xref: /freebsd/sys/compat/linux/linux_file.c (revision 2b743a9e9ddc6736208dc8ca1ce06ce64ad20a19)
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/malloc.h>
44 #include <sys/mount.h>
45 #include <sys/mutex.h>
46 #include <sys/proc.h>
47 #include <sys/stat.h>
48 #include <sys/sx.h>
49 #include <sys/syscallsubr.h>
50 #include <sys/sysproto.h>
51 #include <sys/tty.h>
52 #include <sys/unistd.h>
53 #include <sys/vnode.h>
54 
55 #include <security/mac/mac_framework.h>
56 
57 #include <ufs/ufs/extattr.h>
58 #include <ufs/ufs/quota.h>
59 #include <ufs/ufs/ufsmount.h>
60 
61 #ifdef COMPAT_LINUX32
62 #include <machine/../linux32/linux.h>
63 #include <machine/../linux32/linux32_proto.h>
64 #else
65 #include <machine/../linux/linux.h>
66 #include <machine/../linux/linux_proto.h>
67 #endif
68 #include <compat/linux/linux_util.h>
69 
70 int
71 linux_creat(struct thread *td, struct linux_creat_args *args)
72 {
73     char *path;
74     int error;
75 
76     LCONVPATHEXIST(td, args->path, &path);
77 
78 #ifdef DEBUG
79 	if (ldebug(creat))
80 		printf(ARGS(creat, "%s, %d"), path, args->mode);
81 #endif
82     error = kern_open(td, path, UIO_SYSSPACE, O_WRONLY | O_CREAT | O_TRUNC,
83 	args->mode);
84     LFREEPATH(path);
85     return (error);
86 }
87 
88 int
89 linux_open(struct thread *td, struct linux_open_args *args)
90 {
91     struct proc *p = td->td_proc;
92     struct file *fp;
93     int fd;
94     char *path;
95     int bsd_flags, error;
96 
97     if (args->flags & LINUX_O_CREAT)
98 	LCONVPATHCREAT(td, args->path, &path);
99     else
100 	LCONVPATHEXIST(td, args->path, &path);
101 
102 #ifdef DEBUG
103 	if (ldebug(open))
104 		printf(ARGS(open, "%s, 0x%x, 0x%x"),
105 		    path, args->flags, args->mode);
106 #endif
107     bsd_flags = 0;
108     if (args->flags & LINUX_O_RDONLY)
109 	bsd_flags |= O_RDONLY;
110     if (args->flags & LINUX_O_WRONLY)
111 	bsd_flags |= O_WRONLY;
112     if (args->flags & LINUX_O_RDWR)
113 	bsd_flags |= O_RDWR;
114     if (args->flags & LINUX_O_NDELAY)
115 	bsd_flags |= O_NONBLOCK;
116     if (args->flags & LINUX_O_APPEND)
117 	bsd_flags |= O_APPEND;
118     if (args->flags & LINUX_O_SYNC)
119 	bsd_flags |= O_FSYNC;
120     if (args->flags & LINUX_O_NONBLOCK)
121 	bsd_flags |= O_NONBLOCK;
122     if (args->flags & LINUX_FASYNC)
123 	bsd_flags |= O_ASYNC;
124     if (args->flags & LINUX_O_CREAT)
125 	bsd_flags |= O_CREAT;
126     if (args->flags & LINUX_O_TRUNC)
127 	bsd_flags |= O_TRUNC;
128     if (args->flags & LINUX_O_EXCL)
129 	bsd_flags |= O_EXCL;
130     if (args->flags & LINUX_O_NOCTTY)
131 	bsd_flags |= O_NOCTTY;
132     if (args->flags & LINUX_O_DIRECT)
133 	bsd_flags |= O_DIRECT;
134     if (args->flags & LINUX_O_NOFOLLOW)
135 	bsd_flags |= O_NOFOLLOW;
136     /* XXX LINUX_O_NOATIME: unable to be easily implemented. */
137 
138     error = kern_open(td, path, UIO_SYSSPACE, bsd_flags, args->mode);
139     if (!error) {
140 	    fd = td->td_retval[0];
141 	    /*
142 	     * XXX In between kern_open() and fget(), another process
143 	     * having the same filedesc could use that fd without
144 	     * checking below.
145 	     */
146 	    error = fget(td, fd, &fp);
147 	    if (!error) {
148 		    sx_slock(&proctree_lock);
149 		    PROC_LOCK(p);
150 		    if (!(bsd_flags & O_NOCTTY) &&
151 			SESS_LEADER(p) && !(p->p_flag & P_CONTROLT)) {
152 			    PROC_UNLOCK(p);
153 			    sx_unlock(&proctree_lock);
154 			    if (fp->f_type == DTYPE_VNODE)
155 				    (void) fo_ioctl(fp, TIOCSCTTY, (caddr_t) 0,
156 					     td->td_ucred, td);
157 		    } else {
158 			    PROC_UNLOCK(p);
159 			    sx_sunlock(&proctree_lock);
160 		    }
161 		    if (args->flags & LINUX_O_DIRECTORY) {
162 			    if (fp->f_type != DTYPE_VNODE ||
163 				fp->f_vnode->v_type != VDIR) {
164 				    error = ENOTDIR;
165 			    }
166 		    }
167 		    fdrop(fp, td);
168 		    /*
169 		     * XXX as above, fdrop()/kern_close() pair is racy.
170 		     */
171 		    if (error)
172 			    kern_close(td, fd);
173 	    }
174     }
175 
176 #ifdef DEBUG
177     if (ldebug(open))
178 	    printf(LMSG("open returns error %d"), error);
179 #endif
180     LFREEPATH(path);
181     return error;
182 }
183 
184 int
185 linux_lseek(struct thread *td, struct linux_lseek_args *args)
186 {
187 
188     struct lseek_args /* {
189 	int fd;
190 	int pad;
191 	off_t offset;
192 	int whence;
193     } */ tmp_args;
194     int error;
195 
196 #ifdef DEBUG
197 	if (ldebug(lseek))
198 		printf(ARGS(lseek, "%d, %ld, %d"),
199 		    args->fdes, (long)args->off, args->whence);
200 #endif
201     tmp_args.fd = args->fdes;
202     tmp_args.offset = (off_t)args->off;
203     tmp_args.whence = args->whence;
204     error = lseek(td, &tmp_args);
205     return error;
206 }
207 
208 int
209 linux_llseek(struct thread *td, struct linux_llseek_args *args)
210 {
211 	struct lseek_args bsd_args;
212 	int error;
213 	off_t off;
214 
215 #ifdef DEBUG
216 	if (ldebug(llseek))
217 		printf(ARGS(llseek, "%d, %d:%d, %d"),
218 		    args->fd, args->ohigh, args->olow, args->whence);
219 #endif
220 	off = (args->olow) | (((off_t) args->ohigh) << 32);
221 
222 	bsd_args.fd = args->fd;
223 	bsd_args.offset = off;
224 	bsd_args.whence = args->whence;
225 
226 	if ((error = lseek(td, &bsd_args)))
227 		return error;
228 
229 	if ((error = copyout(td->td_retval, args->res, sizeof (off_t))))
230 		return error;
231 
232 	td->td_retval[0] = 0;
233 	return 0;
234 }
235 
236 int
237 linux_readdir(struct thread *td, struct linux_readdir_args *args)
238 {
239 	struct linux_getdents_args lda;
240 
241 	lda.fd = args->fd;
242 	lda.dent = args->dent;
243 	lda.count = 1;
244 	return linux_getdents(td, &lda);
245 }
246 
247 /*
248  * Note that linux_getdents(2) and linux_getdents64(2) have the same
249  * arguments. They only differ in the definition of struct dirent they
250  * operate on. We use this to common the code, with the exception of
251  * accessing struct dirent. Note that linux_readdir(2) is implemented
252  * by means of linux_getdents(2). In this case we never operate on
253  * struct dirent64 and thus don't need to handle it...
254  */
255 
256 struct l_dirent {
257 	l_long		d_ino;
258 	l_off_t		d_off;
259 	l_ushort	d_reclen;
260 	char		d_name[LINUX_NAME_MAX + 1];
261 };
262 
263 struct l_dirent64 {
264 	uint64_t	d_ino;
265 	int64_t		d_off;
266 	l_ushort	d_reclen;
267 	u_char		d_type;
268 	char		d_name[LINUX_NAME_MAX + 1];
269 };
270 
271 #define LINUX_RECLEN(de,namlen) \
272     ALIGN((((char *)&(de)->d_name - (char *)de) + (namlen) + 1))
273 
274 #define	LINUX_DIRBLKSIZ		512
275 
276 static int
277 getdents_common(struct thread *td, struct linux_getdents64_args *args,
278     int is64bit)
279 {
280 	struct dirent *bdp;
281 	struct vnode *vp;
282 	caddr_t inp, buf;		/* BSD-format */
283 	int len, reclen;		/* BSD-format */
284 	caddr_t outp;			/* Linux-format */
285 	int resid, linuxreclen=0;	/* Linux-format */
286 	struct file *fp;
287 	struct uio auio;
288 	struct iovec aiov;
289 	off_t off;
290 	struct l_dirent linux_dirent;
291 	struct l_dirent64 linux_dirent64;
292 	int buflen, error, eofflag, nbytes, justone;
293 	u_long *cookies = NULL, *cookiep;
294 	int ncookies, vfslocked;
295 
296 	nbytes = args->count;
297 	if (nbytes == 1) {
298 		/* readdir(2) case. Always struct dirent. */
299 		if (is64bit)
300 			return (EINVAL);
301 		nbytes = sizeof(linux_dirent);
302 		justone = 1;
303 	} else
304 		justone = 0;
305 
306 	if ((error = getvnode(td->td_proc->p_fd, args->fd, &fp)) != 0)
307 		return (error);
308 
309 	if ((fp->f_flag & FREAD) == 0) {
310 		fdrop(fp, td);
311 		return (EBADF);
312 	}
313 
314 	vp = fp->f_vnode;
315 	vfslocked = VFS_LOCK_GIANT(vp->v_mount);
316 	if (vp->v_type != VDIR) {
317 		VFS_UNLOCK_GIANT(vfslocked);
318 		fdrop(fp, td);
319 		return (EINVAL);
320 	}
321 
322 	off = fp->f_offset;
323 
324 	buflen = max(LINUX_DIRBLKSIZ, nbytes);
325 	buflen = min(buflen, MAXBSIZE);
326 	buf = malloc(buflen, M_TEMP, M_WAITOK);
327 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
328 
329 again:
330 	aiov.iov_base = buf;
331 	aiov.iov_len = buflen;
332 	auio.uio_iov = &aiov;
333 	auio.uio_iovcnt = 1;
334 	auio.uio_rw = UIO_READ;
335 	auio.uio_segflg = UIO_SYSSPACE;
336 	auio.uio_td = td;
337 	auio.uio_resid = buflen;
338 	auio.uio_offset = off;
339 
340 	if (cookies) {
341 		free(cookies, M_TEMP);
342 		cookies = NULL;
343 	}
344 
345 #ifdef MAC
346 	/*
347 	 * Do directory search MAC check using non-cached credentials.
348 	 */
349 	if ((error = mac_check_vnode_readdir(td->td_ucred, vp)))
350 		goto out;
351 #endif /* MAC */
352 	if ((error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, &ncookies,
353 		 &cookies)))
354 		goto out;
355 
356 	inp = buf;
357 	outp = (caddr_t)args->dirent;
358 	resid = nbytes;
359 	if ((len = buflen - auio.uio_resid) <= 0)
360 		goto eof;
361 
362 	cookiep = cookies;
363 
364 	if (cookies) {
365 		/*
366 		 * When using cookies, the vfs has the option of reading from
367 		 * a different offset than that supplied (UFS truncates the
368 		 * offset to a block boundary to make sure that it never reads
369 		 * partway through a directory entry, even if the directory
370 		 * has been compacted).
371 		 */
372 		while (len > 0 && ncookies > 0 && *cookiep <= off) {
373 			bdp = (struct dirent *) inp;
374 			len -= bdp->d_reclen;
375 			inp += bdp->d_reclen;
376 			cookiep++;
377 			ncookies--;
378 		}
379 	}
380 
381 	while (len > 0) {
382 		if (cookiep && ncookies == 0)
383 			break;
384 		bdp = (struct dirent *) inp;
385 		reclen = bdp->d_reclen;
386 		if (reclen & 3) {
387 			error = EFAULT;
388 			goto out;
389 		}
390 
391 		if (bdp->d_fileno == 0) {
392 			inp += reclen;
393 			if (cookiep) {
394 				off = *cookiep++;
395 				ncookies--;
396 			} else
397 				off += reclen;
398 
399 			len -= reclen;
400 			continue;
401 		}
402 
403 		linuxreclen = (is64bit)
404 		    ? LINUX_RECLEN(&linux_dirent64, bdp->d_namlen)
405 		    : LINUX_RECLEN(&linux_dirent, bdp->d_namlen);
406 
407 		if (reclen > len || resid < linuxreclen) {
408 			outp++;
409 			break;
410 		}
411 
412 		if (justone) {
413 			/* readdir(2) case. */
414 			linux_dirent.d_ino = (l_long)bdp->d_fileno;
415 			linux_dirent.d_off = (l_off_t)linuxreclen;
416 			linux_dirent.d_reclen = (l_ushort)bdp->d_namlen;
417 			strcpy(linux_dirent.d_name, bdp->d_name);
418 			error = copyout(&linux_dirent, outp, linuxreclen);
419 		} else {
420 			if (is64bit) {
421 				linux_dirent64.d_ino = bdp->d_fileno;
422 				linux_dirent64.d_off = (cookiep)
423 				    ? (l_off_t)*cookiep
424 				    : (l_off_t)(off + reclen);
425 				linux_dirent64.d_reclen =
426 				    (l_ushort)linuxreclen;
427 				linux_dirent64.d_type = bdp->d_type;
428 				strcpy(linux_dirent64.d_name, bdp->d_name);
429 				error = copyout(&linux_dirent64, outp,
430 				    linuxreclen);
431 			} else {
432 				linux_dirent.d_ino = bdp->d_fileno;
433 				linux_dirent.d_off = (cookiep)
434 				    ? (l_off_t)*cookiep
435 				    : (l_off_t)(off + reclen);
436 				linux_dirent.d_reclen = (l_ushort)linuxreclen;
437 				strcpy(linux_dirent.d_name, bdp->d_name);
438 				error = copyout(&linux_dirent, outp,
439 				    linuxreclen);
440 			}
441 		}
442 		if (error)
443 			goto out;
444 
445 		inp += reclen;
446 		if (cookiep) {
447 			off = *cookiep++;
448 			ncookies--;
449 		} else
450 			off += reclen;
451 
452 		outp += linuxreclen;
453 		resid -= linuxreclen;
454 		len -= reclen;
455 		if (justone)
456 			break;
457 	}
458 
459 	if (outp == (caddr_t)args->dirent)
460 		goto again;
461 
462 	fp->f_offset = off;
463 	if (justone)
464 		nbytes = resid + linuxreclen;
465 
466 eof:
467 	td->td_retval[0] = nbytes - resid;
468 
469 out:
470 	if (cookies)
471 		free(cookies, M_TEMP);
472 
473 	VOP_UNLOCK(vp, 0, td);
474 	VFS_UNLOCK_GIANT(vfslocked);
475 	fdrop(fp, td);
476 	free(buf, M_TEMP);
477 	return (error);
478 }
479 
480 int
481 linux_getdents(struct thread *td, struct linux_getdents_args *args)
482 {
483 
484 #ifdef DEBUG
485 	if (ldebug(getdents))
486 		printf(ARGS(getdents, "%d, *, %d"), args->fd, args->count);
487 #endif
488 
489 	return (getdents_common(td, (struct linux_getdents64_args*)args, 0));
490 }
491 
492 int
493 linux_getdents64(struct thread *td, struct linux_getdents64_args *args)
494 {
495 
496 #ifdef DEBUG
497 	if (ldebug(getdents64))
498 		printf(ARGS(getdents64, "%d, *, %d"), args->fd, args->count);
499 #endif
500 
501 	return (getdents_common(td, args, 1));
502 }
503 
504 /*
505  * These exist mainly for hooks for doing /compat/linux translation.
506  */
507 
508 int
509 linux_access(struct thread *td, struct linux_access_args *args)
510 {
511 	char *path;
512 	int error;
513 
514 	/* linux convention */
515 	if (args->flags & ~(F_OK | X_OK | W_OK | R_OK))
516 		return (EINVAL);
517 
518 	LCONVPATHEXIST(td, args->path, &path);
519 
520 #ifdef DEBUG
521 	if (ldebug(access))
522 		printf(ARGS(access, "%s, %d"), path, args->flags);
523 #endif
524 	error = kern_access(td, path, UIO_SYSSPACE, args->flags);
525 	LFREEPATH(path);
526 
527 	return (error);
528 }
529 
530 int
531 linux_unlink(struct thread *td, struct linux_unlink_args *args)
532 {
533 	char *path;
534 	int error;
535 	struct stat st;
536 
537 	LCONVPATHEXIST(td, args->path, &path);
538 
539 #ifdef DEBUG
540 	if (ldebug(unlink))
541 		printf(ARGS(unlink, "%s"), path);
542 #endif
543 
544 	error = kern_unlink(td, path, UIO_SYSSPACE);
545 	if (error == EPERM)
546 		/* Introduce POSIX noncompliant behaviour of Linux */
547 		if (kern_stat(td, path, UIO_SYSSPACE, &st) == 0)
548 			if (S_ISDIR(st.st_mode))
549 				error = EISDIR;
550 	LFREEPATH(path);
551 	return (error);
552 }
553 
554 int
555 linux_chdir(struct thread *td, struct linux_chdir_args *args)
556 {
557 	char *path;
558 	int error;
559 
560 	LCONVPATHEXIST(td, args->path, &path);
561 
562 #ifdef DEBUG
563 	if (ldebug(chdir))
564 		printf(ARGS(chdir, "%s"), path);
565 #endif
566 	error = kern_chdir(td, path, UIO_SYSSPACE);
567 	LFREEPATH(path);
568 	return (error);
569 }
570 
571 int
572 linux_chmod(struct thread *td, struct linux_chmod_args *args)
573 {
574 	char *path;
575 	int error;
576 
577 	LCONVPATHEXIST(td, args->path, &path);
578 
579 #ifdef DEBUG
580 	if (ldebug(chmod))
581 		printf(ARGS(chmod, "%s, %d"), path, args->mode);
582 #endif
583 	error = kern_chmod(td, path, UIO_SYSSPACE, args->mode);
584 	LFREEPATH(path);
585 	return (error);
586 }
587 
588 int
589 linux_mkdir(struct thread *td, struct linux_mkdir_args *args)
590 {
591 	char *path;
592 	int error;
593 
594 	LCONVPATHCREAT(td, args->path, &path);
595 
596 #ifdef DEBUG
597 	if (ldebug(mkdir))
598 		printf(ARGS(mkdir, "%s, %d"), path, args->mode);
599 #endif
600 	error = kern_mkdir(td, path, UIO_SYSSPACE, args->mode);
601 	LFREEPATH(path);
602 	return (error);
603 }
604 
605 int
606 linux_rmdir(struct thread *td, struct linux_rmdir_args *args)
607 {
608 	char *path;
609 	int error;
610 
611 	LCONVPATHEXIST(td, args->path, &path);
612 
613 #ifdef DEBUG
614 	if (ldebug(rmdir))
615 		printf(ARGS(rmdir, "%s"), path);
616 #endif
617 	error = kern_rmdir(td, path, UIO_SYSSPACE);
618 	LFREEPATH(path);
619 	return (error);
620 }
621 
622 int
623 linux_rename(struct thread *td, struct linux_rename_args *args)
624 {
625 	char *from, *to;
626 	int error;
627 
628 	LCONVPATHEXIST(td, args->from, &from);
629 	/* Expand LCONVPATHCREATE so that `from' can be freed on errors */
630 	error = linux_emul_convpath(td, args->to, UIO_USERSPACE, &to, 1);
631 	if (to == NULL) {
632 		LFREEPATH(from);
633 		return (error);
634 	}
635 
636 #ifdef DEBUG
637 	if (ldebug(rename))
638 		printf(ARGS(rename, "%s, %s"), from, to);
639 #endif
640 	error = kern_rename(td, from, to, UIO_SYSSPACE);
641 	LFREEPATH(from);
642 	LFREEPATH(to);
643 	return (error);
644 }
645 
646 int
647 linux_symlink(struct thread *td, struct linux_symlink_args *args)
648 {
649 	char *path, *to;
650 	int error;
651 
652 	LCONVPATHEXIST(td, args->path, &path);
653 	/* Expand LCONVPATHCREATE so that `path' can be freed on errors */
654 	error = linux_emul_convpath(td, args->to, UIO_USERSPACE, &to, 1);
655 	if (to == NULL) {
656 		LFREEPATH(path);
657 		return (error);
658 	}
659 
660 #ifdef DEBUG
661 	if (ldebug(symlink))
662 		printf(ARGS(symlink, "%s, %s"), path, to);
663 #endif
664 	error = kern_symlink(td, path, to, UIO_SYSSPACE);
665 	LFREEPATH(path);
666 	LFREEPATH(to);
667 	return (error);
668 }
669 
670 int
671 linux_readlink(struct thread *td, struct linux_readlink_args *args)
672 {
673 	char *name;
674 	int error;
675 
676 	LCONVPATHEXIST(td, args->name, &name);
677 
678 #ifdef DEBUG
679 	if (ldebug(readlink))
680 		printf(ARGS(readlink, "%s, %p, %d"), name, (void *)args->buf,
681 		    args->count);
682 #endif
683 	error = kern_readlink(td, name, UIO_SYSSPACE, args->buf, UIO_USERSPACE,
684 	    args->count);
685 	LFREEPATH(name);
686 	return (error);
687 }
688 
689 int
690 linux_truncate(struct thread *td, struct linux_truncate_args *args)
691 {
692 	char *path;
693 	int error;
694 
695 	LCONVPATHEXIST(td, args->path, &path);
696 
697 #ifdef DEBUG
698 	if (ldebug(truncate))
699 		printf(ARGS(truncate, "%s, %ld"), path, (long)args->length);
700 #endif
701 
702 	error = kern_truncate(td, path, UIO_SYSSPACE, args->length);
703 	LFREEPATH(path);
704 	return (error);
705 }
706 
707 int
708 linux_ftruncate(struct thread *td, struct linux_ftruncate_args *args)
709 {
710 	struct ftruncate_args /* {
711 		int fd;
712 		int pad;
713 		off_t length;
714 		} */ nuap;
715 
716 	nuap.fd = args->fd;
717 	nuap.pad = 0;
718 	nuap.length = args->length;
719 	return (ftruncate(td, &nuap));
720 }
721 
722 int
723 linux_link(struct thread *td, struct linux_link_args *args)
724 {
725 	char *path, *to;
726 	int error;
727 
728 	LCONVPATHEXIST(td, args->path, &path);
729 	/* Expand LCONVPATHCREATE so that `path' can be freed on errors */
730 	error = linux_emul_convpath(td, args->to, UIO_USERSPACE, &to, 1);
731 	if (to == NULL) {
732 		LFREEPATH(path);
733 		return (error);
734 	}
735 
736 #ifdef DEBUG
737 	if (ldebug(link))
738 		printf(ARGS(link, "%s, %s"), path, to);
739 #endif
740 	error = kern_link(td, path, to, UIO_SYSSPACE);
741 	LFREEPATH(path);
742 	LFREEPATH(to);
743 	return (error);
744 }
745 
746 int
747 linux_fdatasync(td, uap)
748 	struct thread *td;
749 	struct linux_fdatasync_args *uap;
750 {
751 	struct fsync_args bsd;
752 
753 	bsd.fd = uap->fd;
754 	return fsync(td, &bsd);
755 }
756 
757 int
758 linux_pread(td, uap)
759 	struct thread *td;
760 	struct linux_pread_args *uap;
761 {
762 	struct pread_args bsd;
763 	struct vnode *vp;
764 	int error;
765 
766 	bsd.fd = uap->fd;
767 	bsd.buf = uap->buf;
768 	bsd.nbyte = uap->nbyte;
769 	bsd.offset = uap->offset;
770 
771 	error = pread(td, &bsd);
772 
773 	if (error == 0) {
774    	   	/* This seems to violate POSIX but linux does it */
775    	   	if ((error = fgetvp(td, uap->fd, &vp)) != 0)
776    		   	return (error);
777 		if (vp->v_type == VDIR) {
778    		   	vrele(vp);
779 			return (EISDIR);
780 		}
781 		vrele(vp);
782 	}
783 
784 	return (error);
785 }
786 
787 int
788 linux_pwrite(td, uap)
789 	struct thread *td;
790 	struct linux_pwrite_args *uap;
791 {
792 	struct pwrite_args bsd;
793 
794 	bsd.fd = uap->fd;
795 	bsd.buf = uap->buf;
796 	bsd.nbyte = uap->nbyte;
797 	bsd.offset = uap->offset;
798 	return pwrite(td, &bsd);
799 }
800 
801 int
802 linux_mount(struct thread *td, struct linux_mount_args *args)
803 {
804 	struct ufs_args ufs;
805 	char fstypename[MFSNAMELEN];
806 	char mntonname[MNAMELEN], mntfromname[MNAMELEN];
807 	int error;
808 	int fsflags;
809 	void *fsdata;
810 
811 	error = copyinstr(args->filesystemtype, fstypename, MFSNAMELEN - 1,
812 	    NULL);
813 	if (error)
814 		return (error);
815 	error = copyinstr(args->specialfile, mntfromname, MNAMELEN - 1, NULL);
816 	if (error)
817 		return (error);
818 	error = copyinstr(args->dir, mntonname, MNAMELEN - 1, NULL);
819 	if (error)
820 		return (error);
821 
822 #ifdef DEBUG
823 	if (ldebug(mount))
824 		printf(ARGS(mount, "%s, %s, %s"),
825 		    fstypename, mntfromname, mntonname);
826 #endif
827 
828 	if (strcmp(fstypename, "ext2") == 0) {
829 		strcpy(fstypename, "ext2fs");
830 		fsdata = &ufs;
831 		ufs.fspec = mntfromname;
832 #define DEFAULT_ROOTID		-2
833 		ufs.export.ex_root = DEFAULT_ROOTID;
834 		ufs.export.ex_flags =
835 		    args->rwflag & LINUX_MS_RDONLY ? MNT_EXRDONLY : 0;
836 	} else if (strcmp(fstypename, "proc") == 0) {
837 		strcpy(fstypename, "linprocfs");
838 		fsdata = NULL;
839 	} else {
840 		return (ENODEV);
841 	}
842 
843 	fsflags = 0;
844 
845 	if ((args->rwflag & 0xffff0000) == 0xc0ed0000) {
846 		/*
847 		 * Linux SYNC flag is not included; the closest equivalent
848 		 * FreeBSD has is !ASYNC, which is our default.
849 		 */
850 		if (args->rwflag & LINUX_MS_RDONLY)
851 			fsflags |= MNT_RDONLY;
852 		if (args->rwflag & LINUX_MS_NOSUID)
853 			fsflags |= MNT_NOSUID;
854 		if (args->rwflag & LINUX_MS_NOEXEC)
855 			fsflags |= MNT_NOEXEC;
856 		if (args->rwflag & LINUX_MS_REMOUNT)
857 			fsflags |= MNT_UPDATE;
858 	}
859 
860 	if (strcmp(fstypename, "linprocfs") == 0) {
861 		error = kernel_vmount(fsflags,
862 			"fstype", fstypename,
863 			"fspath", mntonname,
864 			NULL);
865 	} else
866 		error = EOPNOTSUPP;
867 	return (error);
868 }
869 
870 int
871 linux_oldumount(struct thread *td, struct linux_oldumount_args *args)
872 {
873 	struct linux_umount_args args2;
874 
875 	args2.path = args->path;
876 	args2.flags = 0;
877 	return (linux_umount(td, &args2));
878 }
879 
880 int
881 linux_umount(struct thread *td, struct linux_umount_args *args)
882 {
883 	struct unmount_args bsd;
884 
885 	bsd.path = args->path;
886 	bsd.flags = args->flags;	/* XXX correct? */
887 	return (unmount(td, &bsd));
888 }
889 
890 /*
891  * fcntl family of syscalls
892  */
893 
894 struct l_flock {
895 	l_short		l_type;
896 	l_short		l_whence;
897 	l_off_t		l_start;
898 	l_off_t		l_len;
899 	l_pid_t		l_pid;
900 }
901 #if defined(__amd64__) && defined(COMPAT_LINUX32)
902 __packed
903 #endif
904 ;
905 
906 static void
907 linux_to_bsd_flock(struct l_flock *linux_flock, struct flock *bsd_flock)
908 {
909 	switch (linux_flock->l_type) {
910 	case LINUX_F_RDLCK:
911 		bsd_flock->l_type = F_RDLCK;
912 		break;
913 	case LINUX_F_WRLCK:
914 		bsd_flock->l_type = F_WRLCK;
915 		break;
916 	case LINUX_F_UNLCK:
917 		bsd_flock->l_type = F_UNLCK;
918 		break;
919 	default:
920 		bsd_flock->l_type = -1;
921 		break;
922 	}
923 	bsd_flock->l_whence = linux_flock->l_whence;
924 	bsd_flock->l_start = (off_t)linux_flock->l_start;
925 	bsd_flock->l_len = (off_t)linux_flock->l_len;
926 	bsd_flock->l_pid = (pid_t)linux_flock->l_pid;
927 }
928 
929 static void
930 bsd_to_linux_flock(struct flock *bsd_flock, struct l_flock *linux_flock)
931 {
932 	switch (bsd_flock->l_type) {
933 	case F_RDLCK:
934 		linux_flock->l_type = LINUX_F_RDLCK;
935 		break;
936 	case F_WRLCK:
937 		linux_flock->l_type = LINUX_F_WRLCK;
938 		break;
939 	case F_UNLCK:
940 		linux_flock->l_type = LINUX_F_UNLCK;
941 		break;
942 	}
943 	linux_flock->l_whence = bsd_flock->l_whence;
944 	linux_flock->l_start = (l_off_t)bsd_flock->l_start;
945 	linux_flock->l_len = (l_off_t)bsd_flock->l_len;
946 	linux_flock->l_pid = (l_pid_t)bsd_flock->l_pid;
947 }
948 
949 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
950 struct l_flock64 {
951 	l_short		l_type;
952 	l_short		l_whence;
953 	l_loff_t	l_start;
954 	l_loff_t	l_len;
955 	l_pid_t		l_pid;
956 }
957 #if defined(__amd64__) && defined(COMPAT_LINUX32)
958 __packed
959 #endif
960 ;
961 
962 static void
963 linux_to_bsd_flock64(struct l_flock64 *linux_flock, struct flock *bsd_flock)
964 {
965 	switch (linux_flock->l_type) {
966 	case LINUX_F_RDLCK:
967 		bsd_flock->l_type = F_RDLCK;
968 		break;
969 	case LINUX_F_WRLCK:
970 		bsd_flock->l_type = F_WRLCK;
971 		break;
972 	case LINUX_F_UNLCK:
973 		bsd_flock->l_type = F_UNLCK;
974 		break;
975 	default:
976 		bsd_flock->l_type = -1;
977 		break;
978 	}
979 	bsd_flock->l_whence = linux_flock->l_whence;
980 	bsd_flock->l_start = (off_t)linux_flock->l_start;
981 	bsd_flock->l_len = (off_t)linux_flock->l_len;
982 	bsd_flock->l_pid = (pid_t)linux_flock->l_pid;
983 }
984 
985 static void
986 bsd_to_linux_flock64(struct flock *bsd_flock, struct l_flock64 *linux_flock)
987 {
988 	switch (bsd_flock->l_type) {
989 	case F_RDLCK:
990 		linux_flock->l_type = LINUX_F_RDLCK;
991 		break;
992 	case F_WRLCK:
993 		linux_flock->l_type = LINUX_F_WRLCK;
994 		break;
995 	case F_UNLCK:
996 		linux_flock->l_type = LINUX_F_UNLCK;
997 		break;
998 	}
999 	linux_flock->l_whence = bsd_flock->l_whence;
1000 	linux_flock->l_start = (l_loff_t)bsd_flock->l_start;
1001 	linux_flock->l_len = (l_loff_t)bsd_flock->l_len;
1002 	linux_flock->l_pid = (l_pid_t)bsd_flock->l_pid;
1003 }
1004 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1005 
1006 static int
1007 fcntl_common(struct thread *td, struct linux_fcntl64_args *args)
1008 {
1009 	struct l_flock linux_flock;
1010 	struct flock bsd_flock;
1011 	struct file *fp;
1012 	long arg;
1013 	int error, result;
1014 
1015 	switch (args->cmd) {
1016 	case LINUX_F_DUPFD:
1017 		return (kern_fcntl(td, args->fd, F_DUPFD, args->arg));
1018 
1019 	case LINUX_F_GETFD:
1020 		return (kern_fcntl(td, args->fd, F_GETFD, 0));
1021 
1022 	case LINUX_F_SETFD:
1023 		return (kern_fcntl(td, args->fd, F_SETFD, args->arg));
1024 
1025 	case LINUX_F_GETFL:
1026 		error = kern_fcntl(td, args->fd, F_GETFL, 0);
1027 		result = td->td_retval[0];
1028 		td->td_retval[0] = 0;
1029 		if (result & O_RDONLY)
1030 			td->td_retval[0] |= LINUX_O_RDONLY;
1031 		if (result & O_WRONLY)
1032 			td->td_retval[0] |= LINUX_O_WRONLY;
1033 		if (result & O_RDWR)
1034 			td->td_retval[0] |= LINUX_O_RDWR;
1035 		if (result & O_NDELAY)
1036 			td->td_retval[0] |= LINUX_O_NONBLOCK;
1037 		if (result & O_APPEND)
1038 			td->td_retval[0] |= LINUX_O_APPEND;
1039 		if (result & O_FSYNC)
1040 			td->td_retval[0] |= LINUX_O_SYNC;
1041 		if (result & O_ASYNC)
1042 			td->td_retval[0] |= LINUX_FASYNC;
1043 #ifdef LINUX_O_NOFOLLOW
1044 		if (result & O_NOFOLLOW)
1045 			td->td_retval[0] |= LINUX_O_NOFOLLOW;
1046 #endif
1047 #ifdef LINUX_O_DIRECT
1048 		if (result & O_DIRECT)
1049 			td->td_retval[0] |= LINUX_O_DIRECT;
1050 #endif
1051 		return (error);
1052 
1053 	case LINUX_F_SETFL:
1054 		arg = 0;
1055 		if (args->arg & LINUX_O_NDELAY)
1056 			arg |= O_NONBLOCK;
1057 		if (args->arg & LINUX_O_APPEND)
1058 			arg |= O_APPEND;
1059 		if (args->arg & LINUX_O_SYNC)
1060 			arg |= O_FSYNC;
1061 		if (args->arg & LINUX_FASYNC)
1062 			arg |= O_ASYNC;
1063 #ifdef LINUX_O_NOFOLLOW
1064 		if (args->arg & LINUX_O_NOFOLLOW)
1065 			arg |= O_NOFOLLOW;
1066 #endif
1067 #ifdef LINUX_O_DIRECT
1068 		if (args->arg & LINUX_O_DIRECT)
1069 			arg |= O_DIRECT;
1070 #endif
1071 		return (kern_fcntl(td, args->fd, F_SETFL, arg));
1072 
1073 	case LINUX_F_GETLK:
1074 		error = copyin((void *)args->arg, &linux_flock,
1075 		    sizeof(linux_flock));
1076 		if (error)
1077 			return (error);
1078 		linux_to_bsd_flock(&linux_flock, &bsd_flock);
1079 		error = kern_fcntl(td, args->fd, F_GETLK, (intptr_t)&bsd_flock);
1080 		if (error)
1081 			return (error);
1082 		bsd_to_linux_flock(&bsd_flock, &linux_flock);
1083 		return (copyout(&linux_flock, (void *)args->arg,
1084 		    sizeof(linux_flock)));
1085 
1086 	case LINUX_F_SETLK:
1087 		error = copyin((void *)args->arg, &linux_flock,
1088 		    sizeof(linux_flock));
1089 		if (error)
1090 			return (error);
1091 		linux_to_bsd_flock(&linux_flock, &bsd_flock);
1092 		return (kern_fcntl(td, args->fd, F_SETLK,
1093 		    (intptr_t)&bsd_flock));
1094 
1095 	case LINUX_F_SETLKW:
1096 		error = copyin((void *)args->arg, &linux_flock,
1097 		    sizeof(linux_flock));
1098 		if (error)
1099 			return (error);
1100 		linux_to_bsd_flock(&linux_flock, &bsd_flock);
1101 		return (kern_fcntl(td, args->fd, F_SETLKW,
1102 		     (intptr_t)&bsd_flock));
1103 
1104 	case LINUX_F_GETOWN:
1105 		return (kern_fcntl(td, args->fd, F_GETOWN, 0));
1106 
1107 	case LINUX_F_SETOWN:
1108 		/*
1109 		 * XXX some Linux applications depend on F_SETOWN having no
1110 		 * significant effect for pipes (SIGIO is not delivered for
1111 		 * pipes under Linux-2.2.35 at least).
1112 		 */
1113 		error = fget(td, args->fd, &fp);
1114 		if (error)
1115 			return (error);
1116 		if (fp->f_type == DTYPE_PIPE) {
1117 			fdrop(fp, td);
1118 			return (EINVAL);
1119 		}
1120 		fdrop(fp, td);
1121 
1122 		return (kern_fcntl(td, args->fd, F_SETOWN, args->arg));
1123 	}
1124 
1125 	return (EINVAL);
1126 }
1127 
1128 int
1129 linux_fcntl(struct thread *td, struct linux_fcntl_args *args)
1130 {
1131 	struct linux_fcntl64_args args64;
1132 
1133 #ifdef DEBUG
1134 	if (ldebug(fcntl))
1135 		printf(ARGS(fcntl, "%d, %08x, *"), args->fd, args->cmd);
1136 #endif
1137 
1138 	args64.fd = args->fd;
1139 	args64.cmd = args->cmd;
1140 	args64.arg = args->arg;
1141 	return (fcntl_common(td, &args64));
1142 }
1143 
1144 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1145 int
1146 linux_fcntl64(struct thread *td, struct linux_fcntl64_args *args)
1147 {
1148 	struct l_flock64 linux_flock;
1149 	struct flock bsd_flock;
1150 	int error;
1151 
1152 #ifdef DEBUG
1153 	if (ldebug(fcntl64))
1154 		printf(ARGS(fcntl64, "%d, %08x, *"), args->fd, args->cmd);
1155 #endif
1156 
1157 	switch (args->cmd) {
1158 	case LINUX_F_GETLK64:
1159 		error = copyin((void *)args->arg, &linux_flock,
1160 		    sizeof(linux_flock));
1161 		if (error)
1162 			return (error);
1163 		linux_to_bsd_flock64(&linux_flock, &bsd_flock);
1164 		error = kern_fcntl(td, args->fd, F_GETLK, (intptr_t)&bsd_flock);
1165 		if (error)
1166 			return (error);
1167 		bsd_to_linux_flock64(&bsd_flock, &linux_flock);
1168 		return (copyout(&linux_flock, (void *)args->arg,
1169 			    sizeof(linux_flock)));
1170 
1171 	case LINUX_F_SETLK64:
1172 		error = copyin((void *)args->arg, &linux_flock,
1173 		    sizeof(linux_flock));
1174 		if (error)
1175 			return (error);
1176 		linux_to_bsd_flock64(&linux_flock, &bsd_flock);
1177 		return (kern_fcntl(td, args->fd, F_SETLK,
1178 		    (intptr_t)&bsd_flock));
1179 
1180 	case LINUX_F_SETLKW64:
1181 		error = copyin((void *)args->arg, &linux_flock,
1182 		    sizeof(linux_flock));
1183 		if (error)
1184 			return (error);
1185 		linux_to_bsd_flock64(&linux_flock, &bsd_flock);
1186 		return (kern_fcntl(td, args->fd, F_SETLKW,
1187 		    (intptr_t)&bsd_flock));
1188 	}
1189 
1190 	return (fcntl_common(td, args));
1191 }
1192 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1193 
1194 int
1195 linux_chown(struct thread *td, struct linux_chown_args *args)
1196 {
1197 	char *path;
1198 	int error;
1199 
1200 	LCONVPATHEXIST(td, args->path, &path);
1201 
1202 #ifdef DEBUG
1203 	if (ldebug(chown))
1204 		printf(ARGS(chown, "%s, %d, %d"), path, args->uid, args->gid);
1205 #endif
1206 	error = kern_chown(td, path, UIO_SYSSPACE, args->uid, args->gid);
1207 	LFREEPATH(path);
1208 	return (error);
1209 }
1210 
1211 int
1212 linux_lchown(struct thread *td, struct linux_lchown_args *args)
1213 {
1214 	char *path;
1215 	int error;
1216 
1217 	LCONVPATHEXIST(td, args->path, &path);
1218 
1219 #ifdef DEBUG
1220 	if (ldebug(lchown))
1221 		printf(ARGS(lchown, "%s, %d, %d"), path, args->uid, args->gid);
1222 #endif
1223 	error = kern_lchown(td, path, UIO_SYSSPACE, args->uid, args->gid);
1224 	LFREEPATH(path);
1225 	return (error);
1226 }
1227