1 /*-
2  * SPDX-License-Identifier: BSD-2-Clause-FreeBSD
3  *
4  * Copyright (c) 1994-1995 Søren Schmidt
5  * All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  *
16  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
17  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
18  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
19  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
20  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
21  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
22  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
23  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
24  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
25  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
26  * SUCH DAMAGE.
27  */
28 
29 #include <sys/cdefs.h>
30 __FBSDID("$FreeBSD: stable/12/sys/compat/linux/linux_file.c 369982 2021-06-16 20:24:53Z 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 #ifdef COMPAT_LINUX32
57 #include <compat/freebsd32/freebsd32_misc.h>
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_misc.h>
65 #include <compat/linux/linux_util.h>
66 #include <compat/linux/linux_file.h>
67 
68 static int	linux_common_open(struct thread *, int, char *, int, int);
69 static int	linux_getdents_error(struct thread *, int, int);
70 
71 #ifdef LINUX_LEGACY_SYSCALLS
72 int
linux_creat(struct thread * td,struct linux_creat_args * args)73 linux_creat(struct thread *td, struct linux_creat_args *args)
74 {
75 	char *path;
76 	int error;
77 
78 	LCONVPATHEXIST(td, args->path, &path);
79 
80 	error = kern_openat(td, AT_FDCWD, path, UIO_SYSSPACE,
81 	    O_WRONLY | O_CREAT | O_TRUNC, args->mode);
82 	LFREEPATH(path);
83 	return (error);
84 }
85 #endif
86 
87 static int
linux_common_open(struct thread * td,int dirfd,char * path,int l_flags,int mode)88 linux_common_open(struct thread *td, int dirfd, char *path, int l_flags, int mode)
89 {
90 	struct proc *p = td->td_proc;
91 	struct file *fp;
92 	int fd;
93 	int bsd_flags, error;
94 
95 	bsd_flags = 0;
96 	switch (l_flags & LINUX_O_ACCMODE) {
97 	case LINUX_O_WRONLY:
98 		bsd_flags |= O_WRONLY;
99 		break;
100 	case LINUX_O_RDWR:
101 		bsd_flags |= O_RDWR;
102 		break;
103 	default:
104 		bsd_flags |= O_RDONLY;
105 	}
106 	if (l_flags & LINUX_O_NDELAY)
107 		bsd_flags |= O_NONBLOCK;
108 	if (l_flags & LINUX_O_APPEND)
109 		bsd_flags |= O_APPEND;
110 	if (l_flags & LINUX_O_SYNC)
111 		bsd_flags |= O_FSYNC;
112 	if (l_flags & LINUX_O_CLOEXEC)
113 		bsd_flags |= O_CLOEXEC;
114 	if (l_flags & LINUX_O_NONBLOCK)
115 		bsd_flags |= O_NONBLOCK;
116 	if (l_flags & LINUX_O_ASYNC)
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 		if (error == EMLINK)
137 			error = ELOOP;
138 		goto done;
139 	}
140 	if (bsd_flags & O_NOCTTY)
141 		goto done;
142 
143 	/*
144 	 * XXX In between kern_openat() and fget(), another process
145 	 * having the same filedesc could use that fd without
146 	 * checking below.
147 	*/
148 	fd = td->td_retval[0];
149 	if (fget(td, fd, &cap_ioctl_rights, &fp) == 0) {
150 		if (fp->f_type != DTYPE_VNODE) {
151 			fdrop(fp, td);
152 			goto done;
153 		}
154 		sx_slock(&proctree_lock);
155 		PROC_LOCK(p);
156 		if (SESS_LEADER(p) && !(p->p_flag & P_CONTROLT)) {
157 			PROC_UNLOCK(p);
158 			sx_sunlock(&proctree_lock);
159 			/* XXXPJD: Verify if TIOCSCTTY is allowed. */
160 			(void) fo_ioctl(fp, TIOCSCTTY, (caddr_t) 0,
161 			    td->td_ucred, td);
162 		} else {
163 			PROC_UNLOCK(p);
164 			sx_sunlock(&proctree_lock);
165 		}
166 		fdrop(fp, td);
167 	}
168 
169 done:
170 	LFREEPATH(path);
171 	return (error);
172 }
173 
174 int
linux_openat(struct thread * td,struct linux_openat_args * args)175 linux_openat(struct thread *td, struct linux_openat_args *args)
176 {
177 	char *path;
178 	int dfd;
179 
180 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
181 	if (args->flags & LINUX_O_CREAT)
182 		LCONVPATH_AT(td, args->filename, &path, 1, dfd);
183 	else
184 		LCONVPATH_AT(td, args->filename, &path, 0, dfd);
185 
186 	return (linux_common_open(td, dfd, path, args->flags, args->mode));
187 }
188 
189 #ifdef LINUX_LEGACY_SYSCALLS
190 int
linux_open(struct thread * td,struct linux_open_args * args)191 linux_open(struct thread *td, struct linux_open_args *args)
192 {
193 	char *path;
194 
195 	if (args->flags & LINUX_O_CREAT)
196 		LCONVPATHCREAT(td, args->path, &path);
197 	else
198 		LCONVPATHEXIST(td, args->path, &path);
199 
200 	return (linux_common_open(td, AT_FDCWD, path, args->flags, args->mode));
201 }
202 #endif
203 
204 int
linux_lseek(struct thread * td,struct linux_lseek_args * args)205 linux_lseek(struct thread *td, struct linux_lseek_args *args)
206 {
207 
208 	return (kern_lseek(td, args->fdes, args->off, args->whence));
209 }
210 
211 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
212 int
linux_llseek(struct thread * td,struct linux_llseek_args * args)213 linux_llseek(struct thread *td, struct linux_llseek_args *args)
214 {
215 	int error;
216 	off_t off;
217 
218 	off = (args->olow) | (((off_t) args->ohigh) << 32);
219 
220 	error = kern_lseek(td, args->fd, off, args->whence);
221 	if (error != 0)
222 		return (error);
223 
224 	error = copyout(td->td_retval, args->res, sizeof(off_t));
225 	if (error != 0)
226 		return (error);
227 
228 	td->td_retval[0] = 0;
229 	return (0);
230 }
231 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
232 
233 /*
234  * Note that linux_getdents(2) and linux_getdents64(2) have the same
235  * arguments. They only differ in the definition of struct dirent they
236  * operate on.
237  * Note that linux_readdir(2) is a special case of linux_getdents(2)
238  * where count is always equals 1, meaning that the buffer is one
239  * dirent-structure in size and that the code can't handle more anyway.
240  * Note that linux_readdir(2) can't be implemented by means of linux_getdents(2)
241  * as in case when the *dent buffer size is equal to 1 linux_getdents(2) will
242  * trash user stack.
243  */
244 
245 static int
linux_getdents_error(struct thread * td,int fd,int err)246 linux_getdents_error(struct thread *td, int fd, int err)
247 {
248 	struct vnode *vp;
249 	struct file *fp;
250 	int error;
251 
252 	/* Linux return ENOTDIR in case when fd is not a directory. */
253 	error = getvnode(td, fd, &cap_read_rights, &fp);
254 	if (error != 0)
255 		return (error);
256 	vp = fp->f_vnode;
257 	if (vp->v_type != VDIR) {
258 		fdrop(fp, td);
259 		return (ENOTDIR);
260 	}
261 	fdrop(fp, td);
262 	return (err);
263 }
264 
265 struct l_dirent {
266 	l_ulong		d_ino;
267 	l_off_t		d_off;
268 	l_ushort	d_reclen;
269 	char		d_name[LINUX_NAME_MAX + 1];
270 };
271 
272 struct l_dirent64 {
273 	uint64_t	d_ino;
274 	int64_t		d_off;
275 	l_ushort	d_reclen;
276 	u_char		d_type;
277 	char		d_name[LINUX_NAME_MAX + 1];
278 };
279 
280 /*
281  * Linux uses the last byte in the dirent buffer to store d_type,
282  * at least glibc-2.7 requires it. That is why l_dirent is padded with 2 bytes.
283  */
284 #define LINUX_RECLEN(namlen)						\
285     roundup(offsetof(struct l_dirent, d_name) + (namlen) + 2, sizeof(l_ulong))
286 
287 #define LINUX_RECLEN64(namlen)						\
288     roundup(offsetof(struct l_dirent64, d_name) + (namlen) + 1,		\
289     sizeof(uint64_t))
290 
291 #ifdef LINUX_LEGACY_SYSCALLS
292 int
linux_getdents(struct thread * td,struct linux_getdents_args * args)293 linux_getdents(struct thread *td, struct linux_getdents_args *args)
294 {
295 	struct dirent *bdp;
296 	caddr_t inp, buf;		/* BSD-format */
297 	int len, reclen;		/* BSD-format */
298 	caddr_t outp;			/* Linux-format */
299 	int resid, linuxreclen;		/* Linux-format */
300 	caddr_t lbuf;			/* Linux-format */
301 	off_t base;
302 	struct l_dirent *linux_dirent;
303 	int buflen, error;
304 	size_t retval;
305 
306 	buflen = min(args->count, MAXBSIZE);
307 	buf = malloc(buflen, M_TEMP, M_WAITOK);
308 
309 	error = kern_getdirentries(td, args->fd, buf, buflen,
310 	    &base, NULL, UIO_SYSSPACE);
311 	if (error != 0) {
312 		error = linux_getdents_error(td, args->fd, error);
313 		goto out1;
314 	}
315 
316 	lbuf = malloc(LINUX_RECLEN(LINUX_NAME_MAX), M_TEMP, M_WAITOK | M_ZERO);
317 
318 	len = td->td_retval[0];
319 	inp = buf;
320 	outp = (caddr_t)args->dent;
321 	resid = args->count;
322 	retval = 0;
323 
324 	while (len > 0) {
325 		bdp = (struct dirent *) inp;
326 		reclen = bdp->d_reclen;
327 		linuxreclen = LINUX_RECLEN(bdp->d_namlen);
328 		/*
329 		 * No more space in the user supplied dirent buffer.
330 		 * Return EINVAL.
331 		 */
332 		if (resid < linuxreclen) {
333 			error = EINVAL;
334 			goto out;
335 		}
336 
337 		linux_dirent = (struct l_dirent*)lbuf;
338 		linux_dirent->d_ino = bdp->d_fileno;
339 		linux_dirent->d_off = base + reclen;
340 		linux_dirent->d_reclen = linuxreclen;
341 		/*
342 		 * Copy d_type to last byte of l_dirent buffer
343 		 */
344 		lbuf[linuxreclen - 1] = bdp->d_type;
345 		strlcpy(linux_dirent->d_name, bdp->d_name,
346 		    linuxreclen - offsetof(struct l_dirent, d_name)-1);
347 		error = copyout(linux_dirent, outp, linuxreclen);
348 		if (error != 0)
349 			goto out;
350 
351 		inp += reclen;
352 		base += reclen;
353 		len -= reclen;
354 
355 		retval += linuxreclen;
356 		outp += linuxreclen;
357 		resid -= linuxreclen;
358 	}
359 	td->td_retval[0] = retval;
360 
361 out:
362 	free(lbuf, M_TEMP);
363 out1:
364 	free(buf, M_TEMP);
365 	return (error);
366 }
367 #endif
368 
369 int
linux_getdents64(struct thread * td,struct linux_getdents64_args * args)370 linux_getdents64(struct thread *td, struct linux_getdents64_args *args)
371 {
372 	struct dirent *bdp;
373 	caddr_t inp, buf;		/* BSD-format */
374 	int len, reclen;		/* BSD-format */
375 	caddr_t outp;			/* Linux-format */
376 	int resid, linuxreclen;		/* Linux-format */
377 	caddr_t lbuf;			/* Linux-format */
378 	off_t base;
379 	struct l_dirent64 *linux_dirent64;
380 	int buflen, error;
381 	size_t retval;
382 
383 	buflen = min(args->count, MAXBSIZE);
384 	buf = malloc(buflen, M_TEMP, M_WAITOK);
385 
386 	error = kern_getdirentries(td, args->fd, buf, buflen,
387 	    &base, NULL, UIO_SYSSPACE);
388 	if (error != 0) {
389 		error = linux_getdents_error(td, args->fd, error);
390 		goto out1;
391 	}
392 
393 	lbuf = malloc(LINUX_RECLEN64(LINUX_NAME_MAX), M_TEMP, M_WAITOK | M_ZERO);
394 
395 	len = td->td_retval[0];
396 	inp = buf;
397 	outp = (caddr_t)args->dirent;
398 	resid = args->count;
399 	retval = 0;
400 
401 	while (len > 0) {
402 		bdp = (struct dirent *) inp;
403 		reclen = bdp->d_reclen;
404 		linuxreclen = LINUX_RECLEN64(bdp->d_namlen);
405 		/*
406 		 * No more space in the user supplied dirent buffer.
407 		 * Return EINVAL.
408 		 */
409 		if (resid < linuxreclen) {
410 			error = EINVAL;
411 			goto out;
412 		}
413 
414 		linux_dirent64 = (struct l_dirent64*)lbuf;
415 		linux_dirent64->d_ino = bdp->d_fileno;
416 		linux_dirent64->d_off = base + reclen;
417 		linux_dirent64->d_reclen = linuxreclen;
418 		linux_dirent64->d_type = bdp->d_type;
419 		strlcpy(linux_dirent64->d_name, bdp->d_name,
420 		    linuxreclen - offsetof(struct l_dirent64, d_name));
421 		error = copyout(linux_dirent64, outp, linuxreclen);
422 		if (error != 0)
423 			goto out;
424 
425 		inp += reclen;
426 		base += reclen;
427 		len -= reclen;
428 
429 		retval += linuxreclen;
430 		outp += linuxreclen;
431 		resid -= linuxreclen;
432 	}
433 	td->td_retval[0] = retval;
434 
435 out:
436 	free(lbuf, M_TEMP);
437 out1:
438 	free(buf, M_TEMP);
439 	return (error);
440 }
441 
442 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
443 int
linux_readdir(struct thread * td,struct linux_readdir_args * args)444 linux_readdir(struct thread *td, struct linux_readdir_args *args)
445 {
446 	struct dirent *bdp;
447 	caddr_t buf;			/* BSD-format */
448 	int linuxreclen;		/* Linux-format */
449 	caddr_t lbuf;			/* Linux-format */
450 	off_t base;
451 	struct l_dirent *linux_dirent;
452 	int buflen, error;
453 
454 	buflen = LINUX_RECLEN(LINUX_NAME_MAX);
455 	buf = malloc(buflen, M_TEMP, M_WAITOK);
456 
457 	error = kern_getdirentries(td, args->fd, buf, buflen,
458 	    &base, NULL, UIO_SYSSPACE);
459 	if (error != 0) {
460 		error = linux_getdents_error(td, args->fd, error);
461 		goto out;
462 	}
463 	if (td->td_retval[0] == 0)
464 		goto out;
465 
466 	lbuf = malloc(LINUX_RECLEN(LINUX_NAME_MAX), M_TEMP, M_WAITOK | M_ZERO);
467 
468 	bdp = (struct dirent *) buf;
469 	linuxreclen = LINUX_RECLEN(bdp->d_namlen);
470 
471 	linux_dirent = (struct l_dirent*)lbuf;
472 	linux_dirent->d_ino = bdp->d_fileno;
473 	linux_dirent->d_off = linuxreclen;
474 	linux_dirent->d_reclen = bdp->d_namlen;
475 	strlcpy(linux_dirent->d_name, bdp->d_name,
476 	    linuxreclen - offsetof(struct l_dirent, d_name));
477 	error = copyout(linux_dirent, args->dent, linuxreclen);
478 	if (error == 0)
479 		td->td_retval[0] = linuxreclen;
480 
481 	free(lbuf, M_TEMP);
482 out:
483 	free(buf, M_TEMP);
484 	return (error);
485 }
486 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
487 
488 
489 /*
490  * These exist mainly for hooks for doing /compat/linux translation.
491  */
492 
493 #ifdef LINUX_LEGACY_SYSCALLS
494 int
linux_access(struct thread * td,struct linux_access_args * args)495 linux_access(struct thread *td, struct linux_access_args *args)
496 {
497 	char *path;
498 	int error;
499 
500 	/* Linux convention. */
501 	if (args->amode & ~(F_OK | X_OK | W_OK | R_OK))
502 		return (EINVAL);
503 
504 	LCONVPATHEXIST(td, args->path, &path);
505 
506 	error = kern_accessat(td, AT_FDCWD, path, UIO_SYSSPACE, 0,
507 	    args->amode);
508 	LFREEPATH(path);
509 
510 	return (error);
511 }
512 #endif
513 
514 int
linux_faccessat(struct thread * td,struct linux_faccessat_args * args)515 linux_faccessat(struct thread *td, struct linux_faccessat_args *args)
516 {
517 	char *path;
518 	int error, dfd;
519 
520 	/* Linux convention. */
521 	if (args->amode & ~(F_OK | X_OK | W_OK | R_OK))
522 		return (EINVAL);
523 
524 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
525 	LCONVPATHEXIST_AT(td, args->filename, &path, dfd);
526 
527 	error = kern_accessat(td, dfd, path, UIO_SYSSPACE, 0, args->amode);
528 	LFREEPATH(path);
529 
530 	return (error);
531 }
532 
533 #ifdef LINUX_LEGACY_SYSCALLS
534 int
linux_unlink(struct thread * td,struct linux_unlink_args * args)535 linux_unlink(struct thread *td, struct linux_unlink_args *args)
536 {
537 	char *path;
538 	int error;
539 	struct stat st;
540 
541 	LCONVPATHEXIST(td, args->path, &path);
542 
543 	error = kern_unlinkat(td, AT_FDCWD, path, UIO_SYSSPACE, 0, 0);
544 	if (error == EPERM) {
545 		/* Introduce POSIX noncompliant behaviour of Linux */
546 		if (kern_statat(td, 0, AT_FDCWD, path, UIO_SYSSPACE, &st,
547 		    NULL) == 0) {
548 			if (S_ISDIR(st.st_mode))
549 				error = EISDIR;
550 		}
551 	}
552 	LFREEPATH(path);
553 	return (error);
554 }
555 #endif
556 
557 int
linux_unlinkat(struct thread * td,struct linux_unlinkat_args * args)558 linux_unlinkat(struct thread *td, struct linux_unlinkat_args *args)
559 {
560 	char *path;
561 	int error, dfd;
562 	struct stat st;
563 
564 	if (args->flag & ~LINUX_AT_REMOVEDIR)
565 		return (EINVAL);
566 
567 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
568 	LCONVPATHEXIST_AT(td, args->pathname, &path, dfd);
569 
570 	if (args->flag & LINUX_AT_REMOVEDIR)
571 		error = kern_rmdirat(td, dfd, path, UIO_SYSSPACE, 0);
572 	else
573 		error = kern_unlinkat(td, dfd, path, UIO_SYSSPACE, 0, 0);
574 	if (error == EPERM && !(args->flag & LINUX_AT_REMOVEDIR)) {
575 		/* Introduce POSIX noncompliant behaviour of Linux */
576 		if (kern_statat(td, AT_SYMLINK_NOFOLLOW, dfd, path,
577 		    UIO_SYSSPACE, &st, NULL) == 0 && S_ISDIR(st.st_mode))
578 			error = EISDIR;
579 	}
580 	LFREEPATH(path);
581 	return (error);
582 }
583 int
linux_chdir(struct thread * td,struct linux_chdir_args * args)584 linux_chdir(struct thread *td, struct linux_chdir_args *args)
585 {
586 	char *path;
587 	int error;
588 
589 	LCONVPATHEXIST(td, args->path, &path);
590 
591 	error = kern_chdir(td, path, UIO_SYSSPACE);
592 	LFREEPATH(path);
593 	return (error);
594 }
595 
596 #ifdef LINUX_LEGACY_SYSCALLS
597 int
linux_chmod(struct thread * td,struct linux_chmod_args * args)598 linux_chmod(struct thread *td, struct linux_chmod_args *args)
599 {
600 	char *path;
601 	int error;
602 
603 	LCONVPATHEXIST(td, args->path, &path);
604 
605 	error = kern_fchmodat(td, AT_FDCWD, path, UIO_SYSSPACE,
606 	    args->mode, 0);
607 	LFREEPATH(path);
608 	return (error);
609 }
610 #endif
611 
612 int
linux_fchmodat(struct thread * td,struct linux_fchmodat_args * args)613 linux_fchmodat(struct thread *td, struct linux_fchmodat_args *args)
614 {
615 	char *path;
616 	int error, dfd;
617 
618 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
619 	LCONVPATHEXIST_AT(td, args->filename, &path, dfd);
620 
621 	error = kern_fchmodat(td, dfd, path, UIO_SYSSPACE, args->mode, 0);
622 	LFREEPATH(path);
623 	return (error);
624 }
625 
626 #ifdef LINUX_LEGACY_SYSCALLS
627 int
linux_mkdir(struct thread * td,struct linux_mkdir_args * args)628 linux_mkdir(struct thread *td, struct linux_mkdir_args *args)
629 {
630 	char *path;
631 	int error;
632 
633 	LCONVPATHCREAT(td, args->path, &path);
634 
635 	error = kern_mkdirat(td, AT_FDCWD, path, UIO_SYSSPACE, args->mode);
636 	LFREEPATH(path);
637 	return (error);
638 }
639 #endif
640 
641 int
linux_mkdirat(struct thread * td,struct linux_mkdirat_args * args)642 linux_mkdirat(struct thread *td, struct linux_mkdirat_args *args)
643 {
644 	char *path;
645 	int error, dfd;
646 
647 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
648 	LCONVPATHCREAT_AT(td, args->pathname, &path, dfd);
649 
650 	error = kern_mkdirat(td, dfd, path, UIO_SYSSPACE, args->mode);
651 	LFREEPATH(path);
652 	return (error);
653 }
654 
655 #ifdef LINUX_LEGACY_SYSCALLS
656 int
linux_rmdir(struct thread * td,struct linux_rmdir_args * args)657 linux_rmdir(struct thread *td, struct linux_rmdir_args *args)
658 {
659 	char *path;
660 	int error;
661 
662 	LCONVPATHEXIST(td, args->path, &path);
663 
664 	error = kern_rmdirat(td, AT_FDCWD, path, UIO_SYSSPACE, 0);
665 	LFREEPATH(path);
666 	return (error);
667 }
668 
669 int
linux_rename(struct thread * td,struct linux_rename_args * args)670 linux_rename(struct thread *td, struct linux_rename_args *args)
671 {
672 	char *from, *to;
673 	int error;
674 
675 	LCONVPATHEXIST(td, args->from, &from);
676 	/* Expand LCONVPATHCREATE so that `from' can be freed on errors */
677 	error = linux_emul_convpath(td, args->to, UIO_USERSPACE, &to, 1, AT_FDCWD);
678 	if (to == NULL) {
679 		LFREEPATH(from);
680 		return (error);
681 	}
682 
683 	error = kern_renameat(td, AT_FDCWD, from, AT_FDCWD, to, UIO_SYSSPACE);
684 	LFREEPATH(from);
685 	LFREEPATH(to);
686 	return (error);
687 }
688 #endif
689 
690 int
linux_renameat(struct thread * td,struct linux_renameat_args * args)691 linux_renameat(struct thread *td, struct linux_renameat_args *args)
692 {
693 	struct linux_renameat2_args renameat2_args = {
694 	    .olddfd = args->olddfd,
695 	    .oldname = args->oldname,
696 	    .newdfd = args->newdfd,
697 	    .newname = args->newname,
698 	    .flags = 0
699 	};
700 
701 	return (linux_renameat2(td, &renameat2_args));
702 }
703 
704 int
linux_renameat2(struct thread * td,struct linux_renameat2_args * args)705 linux_renameat2(struct thread *td, struct linux_renameat2_args *args)
706 {
707 	char *from, *to;
708 	int error, olddfd, newdfd;
709 
710 	if (args->flags != 0) {
711 		if (args->flags & ~(LINUX_RENAME_EXCHANGE |
712 		    LINUX_RENAME_NOREPLACE | LINUX_RENAME_WHITEOUT))
713 			return (EINVAL);
714 		if (args->flags & LINUX_RENAME_EXCHANGE &&
715 		    args->flags & (LINUX_RENAME_NOREPLACE |
716 		    LINUX_RENAME_WHITEOUT))
717 			return (EINVAL);
718 #if 0
719 		/*
720 		 * This spams the console on Ubuntu Focal.
721 		 *
722 		 * What's needed here is a general mechanism to let users know
723 		 * about missing features without hogging the system.
724 		 */
725 		linux_msg(td, "renameat2 unsupported flags 0x%x",
726 		    args->flags);
727 #endif
728 		return (EINVAL);
729 	}
730 
731 	olddfd = (args->olddfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->olddfd;
732 	newdfd = (args->newdfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->newdfd;
733 	LCONVPATHEXIST_AT(td, args->oldname, &from, olddfd);
734 	/* Expand LCONVPATHCREATE so that `from' can be freed on errors */
735 	error = linux_emul_convpath(td, args->newname, UIO_USERSPACE, &to, 1, newdfd);
736 	if (to == NULL) {
737 		LFREEPATH(from);
738 		return (error);
739 	}
740 
741 	error = kern_renameat(td, olddfd, from, newdfd, to, UIO_SYSSPACE);
742 	LFREEPATH(from);
743 	LFREEPATH(to);
744 	return (error);
745 }
746 
747 #ifdef LINUX_LEGACY_SYSCALLS
748 int
linux_symlink(struct thread * td,struct linux_symlink_args * args)749 linux_symlink(struct thread *td, struct linux_symlink_args *args)
750 {
751 	char *path, *to;
752 	int error;
753 
754 	LCONVPATHEXIST(td, args->path, &path);
755 	/* Expand LCONVPATHCREATE so that `path' can be freed on errors */
756 	error = linux_emul_convpath(td, args->to, UIO_USERSPACE, &to, 1, AT_FDCWD);
757 	if (to == NULL) {
758 		LFREEPATH(path);
759 		return (error);
760 	}
761 
762 	error = kern_symlinkat(td, path, AT_FDCWD, to, UIO_SYSSPACE);
763 	LFREEPATH(path);
764 	LFREEPATH(to);
765 	return (error);
766 }
767 #endif
768 
769 int
linux_symlinkat(struct thread * td,struct linux_symlinkat_args * args)770 linux_symlinkat(struct thread *td, struct linux_symlinkat_args *args)
771 {
772 	char *path, *to;
773 	int error, dfd;
774 
775 	dfd = (args->newdfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->newdfd;
776 	LCONVPATHEXIST(td, args->oldname, &path);
777 	/* Expand LCONVPATHCREATE so that `path' can be freed on errors */
778 	error = linux_emul_convpath(td, args->newname, UIO_USERSPACE, &to, 1, dfd);
779 	if (to == NULL) {
780 		LFREEPATH(path);
781 		return (error);
782 	}
783 
784 	error = kern_symlinkat(td, path, dfd, to, UIO_SYSSPACE);
785 	LFREEPATH(path);
786 	LFREEPATH(to);
787 	return (error);
788 }
789 
790 #ifdef LINUX_LEGACY_SYSCALLS
791 int
linux_readlink(struct thread * td,struct linux_readlink_args * args)792 linux_readlink(struct thread *td, struct linux_readlink_args *args)
793 {
794 	char *name;
795 	int error;
796 
797 	LCONVPATHEXIST(td, args->name, &name);
798 
799 	error = kern_readlinkat(td, AT_FDCWD, name, UIO_SYSSPACE,
800 	    args->buf, UIO_USERSPACE, args->count);
801 	LFREEPATH(name);
802 	return (error);
803 }
804 #endif
805 
806 int
linux_readlinkat(struct thread * td,struct linux_readlinkat_args * args)807 linux_readlinkat(struct thread *td, struct linux_readlinkat_args *args)
808 {
809 	char *name;
810 	int error, dfd;
811 
812 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
813 	LCONVPATHEXIST_AT(td, args->path, &name, dfd);
814 
815 	error = kern_readlinkat(td, dfd, name, UIO_SYSSPACE, args->buf,
816 	    UIO_USERSPACE, args->bufsiz);
817 	LFREEPATH(name);
818 	return (error);
819 }
820 
821 int
linux_truncate(struct thread * td,struct linux_truncate_args * args)822 linux_truncate(struct thread *td, struct linux_truncate_args *args)
823 {
824 	char *path;
825 	int error;
826 
827 	LCONVPATHEXIST(td, args->path, &path);
828 	error = kern_truncate(td, path, UIO_SYSSPACE, args->length);
829 	LFREEPATH(path);
830 	return (error);
831 }
832 
833 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
834 int
linux_truncate64(struct thread * td,struct linux_truncate64_args * args)835 linux_truncate64(struct thread *td, struct linux_truncate64_args *args)
836 {
837 	char *path;
838 	off_t length;
839 	int error;
840 
841 #if defined(__amd64__) && defined(COMPAT_LINUX32)
842 	length = PAIR32TO64(off_t, args->length);
843 #else
844 	length = args->length;
845 #endif
846 
847 	LCONVPATHEXIST(td, args->path, &path);
848 
849 	error = kern_truncate(td, path, UIO_SYSSPACE, length);
850 	LFREEPATH(path);
851 	return (error);
852 }
853 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
854 
855 int
linux_ftruncate(struct thread * td,struct linux_ftruncate_args * args)856 linux_ftruncate(struct thread *td, struct linux_ftruncate_args *args)
857 {
858 
859 	return (kern_ftruncate(td, args->fd, args->length));
860 }
861 
862 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
863 int
linux_ftruncate64(struct thread * td,struct linux_ftruncate64_args * args)864 linux_ftruncate64(struct thread *td, struct linux_ftruncate64_args *args)
865 {
866 	off_t length;
867 
868 #if defined(__amd64__) && defined(COMPAT_LINUX32)
869 	length = PAIR32TO64(off_t, args->length);
870 #else
871 	length = args->length;
872 #endif
873 
874 	return (kern_ftruncate(td, args->fd, length));
875 }
876 #endif
877 
878 #ifdef LINUX_LEGACY_SYSCALLS
879 int
linux_link(struct thread * td,struct linux_link_args * args)880 linux_link(struct thread *td, struct linux_link_args *args)
881 {
882 	char *path, *to;
883 	int error;
884 
885 	LCONVPATHEXIST(td, args->path, &path);
886 	/* Expand LCONVPATHCREATE so that `path' can be freed on errors */
887 	error = linux_emul_convpath(td, args->to, UIO_USERSPACE, &to, 1, AT_FDCWD);
888 	if (to == NULL) {
889 		LFREEPATH(path);
890 		return (error);
891 	}
892 
893 	error = kern_linkat(td, AT_FDCWD, AT_FDCWD, path, to, UIO_SYSSPACE,
894 	    FOLLOW);
895 	LFREEPATH(path);
896 	LFREEPATH(to);
897 	return (error);
898 }
899 #endif
900 
901 int
linux_linkat(struct thread * td,struct linux_linkat_args * args)902 linux_linkat(struct thread *td, struct linux_linkat_args *args)
903 {
904 	char *path, *to;
905 	int error, olddfd, newdfd, follow;
906 
907 	if (args->flag & ~LINUX_AT_SYMLINK_FOLLOW)
908 		return (EINVAL);
909 
910 	olddfd = (args->olddfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->olddfd;
911 	newdfd = (args->newdfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->newdfd;
912 	LCONVPATHEXIST_AT(td, args->oldname, &path, olddfd);
913 	/* Expand LCONVPATHCREATE so that `path' can be freed on errors */
914 	error = linux_emul_convpath(td, args->newname, UIO_USERSPACE, &to, 1, newdfd);
915 	if (to == NULL) {
916 		LFREEPATH(path);
917 		return (error);
918 	}
919 
920 	follow = (args->flag & LINUX_AT_SYMLINK_FOLLOW) == 0 ? NOFOLLOW :
921 	    FOLLOW;
922 	error = kern_linkat(td, olddfd, newdfd, path, to, UIO_SYSSPACE, follow);
923 	LFREEPATH(path);
924 	LFREEPATH(to);
925 	return (error);
926 }
927 
928 int
linux_fdatasync(struct thread * td,struct linux_fdatasync_args * uap)929 linux_fdatasync(struct thread *td, struct linux_fdatasync_args *uap)
930 {
931 
932 	return (kern_fsync(td, uap->fd, false));
933 }
934 
935 int
linux_sync_file_range(struct thread * td,struct linux_sync_file_range_args * uap)936 linux_sync_file_range(struct thread *td, struct linux_sync_file_range_args *uap)
937 {
938 	off_t nbytes, offset;
939 
940 #if defined(__amd64__) && defined(COMPAT_LINUX32)
941 	nbytes = PAIR32TO64(off_t, uap->nbytes);
942 	offset = PAIR32TO64(off_t, uap->offset);
943 #else
944 	nbytes = uap->nbytes;
945 	offset = uap->offset;
946 #endif
947 
948 	if (offset < 0 || nbytes < 0 ||
949 	    (uap->flags & ~(LINUX_SYNC_FILE_RANGE_WAIT_BEFORE |
950 	    LINUX_SYNC_FILE_RANGE_WRITE |
951 	    LINUX_SYNC_FILE_RANGE_WAIT_AFTER)) != 0) {
952 		return (EINVAL);
953 	}
954 
955 	return (kern_fsync(td, uap->fd, false));
956 }
957 
958 int
linux_pread(struct thread * td,struct linux_pread_args * uap)959 linux_pread(struct thread *td, struct linux_pread_args *uap)
960 {
961 	struct vnode *vp;
962 	off_t offset;
963 	int error;
964 
965 #if defined(__amd64__) && defined(COMPAT_LINUX32)
966 	offset = PAIR32TO64(off_t, uap->offset);
967 #else
968 	offset = uap->offset;
969 #endif
970 
971 	error = kern_pread(td, uap->fd, uap->buf, uap->nbyte, offset);
972 	if (error == 0) {
973 		/* This seems to violate POSIX but Linux does it. */
974 		error = fgetvp(td, uap->fd, &cap_pread_rights, &vp);
975 		if (error != 0)
976 			return (error);
977 		if (vp->v_type == VDIR)
978 			error = EISDIR;
979 		vrele(vp);
980 	}
981 	return (error);
982 }
983 
984 int
linux_pwrite(struct thread * td,struct linux_pwrite_args * uap)985 linux_pwrite(struct thread *td, struct linux_pwrite_args *uap)
986 {
987 	off_t offset;
988 
989 #if defined(__amd64__) && defined(COMPAT_LINUX32)
990 	offset = PAIR32TO64(off_t, uap->offset);
991 #else
992 	offset = uap->offset;
993 #endif
994 
995 	return (kern_pwrite(td, uap->fd, uap->buf, uap->nbyte, offset));
996 }
997 
998 int
linux_preadv(struct thread * td,struct linux_preadv_args * uap)999 linux_preadv(struct thread *td, struct linux_preadv_args *uap)
1000 {
1001 	struct uio *auio;
1002 	int error;
1003 	off_t offset;
1004 
1005 	/*
1006 	 * According http://man7.org/linux/man-pages/man2/preadv.2.html#NOTES
1007 	 * pos_l and pos_h, respectively, contain the
1008 	 * low order and high order 32 bits of offset.
1009 	 */
1010 	offset = (((off_t)uap->pos_h << (sizeof(offset) * 4)) <<
1011 	    (sizeof(offset) * 4)) | uap->pos_l;
1012 	if (offset < 0)
1013 		return (EINVAL);
1014 #ifdef COMPAT_LINUX32
1015 	error = linux32_copyinuio(PTRIN(uap->vec), uap->vlen, &auio);
1016 #else
1017 	error = copyinuio(uap->vec, uap->vlen, &auio);
1018 #endif
1019 	if (error != 0)
1020 		return (error);
1021 	error = kern_preadv(td, uap->fd, auio, offset);
1022 	free(auio, M_IOV);
1023 	return (error);
1024 }
1025 
1026 int
linux_pwritev(struct thread * td,struct linux_pwritev_args * uap)1027 linux_pwritev(struct thread *td, struct linux_pwritev_args *uap)
1028 {
1029 	struct uio *auio;
1030 	int error;
1031 	off_t offset;
1032 
1033 	/*
1034 	 * According http://man7.org/linux/man-pages/man2/pwritev.2.html#NOTES
1035 	 * pos_l and pos_h, respectively, contain the
1036 	 * low order and high order 32 bits of offset.
1037 	 */
1038 	offset = (((off_t)uap->pos_h << (sizeof(offset) * 4)) <<
1039 	    (sizeof(offset) * 4)) | uap->pos_l;
1040 	if (offset < 0)
1041 		return (EINVAL);
1042 #ifdef COMPAT_LINUX32
1043 	error = linux32_copyinuio(PTRIN(uap->vec), uap->vlen, &auio);
1044 #else
1045 	error = copyinuio(uap->vec, uap->vlen, &auio);
1046 #endif
1047 	if (error != 0)
1048 		return (error);
1049 	error = kern_pwritev(td, uap->fd, auio, offset);
1050 	free(auio, M_IOV);
1051 	return (error);
1052 }
1053 
1054 int
linux_mount(struct thread * td,struct linux_mount_args * args)1055 linux_mount(struct thread *td, struct linux_mount_args *args)
1056 {
1057 	char fstypename[MFSNAMELEN];
1058 	char *mntonname, *mntfromname;
1059 	int error, fsflags;
1060 
1061 	mntonname = malloc(MNAMELEN, M_TEMP, M_WAITOK);
1062 	mntfromname = malloc(MNAMELEN, M_TEMP, M_WAITOK);
1063 	error = copyinstr(args->filesystemtype, fstypename, MFSNAMELEN - 1,
1064 	    NULL);
1065 	if (error != 0)
1066 		goto out;
1067 	error = copyinstr(args->specialfile, mntfromname, MNAMELEN - 1, NULL);
1068 	if (error != 0)
1069 		goto out;
1070 	error = copyinstr(args->dir, mntonname, MNAMELEN - 1, NULL);
1071 	if (error != 0)
1072 		goto out;
1073 
1074 	if (strcmp(fstypename, "ext2") == 0) {
1075 		strcpy(fstypename, "ext2fs");
1076 	} else if (strcmp(fstypename, "proc") == 0) {
1077 		strcpy(fstypename, "linprocfs");
1078 	} else if (strcmp(fstypename, "vfat") == 0) {
1079 		strcpy(fstypename, "msdosfs");
1080 	}
1081 
1082 	fsflags = 0;
1083 
1084 	if ((args->rwflag & 0xffff0000) == 0xc0ed0000) {
1085 		/*
1086 		 * Linux SYNC flag is not included; the closest equivalent
1087 		 * FreeBSD has is !ASYNC, which is our default.
1088 		 */
1089 		if (args->rwflag & LINUX_MS_RDONLY)
1090 			fsflags |= MNT_RDONLY;
1091 		if (args->rwflag & LINUX_MS_NOSUID)
1092 			fsflags |= MNT_NOSUID;
1093 		if (args->rwflag & LINUX_MS_NOEXEC)
1094 			fsflags |= MNT_NOEXEC;
1095 		if (args->rwflag & LINUX_MS_REMOUNT)
1096 			fsflags |= MNT_UPDATE;
1097 	}
1098 
1099 	error = kernel_vmount(fsflags,
1100 	    "fstype", fstypename,
1101 	    "fspath", mntonname,
1102 	    "from", mntfromname,
1103 	    NULL);
1104 out:
1105 	free(mntonname, M_TEMP);
1106 	free(mntfromname, M_TEMP);
1107 	return (error);
1108 }
1109 
1110 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1111 int
linux_oldumount(struct thread * td,struct linux_oldumount_args * args)1112 linux_oldumount(struct thread *td, struct linux_oldumount_args *args)
1113 {
1114 	struct linux_umount_args args2;
1115 
1116 	args2.path = args->path;
1117 	args2.flags = 0;
1118 	return (linux_umount(td, &args2));
1119 }
1120 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1121 
1122 #ifdef LINUX_LEGACY_SYSCALLS
1123 int
linux_umount(struct thread * td,struct linux_umount_args * args)1124 linux_umount(struct thread *td, struct linux_umount_args *args)
1125 {
1126 	struct unmount_args bsd;
1127 
1128 	bsd.path = args->path;
1129 	bsd.flags = args->flags;	/* XXX correct? */
1130 	return (sys_unmount(td, &bsd));
1131 }
1132 #endif
1133 
1134 /*
1135  * fcntl family of syscalls
1136  */
1137 
1138 struct l_flock {
1139 	l_short		l_type;
1140 	l_short		l_whence;
1141 	l_off_t		l_start;
1142 	l_off_t		l_len;
1143 	l_pid_t		l_pid;
1144 }
1145 #if defined(__amd64__) && defined(COMPAT_LINUX32)
1146 __packed
1147 #endif
1148 ;
1149 
1150 static void
linux_to_bsd_flock(struct l_flock * linux_flock,struct flock * bsd_flock)1151 linux_to_bsd_flock(struct l_flock *linux_flock, struct flock *bsd_flock)
1152 {
1153 	switch (linux_flock->l_type) {
1154 	case LINUX_F_RDLCK:
1155 		bsd_flock->l_type = F_RDLCK;
1156 		break;
1157 	case LINUX_F_WRLCK:
1158 		bsd_flock->l_type = F_WRLCK;
1159 		break;
1160 	case LINUX_F_UNLCK:
1161 		bsd_flock->l_type = F_UNLCK;
1162 		break;
1163 	default:
1164 		bsd_flock->l_type = -1;
1165 		break;
1166 	}
1167 	bsd_flock->l_whence = linux_flock->l_whence;
1168 	bsd_flock->l_start = (off_t)linux_flock->l_start;
1169 	bsd_flock->l_len = (off_t)linux_flock->l_len;
1170 	bsd_flock->l_pid = (pid_t)linux_flock->l_pid;
1171 	bsd_flock->l_sysid = 0;
1172 }
1173 
1174 static void
bsd_to_linux_flock(struct flock * bsd_flock,struct l_flock * linux_flock)1175 bsd_to_linux_flock(struct flock *bsd_flock, struct l_flock *linux_flock)
1176 {
1177 	switch (bsd_flock->l_type) {
1178 	case F_RDLCK:
1179 		linux_flock->l_type = LINUX_F_RDLCK;
1180 		break;
1181 	case F_WRLCK:
1182 		linux_flock->l_type = LINUX_F_WRLCK;
1183 		break;
1184 	case F_UNLCK:
1185 		linux_flock->l_type = LINUX_F_UNLCK;
1186 		break;
1187 	}
1188 	linux_flock->l_whence = bsd_flock->l_whence;
1189 	linux_flock->l_start = (l_off_t)bsd_flock->l_start;
1190 	linux_flock->l_len = (l_off_t)bsd_flock->l_len;
1191 	linux_flock->l_pid = (l_pid_t)bsd_flock->l_pid;
1192 }
1193 
1194 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1195 struct l_flock64 {
1196 	l_short		l_type;
1197 	l_short		l_whence;
1198 	l_loff_t	l_start;
1199 	l_loff_t	l_len;
1200 	l_pid_t		l_pid;
1201 }
1202 #if defined(__amd64__) && defined(COMPAT_LINUX32)
1203 __packed
1204 #endif
1205 ;
1206 
1207 static void
linux_to_bsd_flock64(struct l_flock64 * linux_flock,struct flock * bsd_flock)1208 linux_to_bsd_flock64(struct l_flock64 *linux_flock, struct flock *bsd_flock)
1209 {
1210 	switch (linux_flock->l_type) {
1211 	case LINUX_F_RDLCK:
1212 		bsd_flock->l_type = F_RDLCK;
1213 		break;
1214 	case LINUX_F_WRLCK:
1215 		bsd_flock->l_type = F_WRLCK;
1216 		break;
1217 	case LINUX_F_UNLCK:
1218 		bsd_flock->l_type = F_UNLCK;
1219 		break;
1220 	default:
1221 		bsd_flock->l_type = -1;
1222 		break;
1223 	}
1224 	bsd_flock->l_whence = linux_flock->l_whence;
1225 	bsd_flock->l_start = (off_t)linux_flock->l_start;
1226 	bsd_flock->l_len = (off_t)linux_flock->l_len;
1227 	bsd_flock->l_pid = (pid_t)linux_flock->l_pid;
1228 	bsd_flock->l_sysid = 0;
1229 }
1230 
1231 static void
bsd_to_linux_flock64(struct flock * bsd_flock,struct l_flock64 * linux_flock)1232 bsd_to_linux_flock64(struct flock *bsd_flock, struct l_flock64 *linux_flock)
1233 {
1234 	switch (bsd_flock->l_type) {
1235 	case F_RDLCK:
1236 		linux_flock->l_type = LINUX_F_RDLCK;
1237 		break;
1238 	case F_WRLCK:
1239 		linux_flock->l_type = LINUX_F_WRLCK;
1240 		break;
1241 	case F_UNLCK:
1242 		linux_flock->l_type = LINUX_F_UNLCK;
1243 		break;
1244 	}
1245 	linux_flock->l_whence = bsd_flock->l_whence;
1246 	linux_flock->l_start = (l_loff_t)bsd_flock->l_start;
1247 	linux_flock->l_len = (l_loff_t)bsd_flock->l_len;
1248 	linux_flock->l_pid = (l_pid_t)bsd_flock->l_pid;
1249 }
1250 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1251 
1252 static int
fcntl_common(struct thread * td,struct linux_fcntl_args * args)1253 fcntl_common(struct thread *td, struct linux_fcntl_args *args)
1254 {
1255 	struct l_flock linux_flock;
1256 	struct flock bsd_flock;
1257 	struct file *fp;
1258 	long arg;
1259 	int error, result;
1260 
1261 	switch (args->cmd) {
1262 	case LINUX_F_DUPFD:
1263 		return (kern_fcntl(td, args->fd, F_DUPFD, args->arg));
1264 
1265 	case LINUX_F_GETFD:
1266 		return (kern_fcntl(td, args->fd, F_GETFD, 0));
1267 
1268 	case LINUX_F_SETFD:
1269 		return (kern_fcntl(td, args->fd, F_SETFD, args->arg));
1270 
1271 	case LINUX_F_GETFL:
1272 		error = kern_fcntl(td, args->fd, F_GETFL, 0);
1273 		result = td->td_retval[0];
1274 		td->td_retval[0] = 0;
1275 		if (result & O_RDONLY)
1276 			td->td_retval[0] |= LINUX_O_RDONLY;
1277 		if (result & O_WRONLY)
1278 			td->td_retval[0] |= LINUX_O_WRONLY;
1279 		if (result & O_RDWR)
1280 			td->td_retval[0] |= LINUX_O_RDWR;
1281 		if (result & O_NDELAY)
1282 			td->td_retval[0] |= LINUX_O_NONBLOCK;
1283 		if (result & O_APPEND)
1284 			td->td_retval[0] |= LINUX_O_APPEND;
1285 		if (result & O_FSYNC)
1286 			td->td_retval[0] |= LINUX_O_SYNC;
1287 		if (result & O_ASYNC)
1288 			td->td_retval[0] |= LINUX_O_ASYNC;
1289 #ifdef LINUX_O_NOFOLLOW
1290 		if (result & O_NOFOLLOW)
1291 			td->td_retval[0] |= LINUX_O_NOFOLLOW;
1292 #endif
1293 #ifdef LINUX_O_DIRECT
1294 		if (result & O_DIRECT)
1295 			td->td_retval[0] |= LINUX_O_DIRECT;
1296 #endif
1297 		return (error);
1298 
1299 	case LINUX_F_SETFL:
1300 		arg = 0;
1301 		if (args->arg & LINUX_O_NDELAY)
1302 			arg |= O_NONBLOCK;
1303 		if (args->arg & LINUX_O_APPEND)
1304 			arg |= O_APPEND;
1305 		if (args->arg & LINUX_O_SYNC)
1306 			arg |= O_FSYNC;
1307 		if (args->arg & LINUX_O_ASYNC)
1308 			arg |= O_ASYNC;
1309 #ifdef LINUX_O_NOFOLLOW
1310 		if (args->arg & LINUX_O_NOFOLLOW)
1311 			arg |= O_NOFOLLOW;
1312 #endif
1313 #ifdef LINUX_O_DIRECT
1314 		if (args->arg & LINUX_O_DIRECT)
1315 			arg |= O_DIRECT;
1316 #endif
1317 		return (kern_fcntl(td, args->fd, F_SETFL, arg));
1318 
1319 	case LINUX_F_GETLK:
1320 		error = copyin((void *)args->arg, &linux_flock,
1321 		    sizeof(linux_flock));
1322 		if (error)
1323 			return (error);
1324 		linux_to_bsd_flock(&linux_flock, &bsd_flock);
1325 		error = kern_fcntl(td, args->fd, F_GETLK, (intptr_t)&bsd_flock);
1326 		if (error)
1327 			return (error);
1328 		bsd_to_linux_flock(&bsd_flock, &linux_flock);
1329 		return (copyout(&linux_flock, (void *)args->arg,
1330 		    sizeof(linux_flock)));
1331 
1332 	case LINUX_F_SETLK:
1333 		error = copyin((void *)args->arg, &linux_flock,
1334 		    sizeof(linux_flock));
1335 		if (error)
1336 			return (error);
1337 		linux_to_bsd_flock(&linux_flock, &bsd_flock);
1338 		return (kern_fcntl(td, args->fd, F_SETLK,
1339 		    (intptr_t)&bsd_flock));
1340 
1341 	case LINUX_F_SETLKW:
1342 		error = copyin((void *)args->arg, &linux_flock,
1343 		    sizeof(linux_flock));
1344 		if (error)
1345 			return (error);
1346 		linux_to_bsd_flock(&linux_flock, &bsd_flock);
1347 		return (kern_fcntl(td, args->fd, F_SETLKW,
1348 		     (intptr_t)&bsd_flock));
1349 
1350 	case LINUX_F_GETOWN:
1351 		return (kern_fcntl(td, args->fd, F_GETOWN, 0));
1352 
1353 	case LINUX_F_SETOWN:
1354 		/*
1355 		 * XXX some Linux applications depend on F_SETOWN having no
1356 		 * significant effect for pipes (SIGIO is not delivered for
1357 		 * pipes under Linux-2.2.35 at least).
1358 		 */
1359 		error = fget(td, args->fd,
1360 		    &cap_fcntl_rights, &fp);
1361 		if (error)
1362 			return (error);
1363 		if (fp->f_type == DTYPE_PIPE) {
1364 			fdrop(fp, td);
1365 			return (EINVAL);
1366 		}
1367 		fdrop(fp, td);
1368 
1369 		return (kern_fcntl(td, args->fd, F_SETOWN, args->arg));
1370 
1371 	case LINUX_F_DUPFD_CLOEXEC:
1372 		return (kern_fcntl(td, args->fd, F_DUPFD_CLOEXEC, args->arg));
1373 	}
1374 
1375 	return (EINVAL);
1376 }
1377 
1378 int
linux_fcntl(struct thread * td,struct linux_fcntl_args * args)1379 linux_fcntl(struct thread *td, struct linux_fcntl_args *args)
1380 {
1381 
1382 	return (fcntl_common(td, args));
1383 }
1384 
1385 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1386 int
linux_fcntl64(struct thread * td,struct linux_fcntl64_args * args)1387 linux_fcntl64(struct thread *td, struct linux_fcntl64_args *args)
1388 {
1389 	struct l_flock64 linux_flock;
1390 	struct flock bsd_flock;
1391 	struct linux_fcntl_args fcntl_args;
1392 	int error;
1393 
1394 	switch (args->cmd) {
1395 	case LINUX_F_GETLK64:
1396 		error = copyin((void *)args->arg, &linux_flock,
1397 		    sizeof(linux_flock));
1398 		if (error)
1399 			return (error);
1400 		linux_to_bsd_flock64(&linux_flock, &bsd_flock);
1401 		error = kern_fcntl(td, args->fd, F_GETLK, (intptr_t)&bsd_flock);
1402 		if (error)
1403 			return (error);
1404 		bsd_to_linux_flock64(&bsd_flock, &linux_flock);
1405 		return (copyout(&linux_flock, (void *)args->arg,
1406 			    sizeof(linux_flock)));
1407 
1408 	case LINUX_F_SETLK64:
1409 		error = copyin((void *)args->arg, &linux_flock,
1410 		    sizeof(linux_flock));
1411 		if (error)
1412 			return (error);
1413 		linux_to_bsd_flock64(&linux_flock, &bsd_flock);
1414 		return (kern_fcntl(td, args->fd, F_SETLK,
1415 		    (intptr_t)&bsd_flock));
1416 
1417 	case LINUX_F_SETLKW64:
1418 		error = copyin((void *)args->arg, &linux_flock,
1419 		    sizeof(linux_flock));
1420 		if (error)
1421 			return (error);
1422 		linux_to_bsd_flock64(&linux_flock, &bsd_flock);
1423 		return (kern_fcntl(td, args->fd, F_SETLKW,
1424 		    (intptr_t)&bsd_flock));
1425 	}
1426 
1427 	fcntl_args.fd = args->fd;
1428 	fcntl_args.cmd = args->cmd;
1429 	fcntl_args.arg = args->arg;
1430 	return (fcntl_common(td, &fcntl_args));
1431 }
1432 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1433 
1434 #ifdef LINUX_LEGACY_SYSCALLS
1435 int
linux_chown(struct thread * td,struct linux_chown_args * args)1436 linux_chown(struct thread *td, struct linux_chown_args *args)
1437 {
1438 	char *path;
1439 	int error;
1440 
1441 	LCONVPATHEXIST(td, args->path, &path);
1442 
1443 	error = kern_fchownat(td, AT_FDCWD, path, UIO_SYSSPACE, args->uid,
1444 	    args->gid, 0);
1445 	LFREEPATH(path);
1446 	return (error);
1447 }
1448 #endif
1449 
1450 int
linux_fchownat(struct thread * td,struct linux_fchownat_args * args)1451 linux_fchownat(struct thread *td, struct linux_fchownat_args *args)
1452 {
1453 	char *path;
1454 	int error, dfd, flag;
1455 
1456 	if (args->flag & ~LINUX_AT_SYMLINK_NOFOLLOW)
1457 		return (EINVAL);
1458 
1459 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD :  args->dfd;
1460 	LCONVPATHEXIST_AT(td, args->filename, &path, dfd);
1461 
1462 	flag = (args->flag & LINUX_AT_SYMLINK_NOFOLLOW) == 0 ? 0 :
1463 	    AT_SYMLINK_NOFOLLOW;
1464 	error = kern_fchownat(td, dfd, path, UIO_SYSSPACE, args->uid, args->gid,
1465 	    flag);
1466 	LFREEPATH(path);
1467 	return (error);
1468 }
1469 
1470 #ifdef LINUX_LEGACY_SYSCALLS
1471 int
linux_lchown(struct thread * td,struct linux_lchown_args * args)1472 linux_lchown(struct thread *td, struct linux_lchown_args *args)
1473 {
1474 	char *path;
1475 	int error;
1476 
1477 	LCONVPATHEXIST(td, args->path, &path);
1478 
1479 	error = kern_fchownat(td, AT_FDCWD, path, UIO_SYSSPACE, args->uid,
1480 	    args->gid, AT_SYMLINK_NOFOLLOW);
1481 	LFREEPATH(path);
1482 	return (error);
1483 }
1484 #endif
1485 
1486 static int
convert_fadvice(int advice)1487 convert_fadvice(int advice)
1488 {
1489 	switch (advice) {
1490 	case LINUX_POSIX_FADV_NORMAL:
1491 		return (POSIX_FADV_NORMAL);
1492 	case LINUX_POSIX_FADV_RANDOM:
1493 		return (POSIX_FADV_RANDOM);
1494 	case LINUX_POSIX_FADV_SEQUENTIAL:
1495 		return (POSIX_FADV_SEQUENTIAL);
1496 	case LINUX_POSIX_FADV_WILLNEED:
1497 		return (POSIX_FADV_WILLNEED);
1498 	case LINUX_POSIX_FADV_DONTNEED:
1499 		return (POSIX_FADV_DONTNEED);
1500 	case LINUX_POSIX_FADV_NOREUSE:
1501 		return (POSIX_FADV_NOREUSE);
1502 	default:
1503 		return (-1);
1504 	}
1505 }
1506 
1507 int
linux_fadvise64(struct thread * td,struct linux_fadvise64_args * args)1508 linux_fadvise64(struct thread *td, struct linux_fadvise64_args *args)
1509 {
1510 	off_t offset;
1511 	int advice;
1512 
1513 #if defined(__amd64__) && defined(COMPAT_LINUX32)
1514 	offset = PAIR32TO64(off_t, args->offset);
1515 #else
1516 	offset = args->offset;
1517 #endif
1518 
1519 	advice = convert_fadvice(args->advice);
1520 	if (advice == -1)
1521 		return (EINVAL);
1522 	return (kern_posix_fadvise(td, args->fd, offset, args->len, advice));
1523 }
1524 
1525 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1526 int
linux_fadvise64_64(struct thread * td,struct linux_fadvise64_64_args * args)1527 linux_fadvise64_64(struct thread *td, struct linux_fadvise64_64_args *args)
1528 {
1529 	off_t len, offset;
1530 	int advice;
1531 
1532 #if defined(__amd64__) && defined(COMPAT_LINUX32)
1533 	len = PAIR32TO64(off_t, args->len);
1534 	offset = PAIR32TO64(off_t, args->offset);
1535 #else
1536 	len = args->len;
1537 	offset = args->offset;
1538 #endif
1539 
1540 	advice = convert_fadvice(args->advice);
1541 	if (advice == -1)
1542 		return (EINVAL);
1543 	return (kern_posix_fadvise(td, args->fd, offset, len, advice));
1544 }
1545 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1546 
1547 #ifdef LINUX_LEGACY_SYSCALLS
1548 int
linux_pipe(struct thread * td,struct linux_pipe_args * args)1549 linux_pipe(struct thread *td, struct linux_pipe_args *args)
1550 {
1551 	int fildes[2];
1552 	int error;
1553 
1554 	error = kern_pipe(td, fildes, 0, NULL, NULL);
1555 	if (error != 0)
1556 		return (error);
1557 
1558 	error = copyout(fildes, args->pipefds, sizeof(fildes));
1559 	if (error != 0) {
1560 		(void)kern_close(td, fildes[0]);
1561 		(void)kern_close(td, fildes[1]);
1562 	}
1563 
1564 	return (error);
1565 }
1566 #endif
1567 
1568 int
linux_pipe2(struct thread * td,struct linux_pipe2_args * args)1569 linux_pipe2(struct thread *td, struct linux_pipe2_args *args)
1570 {
1571 	int fildes[2];
1572 	int error, flags;
1573 
1574 	if ((args->flags & ~(LINUX_O_NONBLOCK | LINUX_O_CLOEXEC)) != 0)
1575 		return (EINVAL);
1576 
1577 	flags = 0;
1578 	if ((args->flags & LINUX_O_NONBLOCK) != 0)
1579 		flags |= O_NONBLOCK;
1580 	if ((args->flags & LINUX_O_CLOEXEC) != 0)
1581 		flags |= O_CLOEXEC;
1582 	error = kern_pipe(td, fildes, flags, NULL, NULL);
1583 	if (error != 0)
1584 		return (error);
1585 
1586 	error = copyout(fildes, args->pipefds, sizeof(fildes));
1587 	if (error != 0) {
1588 		(void)kern_close(td, fildes[0]);
1589 		(void)kern_close(td, fildes[1]);
1590 	}
1591 
1592 	return (error);
1593 }
1594 
1595 int
linux_dup3(struct thread * td,struct linux_dup3_args * args)1596 linux_dup3(struct thread *td, struct linux_dup3_args *args)
1597 {
1598 	int cmd;
1599 	intptr_t newfd;
1600 
1601 	if (args->oldfd == args->newfd)
1602 		return (EINVAL);
1603 	if ((args->flags & ~LINUX_O_CLOEXEC) != 0)
1604 		return (EINVAL);
1605 	if (args->flags & LINUX_O_CLOEXEC)
1606 		cmd = F_DUP2FD_CLOEXEC;
1607 	else
1608 		cmd = F_DUP2FD;
1609 
1610 	newfd = args->newfd;
1611 	return (kern_fcntl(td, args->oldfd, cmd, newfd));
1612 }
1613 
1614 int
linux_fallocate(struct thread * td,struct linux_fallocate_args * args)1615 linux_fallocate(struct thread *td, struct linux_fallocate_args *args)
1616 {
1617 	off_t len, offset;
1618 
1619 	/*
1620 	 * We emulate only posix_fallocate system call for which
1621 	 * mode should be 0.
1622 	 */
1623 	if (args->mode != 0)
1624 		return (EOPNOTSUPP);
1625 
1626 #if defined(__amd64__) && defined(COMPAT_LINUX32)
1627 	len = PAIR32TO64(off_t, args->len);
1628 	offset = PAIR32TO64(off_t, args->offset);
1629 #else
1630 	len = args->len;
1631 	offset = args->offset;
1632 #endif
1633 
1634 	return (kern_posix_fallocate(td, args->fd, offset, len));
1635 }
1636 
1637 int
linux_splice(struct thread * td,struct linux_splice_args * args)1638 linux_splice(struct thread *td, struct linux_splice_args *args)
1639 {
1640 
1641 	linux_msg(td, "syscall splice not really implemented");
1642 
1643 	/*
1644 	 * splice(2) is documented to return EINVAL in various circumstances;
1645 	 * returning it instead of ENOSYS should hint the caller to use fallback
1646 	 * instead.
1647 	 */
1648 	return (EINVAL);
1649 }
1650