xref: /freebsd/sys/compat/linux/linux_file.c (revision 21e0559cbcdf275377da88dec0d7a98e85f00f80)
1 /*-
2  * Copyright (c) 1994-1995 Søren Schmidt
3  * All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer
10  *    in this position and unchanged.
11  * 2. Redistributions in binary form must reproduce the above copyright
12  *    notice, this list of conditions and the following disclaimer in the
13  *    documentation and/or other materials provided with the distribution.
14  * 3. The name of the author may not be used to endorse or promote products
15  *    derived from this software without specific prior written permission
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
18  * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
19  * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
20  * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
21  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
22  * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
23  * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
24  * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
25  * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
26  * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
27  */
28 
29 #include <sys/cdefs.h>
30 __FBSDID("$FreeBSD$");
31 
32 #include "opt_compat.h"
33 
34 #include <sys/param.h>
35 #include <sys/systm.h>
36 #include <sys/capability.h>
37 #include <sys/conf.h>
38 #include <sys/dirent.h>
39 #include <sys/fcntl.h>
40 #include <sys/file.h>
41 #include <sys/filedesc.h>
42 #include <sys/lock.h>
43 #include <sys/malloc.h>
44 #include <sys/mount.h>
45 #include <sys/mutex.h>
46 #include <sys/namei.h>
47 #include <sys/proc.h>
48 #include <sys/stat.h>
49 #include <sys/sx.h>
50 #include <sys/syscallsubr.h>
51 #include <sys/sysproto.h>
52 #include <sys/tty.h>
53 #include <sys/unistd.h>
54 #include <sys/vnode.h>
55 
56 #include <security/mac/mac_framework.h>
57 
58 #include <ufs/ufs/extattr.h>
59 #include <ufs/ufs/quota.h>
60 #include <ufs/ufs/ufsmount.h>
61 
62 #ifdef COMPAT_LINUX32
63 #include <machine/../linux32/linux.h>
64 #include <machine/../linux32/linux32_proto.h>
65 #else
66 #include <machine/../linux/linux.h>
67 #include <machine/../linux/linux_proto.h>
68 #endif
69 #include <compat/linux/linux_misc.h>
70 #include <compat/linux/linux_util.h>
71 #include <compat/linux/linux_file.h>
72 
73 /* XXX */
74 int	do_pipe(struct thread *td, int fildes[2], int flags);
75 
76 int
77 linux_creat(struct thread *td, struct linux_creat_args *args)
78 {
79     char *path;
80     int error;
81 
82     LCONVPATHEXIST(td, args->path, &path);
83 
84 #ifdef DEBUG
85 	if (ldebug(creat))
86 		printf(ARGS(creat, "%s, %d"), path, args->mode);
87 #endif
88     error = kern_open(td, path, UIO_SYSSPACE, O_WRONLY | O_CREAT | O_TRUNC,
89 	args->mode);
90     LFREEPATH(path);
91     return (error);
92 }
93 
94 
95 static int
96 linux_common_open(struct thread *td, int dirfd, char *path, int l_flags, int mode)
97 {
98     struct proc *p = td->td_proc;
99     struct file *fp;
100     int fd;
101     int bsd_flags, error;
102 
103     bsd_flags = 0;
104     switch (l_flags & LINUX_O_ACCMODE) {
105     case LINUX_O_WRONLY:
106 	bsd_flags |= O_WRONLY;
107 	break;
108     case LINUX_O_RDWR:
109 	bsd_flags |= O_RDWR;
110 	break;
111     default:
112 	bsd_flags |= O_RDONLY;
113     }
114     if (l_flags & LINUX_O_NDELAY)
115 	bsd_flags |= O_NONBLOCK;
116     if (l_flags & LINUX_O_APPEND)
117 	bsd_flags |= O_APPEND;
118     if (l_flags & LINUX_O_SYNC)
119 	bsd_flags |= O_FSYNC;
120     if (l_flags & LINUX_O_NONBLOCK)
121 	bsd_flags |= O_NONBLOCK;
122     if (l_flags & LINUX_FASYNC)
123 	bsd_flags |= O_ASYNC;
124     if (l_flags & LINUX_O_CREAT)
125 	bsd_flags |= O_CREAT;
126     if (l_flags & LINUX_O_TRUNC)
127 	bsd_flags |= O_TRUNC;
128     if (l_flags & LINUX_O_EXCL)
129 	bsd_flags |= O_EXCL;
130     if (l_flags & LINUX_O_NOCTTY)
131 	bsd_flags |= O_NOCTTY;
132     if (l_flags & LINUX_O_DIRECT)
133 	bsd_flags |= O_DIRECT;
134     if (l_flags & LINUX_O_NOFOLLOW)
135 	bsd_flags |= O_NOFOLLOW;
136     if (l_flags & LINUX_O_DIRECTORY)
137 	bsd_flags |= O_DIRECTORY;
138     /* XXX LINUX_O_NOATIME: unable to be easily implemented. */
139 
140     error = kern_openat(td, dirfd, path, UIO_SYSSPACE, bsd_flags, mode);
141 
142     if (!error) {
143 	    fd = td->td_retval[0];
144 	    /*
145 	     * XXX In between kern_open() and fget(), another process
146 	     * having the same filedesc could use that fd without
147 	     * checking below.
148 	     */
149 	    error = fget(td, fd, CAP_IOCTL, &fp);
150 	    if (!error) {
151 		    sx_slock(&proctree_lock);
152 		    PROC_LOCK(p);
153 		    if (!(bsd_flags & O_NOCTTY) &&
154 			SESS_LEADER(p) && !(p->p_flag & P_CONTROLT)) {
155 			    PROC_UNLOCK(p);
156 			    sx_unlock(&proctree_lock);
157 			    /* XXXPJD: Verify if TIOCSCTTY is allowed. */
158 			    if (fp->f_type == DTYPE_VNODE)
159 				    (void) fo_ioctl(fp, TIOCSCTTY, (caddr_t) 0,
160 					     td->td_ucred, td);
161 		    } else {
162 			    PROC_UNLOCK(p);
163 			    sx_sunlock(&proctree_lock);
164 		    }
165 		    fdrop(fp, td);
166 		    /*
167 		     * XXX as above, fdrop()/kern_close() pair is racy.
168 		     */
169 		    if (error)
170 			    kern_close(td, fd);
171 	    }
172     }
173 
174 #ifdef DEBUG
175     if (ldebug(open))
176 	    printf(LMSG("open returns error %d"), error);
177 #endif
178     LFREEPATH(path);
179     return (error);
180 }
181 
182 int
183 linux_openat(struct thread *td, struct linux_openat_args *args)
184 {
185 	char *path;
186 	int dfd;
187 
188 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
189 	if (args->flags & LINUX_O_CREAT)
190 		LCONVPATH_AT(td, args->filename, &path, 1, dfd);
191 	else
192 		LCONVPATH_AT(td, args->filename, &path, 0, dfd);
193 #ifdef DEBUG
194 	if (ldebug(openat))
195 		printf(ARGS(openat, "%i, %s, 0x%x, 0x%x"), args->dfd,
196 		    path, args->flags, args->mode);
197 #endif
198 	return (linux_common_open(td, dfd, path, args->flags, args->mode));
199 }
200 
201 int
202 linux_open(struct thread *td, struct linux_open_args *args)
203 {
204     char *path;
205 
206     if (args->flags & LINUX_O_CREAT)
207 	LCONVPATHCREAT(td, args->path, &path);
208     else
209 	LCONVPATHEXIST(td, args->path, &path);
210 
211 #ifdef DEBUG
212 	if (ldebug(open))
213 		printf(ARGS(open, "%s, 0x%x, 0x%x"),
214 		    path, args->flags, args->mode);
215 #endif
216 
217 	return (linux_common_open(td, AT_FDCWD, path, args->flags, args->mode));
218 }
219 
220 int
221 linux_lseek(struct thread *td, struct linux_lseek_args *args)
222 {
223 
224     struct lseek_args /* {
225 	int fd;
226 	int pad;
227 	off_t offset;
228 	int whence;
229     } */ tmp_args;
230     int error;
231 
232 #ifdef DEBUG
233 	if (ldebug(lseek))
234 		printf(ARGS(lseek, "%d, %ld, %d"),
235 		    args->fdes, (long)args->off, args->whence);
236 #endif
237     tmp_args.fd = args->fdes;
238     tmp_args.offset = (off_t)args->off;
239     tmp_args.whence = args->whence;
240     error = sys_lseek(td, &tmp_args);
241     return error;
242 }
243 
244 int
245 linux_llseek(struct thread *td, struct linux_llseek_args *args)
246 {
247 	struct lseek_args bsd_args;
248 	int error;
249 	off_t off;
250 
251 #ifdef DEBUG
252 	if (ldebug(llseek))
253 		printf(ARGS(llseek, "%d, %d:%d, %d"),
254 		    args->fd, args->ohigh, args->olow, args->whence);
255 #endif
256 	off = (args->olow) | (((off_t) args->ohigh) << 32);
257 
258 	bsd_args.fd = args->fd;
259 	bsd_args.offset = off;
260 	bsd_args.whence = args->whence;
261 
262 	if ((error = sys_lseek(td, &bsd_args)))
263 		return error;
264 
265 	if ((error = copyout(td->td_retval, args->res, sizeof (off_t))))
266 		return error;
267 
268 	td->td_retval[0] = 0;
269 	return 0;
270 }
271 
272 int
273 linux_readdir(struct thread *td, struct linux_readdir_args *args)
274 {
275 	struct linux_getdents_args lda;
276 
277 	lda.fd = args->fd;
278 	lda.dent = args->dent;
279 	lda.count = 1;
280 	return linux_getdents(td, &lda);
281 }
282 
283 /*
284  * Note that linux_getdents(2) and linux_getdents64(2) have the same
285  * arguments. They only differ in the definition of struct dirent they
286  * operate on. We use this to common the code, with the exception of
287  * accessing struct dirent. Note that linux_readdir(2) is implemented
288  * by means of linux_getdents(2). In this case we never operate on
289  * struct dirent64 and thus don't need to handle it...
290  */
291 
292 struct l_dirent {
293 	l_ulong		d_ino;
294 	l_off_t		d_off;
295 	l_ushort	d_reclen;
296 	char		d_name[LINUX_NAME_MAX + 1];
297 };
298 
299 struct l_dirent64 {
300 	uint64_t	d_ino;
301 	int64_t		d_off;
302 	l_ushort	d_reclen;
303 	u_char		d_type;
304 	char		d_name[LINUX_NAME_MAX + 1];
305 };
306 
307 /*
308  * Linux uses the last byte in the dirent buffer to store d_type,
309  * at least glibc-2.7 requires it. That is why l_dirent is padded with 2 bytes.
310  */
311 #define LINUX_RECLEN(namlen)						\
312     roundup((offsetof(struct l_dirent, d_name) + (namlen) + 2),		\
313     sizeof(l_ulong))
314 
315 #define LINUX_RECLEN64(namlen)						\
316     roundup((offsetof(struct l_dirent64, d_name) + (namlen) + 1),	\
317     sizeof(uint64_t))
318 
319 #define LINUX_MAXRECLEN		max(LINUX_RECLEN(LINUX_NAME_MAX),	\
320 				    LINUX_RECLEN64(LINUX_NAME_MAX))
321 #define	LINUX_DIRBLKSIZ		512
322 
323 static int
324 getdents_common(struct thread *td, struct linux_getdents64_args *args,
325     int is64bit)
326 {
327 	struct dirent *bdp;
328 	struct vnode *vp;
329 	caddr_t inp, buf;		/* BSD-format */
330 	int len, reclen;		/* BSD-format */
331 	caddr_t outp;			/* Linux-format */
332 	int resid, linuxreclen=0;	/* Linux-format */
333 	caddr_t lbuf;			/* Linux-format */
334 	struct file *fp;
335 	struct uio auio;
336 	struct iovec aiov;
337 	off_t off;
338 	struct l_dirent *linux_dirent;
339 	struct l_dirent64 *linux_dirent64;
340 	int buflen, error, eofflag, nbytes, justone;
341 	u_long *cookies = NULL, *cookiep;
342 	int ncookies;
343 
344 	nbytes = args->count;
345 	if (nbytes == 1) {
346 		/* readdir(2) case. Always struct dirent. */
347 		if (is64bit)
348 			return (EINVAL);
349 		nbytes = sizeof(*linux_dirent);
350 		justone = 1;
351 	} else
352 		justone = 0;
353 
354 	if ((error = getvnode(td->td_proc->p_fd, args->fd, CAP_READ, &fp)) != 0)
355 		return (error);
356 
357 	if ((fp->f_flag & FREAD) == 0) {
358 		fdrop(fp, td);
359 		return (EBADF);
360 	}
361 
362 	off = foffset_lock(fp, 0);
363 	vp = fp->f_vnode;
364 	if (vp->v_type != VDIR) {
365 		foffset_unlock(fp, off, 0);
366 		fdrop(fp, td);
367 		return (EINVAL);
368 	}
369 
370 
371 	buflen = max(LINUX_DIRBLKSIZ, nbytes);
372 	buflen = min(buflen, MAXBSIZE);
373 	buf = malloc(buflen, M_TEMP, M_WAITOK);
374 	lbuf = malloc(LINUX_MAXRECLEN, M_TEMP, M_WAITOK | M_ZERO);
375 	vn_lock(vp, LK_SHARED | LK_RETRY);
376 
377 	aiov.iov_base = buf;
378 	aiov.iov_len = buflen;
379 	auio.uio_iov = &aiov;
380 	auio.uio_iovcnt = 1;
381 	auio.uio_rw = UIO_READ;
382 	auio.uio_segflg = UIO_SYSSPACE;
383 	auio.uio_td = td;
384 	auio.uio_resid = buflen;
385 	auio.uio_offset = off;
386 
387 #ifdef MAC
388 	/*
389 	 * Do directory search MAC check using non-cached credentials.
390 	 */
391 	if ((error = mac_vnode_check_readdir(td->td_ucred, vp)))
392 		goto out;
393 #endif /* MAC */
394 	if ((error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, &ncookies,
395 		 &cookies)))
396 		goto out;
397 
398 	inp = buf;
399 	outp = (caddr_t)args->dirent;
400 	resid = nbytes;
401 	if ((len = buflen - auio.uio_resid) <= 0)
402 		goto eof;
403 
404 	cookiep = cookies;
405 
406 	if (cookies) {
407 		/*
408 		 * When using cookies, the vfs has the option of reading from
409 		 * a different offset than that supplied (UFS truncates the
410 		 * offset to a block boundary to make sure that it never reads
411 		 * partway through a directory entry, even if the directory
412 		 * has been compacted).
413 		 */
414 		while (len > 0 && ncookies > 0 && *cookiep <= off) {
415 			bdp = (struct dirent *) inp;
416 			len -= bdp->d_reclen;
417 			inp += bdp->d_reclen;
418 			cookiep++;
419 			ncookies--;
420 		}
421 	}
422 
423 	while (len > 0) {
424 		if (cookiep && ncookies == 0)
425 			break;
426 		bdp = (struct dirent *) inp;
427 		reclen = bdp->d_reclen;
428 		if (reclen & 3) {
429 			error = EFAULT;
430 			goto out;
431 		}
432 
433 		if (bdp->d_fileno == 0) {
434 			inp += reclen;
435 			if (cookiep) {
436 				off = *cookiep++;
437 				ncookies--;
438 			} else
439 				off += reclen;
440 
441 			len -= reclen;
442 			continue;
443 		}
444 
445 		linuxreclen = (is64bit)
446 		    ? LINUX_RECLEN64(bdp->d_namlen)
447 		    : LINUX_RECLEN(bdp->d_namlen);
448 
449 		if (reclen > len || resid < linuxreclen) {
450 			outp++;
451 			break;
452 		}
453 
454 		if (justone) {
455 			/* readdir(2) case. */
456 			linux_dirent = (struct l_dirent*)lbuf;
457 			linux_dirent->d_ino = bdp->d_fileno;
458 			linux_dirent->d_off = (l_off_t)linuxreclen;
459 			linux_dirent->d_reclen = (l_ushort)bdp->d_namlen;
460 			strlcpy(linux_dirent->d_name, bdp->d_name,
461 			    linuxreclen - offsetof(struct l_dirent, d_name));
462 			error = copyout(linux_dirent, outp, linuxreclen);
463 		}
464 		if (is64bit) {
465 			linux_dirent64 = (struct l_dirent64*)lbuf;
466 			linux_dirent64->d_ino = bdp->d_fileno;
467 			linux_dirent64->d_off = (cookiep)
468 			    ? (l_off_t)*cookiep
469 			    : (l_off_t)(off + reclen);
470 			linux_dirent64->d_reclen = (l_ushort)linuxreclen;
471 			linux_dirent64->d_type = bdp->d_type;
472 			strlcpy(linux_dirent64->d_name, bdp->d_name,
473 			    linuxreclen - offsetof(struct l_dirent64, d_name));
474 			error = copyout(linux_dirent64, outp, linuxreclen);
475 		} else if (!justone) {
476 			linux_dirent = (struct l_dirent*)lbuf;
477 			linux_dirent->d_ino = bdp->d_fileno;
478 			linux_dirent->d_off = (cookiep)
479 			    ? (l_off_t)*cookiep
480 			    : (l_off_t)(off + reclen);
481 			linux_dirent->d_reclen = (l_ushort)linuxreclen;
482 			/*
483 			 * Copy d_type to last byte of l_dirent buffer
484 			 */
485 			lbuf[linuxreclen-1] = bdp->d_type;
486 			strlcpy(linux_dirent->d_name, bdp->d_name,
487 			    linuxreclen - offsetof(struct l_dirent, d_name)-1);
488 			error = copyout(linux_dirent, outp, linuxreclen);
489 		}
490 
491 		if (error)
492 			goto out;
493 
494 		inp += reclen;
495 		if (cookiep) {
496 			off = *cookiep++;
497 			ncookies--;
498 		} else
499 			off += reclen;
500 
501 		outp += linuxreclen;
502 		resid -= linuxreclen;
503 		len -= reclen;
504 		if (justone)
505 			break;
506 	}
507 
508 	if (outp == (caddr_t)args->dirent) {
509 		nbytes = resid;
510 		goto eof;
511 	}
512 
513 	if (justone)
514 		nbytes = resid + linuxreclen;
515 
516 eof:
517 	td->td_retval[0] = nbytes - resid;
518 
519 out:
520 	if (cookies)
521 		free(cookies, M_TEMP);
522 
523 	VOP_UNLOCK(vp, 0);
524 	foffset_unlock(fp, off, 0);
525 	fdrop(fp, td);
526 	free(buf, M_TEMP);
527 	free(lbuf, M_TEMP);
528 	return (error);
529 }
530 
531 int
532 linux_getdents(struct thread *td, struct linux_getdents_args *args)
533 {
534 
535 #ifdef DEBUG
536 	if (ldebug(getdents))
537 		printf(ARGS(getdents, "%d, *, %d"), args->fd, args->count);
538 #endif
539 
540 	return (getdents_common(td, (struct linux_getdents64_args*)args, 0));
541 }
542 
543 int
544 linux_getdents64(struct thread *td, struct linux_getdents64_args *args)
545 {
546 
547 #ifdef DEBUG
548 	if (ldebug(getdents64))
549 		printf(ARGS(getdents64, "%d, *, %d"), args->fd, args->count);
550 #endif
551 
552 	return (getdents_common(td, args, 1));
553 }
554 
555 /*
556  * These exist mainly for hooks for doing /compat/linux translation.
557  */
558 
559 int
560 linux_access(struct thread *td, struct linux_access_args *args)
561 {
562 	char *path;
563 	int error;
564 
565 	/* linux convention */
566 	if (args->amode & ~(F_OK | X_OK | W_OK | R_OK))
567 		return (EINVAL);
568 
569 	LCONVPATHEXIST(td, args->path, &path);
570 
571 #ifdef DEBUG
572 	if (ldebug(access))
573 		printf(ARGS(access, "%s, %d"), path, args->amode);
574 #endif
575 	error = kern_access(td, path, UIO_SYSSPACE, args->amode);
576 	LFREEPATH(path);
577 
578 	return (error);
579 }
580 
581 int
582 linux_faccessat(struct thread *td, struct linux_faccessat_args *args)
583 {
584 	char *path;
585 	int error, dfd, flag;
586 
587 	if (args->flag & ~LINUX_AT_EACCESS)
588 		return (EINVAL);
589 	/* linux convention */
590 	if (args->amode & ~(F_OK | X_OK | W_OK | R_OK))
591 		return (EINVAL);
592 
593 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
594 	LCONVPATHEXIST_AT(td, args->filename, &path, dfd);
595 
596 #ifdef DEBUG
597 	if (ldebug(access))
598 		printf(ARGS(access, "%s, %d"), path, args->amode);
599 #endif
600 
601 	flag = (args->flag & LINUX_AT_EACCESS) == 0 ? 0 : AT_EACCESS;
602 	error = kern_accessat(td, dfd, path, UIO_SYSSPACE, flag, args->amode);
603 	LFREEPATH(path);
604 
605 	return (error);
606 }
607 
608 int
609 linux_unlink(struct thread *td, struct linux_unlink_args *args)
610 {
611 	char *path;
612 	int error;
613 	struct stat st;
614 
615 	LCONVPATHEXIST(td, args->path, &path);
616 
617 #ifdef DEBUG
618 	if (ldebug(unlink))
619 		printf(ARGS(unlink, "%s"), path);
620 #endif
621 
622 	error = kern_unlink(td, path, UIO_SYSSPACE);
623 	if (error == EPERM)
624 		/* Introduce POSIX noncompliant behaviour of Linux */
625 		if (kern_stat(td, path, UIO_SYSSPACE, &st) == 0)
626 			if (S_ISDIR(st.st_mode))
627 				error = EISDIR;
628 	LFREEPATH(path);
629 	return (error);
630 }
631 
632 int
633 linux_unlinkat(struct thread *td, struct linux_unlinkat_args *args)
634 {
635 	char *path;
636 	int error, dfd;
637 	struct stat st;
638 
639 	if (args->flag & ~LINUX_AT_REMOVEDIR)
640 		return (EINVAL);
641 
642 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
643 	LCONVPATHEXIST_AT(td, args->pathname, &path, dfd);
644 
645 #ifdef DEBUG
646 	if (ldebug(unlinkat))
647 		printf(ARGS(unlinkat, "%s"), path);
648 #endif
649 
650 	if (args->flag & LINUX_AT_REMOVEDIR)
651 		error = kern_rmdirat(td, dfd, path, UIO_SYSSPACE);
652 	else
653 		error = kern_unlinkat(td, dfd, path, UIO_SYSSPACE, 0);
654 	if (error == EPERM && !(args->flag & LINUX_AT_REMOVEDIR)) {
655 		/* Introduce POSIX noncompliant behaviour of Linux */
656 		if (kern_statat(td, AT_SYMLINK_NOFOLLOW, dfd, path,
657 		    UIO_SYSSPACE, &st) == 0 && S_ISDIR(st.st_mode))
658 			error = EISDIR;
659 	}
660 	LFREEPATH(path);
661 	return (error);
662 }
663 int
664 linux_chdir(struct thread *td, struct linux_chdir_args *args)
665 {
666 	char *path;
667 	int error;
668 
669 	LCONVPATHEXIST(td, args->path, &path);
670 
671 #ifdef DEBUG
672 	if (ldebug(chdir))
673 		printf(ARGS(chdir, "%s"), path);
674 #endif
675 	error = kern_chdir(td, path, UIO_SYSSPACE);
676 	LFREEPATH(path);
677 	return (error);
678 }
679 
680 int
681 linux_chmod(struct thread *td, struct linux_chmod_args *args)
682 {
683 	char *path;
684 	int error;
685 
686 	LCONVPATHEXIST(td, args->path, &path);
687 
688 #ifdef DEBUG
689 	if (ldebug(chmod))
690 		printf(ARGS(chmod, "%s, %d"), path, args->mode);
691 #endif
692 	error = kern_chmod(td, path, UIO_SYSSPACE, args->mode);
693 	LFREEPATH(path);
694 	return (error);
695 }
696 
697 int
698 linux_fchmodat(struct thread *td, struct linux_fchmodat_args *args)
699 {
700 	char *path;
701 	int error, dfd;
702 
703 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
704 	LCONVPATHEXIST_AT(td, args->filename, &path, dfd);
705 
706 #ifdef DEBUG
707 	if (ldebug(fchmodat))
708 		printf(ARGS(fchmodat, "%s, %d"), path, args->mode);
709 #endif
710 
711 	error = kern_fchmodat(td, dfd, path, UIO_SYSSPACE, args->mode, 0);
712 	LFREEPATH(path);
713 	return (error);
714 }
715 
716 int
717 linux_mkdir(struct thread *td, struct linux_mkdir_args *args)
718 {
719 	char *path;
720 	int error;
721 
722 	LCONVPATHCREAT(td, args->path, &path);
723 
724 #ifdef DEBUG
725 	if (ldebug(mkdir))
726 		printf(ARGS(mkdir, "%s, %d"), path, args->mode);
727 #endif
728 	error = kern_mkdir(td, path, UIO_SYSSPACE, args->mode);
729 	LFREEPATH(path);
730 	return (error);
731 }
732 
733 int
734 linux_mkdirat(struct thread *td, struct linux_mkdirat_args *args)
735 {
736 	char *path;
737 	int error, dfd;
738 
739 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
740 	LCONVPATHCREAT_AT(td, args->pathname, &path, dfd);
741 
742 #ifdef DEBUG
743 	if (ldebug(mkdirat))
744 		printf(ARGS(mkdirat, "%s, %d"), path, args->mode);
745 #endif
746 	error = kern_mkdirat(td, dfd, path, UIO_SYSSPACE, args->mode);
747 	LFREEPATH(path);
748 	return (error);
749 }
750 
751 int
752 linux_rmdir(struct thread *td, struct linux_rmdir_args *args)
753 {
754 	char *path;
755 	int error;
756 
757 	LCONVPATHEXIST(td, args->path, &path);
758 
759 #ifdef DEBUG
760 	if (ldebug(rmdir))
761 		printf(ARGS(rmdir, "%s"), path);
762 #endif
763 	error = kern_rmdir(td, path, UIO_SYSSPACE);
764 	LFREEPATH(path);
765 	return (error);
766 }
767 
768 int
769 linux_rename(struct thread *td, struct linux_rename_args *args)
770 {
771 	char *from, *to;
772 	int error;
773 
774 	LCONVPATHEXIST(td, args->from, &from);
775 	/* Expand LCONVPATHCREATE so that `from' can be freed on errors */
776 	error = linux_emul_convpath(td, args->to, UIO_USERSPACE, &to, 1, AT_FDCWD);
777 	if (to == NULL) {
778 		LFREEPATH(from);
779 		return (error);
780 	}
781 
782 #ifdef DEBUG
783 	if (ldebug(rename))
784 		printf(ARGS(rename, "%s, %s"), from, to);
785 #endif
786 	error = kern_rename(td, from, to, UIO_SYSSPACE);
787 	LFREEPATH(from);
788 	LFREEPATH(to);
789 	return (error);
790 }
791 
792 int
793 linux_renameat(struct thread *td, struct linux_renameat_args *args)
794 {
795 	char *from, *to;
796 	int error, olddfd, newdfd;
797 
798 	olddfd = (args->olddfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->olddfd;
799 	newdfd = (args->newdfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->newdfd;
800 	LCONVPATHEXIST_AT(td, args->oldname, &from, olddfd);
801 	/* Expand LCONVPATHCREATE so that `from' can be freed on errors */
802 	error = linux_emul_convpath(td, args->newname, UIO_USERSPACE, &to, 1, newdfd);
803 	if (to == NULL) {
804 		LFREEPATH(from);
805 		return (error);
806 	}
807 
808 #ifdef DEBUG
809 	if (ldebug(renameat))
810 		printf(ARGS(renameat, "%s, %s"), from, to);
811 #endif
812 	error = kern_renameat(td, olddfd, from, newdfd, to, UIO_SYSSPACE);
813 	LFREEPATH(from);
814 	LFREEPATH(to);
815 	return (error);
816 }
817 
818 int
819 linux_symlink(struct thread *td, struct linux_symlink_args *args)
820 {
821 	char *path, *to;
822 	int error;
823 
824 	LCONVPATHEXIST(td, args->path, &path);
825 	/* Expand LCONVPATHCREATE so that `path' can be freed on errors */
826 	error = linux_emul_convpath(td, args->to, UIO_USERSPACE, &to, 1, AT_FDCWD);
827 	if (to == NULL) {
828 		LFREEPATH(path);
829 		return (error);
830 	}
831 
832 #ifdef DEBUG
833 	if (ldebug(symlink))
834 		printf(ARGS(symlink, "%s, %s"), path, to);
835 #endif
836 	error = kern_symlink(td, path, to, UIO_SYSSPACE);
837 	LFREEPATH(path);
838 	LFREEPATH(to);
839 	return (error);
840 }
841 
842 int
843 linux_symlinkat(struct thread *td, struct linux_symlinkat_args *args)
844 {
845 	char *path, *to;
846 	int error, dfd;
847 
848 	dfd = (args->newdfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->newdfd;
849 	LCONVPATHEXIST_AT(td, args->oldname, &path, dfd);
850 	/* Expand LCONVPATHCREATE so that `path' can be freed on errors */
851 	error = linux_emul_convpath(td, args->newname, UIO_USERSPACE, &to, 1, dfd);
852 	if (to == NULL) {
853 		LFREEPATH(path);
854 		return (error);
855 	}
856 
857 #ifdef DEBUG
858 	if (ldebug(symlinkat))
859 		printf(ARGS(symlinkat, "%s, %s"), path, to);
860 #endif
861 
862 	error = kern_symlinkat(td, path, dfd, to, UIO_SYSSPACE);
863 	LFREEPATH(path);
864 	LFREEPATH(to);
865 	return (error);
866 }
867 
868 int
869 linux_readlink(struct thread *td, struct linux_readlink_args *args)
870 {
871 	char *name;
872 	int error;
873 
874 	LCONVPATHEXIST(td, args->name, &name);
875 
876 #ifdef DEBUG
877 	if (ldebug(readlink))
878 		printf(ARGS(readlink, "%s, %p, %d"), name, (void *)args->buf,
879 		    args->count);
880 #endif
881 	error = kern_readlink(td, name, UIO_SYSSPACE, args->buf, UIO_USERSPACE,
882 	    args->count);
883 	LFREEPATH(name);
884 	return (error);
885 }
886 
887 int
888 linux_readlinkat(struct thread *td, struct linux_readlinkat_args *args)
889 {
890 	char *name;
891 	int error, dfd;
892 
893 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
894 	LCONVPATHEXIST_AT(td, args->path, &name, dfd);
895 
896 #ifdef DEBUG
897 	if (ldebug(readlinkat))
898 		printf(ARGS(readlinkat, "%s, %p, %d"), name, (void *)args->buf,
899 		    args->bufsiz);
900 #endif
901 
902 	error = kern_readlinkat(td, dfd, name, UIO_SYSSPACE, args->buf,
903 	    UIO_USERSPACE, args->bufsiz);
904 	LFREEPATH(name);
905 	return (error);
906 }
907 
908 int
909 linux_truncate(struct thread *td, struct linux_truncate_args *args)
910 {
911 	char *path;
912 	int error;
913 
914 	LCONVPATHEXIST(td, args->path, &path);
915 
916 #ifdef DEBUG
917 	if (ldebug(truncate))
918 		printf(ARGS(truncate, "%s, %ld"), path, (long)args->length);
919 #endif
920 
921 	error = kern_truncate(td, path, UIO_SYSSPACE, args->length);
922 	LFREEPATH(path);
923 	return (error);
924 }
925 
926 int
927 linux_truncate64(struct thread *td, struct linux_truncate64_args *args)
928 {
929 	char *path;
930 	int error;
931 
932 	LCONVPATHEXIST(td, args->path, &path);
933 
934 #ifdef DEBUG
935 	if (ldebug(truncate64))
936 		printf(ARGS(truncate64, "%s, %jd"), path, args->length);
937 #endif
938 
939 	error = kern_truncate(td, path, UIO_SYSSPACE, args->length);
940 	LFREEPATH(path);
941 	return (error);
942 }
943 int
944 linux_ftruncate(struct thread *td, struct linux_ftruncate_args *args)
945 {
946 	struct ftruncate_args /* {
947 		int fd;
948 		int pad;
949 		off_t length;
950 		} */ nuap;
951 
952 	nuap.fd = args->fd;
953 	nuap.length = args->length;
954 	return (sys_ftruncate(td, &nuap));
955 }
956 
957 int
958 linux_link(struct thread *td, struct linux_link_args *args)
959 {
960 	char *path, *to;
961 	int error;
962 
963 	LCONVPATHEXIST(td, args->path, &path);
964 	/* Expand LCONVPATHCREATE so that `path' can be freed on errors */
965 	error = linux_emul_convpath(td, args->to, UIO_USERSPACE, &to, 1, AT_FDCWD);
966 	if (to == NULL) {
967 		LFREEPATH(path);
968 		return (error);
969 	}
970 
971 #ifdef DEBUG
972 	if (ldebug(link))
973 		printf(ARGS(link, "%s, %s"), path, to);
974 #endif
975 	error = kern_link(td, path, to, UIO_SYSSPACE);
976 	LFREEPATH(path);
977 	LFREEPATH(to);
978 	return (error);
979 }
980 
981 int
982 linux_linkat(struct thread *td, struct linux_linkat_args *args)
983 {
984 	char *path, *to;
985 	int error, olddfd, newdfd, follow;
986 
987 	if (args->flag & ~LINUX_AT_SYMLINK_FOLLOW)
988 		return (EINVAL);
989 
990 	olddfd = (args->olddfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->olddfd;
991 	newdfd = (args->newdfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->newdfd;
992 	LCONVPATHEXIST_AT(td, args->oldname, &path, olddfd);
993 	/* Expand LCONVPATHCREATE so that `path' can be freed on errors */
994 	error = linux_emul_convpath(td, args->newname, UIO_USERSPACE, &to, 1, newdfd);
995 	if (to == NULL) {
996 		LFREEPATH(path);
997 		return (error);
998 	}
999 
1000 #ifdef DEBUG
1001 	if (ldebug(linkat))
1002 		printf(ARGS(linkat, "%i, %s, %i, %s, %i"), args->olddfd, path,
1003 			args->newdfd, to, args->flag);
1004 #endif
1005 
1006 	follow = (args->flag & LINUX_AT_SYMLINK_FOLLOW) == 0 ? NOFOLLOW :
1007 	    FOLLOW;
1008 	error = kern_linkat(td, olddfd, newdfd, path, to, UIO_SYSSPACE, follow);
1009 	LFREEPATH(path);
1010 	LFREEPATH(to);
1011 	return (error);
1012 }
1013 
1014 int
1015 linux_fdatasync(td, uap)
1016 	struct thread *td;
1017 	struct linux_fdatasync_args *uap;
1018 {
1019 	struct fsync_args bsd;
1020 
1021 	bsd.fd = uap->fd;
1022 	return sys_fsync(td, &bsd);
1023 }
1024 
1025 int
1026 linux_pread(td, uap)
1027 	struct thread *td;
1028 	struct linux_pread_args *uap;
1029 {
1030 	struct pread_args bsd;
1031 	struct vnode *vp;
1032 	int error;
1033 
1034 	bsd.fd = uap->fd;
1035 	bsd.buf = uap->buf;
1036 	bsd.nbyte = uap->nbyte;
1037 	bsd.offset = uap->offset;
1038 
1039 	error = sys_pread(td, &bsd);
1040 
1041 	if (error == 0) {
1042 		/* This seems to violate POSIX but linux does it */
1043 		if ((error = fgetvp(td, uap->fd, CAP_PREAD, &vp)) != 0)
1044 			return (error);
1045 		if (vp->v_type == VDIR) {
1046 			vrele(vp);
1047 			return (EISDIR);
1048 		}
1049 		vrele(vp);
1050 	}
1051 
1052 	return (error);
1053 }
1054 
1055 int
1056 linux_pwrite(td, uap)
1057 	struct thread *td;
1058 	struct linux_pwrite_args *uap;
1059 {
1060 	struct pwrite_args bsd;
1061 
1062 	bsd.fd = uap->fd;
1063 	bsd.buf = uap->buf;
1064 	bsd.nbyte = uap->nbyte;
1065 	bsd.offset = uap->offset;
1066 	return sys_pwrite(td, &bsd);
1067 }
1068 
1069 int
1070 linux_mount(struct thread *td, struct linux_mount_args *args)
1071 {
1072 	struct ufs_args ufs;
1073 	char fstypename[MFSNAMELEN];
1074 	char mntonname[MNAMELEN], mntfromname[MNAMELEN];
1075 	int error;
1076 	int fsflags;
1077 	void *fsdata;
1078 
1079 	error = copyinstr(args->filesystemtype, fstypename, MFSNAMELEN - 1,
1080 	    NULL);
1081 	if (error)
1082 		return (error);
1083 	error = copyinstr(args->specialfile, mntfromname, MNAMELEN - 1, NULL);
1084 	if (error)
1085 		return (error);
1086 	error = copyinstr(args->dir, mntonname, MNAMELEN - 1, NULL);
1087 	if (error)
1088 		return (error);
1089 
1090 #ifdef DEBUG
1091 	if (ldebug(mount))
1092 		printf(ARGS(mount, "%s, %s, %s"),
1093 		    fstypename, mntfromname, mntonname);
1094 #endif
1095 
1096 	if (strcmp(fstypename, "ext2") == 0) {
1097 		strcpy(fstypename, "ext2fs");
1098 		fsdata = &ufs;
1099 		ufs.fspec = mntfromname;
1100 #define DEFAULT_ROOTID		-2
1101 		ufs.export.ex_root = DEFAULT_ROOTID;
1102 		ufs.export.ex_flags =
1103 		    args->rwflag & LINUX_MS_RDONLY ? MNT_EXRDONLY : 0;
1104 	} else if (strcmp(fstypename, "proc") == 0) {
1105 		strcpy(fstypename, "linprocfs");
1106 		fsdata = NULL;
1107 	} else if (strcmp(fstypename, "vfat") == 0) {
1108 		strcpy(fstypename, "msdosfs");
1109 		fsdata = NULL;
1110 	} else {
1111 		return (ENODEV);
1112 	}
1113 
1114 	fsflags = 0;
1115 
1116 	if ((args->rwflag & 0xffff0000) == 0xc0ed0000) {
1117 		/*
1118 		 * Linux SYNC flag is not included; the closest equivalent
1119 		 * FreeBSD has is !ASYNC, which is our default.
1120 		 */
1121 		if (args->rwflag & LINUX_MS_RDONLY)
1122 			fsflags |= MNT_RDONLY;
1123 		if (args->rwflag & LINUX_MS_NOSUID)
1124 			fsflags |= MNT_NOSUID;
1125 		if (args->rwflag & LINUX_MS_NOEXEC)
1126 			fsflags |= MNT_NOEXEC;
1127 		if (args->rwflag & LINUX_MS_REMOUNT)
1128 			fsflags |= MNT_UPDATE;
1129 	}
1130 
1131 	if (strcmp(fstypename, "linprocfs") == 0) {
1132 		error = kernel_vmount(fsflags,
1133 			"fstype", fstypename,
1134 			"fspath", mntonname,
1135 			NULL);
1136 	} else if (strcmp(fstypename, "msdosfs") == 0) {
1137 		error = kernel_vmount(fsflags,
1138 			"fstype", fstypename,
1139 			"fspath", mntonname,
1140 			"from", mntfromname,
1141 			NULL);
1142 	} else
1143 		error = EOPNOTSUPP;
1144 	return (error);
1145 }
1146 
1147 int
1148 linux_oldumount(struct thread *td, struct linux_oldumount_args *args)
1149 {
1150 	struct linux_umount_args args2;
1151 
1152 	args2.path = args->path;
1153 	args2.flags = 0;
1154 	return (linux_umount(td, &args2));
1155 }
1156 
1157 int
1158 linux_umount(struct thread *td, struct linux_umount_args *args)
1159 {
1160 	struct unmount_args bsd;
1161 
1162 	bsd.path = args->path;
1163 	bsd.flags = args->flags;	/* XXX correct? */
1164 	return (sys_unmount(td, &bsd));
1165 }
1166 
1167 /*
1168  * fcntl family of syscalls
1169  */
1170 
1171 struct l_flock {
1172 	l_short		l_type;
1173 	l_short		l_whence;
1174 	l_off_t		l_start;
1175 	l_off_t		l_len;
1176 	l_pid_t		l_pid;
1177 }
1178 #if defined(__amd64__) && defined(COMPAT_LINUX32)
1179 __packed
1180 #endif
1181 ;
1182 
1183 static void
1184 linux_to_bsd_flock(struct l_flock *linux_flock, struct flock *bsd_flock)
1185 {
1186 	switch (linux_flock->l_type) {
1187 	case LINUX_F_RDLCK:
1188 		bsd_flock->l_type = F_RDLCK;
1189 		break;
1190 	case LINUX_F_WRLCK:
1191 		bsd_flock->l_type = F_WRLCK;
1192 		break;
1193 	case LINUX_F_UNLCK:
1194 		bsd_flock->l_type = F_UNLCK;
1195 		break;
1196 	default:
1197 		bsd_flock->l_type = -1;
1198 		break;
1199 	}
1200 	bsd_flock->l_whence = linux_flock->l_whence;
1201 	bsd_flock->l_start = (off_t)linux_flock->l_start;
1202 	bsd_flock->l_len = (off_t)linux_flock->l_len;
1203 	bsd_flock->l_pid = (pid_t)linux_flock->l_pid;
1204 	bsd_flock->l_sysid = 0;
1205 }
1206 
1207 static void
1208 bsd_to_linux_flock(struct flock *bsd_flock, struct l_flock *linux_flock)
1209 {
1210 	switch (bsd_flock->l_type) {
1211 	case F_RDLCK:
1212 		linux_flock->l_type = LINUX_F_RDLCK;
1213 		break;
1214 	case F_WRLCK:
1215 		linux_flock->l_type = LINUX_F_WRLCK;
1216 		break;
1217 	case F_UNLCK:
1218 		linux_flock->l_type = LINUX_F_UNLCK;
1219 		break;
1220 	}
1221 	linux_flock->l_whence = bsd_flock->l_whence;
1222 	linux_flock->l_start = (l_off_t)bsd_flock->l_start;
1223 	linux_flock->l_len = (l_off_t)bsd_flock->l_len;
1224 	linux_flock->l_pid = (l_pid_t)bsd_flock->l_pid;
1225 }
1226 
1227 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1228 struct l_flock64 {
1229 	l_short		l_type;
1230 	l_short		l_whence;
1231 	l_loff_t	l_start;
1232 	l_loff_t	l_len;
1233 	l_pid_t		l_pid;
1234 }
1235 #if defined(__amd64__) && defined(COMPAT_LINUX32)
1236 __packed
1237 #endif
1238 ;
1239 
1240 static void
1241 linux_to_bsd_flock64(struct l_flock64 *linux_flock, struct flock *bsd_flock)
1242 {
1243 	switch (linux_flock->l_type) {
1244 	case LINUX_F_RDLCK:
1245 		bsd_flock->l_type = F_RDLCK;
1246 		break;
1247 	case LINUX_F_WRLCK:
1248 		bsd_flock->l_type = F_WRLCK;
1249 		break;
1250 	case LINUX_F_UNLCK:
1251 		bsd_flock->l_type = F_UNLCK;
1252 		break;
1253 	default:
1254 		bsd_flock->l_type = -1;
1255 		break;
1256 	}
1257 	bsd_flock->l_whence = linux_flock->l_whence;
1258 	bsd_flock->l_start = (off_t)linux_flock->l_start;
1259 	bsd_flock->l_len = (off_t)linux_flock->l_len;
1260 	bsd_flock->l_pid = (pid_t)linux_flock->l_pid;
1261 	bsd_flock->l_sysid = 0;
1262 }
1263 
1264 static void
1265 bsd_to_linux_flock64(struct flock *bsd_flock, struct l_flock64 *linux_flock)
1266 {
1267 	switch (bsd_flock->l_type) {
1268 	case F_RDLCK:
1269 		linux_flock->l_type = LINUX_F_RDLCK;
1270 		break;
1271 	case F_WRLCK:
1272 		linux_flock->l_type = LINUX_F_WRLCK;
1273 		break;
1274 	case F_UNLCK:
1275 		linux_flock->l_type = LINUX_F_UNLCK;
1276 		break;
1277 	}
1278 	linux_flock->l_whence = bsd_flock->l_whence;
1279 	linux_flock->l_start = (l_loff_t)bsd_flock->l_start;
1280 	linux_flock->l_len = (l_loff_t)bsd_flock->l_len;
1281 	linux_flock->l_pid = (l_pid_t)bsd_flock->l_pid;
1282 }
1283 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1284 
1285 static int
1286 fcntl_common(struct thread *td, struct linux_fcntl64_args *args)
1287 {
1288 	struct l_flock linux_flock;
1289 	struct flock bsd_flock;
1290 	struct file *fp;
1291 	long arg;
1292 	int error, result;
1293 
1294 	switch (args->cmd) {
1295 	case LINUX_F_DUPFD:
1296 		return (kern_fcntl(td, args->fd, F_DUPFD, args->arg));
1297 
1298 	case LINUX_F_GETFD:
1299 		return (kern_fcntl(td, args->fd, F_GETFD, 0));
1300 
1301 	case LINUX_F_SETFD:
1302 		return (kern_fcntl(td, args->fd, F_SETFD, args->arg));
1303 
1304 	case LINUX_F_GETFL:
1305 		error = kern_fcntl(td, args->fd, F_GETFL, 0);
1306 		result = td->td_retval[0];
1307 		td->td_retval[0] = 0;
1308 		if (result & O_RDONLY)
1309 			td->td_retval[0] |= LINUX_O_RDONLY;
1310 		if (result & O_WRONLY)
1311 			td->td_retval[0] |= LINUX_O_WRONLY;
1312 		if (result & O_RDWR)
1313 			td->td_retval[0] |= LINUX_O_RDWR;
1314 		if (result & O_NDELAY)
1315 			td->td_retval[0] |= LINUX_O_NONBLOCK;
1316 		if (result & O_APPEND)
1317 			td->td_retval[0] |= LINUX_O_APPEND;
1318 		if (result & O_FSYNC)
1319 			td->td_retval[0] |= LINUX_O_SYNC;
1320 		if (result & O_ASYNC)
1321 			td->td_retval[0] |= LINUX_FASYNC;
1322 #ifdef LINUX_O_NOFOLLOW
1323 		if (result & O_NOFOLLOW)
1324 			td->td_retval[0] |= LINUX_O_NOFOLLOW;
1325 #endif
1326 #ifdef LINUX_O_DIRECT
1327 		if (result & O_DIRECT)
1328 			td->td_retval[0] |= LINUX_O_DIRECT;
1329 #endif
1330 		return (error);
1331 
1332 	case LINUX_F_SETFL:
1333 		arg = 0;
1334 		if (args->arg & LINUX_O_NDELAY)
1335 			arg |= O_NONBLOCK;
1336 		if (args->arg & LINUX_O_APPEND)
1337 			arg |= O_APPEND;
1338 		if (args->arg & LINUX_O_SYNC)
1339 			arg |= O_FSYNC;
1340 		if (args->arg & LINUX_FASYNC)
1341 			arg |= O_ASYNC;
1342 #ifdef LINUX_O_NOFOLLOW
1343 		if (args->arg & LINUX_O_NOFOLLOW)
1344 			arg |= O_NOFOLLOW;
1345 #endif
1346 #ifdef LINUX_O_DIRECT
1347 		if (args->arg & LINUX_O_DIRECT)
1348 			arg |= O_DIRECT;
1349 #endif
1350 		return (kern_fcntl(td, args->fd, F_SETFL, arg));
1351 
1352 	case LINUX_F_GETLK:
1353 		error = copyin((void *)args->arg, &linux_flock,
1354 		    sizeof(linux_flock));
1355 		if (error)
1356 			return (error);
1357 		linux_to_bsd_flock(&linux_flock, &bsd_flock);
1358 		error = kern_fcntl(td, args->fd, F_GETLK, (intptr_t)&bsd_flock);
1359 		if (error)
1360 			return (error);
1361 		bsd_to_linux_flock(&bsd_flock, &linux_flock);
1362 		return (copyout(&linux_flock, (void *)args->arg,
1363 		    sizeof(linux_flock)));
1364 
1365 	case LINUX_F_SETLK:
1366 		error = copyin((void *)args->arg, &linux_flock,
1367 		    sizeof(linux_flock));
1368 		if (error)
1369 			return (error);
1370 		linux_to_bsd_flock(&linux_flock, &bsd_flock);
1371 		return (kern_fcntl(td, args->fd, F_SETLK,
1372 		    (intptr_t)&bsd_flock));
1373 
1374 	case LINUX_F_SETLKW:
1375 		error = copyin((void *)args->arg, &linux_flock,
1376 		    sizeof(linux_flock));
1377 		if (error)
1378 			return (error);
1379 		linux_to_bsd_flock(&linux_flock, &bsd_flock);
1380 		return (kern_fcntl(td, args->fd, F_SETLKW,
1381 		     (intptr_t)&bsd_flock));
1382 
1383 	case LINUX_F_GETOWN:
1384 		return (kern_fcntl(td, args->fd, F_GETOWN, 0));
1385 
1386 	case LINUX_F_SETOWN:
1387 		/*
1388 		 * XXX some Linux applications depend on F_SETOWN having no
1389 		 * significant effect for pipes (SIGIO is not delivered for
1390 		 * pipes under Linux-2.2.35 at least).
1391 		 */
1392 		error = fget(td, args->fd, CAP_FCNTL, &fp);
1393 		if (error)
1394 			return (error);
1395 		if (fp->f_type == DTYPE_PIPE) {
1396 			fdrop(fp, td);
1397 			return (EINVAL);
1398 		}
1399 		fdrop(fp, td);
1400 
1401 		return (kern_fcntl(td, args->fd, F_SETOWN, args->arg));
1402 	}
1403 
1404 	return (EINVAL);
1405 }
1406 
1407 int
1408 linux_fcntl(struct thread *td, struct linux_fcntl_args *args)
1409 {
1410 	struct linux_fcntl64_args args64;
1411 
1412 #ifdef DEBUG
1413 	if (ldebug(fcntl))
1414 		printf(ARGS(fcntl, "%d, %08x, *"), args->fd, args->cmd);
1415 #endif
1416 
1417 	args64.fd = args->fd;
1418 	args64.cmd = args->cmd;
1419 	args64.arg = args->arg;
1420 	return (fcntl_common(td, &args64));
1421 }
1422 
1423 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1424 int
1425 linux_fcntl64(struct thread *td, struct linux_fcntl64_args *args)
1426 {
1427 	struct l_flock64 linux_flock;
1428 	struct flock bsd_flock;
1429 	int error;
1430 
1431 #ifdef DEBUG
1432 	if (ldebug(fcntl64))
1433 		printf(ARGS(fcntl64, "%d, %08x, *"), args->fd, args->cmd);
1434 #endif
1435 
1436 	switch (args->cmd) {
1437 	case LINUX_F_GETLK64:
1438 		error = copyin((void *)args->arg, &linux_flock,
1439 		    sizeof(linux_flock));
1440 		if (error)
1441 			return (error);
1442 		linux_to_bsd_flock64(&linux_flock, &bsd_flock);
1443 		error = kern_fcntl(td, args->fd, F_GETLK, (intptr_t)&bsd_flock);
1444 		if (error)
1445 			return (error);
1446 		bsd_to_linux_flock64(&bsd_flock, &linux_flock);
1447 		return (copyout(&linux_flock, (void *)args->arg,
1448 			    sizeof(linux_flock)));
1449 
1450 	case LINUX_F_SETLK64:
1451 		error = copyin((void *)args->arg, &linux_flock,
1452 		    sizeof(linux_flock));
1453 		if (error)
1454 			return (error);
1455 		linux_to_bsd_flock64(&linux_flock, &bsd_flock);
1456 		return (kern_fcntl(td, args->fd, F_SETLK,
1457 		    (intptr_t)&bsd_flock));
1458 
1459 	case LINUX_F_SETLKW64:
1460 		error = copyin((void *)args->arg, &linux_flock,
1461 		    sizeof(linux_flock));
1462 		if (error)
1463 			return (error);
1464 		linux_to_bsd_flock64(&linux_flock, &bsd_flock);
1465 		return (kern_fcntl(td, args->fd, F_SETLKW,
1466 		    (intptr_t)&bsd_flock));
1467 	}
1468 
1469 	return (fcntl_common(td, args));
1470 }
1471 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1472 
1473 int
1474 linux_chown(struct thread *td, struct linux_chown_args *args)
1475 {
1476 	char *path;
1477 	int error;
1478 
1479 	LCONVPATHEXIST(td, args->path, &path);
1480 
1481 #ifdef DEBUG
1482 	if (ldebug(chown))
1483 		printf(ARGS(chown, "%s, %d, %d"), path, args->uid, args->gid);
1484 #endif
1485 	error = kern_chown(td, path, UIO_SYSSPACE, args->uid, args->gid);
1486 	LFREEPATH(path);
1487 	return (error);
1488 }
1489 
1490 int
1491 linux_fchownat(struct thread *td, struct linux_fchownat_args *args)
1492 {
1493 	char *path;
1494 	int error, dfd, flag;
1495 
1496 	if (args->flag & ~LINUX_AT_SYMLINK_NOFOLLOW)
1497 		return (EINVAL);
1498 
1499 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD :  args->dfd;
1500 	LCONVPATHEXIST_AT(td, args->filename, &path, dfd);
1501 
1502 #ifdef DEBUG
1503 	if (ldebug(fchownat))
1504 		printf(ARGS(fchownat, "%s, %d, %d"), path, args->uid, args->gid);
1505 #endif
1506 
1507 	flag = (args->flag & LINUX_AT_SYMLINK_NOFOLLOW) == 0 ? 0 :
1508 	    AT_SYMLINK_NOFOLLOW;
1509 	error = kern_fchownat(td, dfd, path, UIO_SYSSPACE, args->uid, args->gid,
1510 	    flag);
1511 	LFREEPATH(path);
1512 	return (error);
1513 }
1514 
1515 int
1516 linux_lchown(struct thread *td, struct linux_lchown_args *args)
1517 {
1518 	char *path;
1519 	int error;
1520 
1521 	LCONVPATHEXIST(td, args->path, &path);
1522 
1523 #ifdef DEBUG
1524 	if (ldebug(lchown))
1525 		printf(ARGS(lchown, "%s, %d, %d"), path, args->uid, args->gid);
1526 #endif
1527 	error = kern_lchown(td, path, UIO_SYSSPACE, args->uid, args->gid);
1528 	LFREEPATH(path);
1529 	return (error);
1530 }
1531 
1532 static int
1533 convert_fadvice(int advice)
1534 {
1535 	switch (advice) {
1536 	case LINUX_POSIX_FADV_NORMAL:
1537 		return (POSIX_FADV_NORMAL);
1538 	case LINUX_POSIX_FADV_RANDOM:
1539 		return (POSIX_FADV_RANDOM);
1540 	case LINUX_POSIX_FADV_SEQUENTIAL:
1541 		return (POSIX_FADV_SEQUENTIAL);
1542 	case LINUX_POSIX_FADV_WILLNEED:
1543 		return (POSIX_FADV_WILLNEED);
1544 	case LINUX_POSIX_FADV_DONTNEED:
1545 		return (POSIX_FADV_DONTNEED);
1546 	case LINUX_POSIX_FADV_NOREUSE:
1547 		return (POSIX_FADV_NOREUSE);
1548 	default:
1549 		return (-1);
1550 	}
1551 }
1552 
1553 int
1554 linux_fadvise64(struct thread *td, struct linux_fadvise64_args *args)
1555 {
1556 	int advice;
1557 
1558 	advice = convert_fadvice(args->advice);
1559 	if (advice == -1)
1560 		return (EINVAL);
1561 	return (kern_posix_fadvise(td, args->fd, args->offset, args->len,
1562 	    advice));
1563 }
1564 
1565 int
1566 linux_fadvise64_64(struct thread *td, struct linux_fadvise64_64_args *args)
1567 {
1568 	int advice;
1569 
1570 	advice = convert_fadvice(args->advice);
1571 	if (advice == -1)
1572 		return (EINVAL);
1573 	return (kern_posix_fadvise(td, args->fd, args->offset, args->len,
1574 	    advice));
1575 }
1576 
1577 int
1578 linux_pipe(struct thread *td, struct linux_pipe_args *args)
1579 {
1580 	int fildes[2];
1581 	int error;
1582 
1583 #ifdef DEBUG
1584 	if (ldebug(pipe))
1585 		printf(ARGS(pipe, "*"));
1586 #endif
1587 
1588 	error = do_pipe(td, fildes, 0);
1589 	if (error)
1590 		return (error);
1591 
1592 	/* XXX: Close descriptors on error. */
1593 	return (copyout(fildes, args->pipefds, sizeof(fildes)));
1594 }
1595 
1596 int
1597 linux_pipe2(struct thread *td, struct linux_pipe2_args *args)
1598 {
1599 	int fildes[2];
1600 	int error, flags;
1601 
1602 #ifdef DEBUG
1603 	if (ldebug(pipe2))
1604 		printf(ARGS(pipe2, "*, %d"), args->flags);
1605 #endif
1606 
1607 	if ((args->flags & ~(LINUX_O_NONBLOCK | LINUX_O_CLOEXEC)) != 0)
1608 		return (EINVAL);
1609 
1610 	flags = 0;
1611 	if ((args->flags & LINUX_O_NONBLOCK) != 0)
1612 		flags |= O_NONBLOCK;
1613 	if ((args->flags & LINUX_O_CLOEXEC) != 0)
1614 		flags |= O_CLOEXEC;
1615 	error = do_pipe(td, fildes, flags);
1616 	if (error)
1617 		return (error);
1618 
1619 	/* XXX: Close descriptors on error. */
1620 	return (copyout(fildes, args->pipefds, sizeof(fildes)));
1621 }
1622