xref: /freebsd/sys/compat/linux/linux_file.c (revision 69f1cb3c91c3377cedc28a9fe37673bda10602cd)
1 /*-
2  * Copyright (c) 1994-1995 S�ren 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 #include "opt_mac.h"
34 
35 #include <sys/param.h>
36 #include <sys/systm.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/mac.h>
44 #include <sys/malloc.h>
45 #include <sys/mount.h>
46 #include <sys/mutex.h>
47 #include <sys/proc.h>
48 #include <sys/syscallsubr.h>
49 #include <sys/sysproto.h>
50 #include <sys/tty.h>
51 #include <sys/vnode.h>
52 
53 #include <ufs/ufs/extattr.h>
54 #include <ufs/ufs/quota.h>
55 #include <ufs/ufs/ufsmount.h>
56 
57 #ifdef COMPAT_LINUX32
58 #include <machine/../linux32/linux.h>
59 #include <machine/../linux32/linux32_proto.h>
60 #else
61 #include <machine/../linux/linux.h>
62 #include <machine/../linux/linux_proto.h>
63 #endif
64 #include <compat/linux/linux_util.h>
65 
66 int
67 linux_creat(struct thread *td, struct linux_creat_args *args)
68 {
69     char *path;
70     int error;
71 
72     LCONVPATHEXIST(td, args->path, &path);
73 
74 #ifdef DEBUG
75 	if (ldebug(creat))
76 		printf(ARGS(creat, "%s, %d"), path, args->mode);
77 #endif
78     error = kern_open(td, path, UIO_SYSSPACE, O_WRONLY | O_CREAT | O_TRUNC,
79 	args->mode);
80     LFREEPATH(path);
81     return (error);
82 }
83 
84 int
85 linux_open(struct thread *td, struct linux_open_args *args)
86 {
87     struct proc *p = td->td_proc;
88     char *path;
89     int bsd_flags, error;
90 
91     if (args->flags & LINUX_O_CREAT)
92 	LCONVPATHCREAT(td, args->path, &path);
93     else
94 	LCONVPATHEXIST(td, args->path, &path);
95 
96 #ifdef DEBUG
97 	if (ldebug(open))
98 		printf(ARGS(open, "%s, 0x%x, 0x%x"),
99 		    path, args->flags, args->mode);
100 #endif
101     bsd_flags = 0;
102     if (args->flags & LINUX_O_RDONLY)
103 	bsd_flags |= O_RDONLY;
104     if (args->flags & LINUX_O_WRONLY)
105 	bsd_flags |= O_WRONLY;
106     if (args->flags & LINUX_O_RDWR)
107 	bsd_flags |= O_RDWR;
108     if (args->flags & LINUX_O_NDELAY)
109 	bsd_flags |= O_NONBLOCK;
110     if (args->flags & LINUX_O_APPEND)
111 	bsd_flags |= O_APPEND;
112     if (args->flags & LINUX_O_SYNC)
113 	bsd_flags |= O_FSYNC;
114     if (args->flags & LINUX_O_NONBLOCK)
115 	bsd_flags |= O_NONBLOCK;
116     if (args->flags & LINUX_FASYNC)
117 	bsd_flags |= O_ASYNC;
118     if (args->flags & LINUX_O_CREAT)
119 	bsd_flags |= O_CREAT;
120     if (args->flags & LINUX_O_TRUNC)
121 	bsd_flags |= O_TRUNC;
122     if (args->flags & LINUX_O_EXCL)
123 	bsd_flags |= O_EXCL;
124     if (args->flags & LINUX_O_NOCTTY)
125 	bsd_flags |= O_NOCTTY;
126 
127     error = kern_open(td, path, UIO_SYSSPACE, bsd_flags, args->mode);
128     PROC_LOCK(p);
129     if (!error && !(bsd_flags & O_NOCTTY) &&
130 	SESS_LEADER(p) && !(p->p_flag & P_CONTROLT)) {
131 	struct file *fp;
132 
133 	PROC_UNLOCK(p);
134 	error = fget(td, td->td_retval[0], &fp);
135 	if (!error) {
136 		if (fp->f_type == DTYPE_VNODE)
137 			fo_ioctl(fp, TIOCSCTTY, (caddr_t) 0, td->td_ucred,
138 			    td);
139 	    fdrop(fp, td);
140 	}
141     } else {
142 	PROC_UNLOCK(p);
143 #ifdef DEBUG
144 	if (ldebug(open))
145 		printf(LMSG("open returns error %d"), error);
146 #endif
147     }
148     LFREEPATH(path);
149     return error;
150 }
151 
152 int
153 linux_lseek(struct thread *td, struct linux_lseek_args *args)
154 {
155 
156     struct lseek_args /* {
157 	int fd;
158 	int pad;
159 	off_t offset;
160 	int whence;
161     } */ tmp_args;
162     int error;
163 
164 #ifdef DEBUG
165 	if (ldebug(lseek))
166 		printf(ARGS(lseek, "%d, %ld, %d"),
167 		    args->fdes, (long)args->off, args->whence);
168 #endif
169     tmp_args.fd = args->fdes;
170     tmp_args.offset = (off_t)args->off;
171     tmp_args.whence = args->whence;
172     error = lseek(td, &tmp_args);
173     return error;
174 }
175 
176 int
177 linux_llseek(struct thread *td, struct linux_llseek_args *args)
178 {
179 	struct lseek_args bsd_args;
180 	int error;
181 	off_t off;
182 
183 #ifdef DEBUG
184 	if (ldebug(llseek))
185 		printf(ARGS(llseek, "%d, %d:%d, %d"),
186 		    args->fd, args->ohigh, args->olow, args->whence);
187 #endif
188 	off = (args->olow) | (((off_t) args->ohigh) << 32);
189 
190 	bsd_args.fd = args->fd;
191 	bsd_args.offset = off;
192 	bsd_args.whence = args->whence;
193 
194 	if ((error = lseek(td, &bsd_args)))
195 		return error;
196 
197 	if ((error = copyout(td->td_retval, args->res, sizeof (off_t))))
198 		return error;
199 
200 	td->td_retval[0] = 0;
201 	return 0;
202 }
203 
204 int
205 linux_readdir(struct thread *td, struct linux_readdir_args *args)
206 {
207 	struct linux_getdents_args lda;
208 
209 	lda.fd = args->fd;
210 	lda.dent = args->dent;
211 	lda.count = 1;
212 	return linux_getdents(td, &lda);
213 }
214 
215 /*
216  * Note that linux_getdents(2) and linux_getdents64(2) have the same
217  * arguments. They only differ in the definition of struct dirent they
218  * operate on. We use this to common the code, with the exception of
219  * accessing struct dirent. Note that linux_readdir(2) is implemented
220  * by means of linux_getdents(2). In this case we never operate on
221  * struct dirent64 and thus don't need to handle it...
222  */
223 
224 struct l_dirent {
225 	l_long		d_ino;
226 	l_off_t		d_off;
227 	l_ushort	d_reclen;
228 	char		d_name[LINUX_NAME_MAX + 1];
229 };
230 
231 struct l_dirent64 {
232 	uint64_t	d_ino;
233 	int64_t		d_off;
234 	l_ushort	d_reclen;
235 	u_char		d_type;
236 	char		d_name[LINUX_NAME_MAX + 1];
237 };
238 
239 #define LINUX_RECLEN(de,namlen) \
240     ALIGN((((char *)&(de)->d_name - (char *)de) + (namlen) + 1))
241 
242 #define	LINUX_DIRBLKSIZ		512
243 
244 static int
245 getdents_common(struct thread *td, struct linux_getdents64_args *args,
246     int is64bit)
247 {
248 	struct dirent *bdp;
249 	struct vnode *vp;
250 	caddr_t inp, buf;		/* BSD-format */
251 	int len, reclen;		/* BSD-format */
252 	caddr_t outp;			/* Linux-format */
253 	int resid, linuxreclen=0;	/* Linux-format */
254 	struct file *fp;
255 	struct uio auio;
256 	struct iovec aiov;
257 	off_t off;
258 	struct l_dirent linux_dirent;
259 	struct l_dirent64 linux_dirent64;
260 	int buflen, error, eofflag, nbytes, justone;
261 	u_long *cookies = NULL, *cookiep;
262 	int ncookies, vfslocked;
263 
264 	nbytes = args->count;
265 	if (nbytes == 1) {
266 		/* readdir(2) case. Always struct dirent. */
267 		if (is64bit)
268 			return (EINVAL);
269 		nbytes = sizeof(linux_dirent);
270 		justone = 1;
271 	} else
272 		justone = 0;
273 
274 	if ((error = getvnode(td->td_proc->p_fd, args->fd, &fp)) != 0)
275 		return (error);
276 
277 	if ((fp->f_flag & FREAD) == 0) {
278 		fdrop(fp, td);
279 		return (EBADF);
280 	}
281 
282 	vp = fp->f_vnode;
283 	vfslocked = VFS_LOCK_GIANT(vp->v_mount);
284 	if (vp->v_type != VDIR) {
285 		VFS_UNLOCK_GIANT(vfslocked);
286 		fdrop(fp, td);
287 		return (EINVAL);
288 	}
289 
290 	off = fp->f_offset;
291 
292 	buflen = max(LINUX_DIRBLKSIZ, nbytes);
293 	buflen = min(buflen, MAXBSIZE);
294 	buf = malloc(buflen, M_TEMP, M_WAITOK);
295 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
296 
297 again:
298 	aiov.iov_base = buf;
299 	aiov.iov_len = buflen;
300 	auio.uio_iov = &aiov;
301 	auio.uio_iovcnt = 1;
302 	auio.uio_rw = UIO_READ;
303 	auio.uio_segflg = UIO_SYSSPACE;
304 	auio.uio_td = td;
305 	auio.uio_resid = buflen;
306 	auio.uio_offset = off;
307 
308 	if (cookies) {
309 		free(cookies, M_TEMP);
310 		cookies = NULL;
311 	}
312 
313 #ifdef MAC
314 	/*
315 	 * Do directory search MAC check using non-cached credentials.
316 	 */
317 	if ((error = mac_check_vnode_readdir(td->td_ucred, vp)))
318 		goto out;
319 #endif /* MAC */
320 	if ((error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, &ncookies,
321 		 &cookies)))
322 		goto out;
323 
324 	inp = buf;
325 	outp = (caddr_t)args->dirent;
326 	resid = nbytes;
327 	if ((len = buflen - auio.uio_resid) <= 0)
328 		goto eof;
329 
330 	cookiep = cookies;
331 
332 	if (cookies) {
333 		/*
334 		 * When using cookies, the vfs has the option of reading from
335 		 * a different offset than that supplied (UFS truncates the
336 		 * offset to a block boundary to make sure that it never reads
337 		 * partway through a directory entry, even if the directory
338 		 * has been compacted).
339 		 */
340 		while (len > 0 && ncookies > 0 && *cookiep <= off) {
341 			bdp = (struct dirent *) inp;
342 			len -= bdp->d_reclen;
343 			inp += bdp->d_reclen;
344 			cookiep++;
345 			ncookies--;
346 		}
347 	}
348 
349 	while (len > 0) {
350 		if (cookiep && ncookies == 0)
351 			break;
352 		bdp = (struct dirent *) inp;
353 		reclen = bdp->d_reclen;
354 		if (reclen & 3) {
355 			error = EFAULT;
356 			goto out;
357 		}
358 
359 		if (bdp->d_fileno == 0) {
360 			inp += reclen;
361 			if (cookiep) {
362 				off = *cookiep++;
363 				ncookies--;
364 			} else
365 				off += reclen;
366 
367 			len -= reclen;
368 			continue;
369 		}
370 
371 		linuxreclen = (is64bit)
372 		    ? LINUX_RECLEN(&linux_dirent64, bdp->d_namlen)
373 		    : LINUX_RECLEN(&linux_dirent, bdp->d_namlen);
374 
375 		if (reclen > len || resid < linuxreclen) {
376 			outp++;
377 			break;
378 		}
379 
380 		if (justone) {
381 			/* readdir(2) case. */
382 			linux_dirent.d_ino = (l_long)bdp->d_fileno;
383 			linux_dirent.d_off = (l_off_t)linuxreclen;
384 			linux_dirent.d_reclen = (l_ushort)bdp->d_namlen;
385 			strcpy(linux_dirent.d_name, bdp->d_name);
386 			error = copyout(&linux_dirent, outp, linuxreclen);
387 		} else {
388 			if (is64bit) {
389 				linux_dirent64.d_ino = bdp->d_fileno;
390 				linux_dirent64.d_off = (cookiep)
391 				    ? (l_off_t)*cookiep
392 				    : (l_off_t)(off + reclen);
393 				linux_dirent64.d_reclen =
394 				    (l_ushort)linuxreclen;
395 				linux_dirent64.d_type = bdp->d_type;
396 				strcpy(linux_dirent64.d_name, bdp->d_name);
397 				error = copyout(&linux_dirent64, outp,
398 				    linuxreclen);
399 			} else {
400 				linux_dirent.d_ino = bdp->d_fileno;
401 				linux_dirent.d_off = (cookiep)
402 				    ? (l_off_t)*cookiep
403 				    : (l_off_t)(off + reclen);
404 				linux_dirent.d_reclen = (l_ushort)linuxreclen;
405 				strcpy(linux_dirent.d_name, bdp->d_name);
406 				error = copyout(&linux_dirent, outp,
407 				    linuxreclen);
408 			}
409 		}
410 		if (error)
411 			goto out;
412 
413 		inp += reclen;
414 		if (cookiep) {
415 			off = *cookiep++;
416 			ncookies--;
417 		} else
418 			off += reclen;
419 
420 		outp += linuxreclen;
421 		resid -= linuxreclen;
422 		len -= reclen;
423 		if (justone)
424 			break;
425 	}
426 
427 	if (outp == (caddr_t)args->dirent)
428 		goto again;
429 
430 	fp->f_offset = off;
431 	if (justone)
432 		nbytes = resid + linuxreclen;
433 
434 eof:
435 	td->td_retval[0] = nbytes - resid;
436 
437 out:
438 	if (cookies)
439 		free(cookies, M_TEMP);
440 
441 	VOP_UNLOCK(vp, 0, td);
442 	VFS_UNLOCK_GIANT(vfslocked);
443 	fdrop(fp, td);
444 	free(buf, M_TEMP);
445 	return (error);
446 }
447 
448 int
449 linux_getdents(struct thread *td, struct linux_getdents_args *args)
450 {
451 
452 #ifdef DEBUG
453 	if (ldebug(getdents))
454 		printf(ARGS(getdents, "%d, *, %d"), args->fd, args->count);
455 #endif
456 
457 	return (getdents_common(td, (struct linux_getdents64_args*)args, 0));
458 }
459 
460 int
461 linux_getdents64(struct thread *td, struct linux_getdents64_args *args)
462 {
463 
464 #ifdef DEBUG
465 	if (ldebug(getdents64))
466 		printf(ARGS(getdents64, "%d, *, %d"), args->fd, args->count);
467 #endif
468 
469 	return (getdents_common(td, args, 1));
470 }
471 
472 /*
473  * These exist mainly for hooks for doing /compat/linux translation.
474  */
475 
476 int
477 linux_access(struct thread *td, struct linux_access_args *args)
478 {
479 	char *path;
480 	int error;
481 
482 	LCONVPATHEXIST(td, args->path, &path);
483 
484 #ifdef DEBUG
485 	if (ldebug(access))
486 		printf(ARGS(access, "%s, %d"), path, args->flags);
487 #endif
488 	error = kern_access(td, path, UIO_SYSSPACE, args->flags);
489 	LFREEPATH(path);
490 	return (error);
491 }
492 
493 int
494 linux_unlink(struct thread *td, struct linux_unlink_args *args)
495 {
496 	char *path;
497 	int error;
498 
499 	LCONVPATHEXIST(td, args->path, &path);
500 
501 #ifdef DEBUG
502 	if (ldebug(unlink))
503 		printf(ARGS(unlink, "%s"), path);
504 #endif
505 
506 	error = kern_unlink(td, path, UIO_SYSSPACE);
507 	LFREEPATH(path);
508 	return (error);
509 }
510 
511 int
512 linux_chdir(struct thread *td, struct linux_chdir_args *args)
513 {
514 	char *path;
515 	int error;
516 
517 	LCONVPATHEXIST(td, args->path, &path);
518 
519 #ifdef DEBUG
520 	if (ldebug(chdir))
521 		printf(ARGS(chdir, "%s"), path);
522 #endif
523 	error = kern_chdir(td, path, UIO_SYSSPACE);
524 	LFREEPATH(path);
525 	return (error);
526 }
527 
528 int
529 linux_chmod(struct thread *td, struct linux_chmod_args *args)
530 {
531 	char *path;
532 	int error;
533 
534 	LCONVPATHEXIST(td, args->path, &path);
535 
536 #ifdef DEBUG
537 	if (ldebug(chmod))
538 		printf(ARGS(chmod, "%s, %d"), path, args->mode);
539 #endif
540 	error = kern_chmod(td, path, UIO_SYSSPACE, args->mode);
541 	LFREEPATH(path);
542 	return (error);
543 }
544 
545 int
546 linux_mkdir(struct thread *td, struct linux_mkdir_args *args)
547 {
548 	char *path;
549 	int error;
550 
551 	LCONVPATHCREAT(td, args->path, &path);
552 
553 #ifdef DEBUG
554 	if (ldebug(mkdir))
555 		printf(ARGS(mkdir, "%s, %d"), path, args->mode);
556 #endif
557 	error = kern_mkdir(td, path, UIO_SYSSPACE, args->mode);
558 	LFREEPATH(path);
559 	return (error);
560 }
561 
562 int
563 linux_rmdir(struct thread *td, struct linux_rmdir_args *args)
564 {
565 	char *path;
566 	int error;
567 
568 	LCONVPATHEXIST(td, args->path, &path);
569 
570 #ifdef DEBUG
571 	if (ldebug(rmdir))
572 		printf(ARGS(rmdir, "%s"), path);
573 #endif
574 	error = kern_rmdir(td, path, UIO_SYSSPACE);
575 	LFREEPATH(path);
576 	return (error);
577 }
578 
579 int
580 linux_rename(struct thread *td, struct linux_rename_args *args)
581 {
582 	char *from, *to;
583 	int error;
584 
585 	LCONVPATHEXIST(td, args->from, &from);
586 	/* Expand LCONVPATHCREATE so that `from' can be freed on errors */
587 	error = linux_emul_convpath(td, args->to, UIO_USERSPACE, &to, 1);
588 	if (to == NULL) {
589 		LFREEPATH(from);
590 		return (error);
591 	}
592 
593 #ifdef DEBUG
594 	if (ldebug(rename))
595 		printf(ARGS(rename, "%s, %s"), from, to);
596 #endif
597 	error = kern_rename(td, from, to, UIO_SYSSPACE);
598 	LFREEPATH(from);
599 	LFREEPATH(to);
600 	return (error);
601 }
602 
603 int
604 linux_symlink(struct thread *td, struct linux_symlink_args *args)
605 {
606 	char *path, *to;
607 	int error;
608 
609 	LCONVPATHEXIST(td, args->path, &path);
610 	/* Expand LCONVPATHCREATE so that `path' can be freed on errors */
611 	error = linux_emul_convpath(td, args->to, UIO_USERSPACE, &to, 1);
612 	if (to == NULL) {
613 		LFREEPATH(path);
614 		return (error);
615 	}
616 
617 #ifdef DEBUG
618 	if (ldebug(symlink))
619 		printf(ARGS(symlink, "%s, %s"), path, to);
620 #endif
621 	error = kern_symlink(td, path, to, UIO_SYSSPACE);
622 	LFREEPATH(path);
623 	LFREEPATH(to);
624 	return (error);
625 }
626 
627 int
628 linux_readlink(struct thread *td, struct linux_readlink_args *args)
629 {
630 	char *name;
631 	int error;
632 
633 	LCONVPATHEXIST(td, args->name, &name);
634 
635 #ifdef DEBUG
636 	if (ldebug(readlink))
637 		printf(ARGS(readlink, "%s, %p, %d"), name, (void *)args->buf,
638 		    args->count);
639 #endif
640 	error = kern_readlink(td, name, UIO_SYSSPACE, args->buf, UIO_USERSPACE,
641 	    args->count);
642 	LFREEPATH(name);
643 	return (error);
644 }
645 
646 int
647 linux_truncate(struct thread *td, struct linux_truncate_args *args)
648 {
649 	char *path;
650 	int error;
651 
652 	LCONVPATHEXIST(td, args->path, &path);
653 
654 #ifdef DEBUG
655 	if (ldebug(truncate))
656 		printf(ARGS(truncate, "%s, %ld"), path, (long)args->length);
657 #endif
658 
659 	error = kern_truncate(td, path, UIO_SYSSPACE, args->length);
660 	LFREEPATH(path);
661 	return (error);
662 }
663 
664 int
665 linux_ftruncate(struct thread *td, struct linux_ftruncate_args *args)
666 {
667 	struct ftruncate_args /* {
668 		int fd;
669 		int pad;
670 		off_t length;
671 		} */ nuap;
672 
673 	nuap.fd = args->fd;
674 	nuap.pad = 0;
675 	nuap.length = args->length;
676 	return (ftruncate(td, &nuap));
677 }
678 
679 int
680 linux_link(struct thread *td, struct linux_link_args *args)
681 {
682 	char *path, *to;
683 	int error;
684 
685 	LCONVPATHEXIST(td, args->path, &path);
686 	/* Expand LCONVPATHCREATE so that `path' can be freed on errors */
687 	error = linux_emul_convpath(td, args->to, UIO_USERSPACE, &to, 1);
688 	if (to == NULL) {
689 		LFREEPATH(path);
690 		return (error);
691 	}
692 
693 #ifdef DEBUG
694 	if (ldebug(link))
695 		printf(ARGS(link, "%s, %s"), path, to);
696 #endif
697 	error = kern_link(td, path, to, UIO_SYSSPACE);
698 	LFREEPATH(path);
699 	LFREEPATH(to);
700 	return (error);
701 }
702 
703 int
704 linux_fdatasync(td, uap)
705 	struct thread *td;
706 	struct linux_fdatasync_args *uap;
707 {
708 	struct fsync_args bsd;
709 
710 	bsd.fd = uap->fd;
711 	return fsync(td, &bsd);
712 }
713 
714 int
715 linux_pread(td, uap)
716 	struct thread *td;
717 	struct linux_pread_args *uap;
718 {
719 	struct pread_args bsd;
720 
721 	bsd.fd = uap->fd;
722 	bsd.buf = uap->buf;
723 	bsd.nbyte = uap->nbyte;
724 	bsd.offset = uap->offset;
725 	return pread(td, &bsd);
726 }
727 
728 int
729 linux_pwrite(td, uap)
730 	struct thread *td;
731 	struct linux_pwrite_args *uap;
732 {
733 	struct pwrite_args bsd;
734 
735 	bsd.fd = uap->fd;
736 	bsd.buf = uap->buf;
737 	bsd.nbyte = uap->nbyte;
738 	bsd.offset = uap->offset;
739 	return pwrite(td, &bsd);
740 }
741 
742 int
743 linux_mount(struct thread *td, struct linux_mount_args *args)
744 {
745 	struct ufs_args ufs;
746 	char fstypename[MFSNAMELEN];
747 	char mntonname[MNAMELEN], mntfromname[MNAMELEN];
748 	int error;
749 	int fsflags;
750 	void *fsdata;
751 
752 	error = copyinstr(args->filesystemtype, fstypename, MFSNAMELEN - 1,
753 	    NULL);
754 	if (error)
755 		return (error);
756 	error = copyinstr(args->specialfile, mntfromname, MNAMELEN - 1, NULL);
757 	if (error)
758 		return (error);
759 	error = copyinstr(args->dir, mntonname, MNAMELEN - 1, NULL);
760 	if (error)
761 		return (error);
762 
763 #ifdef DEBUG
764 	if (ldebug(mount))
765 		printf(ARGS(mount, "%s, %s, %s"),
766 		    fstypename, mntfromname, mntonname);
767 #endif
768 
769 	if (strcmp(fstypename, "ext2") == 0) {
770 		strcpy(fstypename, "ext2fs");
771 		fsdata = &ufs;
772 		ufs.fspec = mntfromname;
773 #define DEFAULT_ROOTID		-2
774 		ufs.export.ex_root = DEFAULT_ROOTID;
775 		ufs.export.ex_flags =
776 		    args->rwflag & LINUX_MS_RDONLY ? MNT_EXRDONLY : 0;
777 	} else if (strcmp(fstypename, "proc") == 0) {
778 		strcpy(fstypename, "linprocfs");
779 		fsdata = NULL;
780 	} else {
781 		return (ENODEV);
782 	}
783 
784 	fsflags = 0;
785 
786 	if ((args->rwflag & 0xffff0000) == 0xc0ed0000) {
787 		/*
788 		 * Linux SYNC flag is not included; the closest equivalent
789 		 * FreeBSD has is !ASYNC, which is our default.
790 		 */
791 		if (args->rwflag & LINUX_MS_RDONLY)
792 			fsflags |= MNT_RDONLY;
793 		if (args->rwflag & LINUX_MS_NOSUID)
794 			fsflags |= MNT_NOSUID;
795 		if (args->rwflag & LINUX_MS_NOEXEC)
796 			fsflags |= MNT_NOEXEC;
797 		if (args->rwflag & LINUX_MS_REMOUNT)
798 			fsflags |= MNT_UPDATE;
799 	}
800 
801 	if (strcmp(fstypename, "linprocfs") == 0) {
802 		error = kernel_vmount(fsflags,
803 			"fstype", fstypename,
804 			"fspath", mntonname,
805 			NULL);
806 	} else
807 		error = EOPNOTSUPP;
808 	return (error);
809 }
810 
811 int
812 linux_oldumount(struct thread *td, struct linux_oldumount_args *args)
813 {
814 	struct linux_umount_args args2;
815 
816 	args2.path = args->path;
817 	args2.flags = 0;
818 	return (linux_umount(td, &args2));
819 }
820 
821 int
822 linux_umount(struct thread *td, struct linux_umount_args *args)
823 {
824 	struct unmount_args bsd;
825 
826 	bsd.path = args->path;
827 	bsd.flags = args->flags;	/* XXX correct? */
828 	return (unmount(td, &bsd));
829 }
830 
831 /*
832  * fcntl family of syscalls
833  */
834 
835 struct l_flock {
836 	l_short		l_type;
837 	l_short		l_whence;
838 	l_off_t		l_start;
839 	l_off_t		l_len;
840 	l_pid_t		l_pid;
841 }
842 #if defined(__amd64__) && defined(COMPAT_LINUX32)
843 __packed
844 #endif
845 ;
846 
847 static void
848 linux_to_bsd_flock(struct l_flock *linux_flock, struct flock *bsd_flock)
849 {
850 	switch (linux_flock->l_type) {
851 	case LINUX_F_RDLCK:
852 		bsd_flock->l_type = F_RDLCK;
853 		break;
854 	case LINUX_F_WRLCK:
855 		bsd_flock->l_type = F_WRLCK;
856 		break;
857 	case LINUX_F_UNLCK:
858 		bsd_flock->l_type = F_UNLCK;
859 		break;
860 	default:
861 		bsd_flock->l_type = -1;
862 		break;
863 	}
864 	bsd_flock->l_whence = linux_flock->l_whence;
865 	bsd_flock->l_start = (off_t)linux_flock->l_start;
866 	bsd_flock->l_len = (off_t)linux_flock->l_len;
867 	bsd_flock->l_pid = (pid_t)linux_flock->l_pid;
868 }
869 
870 static void
871 bsd_to_linux_flock(struct flock *bsd_flock, struct l_flock *linux_flock)
872 {
873 	switch (bsd_flock->l_type) {
874 	case F_RDLCK:
875 		linux_flock->l_type = LINUX_F_RDLCK;
876 		break;
877 	case F_WRLCK:
878 		linux_flock->l_type = LINUX_F_WRLCK;
879 		break;
880 	case F_UNLCK:
881 		linux_flock->l_type = LINUX_F_UNLCK;
882 		break;
883 	}
884 	linux_flock->l_whence = bsd_flock->l_whence;
885 	linux_flock->l_start = (l_off_t)bsd_flock->l_start;
886 	linux_flock->l_len = (l_off_t)bsd_flock->l_len;
887 	linux_flock->l_pid = (l_pid_t)bsd_flock->l_pid;
888 }
889 
890 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
891 struct l_flock64 {
892 	l_short		l_type;
893 	l_short		l_whence;
894 	l_loff_t	l_start;
895 	l_loff_t	l_len;
896 	l_pid_t		l_pid;
897 }
898 #if defined(__amd64__) && defined(COMPAT_LINUX32)
899 __packed
900 #endif
901 ;
902 
903 static void
904 linux_to_bsd_flock64(struct l_flock64 *linux_flock, struct flock *bsd_flock)
905 {
906 	switch (linux_flock->l_type) {
907 	case LINUX_F_RDLCK:
908 		bsd_flock->l_type = F_RDLCK;
909 		break;
910 	case LINUX_F_WRLCK:
911 		bsd_flock->l_type = F_WRLCK;
912 		break;
913 	case LINUX_F_UNLCK:
914 		bsd_flock->l_type = F_UNLCK;
915 		break;
916 	default:
917 		bsd_flock->l_type = -1;
918 		break;
919 	}
920 	bsd_flock->l_whence = linux_flock->l_whence;
921 	bsd_flock->l_start = (off_t)linux_flock->l_start;
922 	bsd_flock->l_len = (off_t)linux_flock->l_len;
923 	bsd_flock->l_pid = (pid_t)linux_flock->l_pid;
924 }
925 
926 static void
927 bsd_to_linux_flock64(struct flock *bsd_flock, struct l_flock64 *linux_flock)
928 {
929 	switch (bsd_flock->l_type) {
930 	case F_RDLCK:
931 		linux_flock->l_type = LINUX_F_RDLCK;
932 		break;
933 	case F_WRLCK:
934 		linux_flock->l_type = LINUX_F_WRLCK;
935 		break;
936 	case F_UNLCK:
937 		linux_flock->l_type = LINUX_F_UNLCK;
938 		break;
939 	}
940 	linux_flock->l_whence = bsd_flock->l_whence;
941 	linux_flock->l_start = (l_loff_t)bsd_flock->l_start;
942 	linux_flock->l_len = (l_loff_t)bsd_flock->l_len;
943 	linux_flock->l_pid = (l_pid_t)bsd_flock->l_pid;
944 }
945 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
946 
947 static int
948 fcntl_common(struct thread *td, struct linux_fcntl64_args *args)
949 {
950 	struct l_flock linux_flock;
951 	struct flock bsd_flock;
952 	struct file *fp;
953 	long arg;
954 	int error, result;
955 
956 	switch (args->cmd) {
957 	case LINUX_F_DUPFD:
958 		return (kern_fcntl(td, args->fd, F_DUPFD, args->arg));
959 
960 	case LINUX_F_GETFD:
961 		return (kern_fcntl(td, args->fd, F_GETFD, 0));
962 
963 	case LINUX_F_SETFD:
964 		return (kern_fcntl(td, args->fd, F_SETFD, args->arg));
965 
966 	case LINUX_F_GETFL:
967 		error = kern_fcntl(td, args->fd, F_GETFL, 0);
968 		result = td->td_retval[0];
969 		td->td_retval[0] = 0;
970 		if (result & O_RDONLY)
971 			td->td_retval[0] |= LINUX_O_RDONLY;
972 		if (result & O_WRONLY)
973 			td->td_retval[0] |= LINUX_O_WRONLY;
974 		if (result & O_RDWR)
975 			td->td_retval[0] |= LINUX_O_RDWR;
976 		if (result & O_NDELAY)
977 			td->td_retval[0] |= LINUX_O_NONBLOCK;
978 		if (result & O_APPEND)
979 			td->td_retval[0] |= LINUX_O_APPEND;
980 		if (result & O_FSYNC)
981 			td->td_retval[0] |= LINUX_O_SYNC;
982 		if (result & O_ASYNC)
983 			td->td_retval[0] |= LINUX_FASYNC;
984 #ifdef LINUX_O_NOFOLLOW
985 		if (result & O_NOFOLLOW)
986 			td->td_retval[0] |= LINUX_O_NOFOLLOW;
987 #endif
988 #ifdef LINUX_O_DIRECT
989 		if (result & O_DIRECT)
990 			td->td_retval[0] |= LINUX_O_DIRECT;
991 #endif
992 		return (error);
993 
994 	case LINUX_F_SETFL:
995 		arg = 0;
996 		if (args->arg & LINUX_O_NDELAY)
997 			arg |= O_NONBLOCK;
998 		if (args->arg & LINUX_O_APPEND)
999 			arg |= O_APPEND;
1000 		if (args->arg & LINUX_O_SYNC)
1001 			arg |= O_FSYNC;
1002 		if (args->arg & LINUX_FASYNC)
1003 			arg |= O_ASYNC;
1004 #ifdef LINUX_O_NOFOLLOW
1005 		if (args->arg & LINUX_O_NOFOLLOW)
1006 			arg |= O_NOFOLLOW;
1007 #endif
1008 #ifdef LINUX_O_DIRECT
1009 		if (args->arg & LINUX_O_DIRECT)
1010 			arg |= O_DIRECT;
1011 #endif
1012 		return (kern_fcntl(td, args->fd, F_SETFL, arg));
1013 
1014 	case LINUX_F_GETLK:
1015 		error = copyin((void *)args->arg, &linux_flock,
1016 		    sizeof(linux_flock));
1017 		if (error)
1018 			return (error);
1019 		linux_to_bsd_flock(&linux_flock, &bsd_flock);
1020 		error = kern_fcntl(td, args->fd, F_GETLK, (intptr_t)&bsd_flock);
1021 		if (error)
1022 			return (error);
1023 		bsd_to_linux_flock(&bsd_flock, &linux_flock);
1024 		return (copyout(&linux_flock, (void *)args->arg,
1025 		    sizeof(linux_flock)));
1026 
1027 	case LINUX_F_SETLK:
1028 		error = copyin((void *)args->arg, &linux_flock,
1029 		    sizeof(linux_flock));
1030 		if (error)
1031 			return (error);
1032 		linux_to_bsd_flock(&linux_flock, &bsd_flock);
1033 		return (kern_fcntl(td, args->fd, F_SETLK,
1034 		    (intptr_t)&bsd_flock));
1035 
1036 	case LINUX_F_SETLKW:
1037 		error = copyin((void *)args->arg, &linux_flock,
1038 		    sizeof(linux_flock));
1039 		if (error)
1040 			return (error);
1041 		linux_to_bsd_flock(&linux_flock, &bsd_flock);
1042 		return (kern_fcntl(td, args->fd, F_SETLKW,
1043 		     (intptr_t)&bsd_flock));
1044 
1045 	case LINUX_F_GETOWN:
1046 		return (kern_fcntl(td, args->fd, F_GETOWN, 0));
1047 
1048 	case LINUX_F_SETOWN:
1049 		/*
1050 		 * XXX some Linux applications depend on F_SETOWN having no
1051 		 * significant effect for pipes (SIGIO is not delivered for
1052 		 * pipes under Linux-2.2.35 at least).
1053 		 */
1054 		error = fget(td, args->fd, &fp);
1055 		if (error)
1056 			return (error);
1057 		if (fp->f_type == DTYPE_PIPE) {
1058 			fdrop(fp, td);
1059 			return (EINVAL);
1060 		}
1061 		fdrop(fp, td);
1062 
1063 		return (kern_fcntl(td, args->fd, F_SETOWN, args->arg));
1064 	}
1065 
1066 	return (EINVAL);
1067 }
1068 
1069 int
1070 linux_fcntl(struct thread *td, struct linux_fcntl_args *args)
1071 {
1072 	struct linux_fcntl64_args args64;
1073 
1074 #ifdef DEBUG
1075 	if (ldebug(fcntl))
1076 		printf(ARGS(fcntl, "%d, %08x, *"), args->fd, args->cmd);
1077 #endif
1078 
1079 	args64.fd = args->fd;
1080 	args64.cmd = args->cmd;
1081 	args64.arg = args->arg;
1082 	return (fcntl_common(td, &args64));
1083 }
1084 
1085 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1086 int
1087 linux_fcntl64(struct thread *td, struct linux_fcntl64_args *args)
1088 {
1089 	struct l_flock64 linux_flock;
1090 	struct flock bsd_flock;
1091 	int error;
1092 
1093 #ifdef DEBUG
1094 	if (ldebug(fcntl64))
1095 		printf(ARGS(fcntl64, "%d, %08x, *"), args->fd, args->cmd);
1096 #endif
1097 
1098 	switch (args->cmd) {
1099 	case LINUX_F_GETLK64:
1100 		error = copyin((void *)args->arg, &linux_flock,
1101 		    sizeof(linux_flock));
1102 		if (error)
1103 			return (error);
1104 		linux_to_bsd_flock64(&linux_flock, &bsd_flock);
1105 		error = kern_fcntl(td, args->fd, F_GETLK, (intptr_t)&bsd_flock);
1106 		if (error)
1107 			return (error);
1108 		bsd_to_linux_flock64(&bsd_flock, &linux_flock);
1109 		return (copyout(&linux_flock, (void *)args->arg,
1110 			    sizeof(linux_flock)));
1111 
1112 	case LINUX_F_SETLK64:
1113 		error = copyin((void *)args->arg, &linux_flock,
1114 		    sizeof(linux_flock));
1115 		if (error)
1116 			return (error);
1117 		linux_to_bsd_flock64(&linux_flock, &bsd_flock);
1118 		return (kern_fcntl(td, args->fd, F_SETLK,
1119 		    (intptr_t)&bsd_flock));
1120 
1121 	case LINUX_F_SETLKW64:
1122 		error = copyin((void *)args->arg, &linux_flock,
1123 		    sizeof(linux_flock));
1124 		if (error)
1125 			return (error);
1126 		linux_to_bsd_flock64(&linux_flock, &bsd_flock);
1127 		return (kern_fcntl(td, args->fd, F_SETLKW,
1128 		    (intptr_t)&bsd_flock));
1129 	}
1130 
1131 	return (fcntl_common(td, args));
1132 }
1133 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1134 
1135 int
1136 linux_chown(struct thread *td, struct linux_chown_args *args)
1137 {
1138 	char *path;
1139 	int error;
1140 
1141 	LCONVPATHEXIST(td, args->path, &path);
1142 
1143 #ifdef DEBUG
1144 	if (ldebug(chown))
1145 		printf(ARGS(chown, "%s, %d, %d"), path, args->uid, args->gid);
1146 #endif
1147 	error = kern_chown(td, path, UIO_SYSSPACE, args->uid, args->gid);
1148 	LFREEPATH(path);
1149 	return (error);
1150 }
1151 
1152 int
1153 linux_lchown(struct thread *td, struct linux_lchown_args *args)
1154 {
1155 	char *path;
1156 	int error;
1157 
1158 	LCONVPATHEXIST(td, args->path, &path);
1159 
1160 #ifdef DEBUG
1161 	if (ldebug(lchown))
1162 		printf(ARGS(lchown, "%s, %d, %d"), path, args->uid, args->gid);
1163 #endif
1164 	error = kern_lchown(td, path, UIO_SYSSPACE, args->uid, args->gid);
1165 	LFREEPATH(path);
1166 	return (error);
1167 }
1168