xref: /freebsd/sys/fs/fuse/fuse_vfsops.c (revision 54e9e4e72d711fb41f88f793f6c64df1126112f9)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 2007-2009 Google Inc. and Amit Singh
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 are
9  * met:
10  *
11  * * Redistributions of source code must retain the above copyright
12  *   notice, this list of conditions and the following disclaimer.
13  * * Redistributions in binary form must reproduce the above
14  *   copyright notice, this list of conditions and the following disclaimer
15  *   in the documentation and/or other materials provided with the
16  *   distribution.
17  * * Neither the name of Google Inc. nor the names of its
18  *   contributors may be used to endorse or promote products derived from
19  *   this software without specific prior written permission.
20  *
21  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
22  * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
23  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
24  * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
25  * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
26  * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
27  * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
28  * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
29  * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
30  * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
31  * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
32  *
33  * Copyright (C) 2005 Csaba Henk.
34  * All rights reserved.
35  *
36  * Redistribution and use in source and binary forms, with or without
37  * modification, are permitted provided that the following conditions
38  * are met:
39  * 1. Redistributions of source code must retain the above copyright
40  *    notice, this list of conditions and the following disclaimer.
41  * 2. Redistributions in binary form must reproduce the above copyright
42  *    notice, this list of conditions and the following disclaimer in the
43  *    documentation and/or other materials provided with the distribution.
44  *
45  * THIS SOFTWARE IS PROVIDED BY AUTHOR AND CONTRIBUTORS ``AS IS'' AND
46  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
47  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
48  * ARE DISCLAIMED.  IN NO EVENT SHALL AUTHOR OR CONTRIBUTORS BE LIABLE
49  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
50  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
51  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
52  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
53  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
54  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
55  * SUCH DAMAGE.
56  */
57 
58 #include <sys/cdefs.h>
59 __FBSDID("$FreeBSD$");
60 
61 #include <sys/param.h>
62 #include <sys/buf.h>
63 #include <sys/module.h>
64 #include <sys/systm.h>
65 #include <sys/errno.h>
66 #include <sys/kernel.h>
67 #include <sys/capsicum.h>
68 #include <sys/conf.h>
69 #include <sys/filedesc.h>
70 #include <sys/uio.h>
71 #include <sys/malloc.h>
72 #include <sys/queue.h>
73 #include <sys/lock.h>
74 #include <sys/sx.h>
75 #include <sys/mutex.h>
76 #include <sys/proc.h>
77 #include <sys/vnode.h>
78 #include <sys/namei.h>
79 #include <sys/mount.h>
80 #include <sys/sysctl.h>
81 #include <sys/fcntl.h>
82 
83 #include "fuse.h"
84 #include "fuse_param.h"
85 #include "fuse_node.h"
86 #include "fuse_ipc.h"
87 #include "fuse_internal.h"
88 
89 #include <sys/priv.h>
90 #include <security/mac/mac_framework.h>
91 
92 SDT_PROVIDER_DECLARE(fuse);
93 /*
94  * Fuse trace probe:
95  * arg0: verbosity.  Higher numbers give more verbose messages
96  * arg1: Textual message
97  */
98 SDT_PROBE_DEFINE2(fuse, , vfsops, trace, "int", "char*");
99 
100 /* This will do for privilege types for now */
101 #ifndef PRIV_VFS_FUSE_ALLOWOTHER
102 #define PRIV_VFS_FUSE_ALLOWOTHER PRIV_VFS_MOUNT_NONUSER
103 #endif
104 #ifndef PRIV_VFS_FUSE_MOUNT_NONUSER
105 #define PRIV_VFS_FUSE_MOUNT_NONUSER PRIV_VFS_MOUNT_NONUSER
106 #endif
107 #ifndef PRIV_VFS_FUSE_SYNC_UNMOUNT
108 #define PRIV_VFS_FUSE_SYNC_UNMOUNT PRIV_VFS_MOUNT_NONUSER
109 #endif
110 
111 static vfs_mount_t fuse_vfsop_mount;
112 static vfs_unmount_t fuse_vfsop_unmount;
113 static vfs_root_t fuse_vfsop_root;
114 static vfs_statfs_t fuse_vfsop_statfs;
115 
116 struct vfsops fuse_vfsops = {
117 	.vfs_mount = fuse_vfsop_mount,
118 	.vfs_unmount = fuse_vfsop_unmount,
119 	.vfs_root = fuse_vfsop_root,
120 	.vfs_statfs = fuse_vfsop_statfs,
121 };
122 
123 SYSCTL_INT(_vfs_fusefs, OID_AUTO, init_backgrounded, CTLFLAG_RD,
124     SYSCTL_NULL_INT_PTR, 1, "indicate async handshake");
125 static int fuse_enforce_dev_perms = 0;
126 
127 SYSCTL_INT(_vfs_fusefs, OID_AUTO, enforce_dev_perms, CTLFLAG_RW,
128     &fuse_enforce_dev_perms, 0,
129     "enforce fuse device permissions for secondary mounts");
130 static unsigned sync_unmount = 1;
131 
132 SYSCTL_UINT(_vfs_fusefs, OID_AUTO, sync_unmount, CTLFLAG_RW,
133     &sync_unmount, 0, "specify when to use synchronous unmount");
134 
135 MALLOC_DEFINE(M_FUSEVFS, "fuse_filesystem", "buffer for fuse vfs layer");
136 
137 static int
138 fuse_getdevice(const char *fspec, struct thread *td, struct cdev **fdevp)
139 {
140 	struct nameidata nd, *ndp = &nd;
141 	struct vnode *devvp;
142 	struct cdev *fdev;
143 	int err;
144 
145 	/*
146 	 * Not an update, or updating the name: look up the name
147 	 * and verify that it refers to a sensible disk device.
148 	 */
149 
150 	NDINIT(ndp, LOOKUP, FOLLOW, UIO_SYSSPACE, fspec, td);
151 	if ((err = namei(ndp)) != 0)
152 		return err;
153 	NDFREE(ndp, NDF_ONLY_PNBUF);
154 	devvp = ndp->ni_vp;
155 
156 	if (devvp->v_type != VCHR) {
157 		vrele(devvp);
158 		return ENXIO;
159 	}
160 	fdev = devvp->v_rdev;
161 	dev_ref(fdev);
162 
163 	if (fuse_enforce_dev_perms) {
164 		/*
165 	         * Check if mounter can open the fuse device.
166 	         *
167 	         * This has significance only if we are doing a secondary mount
168 	         * which doesn't involve actually opening fuse devices, but we
169 	         * still want to enforce the permissions of the device (in
170 	         * order to keep control over the circle of fuse users).
171 	         *
172 	         * (In case of primary mounts, we are either the superuser so
173 	         * we can do anything anyway, or we can mount only if the
174 	         * device is already opened by us, ie. we are permitted to open
175 	         * the device.)
176 	         */
177 #if 0
178 #ifdef MAC
179 		err = mac_check_vnode_open(td->td_ucred, devvp, VREAD | VWRITE);
180 		if (!err)
181 #endif
182 #endif /* 0 */
183 			err = VOP_ACCESS(devvp, VREAD | VWRITE, td->td_ucred, td);
184 		if (err) {
185 			vrele(devvp);
186 			dev_rel(fdev);
187 			return err;
188 		}
189 	}
190 	/*
191 	 * according to coda code, no extra lock is needed --
192 	 * although in sys/vnode.h this field is marked "v"
193 	 */
194 	vrele(devvp);
195 
196 	if (!fdev->si_devsw ||
197 	    strcmp("fuse", fdev->si_devsw->d_name)) {
198 		dev_rel(fdev);
199 		return ENXIO;
200 	}
201 	*fdevp = fdev;
202 
203 	return 0;
204 }
205 
206 #define FUSE_FLAGOPT(fnam, fval) do {				\
207 	vfs_flagopt(opts, #fnam, &mntopts, fval);		\
208 	vfs_flagopt(opts, "__" #fnam, &__mntopts, fval);	\
209 } while (0)
210 
211 SDT_PROBE_DEFINE1(fuse, , vfsops, mntopts, "uint64_t");
212 SDT_PROBE_DEFINE4(fuse, , vfsops, mount_err, "char*", "struct fuse_data*",
213 	"struct mount*", "int");
214 
215 static int
216 fuse_vfsop_mount(struct mount *mp)
217 {
218 	int err;
219 
220 	uint64_t mntopts, __mntopts;
221 	uint32_t max_read;
222 	int daemon_timeout;
223 	int fd;
224 
225 	size_t len;
226 
227 	struct cdev *fdev;
228 	struct fuse_data *data = NULL;
229 	struct thread *td;
230 	struct file *fp, *fptmp;
231 	char *fspec, *subtype;
232 	struct vfsoptlist *opts;
233 
234 	subtype = NULL;
235 	max_read = ~0;
236 	err = 0;
237 	mntopts = 0;
238 	__mntopts = 0;
239 	td = curthread;
240 
241 	if (mp->mnt_flag & MNT_UPDATE)
242 		return EOPNOTSUPP;
243 
244 	MNT_ILOCK(mp);
245 	mp->mnt_flag |= MNT_SYNCHRONOUS;
246 	mp->mnt_data = NULL;
247 	MNT_IUNLOCK(mp);
248 	/* Get the new options passed to mount */
249 	opts = mp->mnt_optnew;
250 
251 	if (!opts)
252 		return EINVAL;
253 
254 	/* `fspath' contains the mount point (eg. /mnt/fuse/sshfs); REQUIRED */
255 	if (!vfs_getopts(opts, "fspath", &err))
256 		return err;
257 
258 	/* `from' contains the device name (eg. /dev/fuse0); REQUIRED */
259 	fspec = vfs_getopts(opts, "from", &err);
260 	if (!fspec)
261 		return err;
262 
263 	/* `fd' contains the filedescriptor for this session; REQUIRED */
264 	if (vfs_scanopt(opts, "fd", "%d", &fd) != 1)
265 		return EINVAL;
266 
267 	err = fuse_getdevice(fspec, td, &fdev);
268 	if (err != 0)
269 		return err;
270 
271 	/*
272 	 * With the help of underscored options the mount program
273 	 * can inform us from the flags it sets by default
274 	 */
275 	FUSE_FLAGOPT(allow_other, FSESS_DAEMON_CAN_SPY);
276 	FUSE_FLAGOPT(push_symlinks_in, FSESS_PUSH_SYMLINKS_IN);
277 	FUSE_FLAGOPT(default_permissions, FSESS_DEFAULT_PERMISSIONS);
278 	FUSE_FLAGOPT(no_attrcache, FSESS_NO_ATTRCACHE);
279 	FUSE_FLAGOPT(no_readahed, FSESS_NO_READAHEAD);
280 	FUSE_FLAGOPT(no_datacache, FSESS_NO_DATACACHE);
281 	FUSE_FLAGOPT(no_namecache, FSESS_NO_NAMECACHE);
282 	FUSE_FLAGOPT(no_mmap, FSESS_NO_MMAP);
283 	FUSE_FLAGOPT(brokenio, FSESS_BROKENIO);
284 
285 	(void)vfs_scanopt(opts, "max_read=", "%u", &max_read);
286 	if (vfs_scanopt(opts, "timeout=", "%u", &daemon_timeout) == 1) {
287 		if (daemon_timeout < FUSE_MIN_DAEMON_TIMEOUT)
288 			daemon_timeout = FUSE_MIN_DAEMON_TIMEOUT;
289 		else if (daemon_timeout > FUSE_MAX_DAEMON_TIMEOUT)
290 			daemon_timeout = FUSE_MAX_DAEMON_TIMEOUT;
291 	} else {
292 		daemon_timeout = FUSE_DEFAULT_DAEMON_TIMEOUT;
293 	}
294 	subtype = vfs_getopts(opts, "subtype=", &err);
295 
296 	SDT_PROBE1(fuse, , vfsops, mntopts, mntopts);
297 
298 	err = fget(td, fd, &cap_read_rights, &fp);
299 	if (err != 0) {
300 		SDT_PROBE2(fuse, , vfsops, trace, 1,
301 			"invalid or not opened device");
302 		goto out;
303 	}
304 	fptmp = td->td_fpop;
305 	td->td_fpop = fp;
306 	err = devfs_get_cdevpriv((void **)&data);
307 	td->td_fpop = fptmp;
308 	fdrop(fp, td);
309 	FUSE_LOCK();
310 	if (err != 0 || data == NULL || data->mp != NULL) {
311 		err = ENXIO;
312 		SDT_PROBE4(fuse, , vfsops, mount_err,
313 			"invalid or not opened device", data, mp, err);
314 		FUSE_UNLOCK();
315 		goto out;
316 	}
317 	if (fdata_get_dead(data)) {
318 		err = ENOTCONN;
319 		SDT_PROBE4(fuse, , vfsops, mount_err,
320 			"device is dead during mount", data, mp, err);
321 		FUSE_UNLOCK();
322 		goto out;
323 	}
324 	/* Sanity + permission checks */
325 	if (!data->daemoncred)
326 		panic("fuse daemon found, but identity unknown");
327 	if (mntopts & FSESS_DAEMON_CAN_SPY)
328 		err = priv_check(td, PRIV_VFS_FUSE_ALLOWOTHER);
329 	if (err == 0 && td->td_ucred->cr_uid != data->daemoncred->cr_uid)
330 		/* are we allowed to do the first mount? */
331 		err = priv_check(td, PRIV_VFS_FUSE_MOUNT_NONUSER);
332 	if (err) {
333 		FUSE_UNLOCK();
334 		goto out;
335 	}
336 	data->ref++;
337 	data->mp = mp;
338 	data->dataflags |= mntopts;
339 	data->max_read = max_read;
340 	data->daemon_timeout = daemon_timeout;
341 	FUSE_UNLOCK();
342 
343 	vfs_getnewfsid(mp);
344 	MNT_ILOCK(mp);
345 	mp->mnt_data = data;
346 	mp->mnt_flag |= MNT_LOCAL;
347 	mp->mnt_kern_flag |= MNTK_USES_BCACHE;
348 	MNT_IUNLOCK(mp);
349 	/* We need this here as this slot is used by getnewvnode() */
350 	mp->mnt_stat.f_iosize = maxbcachebuf;
351 	if (subtype) {
352 		strlcat(mp->mnt_stat.f_fstypename, ".", MFSNAMELEN);
353 		strlcat(mp->mnt_stat.f_fstypename, subtype, MFSNAMELEN);
354 	}
355 	copystr(fspec, mp->mnt_stat.f_mntfromname, MNAMELEN - 1, &len);
356 	bzero(mp->mnt_stat.f_mntfromname + len, MNAMELEN - len);
357 
358 	/* Now handshaking with daemon */
359 	fuse_internal_send_init(data, td);
360 
361 out:
362 	if (err) {
363 		FUSE_LOCK();
364 		if (data != NULL && data->mp == mp) {
365 			/*
366 			 * Destroy device only if we acquired reference to
367 			 * it
368 			 */
369 			SDT_PROBE4(fuse, , vfsops, mount_err,
370 				"mount failed, destroy device", data, mp, err);
371 			data->mp = NULL;
372 			fdata_trydestroy(data);
373 		}
374 		FUSE_UNLOCK();
375 		dev_rel(fdev);
376 	}
377 	return err;
378 }
379 
380 static int
381 fuse_vfsop_unmount(struct mount *mp, int mntflags)
382 {
383 	int err = 0;
384 	int flags = 0;
385 
386 	struct cdev *fdev;
387 	struct fuse_data *data;
388 	struct fuse_dispatcher fdi;
389 	struct thread *td = curthread;
390 
391 	if (mntflags & MNT_FORCE) {
392 		flags |= FORCECLOSE;
393 	}
394 	data = fuse_get_mpdata(mp);
395 	if (!data) {
396 		panic("no private data for mount point?");
397 	}
398 	/* There is 1 extra root vnode reference (mp->mnt_data). */
399 	FUSE_LOCK();
400 	if (data->vroot != NULL) {
401 		struct vnode *vroot = data->vroot;
402 
403 		data->vroot = NULL;
404 		FUSE_UNLOCK();
405 		vrele(vroot);
406 	} else
407 		FUSE_UNLOCK();
408 	err = vflush(mp, 0, flags, td);
409 	if (err) {
410 		return err;
411 	}
412 	if (fdata_get_dead(data)) {
413 		goto alreadydead;
414 	}
415 	fdisp_init(&fdi, 0);
416 	fdisp_make(&fdi, FUSE_DESTROY, mp, 0, td, NULL);
417 
418 	err = fdisp_wait_answ(&fdi);
419 	fdisp_destroy(&fdi);
420 
421 	fdata_set_dead(data);
422 
423 alreadydead:
424 	FUSE_LOCK();
425 	data->mp = NULL;
426 	fdev = data->fdev;
427 	fdata_trydestroy(data);
428 	FUSE_UNLOCK();
429 
430 	MNT_ILOCK(mp);
431 	mp->mnt_data = NULL;
432 	mp->mnt_flag &= ~MNT_LOCAL;
433 	MNT_IUNLOCK(mp);
434 
435 	dev_rel(fdev);
436 
437 	return 0;
438 }
439 
440 static int
441 fuse_vfsop_root(struct mount *mp, int lkflags, struct vnode **vpp)
442 {
443 	struct fuse_data *data = fuse_get_mpdata(mp);
444 	int err = 0;
445 
446 	if (data->vroot != NULL) {
447 		err = vget(data->vroot, lkflags, curthread);
448 		if (err == 0)
449 			*vpp = data->vroot;
450 	} else {
451 		err = fuse_vnode_get(mp, NULL, FUSE_ROOT_ID, NULL, vpp, NULL,
452 		    VDIR);
453 		if (err == 0) {
454 			FUSE_LOCK();
455 			MPASS(data->vroot == NULL || data->vroot == *vpp);
456 			if (data->vroot == NULL) {
457 				SDT_PROBE2(fuse, , vfsops, trace, 1,
458 					"new root vnode");
459 				data->vroot = *vpp;
460 				FUSE_UNLOCK();
461 				vref(*vpp);
462 			} else if (data->vroot != *vpp) {
463 				SDT_PROBE2(fuse, , vfsops, trace, 1,
464 					"root vnode race");
465 				FUSE_UNLOCK();
466 				VOP_UNLOCK(*vpp, 0);
467 				vrele(*vpp);
468 				vrecycle(*vpp);
469 				*vpp = data->vroot;
470 			} else
471 				FUSE_UNLOCK();
472 		}
473 	}
474 	return err;
475 }
476 
477 static int
478 fuse_vfsop_statfs(struct mount *mp, struct statfs *sbp)
479 {
480 	struct fuse_dispatcher fdi;
481 	int err = 0;
482 
483 	struct fuse_statfs_out *fsfo;
484 	struct fuse_data *data;
485 
486 	data = fuse_get_mpdata(mp);
487 
488 	if (!(data->dataflags & FSESS_INITED))
489 		goto fake;
490 
491 	fdisp_init(&fdi, 0);
492 	fdisp_make(&fdi, FUSE_STATFS, mp, FUSE_ROOT_ID, NULL, NULL);
493 	err = fdisp_wait_answ(&fdi);
494 	if (err) {
495 		fdisp_destroy(&fdi);
496 		if (err == ENOTCONN) {
497 			/*
498 	                 * We want to seem a legitimate fs even if the daemon
499 	                 * is stiff dead... (so that, eg., we can still do path
500 	                 * based unmounting after the daemon dies).
501 	                 */
502 			goto fake;
503 		}
504 		return err;
505 	}
506 	fsfo = fdi.answ;
507 
508 	sbp->f_blocks = fsfo->st.blocks;
509 	sbp->f_bfree = fsfo->st.bfree;
510 	sbp->f_bavail = fsfo->st.bavail;
511 	sbp->f_files = fsfo->st.files;
512 	sbp->f_ffree = fsfo->st.ffree;	/* cast from uint64_t to int64_t */
513 	sbp->f_namemax = fsfo->st.namelen;
514 	sbp->f_bsize = fsfo->st.frsize;	/* cast from uint32_t to uint64_t */
515 
516 	fdisp_destroy(&fdi);
517 	return 0;
518 
519 fake:
520 	sbp->f_blocks = 0;
521 	sbp->f_bfree = 0;
522 	sbp->f_bavail = 0;
523 	sbp->f_files = 0;
524 	sbp->f_ffree = 0;
525 	sbp->f_namemax = 0;
526 	sbp->f_bsize = FUSE_DEFAULT_BLOCKSIZE;
527 
528 	return 0;
529 }
530