2 * SPDX-License-Identifier: BSD-3-Clause
4 * Copyright (c) 2007-2009 Google Inc. and Amit Singh
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions are
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
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.
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.
33 * Copyright (C) 2005 Csaba Henk.
34 * All rights reserved.
36 * Copyright (c) 2019 The FreeBSD Foundation
38 * Portions of this software were developed by BFF Storage Systems, LLC under
39 * sponsorship from the FreeBSD Foundation.
41 * Redistribution and use in source and binary forms, with or without
42 * modification, are permitted provided that the following conditions
44 * 1. Redistributions of source code must retain the above copyright
45 * notice, this list of conditions and the following disclaimer.
46 * 2. Redistributions in binary form must reproduce the above copyright
47 * notice, this list of conditions and the following disclaimer in the
48 * documentation and/or other materials provided with the distribution.
50 * THIS SOFTWARE IS PROVIDED BY AUTHOR AND CONTRIBUTORS ``AS IS'' AND
51 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
52 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
53 * ARE DISCLAIMED. IN NO EVENT SHALL AUTHOR OR CONTRIBUTORS BE LIABLE
54 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
55 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
56 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
57 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
58 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
59 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
63 #include <sys/cdefs.h>
64 __FBSDID("$FreeBSD$");
66 #include <sys/types.h>
67 #include <sys/systm.h>
68 #include <sys/counter.h>
69 #include <sys/module.h>
70 #include <sys/errno.h>
71 #include <sys/param.h>
72 #include <sys/kernel.h>
75 #include <sys/malloc.h>
76 #include <sys/queue.h>
79 #include <sys/mutex.h>
81 #include <sys/vnode.h>
82 #include <sys/namei.h>
83 #include <sys/mount.h>
84 #include <sys/sysctl.h>
85 #include <sys/fcntl.h>
88 #include <security/mac/mac_framework.h>
90 #include <vm/vm_extern.h>
93 #include "fuse_node.h"
94 #include "fuse_internal.h"
98 SDT_PROVIDER_DECLARE(fusefs);
101 * arg0: verbosity. Higher numbers give more verbose messages
102 * arg1: Textual message
104 SDT_PROBE_DEFINE2(fusefs, , node, trace, "int", "char*");
106 MALLOC_DEFINE(M_FUSEVN, "fuse_vnode", "fuse vnode private data");
108 static int sysctl_fuse_cache_mode(SYSCTL_HANDLER_ARGS);
110 static counter_u64_t fuse_node_count;
112 SYSCTL_COUNTER_U64(_vfs_fusefs_stats, OID_AUTO, node_count, CTLFLAG_RD,
113 &fuse_node_count, "Count of FUSE vnodes");
115 int fuse_data_cache_mode = FUSE_CACHE_WT;
119 * This sysctl is no longer needed as of fuse protocol 7.23. Individual
120 * servers can select the cache behavior they need for each mountpoint:
121 * - writethrough: the default
122 * - writeback: set FUSE_WRITEBACK_CACHE in fuse_init_out.flags
123 * - uncached: set FOPEN_DIRECT_IO for every file
124 * The sysctl is retained primarily for use by jails supporting older FUSE
125 * protocols. It may be removed entirely once FreeBSD 11.3 and 12.0 are EOL.
127 SYSCTL_PROC(_vfs_fusefs, OID_AUTO, data_cache_mode,
128 CTLTYPE_INT | CTLFLAG_MPSAFE | CTLFLAG_RW,
129 &fuse_data_cache_mode, 0, sysctl_fuse_cache_mode, "I",
130 "Zero: disable caching of FUSE file data; One: write-through caching "
131 "(default); Two: write-back caching (generally unsafe)");
134 sysctl_fuse_cache_mode(SYSCTL_HANDLER_ARGS)
139 error = sysctl_handle_int(oidp, &val, 0, req);
140 if (error || !req->newptr)
156 fuse_vnode_init(struct vnode *vp, struct fuse_vnode_data *fvdat,
157 uint64_t nodeid, enum vtype vtyp)
160 LIST_INIT(&fvdat->handles);
161 vattr_null(&fvdat->cached_attrs);
162 if (nodeid == FUSE_ROOT_ID) {
163 vp->v_vflag |= VV_ROOT;
168 counter_u64_add(fuse_node_count, 1);
172 fuse_vnode_destroy(struct vnode *vp)
174 struct fuse_vnode_data *fvdat = vp->v_data;
177 KASSERT(LIST_EMPTY(&fvdat->handles),
178 ("Destroying fuse vnode with open files!"));
179 free(fvdat, M_FUSEVN);
181 counter_u64_add(fuse_node_count, -1);
185 fuse_vnode_cmp(struct vnode *vp, void *nidp)
187 return (VTOI(vp) != *((uint64_t *)nidp));
190 SDT_PROBE_DEFINE3(fusefs, , node, stale_vnode, "struct vnode*", "enum vtype",
193 fuse_vnode_alloc(struct mount *mp,
199 struct fuse_data *data;
200 struct fuse_vnode_data *fvdat;
204 data = fuse_get_mpdata(mp);
209 err = vfs_hash_get(mp, fuse_vnode_hash(nodeid), LK_EXCLUSIVE, td, vpp,
210 fuse_vnode_cmp, &nodeid);
215 if ((*vpp)->v_type != vtyp) {
217 * STALE vnode! This probably indicates a buggy
218 * server, but it could also be the result of a race
219 * between FUSE_LOOKUP and another client's
220 * FUSE_UNLINK/FUSE_CREATE
222 SDT_PROBE3(fusefs, , node, stale_vnode, *vpp, vtyp,
224 fuse_internal_vnode_disappear(*vpp);
225 lockmgr((*vpp)->v_vnlock, LK_RELEASE, NULL);
229 MPASS((*vpp)->v_data != NULL);
230 MPASS(VTOFUD(*vpp)->nid == nodeid);
231 SDT_PROBE2(fusefs, , node, trace, 1, "vnode taken from hash");
234 fvdat = malloc(sizeof(*fvdat), M_FUSEVN, M_WAITOK | M_ZERO);
237 err = getnewvnode("fuse", mp, &fuse_fifoops, vpp);
240 err = getnewvnode("fuse", mp, &fuse_vnops, vpp);
244 free(fvdat, M_FUSEVN);
247 lockmgr((*vpp)->v_vnlock, LK_EXCLUSIVE, NULL);
248 fuse_vnode_init(*vpp, fvdat, nodeid, vtyp);
249 err = insmntque(*vpp, mp);
250 ASSERT_VOP_ELOCKED(*vpp, "fuse_vnode_alloc");
252 lockmgr((*vpp)->v_vnlock, LK_RELEASE, NULL);
253 free(fvdat, M_FUSEVN);
257 /* Disallow async reads for fifos because UFS does. I don't know why */
258 if (data->dataflags & FSESS_ASYNC_READ && vtyp != VFIFO)
259 VN_LOCK_ASHARE(*vpp);
261 err = vfs_hash_insert(*vpp, fuse_vnode_hash(nodeid), LK_EXCLUSIVE,
262 td, &vp2, fuse_vnode_cmp, &nodeid);
264 lockmgr((*vpp)->v_vnlock, LK_RELEASE, NULL);
265 free(fvdat, M_FUSEVN);
274 ASSERT_VOP_ELOCKED(*vpp, "fuse_vnode_alloc");
280 fuse_vnode_get(struct mount *mp,
281 struct fuse_entry_out *feo,
285 struct componentname *cnp,
288 struct thread *td = (cnp != NULL ? cnp->cn_thread : curthread);
290 * feo should only be NULL for the root directory, which (when libfuse
291 * is used) always has generation 0
293 uint64_t generation = feo ? feo->generation : 0;
296 err = fuse_vnode_alloc(mp, td, nodeid, vtyp, vpp);
301 MPASS(cnp && (cnp->cn_flags & ISDOTDOT) == 0);
303 !(cnp->cn_namelen == 1 && cnp->cn_nameptr[0] == '.'));
304 fuse_vnode_setparent(*vpp, dvp);
306 if (dvp != NULL && cnp != NULL && (cnp->cn_flags & MAKEENTRY) != 0 &&
308 (feo->entry_valid != 0 || feo->entry_valid_nsec != 0)) {
309 struct timespec timeout;
311 ASSERT_VOP_LOCKED(*vpp, "fuse_vnode_get");
312 ASSERT_VOP_LOCKED(dvp, "fuse_vnode_get");
314 fuse_validity_2_timespec(feo, &timeout);
315 cache_enter_time(dvp, *vpp, cnp, &timeout, NULL);
318 VTOFUD(*vpp)->generation = generation;
320 * In userland, libfuse uses cached lookups for dot and dotdot entries,
321 * thus it does not really bump the nlookup counter for forget.
322 * Follow the same semantic and avoid the bump in order to keep
323 * nlookup counters consistent.
325 if (cnp == NULL || ((cnp->cn_flags & ISDOTDOT) == 0 &&
326 (cnp->cn_namelen != 1 || cnp->cn_nameptr[0] != '.')))
327 VTOFUD(*vpp)->nlookup++;
333 * Called for every fusefs vnode open to initialize the vnode (not
334 * fuse_filehandle) for use
337 fuse_vnode_open(struct vnode *vp, int32_t fuse_open_flags, struct thread *td)
339 if (vnode_vtype(vp) == VREG)
340 vnode_create_vobject(vp, 0, td);
344 fuse_vnode_savesize(struct vnode *vp, struct ucred *cred, pid_t pid)
346 struct fuse_vnode_data *fvdat = VTOFUD(vp);
347 struct thread *td = curthread;
348 struct fuse_filehandle *fufh = NULL;
349 struct fuse_dispatcher fdi;
350 struct fuse_setattr_in *fsai;
353 ASSERT_VOP_ELOCKED(vp, "fuse_io_extend");
355 if (fuse_isdeadfs(vp)) {
358 if (vnode_vtype(vp) == VDIR) {
361 if (vfs_isrdonly(vnode_mount(vp))) {
367 fdisp_init(&fdi, sizeof(*fsai));
368 fdisp_make_vp(&fdi, FUSE_SETATTR, vp, td, cred);
372 /* Truncate to a new value. */
373 MPASS((fvdat->flag & FN_SIZECHANGE) != 0);
374 fsai->size = fvdat->cached_attrs.va_size;
375 fsai->valid |= FATTR_SIZE;
377 fuse_filehandle_getrw(vp, FWRITE, &fufh, cred, pid);
379 fsai->fh = fufh->fh_id;
380 fsai->valid |= FATTR_FH;
382 err = fdisp_wait_answ(&fdi);
385 fvdat->flag &= ~FN_SIZECHANGE;
391 * Adjust the vnode's size to a new value, such as that provided by
395 fuse_vnode_setsize(struct vnode *vp, off_t newsize)
397 struct fuse_vnode_data *fvdat = VTOFUD(vp);
401 struct buf *bp = NULL;
404 ASSERT_VOP_ELOCKED(vp, "fuse_vnode_setsize");
406 iosize = fuse_iosize(vp);
407 oldsize = fvdat->cached_attrs.va_size;
408 fvdat->cached_attrs.va_size = newsize;
409 if ((attrs = VTOVA(vp)) != NULL)
410 attrs->va_size = newsize;
412 if (newsize < oldsize) {
415 err = vtruncbuf(vp, newsize, fuse_iosize(vp));
418 if (newsize % iosize == 0)
421 * Zero the contents of the last partial block.
422 * Sure seems like vtruncbuf should do this for us.
425 lbn = newsize / iosize;
426 bp = getblk(vp, lbn, iosize, PCATCH, 0, 0);
431 if (!(bp->b_flags & B_CACHE))
432 goto out; /* Nothing to do */
433 MPASS(bp->b_flags & B_VMIO);
435 bp->b_dirtyend = MIN(bp->b_dirtyend, newsize - lbn * iosize);
440 vnode_pager_setsize(vp, newsize);
444 /* Get the current, possibly dirty, size of the file */
446 fuse_vnode_size(struct vnode *vp, off_t *filesize, struct ucred *cred,
449 struct fuse_vnode_data *fvdat = VTOFUD(vp);
452 if (!(fvdat->flag & FN_SIZECHANGE) &&
453 (!fuse_vnode_attr_cache_valid(vp) ||
454 fvdat->cached_attrs.va_size == VNOVAL))
455 error = fuse_internal_do_getattr(vp, NULL, cred, td);
458 *filesize = fvdat->cached_attrs.va_size;
464 fuse_vnode_undirty_cached_timestamps(struct vnode *vp)
466 struct fuse_vnode_data *fvdat = VTOFUD(vp);
468 fvdat->flag &= ~(FN_MTIMECHANGE | FN_CTIMECHANGE);
471 /* Update a fuse file's cached timestamps */
473 fuse_vnode_update(struct vnode *vp, int flags)
475 struct fuse_vnode_data *fvdat = VTOFUD(vp);
476 struct fuse_data *data = fuse_get_mpdata(vnode_mount(vp));
481 if (data->time_gran > 1)
482 ts.tv_nsec = rounddown(ts.tv_nsec, data->time_gran);
484 if (flags & FN_MTIMECHANGE)
485 fvdat->cached_attrs.va_mtime = ts;
486 if (flags & FN_CTIMECHANGE)
487 fvdat->cached_attrs.va_ctime = ts;
489 fvdat->flag |= flags;
495 fuse_node_count = counter_u64_alloc(M_WAITOK);
499 fuse_node_destroy(void)
501 counter_u64_free(fuse_node_count);