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/conf.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 #include <sys/fnv_hash.h> 81 #include <sys/priv.h> 82 #include <security/mac/mac_framework.h> 83 #include <vm/vm.h> 84 #include <vm/vm_extern.h> 85 86 #include "fuse.h" 87 #include "fuse_node.h" 88 #include "fuse_internal.h" 89 #include "fuse_io.h" 90 #include "fuse_ipc.h" 91 92 #define FUSE_DEBUG_MODULE VNOPS 93 #include "fuse_debug.h" 94 95 MALLOC_DEFINE(M_FUSEVN, "fuse_vnode", "fuse vnode private data"); 96 97 static int fuse_node_count = 0; 98 99 SYSCTL_INT(_vfs_fuse, OID_AUTO, node_count, CTLFLAG_RD, 100 &fuse_node_count, 0, "Count of FUSE vnodes"); 101 102 int fuse_data_cache_enable = 1; 103 104 SYSCTL_INT(_vfs_fuse, OID_AUTO, data_cache_enable, CTLFLAG_RW, 105 &fuse_data_cache_enable, 0, 106 "enable caching of FUSE file data (including dirty data)"); 107 108 int fuse_data_cache_invalidate = 0; 109 110 SYSCTL_INT(_vfs_fuse, OID_AUTO, data_cache_invalidate, CTLFLAG_RW, 111 &fuse_data_cache_invalidate, 0, 112 "If non-zero, discard cached clean file data when there are no active file" 113 " users"); 114 115 int fuse_mmap_enable = 1; 116 117 SYSCTL_INT(_vfs_fuse, OID_AUTO, mmap_enable, CTLFLAG_RW, 118 &fuse_mmap_enable, 0, 119 "If non-zero, and data_cache_enable is also non-zero, enable mmap(2) of " 120 "FUSE files"); 121 122 int fuse_refresh_size = 0; 123 124 SYSCTL_INT(_vfs_fuse, OID_AUTO, refresh_size, CTLFLAG_RW, 125 &fuse_refresh_size, 0, 126 "If non-zero, and no dirty file extension data is buffered, fetch file " 127 "size before write operations"); 128 129 int fuse_sync_resize = 1; 130 131 SYSCTL_INT(_vfs_fuse, OID_AUTO, sync_resize, CTLFLAG_RW, 132 &fuse_sync_resize, 0, 133 "If a cached write extended a file, inform FUSE filesystem of the changed" 134 "size immediately subsequent to the issued writes"); 135 136 int fuse_fix_broken_io = 0; 137 138 SYSCTL_INT(_vfs_fuse, OID_AUTO, fix_broken_io, CTLFLAG_RW, 139 &fuse_fix_broken_io, 0, 140 "If non-zero, print a diagnostic warning if a userspace filesystem returns" 141 " EIO on reads of recently extended portions of files"); 142 143 static void 144 fuse_vnode_init(struct vnode *vp, struct fuse_vnode_data *fvdat, 145 uint64_t nodeid, enum vtype vtyp) 146 { 147 int i; 148 149 fvdat->nid = nodeid; 150 if (nodeid == FUSE_ROOT_ID) { 151 vp->v_vflag |= VV_ROOT; 152 } 153 vp->v_type = vtyp; 154 vp->v_data = fvdat; 155 156 for (i = 0; i < FUFH_MAXTYPE; i++) 157 fvdat->fufh[i].fh_type = FUFH_INVALID; 158 159 atomic_add_acq_int(&fuse_node_count, 1); 160 } 161 162 void 163 fuse_vnode_destroy(struct vnode *vp) 164 { 165 struct fuse_vnode_data *fvdat = vp->v_data; 166 167 vp->v_data = NULL; 168 free(fvdat, M_FUSEVN); 169 170 atomic_subtract_acq_int(&fuse_node_count, 1); 171 } 172 173 static int 174 fuse_vnode_cmp(struct vnode *vp, void *nidp) 175 { 176 return (VTOI(vp) != *((uint64_t *)nidp)); 177 } 178 179 static uint32_t __inline 180 fuse_vnode_hash(uint64_t id) 181 { 182 return (fnv_32_buf(&id, sizeof(id), FNV1_32_INIT)); 183 } 184 185 static int 186 fuse_vnode_alloc(struct mount *mp, 187 struct thread *td, 188 uint64_t nodeid, 189 enum vtype vtyp, 190 struct vnode **vpp) 191 { 192 struct fuse_vnode_data *fvdat; 193 struct vnode *vp2; 194 int err = 0; 195 196 FS_DEBUG("been asked for vno #%ju\n", (uintmax_t)nodeid); 197 198 if (vtyp == VNON) { 199 return EINVAL; 200 } 201 *vpp = NULL; 202 err = vfs_hash_get(mp, fuse_vnode_hash(nodeid), LK_EXCLUSIVE, td, vpp, 203 fuse_vnode_cmp, &nodeid); 204 if (err) 205 return (err); 206 207 if (*vpp) { 208 MPASS((*vpp)->v_type == vtyp && (*vpp)->v_data != NULL); 209 FS_DEBUG("vnode taken from hash\n"); 210 return (0); 211 } 212 fvdat = malloc(sizeof(*fvdat), M_FUSEVN, M_WAITOK | M_ZERO); 213 err = getnewvnode("fuse", mp, &fuse_vnops, vpp); 214 if (err) { 215 free(fvdat, M_FUSEVN); 216 return (err); 217 } 218 lockmgr((*vpp)->v_vnlock, LK_EXCLUSIVE, NULL); 219 fuse_vnode_init(*vpp, fvdat, nodeid, vtyp); 220 err = insmntque(*vpp, mp); 221 ASSERT_VOP_ELOCKED(*vpp, "fuse_vnode_alloc"); 222 if (err) { 223 free(fvdat, M_FUSEVN); 224 *vpp = NULL; 225 return (err); 226 } 227 err = vfs_hash_insert(*vpp, fuse_vnode_hash(nodeid), LK_EXCLUSIVE, 228 td, &vp2, fuse_vnode_cmp, &nodeid); 229 if (err) 230 return (err); 231 if (vp2 != NULL) { 232 *vpp = vp2; 233 return (0); 234 } 235 236 ASSERT_VOP_ELOCKED(*vpp, "fuse_vnode_alloc"); 237 238 return (0); 239 } 240 241 int 242 fuse_vnode_get(struct mount *mp, 243 uint64_t nodeid, 244 struct vnode *dvp, 245 struct vnode **vpp, 246 struct componentname *cnp, 247 enum vtype vtyp) 248 { 249 struct thread *td = (cnp != NULL ? cnp->cn_thread : curthread); 250 int err = 0; 251 252 debug_printf("dvp=%p\n", dvp); 253 254 err = fuse_vnode_alloc(mp, td, nodeid, vtyp, vpp); 255 if (err) { 256 return err; 257 } 258 if (dvp != NULL) { 259 MPASS((cnp->cn_flags & ISDOTDOT) == 0); 260 MPASS(!(cnp->cn_namelen == 1 && cnp->cn_nameptr[0] == '.')); 261 fuse_vnode_setparent(*vpp, dvp); 262 } 263 if (dvp != NULL && cnp != NULL && (cnp->cn_flags & MAKEENTRY) != 0) { 264 ASSERT_VOP_LOCKED(*vpp, "fuse_vnode_get"); 265 ASSERT_VOP_LOCKED(dvp, "fuse_vnode_get"); 266 cache_enter(dvp, *vpp, cnp); 267 } 268 269 /* 270 * In userland, libfuse uses cached lookups for dot and dotdot entries, 271 * thus it does not really bump the nlookup counter for forget. 272 * Follow the same semantic and avoid tu bump it in order to keep 273 * nlookup counters consistent. 274 */ 275 if (cnp == NULL || ((cnp->cn_flags & ISDOTDOT) == 0 && 276 (cnp->cn_namelen != 1 || cnp->cn_nameptr[0] != '.'))) 277 VTOFUD(*vpp)->nlookup++; 278 279 return 0; 280 } 281 282 void 283 fuse_vnode_open(struct vnode *vp, int32_t fuse_open_flags, struct thread *td) 284 { 285 /* 286 * Funcation is called for every vnode open. 287 * Merge fuse_open_flags it may be 0 288 */ 289 /* 290 * Ideally speaking, direct io should be enabled on 291 * fd's but do not see of any way of providing that 292 * this implementation. 293 * 294 * Also cannot think of a reason why would two 295 * different fd's on same vnode would like 296 * have DIRECT_IO turned on and off. But linux 297 * based implementation works on an fd not an 298 * inode and provides such a feature. 299 * 300 * XXXIP: Handle fd based DIRECT_IO 301 */ 302 if (fuse_open_flags & FOPEN_DIRECT_IO) { 303 ASSERT_VOP_ELOCKED(vp, __func__); 304 VTOFUD(vp)->flag |= FN_DIRECTIO; 305 fuse_io_invalbuf(vp, td); 306 } else { 307 if ((fuse_open_flags & FOPEN_KEEP_CACHE) == 0) 308 fuse_io_invalbuf(vp, td); 309 VTOFUD(vp)->flag &= ~FN_DIRECTIO; 310 } 311 312 if (vnode_vtype(vp) == VREG) { 313 /* XXXIP prevent getattr, by using cached node size */ 314 vnode_create_vobject(vp, 0, td); 315 } 316 } 317 318 int 319 fuse_vnode_savesize(struct vnode *vp, struct ucred *cred) 320 { 321 struct fuse_vnode_data *fvdat = VTOFUD(vp); 322 struct thread *td = curthread; 323 struct fuse_filehandle *fufh = NULL; 324 struct fuse_dispatcher fdi; 325 struct fuse_setattr_in *fsai; 326 int err = 0; 327 328 FS_DEBUG("inode=%ju size=%ju\n", (uintmax_t)VTOI(vp), 329 (uintmax_t)fvdat->filesize); 330 ASSERT_VOP_ELOCKED(vp, "fuse_io_extend"); 331 332 if (fuse_isdeadfs(vp)) { 333 return EBADF; 334 } 335 if (vnode_vtype(vp) == VDIR) { 336 return EISDIR; 337 } 338 if (vfs_isrdonly(vnode_mount(vp))) { 339 return EROFS; 340 } 341 if (cred == NULL) { 342 cred = td->td_ucred; 343 } 344 fdisp_init(&fdi, sizeof(*fsai)); 345 fdisp_make_vp(&fdi, FUSE_SETATTR, vp, td, cred); 346 fsai = fdi.indata; 347 fsai->valid = 0; 348 349 /* Truncate to a new value. */ 350 fsai->size = fvdat->filesize; 351 fsai->valid |= FATTR_SIZE; 352 353 fuse_filehandle_getrw(vp, FUFH_WRONLY, &fufh); 354 if (fufh) { 355 fsai->fh = fufh->fh_id; 356 fsai->valid |= FATTR_FH; 357 } 358 err = fdisp_wait_answ(&fdi); 359 fdisp_destroy(&fdi); 360 if (err == 0) 361 fvdat->flag &= ~FN_SIZECHANGE; 362 363 return err; 364 } 365 366 void 367 fuse_vnode_refreshsize(struct vnode *vp, struct ucred *cred) 368 { 369 370 struct fuse_vnode_data *fvdat = VTOFUD(vp); 371 struct vattr va; 372 373 if ((fvdat->flag & FN_SIZECHANGE) != 0 || 374 (fuse_refresh_size == 0 && fvdat->filesize != 0)) 375 return; 376 377 VOP_GETATTR(vp, &va, cred); 378 FS_DEBUG("refreshed file size: %jd\n", (intmax_t)VTOFUD(vp)->filesize); 379 } 380 381 int 382 fuse_vnode_setsize(struct vnode *vp, struct ucred *cred, off_t newsize) 383 { 384 struct fuse_vnode_data *fvdat = VTOFUD(vp); 385 off_t oldsize; 386 int err = 0; 387 388 FS_DEBUG("inode=%ju oldsize=%ju newsize=%ju\n", 389 (uintmax_t)VTOI(vp), (uintmax_t)fvdat->filesize, 390 (uintmax_t)newsize); 391 ASSERT_VOP_ELOCKED(vp, "fuse_vnode_setsize"); 392 393 oldsize = fvdat->filesize; 394 fvdat->filesize = newsize; 395 fvdat->flag |= FN_SIZECHANGE; 396 397 if (newsize < oldsize) { 398 err = vtruncbuf(vp, cred, newsize, fuse_iosize(vp)); 399 } 400 vnode_pager_setsize(vp, newsize); 401 return err; 402 } 403