xref: /freebsd/sys/fs/fuse/fuse_vfsops.c (revision 41059135ce931c0f1014a999ffabc6bc470ce856)
1 /*
2  * Copyright (c) 2007-2009 Google Inc. and Amit Singh
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 are
7  * met:
8  *
9  * * Redistributions of source code must retain the above copyright
10  *   notice, this list of conditions and the following disclaimer.
11  * * Redistributions in binary form must reproduce the above
12  *   copyright notice, this list of conditions and the following disclaimer
13  *   in the documentation and/or other materials provided with the
14  *   distribution.
15  * * Neither the name of Google Inc. nor the names of its
16  *   contributors may be used to endorse or promote products derived from
17  *   this software without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
20  * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
21  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
22  * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
23  * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
24  * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
25  * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
26  * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
27  * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
28  * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
29  * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
30  *
31  * Copyright (C) 2005 Csaba Henk.
32  * All rights reserved.
33  *
34  * Redistribution and use in source and binary forms, with or without
35  * modification, are permitted provided that the following conditions
36  * are met:
37  * 1. Redistributions of source code must retain the above copyright
38  *    notice, this list of conditions and the following disclaimer.
39  * 2. Redistributions in binary form must reproduce the above copyright
40  *    notice, this list of conditions and the following disclaimer in the
41  *    documentation and/or other materials provided with the distribution.
42  *
43  * THIS SOFTWARE IS PROVIDED BY AUTHOR AND CONTRIBUTORS ``AS IS'' AND
44  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
45  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
46  * ARE DISCLAIMED.  IN NO EVENT SHALL AUTHOR OR CONTRIBUTORS BE LIABLE
47  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
48  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
49  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
50  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
51  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
52  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
53  * SUCH DAMAGE.
54  */
55 
56 #include <sys/cdefs.h>
57 __FBSDID("$FreeBSD$");
58 
59 #include <sys/types.h>
60 #include <sys/module.h>
61 #include <sys/systm.h>
62 #include <sys/errno.h>
63 #include <sys/param.h>
64 #include <sys/kernel.h>
65 #include <sys/capsicum.h>
66 #include <sys/conf.h>
67 #include <sys/filedesc.h>
68 #include <sys/uio.h>
69 #include <sys/malloc.h>
70 #include <sys/queue.h>
71 #include <sys/lock.h>
72 #include <sys/sx.h>
73 #include <sys/mutex.h>
74 #include <sys/proc.h>
75 #include <sys/vnode.h>
76 #include <sys/namei.h>
77 #include <sys/mount.h>
78 #include <sys/sysctl.h>
79 #include <sys/fcntl.h>
80 
81 #include "fuse.h"
82 #include "fuse_param.h"
83 #include "fuse_node.h"
84 #include "fuse_ipc.h"
85 #include "fuse_internal.h"
86 
87 #include <sys/priv.h>
88 #include <security/mac/mac_framework.h>
89 
90 #define FUSE_DEBUG_MODULE VFSOPS
91 #include "fuse_debug.h"
92 
93 /* This will do for privilege types for now */
94 #ifndef PRIV_VFS_FUSE_ALLOWOTHER
95 #define PRIV_VFS_FUSE_ALLOWOTHER PRIV_VFS_MOUNT_NONUSER
96 #endif
97 #ifndef PRIV_VFS_FUSE_MOUNT_NONUSER
98 #define PRIV_VFS_FUSE_MOUNT_NONUSER PRIV_VFS_MOUNT_NONUSER
99 #endif
100 #ifndef PRIV_VFS_FUSE_SYNC_UNMOUNT
101 #define PRIV_VFS_FUSE_SYNC_UNMOUNT PRIV_VFS_MOUNT_NONUSER
102 #endif
103 
104 static vfs_mount_t fuse_vfsop_mount;
105 static vfs_unmount_t fuse_vfsop_unmount;
106 static vfs_root_t fuse_vfsop_root;
107 static vfs_statfs_t fuse_vfsop_statfs;
108 
109 struct vfsops fuse_vfsops = {
110 	.vfs_mount = fuse_vfsop_mount,
111 	.vfs_unmount = fuse_vfsop_unmount,
112 	.vfs_root = fuse_vfsop_root,
113 	.vfs_statfs = fuse_vfsop_statfs,
114 };
115 
116 SYSCTL_INT(_vfs_fuse, OID_AUTO, init_backgrounded, CTLFLAG_RD,
117     SYSCTL_NULL_INT_PTR, 1, "indicate async handshake");
118 static int fuse_enforce_dev_perms = 0;
119 
120 SYSCTL_INT(_vfs_fuse, OID_AUTO, enforce_dev_perms, CTLFLAG_RW,
121     &fuse_enforce_dev_perms, 0,
122     "enforce fuse device permissions for secondary mounts");
123 static unsigned sync_unmount = 1;
124 
125 SYSCTL_UINT(_vfs_fuse, OID_AUTO, sync_unmount, CTLFLAG_RW,
126     &sync_unmount, 0, "specify when to use synchronous unmount");
127 
128 MALLOC_DEFINE(M_FUSEVFS, "fuse_filesystem", "buffer for fuse vfs layer");
129 
130 static int
131 fuse_getdevice(const char *fspec, struct thread *td, struct cdev **fdevp)
132 {
133 	struct nameidata nd, *ndp = &nd;
134 	struct vnode *devvp;
135 	struct cdev *fdev;
136 	int err;
137 
138 	/*
139          * Not an update, or updating the name: look up the name
140          * and verify that it refers to a sensible disk device.
141          */
142 
143 	NDINIT(ndp, LOOKUP, FOLLOW, UIO_SYSSPACE, fspec, td);
144 	if ((err = namei(ndp)) != 0)
145 		return err;
146 	NDFREE(ndp, NDF_ONLY_PNBUF);
147 	devvp = ndp->ni_vp;
148 
149 	if (devvp->v_type != VCHR) {
150 		vrele(devvp);
151 		return ENXIO;
152 	}
153 	fdev = devvp->v_rdev;
154 	dev_ref(fdev);
155 
156 	if (fuse_enforce_dev_perms) {
157 		/*
158 	         * Check if mounter can open the fuse device.
159 	         *
160 	         * This has significance only if we are doing a secondary mount
161 	         * which doesn't involve actually opening fuse devices, but we
162 	         * still want to enforce the permissions of the device (in
163 	         * order to keep control over the circle of fuse users).
164 	         *
165 	         * (In case of primary mounts, we are either the superuser so
166 	         * we can do anything anyway, or we can mount only if the
167 	         * device is already opened by us, ie. we are permitted to open
168 	         * the device.)
169 	         */
170 #if 0
171 #ifdef MAC
172 		err = mac_check_vnode_open(td->td_ucred, devvp, VREAD | VWRITE);
173 		if (!err)
174 #endif
175 #endif /* 0 */
176 			err = VOP_ACCESS(devvp, VREAD | VWRITE, td->td_ucred, td);
177 		if (err) {
178 			vrele(devvp);
179 			dev_rel(fdev);
180 			return err;
181 		}
182 	}
183 	/*
184          * according to coda code, no extra lock is needed --
185          * although in sys/vnode.h this field is marked "v"
186          */
187 	vrele(devvp);
188 
189 	if (!fdev->si_devsw ||
190 	    strcmp("fuse", fdev->si_devsw->d_name)) {
191 		dev_rel(fdev);
192 		return ENXIO;
193 	}
194 	*fdevp = fdev;
195 
196 	return 0;
197 }
198 
199 #define FUSE_FLAGOPT(fnam, fval) do {				\
200     vfs_flagopt(opts, #fnam, &mntopts, fval);		\
201     vfs_flagopt(opts, "__" #fnam, &__mntopts, fval);	\
202 } while (0)
203 
204 static int
205 fuse_vfsop_mount(struct mount *mp)
206 {
207 	int err;
208 
209 	uint64_t mntopts, __mntopts;
210 	int max_read_set;
211 	uint32_t max_read;
212 	int daemon_timeout;
213 	int fd;
214 
215 	size_t len;
216 
217 	struct cdev *fdev;
218 	struct fuse_data *data;
219 	struct thread *td;
220 	struct file *fp, *fptmp;
221 	char *fspec, *subtype;
222 	struct vfsoptlist *opts;
223 	cap_rights_t rights;
224 
225 	subtype = NULL;
226 	max_read_set = 0;
227 	max_read = ~0;
228 	err = 0;
229 	mntopts = 0;
230 	__mntopts = 0;
231 	td = curthread;
232 
233 	fuse_trace_printf_vfsop();
234 
235 	if (mp->mnt_flag & MNT_UPDATE)
236 		return EOPNOTSUPP;
237 
238 	MNT_ILOCK(mp);
239 	mp->mnt_flag |= MNT_SYNCHRONOUS;
240 	mp->mnt_data = NULL;
241 	MNT_IUNLOCK(mp);
242 	/* Get the new options passed to mount */
243 	opts = mp->mnt_optnew;
244 
245 	if (!opts)
246 		return EINVAL;
247 
248 	/* `fspath' contains the mount point (eg. /mnt/fuse/sshfs); REQUIRED */
249 	if (!vfs_getopts(opts, "fspath", &err))
250 		return err;
251 
252 	/* `from' contains the device name (eg. /dev/fuse0); REQUIRED */
253 	fspec = vfs_getopts(opts, "from", &err);
254 	if (!fspec)
255 		return err;
256 
257 	/* `fd' contains the filedescriptor for this session; REQUIRED */
258 	if (vfs_scanopt(opts, "fd", "%d", &fd) != 1)
259 		return EINVAL;
260 
261 	err = fuse_getdevice(fspec, td, &fdev);
262 	if (err != 0)
263 		return err;
264 
265 	/*
266          * With the help of underscored options the mount program
267          * can inform us from the flags it sets by default
268          */
269 	FUSE_FLAGOPT(allow_other, FSESS_DAEMON_CAN_SPY);
270 	FUSE_FLAGOPT(push_symlinks_in, FSESS_PUSH_SYMLINKS_IN);
271 	FUSE_FLAGOPT(default_permissions, FSESS_DEFAULT_PERMISSIONS);
272 	FUSE_FLAGOPT(no_attrcache, FSESS_NO_ATTRCACHE);
273 	FUSE_FLAGOPT(no_readahed, FSESS_NO_READAHEAD);
274 	FUSE_FLAGOPT(no_datacache, FSESS_NO_DATACACHE);
275 	FUSE_FLAGOPT(no_namecache, FSESS_NO_NAMECACHE);
276 	FUSE_FLAGOPT(no_mmap, FSESS_NO_MMAP);
277 	FUSE_FLAGOPT(brokenio, FSESS_BROKENIO);
278 
279 	if (vfs_scanopt(opts, "max_read=", "%u", &max_read) == 1)
280 		max_read_set = 1;
281 	if (vfs_scanopt(opts, "timeout=", "%u", &daemon_timeout) == 1) {
282 		if (daemon_timeout < FUSE_MIN_DAEMON_TIMEOUT)
283 			daemon_timeout = FUSE_MIN_DAEMON_TIMEOUT;
284 		else if (daemon_timeout > FUSE_MAX_DAEMON_TIMEOUT)
285 			daemon_timeout = FUSE_MAX_DAEMON_TIMEOUT;
286 	} else {
287 		daemon_timeout = FUSE_DEFAULT_DAEMON_TIMEOUT;
288 	}
289 	subtype = vfs_getopts(opts, "subtype=", &err);
290 
291 	FS_DEBUG2G("mntopts 0x%jx\n", (uintmax_t)mntopts);
292 
293 	err = fget(td, fd, cap_rights_init(&rights, CAP_READ), &fp);
294 	if (err != 0) {
295 		FS_DEBUG("invalid or not opened device: data=%p\n", data);
296 		goto out;
297 	}
298 	fptmp = td->td_fpop;
299 	td->td_fpop = fp;
300         err = devfs_get_cdevpriv((void **)&data);
301 	td->td_fpop = fptmp;
302 	fdrop(fp, td);
303 	FUSE_LOCK();
304 	if (err != 0 || data == NULL || data->mp != NULL) {
305 		FS_DEBUG("invalid or not opened device: data=%p data.mp=%p\n",
306 		    data, data != NULL ? data->mp : NULL);
307 		err = ENXIO;
308 		FUSE_UNLOCK();
309 		goto out;
310 	}
311 	if (fdata_get_dead(data)) {
312 		FS_DEBUG("device is dead during mount: data=%p\n", data);
313 		err = ENOTCONN;
314 		FUSE_UNLOCK();
315 		goto out;
316 	}
317 	/* Sanity + permission checks */
318 	if (!data->daemoncred)
319 		panic("fuse daemon found, but identity unknown");
320 	if (mntopts & FSESS_DAEMON_CAN_SPY)
321 		err = priv_check(td, PRIV_VFS_FUSE_ALLOWOTHER);
322 	if (err == 0 && td->td_ucred->cr_uid != data->daemoncred->cr_uid)
323 		/* are we allowed to do the first mount? */
324 		err = priv_check(td, PRIV_VFS_FUSE_MOUNT_NONUSER);
325 	if (err) {
326 		FUSE_UNLOCK();
327 		goto out;
328 	}
329 	data->ref++;
330 	data->mp = mp;
331 	data->dataflags |= mntopts;
332 	data->max_read = max_read;
333 	data->daemon_timeout = daemon_timeout;
334 	FUSE_UNLOCK();
335 
336 	vfs_getnewfsid(mp);
337 	MNT_ILOCK(mp);
338 	mp->mnt_data = data;
339 	mp->mnt_flag |= MNT_LOCAL;
340 	mp->mnt_kern_flag |= MNTK_USES_BCACHE;
341 	MNT_IUNLOCK(mp);
342 	/* We need this here as this slot is used by getnewvnode() */
343 	mp->mnt_stat.f_iosize = PAGE_SIZE;
344 	if (subtype) {
345 		strlcat(mp->mnt_stat.f_fstypename, ".", MFSNAMELEN);
346 		strlcat(mp->mnt_stat.f_fstypename, subtype, MFSNAMELEN);
347 	}
348 	copystr(fspec, mp->mnt_stat.f_mntfromname, MNAMELEN - 1, &len);
349 	bzero(mp->mnt_stat.f_mntfromname + len, MNAMELEN - len);
350 	FS_DEBUG2G("mp %p: %s\n", mp, mp->mnt_stat.f_mntfromname);
351 
352 	/* Now handshaking with daemon */
353 	fuse_internal_send_init(data, td);
354 
355 out:
356 	if (err) {
357 		FUSE_LOCK();
358 		if (data->mp == mp) {
359 			/*
360 			 * Destroy device only if we acquired reference to
361 			 * it
362 			 */
363 			FS_DEBUG("mount failed, destroy device: data=%p mp=%p"
364 			      " err=%d\n",
365 			    data, mp, err);
366 			data->mp = NULL;
367 			fdata_trydestroy(data);
368 		}
369 		FUSE_UNLOCK();
370 		dev_rel(fdev);
371 	}
372 	return err;
373 }
374 
375 static int
376 fuse_vfsop_unmount(struct mount *mp, int mntflags)
377 {
378 	int err = 0;
379 	int flags = 0;
380 
381 	struct cdev *fdev;
382 	struct fuse_data *data;
383 	struct fuse_dispatcher fdi;
384 	struct thread *td = curthread;
385 
386 	fuse_trace_printf_vfsop();
387 
388 	if (mntflags & MNT_FORCE) {
389 		flags |= FORCECLOSE;
390 	}
391 	data = fuse_get_mpdata(mp);
392 	if (!data) {
393 		panic("no private data for mount point?");
394 	}
395 	/* There is 1 extra root vnode reference (mp->mnt_data). */
396 	FUSE_LOCK();
397 	if (data->vroot != NULL) {
398 		struct vnode *vroot = data->vroot;
399 
400 		data->vroot = NULL;
401 		FUSE_UNLOCK();
402 		vrele(vroot);
403 	} else
404 		FUSE_UNLOCK();
405 	err = vflush(mp, 0, flags, td);
406 	if (err) {
407 		debug_printf("vflush failed");
408 		return err;
409 	}
410 	if (fdata_get_dead(data)) {
411 		goto alreadydead;
412 	}
413 	fdisp_init(&fdi, 0);
414 	fdisp_make(&fdi, FUSE_DESTROY, mp, 0, td, NULL);
415 
416 	err = fdisp_wait_answ(&fdi);
417 	fdisp_destroy(&fdi);
418 
419 	fdata_set_dead(data);
420 
421 alreadydead:
422 	FUSE_LOCK();
423 	data->mp = NULL;
424 	fdev = data->fdev;
425 	fdata_trydestroy(data);
426 	FUSE_UNLOCK();
427 
428 	MNT_ILOCK(mp);
429 	mp->mnt_data = NULL;
430 	mp->mnt_flag &= ~MNT_LOCAL;
431 	MNT_IUNLOCK(mp);
432 
433 	dev_rel(fdev);
434 
435 	return 0;
436 }
437 
438 static int
439 fuse_vfsop_root(struct mount *mp, int lkflags, struct vnode **vpp)
440 {
441 	struct fuse_data *data = fuse_get_mpdata(mp);
442 	int err = 0;
443 
444 	if (data->vroot != NULL) {
445 		err = vget(data->vroot, lkflags, curthread);
446 		if (err == 0)
447 			*vpp = data->vroot;
448 	} else {
449 		err = fuse_vnode_get(mp, FUSE_ROOT_ID, NULL, vpp, NULL, VDIR);
450 		if (err == 0) {
451 			FUSE_LOCK();
452 			MPASS(data->vroot == NULL || data->vroot == *vpp);
453 			if (data->vroot == NULL) {
454 				FS_DEBUG("new root vnode\n");
455 				data->vroot = *vpp;
456 				FUSE_UNLOCK();
457 				vref(*vpp);
458 			} else if (data->vroot != *vpp) {
459 				FS_DEBUG("root vnode race\n");
460 				FUSE_UNLOCK();
461 				VOP_UNLOCK(*vpp, 0);
462 				vrele(*vpp);
463 				vrecycle(*vpp);
464 				*vpp = data->vroot;
465 			} else
466 				FUSE_UNLOCK();
467 		}
468 	}
469 	return err;
470 }
471 
472 static int
473 fuse_vfsop_statfs(struct mount *mp, struct statfs *sbp)
474 {
475 	struct fuse_dispatcher fdi;
476 	int err = 0;
477 
478 	struct fuse_statfs_out *fsfo;
479 	struct fuse_data *data;
480 
481 	FS_DEBUG2G("mp %p: %s\n", mp, mp->mnt_stat.f_mntfromname);
482 	data = fuse_get_mpdata(mp);
483 
484 	if (!(data->dataflags & FSESS_INITED))
485 		goto fake;
486 
487 	fdisp_init(&fdi, 0);
488 	fdisp_make(&fdi, FUSE_STATFS, mp, FUSE_ROOT_ID, NULL, NULL);
489 	err = fdisp_wait_answ(&fdi);
490 	if (err) {
491 		fdisp_destroy(&fdi);
492 		if (err == ENOTCONN) {
493 			/*
494 	                 * We want to seem a legitimate fs even if the daemon
495 	                 * is stiff dead... (so that, eg., we can still do path
496 	                 * based unmounting after the daemon dies).
497 	                 */
498 			goto fake;
499 		}
500 		return err;
501 	}
502 	fsfo = fdi.answ;
503 
504 	sbp->f_blocks = fsfo->st.blocks;
505 	sbp->f_bfree = fsfo->st.bfree;
506 	sbp->f_bavail = fsfo->st.bavail;
507 	sbp->f_files = fsfo->st.files;
508 	sbp->f_ffree = fsfo->st.ffree;	/* cast from uint64_t to int64_t */
509 	sbp->f_namemax = fsfo->st.namelen;
510 	sbp->f_bsize = fsfo->st.frsize;	/* cast from uint32_t to uint64_t */
511 
512 	FS_DEBUG("fuse_statfs_out -- blocks: %llu, bfree: %llu, bavail: %llu, "
513 	      "fil	es: %llu, ffree: %llu, bsize: %i, namelen: %i\n",
514 	      (unsigned long long)fsfo->st.blocks,
515 	      (unsigned long long)fsfo->st.bfree,
516 	      (unsigned long long)fsfo->st.bavail,
517 	      (unsigned long long)fsfo->st.files,
518 	      (unsigned long long)fsfo->st.ffree, fsfo->st.bsize,
519 	      fsfo->st.namelen);
520 
521 	fdisp_destroy(&fdi);
522 	return 0;
523 
524 fake:
525 	sbp->f_blocks = 0;
526 	sbp->f_bfree = 0;
527 	sbp->f_bavail = 0;
528 	sbp->f_files = 0;
529 	sbp->f_ffree = 0;
530 	sbp->f_namemax = 0;
531 	sbp->f_bsize = FUSE_DEFAULT_BLOCKSIZE;
532 
533 	return 0;
534 }
535