1/* $NetBSD: tmpfs_vfsops.c,v 1.52 2011/09/27 01:10:43 christos 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 1.52 2011/09/27 01:10:43 christos 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 60MODULE(MODULE_CLASS_VFS, tmpfs, NULL); 61 62struct pool tmpfs_dirent_pool; 63struct pool tmpfs_node_pool; 64 65static int tmpfs_mount(struct mount *, const char *, void *, size_t *); 66static int tmpfs_start(struct mount *, int); 67static int tmpfs_unmount(struct mount *, int); 68static int tmpfs_root(struct mount *, vnode_t **); 69static int tmpfs_vget(struct mount *, ino_t, vnode_t **); 70static int tmpfs_fhtovp(struct mount *, struct fid *, vnode_t **); 71static int tmpfs_vptofh(struct vnode *, struct fid *, size_t *); 72static int tmpfs_statvfs(struct mount *, struct statvfs *); 73static int tmpfs_sync(struct mount *, int, kauth_cred_t); 74static void tmpfs_init(void); 75static void tmpfs_done(void); 76static int tmpfs_snapshot(struct mount *, vnode_t *, struct timespec *); 77 78static void 79tmpfs_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 88static void 89tmpfs_done(void) 90{ 91 92 pool_destroy(&tmpfs_dirent_pool); 93 pool_destroy(&tmpfs_node_pool); 94} 95 96static int 97tmpfs_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 201static int 202tmpfs_start(struct mount *mp, int flags) 203{ 204 205 return 0; 206} 207 208static int 209tmpfs_unmount(struct mount *mp, int mntflags) 210{ 211 tmpfs_mount_t *tmp; 212 tmpfs_node_t *node; 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 tmp = VFS_TO_TMPFS(mp); 225 226 /* Destroy any existing inodes. */ 227 while ((node = LIST_FIRST(&tmp->tm_nodes)) != NULL) { 228 if (node->tn_type == VDIR) { 229 tmpfs_dirent_t *de; 230 231 /* Destroy any directory entries. */ 232 de = TAILQ_FIRST(&node->tn_spec.tn_dir.tn_dir); 233 while (de != NULL) { 234 tmpfs_dirent_t *nde; 235 236 nde = TAILQ_NEXT(de, td_entries); 237 tmpfs_free_dirent(tmp, de); 238 node->tn_size -= sizeof(tmpfs_dirent_t); 239 de = nde; 240 } 241 } 242 /* Removes inode from the list. */ 243 tmpfs_free_node(tmp, node); 244 } 245 246 /* Throw away the tmpfs_mount structure. */ 247 tmpfs_mntmem_destroy(tmp); 248 mutex_destroy(&tmp->tm_lock); 249 kmem_free(tmp, sizeof(*tmp)); 250 mp->mnt_data = NULL; 251 252 return 0; 253} 254 255static int 256tmpfs_root(struct mount *mp, vnode_t **vpp) 257{ 258 tmpfs_node_t *node = VFS_TO_TMPFS(mp)->tm_root; 259 260 mutex_enter(&node->tn_vlock); 261 return tmpfs_vnode_get(mp, node, vpp); 262} 263 264static int 265tmpfs_vget(struct mount *mp, ino_t ino, vnode_t **vpp) 266{ 267 268 printf("tmpfs_vget called; need for it unknown yet\n"); 269 return EOPNOTSUPP; 270} 271 272static int 273tmpfs_fhtovp(struct mount *mp, struct fid *fhp, vnode_t **vpp) 274{ 275 tmpfs_mount_t *tmp = VFS_TO_TMPFS(mp); 276 tmpfs_node_t *node; 277 tmpfs_fid_t tfh; 278 279 if (fhp->fid_len != sizeof(tmpfs_fid_t)) { 280 return EINVAL; 281 } 282 memcpy(&tfh, fhp, sizeof(tmpfs_fid_t)); 283 284 mutex_enter(&tmp->tm_lock); 285 LIST_FOREACH(node, &tmp->tm_nodes, tn_entries) { 286 if (node->tn_id != tfh.tf_id) { 287 continue; 288 } 289 if (TMPFS_NODE_GEN(node) != tfh.tf_gen) { 290 continue; 291 } 292 mutex_enter(&node->tn_vlock); 293 break; 294 } 295 mutex_exit(&tmp->tm_lock); 296 297 /* Will release the tn_vlock. */ 298 return node ? tmpfs_vnode_get(mp, node, vpp) : ESTALE; 299} 300 301static int 302tmpfs_vptofh(vnode_t *vp, struct fid *fhp, size_t *fh_size) 303{ 304 tmpfs_fid_t tfh; 305 tmpfs_node_t *node; 306 307 if (*fh_size < sizeof(tmpfs_fid_t)) { 308 *fh_size = sizeof(tmpfs_fid_t); 309 return E2BIG; 310 } 311 *fh_size = sizeof(tmpfs_fid_t); 312 node = VP_TO_TMPFS_NODE(vp); 313 314 memset(&tfh, 0, sizeof(tfh)); 315 tfh.tf_len = sizeof(tmpfs_fid_t); 316 tfh.tf_gen = TMPFS_NODE_GEN(node); 317 tfh.tf_id = node->tn_id; 318 memcpy(fhp, &tfh, sizeof(tfh)); 319 320 return 0; 321} 322 323static int 324tmpfs_statvfs(struct mount *mp, struct statvfs *sbp) 325{ 326 tmpfs_mount_t *tmp; 327 fsfilcnt_t freenodes; 328 size_t avail; 329 330 tmp = VFS_TO_TMPFS(mp); 331 332 sbp->f_iosize = sbp->f_frsize = sbp->f_bsize = PAGE_SIZE; 333 334 mutex_enter(&tmp->tm_acc_lock); 335 avail = tmpfs_pages_avail(tmp); 336 sbp->f_blocks = (tmpfs_bytes_max(tmp) >> PAGE_SHIFT); 337 sbp->f_bavail = sbp->f_bfree = avail; 338 sbp->f_bresvd = 0; 339 340 freenodes = MIN(tmp->tm_nodes_max - tmp->tm_nodes_cnt, 341 avail * PAGE_SIZE / sizeof(tmpfs_node_t)); 342 343 sbp->f_files = tmp->tm_nodes_cnt + freenodes; 344 sbp->f_favail = sbp->f_ffree = freenodes; 345 sbp->f_fresvd = 0; 346 mutex_exit(&tmp->tm_acc_lock); 347 348 copy_statvfs_info(sbp, mp); 349 350 return 0; 351} 352 353static int 354tmpfs_sync(struct mount *mp, int waitfor, kauth_cred_t uc) 355{ 356 357 return 0; 358} 359 360static int 361tmpfs_snapshot(struct mount *mp, vnode_t *vp, struct timespec *ctime) 362{ 363 364 return EOPNOTSUPP; 365} 366 367/* 368 * tmpfs vfs operations. 369 */ 370 371extern const struct vnodeopv_desc tmpfs_fifoop_opv_desc; 372extern const struct vnodeopv_desc tmpfs_specop_opv_desc; 373extern const struct vnodeopv_desc tmpfs_vnodeop_opv_desc; 374 375const struct vnodeopv_desc * const tmpfs_vnodeopv_descs[] = { 376 &tmpfs_fifoop_opv_desc, 377 &tmpfs_specop_opv_desc, 378 &tmpfs_vnodeop_opv_desc, 379 NULL, 380}; 381 382struct vfsops tmpfs_vfsops = { 383 MOUNT_TMPFS, /* vfs_name */ 384 sizeof (struct tmpfs_args), 385 tmpfs_mount, /* vfs_mount */ 386 tmpfs_start, /* vfs_start */ 387 tmpfs_unmount, /* vfs_unmount */ 388 tmpfs_root, /* vfs_root */ 389 (void *)eopnotsupp, /* vfs_quotactl */ 390 tmpfs_statvfs, /* vfs_statvfs */ 391 tmpfs_sync, /* vfs_sync */ 392 tmpfs_vget, /* vfs_vget */ 393 tmpfs_fhtovp, /* vfs_fhtovp */ 394 tmpfs_vptofh, /* vfs_vptofh */ 395 tmpfs_init, /* vfs_init */ 396 NULL, /* vfs_reinit */ 397 tmpfs_done, /* vfs_done */ 398 NULL, /* vfs_mountroot */ 399 tmpfs_snapshot, /* vfs_snapshot */ 400 vfs_stdextattrctl, /* vfs_extattrctl */ 401 (void *)eopnotsupp, /* vfs_suspendctl */ 402 genfs_renamelock_enter, 403 genfs_renamelock_exit, 404 (void *)eopnotsupp, 405 tmpfs_vnodeopv_descs, 406 0, /* vfs_refcount */ 407 { NULL, NULL }, 408}; 409 410static int 411tmpfs_modcmd(modcmd_t cmd, void *arg) 412{ 413 414 switch (cmd) { 415 case MODULE_CMD_INIT: 416 return vfs_attach(&tmpfs_vfsops); 417 case MODULE_CMD_FINI: 418 return vfs_detach(&tmpfs_vfsops); 419 default: 420 return ENOTTY; 421 } 422} 423