1 /* $NetBSD: tmpfs_vfsops.c,v 1.10 2005/12/11 12:24:29 christos Exp $ */ 2 3 /*- 4 * Copyright (c) 2005 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 #include <sys/cdefs.h> 44 __FBSDID("$FreeBSD$"); 45 46 #include <sys/param.h> 47 #include <sys/limits.h> 48 #include <sys/lock.h> 49 #include <sys/mutex.h> 50 #include <sys/kernel.h> 51 #include <sys/stat.h> 52 #include <sys/systm.h> 53 #include <sys/sysctl.h> 54 55 #include <vm/vm.h> 56 #include <vm/vm_object.h> 57 #include <vm/vm_param.h> 58 59 #include <fs/tmpfs/tmpfs.h> 60 61 /* 62 * Default permission for root node 63 */ 64 #define TMPFS_DEFAULT_ROOT_MODE (S_IRWXU|S_IRGRP|S_IXGRP|S_IROTH|S_IXOTH) 65 66 MALLOC_DEFINE(M_TMPFSMNT, "tmpfs mount", "tmpfs mount structures"); 67 MALLOC_DEFINE(M_TMPFSNAME, "tmpfs name", "tmpfs file names"); 68 69 /* --------------------------------------------------------------------- */ 70 71 static int tmpfs_mount(struct mount *); 72 static int tmpfs_unmount(struct mount *, int); 73 static int tmpfs_root(struct mount *, int flags, struct vnode **); 74 static int tmpfs_fhtovp(struct mount *, struct fid *, int, 75 struct vnode **); 76 static int tmpfs_statfs(struct mount *, struct statfs *); 77 78 /* --------------------------------------------------------------------- */ 79 80 static const char *tmpfs_opts[] = { 81 "from", "size", "maxfilesize", "inodes", "uid", "gid", "mode", "export", 82 NULL 83 }; 84 85 /* --------------------------------------------------------------------- */ 86 87 static int 88 tmpfs_node_ctor(void *mem, int size, void *arg, int flags) 89 { 90 struct tmpfs_node *node = (struct tmpfs_node *)mem; 91 92 node->tn_gen++; 93 node->tn_size = 0; 94 node->tn_status = 0; 95 node->tn_flags = 0; 96 node->tn_links = 0; 97 node->tn_vnode = NULL; 98 node->tn_vpstate = 0; 99 100 return (0); 101 } 102 103 static void 104 tmpfs_node_dtor(void *mem, int size, void *arg) 105 { 106 struct tmpfs_node *node = (struct tmpfs_node *)mem; 107 node->tn_type = VNON; 108 } 109 110 static int 111 tmpfs_node_init(void *mem, int size, int flags) 112 { 113 struct tmpfs_node *node = (struct tmpfs_node *)mem; 114 node->tn_id = 0; 115 116 mtx_init(&node->tn_interlock, "tmpfs node interlock", NULL, MTX_DEF); 117 node->tn_gen = arc4random(); 118 119 return (0); 120 } 121 122 static void 123 tmpfs_node_fini(void *mem, int size) 124 { 125 struct tmpfs_node *node = (struct tmpfs_node *)mem; 126 127 mtx_destroy(&node->tn_interlock); 128 } 129 130 static int 131 tmpfs_mount(struct mount *mp) 132 { 133 struct tmpfs_mount *tmp; 134 struct tmpfs_node *root; 135 size_t pages; 136 uint32_t nodes; 137 int error; 138 /* Size counters. */ 139 u_int nodes_max; 140 u_quad_t size_max, maxfilesize; 141 142 /* Root node attributes. */ 143 uid_t root_uid; 144 gid_t root_gid; 145 mode_t root_mode; 146 147 struct vattr va; 148 149 if (vfs_filteropt(mp->mnt_optnew, tmpfs_opts)) 150 return (EINVAL); 151 152 if (mp->mnt_flag & MNT_UPDATE) { 153 /* 154 * Only support update mounts for NFS export. 155 */ 156 if (vfs_flagopt(mp->mnt_optnew, "export", NULL, 0)) 157 return (0); 158 return (EOPNOTSUPP); 159 } 160 161 vn_lock(mp->mnt_vnodecovered, LK_SHARED | LK_RETRY); 162 error = VOP_GETATTR(mp->mnt_vnodecovered, &va, mp->mnt_cred); 163 VOP_UNLOCK(mp->mnt_vnodecovered, 0); 164 if (error) 165 return (error); 166 167 if (mp->mnt_cred->cr_ruid != 0 || 168 vfs_scanopt(mp->mnt_optnew, "gid", "%d", &root_gid) != 1) 169 root_gid = va.va_gid; 170 if (mp->mnt_cred->cr_ruid != 0 || 171 vfs_scanopt(mp->mnt_optnew, "uid", "%d", &root_uid) != 1) 172 root_uid = va.va_uid; 173 if (mp->mnt_cred->cr_ruid != 0 || 174 vfs_scanopt(mp->mnt_optnew, "mode", "%ho", &root_mode) != 1) 175 root_mode = va.va_mode; 176 if (vfs_scanopt(mp->mnt_optnew, "inodes", "%u", &nodes_max) != 1) 177 nodes_max = 0; 178 if (vfs_scanopt(mp->mnt_optnew, "size", "%qu", &size_max) != 1) 179 size_max = 0; 180 if (vfs_scanopt(mp->mnt_optnew, "maxfilesize", "%qu", 181 &maxfilesize) != 1) 182 maxfilesize = 0; 183 184 /* Do not allow mounts if we do not have enough memory to preserve 185 * the minimum reserved pages. */ 186 if (tmpfs_mem_info() < TMPFS_PAGES_RESERVED) 187 return ENOSPC; 188 189 /* Get the maximum number of memory pages this file system is 190 * allowed to use, based on the maximum size the user passed in 191 * the mount structure. A value of zero is treated as if the 192 * maximum available space was requested. */ 193 if (size_max < PAGE_SIZE || size_max > SIZE_MAX - PAGE_SIZE) 194 pages = SIZE_MAX; 195 else 196 pages = howmany(size_max, PAGE_SIZE); 197 MPASS(pages > 0); 198 199 if (nodes_max <= 3) { 200 if (pages > UINT32_MAX - 3) 201 nodes = UINT32_MAX; 202 else 203 nodes = pages + 3; 204 } else 205 nodes = nodes_max; 206 MPASS(nodes >= 3); 207 208 /* Allocate the tmpfs mount structure and fill it. */ 209 tmp = (struct tmpfs_mount *)malloc(sizeof(struct tmpfs_mount), 210 M_TMPFSMNT, M_WAITOK | M_ZERO); 211 212 mtx_init(&tmp->allnode_lock, "tmpfs allnode lock", NULL, MTX_DEF); 213 tmp->tm_nodes_max = nodes; 214 tmp->tm_nodes_inuse = 0; 215 tmp->tm_maxfilesize = maxfilesize > 0 ? maxfilesize : UINT64_MAX; 216 LIST_INIT(&tmp->tm_nodes_used); 217 218 tmp->tm_pages_max = pages; 219 tmp->tm_pages_used = 0; 220 tmp->tm_ino_unr = new_unrhdr(2, INT_MAX, &tmp->allnode_lock); 221 tmp->tm_dirent_pool = uma_zcreate("TMPFS dirent", 222 sizeof(struct tmpfs_dirent), 223 NULL, NULL, NULL, NULL, 224 UMA_ALIGN_PTR, 0); 225 tmp->tm_node_pool = uma_zcreate("TMPFS node", 226 sizeof(struct tmpfs_node), 227 tmpfs_node_ctor, tmpfs_node_dtor, 228 tmpfs_node_init, tmpfs_node_fini, 229 UMA_ALIGN_PTR, 0); 230 231 /* Allocate the root node. */ 232 error = tmpfs_alloc_node(tmp, VDIR, root_uid, 233 root_gid, root_mode & ALLPERMS, NULL, NULL, 234 VNOVAL, &root); 235 236 if (error != 0 || root == NULL) { 237 uma_zdestroy(tmp->tm_node_pool); 238 uma_zdestroy(tmp->tm_dirent_pool); 239 delete_unrhdr(tmp->tm_ino_unr); 240 free(tmp, M_TMPFSMNT); 241 return error; 242 } 243 KASSERT(root->tn_id == 2, ("tmpfs root with invalid ino: %d", root->tn_id)); 244 tmp->tm_root = root; 245 246 MNT_ILOCK(mp); 247 mp->mnt_flag |= MNT_LOCAL; 248 mp->mnt_kern_flag |= MNTK_MPSAFE; 249 MNT_IUNLOCK(mp); 250 251 mp->mnt_data = tmp; 252 mp->mnt_stat.f_namemax = MAXNAMLEN; 253 vfs_getnewfsid(mp); 254 vfs_mountedfrom(mp, "tmpfs"); 255 256 return 0; 257 } 258 259 /* --------------------------------------------------------------------- */ 260 261 /* ARGSUSED2 */ 262 static int 263 tmpfs_unmount(struct mount *mp, int mntflags) 264 { 265 int error; 266 int flags = 0; 267 struct tmpfs_mount *tmp; 268 struct tmpfs_node *node; 269 270 /* Handle forced unmounts. */ 271 if (mntflags & MNT_FORCE) 272 flags |= FORCECLOSE; 273 274 /* Finalize all pending I/O. */ 275 error = vflush(mp, 0, flags, curthread); 276 if (error != 0) 277 return error; 278 279 tmp = VFS_TO_TMPFS(mp); 280 281 /* Free all associated data. The loop iterates over the linked list 282 * we have containing all used nodes. For each of them that is 283 * a directory, we free all its directory entries. Note that after 284 * freeing a node, it will automatically go to the available list, 285 * so we will later have to iterate over it to release its items. */ 286 node = LIST_FIRST(&tmp->tm_nodes_used); 287 while (node != NULL) { 288 struct tmpfs_node *next; 289 290 if (node->tn_type == VDIR) { 291 struct tmpfs_dirent *de; 292 293 de = TAILQ_FIRST(&node->tn_dir.tn_dirhead); 294 while (de != NULL) { 295 struct tmpfs_dirent *nde; 296 297 nde = TAILQ_NEXT(de, td_entries); 298 tmpfs_free_dirent(tmp, de, FALSE); 299 de = nde; 300 node->tn_size -= sizeof(struct tmpfs_dirent); 301 } 302 } 303 304 next = LIST_NEXT(node, tn_entries); 305 tmpfs_free_node(tmp, node); 306 node = next; 307 } 308 309 uma_zdestroy(tmp->tm_dirent_pool); 310 uma_zdestroy(tmp->tm_node_pool); 311 delete_unrhdr(tmp->tm_ino_unr); 312 313 mtx_destroy(&tmp->allnode_lock); 314 MPASS(tmp->tm_pages_used == 0); 315 MPASS(tmp->tm_nodes_inuse == 0); 316 317 /* Throw away the tmpfs_mount structure. */ 318 free(mp->mnt_data, M_TMPFSMNT); 319 mp->mnt_data = NULL; 320 321 MNT_ILOCK(mp); 322 mp->mnt_flag &= ~MNT_LOCAL; 323 MNT_IUNLOCK(mp); 324 return 0; 325 } 326 327 /* --------------------------------------------------------------------- */ 328 329 static int 330 tmpfs_root(struct mount *mp, int flags, struct vnode **vpp) 331 { 332 int error; 333 error = tmpfs_alloc_vp(mp, VFS_TO_TMPFS(mp)->tm_root, flags, vpp); 334 335 if (!error) 336 (*vpp)->v_vflag |= VV_ROOT; 337 338 return error; 339 } 340 341 /* --------------------------------------------------------------------- */ 342 343 static int 344 tmpfs_fhtovp(struct mount *mp, struct fid *fhp, int flags, 345 struct vnode **vpp) 346 { 347 boolean_t found; 348 struct tmpfs_fid *tfhp; 349 struct tmpfs_mount *tmp; 350 struct tmpfs_node *node; 351 352 tmp = VFS_TO_TMPFS(mp); 353 354 tfhp = (struct tmpfs_fid *)fhp; 355 if (tfhp->tf_len != sizeof(struct tmpfs_fid)) 356 return EINVAL; 357 358 if (tfhp->tf_id >= tmp->tm_nodes_max) 359 return EINVAL; 360 361 found = FALSE; 362 363 TMPFS_LOCK(tmp); 364 LIST_FOREACH(node, &tmp->tm_nodes_used, tn_entries) { 365 if (node->tn_id == tfhp->tf_id && 366 node->tn_gen == tfhp->tf_gen) { 367 found = TRUE; 368 break; 369 } 370 } 371 TMPFS_UNLOCK(tmp); 372 373 if (found) 374 return (tmpfs_alloc_vp(mp, node, LK_EXCLUSIVE, vpp)); 375 376 return (EINVAL); 377 } 378 379 /* --------------------------------------------------------------------- */ 380 381 /* ARGSUSED2 */ 382 static int 383 tmpfs_statfs(struct mount *mp, struct statfs *sbp) 384 { 385 fsfilcnt_t freenodes; 386 struct tmpfs_mount *tmp; 387 388 tmp = VFS_TO_TMPFS(mp); 389 390 sbp->f_iosize = PAGE_SIZE; 391 sbp->f_bsize = PAGE_SIZE; 392 393 sbp->f_blocks = TMPFS_PAGES_MAX(tmp); 394 sbp->f_bavail = sbp->f_bfree = TMPFS_PAGES_AVAIL(tmp); 395 396 freenodes = MIN(tmp->tm_nodes_max - tmp->tm_nodes_inuse, 397 TMPFS_PAGES_AVAIL(tmp) * PAGE_SIZE / sizeof(struct tmpfs_node)); 398 399 sbp->f_files = freenodes + tmp->tm_nodes_inuse; 400 sbp->f_ffree = freenodes; 401 /* sbp->f_owner = tmp->tn_uid; */ 402 403 return 0; 404 } 405 406 /* --------------------------------------------------------------------- */ 407 408 /* 409 * tmpfs vfs operations. 410 */ 411 412 struct vfsops tmpfs_vfsops = { 413 .vfs_mount = tmpfs_mount, 414 .vfs_unmount = tmpfs_unmount, 415 .vfs_root = tmpfs_root, 416 .vfs_statfs = tmpfs_statfs, 417 .vfs_fhtovp = tmpfs_fhtovp, 418 }; 419 VFS_SET(tmpfs_vfsops, tmpfs, 0); 420