1 /* $NetBSD: tmpfs_vfsops.c,v 1.10 2005/12/11 12:24:29 christos Exp $ */ 2 3 /*- 4 * SPDX-License-Identifier: BSD-2-Clause-NetBSD 5 * 6 * Copyright (c) 2005 The NetBSD Foundation, Inc. 7 * All rights reserved. 8 * 9 * This code is derived from software contributed to The NetBSD Foundation 10 * by Julio M. Merino Vidal, developed as part of Google's Summer of Code 11 * 2005 program. 12 * 13 * Redistribution and use in source and binary forms, with or without 14 * modification, are permitted provided that the following conditions 15 * are met: 16 * 1. Redistributions of source code must retain the above copyright 17 * notice, this list of conditions and the following disclaimer. 18 * 2. Redistributions in binary form must reproduce the above copyright 19 * notice, this list of conditions and the following disclaimer in the 20 * documentation and/or other materials provided with the distribution. 21 * 22 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 23 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 24 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 25 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 26 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 27 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 28 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 29 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 30 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 31 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 32 * POSSIBILITY OF SUCH DAMAGE. 33 */ 34 35 /* 36 * Efficient memory file system. 37 * 38 * tmpfs is a file system that uses FreeBSD's virtual memory 39 * sub-system to store file data and metadata in an efficient way. 40 * This means that it does not follow the structure of an on-disk file 41 * system because it simply does not need to. Instead, it uses 42 * memory-specific data structures and algorithms to automatically 43 * allocate and release resources. 44 */ 45 #include <sys/cdefs.h> 46 __FBSDID("$FreeBSD$"); 47 48 #include <sys/param.h> 49 #include <sys/dirent.h> 50 #include <sys/limits.h> 51 #include <sys/lock.h> 52 #include <sys/mount.h> 53 #include <sys/mutex.h> 54 #include <sys/proc.h> 55 #include <sys/jail.h> 56 #include <sys/kernel.h> 57 #include <sys/rwlock.h> 58 #include <sys/stat.h> 59 #include <sys/systm.h> 60 #include <sys/sysctl.h> 61 #include <sys/vnode.h> 62 63 #include <vm/vm.h> 64 #include <vm/vm_object.h> 65 #include <vm/vm_param.h> 66 67 #include <fs/tmpfs/tmpfs.h> 68 69 /* 70 * Default permission for root node 71 */ 72 #define TMPFS_DEFAULT_ROOT_MODE (S_IRWXU|S_IRGRP|S_IXGRP|S_IROTH|S_IXOTH) 73 74 MALLOC_DEFINE(M_TMPFSMNT, "tmpfs mount", "tmpfs mount structures"); 75 MALLOC_DEFINE(M_TMPFSNAME, "tmpfs name", "tmpfs file names"); 76 77 static int tmpfs_mount(struct mount *); 78 static int tmpfs_unmount(struct mount *, int); 79 static int tmpfs_root(struct mount *, int flags, struct vnode **); 80 static int tmpfs_fhtovp(struct mount *, struct fid *, int, 81 struct vnode **); 82 static int tmpfs_statfs(struct mount *, struct statfs *); 83 static void tmpfs_susp_clean(struct mount *); 84 85 static const char *tmpfs_opts[] = { 86 "from", "size", "maxfilesize", "inodes", "uid", "gid", "mode", "export", 87 "union", "nonc", NULL 88 }; 89 90 static const char *tmpfs_updateopts[] = { 91 "from", "export", NULL 92 }; 93 94 static int 95 tmpfs_node_ctor(void *mem, int size, void *arg, int flags) 96 { 97 struct tmpfs_node *node = (struct tmpfs_node *)mem; 98 99 node->tn_gen++; 100 node->tn_size = 0; 101 node->tn_status = 0; 102 node->tn_flags = 0; 103 node->tn_links = 0; 104 node->tn_vnode = NULL; 105 node->tn_vpstate = 0; 106 107 return (0); 108 } 109 110 static void 111 tmpfs_node_dtor(void *mem, int size, void *arg) 112 { 113 struct tmpfs_node *node = (struct tmpfs_node *)mem; 114 node->tn_type = VNON; 115 } 116 117 static int 118 tmpfs_node_init(void *mem, int size, int flags) 119 { 120 struct tmpfs_node *node = (struct tmpfs_node *)mem; 121 node->tn_id = 0; 122 123 mtx_init(&node->tn_interlock, "tmpfs node interlock", NULL, MTX_DEF); 124 node->tn_gen = arc4random(); 125 126 return (0); 127 } 128 129 static void 130 tmpfs_node_fini(void *mem, int size) 131 { 132 struct tmpfs_node *node = (struct tmpfs_node *)mem; 133 134 mtx_destroy(&node->tn_interlock); 135 } 136 137 static int 138 tmpfs_mount(struct mount *mp) 139 { 140 const size_t nodes_per_page = howmany(PAGE_SIZE, 141 sizeof(struct tmpfs_dirent) + sizeof(struct tmpfs_node)); 142 struct tmpfs_mount *tmp; 143 struct tmpfs_node *root; 144 struct thread *td = curthread; 145 int error; 146 bool nonc; 147 /* Size counters. */ 148 u_quad_t pages; 149 off_t nodes_max, size_max, maxfilesize; 150 151 /* Root node attributes. */ 152 uid_t root_uid; 153 gid_t root_gid; 154 mode_t root_mode; 155 156 struct vattr va; 157 158 if (!prison_allow(td->td_ucred, PR_ALLOW_MOUNT_TMPFS)) 159 return (EPERM); 160 161 if (vfs_filteropt(mp->mnt_optnew, tmpfs_opts)) 162 return (EINVAL); 163 164 if (mp->mnt_flag & MNT_UPDATE) { 165 /* Only support update mounts for certain options. */ 166 if (vfs_filteropt(mp->mnt_optnew, tmpfs_updateopts) != 0) 167 return (EOPNOTSUPP); 168 if (vfs_flagopt(mp->mnt_optnew, "ro", NULL, 0) != 169 ((struct tmpfs_mount *)mp->mnt_data)->tm_ronly) 170 return (EOPNOTSUPP); 171 return (0); 172 } 173 174 vn_lock(mp->mnt_vnodecovered, LK_SHARED | LK_RETRY); 175 error = VOP_GETATTR(mp->mnt_vnodecovered, &va, mp->mnt_cred); 176 VOP_UNLOCK(mp->mnt_vnodecovered, 0); 177 if (error) 178 return (error); 179 180 if (mp->mnt_cred->cr_ruid != 0 || 181 vfs_scanopt(mp->mnt_optnew, "gid", "%d", &root_gid) != 1) 182 root_gid = va.va_gid; 183 if (mp->mnt_cred->cr_ruid != 0 || 184 vfs_scanopt(mp->mnt_optnew, "uid", "%d", &root_uid) != 1) 185 root_uid = va.va_uid; 186 if (mp->mnt_cred->cr_ruid != 0 || 187 vfs_scanopt(mp->mnt_optnew, "mode", "%ho", &root_mode) != 1) 188 root_mode = va.va_mode; 189 if (vfs_getopt_size(mp->mnt_optnew, "inodes", &nodes_max) != 0) 190 nodes_max = 0; 191 if (vfs_getopt_size(mp->mnt_optnew, "size", &size_max) != 0) 192 size_max = 0; 193 if (vfs_getopt_size(mp->mnt_optnew, "maxfilesize", &maxfilesize) != 0) 194 maxfilesize = 0; 195 nonc = vfs_getopt(mp->mnt_optnew, "nonc", NULL, NULL) == 0; 196 197 /* Do not allow mounts if we do not have enough memory to preserve 198 * the minimum reserved pages. */ 199 if (tmpfs_mem_avail() < TMPFS_PAGES_MINRESERVED) 200 return (ENOSPC); 201 202 /* Get the maximum number of memory pages this file system is 203 * allowed to use, based on the maximum size the user passed in 204 * the mount structure. A value of zero is treated as if the 205 * maximum available space was requested. */ 206 if (size_max == 0 || size_max > OFF_MAX - PAGE_SIZE || 207 (SIZE_MAX < OFF_MAX && size_max / PAGE_SIZE >= SIZE_MAX)) 208 pages = SIZE_MAX; 209 else { 210 size_max = roundup(size_max, PAGE_SIZE); 211 pages = howmany(size_max, PAGE_SIZE); 212 } 213 MPASS(pages > 0); 214 215 if (nodes_max <= 3) { 216 if (pages < INT_MAX / nodes_per_page) 217 nodes_max = pages * nodes_per_page; 218 else 219 nodes_max = INT_MAX; 220 } 221 if (nodes_max > INT_MAX) 222 nodes_max = INT_MAX; 223 MPASS(nodes_max >= 3); 224 225 /* Allocate the tmpfs mount structure and fill it. */ 226 tmp = (struct tmpfs_mount *)malloc(sizeof(struct tmpfs_mount), 227 M_TMPFSMNT, M_WAITOK | M_ZERO); 228 229 mtx_init(&tmp->tm_allnode_lock, "tmpfs allnode lock", NULL, MTX_DEF); 230 tmp->tm_nodes_max = nodes_max; 231 tmp->tm_nodes_inuse = 0; 232 tmp->tm_refcount = 1; 233 tmp->tm_maxfilesize = maxfilesize > 0 ? maxfilesize : OFF_MAX; 234 LIST_INIT(&tmp->tm_nodes_used); 235 236 tmp->tm_pages_max = pages; 237 tmp->tm_pages_used = 0; 238 tmp->tm_ino_unr = new_unrhdr(2, INT_MAX, &tmp->tm_allnode_lock); 239 tmp->tm_dirent_pool = uma_zcreate("TMPFS dirent", 240 sizeof(struct tmpfs_dirent), NULL, NULL, NULL, NULL, 241 UMA_ALIGN_PTR, 0); 242 tmp->tm_node_pool = uma_zcreate("TMPFS node", 243 sizeof(struct tmpfs_node), tmpfs_node_ctor, tmpfs_node_dtor, 244 tmpfs_node_init, tmpfs_node_fini, UMA_ALIGN_PTR, 0); 245 tmp->tm_ronly = (mp->mnt_flag & MNT_RDONLY) != 0; 246 tmp->tm_nonc = nonc; 247 248 /* Allocate the root node. */ 249 error = tmpfs_alloc_node(mp, tmp, VDIR, root_uid, root_gid, 250 root_mode & ALLPERMS, NULL, NULL, VNOVAL, &root); 251 252 if (error != 0 || root == NULL) { 253 uma_zdestroy(tmp->tm_node_pool); 254 uma_zdestroy(tmp->tm_dirent_pool); 255 delete_unrhdr(tmp->tm_ino_unr); 256 free(tmp, M_TMPFSMNT); 257 return (error); 258 } 259 KASSERT(root->tn_id == 2, 260 ("tmpfs root with invalid ino: %ju", (uintmax_t)root->tn_id)); 261 tmp->tm_root = root; 262 263 MNT_ILOCK(mp); 264 mp->mnt_flag |= MNT_LOCAL; 265 mp->mnt_kern_flag |= MNTK_LOOKUP_SHARED | MNTK_EXTENDED_SHARED; 266 MNT_IUNLOCK(mp); 267 268 mp->mnt_data = tmp; 269 mp->mnt_stat.f_namemax = MAXNAMLEN; 270 vfs_getnewfsid(mp); 271 vfs_mountedfrom(mp, "tmpfs"); 272 273 return 0; 274 } 275 276 /* ARGSUSED2 */ 277 static int 278 tmpfs_unmount(struct mount *mp, int mntflags) 279 { 280 struct tmpfs_mount *tmp; 281 struct tmpfs_node *node; 282 int error, flags; 283 284 flags = (mntflags & MNT_FORCE) != 0 ? FORCECLOSE : 0; 285 tmp = VFS_TO_TMPFS(mp); 286 287 /* Stop writers */ 288 error = vfs_write_suspend_umnt(mp); 289 if (error != 0) 290 return (error); 291 /* 292 * At this point, nodes cannot be destroyed by any other 293 * thread because write suspension is started. 294 */ 295 296 for (;;) { 297 error = vflush(mp, 0, flags, curthread); 298 if (error != 0) { 299 vfs_write_resume(mp, VR_START_WRITE); 300 return (error); 301 } 302 MNT_ILOCK(mp); 303 if (mp->mnt_nvnodelistsize == 0) { 304 MNT_IUNLOCK(mp); 305 break; 306 } 307 MNT_IUNLOCK(mp); 308 if ((mntflags & MNT_FORCE) == 0) { 309 vfs_write_resume(mp, VR_START_WRITE); 310 return (EBUSY); 311 } 312 } 313 314 TMPFS_LOCK(tmp); 315 while ((node = LIST_FIRST(&tmp->tm_nodes_used)) != NULL) { 316 TMPFS_NODE_LOCK(node); 317 if (node->tn_type == VDIR) 318 tmpfs_dir_destroy(tmp, node); 319 if (tmpfs_free_node_locked(tmp, node, true)) 320 TMPFS_LOCK(tmp); 321 else 322 TMPFS_NODE_UNLOCK(node); 323 } 324 325 mp->mnt_data = NULL; 326 tmpfs_free_tmp(tmp); 327 vfs_write_resume(mp, VR_START_WRITE); 328 329 MNT_ILOCK(mp); 330 mp->mnt_flag &= ~MNT_LOCAL; 331 MNT_IUNLOCK(mp); 332 333 return (0); 334 } 335 336 void 337 tmpfs_free_tmp(struct tmpfs_mount *tmp) 338 { 339 340 MPASS(tmp->tm_refcount > 0); 341 tmp->tm_refcount--; 342 if (tmp->tm_refcount > 0) { 343 TMPFS_UNLOCK(tmp); 344 return; 345 } 346 TMPFS_UNLOCK(tmp); 347 348 uma_zdestroy(tmp->tm_dirent_pool); 349 uma_zdestroy(tmp->tm_node_pool); 350 clear_unrhdr(tmp->tm_ino_unr); 351 delete_unrhdr(tmp->tm_ino_unr); 352 353 mtx_destroy(&tmp->tm_allnode_lock); 354 MPASS(tmp->tm_pages_used == 0); 355 MPASS(tmp->tm_nodes_inuse == 0); 356 357 free(tmp, M_TMPFSMNT); 358 } 359 360 static int 361 tmpfs_root(struct mount *mp, int flags, struct vnode **vpp) 362 { 363 int error; 364 365 error = tmpfs_alloc_vp(mp, VFS_TO_TMPFS(mp)->tm_root, flags, vpp); 366 if (error == 0) 367 (*vpp)->v_vflag |= VV_ROOT; 368 return (error); 369 } 370 371 static int 372 tmpfs_fhtovp(struct mount *mp, struct fid *fhp, int flags, 373 struct vnode **vpp) 374 { 375 struct tmpfs_fid *tfhp; 376 struct tmpfs_mount *tmp; 377 struct tmpfs_node *node; 378 int error; 379 380 tmp = VFS_TO_TMPFS(mp); 381 382 tfhp = (struct tmpfs_fid *)fhp; 383 if (tfhp->tf_len != sizeof(struct tmpfs_fid)) 384 return (EINVAL); 385 386 if (tfhp->tf_id >= tmp->tm_nodes_max) 387 return (EINVAL); 388 389 TMPFS_LOCK(tmp); 390 LIST_FOREACH(node, &tmp->tm_nodes_used, tn_entries) { 391 if (node->tn_id == tfhp->tf_id && 392 node->tn_gen == tfhp->tf_gen) { 393 tmpfs_ref_node(node); 394 break; 395 } 396 } 397 TMPFS_UNLOCK(tmp); 398 399 if (node != NULL) { 400 error = tmpfs_alloc_vp(mp, node, LK_EXCLUSIVE, vpp); 401 tmpfs_free_node(tmp, node); 402 } else 403 error = EINVAL; 404 return (error); 405 } 406 407 /* ARGSUSED2 */ 408 static int 409 tmpfs_statfs(struct mount *mp, struct statfs *sbp) 410 { 411 struct tmpfs_mount *tmp; 412 size_t used; 413 414 tmp = VFS_TO_TMPFS(mp); 415 416 sbp->f_iosize = PAGE_SIZE; 417 sbp->f_bsize = PAGE_SIZE; 418 419 used = tmpfs_pages_used(tmp); 420 if (tmp->tm_pages_max != ULONG_MAX) 421 sbp->f_blocks = tmp->tm_pages_max; 422 else 423 sbp->f_blocks = used + tmpfs_mem_avail(); 424 if (sbp->f_blocks <= used) 425 sbp->f_bavail = 0; 426 else 427 sbp->f_bavail = sbp->f_blocks - used; 428 sbp->f_bfree = sbp->f_bavail; 429 used = tmp->tm_nodes_inuse; 430 sbp->f_files = tmp->tm_nodes_max; 431 if (sbp->f_files <= used) 432 sbp->f_ffree = 0; 433 else 434 sbp->f_ffree = sbp->f_files - used; 435 /* sbp->f_owner = tmp->tn_uid; */ 436 437 return 0; 438 } 439 440 static int 441 tmpfs_sync(struct mount *mp, int waitfor) 442 { 443 struct vnode *vp, *mvp; 444 struct vm_object *obj; 445 446 if (waitfor == MNT_SUSPEND) { 447 MNT_ILOCK(mp); 448 mp->mnt_kern_flag |= MNTK_SUSPEND2 | MNTK_SUSPENDED; 449 MNT_IUNLOCK(mp); 450 } else if (waitfor == MNT_LAZY) { 451 /* 452 * Handle lazy updates of mtime from writes to mmaped 453 * regions. Use MNT_VNODE_FOREACH_ALL instead of 454 * MNT_VNODE_FOREACH_ACTIVE, since unmap of the 455 * tmpfs-backed vnode does not call vinactive(), due 456 * to vm object type is OBJT_SWAP. 457 */ 458 MNT_VNODE_FOREACH_ALL(vp, mp, mvp) { 459 if (vp->v_type != VREG) { 460 VI_UNLOCK(vp); 461 continue; 462 } 463 obj = vp->v_object; 464 KASSERT((obj->flags & (OBJ_TMPFS_NODE | OBJ_TMPFS)) == 465 (OBJ_TMPFS_NODE | OBJ_TMPFS), ("non-tmpfs obj")); 466 467 /* 468 * Unlocked read, avoid taking vnode lock if 469 * not needed. Lost update will be handled on 470 * the next call. 471 */ 472 if ((obj->flags & OBJ_TMPFS_DIRTY) == 0) { 473 VI_UNLOCK(vp); 474 continue; 475 } 476 if (vget(vp, LK_EXCLUSIVE | LK_RETRY | LK_INTERLOCK, 477 curthread) != 0) 478 continue; 479 tmpfs_check_mtime(vp); 480 vput(vp); 481 } 482 } 483 return (0); 484 } 485 486 /* 487 * The presence of a susp_clean method tells the VFS to track writes. 488 */ 489 static void 490 tmpfs_susp_clean(struct mount *mp __unused) 491 { 492 } 493 494 /* 495 * tmpfs vfs operations. 496 */ 497 498 struct vfsops tmpfs_vfsops = { 499 .vfs_mount = tmpfs_mount, 500 .vfs_unmount = tmpfs_unmount, 501 .vfs_root = tmpfs_root, 502 .vfs_statfs = tmpfs_statfs, 503 .vfs_fhtovp = tmpfs_fhtovp, 504 .vfs_sync = tmpfs_sync, 505 .vfs_susp_clean = tmpfs_susp_clean, 506 }; 507 VFS_SET(tmpfs_vfsops, tmpfs, VFCF_JAIL); 508