linux_file.c revision 293521
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: stable/10/sys/compat/linux/linux_file.c 293521 2016-01-09 15:54:55Z dchagin $");
31
32#include "opt_compat.h"
33
34#include <sys/param.h>
35#include <sys/systm.h>
36#include <sys/capsicum.h>
37#include <sys/conf.h>
38#include <sys/dirent.h>
39#include <sys/fcntl.h>
40#include <sys/file.h>
41#include <sys/filedesc.h>
42#include <sys/lock.h>
43#include <sys/malloc.h>
44#include <sys/mount.h>
45#include <sys/mutex.h>
46#include <sys/namei.h>
47#include <sys/proc.h>
48#include <sys/stat.h>
49#include <sys/sx.h>
50#include <sys/syscallsubr.h>
51#include <sys/sysproto.h>
52#include <sys/tty.h>
53#include <sys/unistd.h>
54#include <sys/vnode.h>
55
56#include <security/mac/mac_framework.h>
57
58#ifdef COMPAT_LINUX32
59#include <machine/../linux32/linux.h>
60#include <machine/../linux32/linux32_proto.h>
61#else
62#include <machine/../linux/linux.h>
63#include <machine/../linux/linux_proto.h>
64#endif
65#include <compat/linux/linux_misc.h>
66#include <compat/linux/linux_util.h>
67#include <compat/linux/linux_file.h>
68
69int
70linux_creat(struct thread *td, struct linux_creat_args *args)
71{
72    char *path;
73    int error;
74
75    LCONVPATHEXIST(td, args->path, &path);
76
77#ifdef DEBUG
78	if (ldebug(creat))
79		printf(ARGS(creat, "%s, %d"), path, args->mode);
80#endif
81    error = kern_open(td, path, UIO_SYSSPACE, O_WRONLY | O_CREAT | O_TRUNC,
82	args->mode);
83    LFREEPATH(path);
84    return (error);
85}
86
87
88static int
89linux_common_open(struct thread *td, int dirfd, char *path, int l_flags, int mode)
90{
91    cap_rights_t rights;
92    struct proc *p = td->td_proc;
93    struct file *fp;
94    int fd;
95    int bsd_flags, error;
96
97    bsd_flags = 0;
98    switch (l_flags & LINUX_O_ACCMODE) {
99    case LINUX_O_WRONLY:
100	bsd_flags |= O_WRONLY;
101	break;
102    case LINUX_O_RDWR:
103	bsd_flags |= O_RDWR;
104	break;
105    default:
106	bsd_flags |= O_RDONLY;
107    }
108    if (l_flags & LINUX_O_NDELAY)
109	bsd_flags |= O_NONBLOCK;
110    if (l_flags & LINUX_O_APPEND)
111	bsd_flags |= O_APPEND;
112    if (l_flags & LINUX_O_SYNC)
113	bsd_flags |= O_FSYNC;
114    if (l_flags & LINUX_O_NONBLOCK)
115	bsd_flags |= O_NONBLOCK;
116    if (l_flags & LINUX_FASYNC)
117	bsd_flags |= O_ASYNC;
118    if (l_flags & LINUX_O_CREAT)
119	bsd_flags |= O_CREAT;
120    if (l_flags & LINUX_O_TRUNC)
121	bsd_flags |= O_TRUNC;
122    if (l_flags & LINUX_O_EXCL)
123	bsd_flags |= O_EXCL;
124    if (l_flags & LINUX_O_NOCTTY)
125	bsd_flags |= O_NOCTTY;
126    if (l_flags & LINUX_O_DIRECT)
127	bsd_flags |= O_DIRECT;
128    if (l_flags & LINUX_O_NOFOLLOW)
129	bsd_flags |= O_NOFOLLOW;
130    if (l_flags & LINUX_O_DIRECTORY)
131	bsd_flags |= O_DIRECTORY;
132    /* XXX LINUX_O_NOATIME: unable to be easily implemented. */
133
134    error = kern_openat(td, dirfd, path, UIO_SYSSPACE, bsd_flags, mode);
135    if (error != 0)
136	    goto done;
137
138    if (bsd_flags & O_NOCTTY)
139	    goto done;
140
141    /*
142     * XXX In between kern_open() and fget(), another process
143     * having the same filedesc could use that fd without
144     * checking below.
145     */
146    fd = td->td_retval[0];
147    if (fget(td, fd, cap_rights_init(&rights, CAP_IOCTL), &fp) == 0) {
148	    if (fp->f_type != DTYPE_VNODE) {
149		    fdrop(fp, td);
150		    goto done;
151	    }
152	    sx_slock(&proctree_lock);
153	    PROC_LOCK(p);
154	    if (SESS_LEADER(p) && !(p->p_flag & P_CONTROLT)) {
155		    PROC_UNLOCK(p);
156		    sx_sunlock(&proctree_lock);
157		    /* XXXPJD: Verify if TIOCSCTTY is allowed. */
158		    (void) fo_ioctl(fp, TIOCSCTTY, (caddr_t) 0,
159			td->td_ucred, td);
160	    } else {
161		    PROC_UNLOCK(p);
162		    sx_sunlock(&proctree_lock);
163	    }
164	    fdrop(fp, td);
165    }
166
167done:
168#ifdef DEBUG
169    if (ldebug(open))
170	    printf(LMSG("open returns error %d"), error);
171#endif
172    LFREEPATH(path);
173    return (error);
174}
175
176int
177linux_openat(struct thread *td, struct linux_openat_args *args)
178{
179	char *path;
180	int dfd;
181
182	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
183	if (args->flags & LINUX_O_CREAT)
184		LCONVPATH_AT(td, args->filename, &path, 1, dfd);
185	else
186		LCONVPATH_AT(td, args->filename, &path, 0, dfd);
187#ifdef DEBUG
188	if (ldebug(openat))
189		printf(ARGS(openat, "%i, %s, 0x%x, 0x%x"), args->dfd,
190		    path, args->flags, args->mode);
191#endif
192	return (linux_common_open(td, dfd, path, args->flags, args->mode));
193}
194
195int
196linux_open(struct thread *td, struct linux_open_args *args)
197{
198    char *path;
199
200    if (args->flags & LINUX_O_CREAT)
201	LCONVPATHCREAT(td, args->path, &path);
202    else
203	LCONVPATHEXIST(td, args->path, &path);
204
205#ifdef DEBUG
206	if (ldebug(open))
207		printf(ARGS(open, "%s, 0x%x, 0x%x"),
208		    path, args->flags, args->mode);
209#endif
210
211	return (linux_common_open(td, AT_FDCWD, path, args->flags, args->mode));
212}
213
214int
215linux_lseek(struct thread *td, struct linux_lseek_args *args)
216{
217
218    struct lseek_args /* {
219	int fd;
220	int pad;
221	off_t offset;
222	int whence;
223    } */ tmp_args;
224    int error;
225
226#ifdef DEBUG
227	if (ldebug(lseek))
228		printf(ARGS(lseek, "%d, %ld, %d"),
229		    args->fdes, (long)args->off, args->whence);
230#endif
231    tmp_args.fd = args->fdes;
232    tmp_args.offset = (off_t)args->off;
233    tmp_args.whence = args->whence;
234    error = sys_lseek(td, &tmp_args);
235    return error;
236}
237
238#if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
239int
240linux_llseek(struct thread *td, struct linux_llseek_args *args)
241{
242	struct lseek_args bsd_args;
243	int error;
244	off_t off;
245
246#ifdef DEBUG
247	if (ldebug(llseek))
248		printf(ARGS(llseek, "%d, %d:%d, %d"),
249		    args->fd, args->ohigh, args->olow, args->whence);
250#endif
251	off = (args->olow) | (((off_t) args->ohigh) << 32);
252
253	bsd_args.fd = args->fd;
254	bsd_args.offset = off;
255	bsd_args.whence = args->whence;
256
257	if ((error = sys_lseek(td, &bsd_args)))
258		return error;
259
260	if ((error = copyout(td->td_retval, args->res, sizeof (off_t))))
261		return error;
262
263	td->td_retval[0] = 0;
264	return 0;
265}
266
267int
268linux_readdir(struct thread *td, struct linux_readdir_args *args)
269{
270	struct linux_getdents_args lda;
271
272	lda.fd = args->fd;
273	lda.dent = args->dent;
274	lda.count = 1;
275	return linux_getdents(td, &lda);
276}
277#endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
278
279/*
280 * Note that linux_getdents(2) and linux_getdents64(2) have the same
281 * arguments. They only differ in the definition of struct dirent they
282 * operate on. We use this to common the code, with the exception of
283 * accessing struct dirent. Note that linux_readdir(2) is implemented
284 * by means of linux_getdents(2). In this case we never operate on
285 * struct dirent64 and thus don't need to handle it...
286 */
287
288struct l_dirent {
289	l_ulong		d_ino;
290	l_off_t		d_off;
291	l_ushort	d_reclen;
292	char		d_name[LINUX_NAME_MAX + 1];
293};
294
295struct l_dirent64 {
296	uint64_t	d_ino;
297	int64_t		d_off;
298	l_ushort	d_reclen;
299	u_char		d_type;
300	char		d_name[LINUX_NAME_MAX + 1];
301};
302
303/*
304 * Linux uses the last byte in the dirent buffer to store d_type,
305 * at least glibc-2.7 requires it. That is why l_dirent is padded with 2 bytes.
306 */
307#define LINUX_RECLEN(namlen)						\
308    roundup((offsetof(struct l_dirent, d_name) + (namlen) + 2),		\
309    sizeof(l_ulong))
310
311#define LINUX_RECLEN64(namlen)						\
312    roundup((offsetof(struct l_dirent64, d_name) + (namlen) + 1),	\
313    sizeof(uint64_t))
314
315#define LINUX_MAXRECLEN		max(LINUX_RECLEN(LINUX_NAME_MAX),	\
316				    LINUX_RECLEN64(LINUX_NAME_MAX))
317#define	LINUX_DIRBLKSIZ		512
318
319static int
320getdents_common(struct thread *td, struct linux_getdents64_args *args,
321    int is64bit)
322{
323	struct dirent *bdp;
324	struct vnode *vp;
325	caddr_t inp, buf;		/* BSD-format */
326	int len, reclen;		/* BSD-format */
327	caddr_t outp;			/* Linux-format */
328	int resid, linuxreclen=0;	/* Linux-format */
329	caddr_t lbuf;			/* Linux-format */
330	cap_rights_t rights;
331	struct file *fp;
332	struct uio auio;
333	struct iovec aiov;
334	off_t off;
335	struct l_dirent *linux_dirent;
336	struct l_dirent64 *linux_dirent64;
337	int buflen, error, eofflag, nbytes, justone;
338	u_long *cookies = NULL, *cookiep;
339	int ncookies;
340
341	nbytes = args->count;
342	if (nbytes == 1) {
343		/* readdir(2) case. Always struct dirent. */
344		if (is64bit)
345			return (EINVAL);
346		nbytes = sizeof(*linux_dirent);
347		justone = 1;
348	} else
349		justone = 0;
350
351	error = getvnode(td->td_proc->p_fd, args->fd,
352	    cap_rights_init(&rights, CAP_READ), &fp);
353	if (error != 0)
354		return (error);
355
356	if ((fp->f_flag & FREAD) == 0) {
357		fdrop(fp, td);
358		return (EBADF);
359	}
360
361	off = foffset_lock(fp, 0);
362	vp = fp->f_vnode;
363	if (vp->v_type != VDIR) {
364		foffset_unlock(fp, off, 0);
365		fdrop(fp, td);
366		return (EINVAL);
367	}
368
369
370	buflen = max(LINUX_DIRBLKSIZ, nbytes);
371	buflen = min(buflen, MAXBSIZE);
372	buf = malloc(buflen, M_TEMP, M_WAITOK);
373	lbuf = malloc(LINUX_MAXRECLEN, M_TEMP, M_WAITOK | M_ZERO);
374	vn_lock(vp, LK_SHARED | LK_RETRY);
375
376	aiov.iov_base = buf;
377	aiov.iov_len = buflen;
378	auio.uio_iov = &aiov;
379	auio.uio_iovcnt = 1;
380	auio.uio_rw = UIO_READ;
381	auio.uio_segflg = UIO_SYSSPACE;
382	auio.uio_td = td;
383	auio.uio_resid = buflen;
384	auio.uio_offset = off;
385
386#ifdef MAC
387	/*
388	 * Do directory search MAC check using non-cached credentials.
389	 */
390	if ((error = mac_vnode_check_readdir(td->td_ucred, vp)))
391		goto out;
392#endif /* MAC */
393	if ((error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, &ncookies,
394		 &cookies)))
395		goto out;
396
397	inp = buf;
398	outp = (caddr_t)args->dirent;
399	resid = nbytes;
400	if ((len = buflen - auio.uio_resid) <= 0)
401		goto eof;
402
403	cookiep = cookies;
404
405	if (cookies) {
406		/*
407		 * When using cookies, the vfs has the option of reading from
408		 * a different offset than that supplied (UFS truncates the
409		 * offset to a block boundary to make sure that it never reads
410		 * partway through a directory entry, even if the directory
411		 * has been compacted).
412		 */
413		while (len > 0 && ncookies > 0 && *cookiep <= off) {
414			bdp = (struct dirent *) inp;
415			len -= bdp->d_reclen;
416			inp += bdp->d_reclen;
417			cookiep++;
418			ncookies--;
419		}
420	}
421
422	while (len > 0) {
423		if (cookiep && ncookies == 0)
424			break;
425		bdp = (struct dirent *) inp;
426		reclen = bdp->d_reclen;
427		if (reclen & 3) {
428			error = EFAULT;
429			goto out;
430		}
431
432		if (bdp->d_fileno == 0) {
433			inp += reclen;
434			if (cookiep) {
435				off = *cookiep++;
436				ncookies--;
437			} else
438				off += reclen;
439
440			len -= reclen;
441			continue;
442		}
443
444		linuxreclen = (is64bit)
445		    ? LINUX_RECLEN64(bdp->d_namlen)
446		    : LINUX_RECLEN(bdp->d_namlen);
447
448		if (reclen > len || resid < linuxreclen) {
449			outp++;
450			break;
451		}
452
453		if (justone) {
454			/* readdir(2) case. */
455			linux_dirent = (struct l_dirent*)lbuf;
456			linux_dirent->d_ino = bdp->d_fileno;
457			linux_dirent->d_off = (l_off_t)linuxreclen;
458			linux_dirent->d_reclen = (l_ushort)bdp->d_namlen;
459			strlcpy(linux_dirent->d_name, bdp->d_name,
460			    linuxreclen - offsetof(struct l_dirent, d_name));
461			error = copyout(linux_dirent, outp, linuxreclen);
462		}
463		if (is64bit) {
464			linux_dirent64 = (struct l_dirent64*)lbuf;
465			linux_dirent64->d_ino = bdp->d_fileno;
466			linux_dirent64->d_off = (cookiep)
467			    ? (l_off_t)*cookiep
468			    : (l_off_t)(off + reclen);
469			linux_dirent64->d_reclen = (l_ushort)linuxreclen;
470			linux_dirent64->d_type = bdp->d_type;
471			strlcpy(linux_dirent64->d_name, bdp->d_name,
472			    linuxreclen - offsetof(struct l_dirent64, d_name));
473			error = copyout(linux_dirent64, outp, linuxreclen);
474		} else if (!justone) {
475			linux_dirent = (struct l_dirent*)lbuf;
476			linux_dirent->d_ino = bdp->d_fileno;
477			linux_dirent->d_off = (cookiep)
478			    ? (l_off_t)*cookiep
479			    : (l_off_t)(off + reclen);
480			linux_dirent->d_reclen = (l_ushort)linuxreclen;
481			/*
482			 * Copy d_type to last byte of l_dirent buffer
483			 */
484			lbuf[linuxreclen-1] = bdp->d_type;
485			strlcpy(linux_dirent->d_name, bdp->d_name,
486			    linuxreclen - offsetof(struct l_dirent, d_name)-1);
487			error = copyout(linux_dirent, outp, linuxreclen);
488		}
489
490		if (error)
491			goto out;
492
493		inp += reclen;
494		if (cookiep) {
495			off = *cookiep++;
496			ncookies--;
497		} else
498			off += reclen;
499
500		outp += linuxreclen;
501		resid -= linuxreclen;
502		len -= reclen;
503		if (justone)
504			break;
505	}
506
507	if (outp == (caddr_t)args->dirent) {
508		nbytes = resid;
509		goto eof;
510	}
511
512	if (justone)
513		nbytes = resid + linuxreclen;
514
515eof:
516	td->td_retval[0] = nbytes - resid;
517
518out:
519	free(cookies, M_TEMP);
520
521	VOP_UNLOCK(vp, 0);
522	foffset_unlock(fp, off, 0);
523	fdrop(fp, td);
524	free(buf, M_TEMP);
525	free(lbuf, M_TEMP);
526	return (error);
527}
528
529int
530linux_getdents(struct thread *td, struct linux_getdents_args *args)
531{
532
533#ifdef DEBUG
534	if (ldebug(getdents))
535		printf(ARGS(getdents, "%d, *, %d"), args->fd, args->count);
536#endif
537
538	return (getdents_common(td, (struct linux_getdents64_args*)args, 0));
539}
540
541int
542linux_getdents64(struct thread *td, struct linux_getdents64_args *args)
543{
544
545#ifdef DEBUG
546	if (ldebug(getdents64))
547		printf(ARGS(getdents64, "%d, *, %d"), args->fd, args->count);
548#endif
549
550	return (getdents_common(td, args, 1));
551}
552
553/*
554 * These exist mainly for hooks for doing /compat/linux translation.
555 */
556
557int
558linux_access(struct thread *td, struct linux_access_args *args)
559{
560	char *path;
561	int error;
562
563	/* linux convention */
564	if (args->amode & ~(F_OK | X_OK | W_OK | R_OK))
565		return (EINVAL);
566
567	LCONVPATHEXIST(td, args->path, &path);
568
569#ifdef DEBUG
570	if (ldebug(access))
571		printf(ARGS(access, "%s, %d"), path, args->amode);
572#endif
573	error = kern_access(td, path, UIO_SYSSPACE, args->amode);
574	LFREEPATH(path);
575
576	return (error);
577}
578
579int
580linux_faccessat(struct thread *td, struct linux_faccessat_args *args)
581{
582	char *path;
583	int error, dfd, flag;
584
585	if (args->flag & ~LINUX_AT_EACCESS)
586		return (EINVAL);
587	/* linux convention */
588	if (args->amode & ~(F_OK | X_OK | W_OK | R_OK))
589		return (EINVAL);
590
591	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
592	LCONVPATHEXIST_AT(td, args->filename, &path, dfd);
593
594#ifdef DEBUG
595	if (ldebug(access))
596		printf(ARGS(access, "%s, %d"), path, args->amode);
597#endif
598
599	flag = (args->flag & LINUX_AT_EACCESS) == 0 ? 0 : AT_EACCESS;
600	error = kern_accessat(td, dfd, path, UIO_SYSSPACE, flag, args->amode);
601	LFREEPATH(path);
602
603	return (error);
604}
605
606int
607linux_unlink(struct thread *td, struct linux_unlink_args *args)
608{
609	char *path;
610	int error;
611	struct stat st;
612
613	LCONVPATHEXIST(td, args->path, &path);
614
615#ifdef DEBUG
616	if (ldebug(unlink))
617		printf(ARGS(unlink, "%s"), path);
618#endif
619
620	error = kern_unlink(td, path, UIO_SYSSPACE);
621	if (error == EPERM)
622		/* Introduce POSIX noncompliant behaviour of Linux */
623		if (kern_stat(td, path, UIO_SYSSPACE, &st) == 0)
624			if (S_ISDIR(st.st_mode))
625				error = EISDIR;
626	LFREEPATH(path);
627	return (error);
628}
629
630int
631linux_unlinkat(struct thread *td, struct linux_unlinkat_args *args)
632{
633	char *path;
634	int error, dfd;
635	struct stat st;
636
637	if (args->flag & ~LINUX_AT_REMOVEDIR)
638		return (EINVAL);
639
640	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
641	LCONVPATHEXIST_AT(td, args->pathname, &path, dfd);
642
643#ifdef DEBUG
644	if (ldebug(unlinkat))
645		printf(ARGS(unlinkat, "%s"), path);
646#endif
647
648	if (args->flag & LINUX_AT_REMOVEDIR)
649		error = kern_rmdirat(td, dfd, path, UIO_SYSSPACE);
650	else
651		error = kern_unlinkat(td, dfd, path, UIO_SYSSPACE, 0);
652	if (error == EPERM && !(args->flag & LINUX_AT_REMOVEDIR)) {
653		/* Introduce POSIX noncompliant behaviour of Linux */
654		if (kern_statat(td, AT_SYMLINK_NOFOLLOW, dfd, path,
655		    UIO_SYSSPACE, &st) == 0 && S_ISDIR(st.st_mode))
656			error = EISDIR;
657	}
658	LFREEPATH(path);
659	return (error);
660}
661int
662linux_chdir(struct thread *td, struct linux_chdir_args *args)
663{
664	char *path;
665	int error;
666
667	LCONVPATHEXIST(td, args->path, &path);
668
669#ifdef DEBUG
670	if (ldebug(chdir))
671		printf(ARGS(chdir, "%s"), path);
672#endif
673	error = kern_chdir(td, path, UIO_SYSSPACE);
674	LFREEPATH(path);
675	return (error);
676}
677
678int
679linux_chmod(struct thread *td, struct linux_chmod_args *args)
680{
681	char *path;
682	int error;
683
684	LCONVPATHEXIST(td, args->path, &path);
685
686#ifdef DEBUG
687	if (ldebug(chmod))
688		printf(ARGS(chmod, "%s, %d"), path, args->mode);
689#endif
690	error = kern_chmod(td, path, UIO_SYSSPACE, args->mode);
691	LFREEPATH(path);
692	return (error);
693}
694
695int
696linux_fchmodat(struct thread *td, struct linux_fchmodat_args *args)
697{
698	char *path;
699	int error, dfd;
700
701	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
702	LCONVPATHEXIST_AT(td, args->filename, &path, dfd);
703
704#ifdef DEBUG
705	if (ldebug(fchmodat))
706		printf(ARGS(fchmodat, "%s, %d"), path, args->mode);
707#endif
708
709	error = kern_fchmodat(td, dfd, path, UIO_SYSSPACE, args->mode, 0);
710	LFREEPATH(path);
711	return (error);
712}
713
714int
715linux_mkdir(struct thread *td, struct linux_mkdir_args *args)
716{
717	char *path;
718	int error;
719
720	LCONVPATHCREAT(td, args->path, &path);
721
722#ifdef DEBUG
723	if (ldebug(mkdir))
724		printf(ARGS(mkdir, "%s, %d"), path, args->mode);
725#endif
726	error = kern_mkdir(td, path, UIO_SYSSPACE, args->mode);
727	LFREEPATH(path);
728	return (error);
729}
730
731int
732linux_mkdirat(struct thread *td, struct linux_mkdirat_args *args)
733{
734	char *path;
735	int error, dfd;
736
737	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
738	LCONVPATHCREAT_AT(td, args->pathname, &path, dfd);
739
740#ifdef DEBUG
741	if (ldebug(mkdirat))
742		printf(ARGS(mkdirat, "%s, %d"), path, args->mode);
743#endif
744	error = kern_mkdirat(td, dfd, path, UIO_SYSSPACE, args->mode);
745	LFREEPATH(path);
746	return (error);
747}
748
749int
750linux_rmdir(struct thread *td, struct linux_rmdir_args *args)
751{
752	char *path;
753	int error;
754
755	LCONVPATHEXIST(td, args->path, &path);
756
757#ifdef DEBUG
758	if (ldebug(rmdir))
759		printf(ARGS(rmdir, "%s"), path);
760#endif
761	error = kern_rmdir(td, path, UIO_SYSSPACE);
762	LFREEPATH(path);
763	return (error);
764}
765
766int
767linux_rename(struct thread *td, struct linux_rename_args *args)
768{
769	char *from, *to;
770	int error;
771
772	LCONVPATHEXIST(td, args->from, &from);
773	/* Expand LCONVPATHCREATE so that `from' can be freed on errors */
774	error = linux_emul_convpath(td, args->to, UIO_USERSPACE, &to, 1, AT_FDCWD);
775	if (to == NULL) {
776		LFREEPATH(from);
777		return (error);
778	}
779
780#ifdef DEBUG
781	if (ldebug(rename))
782		printf(ARGS(rename, "%s, %s"), from, to);
783#endif
784	error = kern_rename(td, from, to, UIO_SYSSPACE);
785	LFREEPATH(from);
786	LFREEPATH(to);
787	return (error);
788}
789
790int
791linux_renameat(struct thread *td, struct linux_renameat_args *args)
792{
793	char *from, *to;
794	int error, olddfd, newdfd;
795
796	olddfd = (args->olddfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->olddfd;
797	newdfd = (args->newdfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->newdfd;
798	LCONVPATHEXIST_AT(td, args->oldname, &from, olddfd);
799	/* Expand LCONVPATHCREATE so that `from' can be freed on errors */
800	error = linux_emul_convpath(td, args->newname, UIO_USERSPACE, &to, 1, newdfd);
801	if (to == NULL) {
802		LFREEPATH(from);
803		return (error);
804	}
805
806#ifdef DEBUG
807	if (ldebug(renameat))
808		printf(ARGS(renameat, "%s, %s"), from, to);
809#endif
810	error = kern_renameat(td, olddfd, from, newdfd, to, UIO_SYSSPACE);
811	LFREEPATH(from);
812	LFREEPATH(to);
813	return (error);
814}
815
816int
817linux_symlink(struct thread *td, struct linux_symlink_args *args)
818{
819	char *path, *to;
820	int error;
821
822	LCONVPATHEXIST(td, args->path, &path);
823	/* Expand LCONVPATHCREATE so that `path' can be freed on errors */
824	error = linux_emul_convpath(td, args->to, UIO_USERSPACE, &to, 1, AT_FDCWD);
825	if (to == NULL) {
826		LFREEPATH(path);
827		return (error);
828	}
829
830#ifdef DEBUG
831	if (ldebug(symlink))
832		printf(ARGS(symlink, "%s, %s"), path, to);
833#endif
834	error = kern_symlink(td, path, to, UIO_SYSSPACE);
835	LFREEPATH(path);
836	LFREEPATH(to);
837	return (error);
838}
839
840int
841linux_symlinkat(struct thread *td, struct linux_symlinkat_args *args)
842{
843	char *path, *to;
844	int error, dfd;
845
846	dfd = (args->newdfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->newdfd;
847	LCONVPATHEXIST_AT(td, args->oldname, &path, dfd);
848	/* Expand LCONVPATHCREATE so that `path' can be freed on errors */
849	error = linux_emul_convpath(td, args->newname, UIO_USERSPACE, &to, 1, dfd);
850	if (to == NULL) {
851		LFREEPATH(path);
852		return (error);
853	}
854
855#ifdef DEBUG
856	if (ldebug(symlinkat))
857		printf(ARGS(symlinkat, "%s, %s"), path, to);
858#endif
859
860	error = kern_symlinkat(td, path, dfd, to, UIO_SYSSPACE);
861	LFREEPATH(path);
862	LFREEPATH(to);
863	return (error);
864}
865
866int
867linux_readlink(struct thread *td, struct linux_readlink_args *args)
868{
869	char *name;
870	int error;
871
872	LCONVPATHEXIST(td, args->name, &name);
873
874#ifdef DEBUG
875	if (ldebug(readlink))
876		printf(ARGS(readlink, "%s, %p, %d"), name, (void *)args->buf,
877		    args->count);
878#endif
879	error = kern_readlink(td, name, UIO_SYSSPACE, args->buf, UIO_USERSPACE,
880	    args->count);
881	LFREEPATH(name);
882	return (error);
883}
884
885int
886linux_readlinkat(struct thread *td, struct linux_readlinkat_args *args)
887{
888	char *name;
889	int error, dfd;
890
891	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
892	LCONVPATHEXIST_AT(td, args->path, &name, dfd);
893
894#ifdef DEBUG
895	if (ldebug(readlinkat))
896		printf(ARGS(readlinkat, "%s, %p, %d"), name, (void *)args->buf,
897		    args->bufsiz);
898#endif
899
900	error = kern_readlinkat(td, dfd, name, UIO_SYSSPACE, args->buf,
901	    UIO_USERSPACE, args->bufsiz);
902	LFREEPATH(name);
903	return (error);
904}
905
906int
907linux_truncate(struct thread *td, struct linux_truncate_args *args)
908{
909	char *path;
910	int error;
911
912	LCONVPATHEXIST(td, args->path, &path);
913
914#ifdef DEBUG
915	if (ldebug(truncate))
916		printf(ARGS(truncate, "%s, %ld"), path, (long)args->length);
917#endif
918
919	error = kern_truncate(td, path, UIO_SYSSPACE, args->length);
920	LFREEPATH(path);
921	return (error);
922}
923
924#if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
925int
926linux_truncate64(struct thread *td, struct linux_truncate64_args *args)
927{
928	char *path;
929	int error;
930
931	LCONVPATHEXIST(td, args->path, &path);
932
933#ifdef DEBUG
934	if (ldebug(truncate64))
935		printf(ARGS(truncate64, "%s, %jd"), path, args->length);
936#endif
937
938	error = kern_truncate(td, path, UIO_SYSSPACE, args->length);
939	LFREEPATH(path);
940	return (error);
941}
942#endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
943
944int
945linux_ftruncate(struct thread *td, struct linux_ftruncate_args *args)
946{
947	struct ftruncate_args /* {
948		int fd;
949		int pad;
950		off_t length;
951		} */ nuap;
952
953	nuap.fd = args->fd;
954	nuap.length = args->length;
955	return (sys_ftruncate(td, &nuap));
956}
957
958int
959linux_link(struct thread *td, struct linux_link_args *args)
960{
961	char *path, *to;
962	int error;
963
964	LCONVPATHEXIST(td, args->path, &path);
965	/* Expand LCONVPATHCREATE so that `path' can be freed on errors */
966	error = linux_emul_convpath(td, args->to, UIO_USERSPACE, &to, 1, AT_FDCWD);
967	if (to == NULL) {
968		LFREEPATH(path);
969		return (error);
970	}
971
972#ifdef DEBUG
973	if (ldebug(link))
974		printf(ARGS(link, "%s, %s"), path, to);
975#endif
976	error = kern_link(td, path, to, UIO_SYSSPACE);
977	LFREEPATH(path);
978	LFREEPATH(to);
979	return (error);
980}
981
982int
983linux_linkat(struct thread *td, struct linux_linkat_args *args)
984{
985	char *path, *to;
986	int error, olddfd, newdfd, follow;
987
988	if (args->flag & ~LINUX_AT_SYMLINK_FOLLOW)
989		return (EINVAL);
990
991	olddfd = (args->olddfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->olddfd;
992	newdfd = (args->newdfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->newdfd;
993	LCONVPATHEXIST_AT(td, args->oldname, &path, olddfd);
994	/* Expand LCONVPATHCREATE so that `path' can be freed on errors */
995	error = linux_emul_convpath(td, args->newname, UIO_USERSPACE, &to, 1, newdfd);
996	if (to == NULL) {
997		LFREEPATH(path);
998		return (error);
999	}
1000
1001#ifdef DEBUG
1002	if (ldebug(linkat))
1003		printf(ARGS(linkat, "%i, %s, %i, %s, %i"), args->olddfd, path,
1004			args->newdfd, to, args->flag);
1005#endif
1006
1007	follow = (args->flag & LINUX_AT_SYMLINK_FOLLOW) == 0 ? NOFOLLOW :
1008	    FOLLOW;
1009	error = kern_linkat(td, olddfd, newdfd, path, to, UIO_SYSSPACE, follow);
1010	LFREEPATH(path);
1011	LFREEPATH(to);
1012	return (error);
1013}
1014
1015int
1016linux_fdatasync(td, uap)
1017	struct thread *td;
1018	struct linux_fdatasync_args *uap;
1019{
1020	struct fsync_args bsd;
1021
1022	bsd.fd = uap->fd;
1023	return sys_fsync(td, &bsd);
1024}
1025
1026int
1027linux_pread(td, uap)
1028	struct thread *td;
1029	struct linux_pread_args *uap;
1030{
1031	struct pread_args bsd;
1032	cap_rights_t rights;
1033	struct vnode *vp;
1034	int error;
1035
1036	bsd.fd = uap->fd;
1037	bsd.buf = uap->buf;
1038	bsd.nbyte = uap->nbyte;
1039	bsd.offset = uap->offset;
1040
1041	error = sys_pread(td, &bsd);
1042
1043	if (error == 0) {
1044		/* This seems to violate POSIX but linux does it */
1045		error = fgetvp(td, uap->fd,
1046		    cap_rights_init(&rights, CAP_PREAD), &vp);
1047		if (error != 0)
1048			return (error);
1049		if (vp->v_type == VDIR) {
1050			vrele(vp);
1051			return (EISDIR);
1052		}
1053		vrele(vp);
1054	}
1055
1056	return (error);
1057}
1058
1059int
1060linux_pwrite(td, uap)
1061	struct thread *td;
1062	struct linux_pwrite_args *uap;
1063{
1064	struct pwrite_args bsd;
1065
1066	bsd.fd = uap->fd;
1067	bsd.buf = uap->buf;
1068	bsd.nbyte = uap->nbyte;
1069	bsd.offset = uap->offset;
1070	return sys_pwrite(td, &bsd);
1071}
1072
1073int
1074linux_mount(struct thread *td, struct linux_mount_args *args)
1075{
1076	char fstypename[MFSNAMELEN];
1077	char mntonname[MNAMELEN], mntfromname[MNAMELEN];
1078	int error;
1079	int fsflags;
1080
1081	error = copyinstr(args->filesystemtype, fstypename, MFSNAMELEN - 1,
1082	    NULL);
1083	if (error)
1084		return (error);
1085	error = copyinstr(args->specialfile, mntfromname, MNAMELEN - 1, NULL);
1086	if (error)
1087		return (error);
1088	error = copyinstr(args->dir, mntonname, MNAMELEN - 1, NULL);
1089	if (error)
1090		return (error);
1091
1092#ifdef DEBUG
1093	if (ldebug(mount))
1094		printf(ARGS(mount, "%s, %s, %s"),
1095		    fstypename, mntfromname, mntonname);
1096#endif
1097
1098	if (strcmp(fstypename, "ext2") == 0) {
1099		strcpy(fstypename, "ext2fs");
1100	} else if (strcmp(fstypename, "proc") == 0) {
1101		strcpy(fstypename, "linprocfs");
1102	} else if (strcmp(fstypename, "vfat") == 0) {
1103		strcpy(fstypename, "msdosfs");
1104	}
1105
1106	fsflags = 0;
1107
1108	if ((args->rwflag & 0xffff0000) == 0xc0ed0000) {
1109		/*
1110		 * Linux SYNC flag is not included; the closest equivalent
1111		 * FreeBSD has is !ASYNC, which is our default.
1112		 */
1113		if (args->rwflag & LINUX_MS_RDONLY)
1114			fsflags |= MNT_RDONLY;
1115		if (args->rwflag & LINUX_MS_NOSUID)
1116			fsflags |= MNT_NOSUID;
1117		if (args->rwflag & LINUX_MS_NOEXEC)
1118			fsflags |= MNT_NOEXEC;
1119		if (args->rwflag & LINUX_MS_REMOUNT)
1120			fsflags |= MNT_UPDATE;
1121	}
1122
1123	error = kernel_vmount(fsflags,
1124	    "fstype", fstypename,
1125	    "fspath", mntonname,
1126	    "from", mntfromname,
1127	    NULL);
1128	return (error);
1129}
1130
1131#if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1132int
1133linux_oldumount(struct thread *td, struct linux_oldumount_args *args)
1134{
1135	struct linux_umount_args args2;
1136
1137	args2.path = args->path;
1138	args2.flags = 0;
1139	return (linux_umount(td, &args2));
1140}
1141#endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1142
1143int
1144linux_umount(struct thread *td, struct linux_umount_args *args)
1145{
1146	struct unmount_args bsd;
1147
1148	bsd.path = args->path;
1149	bsd.flags = args->flags;	/* XXX correct? */
1150	return (sys_unmount(td, &bsd));
1151}
1152
1153/*
1154 * fcntl family of syscalls
1155 */
1156
1157struct l_flock {
1158	l_short		l_type;
1159	l_short		l_whence;
1160	l_off_t		l_start;
1161	l_off_t		l_len;
1162	l_pid_t		l_pid;
1163}
1164#if defined(__amd64__) && defined(COMPAT_LINUX32)
1165__packed
1166#endif
1167;
1168
1169static void
1170linux_to_bsd_flock(struct l_flock *linux_flock, struct flock *bsd_flock)
1171{
1172	switch (linux_flock->l_type) {
1173	case LINUX_F_RDLCK:
1174		bsd_flock->l_type = F_RDLCK;
1175		break;
1176	case LINUX_F_WRLCK:
1177		bsd_flock->l_type = F_WRLCK;
1178		break;
1179	case LINUX_F_UNLCK:
1180		bsd_flock->l_type = F_UNLCK;
1181		break;
1182	default:
1183		bsd_flock->l_type = -1;
1184		break;
1185	}
1186	bsd_flock->l_whence = linux_flock->l_whence;
1187	bsd_flock->l_start = (off_t)linux_flock->l_start;
1188	bsd_flock->l_len = (off_t)linux_flock->l_len;
1189	bsd_flock->l_pid = (pid_t)linux_flock->l_pid;
1190	bsd_flock->l_sysid = 0;
1191}
1192
1193static void
1194bsd_to_linux_flock(struct flock *bsd_flock, struct l_flock *linux_flock)
1195{
1196	switch (bsd_flock->l_type) {
1197	case F_RDLCK:
1198		linux_flock->l_type = LINUX_F_RDLCK;
1199		break;
1200	case F_WRLCK:
1201		linux_flock->l_type = LINUX_F_WRLCK;
1202		break;
1203	case F_UNLCK:
1204		linux_flock->l_type = LINUX_F_UNLCK;
1205		break;
1206	}
1207	linux_flock->l_whence = bsd_flock->l_whence;
1208	linux_flock->l_start = (l_off_t)bsd_flock->l_start;
1209	linux_flock->l_len = (l_off_t)bsd_flock->l_len;
1210	linux_flock->l_pid = (l_pid_t)bsd_flock->l_pid;
1211}
1212
1213#if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1214struct l_flock64 {
1215	l_short		l_type;
1216	l_short		l_whence;
1217	l_loff_t	l_start;
1218	l_loff_t	l_len;
1219	l_pid_t		l_pid;
1220}
1221#if defined(__amd64__) && defined(COMPAT_LINUX32)
1222__packed
1223#endif
1224;
1225
1226static void
1227linux_to_bsd_flock64(struct l_flock64 *linux_flock, struct flock *bsd_flock)
1228{
1229	switch (linux_flock->l_type) {
1230	case LINUX_F_RDLCK:
1231		bsd_flock->l_type = F_RDLCK;
1232		break;
1233	case LINUX_F_WRLCK:
1234		bsd_flock->l_type = F_WRLCK;
1235		break;
1236	case LINUX_F_UNLCK:
1237		bsd_flock->l_type = F_UNLCK;
1238		break;
1239	default:
1240		bsd_flock->l_type = -1;
1241		break;
1242	}
1243	bsd_flock->l_whence = linux_flock->l_whence;
1244	bsd_flock->l_start = (off_t)linux_flock->l_start;
1245	bsd_flock->l_len = (off_t)linux_flock->l_len;
1246	bsd_flock->l_pid = (pid_t)linux_flock->l_pid;
1247	bsd_flock->l_sysid = 0;
1248}
1249
1250static void
1251bsd_to_linux_flock64(struct flock *bsd_flock, struct l_flock64 *linux_flock)
1252{
1253	switch (bsd_flock->l_type) {
1254	case F_RDLCK:
1255		linux_flock->l_type = LINUX_F_RDLCK;
1256		break;
1257	case F_WRLCK:
1258		linux_flock->l_type = LINUX_F_WRLCK;
1259		break;
1260	case F_UNLCK:
1261		linux_flock->l_type = LINUX_F_UNLCK;
1262		break;
1263	}
1264	linux_flock->l_whence = bsd_flock->l_whence;
1265	linux_flock->l_start = (l_loff_t)bsd_flock->l_start;
1266	linux_flock->l_len = (l_loff_t)bsd_flock->l_len;
1267	linux_flock->l_pid = (l_pid_t)bsd_flock->l_pid;
1268}
1269#endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1270
1271static int
1272fcntl_common(struct thread *td, struct linux_fcntl_args *args)
1273{
1274	struct l_flock linux_flock;
1275	struct flock bsd_flock;
1276	cap_rights_t rights;
1277	struct file *fp;
1278	long arg;
1279	int error, result;
1280
1281	switch (args->cmd) {
1282	case LINUX_F_DUPFD:
1283		return (kern_fcntl(td, args->fd, F_DUPFD, args->arg));
1284
1285	case LINUX_F_GETFD:
1286		return (kern_fcntl(td, args->fd, F_GETFD, 0));
1287
1288	case LINUX_F_SETFD:
1289		return (kern_fcntl(td, args->fd, F_SETFD, args->arg));
1290
1291	case LINUX_F_GETFL:
1292		error = kern_fcntl(td, args->fd, F_GETFL, 0);
1293		result = td->td_retval[0];
1294		td->td_retval[0] = 0;
1295		if (result & O_RDONLY)
1296			td->td_retval[0] |= LINUX_O_RDONLY;
1297		if (result & O_WRONLY)
1298			td->td_retval[0] |= LINUX_O_WRONLY;
1299		if (result & O_RDWR)
1300			td->td_retval[0] |= LINUX_O_RDWR;
1301		if (result & O_NDELAY)
1302			td->td_retval[0] |= LINUX_O_NONBLOCK;
1303		if (result & O_APPEND)
1304			td->td_retval[0] |= LINUX_O_APPEND;
1305		if (result & O_FSYNC)
1306			td->td_retval[0] |= LINUX_O_SYNC;
1307		if (result & O_ASYNC)
1308			td->td_retval[0] |= LINUX_FASYNC;
1309#ifdef LINUX_O_NOFOLLOW
1310		if (result & O_NOFOLLOW)
1311			td->td_retval[0] |= LINUX_O_NOFOLLOW;
1312#endif
1313#ifdef LINUX_O_DIRECT
1314		if (result & O_DIRECT)
1315			td->td_retval[0] |= LINUX_O_DIRECT;
1316#endif
1317		return (error);
1318
1319	case LINUX_F_SETFL:
1320		arg = 0;
1321		if (args->arg & LINUX_O_NDELAY)
1322			arg |= O_NONBLOCK;
1323		if (args->arg & LINUX_O_APPEND)
1324			arg |= O_APPEND;
1325		if (args->arg & LINUX_O_SYNC)
1326			arg |= O_FSYNC;
1327		if (args->arg & LINUX_FASYNC)
1328			arg |= O_ASYNC;
1329#ifdef LINUX_O_NOFOLLOW
1330		if (args->arg & LINUX_O_NOFOLLOW)
1331			arg |= O_NOFOLLOW;
1332#endif
1333#ifdef LINUX_O_DIRECT
1334		if (args->arg & LINUX_O_DIRECT)
1335			arg |= O_DIRECT;
1336#endif
1337		return (kern_fcntl(td, args->fd, F_SETFL, arg));
1338
1339	case LINUX_F_GETLK:
1340		error = copyin((void *)args->arg, &linux_flock,
1341		    sizeof(linux_flock));
1342		if (error)
1343			return (error);
1344		linux_to_bsd_flock(&linux_flock, &bsd_flock);
1345		error = kern_fcntl(td, args->fd, F_GETLK, (intptr_t)&bsd_flock);
1346		if (error)
1347			return (error);
1348		bsd_to_linux_flock(&bsd_flock, &linux_flock);
1349		return (copyout(&linux_flock, (void *)args->arg,
1350		    sizeof(linux_flock)));
1351
1352	case LINUX_F_SETLK:
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		return (kern_fcntl(td, args->fd, F_SETLK,
1359		    (intptr_t)&bsd_flock));
1360
1361	case LINUX_F_SETLKW:
1362		error = copyin((void *)args->arg, &linux_flock,
1363		    sizeof(linux_flock));
1364		if (error)
1365			return (error);
1366		linux_to_bsd_flock(&linux_flock, &bsd_flock);
1367		return (kern_fcntl(td, args->fd, F_SETLKW,
1368		     (intptr_t)&bsd_flock));
1369
1370	case LINUX_F_GETOWN:
1371		return (kern_fcntl(td, args->fd, F_GETOWN, 0));
1372
1373	case LINUX_F_SETOWN:
1374		/*
1375		 * XXX some Linux applications depend on F_SETOWN having no
1376		 * significant effect for pipes (SIGIO is not delivered for
1377		 * pipes under Linux-2.2.35 at least).
1378		 */
1379		error = fget(td, args->fd,
1380		    cap_rights_init(&rights, CAP_FCNTL), &fp);
1381		if (error)
1382			return (error);
1383		if (fp->f_type == DTYPE_PIPE) {
1384			fdrop(fp, td);
1385			return (EINVAL);
1386		}
1387		fdrop(fp, td);
1388
1389		return (kern_fcntl(td, args->fd, F_SETOWN, args->arg));
1390	}
1391
1392	return (EINVAL);
1393}
1394
1395int
1396linux_fcntl(struct thread *td, struct linux_fcntl_args *args)
1397{
1398
1399#ifdef DEBUG
1400	if (ldebug(fcntl))
1401		printf(ARGS(fcntl, "%d, %08x, *"), args->fd, args->cmd);
1402#endif
1403
1404	return (fcntl_common(td, args));
1405}
1406
1407#if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1408int
1409linux_fcntl64(struct thread *td, struct linux_fcntl64_args *args)
1410{
1411	struct l_flock64 linux_flock;
1412	struct flock bsd_flock;
1413	struct linux_fcntl_args fcntl_args;
1414	int error;
1415
1416#ifdef DEBUG
1417	if (ldebug(fcntl64))
1418		printf(ARGS(fcntl64, "%d, %08x, *"), args->fd, args->cmd);
1419#endif
1420
1421	switch (args->cmd) {
1422	case LINUX_F_GETLK64:
1423		error = copyin((void *)args->arg, &linux_flock,
1424		    sizeof(linux_flock));
1425		if (error)
1426			return (error);
1427		linux_to_bsd_flock64(&linux_flock, &bsd_flock);
1428		error = kern_fcntl(td, args->fd, F_GETLK, (intptr_t)&bsd_flock);
1429		if (error)
1430			return (error);
1431		bsd_to_linux_flock64(&bsd_flock, &linux_flock);
1432		return (copyout(&linux_flock, (void *)args->arg,
1433			    sizeof(linux_flock)));
1434
1435	case LINUX_F_SETLK64:
1436		error = copyin((void *)args->arg, &linux_flock,
1437		    sizeof(linux_flock));
1438		if (error)
1439			return (error);
1440		linux_to_bsd_flock64(&linux_flock, &bsd_flock);
1441		return (kern_fcntl(td, args->fd, F_SETLK,
1442		    (intptr_t)&bsd_flock));
1443
1444	case LINUX_F_SETLKW64:
1445		error = copyin((void *)args->arg, &linux_flock,
1446		    sizeof(linux_flock));
1447		if (error)
1448			return (error);
1449		linux_to_bsd_flock64(&linux_flock, &bsd_flock);
1450		return (kern_fcntl(td, args->fd, F_SETLKW,
1451		    (intptr_t)&bsd_flock));
1452	}
1453
1454	fcntl_args.fd = args->fd;
1455	fcntl_args.cmd = args->cmd;
1456	fcntl_args.arg = args->arg;
1457	return (fcntl_common(td, &fcntl_args));
1458}
1459#endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1460
1461int
1462linux_chown(struct thread *td, struct linux_chown_args *args)
1463{
1464	char *path;
1465	int error;
1466
1467	LCONVPATHEXIST(td, args->path, &path);
1468
1469#ifdef DEBUG
1470	if (ldebug(chown))
1471		printf(ARGS(chown, "%s, %d, %d"), path, args->uid, args->gid);
1472#endif
1473	error = kern_chown(td, path, UIO_SYSSPACE, args->uid, args->gid);
1474	LFREEPATH(path);
1475	return (error);
1476}
1477
1478int
1479linux_fchownat(struct thread *td, struct linux_fchownat_args *args)
1480{
1481	char *path;
1482	int error, dfd, flag;
1483
1484	if (args->flag & ~LINUX_AT_SYMLINK_NOFOLLOW)
1485		return (EINVAL);
1486
1487	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD :  args->dfd;
1488	LCONVPATHEXIST_AT(td, args->filename, &path, dfd);
1489
1490#ifdef DEBUG
1491	if (ldebug(fchownat))
1492		printf(ARGS(fchownat, "%s, %d, %d"), path, args->uid, args->gid);
1493#endif
1494
1495	flag = (args->flag & LINUX_AT_SYMLINK_NOFOLLOW) == 0 ? 0 :
1496	    AT_SYMLINK_NOFOLLOW;
1497	error = kern_fchownat(td, dfd, path, UIO_SYSSPACE, args->uid, args->gid,
1498	    flag);
1499	LFREEPATH(path);
1500	return (error);
1501}
1502
1503int
1504linux_lchown(struct thread *td, struct linux_lchown_args *args)
1505{
1506	char *path;
1507	int error;
1508
1509	LCONVPATHEXIST(td, args->path, &path);
1510
1511#ifdef DEBUG
1512	if (ldebug(lchown))
1513		printf(ARGS(lchown, "%s, %d, %d"), path, args->uid, args->gid);
1514#endif
1515	error = kern_lchown(td, path, UIO_SYSSPACE, args->uid, args->gid);
1516	LFREEPATH(path);
1517	return (error);
1518}
1519
1520static int
1521convert_fadvice(int advice)
1522{
1523	switch (advice) {
1524	case LINUX_POSIX_FADV_NORMAL:
1525		return (POSIX_FADV_NORMAL);
1526	case LINUX_POSIX_FADV_RANDOM:
1527		return (POSIX_FADV_RANDOM);
1528	case LINUX_POSIX_FADV_SEQUENTIAL:
1529		return (POSIX_FADV_SEQUENTIAL);
1530	case LINUX_POSIX_FADV_WILLNEED:
1531		return (POSIX_FADV_WILLNEED);
1532	case LINUX_POSIX_FADV_DONTNEED:
1533		return (POSIX_FADV_DONTNEED);
1534	case LINUX_POSIX_FADV_NOREUSE:
1535		return (POSIX_FADV_NOREUSE);
1536	default:
1537		return (-1);
1538	}
1539}
1540
1541int
1542linux_fadvise64(struct thread *td, struct linux_fadvise64_args *args)
1543{
1544	int advice;
1545
1546	advice = convert_fadvice(args->advice);
1547	if (advice == -1)
1548		return (EINVAL);
1549	return (kern_posix_fadvise(td, args->fd, args->offset, args->len,
1550	    advice));
1551}
1552
1553#if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1554int
1555linux_fadvise64_64(struct thread *td, struct linux_fadvise64_64_args *args)
1556{
1557	int advice;
1558
1559	advice = convert_fadvice(args->advice);
1560	if (advice == -1)
1561		return (EINVAL);
1562	return (kern_posix_fadvise(td, args->fd, args->offset, args->len,
1563	    advice));
1564}
1565#endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1566
1567int
1568linux_pipe(struct thread *td, struct linux_pipe_args *args)
1569{
1570	int fildes[2];
1571	int error;
1572
1573#ifdef DEBUG
1574	if (ldebug(pipe))
1575		printf(ARGS(pipe, "*"));
1576#endif
1577
1578	error = kern_pipe2(td, fildes, 0);
1579	if (error)
1580		return (error);
1581
1582	/* XXX: Close descriptors on error. */
1583	return (copyout(fildes, args->pipefds, sizeof(fildes)));
1584}
1585
1586int
1587linux_pipe2(struct thread *td, struct linux_pipe2_args *args)
1588{
1589	int fildes[2];
1590	int error, flags;
1591
1592#ifdef DEBUG
1593	if (ldebug(pipe2))
1594		printf(ARGS(pipe2, "*, %d"), args->flags);
1595#endif
1596
1597	if ((args->flags & ~(LINUX_O_NONBLOCK | LINUX_O_CLOEXEC)) != 0)
1598		return (EINVAL);
1599
1600	flags = 0;
1601	if ((args->flags & LINUX_O_NONBLOCK) != 0)
1602		flags |= O_NONBLOCK;
1603	if ((args->flags & LINUX_O_CLOEXEC) != 0)
1604		flags |= O_CLOEXEC;
1605	error = kern_pipe2(td, fildes, flags);
1606	if (error)
1607		return (error);
1608
1609	/* XXX: Close descriptors on error. */
1610	return (copyout(fildes, args->pipefds, sizeof(fildes)));
1611}
1612
1613int
1614linux_dup3(struct thread *td, struct linux_dup3_args *args)
1615{
1616	int cmd;
1617	intptr_t newfd;
1618
1619	if (args->oldfd == args->newfd)
1620		return (EINVAL);
1621	if ((args->flags & ~LINUX_O_CLOEXEC) != 0)
1622		return (EINVAL);
1623	if (args->flags & LINUX_O_CLOEXEC)
1624		cmd = F_DUP2FD_CLOEXEC;
1625	else
1626		cmd = F_DUP2FD;
1627
1628	newfd = args->newfd;
1629	return (kern_fcntl(td, args->oldfd, cmd, newfd));
1630}
1631