tmpfs_vfsops.c revision 1.51
1/* $NetBSD: tmpfs_vfsops.c,v 1.51 2011/05/29 22:29:07 rmind 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 1.51 2011/05/29 22:29:07 rmind 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 /* Validate the version. */ 107 if (*data_len < sizeof(*args) || 108 args->ta_version != TMPFS_ARGS_VERSION) 109 return EINVAL; 110 111 /* Handle retrieval of mount point arguments. */ 112 if (mp->mnt_flag & MNT_GETARGS) { 113 if (mp->mnt_data == NULL) 114 return EIO; 115 tmp = VFS_TO_TMPFS(mp); 116 117 args->ta_version = TMPFS_ARGS_VERSION; 118 args->ta_nodes_max = tmp->tm_nodes_max; 119 args->ta_size_max = tmp->tm_mem_limit; 120 121 root = tmp->tm_root; 122 args->ta_root_uid = root->tn_uid; 123 args->ta_root_gid = root->tn_gid; 124 args->ta_root_mode = root->tn_mode; 125 126 *data_len = sizeof(*args); 127 return 0; 128 } 129 130 if (mp->mnt_flag & MNT_UPDATE) { 131 /* TODO */ 132 return EOPNOTSUPP; 133 } 134 135 /* Prohibit mounts if there is not enough memory. */ 136 if (tmpfs_mem_info(true) < TMPFS_PAGES_RESERVED) 137 return EINVAL; 138 139 /* Get the memory usage limit for this file-system. */ 140 if (args->ta_size_max < PAGE_SIZE) { 141 memlimit = UINT64_MAX; 142 } else { 143 memlimit = args->ta_size_max; 144 } 145 KASSERT(memlimit > 0); 146 147 if (args->ta_nodes_max <= 3) { 148 nodes = 3 + (memlimit / 1024); 149 } else { 150 nodes = args->ta_nodes_max; 151 } 152 nodes = MIN(nodes, INT_MAX); 153 KASSERT(nodes >= 3); 154 155 /* Allocate the tmpfs mount structure and fill it. */ 156 tmp = kmem_zalloc(sizeof(tmpfs_mount_t), KM_SLEEP); 157 if (tmp == NULL) 158 return ENOMEM; 159 160 tmp->tm_nodes_max = nodes; 161 tmp->tm_nodes_cnt = 0; 162 LIST_INIT(&tmp->tm_nodes); 163 164 mutex_init(&tmp->tm_lock, MUTEX_DEFAULT, IPL_NONE); 165 tmpfs_mntmem_init(tmp, memlimit); 166 167 /* Allocate the root node. */ 168 error = tmpfs_alloc_node(tmp, VDIR, args->ta_root_uid, 169 args->ta_root_gid, args->ta_root_mode & ALLPERMS, NULL, 170 VNOVAL, &root); 171 KASSERT(error == 0 && root != NULL); 172 173 /* 174 * Parent of the root inode is itself. Also, root inode has no 175 * directory entry (i.e. is never attached), thus hold an extra 176 * reference (link) for it. 177 */ 178 root->tn_links++; 179 root->tn_spec.tn_dir.tn_parent = root; 180 tmp->tm_root = root; 181 182 mp->mnt_data = tmp; 183 mp->mnt_flag |= MNT_LOCAL; 184 mp->mnt_stat.f_namemax = MAXNAMLEN; 185 mp->mnt_fs_bshift = PAGE_SHIFT; 186 mp->mnt_dev_bshift = DEV_BSHIFT; 187 mp->mnt_iflag |= IMNT_MPSAFE; 188 vfs_getnewfsid(mp); 189 190 error = set_statvfs_info(path, UIO_USERSPACE, "tmpfs", UIO_SYSSPACE, 191 mp->mnt_op->vfs_name, mp, curlwp); 192 if (error) { 193 (void)tmpfs_unmount(mp, MNT_FORCE); 194 } 195 return error; 196} 197 198static int 199tmpfs_start(struct mount *mp, int flags) 200{ 201 202 return 0; 203} 204 205static int 206tmpfs_unmount(struct mount *mp, int mntflags) 207{ 208 tmpfs_mount_t *tmp; 209 tmpfs_node_t *node; 210 int error, flags = 0; 211 212 /* Handle forced unmounts. */ 213 if (mntflags & MNT_FORCE) 214 flags |= FORCECLOSE; 215 216 /* Finalize all pending I/O. */ 217 error = vflush(mp, NULL, flags); 218 if (error != 0) 219 return error; 220 221 tmp = VFS_TO_TMPFS(mp); 222 223 /* Destroy any existing inodes. */ 224 while ((node = LIST_FIRST(&tmp->tm_nodes)) != NULL) { 225 if (node->tn_type == VDIR) { 226 tmpfs_dirent_t *de; 227 228 /* Destroy any directory entries. */ 229 de = TAILQ_FIRST(&node->tn_spec.tn_dir.tn_dir); 230 while (de != NULL) { 231 tmpfs_dirent_t *nde; 232 233 nde = TAILQ_NEXT(de, td_entries); 234 tmpfs_free_dirent(tmp, de); 235 node->tn_size -= sizeof(tmpfs_dirent_t); 236 de = nde; 237 } 238 } 239 /* Removes inode from the list. */ 240 tmpfs_free_node(tmp, node); 241 } 242 243 /* Throw away the tmpfs_mount structure. */ 244 tmpfs_mntmem_destroy(tmp); 245 mutex_destroy(&tmp->tm_lock); 246 kmem_free(tmp, sizeof(*tmp)); 247 mp->mnt_data = NULL; 248 249 return 0; 250} 251 252static int 253tmpfs_root(struct mount *mp, vnode_t **vpp) 254{ 255 tmpfs_node_t *node = VFS_TO_TMPFS(mp)->tm_root; 256 257 mutex_enter(&node->tn_vlock); 258 return tmpfs_vnode_get(mp, node, vpp); 259} 260 261static int 262tmpfs_vget(struct mount *mp, ino_t ino, vnode_t **vpp) 263{ 264 265 printf("tmpfs_vget called; need for it unknown yet\n"); 266 return EOPNOTSUPP; 267} 268 269static int 270tmpfs_fhtovp(struct mount *mp, struct fid *fhp, vnode_t **vpp) 271{ 272 tmpfs_mount_t *tmp = VFS_TO_TMPFS(mp); 273 tmpfs_node_t *node; 274 tmpfs_fid_t tfh; 275 276 if (fhp->fid_len != sizeof(tmpfs_fid_t)) { 277 return EINVAL; 278 } 279 memcpy(&tfh, fhp, sizeof(tmpfs_fid_t)); 280 281 mutex_enter(&tmp->tm_lock); 282 LIST_FOREACH(node, &tmp->tm_nodes, tn_entries) { 283 if (node->tn_id != tfh.tf_id) { 284 continue; 285 } 286 if (TMPFS_NODE_GEN(node) != tfh.tf_gen) { 287 continue; 288 } 289 mutex_enter(&node->tn_vlock); 290 break; 291 } 292 mutex_exit(&tmp->tm_lock); 293 294 /* Will release the tn_vlock. */ 295 return node ? tmpfs_vnode_get(mp, node, vpp) : ESTALE; 296} 297 298static int 299tmpfs_vptofh(vnode_t *vp, struct fid *fhp, size_t *fh_size) 300{ 301 tmpfs_fid_t tfh; 302 tmpfs_node_t *node; 303 304 if (*fh_size < sizeof(tmpfs_fid_t)) { 305 *fh_size = sizeof(tmpfs_fid_t); 306 return E2BIG; 307 } 308 *fh_size = sizeof(tmpfs_fid_t); 309 node = VP_TO_TMPFS_NODE(vp); 310 311 memset(&tfh, 0, sizeof(tfh)); 312 tfh.tf_len = sizeof(tmpfs_fid_t); 313 tfh.tf_gen = TMPFS_NODE_GEN(node); 314 tfh.tf_id = node->tn_id; 315 memcpy(fhp, &tfh, sizeof(tfh)); 316 317 return 0; 318} 319 320static int 321tmpfs_statvfs(struct mount *mp, struct statvfs *sbp) 322{ 323 tmpfs_mount_t *tmp; 324 fsfilcnt_t freenodes; 325 size_t avail; 326 327 tmp = VFS_TO_TMPFS(mp); 328 329 sbp->f_iosize = sbp->f_frsize = sbp->f_bsize = PAGE_SIZE; 330 331 mutex_enter(&tmp->tm_acc_lock); 332 avail = tmpfs_pages_avail(tmp); 333 sbp->f_blocks = (tmpfs_bytes_max(tmp) >> PAGE_SHIFT); 334 sbp->f_bavail = sbp->f_bfree = avail; 335 sbp->f_bresvd = 0; 336 337 freenodes = MIN(tmp->tm_nodes_max - tmp->tm_nodes_cnt, 338 avail * PAGE_SIZE / sizeof(tmpfs_node_t)); 339 340 sbp->f_files = tmp->tm_nodes_cnt + freenodes; 341 sbp->f_favail = sbp->f_ffree = freenodes; 342 sbp->f_fresvd = 0; 343 mutex_exit(&tmp->tm_acc_lock); 344 345 copy_statvfs_info(sbp, mp); 346 347 return 0; 348} 349 350static int 351tmpfs_sync(struct mount *mp, int waitfor, kauth_cred_t uc) 352{ 353 354 return 0; 355} 356 357static int 358tmpfs_snapshot(struct mount *mp, vnode_t *vp, struct timespec *ctime) 359{ 360 361 return EOPNOTSUPP; 362} 363 364/* 365 * tmpfs vfs operations. 366 */ 367 368extern const struct vnodeopv_desc tmpfs_fifoop_opv_desc; 369extern const struct vnodeopv_desc tmpfs_specop_opv_desc; 370extern const struct vnodeopv_desc tmpfs_vnodeop_opv_desc; 371 372const struct vnodeopv_desc * const tmpfs_vnodeopv_descs[] = { 373 &tmpfs_fifoop_opv_desc, 374 &tmpfs_specop_opv_desc, 375 &tmpfs_vnodeop_opv_desc, 376 NULL, 377}; 378 379struct vfsops tmpfs_vfsops = { 380 MOUNT_TMPFS, /* vfs_name */ 381 sizeof (struct tmpfs_args), 382 tmpfs_mount, /* vfs_mount */ 383 tmpfs_start, /* vfs_start */ 384 tmpfs_unmount, /* vfs_unmount */ 385 tmpfs_root, /* vfs_root */ 386 (void *)eopnotsupp, /* vfs_quotactl */ 387 tmpfs_statvfs, /* vfs_statvfs */ 388 tmpfs_sync, /* vfs_sync */ 389 tmpfs_vget, /* vfs_vget */ 390 tmpfs_fhtovp, /* vfs_fhtovp */ 391 tmpfs_vptofh, /* vfs_vptofh */ 392 tmpfs_init, /* vfs_init */ 393 NULL, /* vfs_reinit */ 394 tmpfs_done, /* vfs_done */ 395 NULL, /* vfs_mountroot */ 396 tmpfs_snapshot, /* vfs_snapshot */ 397 vfs_stdextattrctl, /* vfs_extattrctl */ 398 (void *)eopnotsupp, /* vfs_suspendctl */ 399 genfs_renamelock_enter, 400 genfs_renamelock_exit, 401 (void *)eopnotsupp, 402 tmpfs_vnodeopv_descs, 403 0, /* vfs_refcount */ 404 { NULL, NULL }, 405}; 406 407static int 408tmpfs_modcmd(modcmd_t cmd, void *arg) 409{ 410 411 switch (cmd) { 412 case MODULE_CMD_INIT: 413 return vfs_attach(&tmpfs_vfsops); 414 case MODULE_CMD_FINI: 415 return vfs_detach(&tmpfs_vfsops); 416 default: 417 return ENOTTY; 418 } 419} 420