xref: /freebsd/sys/fs/fuse/fuse_vfsops.c (revision ab40f58ccfe6c07ebefddc72f4661a52fe746353)
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/types.h>
62 #include <sys/module.h>
63 #include <sys/systm.h>
64 #include <sys/errno.h>
65 #include <sys/param.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 #define FUSE_DEBUG_MODULE VFSOPS
93 #include "fuse_debug.h"
94 
95 /* This will do for privilege types for now */
96 #ifndef PRIV_VFS_FUSE_ALLOWOTHER
97 #define PRIV_VFS_FUSE_ALLOWOTHER PRIV_VFS_MOUNT_NONUSER
98 #endif
99 #ifndef PRIV_VFS_FUSE_MOUNT_NONUSER
100 #define PRIV_VFS_FUSE_MOUNT_NONUSER PRIV_VFS_MOUNT_NONUSER
101 #endif
102 #ifndef PRIV_VFS_FUSE_SYNC_UNMOUNT
103 #define PRIV_VFS_FUSE_SYNC_UNMOUNT PRIV_VFS_MOUNT_NONUSER
104 #endif
105 
106 static vfs_mount_t fuse_vfsop_mount;
107 static vfs_unmount_t fuse_vfsop_unmount;
108 static vfs_root_t fuse_vfsop_root;
109 static vfs_statfs_t fuse_vfsop_statfs;
110 
111 struct vfsops fuse_vfsops = {
112 	.vfs_mount = fuse_vfsop_mount,
113 	.vfs_unmount = fuse_vfsop_unmount,
114 	.vfs_root = fuse_vfsop_root,
115 	.vfs_statfs = fuse_vfsop_statfs,
116 };
117 
118 SYSCTL_INT(_vfs_fuse, OID_AUTO, init_backgrounded, CTLFLAG_RD,
119     SYSCTL_NULL_INT_PTR, 1, "indicate async handshake");
120 static int fuse_enforce_dev_perms = 0;
121 
122 SYSCTL_INT(_vfs_fuse, OID_AUTO, enforce_dev_perms, CTLFLAG_RW,
123     &fuse_enforce_dev_perms, 0,
124     "enforce fuse device permissions for secondary mounts");
125 static unsigned sync_unmount = 1;
126 
127 SYSCTL_UINT(_vfs_fuse, OID_AUTO, sync_unmount, CTLFLAG_RW,
128     &sync_unmount, 0, "specify when to use synchronous unmount");
129 
130 MALLOC_DEFINE(M_FUSEVFS, "fuse_filesystem", "buffer for fuse vfs layer");
131 
132 static int
133 fuse_getdevice(const char *fspec, struct thread *td, struct cdev **fdevp)
134 {
135 	struct nameidata nd, *ndp = &nd;
136 	struct vnode *devvp;
137 	struct cdev *fdev;
138 	int err;
139 
140 	/*
141          * Not an update, or updating the name: look up the name
142          * and verify that it refers to a sensible disk device.
143          */
144 
145 	NDINIT(ndp, LOOKUP, FOLLOW, UIO_SYSSPACE, fspec, td);
146 	if ((err = namei(ndp)) != 0)
147 		return err;
148 	NDFREE(ndp, NDF_ONLY_PNBUF);
149 	devvp = ndp->ni_vp;
150 
151 	if (devvp->v_type != VCHR) {
152 		vrele(devvp);
153 		return ENXIO;
154 	}
155 	fdev = devvp->v_rdev;
156 	dev_ref(fdev);
157 
158 	if (fuse_enforce_dev_perms) {
159 		/*
160 	         * Check if mounter can open the fuse device.
161 	         *
162 	         * This has significance only if we are doing a secondary mount
163 	         * which doesn't involve actually opening fuse devices, but we
164 	         * still want to enforce the permissions of the device (in
165 	         * order to keep control over the circle of fuse users).
166 	         *
167 	         * (In case of primary mounts, we are either the superuser so
168 	         * we can do anything anyway, or we can mount only if the
169 	         * device is already opened by us, ie. we are permitted to open
170 	         * the device.)
171 	         */
172 #if 0
173 #ifdef MAC
174 		err = mac_check_vnode_open(td->td_ucred, devvp, VREAD | VWRITE);
175 		if (!err)
176 #endif
177 #endif /* 0 */
178 			err = VOP_ACCESS(devvp, VREAD | VWRITE, td->td_ucred, td);
179 		if (err) {
180 			vrele(devvp);
181 			dev_rel(fdev);
182 			return err;
183 		}
184 	}
185 	/*
186          * according to coda code, no extra lock is needed --
187          * although in sys/vnode.h this field is marked "v"
188          */
189 	vrele(devvp);
190 
191 	if (!fdev->si_devsw ||
192 	    strcmp("fuse", fdev->si_devsw->d_name)) {
193 		dev_rel(fdev);
194 		return ENXIO;
195 	}
196 	*fdevp = fdev;
197 
198 	return 0;
199 }
200 
201 #define FUSE_FLAGOPT(fnam, fval) do {				\
202     vfs_flagopt(opts, #fnam, &mntopts, fval);		\
203     vfs_flagopt(opts, "__" #fnam, &__mntopts, fval);	\
204 } while (0)
205 
206 static int
207 fuse_vfsop_mount(struct mount *mp)
208 {
209 	int err;
210 
211 	uint64_t mntopts, __mntopts;
212 	int max_read_set;
213 	uint32_t max_read;
214 	int daemon_timeout;
215 	int fd;
216 
217 	size_t len;
218 
219 	struct cdev *fdev;
220 	struct fuse_data *data;
221 	struct thread *td;
222 	struct file *fp, *fptmp;
223 	char *fspec, *subtype;
224 	struct vfsoptlist *opts;
225 
226 	subtype = NULL;
227 	max_read_set = 0;
228 	max_read = ~0;
229 	err = 0;
230 	mntopts = 0;
231 	__mntopts = 0;
232 	td = curthread;
233 
234 	fuse_trace_printf_vfsop();
235 
236 	if (mp->mnt_flag & MNT_UPDATE)
237 		return EOPNOTSUPP;
238 
239 	MNT_ILOCK(mp);
240 	mp->mnt_flag |= MNT_SYNCHRONOUS;
241 	mp->mnt_data = NULL;
242 	MNT_IUNLOCK(mp);
243 	/* Get the new options passed to mount */
244 	opts = mp->mnt_optnew;
245 
246 	if (!opts)
247 		return EINVAL;
248 
249 	/* `fspath' contains the mount point (eg. /mnt/fuse/sshfs); REQUIRED */
250 	if (!vfs_getopts(opts, "fspath", &err))
251 		return err;
252 
253 	/* `from' contains the device name (eg. /dev/fuse0); REQUIRED */
254 	fspec = vfs_getopts(opts, "from", &err);
255 	if (!fspec)
256 		return err;
257 
258 	/* `fd' contains the filedescriptor for this session; REQUIRED */
259 	if (vfs_scanopt(opts, "fd", "%d", &fd) != 1)
260 		return EINVAL;
261 
262 	err = fuse_getdevice(fspec, td, &fdev);
263 	if (err != 0)
264 		return err;
265 
266 	/*
267          * With the help of underscored options the mount program
268          * can inform us from the flags it sets by default
269          */
270 	FUSE_FLAGOPT(allow_other, FSESS_DAEMON_CAN_SPY);
271 	FUSE_FLAGOPT(push_symlinks_in, FSESS_PUSH_SYMLINKS_IN);
272 	FUSE_FLAGOPT(default_permissions, FSESS_DEFAULT_PERMISSIONS);
273 	FUSE_FLAGOPT(no_attrcache, FSESS_NO_ATTRCACHE);
274 	FUSE_FLAGOPT(no_readahed, FSESS_NO_READAHEAD);
275 	FUSE_FLAGOPT(no_datacache, FSESS_NO_DATACACHE);
276 	FUSE_FLAGOPT(no_namecache, FSESS_NO_NAMECACHE);
277 	FUSE_FLAGOPT(no_mmap, FSESS_NO_MMAP);
278 	FUSE_FLAGOPT(brokenio, FSESS_BROKENIO);
279 
280 	if (vfs_scanopt(opts, "max_read=", "%u", &max_read) == 1)
281 		max_read_set = 1;
282 	if (vfs_scanopt(opts, "timeout=", "%u", &daemon_timeout) == 1) {
283 		if (daemon_timeout < FUSE_MIN_DAEMON_TIMEOUT)
284 			daemon_timeout = FUSE_MIN_DAEMON_TIMEOUT;
285 		else if (daemon_timeout > FUSE_MAX_DAEMON_TIMEOUT)
286 			daemon_timeout = FUSE_MAX_DAEMON_TIMEOUT;
287 	} else {
288 		daemon_timeout = FUSE_DEFAULT_DAEMON_TIMEOUT;
289 	}
290 	subtype = vfs_getopts(opts, "subtype=", &err);
291 
292 	FS_DEBUG2G("mntopts 0x%jx\n", (uintmax_t)mntopts);
293 
294 	err = fget(td, fd, &cap_read_rights, &fp);
295 	if (err != 0) {
296 		FS_DEBUG("invalid or not opened device: data=%p\n", data);
297 		goto out;
298 	}
299 	fptmp = td->td_fpop;
300 	td->td_fpop = fp;
301         err = devfs_get_cdevpriv((void **)&data);
302 	td->td_fpop = fptmp;
303 	fdrop(fp, td);
304 	FUSE_LOCK();
305 	if (err != 0 || data == NULL || data->mp != NULL) {
306 		FS_DEBUG("invalid or not opened device: data=%p data.mp=%p\n",
307 		    data, data != NULL ? data->mp : NULL);
308 		err = ENXIO;
309 		FUSE_UNLOCK();
310 		goto out;
311 	}
312 	if (fdata_get_dead(data)) {
313 		FS_DEBUG("device is dead during mount: data=%p\n", data);
314 		err = ENOTCONN;
315 		FUSE_UNLOCK();
316 		goto out;
317 	}
318 	/* Sanity + permission checks */
319 	if (!data->daemoncred)
320 		panic("fuse daemon found, but identity unknown");
321 	if (mntopts & FSESS_DAEMON_CAN_SPY)
322 		err = priv_check(td, PRIV_VFS_FUSE_ALLOWOTHER);
323 	if (err == 0 && td->td_ucred->cr_uid != data->daemoncred->cr_uid)
324 		/* are we allowed to do the first mount? */
325 		err = priv_check(td, PRIV_VFS_FUSE_MOUNT_NONUSER);
326 	if (err) {
327 		FUSE_UNLOCK();
328 		goto out;
329 	}
330 	data->ref++;
331 	data->mp = mp;
332 	data->dataflags |= mntopts;
333 	data->max_read = max_read;
334 	data->daemon_timeout = daemon_timeout;
335 	FUSE_UNLOCK();
336 
337 	vfs_getnewfsid(mp);
338 	MNT_ILOCK(mp);
339 	mp->mnt_data = data;
340 	mp->mnt_flag |= MNT_LOCAL;
341 	mp->mnt_kern_flag |= MNTK_USES_BCACHE;
342 	MNT_IUNLOCK(mp);
343 	/* We need this here as this slot is used by getnewvnode() */
344 	mp->mnt_stat.f_iosize = PAGE_SIZE;
345 	if (subtype) {
346 		strlcat(mp->mnt_stat.f_fstypename, ".", MFSNAMELEN);
347 		strlcat(mp->mnt_stat.f_fstypename, subtype, MFSNAMELEN);
348 	}
349 	copystr(fspec, mp->mnt_stat.f_mntfromname, MNAMELEN - 1, &len);
350 	bzero(mp->mnt_stat.f_mntfromname + len, MNAMELEN - len);
351 	FS_DEBUG2G("mp %p: %s\n", mp, mp->mnt_stat.f_mntfromname);
352 
353 	/* Now handshaking with daemon */
354 	fuse_internal_send_init(data, td);
355 
356 out:
357 	if (err) {
358 		FUSE_LOCK();
359 		if (data->mp == mp) {
360 			/*
361 			 * Destroy device only if we acquired reference to
362 			 * it
363 			 */
364 			FS_DEBUG("mount failed, destroy device: data=%p mp=%p"
365 			      " err=%d\n",
366 			    data, mp, err);
367 			data->mp = NULL;
368 			fdata_trydestroy(data);
369 		}
370 		FUSE_UNLOCK();
371 		dev_rel(fdev);
372 	}
373 	return err;
374 }
375 
376 static int
377 fuse_vfsop_unmount(struct mount *mp, int mntflags)
378 {
379 	int err = 0;
380 	int flags = 0;
381 
382 	struct cdev *fdev;
383 	struct fuse_data *data;
384 	struct fuse_dispatcher fdi;
385 	struct thread *td = curthread;
386 
387 	fuse_trace_printf_vfsop();
388 
389 	if (mntflags & MNT_FORCE) {
390 		flags |= FORCECLOSE;
391 	}
392 	data = fuse_get_mpdata(mp);
393 	if (!data) {
394 		panic("no private data for mount point?");
395 	}
396 	/* There is 1 extra root vnode reference (mp->mnt_data). */
397 	FUSE_LOCK();
398 	if (data->vroot != NULL) {
399 		struct vnode *vroot = data->vroot;
400 
401 		data->vroot = NULL;
402 		FUSE_UNLOCK();
403 		vrele(vroot);
404 	} else
405 		FUSE_UNLOCK();
406 	err = vflush(mp, 0, flags, td);
407 	if (err) {
408 		debug_printf("vflush failed");
409 		return err;
410 	}
411 	if (fdata_get_dead(data)) {
412 		goto alreadydead;
413 	}
414 	fdisp_init(&fdi, 0);
415 	fdisp_make(&fdi, FUSE_DESTROY, mp, 0, td, NULL);
416 
417 	err = fdisp_wait_answ(&fdi);
418 	fdisp_destroy(&fdi);
419 
420 	fdata_set_dead(data);
421 
422 alreadydead:
423 	FUSE_LOCK();
424 	data->mp = NULL;
425 	fdev = data->fdev;
426 	fdata_trydestroy(data);
427 	FUSE_UNLOCK();
428 
429 	MNT_ILOCK(mp);
430 	mp->mnt_data = NULL;
431 	mp->mnt_flag &= ~MNT_LOCAL;
432 	MNT_IUNLOCK(mp);
433 
434 	dev_rel(fdev);
435 
436 	return 0;
437 }
438 
439 static int
440 fuse_vfsop_root(struct mount *mp, int lkflags, struct vnode **vpp)
441 {
442 	struct fuse_data *data = fuse_get_mpdata(mp);
443 	int err = 0;
444 
445 	if (data->vroot != NULL) {
446 		err = vget(data->vroot, lkflags, curthread);
447 		if (err == 0)
448 			*vpp = data->vroot;
449 	} else {
450 		err = fuse_vnode_get(mp, FUSE_ROOT_ID, NULL, vpp, NULL, VDIR);
451 		if (err == 0) {
452 			FUSE_LOCK();
453 			MPASS(data->vroot == NULL || data->vroot == *vpp);
454 			if (data->vroot == NULL) {
455 				FS_DEBUG("new root vnode\n");
456 				data->vroot = *vpp;
457 				FUSE_UNLOCK();
458 				vref(*vpp);
459 			} else if (data->vroot != *vpp) {
460 				FS_DEBUG("root vnode race\n");
461 				FUSE_UNLOCK();
462 				VOP_UNLOCK(*vpp, 0);
463 				vrele(*vpp);
464 				vrecycle(*vpp);
465 				*vpp = data->vroot;
466 			} else
467 				FUSE_UNLOCK();
468 		}
469 	}
470 	return err;
471 }
472 
473 static int
474 fuse_vfsop_statfs(struct mount *mp, struct statfs *sbp)
475 {
476 	struct fuse_dispatcher fdi;
477 	int err = 0;
478 
479 	struct fuse_statfs_out *fsfo;
480 	struct fuse_data *data;
481 
482 	FS_DEBUG2G("mp %p: %s\n", mp, mp->mnt_stat.f_mntfromname);
483 	data = fuse_get_mpdata(mp);
484 
485 	if (!(data->dataflags & FSESS_INITED))
486 		goto fake;
487 
488 	fdisp_init(&fdi, 0);
489 	fdisp_make(&fdi, FUSE_STATFS, mp, FUSE_ROOT_ID, NULL, NULL);
490 	err = fdisp_wait_answ(&fdi);
491 	if (err) {
492 		fdisp_destroy(&fdi);
493 		if (err == ENOTCONN) {
494 			/*
495 	                 * We want to seem a legitimate fs even if the daemon
496 	                 * is stiff dead... (so that, eg., we can still do path
497 	                 * based unmounting after the daemon dies).
498 	                 */
499 			goto fake;
500 		}
501 		return err;
502 	}
503 	fsfo = fdi.answ;
504 
505 	sbp->f_blocks = fsfo->st.blocks;
506 	sbp->f_bfree = fsfo->st.bfree;
507 	sbp->f_bavail = fsfo->st.bavail;
508 	sbp->f_files = fsfo->st.files;
509 	sbp->f_ffree = fsfo->st.ffree;	/* cast from uint64_t to int64_t */
510 	sbp->f_namemax = fsfo->st.namelen;
511 	sbp->f_bsize = fsfo->st.frsize;	/* cast from uint32_t to uint64_t */
512 
513 	FS_DEBUG("fuse_statfs_out -- blocks: %llu, bfree: %llu, bavail: %llu, "
514 	      "fil	es: %llu, ffree: %llu, bsize: %i, namelen: %i\n",
515 	      (unsigned long long)fsfo->st.blocks,
516 	      (unsigned long long)fsfo->st.bfree,
517 	      (unsigned long long)fsfo->st.bavail,
518 	      (unsigned long long)fsfo->st.files,
519 	      (unsigned long long)fsfo->st.ffree, fsfo->st.bsize,
520 	      fsfo->st.namelen);
521 
522 	fdisp_destroy(&fdi);
523 	return 0;
524 
525 fake:
526 	sbp->f_blocks = 0;
527 	sbp->f_bfree = 0;
528 	sbp->f_bavail = 0;
529 	sbp->f_files = 0;
530 	sbp->f_ffree = 0;
531 	sbp->f_namemax = 0;
532 	sbp->f_bsize = FUSE_DEFAULT_BLOCKSIZE;
533 
534 	return 0;
535 }
536