 	int error, dfd;
634 	struct stat st;
635 
636 	if (args->flag & ~LINUX_AT_REMOVEDIR)
637 		return (EINVAL);
638 
639 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
640 	LCONVPATHEXIST_AT(td, args->pathname, &path, dfd);
641 
642 #ifdef DEBUG
643 	if (ldebug(unlinkat))
644 		printf(ARGS(unlinkat, "%s"), path);
645 #endif
646 
647 	if (args->flag & LINUX_AT_REMOVEDIR)
648 		error = kern_rmdirat(td, dfd, path, UIO_SYSSPACE);
649 	else
650 		error = kern_unlinkat(td, dfd, path, UIO_SYSSPACE, 0);
651 	if (error == EPERM && !(args->flag & LINUX_AT_REMOVEDIR)) {
652 		/* Introduce POSIX noncompliant behaviour of Linux */
653 		if (kern_statat(td, AT_SYMLINK_NOFOLLOW, dfd, path,
654 		    UIO_SYSSPACE, &st, NULL) == 0 && S_ISDIR(st.st_mode))
655 			error = EISDIR;
656 	}
657 	LFREEPATH(path);
658 	return (error);
659 }
660 int
661 linux_chdir(struct thread *td, struct linux_chdir_args *args)
662 {
663 	char *path;
664 	int error;
665 
666 	LCONVPATHEXIST(td, args->path, &path);
667 
668 #ifdef DEBUG
669 	if (ldebug(chdir))
670 		printf(ARGS(chdir, "%s"), path);
671 #endif
672 	error = kern_chdir(td, path, UIO_SYSSPACE);
673 	LFREEPATH(path);
674 	return (error);
675 }
676 
677 int
678 linux_chmod(struct thread *td, struct linux_chmod_args *args)
679 {
680 	char *path;
681 	int error;
682 
683 	LCONVPATHEXIST(td, args->path, &path);
684 
685 #ifdef DEBUG
686 	if (ldebug(chmod))
687 		printf(ARGS(chmod, "%s, %d"), path, args->mode);
688 #endif
689 	error = kern_fchmodat(td, AT_FDCWD, path, UIO_SYSSPACE,
690 	    args->mode, 0);
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_mkdirat(td, AT_FDCWD, 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_rmdirat(td, AT_FDCWD, 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_renameat(td, AT_FDCWD, from, AT_FDCWD, 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_symlinkat(td, path, AT_FDCWD, 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_readlinkat(td, AT_FDCWD, name, UIO_SYSSPACE,
880 	    args->buf, UIO_USERSPACE, 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 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
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 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
943 
944 int
945 linux_ftruncate(struct thread *td, struct linux_ftruncate_args *args)
946 {
947 	struct ftruncate_args /* {
948 		int fd;
949 		int pad;
950 		off_t length;
951 		} */ nuap;
952 
953 	nuap.fd = args->fd;
954 	nuap.length = args->length;
955 	return (sys_ftruncate(td, &nuap));
956 }
957 
958 int
959 linux_link(struct thread *td, struct linux_link_args *args)
960 {
961 	char *path, *to;
962 	int error;
963 
964 	LCONVPATHEXIST(td, args->path, &path);
965 	/* Expand LCONVPATHCREATE so that `path' can be freed on errors */
966 	error = linux_emul_convpath(td, args->to, UIO_USERSPACE, &to, 1, AT_FDCWD);
967 	if (to == NULL) {
968 		LFREEPATH(path);
969 		return (error);
970 	}
971 
972 #ifdef DEBUG
973 	if (ldebug(link))
974 		printf(ARGS(link, "%s, %s"), path, to);
975 #endif
976 	error = kern_linkat(td, AT_FDCWD, AT_FDCWD, path, to, UIO_SYSSPACE,
977 	    FOLLOW);
978 	LFREEPATH(path);
979 	LFREEPATH(to);
980 	return (error);
981 }
982 
983 int
984 linux_linkat(struct thread *td, struct linux_linkat_args *args)
985 {
986 	char *path, *to;
987 	int error, olddfd, newdfd, follow;
988 
989 	if (args->flag & ~LINUX_AT_SYMLINK_FOLLOW)
990 		return (EINVAL);
991 
992 	olddfd = (args->olddfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->olddfd;
993 	newdfd = (args->newdfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->newdfd;
994 	LCONVPATHEXIST_AT(td, args->oldname, &path, olddfd);
995 	/* Expand LCONVPATHCREATE so that `path' can be freed on errors */
996 	error = linux_emul_convpath(td, args->newname, UIO_USERSPACE, &to, 1, newdfd);
997 	if (to == NULL) {
998 		LFREEPATH(path);
999 		return (error);
1000 	}
1001 
1002 #ifdef DEBUG
1003 	if (ldebug(linkat))
1004 		printf(ARGS(linkat, "%i, %s, %i, %s, %i"), args->olddfd, path,
1005 			args->newdfd, to, args->flag);
1006 #endif
1007 
1008 	follow = (args->flag & LINUX_AT_SYMLINK_FOLLOW) == 0 ? NOFOLLOW :
1009 	    FOLLOW;
1010 	error = kern_linkat(td, olddfd, newdfd, path, to, UIO_SYSSPACE, follow);
1011 	LFREEPATH(path);
1012 	LFREEPATH(to);
1013 	return (error);
1014 }
1015 
1016 int
1017 linux_fdatasync(td, uap)
1018 	struct thread *td;
1019 	struct linux_fdatasync_args *uap;
1020 {
1021 	struct fsync_args bsd;
1022 
1023 	bsd.fd = uap->fd;
1024 	return sys_fsync(td, &bsd);
1025 }
1026 
1027 int
1028 linux_pread(td, uap)
1029 	struct thread *td;
1030 	struct linux_pread_args *uap;
1031 {
1032 	struct pread_args bsd;
1033 	cap_rights_t rights;
1034 	struct vnode *vp;
1035 	int error;
1036 
1037 	bsd.fd = uap->fd;
1038 	bsd.buf = uap->buf;
1039 	bsd.nbyte = uap->nbyte;
1040 	bsd.offset = uap->offset;
1041 
1042 	error = sys_pread(td, &bsd);
1043 
1044 	if (error == 0) {
1045 		/* This seems to violate POSIX but linux does it */
1046 		error = fgetvp(td, uap->fd,
1047 		    cap_rights_init(&rights, CAP_PREAD), &vp);
1048 		if (error != 0)
1049 			return (error);
1050 		if (vp->v_type == VDIR) {
1051 			vrele(vp);
1052 			return (EISDIR);
1053 		}
1054 		vrele(vp);
1055 	}
1056 
1057 	return (error);
1058 }
1059 
1060 int
1061 linux_pwrite(td, uap)
1062 	struct thread *td;
1063 	struct linux_pwrite_args *uap;
1064 {
1065 	struct pwrite_args bsd;
1066 
1067 	bsd.fd = uap->fd;
1068 	bsd.buf = uap->buf;
1069 	bsd.nbyte = uap->nbyte;
1070 	bsd.offset = uap->offset;
1071 	return sys_pwrite(td, &bsd);
1072 }
1073 
1074 int
1075 linux_mount(struct thread *td, struct linux_mount_args *args)
1076 {
1077 	char fstypename[MFSNAMELEN];
1078 	char mntonname[MNAMELEN], mntfromname[MNAMELEN];
1079 	int error;
1080 	int fsflags;
1081 
1082 	error = copyinstr(args->filesystemtype, fstypename, MFSNAMELEN - 1,
1083 	    NULL);
1084 	if (error)
1085 		return (error);
1086 	error = copyinstr(args->specialfile, mntfromname, MNAMELEN - 1, NULL);
1087 	if (error)
1088 		return (error);
1089 	error = copyinstr(args->dir, mntonname, MNAMELEN - 1, NULL);
1090 	if (error)
1091 		return (error);
1092 
1093 #ifdef DEBUG
1094 	if (ldebug(mount))
1095 		printf(ARGS(mount, "%s, %s, %s"),
1096 		    fstypename, mntfromname, mntonname);
1097 #endif
1098 
1099 	if (strcmp(fstypename, "ext2") == 0) {
1100 		strcpy(fstypename, "ext2fs");
1101 	} else if (strcmp(fstypename, "proc") == 0) {
1102 		strcpy(fstypename, "linprocfs");
1103 	} else if (strcmp(fstypename, "vfat") == 0) {
1104 		strcpy(fstypename, "msdosfs");
1105 	}
1106 
1107 	fsflags = 0;
1108 
1109 	if ((args->rwflag & 0xffff0000) == 0xc0ed0000) {
1110 		/*
1111 		 * Linux SYNC flag is not included; the closest equivalent
1112 		 * FreeBSD has is !ASYNC, which is our default.
1113 		 */
1114 		if (args->rwflag & LINUX_MS_RDONLY)
1115 			fsflags |= MNT_RDONLY;
1116 		if (args->rwflag & LINUX_MS_NOSUID)
1117 			fsflags |= MNT_NOSUID;
1118 		if (args->rwflag & LINUX_MS_NOEXEC)
1119 			fsflags |= MNT_NOEXEC;
1120 		if (args->rwflag & LINUX_MS_REMOUNT)
1121 			fsflags |= MNT_UPDATE;
1122 	}
1123 
1124 	error = kernel_vmount(fsflags,
1125 	    "fstype", fstypename,
1126 	    "fspath", mntonname,
1127 	    "from", mntfromname,
1128 	    NULL);
1129 	return (error);
1130 }
1131 
1132 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1133 int
1134 linux_oldumount(struct thread *td, struct linux_oldumount_args *args)
1135 {
1136 	struct linux_umount_args args2;
1137 
1138 	args2.path = args->path;
1139 	args2.flags = 0;
1140 	return (linux_umount(td, &args2));
1141 }
1142 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1143 
1144 int
1145 linux_umount(struct thread *td, struct linux_umount_args *args)
1146 {
1147 	struct unmount_args bsd;
1148 
1149 	bsd.path = args->path;
1150 	bsd.flags = args->flags;	/* XXX correct? */
1151 	return (sys_unmount(td, &bsd));
1152 }
1153 
1154 /*
1155  * fcntl family of syscalls
1156  */
1157 
1158 struct l_flock {
1159 	l_short		l_type;
1160 	l_short		l_whence;
1161 	l_off_t		l_start;
1162 	l_off_t		l_len;
1163 	l_pid_t		l_pid;
1164 }
1165 #if defined(__amd64__) && defined(COMPAT_LINUX32)
1166 __packed
1167 #endif
1168 ;
1169 
1170 static void
1171 linux_to_bsd_flock(struct l_flock *linux_flock, struct flock *bsd_flock)
1172 {
1173 	switch (linux_flock->l_type) {
1174 	case LINUX_F_RDLCK:
1175 		bsd_flock->l_type = F_RDLCK;
1176 		break;
1177 	case LINUX_F_WRLCK:
1178 		bsd_flock->l_type = F_WRLCK;
1179 		break;
1180 	case LINUX_F_UNLCK:
1181 		bsd_flock->l_type = F_UNLCK;
1182 		break;
1183 	default:
1184 		bsd_flock->l_type = -1;
1185 		break;
1186 	}
1187 	bsd_flock->l_whence = linux_flock->l_whence;
1188 	bsd_flock->l_start = (off_t)linux_flock->l_start;
1189 	bsd_flock->l_len = (off_t)linux_flock->l_len;
1190 	bsd_flock->l_pid = (pid_t)linux_flock->l_pid;
1191 	bsd_flock->l_sysid = 0;
1192 }
1193 
1194 static void
1195 bsd_to_linux_flock(struct flock *bsd_flock, struct l_flock *linux_flock)
1196 {
1197 	switch (bsd_flock->l_type) {
1198 	case F_RDLCK:
1199 		linux_flock->l_type = LINUX_F_RDLCK;
1200 		break;
1201 	case F_WRLCK:
1202 		linux_flock->l_type = LINUX_F_WRLCK;
1203 		break;
1204 	case F_UNLCK:
1205 		linux_flock->l_type = LINUX_F_UNLCK;
1206 		break;
1207 	}
1208 	linux_flock->l_whence = bsd_flock->l_whence;
1209 	linux_flock->l_start = (l_off_t)bsd_flock->l_start;
1210 	linux_flock->l_len = (l_off_t)bsd_flock->l_len;
1211 	linux_flock->l_pid = (l_pid_t)bsd_flock->l_pid;
1212 }
1213 
1214 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1215 struct l_flock64 {
1216 	l_short		l_type;
1217 	l_short		l_whence;
1218 	l_loff_t	l_start;
1219 	l_loff_t	l_len;
1220 	l_pid_t		l_pid;
1221 }
1222 #if defined(__amd64__) && defined(COMPAT_LINUX32)
1223 __packed
1224 #endif
1225 ;
1226 
1227 static void
1228 linux_to_bsd_flock64(struct l_flock64 *linux_flock, struct flock *bsd_flock)
1229 {
1230 	switch (linux_flock->l_type) {
1231 	case LINUX_F_RDLCK:
1232 		bsd_flock->l_type = F_RDLCK;
1233 		break;
1234 	case LINUX_F_WRLCK:
1235 		bsd_flock->l_type = F_WRLCK;
1236 		break;
1237 	case LINUX_F_UNLCK:
1238 		bsd_flock->l_type = F_UNLCK;
1239 		break;
1240 	default:
1241 		bsd_flock->l_type = -1;
1242 		break;
1243 	}
1244 	bsd_flock->l_whence = linux_flock->l_whence;
1245 	bsd_flock->l_start = (off_t)linux_flock->l_start;
1246 	bsd_flock->l_len = (off_t)linux_flock->l_len;
1247 	bsd_flock->l_pid = (pid_t)linux_flock->l_pid;
1248 	bsd_flock->l_sysid = 0;
1249 }
1250 
1251 static void
1252 bsd_to_linux_flock64(struct flock *bsd_flock, struct l_flock64 *linux_flock)
1253 {
1254 	switch (bsd_flock->l_type) {
1255 	case F_RDLCK:
1256 		linux_flock->l_type = LINUX_F_RDLCK;
1257 		break;
1258 	case F_WRLCK:
1259 		linux_flock->l_type = LINUX_F_WRLCK;
1260 		break;
1261 	case F_UNLCK:
1262 		linux_flock->l_type = LINUX_F_UNLCK;
1263 		break;
1264 	}
1265 	linux_flock->l_whence = bsd_flock->l_whence;
1266 	linux_flock->l_start = (l_loff_t)bsd_flock->l_start;
1267 	linux_flock->l_len = (l_loff_t)bsd_flock->l_len;
1268 	linux_flock->l_pid = (l_pid_t)bsd_flock->l_pid;
1269 }
1270 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1271 
1272 static int
1273 fcntl_common(struct thread *td, struct linux_fcntl_args *args)
1274 {
1275 	struct l_flock linux_flock;
1276 	struct flock bsd_flock;
1277 	cap_rights_t rights;
1278 	struct file *fp;
1279 	long arg;
1280 	int error, result;
1281 
1282 	switch (args->cmd) {
1283 	case LINUX_F_DUPFD:
1284 		return (kern_fcntl(td, args->fd, F_DUPFD, args->arg));
1285 
1286 	case LINUX_F_GETFD:
1287 		return (kern_fcntl(td, args->fd, F_GETFD, 0));
1288 
1289 	case LINUX_F_SETFD:
1290 		return (kern_fcntl(td, args->fd, F_SETFD, args->arg));
1291 
1292 	case LINUX_F_GETFL:
1293 		error = kern_fcntl(td, args->fd, F_GETFL, 0);
1294 		result = td->td_retval[0];
1295 		td->td_retval[0] = 0;
1296 		if (result & O_RDONLY)
1297 			td->td_retval[0] |= LINUX_O_RDONLY;
1298 		if (result & O_WRONLY)
1299 			td->td_retval[0] |= LINUX_O_WRONLY;
1300 		if (result & O_RDWR)
1301 			td->td_retval[0] |= LINUX_O_RDWR;
1302 		if (result & O_NDELAY)
1303 			td->td_retval[0] |= LINUX_O_NONBLOCK;
1304 		if (result & O_APPEND)
1305 			td->td_retval[0] |= LINUX_O_APPEND;
1306 		if (result & O_FSYNC)
1307 			td->td_retval[0] |= LINUX_O_SYNC;
1308 		if (result & O_ASYNC)
1309 			td->td_retval[0] |= LINUX_FASYNC;
1310 #ifdef LINUX_O_NOFOLLOW
1311 		if (result & O_NOFOLLOW)
1312 			td->td_retval[0] |= LINUX_O_NOFOLLOW;
1313 #endif
1314 #ifdef LINUX_O_DIRECT
1315 		if (result & O_DIRECT)
1316 			td->td_retval[0] |= LINUX_O_DIRECT;
1317 #endif
1318 		return (error);
1319 
1320 	case LINUX_F_SETFL:
1321 		arg = 0;
1322 		if (args->arg & LINUX_O_NDELAY)
1323 			arg |= O_NONBLOCK;
1324 		if (args->arg & LINUX_O_APPEND)
1325 			arg |= O_APPEND;
1326 		if (args->arg & LINUX_O_SYNC)
1327 			arg |= O_FSYNC;
1328 		if (args->arg & LINUX_FASYNC)
1329 			arg |= O_ASYNC;
1330 #ifdef LINUX_O_NOFOLLOW
1331 		if (args->arg & LINUX_O_NOFOLLOW)
1332 			arg |= O_NOFOLLOW;
1333 #endif
1334 #ifdef LINUX_O_DIRECT
1335 		if (args->arg & LINUX_O_DIRECT)
1336 			arg |= O_DIRECT;
1337 #endif
1338 		return (kern_fcntl(td, args->fd, F_SETFL, arg));
1339 
1340 	case LINUX_F_GETLK:
1341 		error = copyin((void *)args->arg, &linux_flock,
1342 		    sizeof(linux_flock));
1343 		if (error)
1344 			return (error);
1345 		linux_to_bsd_flock(&linux_flock, &bsd_flock);
1346 		error = kern_fcntl(td, args->fd, F_GETLK, (intptr_t)&bsd_flock);
1347 		if (error)
1348 			return (error);
1349 		bsd_to_linux_flock(&bsd_flock, &linux_flock);
1350 		return (copyout(&linux_flock, (void *)args->arg,
1351 		    sizeof(linux_flock)));
1352 
1353 	case LINUX_F_SETLK:
1354 		error = copyin((void *)args->arg, &linux_flock,
1355 		    sizeof(linux_flock));
1356 		if (error)
1357 			return (error);
1358 		linux_to_bsd_flock(&linux_flock, &bsd_flock);
1359 		return (kern_fcntl(td, args->fd, F_SETLK,
1360 		    (intptr_t)&bsd_flock));
1361 
1362 	case LINUX_F_SETLKW:
1363 		error = copyin((void *)args->arg, &linux_flock,
1364 		    sizeof(linux_flock));
1365 		if (error)
1366 			return (error);
1367 		linux_to_bsd_flock(&linux_flock, &bsd_flock);
1368 		return (kern_fcntl(td, args->fd, F_SETLKW,
1369 		     (intptr_t)&bsd_flock));
1370 
1371 	case LINUX_F_GETOWN:
1372 		return (kern_fcntl(td, args->fd, F_GETOWN, 0));
1373 
1374 	case LINUX_F_SETOWN:
1375 		/*
1376 		 * XXX some Linux applications depend on F_SETOWN having no
1377 		 * significant effect for pipes (SIGIO is not delivered for
1378 		 * pipes under Linux-2.2.35 at least).
1379 		 */
1380 		error = fget(td, args->fd,
1381 		    cap_rights_init(&rights, CAP_FCNTL), &fp);
1382 		if (error)
1383 			return (error);
1384 		if (fp->f_type == DTYPE_PIPE) {
1385 			fdrop(fp, td);
1386 			return (EINVAL);
1387 		}
1388 		fdrop(fp, td);
1389 
1390 		return (kern_fcntl(td, args->fd, F_SETOWN, args->arg));
1391 
1392 	case LINUX_F_DUPFD_CLOEXEC:
1393 		return (kern_fcntl(td, args->fd, F_DUPFD_CLOEXEC, args->arg));
1394 	}
1395 
1396 	return (EINVAL);
1397 }
1398 
1399 int
1400 linux_fcntl(struct thread *td, struct linux_fcntl_args *args)
1401 {
1402 
1403 #ifdef DEBUG
1404 	if (ldebug(fcntl))
1405 		printf(ARGS(fcntl, "%d, %08x, *"), args->fd, args->cmd);
1406 #endif
1407 
1408 	return (fcntl_common(td, args));
1409 }
1410 
1411 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1412 int
1413 linux_fcntl64(struct thread *td, struct linux_fcntl64_args *args)
1414 {
1415 	struct l_flock64 linux_flock;
1416 	struct flock bsd_flock;
1417 	struct linux_fcntl_args fcntl_args;
1418 	int error;
1419 
1420 #ifdef DEBUG
1421 	if (ldebug(fcntl64))
1422 		printf(ARGS(fcntl64, "%d, %08x, *"), args->fd, args->cmd);
1423 #endif
1424 
1425 	switch (args->cmd) {
1426 	case LINUX_F_GETLK64:
1427 		error = copyin((void *)args->arg, &linux_flock,
1428 		    sizeof(linux_flock));
1429 		if (error)
1430 			return (error);
1431 		linux_to_bsd_flock64(&linux_flock, &bsd_flock);
1432 		error = kern_fcntl(td, args->fd, F_GETLK, (intptr_t)&bsd_flock);
1433 		if (error)
1434 			return (error);
1435 		bsd_to_linux_flock64(&bsd_flock, &linux_flock);
1436 		return (copyout(&linux_flock, (void *)args->arg,
1437 			    sizeof(linux_flock)));
1438 
1439 	case LINUX_F_SETLK64:
1440 		error = copyin((void *)args->arg, &linux_flock,
1441 		    sizeof(linux_flock));
1442 		if (error)
1443 			return (error);
1444 		linux_to_bsd_flock64(&linux_flock, &bsd_flock);
1445 		return (kern_fcntl(td, args->fd, F_SETLK,
1446 		    (intptr_t)&bsd_flock));
1447 
1448 	case LINUX_F_SETLKW64:
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_SETLKW,
1455 		    (intptr_t)&bsd_flock));
1456 	}
1457 
1458 	fcntl_args.fd = args->fd;
1459 	fcntl_args.cmd = args->cmd;
1460 	fcntl_args.arg = args->arg;
1461 	return (fcntl_common(td, &fcntl_args));
1462 }
1463 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1464 
1465 int
1466 linux_chown(struct thread *td, struct linux_chown_args *args)
1467 {
1468 	char *path;
1469 	int error;
1470 
1471 	LCONVPATHEXIST(td, args->path, &path);
1472 
1473 #ifdef DEBUG
1474 	if (ldebug(chown))
1475 		printf(ARGS(chown, "%s, %d, %d"), path, args->uid, args->gid);
1476 #endif
1477 	error = kern_fchownat(td, AT_FDCWD, path, UIO_SYSSPACE, args->uid,
1478 	    args->gid, 0);
1479 	LFREEPATH(path);
1480 	return (error);
1481 }
1482 
1483 int
1484 linux_fchownat(struct thread *td, struct linux_fchownat_args *args)
1485 {
1486 	char *path;
1487 	int error, dfd, flag;
1488 
1489 	if (args->flag & ~LINUX_AT_SYMLINK_NOFOLLOW)
1490 		return (EINVAL);
1491 
1492 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD :  args->dfd;
1493 	LCONVPATHEXIST_AT(td, args->filename, &path, dfd);
1494 
1495 #ifdef DEBUG
1496 	if (ldebug(fchownat))
1497 		printf(ARGS(fchownat, "%s, %d, %d"), path, args->uid, args->gid);
1498 #endif
1499 
1500 	flag = (args->flag & LINUX_AT_SYMLINK_NOFOLLOW) == 0 ? 0 :
1501 	    AT_SYMLINK_NOFOLLOW;
1502 	error = kern_fchownat(td, dfd, path, UIO_SYSSPACE, args->uid, args->gid,
1503 	    flag);
1504 	LFREEPATH(path);
1505 	return (error);
1506 }
1507 
1508 int
1509 linux_lchown(struct thread *td, struct linux_lchown_args *args)
1510 {
1511 	char *path;
1512 	int error;
1513 
1514 	LCONVPATHEXIST(td, args->path, &path);
1515 
1516 #ifdef DEBUG
1517 	if (ldebug(lchown))
1518 		printf(ARGS(lchown, "%s, %d, %d"), path, args->uid, args->gid);
1519 #endif
1520 	error = kern_fchownat(td, AT_FDCWD, path, UIO_SYSSPACE, args->uid,
1521 	    args->gid, AT_SYMLINK_NOFOLLOW);
1522 	LFREEPATH(path);
1523 	return (error);
1524 }
1525 
1526 static int
1527 convert_fadvice(int advice)
1528 {
1529 	switch (advice) {
1530 	case LINUX_POSIX_FADV_NORMAL:
1531 		return (POSIX_FADV_NORMAL);
1532 	case LINUX_POSIX_FADV_RANDOM:
1533 		return (POSIX_FADV_RANDOM);
1534 	case LINUX_POSIX_FADV_SEQUENTIAL:
1535 		return (POSIX_FADV_SEQUENTIAL);
1536 	case LINUX_POSIX_FADV_WILLNEED:
1537 		return (POSIX_FADV_WILLNEED);
1538 	case LINUX_POSIX_FADV_DONTNEED:
1539 		return (POSIX_FADV_DONTNEED);
1540 	case LINUX_POSIX_FADV_NOREUSE:
1541 		return (POSIX_FADV_NOREUSE);
1542 	default:
1543 		return (-1);
1544 	}
1545 }
1546 
1547 int
1548 linux_fadvise64(struct thread *td, struct linux_fadvise64_args *args)
1549 {
1550 	int advice;
1551 
1552 	advice = convert_fadvice(args->advice);
1553 	if (advice == -1)
1554 		return (EINVAL);
1555 	return (kern_posix_fadvise(td, args->fd, args->offset, args->len,
1556 	    advice));
1557 }
1558 
1559 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1560 int
1561 linux_fadvise64_64(struct thread *td, struct linux_fadvise64_64_args *args)
1562 {
1563 	int advice;
1564 
1565 	advice = convert_fadvice(args->advice);
1566 	if (advice == -1)
1567 		return (EINVAL);
1568 	return (kern_posix_fadvise(td, args->fd, args->offset, args->len,
1569 	    advice));
1570 }
1571 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
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_pipe(td, fildes, 0, NULL, NULL);
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_pipe(td, fildes, flags, NULL, NULL);
1612 	if (error)
1613 		return (error);
1614 
1615 	/* XXX: Close descriptors on error. */
1616 	return (copyout(fildes, args->pipefds, sizeof(fildes)));
1617 }
1618 
1619 int
1620 linux_dup3(struct thread *td, struct linux_dup3_args *args)
1621 {
1622 	int cmd;
1623 	intptr_t newfd;
1624 
1625 	if (args->oldfd == args->newfd)
1626 		return (EINVAL);
1627 	if ((args->flags & ~LINUX_O_CLOEXEC) != 0)
1628 		return (EINVAL);
1629 	if (args->flags & LINUX_O_CLOEXEC)
1630 		cmd = F_DUP2FD_CLOEXEC;
1631 	else
1632 		cmd = F_DUP2FD;
1633 
1634 	newfd = args->newfd;
1635 	return (kern_fcntl(td, args->oldfd, cmd, newfd));
1636 }
1637 
1638 int
1639 linux_fallocate(struct thread *td, struct linux_fallocate_args *args)
1640 {
1641 
1642 	/*
1643 	 * We emulate only posix_fallocate system call for which
1644 	 * mode should be 0.
1645 	 */
1646 	if (args->mode != 0)
1647 		return (ENOSYS);
1648 
1649 	return (kern_posix_fallocate(td, args->fd, args->offset,
1650 	    args->len));
1651 }
1652