1 /* $NetBSD: tmpfs_vfsops.c,v 1.59 2014/04/16 18:55:19 maxv Exp $ */ 2 3 /* 4 * Copyright (c) 2005, 2006, 2007 The NetBSD Foundation, Inc. 5 * All rights reserved. 6 * 7 * This code is derived from software contributed to The NetBSD Foundation 8 * by Julio M. Merino Vidal, developed as part of Google's Summer of Code 9 * 2005 program. 10 * 11 * Redistribution and use in source and binary forms, with or without 12 * modification, are permitted provided that the following conditions 13 * are met: 14 * 1. Redistributions of source code must retain the above copyright 15 * notice, this list of conditions and the following disclaimer. 16 * 2. Redistributions in binary form must reproduce the above copyright 17 * notice, this list of conditions and the following disclaimer in the 18 * documentation and/or other materials provided with the distribution. 19 * 20 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 21 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 22 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 23 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 24 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 25 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 26 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 27 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 28 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 29 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 30 * POSSIBILITY OF SUCH DAMAGE. 31 */ 32 33 /* 34 * Efficient memory file system. 35 * 36 * tmpfs is a file system that uses NetBSD's virtual memory sub-system 37 * (the well-known UVM) to store file data and metadata in an efficient 38 * way. This means that it does not follow the structure of an on-disk 39 * file system because it simply does not need to. Instead, it uses 40 * memory-specific data structures and algorithms to automatically 41 * allocate and release resources. 42 */ 43 44 #include <sys/cdefs.h> 45 __KERNEL_RCSID(0, "$NetBSD: tmpfs_vfsops.c,v 1.59 2014/04/16 18:55:19 maxv Exp $"); 46 47 #include <sys/param.h> 48 #include <sys/types.h> 49 #include <sys/kmem.h> 50 #include <sys/mount.h> 51 #include <sys/stat.h> 52 #include <sys/systm.h> 53 #include <sys/vnode.h> 54 #include <sys/module.h> 55 56 #include <miscfs/genfs/genfs.h> 57 #include <fs/tmpfs/tmpfs.h> 58 #include <fs/tmpfs/tmpfs_args.h> 59 60 MODULE(MODULE_CLASS_VFS, tmpfs, NULL); 61 62 struct pool tmpfs_dirent_pool; 63 struct pool tmpfs_node_pool; 64 65 static int tmpfs_mount(struct mount *, const char *, void *, size_t *); 66 static int tmpfs_start(struct mount *, int); 67 static int tmpfs_unmount(struct mount *, int); 68 static int tmpfs_root(struct mount *, vnode_t **); 69 static int tmpfs_vget(struct mount *, ino_t, vnode_t **); 70 static int tmpfs_fhtovp(struct mount *, struct fid *, vnode_t **); 71 static int tmpfs_vptofh(struct vnode *, struct fid *, size_t *); 72 static int tmpfs_statvfs(struct mount *, struct statvfs *); 73 static int tmpfs_sync(struct mount *, int, kauth_cred_t); 74 static void tmpfs_init(void); 75 static void tmpfs_done(void); 76 static int tmpfs_snapshot(struct mount *, vnode_t *, struct timespec *); 77 78 static void 79 tmpfs_init(void) 80 { 81 82 pool_init(&tmpfs_dirent_pool, sizeof(tmpfs_dirent_t), 0, 0, 0, 83 "tmpfs_dirent", &pool_allocator_nointr, IPL_NONE); 84 pool_init(&tmpfs_node_pool, sizeof(tmpfs_node_t), 0, 0, 0, 85 "tmpfs_node", &pool_allocator_nointr, IPL_NONE); 86 } 87 88 static void 89 tmpfs_done(void) 90 { 91 92 pool_destroy(&tmpfs_dirent_pool); 93 pool_destroy(&tmpfs_node_pool); 94 } 95 96 static int 97 tmpfs_mount(struct mount *mp, const char *path, void *data, size_t *data_len) 98 { 99 struct tmpfs_args *args = data; 100 tmpfs_mount_t *tmp; 101 tmpfs_node_t *root; 102 uint64_t memlimit; 103 ino_t nodes; 104 int error; 105 106 if (args == NULL) 107 return EINVAL; 108 109 /* Validate the version. */ 110 if (*data_len < sizeof(*args) || 111 args->ta_version != TMPFS_ARGS_VERSION) 112 return EINVAL; 113 114 /* Handle retrieval of mount point arguments. */ 115 if (mp->mnt_flag & MNT_GETARGS) { 116 if (mp->mnt_data == NULL) 117 return EIO; 118 tmp = VFS_TO_TMPFS(mp); 119 120 args->ta_version = TMPFS_ARGS_VERSION; 121 args->ta_nodes_max = tmp->tm_nodes_max; 122 args->ta_size_max = tmp->tm_mem_limit; 123 124 root = tmp->tm_root; 125 args->ta_root_uid = root->tn_uid; 126 args->ta_root_gid = root->tn_gid; 127 args->ta_root_mode = root->tn_mode; 128 129 *data_len = sizeof(*args); 130 return 0; 131 } 132 133 if (mp->mnt_flag & MNT_UPDATE) { 134 /* TODO */ 135 return EOPNOTSUPP; 136 } 137 138 /* Prohibit mounts if there is not enough memory. */ 139 if (tmpfs_mem_info(true) < TMPFS_PAGES_RESERVED) 140 return EINVAL; 141 142 /* Get the memory usage limit for this file-system. */ 143 if (args->ta_size_max < PAGE_SIZE) { 144 memlimit = UINT64_MAX; 145 } else { 146 memlimit = args->ta_size_max; 147 } 148 KASSERT(memlimit > 0); 149 150 if (args->ta_nodes_max <= 3) { 151 nodes = 3 + (memlimit / 1024); 152 } else { 153 nodes = args->ta_nodes_max; 154 } 155 nodes = MIN(nodes, INT_MAX); 156 KASSERT(nodes >= 3); 157 158 /* Allocate the tmpfs mount structure and fill it. */ 159 tmp = kmem_zalloc(sizeof(tmpfs_mount_t), KM_SLEEP); 160 if (tmp == NULL) 161 return ENOMEM; 162 163 tmp->tm_nodes_max = nodes; 164 tmp->tm_nodes_cnt = 0; 165 LIST_INIT(&tmp->tm_nodes); 166 167 mutex_init(&tmp->tm_lock, MUTEX_DEFAULT, IPL_NONE); 168 tmpfs_mntmem_init(tmp, memlimit); 169 170 /* Allocate the root node. */ 171 error = tmpfs_alloc_node(tmp, VDIR, args->ta_root_uid, 172 args->ta_root_gid, args->ta_root_mode & ALLPERMS, NULL, 173 VNOVAL, &root); 174 KASSERT(error == 0 && root != NULL); 175 176 /* 177 * Parent of the root inode is itself. Also, root inode has no 178 * directory entry (i.e. is never attached), thus hold an extra 179 * reference (link) for it. 180 */ 181 root->tn_links++; 182 root->tn_spec.tn_dir.tn_parent = root; 183 tmp->tm_root = root; 184 185 mp->mnt_data = tmp; 186 mp->mnt_flag |= MNT_LOCAL; 187 mp->mnt_stat.f_namemax = TMPFS_MAXNAMLEN; 188 mp->mnt_fs_bshift = PAGE_SHIFT; 189 mp->mnt_dev_bshift = DEV_BSHIFT; 190 mp->mnt_iflag |= IMNT_MPSAFE; 191 vfs_getnewfsid(mp); 192 193 error = set_statvfs_info(path, UIO_USERSPACE, "tmpfs", UIO_SYSSPACE, 194 mp->mnt_op->vfs_name, mp, curlwp); 195 if (error) { 196 (void)tmpfs_unmount(mp, MNT_FORCE); 197 } 198 return error; 199 } 200 201 static int 202 tmpfs_start(struct mount *mp, int flags) 203 { 204 205 return 0; 206 } 207 208 static int 209 tmpfs_unmount(struct mount *mp, int mntflags) 210 { 211 tmpfs_mount_t *tmp = VFS_TO_TMPFS(mp); 212 tmpfs_node_t *node, *cnode; 213 int error, flags = 0; 214 215 /* Handle forced unmounts. */ 216 if (mntflags & MNT_FORCE) 217 flags |= FORCECLOSE; 218 219 /* Finalize all pending I/O. */ 220 error = vflush(mp, NULL, flags); 221 if (error != 0) 222 return error; 223 224 /* 225 * First round, detach and destroy all directory entries. 226 * Also, clear the pointers to the vnodes - they are gone. 227 */ 228 LIST_FOREACH(node, &tmp->tm_nodes, tn_entries) { 229 tmpfs_dirent_t *de; 230 231 node->tn_vnode = NULL; 232 if (node->tn_type != VDIR) { 233 continue; 234 } 235 while ((de = TAILQ_FIRST(&node->tn_spec.tn_dir.tn_dir)) != NULL) { 236 cnode = de->td_node; 237 if (cnode && cnode != TMPFS_NODE_WHITEOUT) { 238 cnode->tn_vnode = NULL; 239 } 240 tmpfs_dir_detach(node, de); 241 tmpfs_free_dirent(tmp, de); 242 } 243 /* Extra virtual entry (itself for the root). */ 244 node->tn_links--; 245 } 246 247 /* Release the reference on root (diagnostic). */ 248 node = tmp->tm_root; 249 node->tn_links--; 250 251 /* Second round, destroy all inodes. */ 252 while ((node = LIST_FIRST(&tmp->tm_nodes)) != NULL) { 253 tmpfs_free_node(tmp, node); 254 } 255 256 /* Throw away the tmpfs_mount structure. */ 257 tmpfs_mntmem_destroy(tmp); 258 mutex_destroy(&tmp->tm_lock); 259 kmem_free(tmp, sizeof(*tmp)); 260 mp->mnt_data = NULL; 261 262 return 0; 263 } 264 265 static int 266 tmpfs_root(struct mount *mp, vnode_t **vpp) 267 { 268 tmpfs_node_t *node = VFS_TO_TMPFS(mp)->tm_root; 269 270 mutex_enter(&node->tn_vlock); 271 return tmpfs_vnode_get(mp, node, vpp); 272 } 273 274 static int 275 tmpfs_vget(struct mount *mp, ino_t ino, vnode_t **vpp) 276 { 277 278 return EOPNOTSUPP; 279 } 280 281 static int 282 tmpfs_fhtovp(struct mount *mp, struct fid *fhp, vnode_t **vpp) 283 { 284 tmpfs_mount_t *tmp = VFS_TO_TMPFS(mp); 285 tmpfs_node_t *node; 286 tmpfs_fid_t tfh; 287 int error; 288 289 if (fhp->fid_len != sizeof(tmpfs_fid_t)) { 290 return EINVAL; 291 } 292 memcpy(&tfh, fhp, sizeof(tmpfs_fid_t)); 293 294 mutex_enter(&tmp->tm_lock); 295 LIST_FOREACH(node, &tmp->tm_nodes, tn_entries) { 296 if (node->tn_id == tfh.tf_id) { 297 mutex_enter(&node->tn_vlock); 298 break; 299 } 300 } 301 mutex_exit(&tmp->tm_lock); 302 303 if (node == NULL) 304 return ESTALE; 305 /* Will release the tn_vlock. */ 306 if ((error = tmpfs_vnode_get(mp, node, vpp)) != 0) 307 return error; 308 if (TMPFS_NODE_GEN(node) != tfh.tf_gen) { 309 vput(*vpp); 310 *vpp = NULL; 311 return ESTALE; 312 } 313 314 return 0; 315 } 316 317 static int 318 tmpfs_vptofh(vnode_t *vp, struct fid *fhp, size_t *fh_size) 319 { 320 tmpfs_fid_t tfh; 321 tmpfs_node_t *node; 322 323 if (*fh_size < sizeof(tmpfs_fid_t)) { 324 *fh_size = sizeof(tmpfs_fid_t); 325 return E2BIG; 326 } 327 *fh_size = sizeof(tmpfs_fid_t); 328 node = VP_TO_TMPFS_NODE(vp); 329 330 memset(&tfh, 0, sizeof(tfh)); 331 tfh.tf_len = sizeof(tmpfs_fid_t); 332 tfh.tf_gen = TMPFS_NODE_GEN(node); 333 tfh.tf_id = node->tn_id; 334 memcpy(fhp, &tfh, sizeof(tfh)); 335 336 return 0; 337 } 338 339 static int 340 tmpfs_statvfs(struct mount *mp, struct statvfs *sbp) 341 { 342 tmpfs_mount_t *tmp; 343 fsfilcnt_t freenodes; 344 size_t avail; 345 346 tmp = VFS_TO_TMPFS(mp); 347 348 sbp->f_iosize = sbp->f_frsize = sbp->f_bsize = PAGE_SIZE; 349 350 mutex_enter(&tmp->tm_acc_lock); 351 avail = tmpfs_pages_avail(tmp); 352 sbp->f_blocks = (tmpfs_bytes_max(tmp) >> PAGE_SHIFT); 353 sbp->f_bavail = sbp->f_bfree = avail; 354 sbp->f_bresvd = 0; 355 356 freenodes = MIN(tmp->tm_nodes_max - tmp->tm_nodes_cnt, 357 avail * PAGE_SIZE / sizeof(tmpfs_node_t)); 358 359 sbp->f_files = tmp->tm_nodes_cnt + freenodes; 360 sbp->f_favail = sbp->f_ffree = freenodes; 361 sbp->f_fresvd = 0; 362 mutex_exit(&tmp->tm_acc_lock); 363 364 copy_statvfs_info(sbp, mp); 365 366 return 0; 367 } 368 369 static int 370 tmpfs_sync(struct mount *mp, int waitfor, kauth_cred_t uc) 371 { 372 373 return 0; 374 } 375 376 static int 377 tmpfs_snapshot(struct mount *mp, vnode_t *vp, struct timespec *ctime) 378 { 379 380 return EOPNOTSUPP; 381 } 382 383 /* 384 * tmpfs vfs operations. 385 */ 386 387 extern const struct vnodeopv_desc tmpfs_fifoop_opv_desc; 388 extern const struct vnodeopv_desc tmpfs_specop_opv_desc; 389 extern const struct vnodeopv_desc tmpfs_vnodeop_opv_desc; 390 391 const struct vnodeopv_desc * const tmpfs_vnodeopv_descs[] = { 392 &tmpfs_fifoop_opv_desc, 393 &tmpfs_specop_opv_desc, 394 &tmpfs_vnodeop_opv_desc, 395 NULL, 396 }; 397 398 struct vfsops tmpfs_vfsops = { 399 .vfs_name = MOUNT_TMPFS, 400 .vfs_min_mount_data = sizeof (struct tmpfs_args), 401 .vfs_mount = tmpfs_mount, 402 .vfs_start = tmpfs_start, 403 .vfs_unmount = tmpfs_unmount, 404 .vfs_root = tmpfs_root, 405 .vfs_quotactl = (void *)eopnotsupp, 406 .vfs_statvfs = tmpfs_statvfs, 407 .vfs_sync = tmpfs_sync, 408 .vfs_vget = tmpfs_vget, 409 .vfs_fhtovp = tmpfs_fhtovp, 410 .vfs_vptofh = tmpfs_vptofh, 411 .vfs_init = tmpfs_init, 412 .vfs_done = tmpfs_done, 413 .vfs_snapshot = tmpfs_snapshot, 414 .vfs_extattrctl = vfs_stdextattrctl, 415 .vfs_suspendctl = (void *)eopnotsupp, 416 .vfs_renamelock_enter = genfs_renamelock_enter, 417 .vfs_renamelock_exit = genfs_renamelock_exit, 418 .vfs_fsync = (void *)eopnotsupp, 419 .vfs_opv_descs = tmpfs_vnodeopv_descs 420 }; 421 422 static int 423 tmpfs_modcmd(modcmd_t cmd, void *arg) 424 { 425 426 switch (cmd) { 427 case MODULE_CMD_INIT: 428 return vfs_attach(&tmpfs_vfsops); 429 case MODULE_CMD_FINI: 430 return vfs_detach(&tmpfs_vfsops); 431 default: 432 return ENOTTY; 433 } 434 } 435