1 /* $NetBSD: tmpfs_subr.c,v 1.35 2007/07/09 21:10:50 ad 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 supporting functions. 35 */ 36 #include <sys/cdefs.h> 37 __FBSDID("$FreeBSD$"); 38 39 #include <sys/param.h> 40 #include <sys/fnv_hash.h> 41 #include <sys/lock.h> 42 #include <sys/namei.h> 43 #include <sys/priv.h> 44 #include <sys/proc.h> 45 #include <sys/rwlock.h> 46 #include <sys/stat.h> 47 #include <sys/systm.h> 48 #include <sys/sysctl.h> 49 #include <sys/vnode.h> 50 #include <sys/vmmeter.h> 51 52 #include <vm/vm.h> 53 #include <vm/vm_param.h> 54 #include <vm/vm_object.h> 55 #include <vm/vm_page.h> 56 #include <vm/vm_pageout.h> 57 #include <vm/vm_pager.h> 58 #include <vm/vm_extern.h> 59 60 #include <fs/tmpfs/tmpfs.h> 61 #include <fs/tmpfs/tmpfs_fifoops.h> 62 #include <fs/tmpfs/tmpfs_vnops.h> 63 64 struct tmpfs_dir_cursor { 65 struct tmpfs_dirent *tdc_current; 66 struct tmpfs_dirent *tdc_tree; 67 }; 68 69 SYSCTL_NODE(_vfs, OID_AUTO, tmpfs, CTLFLAG_RW, 0, "tmpfs file system"); 70 71 static long tmpfs_pages_reserved = TMPFS_PAGES_MINRESERVED; 72 73 static int 74 sysctl_mem_reserved(SYSCTL_HANDLER_ARGS) 75 { 76 int error; 77 long pages, bytes; 78 79 pages = *(long *)arg1; 80 bytes = pages * PAGE_SIZE; 81 82 error = sysctl_handle_long(oidp, &bytes, 0, req); 83 if (error || !req->newptr) 84 return (error); 85 86 pages = bytes / PAGE_SIZE; 87 if (pages < TMPFS_PAGES_MINRESERVED) 88 return (EINVAL); 89 90 *(long *)arg1 = pages; 91 return (0); 92 } 93 94 SYSCTL_PROC(_vfs_tmpfs, OID_AUTO, memory_reserved, CTLTYPE_LONG|CTLFLAG_RW, 95 &tmpfs_pages_reserved, 0, sysctl_mem_reserved, "L", 96 "Amount of available memory and swap below which tmpfs growth stops"); 97 98 static __inline int tmpfs_dirtree_cmp(struct tmpfs_dirent *a, 99 struct tmpfs_dirent *b); 100 RB_PROTOTYPE_STATIC(tmpfs_dir, tmpfs_dirent, uh.td_entries, tmpfs_dirtree_cmp); 101 102 size_t 103 tmpfs_mem_avail(void) 104 { 105 vm_ooffset_t avail; 106 107 avail = swap_pager_avail + vm_cnt.v_free_count + vm_cnt.v_cache_count - 108 tmpfs_pages_reserved; 109 if (__predict_false(avail < 0)) 110 avail = 0; 111 return (avail); 112 } 113 114 size_t 115 tmpfs_pages_used(struct tmpfs_mount *tmp) 116 { 117 const size_t node_size = sizeof(struct tmpfs_node) + 118 sizeof(struct tmpfs_dirent); 119 size_t meta_pages; 120 121 meta_pages = howmany((uintmax_t)tmp->tm_nodes_inuse * node_size, 122 PAGE_SIZE); 123 return (meta_pages + tmp->tm_pages_used); 124 } 125 126 static size_t 127 tmpfs_pages_check_avail(struct tmpfs_mount *tmp, size_t req_pages) 128 { 129 if (tmpfs_mem_avail() < req_pages) 130 return (0); 131 132 if (tmp->tm_pages_max != SIZE_MAX && 133 tmp->tm_pages_max < req_pages + tmpfs_pages_used(tmp)) 134 return (0); 135 136 return (1); 137 } 138 139 /* 140 * Allocates a new node of type 'type' inside the 'tmp' mount point, with 141 * its owner set to 'uid', its group to 'gid' and its mode set to 'mode', 142 * using the credentials of the process 'p'. 143 * 144 * If the node type is set to 'VDIR', then the parent parameter must point 145 * to the parent directory of the node being created. It may only be NULL 146 * while allocating the root node. 147 * 148 * If the node type is set to 'VBLK' or 'VCHR', then the rdev parameter 149 * specifies the device the node represents. 150 * 151 * If the node type is set to 'VLNK', then the parameter target specifies 152 * the file name of the target file for the symbolic link that is being 153 * created. 154 * 155 * Note that new nodes are retrieved from the available list if it has 156 * items or, if it is empty, from the node pool as long as there is enough 157 * space to create them. 158 * 159 * Returns zero on success or an appropriate error code on failure. 160 */ 161 int 162 tmpfs_alloc_node(struct mount *mp, struct tmpfs_mount *tmp, enum vtype type, 163 uid_t uid, gid_t gid, mode_t mode, struct tmpfs_node *parent, 164 char *target, dev_t rdev, struct tmpfs_node **node) 165 { 166 struct tmpfs_node *nnode; 167 vm_object_t obj; 168 169 /* If the root directory of the 'tmp' file system is not yet 170 * allocated, this must be the request to do it. */ 171 MPASS(IMPLIES(tmp->tm_root == NULL, parent == NULL && type == VDIR)); 172 KASSERT(tmp->tm_root == NULL || mp->mnt_writeopcount > 0, 173 ("creating node not under vn_start_write")); 174 175 MPASS(IFF(type == VLNK, target != NULL)); 176 MPASS(IFF(type == VBLK || type == VCHR, rdev != VNOVAL)); 177 178 if (tmp->tm_nodes_inuse >= tmp->tm_nodes_max) 179 return (ENOSPC); 180 if (tmpfs_pages_check_avail(tmp, 1) == 0) 181 return (ENOSPC); 182 183 if ((mp->mnt_kern_flag & MNTK_UNMOUNT) != 0) { 184 /* 185 * When a new tmpfs node is created for fully 186 * constructed mount point, there must be a parent 187 * node, which vnode is locked exclusively. As 188 * consequence, if the unmount is executing in 189 * parallel, vflush() cannot reclaim the parent vnode. 190 * Due to this, the check for MNTK_UNMOUNT flag is not 191 * racy: if we did not see MNTK_UNMOUNT flag, then tmp 192 * cannot be destroyed until node construction is 193 * finished and the parent vnode unlocked. 194 * 195 * Tmpfs does not need to instantiate new nodes during 196 * unmount. 197 */ 198 return (EBUSY); 199 } 200 201 nnode = (struct tmpfs_node *)uma_zalloc_arg( 202 tmp->tm_node_pool, tmp, M_WAITOK); 203 204 /* Generic initialization. */ 205 nnode->tn_type = type; 206 vfs_timestamp(&nnode->tn_atime); 207 nnode->tn_birthtime = nnode->tn_ctime = nnode->tn_mtime = 208 nnode->tn_atime; 209 nnode->tn_uid = uid; 210 nnode->tn_gid = gid; 211 nnode->tn_mode = mode; 212 nnode->tn_id = alloc_unr(tmp->tm_ino_unr); 213 214 /* Type-specific initialization. */ 215 switch (nnode->tn_type) { 216 case VBLK: 217 case VCHR: 218 nnode->tn_rdev = rdev; 219 break; 220 221 case VDIR: 222 RB_INIT(&nnode->tn_dir.tn_dirhead); 223 LIST_INIT(&nnode->tn_dir.tn_dupindex); 224 MPASS(parent != nnode); 225 MPASS(IMPLIES(parent == NULL, tmp->tm_root == NULL)); 226 nnode->tn_dir.tn_parent = (parent == NULL) ? nnode : parent; 227 nnode->tn_dir.tn_readdir_lastn = 0; 228 nnode->tn_dir.tn_readdir_lastp = NULL; 229 nnode->tn_links++; 230 TMPFS_NODE_LOCK(nnode->tn_dir.tn_parent); 231 nnode->tn_dir.tn_parent->tn_links++; 232 TMPFS_NODE_UNLOCK(nnode->tn_dir.tn_parent); 233 break; 234 235 case VFIFO: 236 /* FALLTHROUGH */ 237 case VSOCK: 238 break; 239 240 case VLNK: 241 MPASS(strlen(target) < MAXPATHLEN); 242 nnode->tn_size = strlen(target); 243 nnode->tn_link = malloc(nnode->tn_size, M_TMPFSNAME, 244 M_WAITOK); 245 memcpy(nnode->tn_link, target, nnode->tn_size); 246 break; 247 248 case VREG: 249 obj = nnode->tn_reg.tn_aobj = 250 vm_pager_allocate(OBJT_SWAP, NULL, 0, VM_PROT_DEFAULT, 0, 251 NULL /* XXXKIB - tmpfs needs swap reservation */); 252 VM_OBJECT_WLOCK(obj); 253 /* OBJ_TMPFS is set together with the setting of vp->v_object */ 254 vm_object_set_flag(obj, OBJ_NOSPLIT | OBJ_TMPFS_NODE); 255 vm_object_clear_flag(obj, OBJ_ONEMAPPING); 256 VM_OBJECT_WUNLOCK(obj); 257 break; 258 259 default: 260 panic("tmpfs_alloc_node: type %p %d", nnode, (int)nnode->tn_type); 261 } 262 263 TMPFS_LOCK(tmp); 264 LIST_INSERT_HEAD(&tmp->tm_nodes_used, nnode, tn_entries); 265 tmp->tm_nodes_inuse++; 266 TMPFS_UNLOCK(tmp); 267 268 *node = nnode; 269 return 0; 270 } 271 272 /* 273 * Destroys the node pointed to by node from the file system 'tmp'. 274 * If the node does not belong to the given mount point, the results are 275 * unpredicted. 276 * 277 * If the node references a directory; no entries are allowed because 278 * their removal could need a recursive algorithm, something forbidden in 279 * kernel space. Furthermore, there is not need to provide such 280 * functionality (recursive removal) because the only primitives offered 281 * to the user are the removal of empty directories and the deletion of 282 * individual files. 283 * 284 * Note that nodes are not really deleted; in fact, when a node has been 285 * allocated, it cannot be deleted during the whole life of the file 286 * system. Instead, they are moved to the available list and remain there 287 * until reused. 288 */ 289 void 290 tmpfs_free_node(struct tmpfs_mount *tmp, struct tmpfs_node *node) 291 { 292 vm_object_t uobj; 293 294 #ifdef INVARIANTS 295 TMPFS_NODE_LOCK(node); 296 MPASS(node->tn_vnode == NULL); 297 MPASS((node->tn_vpstate & TMPFS_VNODE_ALLOCATING) == 0); 298 TMPFS_NODE_UNLOCK(node); 299 #endif 300 301 TMPFS_LOCK(tmp); 302 LIST_REMOVE(node, tn_entries); 303 tmp->tm_nodes_inuse--; 304 TMPFS_UNLOCK(tmp); 305 306 switch (node->tn_type) { 307 case VNON: 308 /* Do not do anything. VNON is provided to let the 309 * allocation routine clean itself easily by avoiding 310 * duplicating code in it. */ 311 /* FALLTHROUGH */ 312 case VBLK: 313 /* FALLTHROUGH */ 314 case VCHR: 315 /* FALLTHROUGH */ 316 case VDIR: 317 /* FALLTHROUGH */ 318 case VFIFO: 319 /* FALLTHROUGH */ 320 case VSOCK: 321 break; 322 323 case VLNK: 324 free(node->tn_link, M_TMPFSNAME); 325 break; 326 327 case VREG: 328 uobj = node->tn_reg.tn_aobj; 329 if (uobj != NULL) { 330 TMPFS_LOCK(tmp); 331 tmp->tm_pages_used -= uobj->size; 332 TMPFS_UNLOCK(tmp); 333 KASSERT((uobj->flags & OBJ_TMPFS) == 0, 334 ("leaked OBJ_TMPFS node %p vm_obj %p", node, uobj)); 335 vm_object_deallocate(uobj); 336 } 337 break; 338 339 default: 340 panic("tmpfs_free_node: type %p %d", node, (int)node->tn_type); 341 } 342 343 free_unr(tmp->tm_ino_unr, node->tn_id); 344 uma_zfree(tmp->tm_node_pool, node); 345 } 346 347 static __inline uint32_t 348 tmpfs_dirent_hash(const char *name, u_int len) 349 { 350 uint32_t hash; 351 352 hash = fnv_32_buf(name, len, FNV1_32_INIT + len) & TMPFS_DIRCOOKIE_MASK; 353 #ifdef TMPFS_DEBUG_DIRCOOKIE_DUP 354 hash &= 0xf; 355 #endif 356 if (hash < TMPFS_DIRCOOKIE_MIN) 357 hash += TMPFS_DIRCOOKIE_MIN; 358 359 return (hash); 360 } 361 362 static __inline off_t 363 tmpfs_dirent_cookie(struct tmpfs_dirent *de) 364 { 365 if (de == NULL) 366 return (TMPFS_DIRCOOKIE_EOF); 367 368 MPASS(de->td_cookie >= TMPFS_DIRCOOKIE_MIN); 369 370 return (de->td_cookie); 371 } 372 373 static __inline boolean_t 374 tmpfs_dirent_dup(struct tmpfs_dirent *de) 375 { 376 return ((de->td_cookie & TMPFS_DIRCOOKIE_DUP) != 0); 377 } 378 379 static __inline boolean_t 380 tmpfs_dirent_duphead(struct tmpfs_dirent *de) 381 { 382 return ((de->td_cookie & TMPFS_DIRCOOKIE_DUPHEAD) != 0); 383 } 384 385 void 386 tmpfs_dirent_init(struct tmpfs_dirent *de, const char *name, u_int namelen) 387 { 388 de->td_hash = de->td_cookie = tmpfs_dirent_hash(name, namelen); 389 memcpy(de->ud.td_name, name, namelen); 390 de->td_namelen = namelen; 391 } 392 393 /* 394 * Allocates a new directory entry for the node node with a name of name. 395 * The new directory entry is returned in *de. 396 * 397 * The link count of node is increased by one to reflect the new object 398 * referencing it. 399 * 400 * Returns zero on success or an appropriate error code on failure. 401 */ 402 int 403 tmpfs_alloc_dirent(struct tmpfs_mount *tmp, struct tmpfs_node *node, 404 const char *name, u_int len, struct tmpfs_dirent **de) 405 { 406 struct tmpfs_dirent *nde; 407 408 nde = uma_zalloc(tmp->tm_dirent_pool, M_WAITOK); 409 nde->td_node = node; 410 if (name != NULL) { 411 nde->ud.td_name = malloc(len, M_TMPFSNAME, M_WAITOK); 412 tmpfs_dirent_init(nde, name, len); 413 } else 414 nde->td_namelen = 0; 415 if (node != NULL) 416 node->tn_links++; 417 418 *de = nde; 419 420 return 0; 421 } 422 423 /* 424 * Frees a directory entry. It is the caller's responsibility to destroy 425 * the node referenced by it if needed. 426 * 427 * The link count of node is decreased by one to reflect the removal of an 428 * object that referenced it. This only happens if 'node_exists' is true; 429 * otherwise the function will not access the node referred to by the 430 * directory entry, as it may already have been released from the outside. 431 */ 432 void 433 tmpfs_free_dirent(struct tmpfs_mount *tmp, struct tmpfs_dirent *de) 434 { 435 struct tmpfs_node *node; 436 437 node = de->td_node; 438 if (node != NULL) { 439 MPASS(node->tn_links > 0); 440 node->tn_links--; 441 } 442 if (!tmpfs_dirent_duphead(de) && de->ud.td_name != NULL) 443 free(de->ud.td_name, M_TMPFSNAME); 444 uma_zfree(tmp->tm_dirent_pool, de); 445 } 446 447 void 448 tmpfs_destroy_vobject(struct vnode *vp, vm_object_t obj) 449 { 450 451 ASSERT_VOP_ELOCKED(vp, "tmpfs_destroy_vobject"); 452 if (vp->v_type != VREG || obj == NULL) 453 return; 454 455 VM_OBJECT_WLOCK(obj); 456 VI_LOCK(vp); 457 vm_object_clear_flag(obj, OBJ_TMPFS); 458 obj->un_pager.swp.swp_tmpfs = NULL; 459 VI_UNLOCK(vp); 460 VM_OBJECT_WUNLOCK(obj); 461 } 462 463 /* 464 * Need to clear v_object for insmntque failure. 465 */ 466 static void 467 tmpfs_insmntque_dtr(struct vnode *vp, void *dtr_arg) 468 { 469 470 tmpfs_destroy_vobject(vp, vp->v_object); 471 vp->v_object = NULL; 472 vp->v_data = NULL; 473 vp->v_op = &dead_vnodeops; 474 vgone(vp); 475 vput(vp); 476 } 477 478 /* 479 * Allocates a new vnode for the node node or returns a new reference to 480 * an existing one if the node had already a vnode referencing it. The 481 * resulting locked vnode is returned in *vpp. 482 * 483 * Returns zero on success or an appropriate error code on failure. 484 */ 485 int 486 tmpfs_alloc_vp(struct mount *mp, struct tmpfs_node *node, int lkflag, 487 struct vnode **vpp) 488 { 489 struct vnode *vp; 490 vm_object_t object; 491 int error; 492 493 error = 0; 494 loop: 495 TMPFS_NODE_LOCK(node); 496 loop1: 497 if ((vp = node->tn_vnode) != NULL) { 498 MPASS((node->tn_vpstate & TMPFS_VNODE_DOOMED) == 0); 499 VI_LOCK(vp); 500 if ((node->tn_type == VDIR && node->tn_dir.tn_parent == NULL) || 501 ((vp->v_iflag & VI_DOOMED) != 0 && 502 (lkflag & LK_NOWAIT) != 0)) { 503 VI_UNLOCK(vp); 504 TMPFS_NODE_UNLOCK(node); 505 error = ENOENT; 506 vp = NULL; 507 goto out; 508 } 509 if ((vp->v_iflag & VI_DOOMED) != 0) { 510 VI_UNLOCK(vp); 511 node->tn_vpstate |= TMPFS_VNODE_WRECLAIM; 512 while ((node->tn_vpstate & TMPFS_VNODE_WRECLAIM) != 0) { 513 msleep(&node->tn_vnode, TMPFS_NODE_MTX(node), 514 0, "tmpfsE", 0); 515 } 516 goto loop1; 517 } 518 TMPFS_NODE_UNLOCK(node); 519 error = vget(vp, lkflag | LK_INTERLOCK, curthread); 520 if (error == ENOENT) 521 goto loop; 522 if (error != 0) { 523 vp = NULL; 524 goto out; 525 } 526 527 /* 528 * Make sure the vnode is still there after 529 * getting the interlock to avoid racing a free. 530 */ 531 if (node->tn_vnode == NULL || node->tn_vnode != vp) { 532 vput(vp); 533 goto loop; 534 } 535 536 goto out; 537 } 538 539 if ((node->tn_vpstate & TMPFS_VNODE_DOOMED) || 540 (node->tn_type == VDIR && node->tn_dir.tn_parent == NULL)) { 541 TMPFS_NODE_UNLOCK(node); 542 error = ENOENT; 543 vp = NULL; 544 goto out; 545 } 546 547 /* 548 * otherwise lock the vp list while we call getnewvnode 549 * since that can block. 550 */ 551 if (node->tn_vpstate & TMPFS_VNODE_ALLOCATING) { 552 node->tn_vpstate |= TMPFS_VNODE_WANT; 553 error = msleep((caddr_t) &node->tn_vpstate, 554 TMPFS_NODE_MTX(node), PDROP | PCATCH, 555 "tmpfs_alloc_vp", 0); 556 if (error) 557 return error; 558 559 goto loop; 560 } else 561 node->tn_vpstate |= TMPFS_VNODE_ALLOCATING; 562 563 TMPFS_NODE_UNLOCK(node); 564 565 /* Get a new vnode and associate it with our node. */ 566 error = getnewvnode("tmpfs", mp, &tmpfs_vnodeop_entries, &vp); 567 if (error != 0) 568 goto unlock; 569 MPASS(vp != NULL); 570 571 /* lkflag is ignored, the lock is exclusive */ 572 (void) vn_lock(vp, lkflag | LK_RETRY); 573 574 vp->v_data = node; 575 vp->v_type = node->tn_type; 576 577 /* Type-specific initialization. */ 578 switch (node->tn_type) { 579 case VBLK: 580 /* FALLTHROUGH */ 581 case VCHR: 582 /* FALLTHROUGH */ 583 case VLNK: 584 /* FALLTHROUGH */ 585 case VSOCK: 586 break; 587 case VFIFO: 588 vp->v_op = &tmpfs_fifoop_entries; 589 break; 590 case VREG: 591 object = node->tn_reg.tn_aobj; 592 VM_OBJECT_WLOCK(object); 593 VI_LOCK(vp); 594 KASSERT(vp->v_object == NULL, ("Not NULL v_object in tmpfs")); 595 vp->v_object = object; 596 object->un_pager.swp.swp_tmpfs = vp; 597 vm_object_set_flag(object, OBJ_TMPFS); 598 VI_UNLOCK(vp); 599 VM_OBJECT_WUNLOCK(object); 600 break; 601 case VDIR: 602 MPASS(node->tn_dir.tn_parent != NULL); 603 if (node->tn_dir.tn_parent == node) 604 vp->v_vflag |= VV_ROOT; 605 break; 606 607 default: 608 panic("tmpfs_alloc_vp: type %p %d", node, (int)node->tn_type); 609 } 610 if (vp->v_type != VFIFO) 611 VN_LOCK_ASHARE(vp); 612 613 error = insmntque1(vp, mp, tmpfs_insmntque_dtr, NULL); 614 if (error) 615 vp = NULL; 616 617 unlock: 618 TMPFS_NODE_LOCK(node); 619 620 MPASS(node->tn_vpstate & TMPFS_VNODE_ALLOCATING); 621 node->tn_vpstate &= ~TMPFS_VNODE_ALLOCATING; 622 node->tn_vnode = vp; 623 624 if (node->tn_vpstate & TMPFS_VNODE_WANT) { 625 node->tn_vpstate &= ~TMPFS_VNODE_WANT; 626 TMPFS_NODE_UNLOCK(node); 627 wakeup((caddr_t) &node->tn_vpstate); 628 } else 629 TMPFS_NODE_UNLOCK(node); 630 631 out: 632 *vpp = vp; 633 634 #ifdef INVARIANTS 635 if (error == 0) { 636 MPASS(*vpp != NULL && VOP_ISLOCKED(*vpp)); 637 TMPFS_NODE_LOCK(node); 638 MPASS(*vpp == node->tn_vnode); 639 TMPFS_NODE_UNLOCK(node); 640 } 641 #endif 642 643 return error; 644 } 645 646 /* 647 * Destroys the association between the vnode vp and the node it 648 * references. 649 */ 650 void 651 tmpfs_free_vp(struct vnode *vp) 652 { 653 struct tmpfs_node *node; 654 655 node = VP_TO_TMPFS_NODE(vp); 656 657 TMPFS_NODE_ASSERT_LOCKED(node); 658 node->tn_vnode = NULL; 659 if ((node->tn_vpstate & TMPFS_VNODE_WRECLAIM) != 0) 660 wakeup(&node->tn_vnode); 661 node->tn_vpstate &= ~TMPFS_VNODE_WRECLAIM; 662 vp->v_data = NULL; 663 } 664 665 /* 666 * Allocates a new file of type 'type' and adds it to the parent directory 667 * 'dvp'; this addition is done using the component name given in 'cnp'. 668 * The ownership of the new file is automatically assigned based on the 669 * credentials of the caller (through 'cnp'), the group is set based on 670 * the parent directory and the mode is determined from the 'vap' argument. 671 * If successful, *vpp holds a vnode to the newly created file and zero 672 * is returned. Otherwise *vpp is NULL and the function returns an 673 * appropriate error code. 674 */ 675 int 676 tmpfs_alloc_file(struct vnode *dvp, struct vnode **vpp, struct vattr *vap, 677 struct componentname *cnp, char *target) 678 { 679 int error; 680 struct tmpfs_dirent *de; 681 struct tmpfs_mount *tmp; 682 struct tmpfs_node *dnode; 683 struct tmpfs_node *node; 684 struct tmpfs_node *parent; 685 686 MPASS(VOP_ISLOCKED(dvp)); 687 MPASS(cnp->cn_flags & HASBUF); 688 689 tmp = VFS_TO_TMPFS(dvp->v_mount); 690 dnode = VP_TO_TMPFS_DIR(dvp); 691 *vpp = NULL; 692 693 /* If the entry we are creating is a directory, we cannot overflow 694 * the number of links of its parent, because it will get a new 695 * link. */ 696 if (vap->va_type == VDIR) { 697 /* Ensure that we do not overflow the maximum number of links 698 * imposed by the system. */ 699 MPASS(dnode->tn_links <= LINK_MAX); 700 if (dnode->tn_links == LINK_MAX) { 701 return (EMLINK); 702 } 703 704 parent = dnode; 705 MPASS(parent != NULL); 706 } else 707 parent = NULL; 708 709 /* Allocate a node that represents the new file. */ 710 error = tmpfs_alloc_node(dvp->v_mount, tmp, vap->va_type, 711 cnp->cn_cred->cr_uid, 712 dnode->tn_gid, vap->va_mode, parent, target, vap->va_rdev, &node); 713 if (error != 0) 714 return (error); 715 716 /* Allocate a directory entry that points to the new file. */ 717 error = tmpfs_alloc_dirent(tmp, node, cnp->cn_nameptr, cnp->cn_namelen, 718 &de); 719 if (error != 0) { 720 tmpfs_free_node(tmp, node); 721 return (error); 722 } 723 724 /* Allocate a vnode for the new file. */ 725 error = tmpfs_alloc_vp(dvp->v_mount, node, LK_EXCLUSIVE, vpp); 726 if (error != 0) { 727 tmpfs_free_dirent(tmp, de); 728 tmpfs_free_node(tmp, node); 729 return (error); 730 } 731 732 /* Now that all required items are allocated, we can proceed to 733 * insert the new node into the directory, an operation that 734 * cannot fail. */ 735 if (cnp->cn_flags & ISWHITEOUT) 736 tmpfs_dir_whiteout_remove(dvp, cnp); 737 tmpfs_dir_attach(dvp, de); 738 return (0); 739 } 740 741 static struct tmpfs_dirent * 742 tmpfs_dir_first(struct tmpfs_node *dnode, struct tmpfs_dir_cursor *dc) 743 { 744 struct tmpfs_dirent *de; 745 746 de = RB_MIN(tmpfs_dir, &dnode->tn_dir.tn_dirhead); 747 dc->tdc_tree = de; 748 if (de != NULL && tmpfs_dirent_duphead(de)) 749 de = LIST_FIRST(&de->ud.td_duphead); 750 dc->tdc_current = de; 751 752 return (dc->tdc_current); 753 } 754 755 static struct tmpfs_dirent * 756 tmpfs_dir_next(struct tmpfs_node *dnode, struct tmpfs_dir_cursor *dc) 757 { 758 struct tmpfs_dirent *de; 759 760 MPASS(dc->tdc_tree != NULL); 761 if (tmpfs_dirent_dup(dc->tdc_current)) { 762 dc->tdc_current = LIST_NEXT(dc->tdc_current, uh.td_dup.entries); 763 if (dc->tdc_current != NULL) 764 return (dc->tdc_current); 765 } 766 dc->tdc_tree = dc->tdc_current = RB_NEXT(tmpfs_dir, 767 &dnode->tn_dir.tn_dirhead, dc->tdc_tree); 768 if ((de = dc->tdc_current) != NULL && tmpfs_dirent_duphead(de)) { 769 dc->tdc_current = LIST_FIRST(&de->ud.td_duphead); 770 MPASS(dc->tdc_current != NULL); 771 } 772 773 return (dc->tdc_current); 774 } 775 776 /* Lookup directory entry in RB-Tree. Function may return duphead entry. */ 777 static struct tmpfs_dirent * 778 tmpfs_dir_xlookup_hash(struct tmpfs_node *dnode, uint32_t hash) 779 { 780 struct tmpfs_dirent *de, dekey; 781 782 dekey.td_hash = hash; 783 de = RB_FIND(tmpfs_dir, &dnode->tn_dir.tn_dirhead, &dekey); 784 return (de); 785 } 786 787 /* Lookup directory entry by cookie, initialize directory cursor accordingly. */ 788 static struct tmpfs_dirent * 789 tmpfs_dir_lookup_cookie(struct tmpfs_node *node, off_t cookie, 790 struct tmpfs_dir_cursor *dc) 791 { 792 struct tmpfs_dir *dirhead = &node->tn_dir.tn_dirhead; 793 struct tmpfs_dirent *de, dekey; 794 795 MPASS(cookie >= TMPFS_DIRCOOKIE_MIN); 796 797 if (cookie == node->tn_dir.tn_readdir_lastn && 798 (de = node->tn_dir.tn_readdir_lastp) != NULL) { 799 /* Protect against possible race, tn_readdir_last[pn] 800 * may be updated with only shared vnode lock held. */ 801 if (cookie == tmpfs_dirent_cookie(de)) 802 goto out; 803 } 804 805 if ((cookie & TMPFS_DIRCOOKIE_DUP) != 0) { 806 LIST_FOREACH(de, &node->tn_dir.tn_dupindex, 807 uh.td_dup.index_entries) { 808 MPASS(tmpfs_dirent_dup(de)); 809 if (de->td_cookie == cookie) 810 goto out; 811 /* dupindex list is sorted. */ 812 if (de->td_cookie < cookie) { 813 de = NULL; 814 goto out; 815 } 816 } 817 MPASS(de == NULL); 818 goto out; 819 } 820 821 MPASS((cookie & TMPFS_DIRCOOKIE_MASK) == cookie); 822 dekey.td_hash = cookie; 823 /* Recover if direntry for cookie was removed */ 824 de = RB_NFIND(tmpfs_dir, dirhead, &dekey); 825 dc->tdc_tree = de; 826 dc->tdc_current = de; 827 if (de != NULL && tmpfs_dirent_duphead(de)) { 828 dc->tdc_current = LIST_FIRST(&de->ud.td_duphead); 829 MPASS(dc->tdc_current != NULL); 830 } 831 return (dc->tdc_current); 832 833 out: 834 dc->tdc_tree = de; 835 dc->tdc_current = de; 836 if (de != NULL && tmpfs_dirent_dup(de)) 837 dc->tdc_tree = tmpfs_dir_xlookup_hash(node, 838 de->td_hash); 839 return (dc->tdc_current); 840 } 841 842 /* 843 * Looks for a directory entry in the directory represented by node. 844 * 'cnp' describes the name of the entry to look for. Note that the . 845 * and .. components are not allowed as they do not physically exist 846 * within directories. 847 * 848 * Returns a pointer to the entry when found, otherwise NULL. 849 */ 850 struct tmpfs_dirent * 851 tmpfs_dir_lookup(struct tmpfs_node *node, struct tmpfs_node *f, 852 struct componentname *cnp) 853 { 854 struct tmpfs_dir_duphead *duphead; 855 struct tmpfs_dirent *de; 856 uint32_t hash; 857 858 MPASS(IMPLIES(cnp->cn_namelen == 1, cnp->cn_nameptr[0] != '.')); 859 MPASS(IMPLIES(cnp->cn_namelen == 2, !(cnp->cn_nameptr[0] == '.' && 860 cnp->cn_nameptr[1] == '.'))); 861 TMPFS_VALIDATE_DIR(node); 862 863 hash = tmpfs_dirent_hash(cnp->cn_nameptr, cnp->cn_namelen); 864 de = tmpfs_dir_xlookup_hash(node, hash); 865 if (de != NULL && tmpfs_dirent_duphead(de)) { 866 duphead = &de->ud.td_duphead; 867 LIST_FOREACH(de, duphead, uh.td_dup.entries) { 868 if (TMPFS_DIRENT_MATCHES(de, cnp->cn_nameptr, 869 cnp->cn_namelen)) 870 break; 871 } 872 } else if (de != NULL) { 873 if (!TMPFS_DIRENT_MATCHES(de, cnp->cn_nameptr, 874 cnp->cn_namelen)) 875 de = NULL; 876 } 877 if (de != NULL && f != NULL && de->td_node != f) 878 de = NULL; 879 880 return (de); 881 } 882 883 /* 884 * Attach duplicate-cookie directory entry nde to dnode and insert to dupindex 885 * list, allocate new cookie value. 886 */ 887 static void 888 tmpfs_dir_attach_dup(struct tmpfs_node *dnode, 889 struct tmpfs_dir_duphead *duphead, struct tmpfs_dirent *nde) 890 { 891 struct tmpfs_dir_duphead *dupindex; 892 struct tmpfs_dirent *de, *pde; 893 894 dupindex = &dnode->tn_dir.tn_dupindex; 895 de = LIST_FIRST(dupindex); 896 if (de == NULL || de->td_cookie < TMPFS_DIRCOOKIE_DUP_MAX) { 897 if (de == NULL) 898 nde->td_cookie = TMPFS_DIRCOOKIE_DUP_MIN; 899 else 900 nde->td_cookie = de->td_cookie + 1; 901 MPASS(tmpfs_dirent_dup(nde)); 902 LIST_INSERT_HEAD(dupindex, nde, uh.td_dup.index_entries); 903 LIST_INSERT_HEAD(duphead, nde, uh.td_dup.entries); 904 return; 905 } 906 907 /* 908 * Cookie numbers are near exhaustion. Scan dupindex list for unused 909 * numbers. dupindex list is sorted in descending order. Keep it so 910 * after inserting nde. 911 */ 912 while (1) { 913 pde = de; 914 de = LIST_NEXT(de, uh.td_dup.index_entries); 915 if (de == NULL && pde->td_cookie != TMPFS_DIRCOOKIE_DUP_MIN) { 916 /* 917 * Last element of the index doesn't have minimal cookie 918 * value, use it. 919 */ 920 nde->td_cookie = TMPFS_DIRCOOKIE_DUP_MIN; 921 LIST_INSERT_AFTER(pde, nde, uh.td_dup.index_entries); 922 LIST_INSERT_HEAD(duphead, nde, uh.td_dup.entries); 923 return; 924 } else if (de == NULL) { 925 /* 926 * We are so lucky have 2^30 hash duplicates in single 927 * directory :) Return largest possible cookie value. 928 * It should be fine except possible issues with 929 * VOP_READDIR restart. 930 */ 931 nde->td_cookie = TMPFS_DIRCOOKIE_DUP_MAX; 932 LIST_INSERT_HEAD(dupindex, nde, 933 uh.td_dup.index_entries); 934 LIST_INSERT_HEAD(duphead, nde, uh.td_dup.entries); 935 return; 936 } 937 if (de->td_cookie + 1 == pde->td_cookie || 938 de->td_cookie >= TMPFS_DIRCOOKIE_DUP_MAX) 939 continue; /* No hole or invalid cookie. */ 940 nde->td_cookie = de->td_cookie + 1; 941 MPASS(tmpfs_dirent_dup(nde)); 942 MPASS(pde->td_cookie > nde->td_cookie); 943 MPASS(nde->td_cookie > de->td_cookie); 944 LIST_INSERT_BEFORE(de, nde, uh.td_dup.index_entries); 945 LIST_INSERT_HEAD(duphead, nde, uh.td_dup.entries); 946 return; 947 }; 948 } 949 950 /* 951 * Attaches the directory entry de to the directory represented by vp. 952 * Note that this does not change the link count of the node pointed by 953 * the directory entry, as this is done by tmpfs_alloc_dirent. 954 */ 955 void 956 tmpfs_dir_attach(struct vnode *vp, struct tmpfs_dirent *de) 957 { 958 struct tmpfs_node *dnode; 959 struct tmpfs_dirent *xde, *nde; 960 961 ASSERT_VOP_ELOCKED(vp, __func__); 962 MPASS(de->td_namelen > 0); 963 MPASS(de->td_hash >= TMPFS_DIRCOOKIE_MIN); 964 MPASS(de->td_cookie == de->td_hash); 965 966 dnode = VP_TO_TMPFS_DIR(vp); 967 dnode->tn_dir.tn_readdir_lastn = 0; 968 dnode->tn_dir.tn_readdir_lastp = NULL; 969 970 MPASS(!tmpfs_dirent_dup(de)); 971 xde = RB_INSERT(tmpfs_dir, &dnode->tn_dir.tn_dirhead, de); 972 if (xde != NULL && tmpfs_dirent_duphead(xde)) 973 tmpfs_dir_attach_dup(dnode, &xde->ud.td_duphead, de); 974 else if (xde != NULL) { 975 /* 976 * Allocate new duphead. Swap xde with duphead to avoid 977 * adding/removing elements with the same hash. 978 */ 979 MPASS(!tmpfs_dirent_dup(xde)); 980 tmpfs_alloc_dirent(VFS_TO_TMPFS(vp->v_mount), NULL, NULL, 0, 981 &nde); 982 /* *nde = *xde; XXX gcc 4.2.1 may generate invalid code. */ 983 memcpy(nde, xde, sizeof(*xde)); 984 xde->td_cookie |= TMPFS_DIRCOOKIE_DUPHEAD; 985 LIST_INIT(&xde->ud.td_duphead); 986 xde->td_namelen = 0; 987 xde->td_node = NULL; 988 tmpfs_dir_attach_dup(dnode, &xde->ud.td_duphead, nde); 989 tmpfs_dir_attach_dup(dnode, &xde->ud.td_duphead, de); 990 } 991 dnode->tn_size += sizeof(struct tmpfs_dirent); 992 dnode->tn_status |= TMPFS_NODE_ACCESSED | TMPFS_NODE_CHANGED | \ 993 TMPFS_NODE_MODIFIED; 994 tmpfs_update(vp); 995 } 996 997 /* 998 * Detaches the directory entry de from the directory represented by vp. 999 * Note that this does not change the link count of the node pointed by 1000 * the directory entry, as this is done by tmpfs_free_dirent. 1001 */ 1002 void 1003 tmpfs_dir_detach(struct vnode *vp, struct tmpfs_dirent *de) 1004 { 1005 struct tmpfs_mount *tmp; 1006 struct tmpfs_dir *head; 1007 struct tmpfs_node *dnode; 1008 struct tmpfs_dirent *xde; 1009 1010 ASSERT_VOP_ELOCKED(vp, __func__); 1011 1012 dnode = VP_TO_TMPFS_DIR(vp); 1013 head = &dnode->tn_dir.tn_dirhead; 1014 dnode->tn_dir.tn_readdir_lastn = 0; 1015 dnode->tn_dir.tn_readdir_lastp = NULL; 1016 1017 if (tmpfs_dirent_dup(de)) { 1018 /* Remove duphead if de was last entry. */ 1019 if (LIST_NEXT(de, uh.td_dup.entries) == NULL) { 1020 xde = tmpfs_dir_xlookup_hash(dnode, de->td_hash); 1021 MPASS(tmpfs_dirent_duphead(xde)); 1022 } else 1023 xde = NULL; 1024 LIST_REMOVE(de, uh.td_dup.entries); 1025 LIST_REMOVE(de, uh.td_dup.index_entries); 1026 if (xde != NULL) { 1027 if (LIST_EMPTY(&xde->ud.td_duphead)) { 1028 RB_REMOVE(tmpfs_dir, head, xde); 1029 tmp = VFS_TO_TMPFS(vp->v_mount); 1030 MPASS(xde->td_node == NULL); 1031 tmpfs_free_dirent(tmp, xde); 1032 } 1033 } 1034 de->td_cookie = de->td_hash; 1035 } else 1036 RB_REMOVE(tmpfs_dir, head, de); 1037 1038 dnode->tn_size -= sizeof(struct tmpfs_dirent); 1039 dnode->tn_status |= TMPFS_NODE_ACCESSED | TMPFS_NODE_CHANGED | \ 1040 TMPFS_NODE_MODIFIED; 1041 tmpfs_update(vp); 1042 } 1043 1044 void 1045 tmpfs_dir_destroy(struct tmpfs_mount *tmp, struct tmpfs_node *dnode) 1046 { 1047 struct tmpfs_dirent *de, *dde, *nde; 1048 1049 RB_FOREACH_SAFE(de, tmpfs_dir, &dnode->tn_dir.tn_dirhead, nde) { 1050 RB_REMOVE(tmpfs_dir, &dnode->tn_dir.tn_dirhead, de); 1051 /* Node may already be destroyed. */ 1052 de->td_node = NULL; 1053 if (tmpfs_dirent_duphead(de)) { 1054 while ((dde = LIST_FIRST(&de->ud.td_duphead)) != NULL) { 1055 LIST_REMOVE(dde, uh.td_dup.entries); 1056 dde->td_node = NULL; 1057 tmpfs_free_dirent(tmp, dde); 1058 } 1059 } 1060 tmpfs_free_dirent(tmp, de); 1061 } 1062 } 1063 1064 /* 1065 * Helper function for tmpfs_readdir. Creates a '.' entry for the given 1066 * directory and returns it in the uio space. The function returns 0 1067 * on success, -1 if there was not enough space in the uio structure to 1068 * hold the directory entry or an appropriate error code if another 1069 * error happens. 1070 */ 1071 static int 1072 tmpfs_dir_getdotdent(struct tmpfs_node *node, struct uio *uio) 1073 { 1074 int error; 1075 struct dirent dent; 1076 1077 TMPFS_VALIDATE_DIR(node); 1078 MPASS(uio->uio_offset == TMPFS_DIRCOOKIE_DOT); 1079 1080 dent.d_fileno = node->tn_id; 1081 dent.d_type = DT_DIR; 1082 dent.d_namlen = 1; 1083 dent.d_name[0] = '.'; 1084 dent.d_name[1] = '\0'; 1085 dent.d_reclen = GENERIC_DIRSIZ(&dent); 1086 1087 if (dent.d_reclen > uio->uio_resid) 1088 error = EJUSTRETURN; 1089 else 1090 error = uiomove(&dent, dent.d_reclen, uio); 1091 1092 node->tn_status |= TMPFS_NODE_ACCESSED; 1093 1094 return error; 1095 } 1096 1097 /* 1098 * Helper function for tmpfs_readdir. Creates a '..' entry for the given 1099 * directory and returns it in the uio space. The function returns 0 1100 * on success, -1 if there was not enough space in the uio structure to 1101 * hold the directory entry or an appropriate error code if another 1102 * error happens. 1103 */ 1104 static int 1105 tmpfs_dir_getdotdotdent(struct tmpfs_node *node, struct uio *uio) 1106 { 1107 int error; 1108 struct dirent dent; 1109 1110 TMPFS_VALIDATE_DIR(node); 1111 MPASS(uio->uio_offset == TMPFS_DIRCOOKIE_DOTDOT); 1112 1113 /* 1114 * Return ENOENT if the current node is already removed. 1115 */ 1116 TMPFS_ASSERT_LOCKED(node); 1117 if (node->tn_dir.tn_parent == NULL) { 1118 return (ENOENT); 1119 } 1120 1121 TMPFS_NODE_LOCK(node->tn_dir.tn_parent); 1122 dent.d_fileno = node->tn_dir.tn_parent->tn_id; 1123 TMPFS_NODE_UNLOCK(node->tn_dir.tn_parent); 1124 1125 dent.d_type = DT_DIR; 1126 dent.d_namlen = 2; 1127 dent.d_name[0] = '.'; 1128 dent.d_name[1] = '.'; 1129 dent.d_name[2] = '\0'; 1130 dent.d_reclen = GENERIC_DIRSIZ(&dent); 1131 1132 if (dent.d_reclen > uio->uio_resid) 1133 error = EJUSTRETURN; 1134 else 1135 error = uiomove(&dent, dent.d_reclen, uio); 1136 1137 node->tn_status |= TMPFS_NODE_ACCESSED; 1138 1139 return error; 1140 } 1141 1142 /* 1143 * Helper function for tmpfs_readdir. Returns as much directory entries 1144 * as can fit in the uio space. The read starts at uio->uio_offset. 1145 * The function returns 0 on success, -1 if there was not enough space 1146 * in the uio structure to hold the directory entry or an appropriate 1147 * error code if another error happens. 1148 */ 1149 int 1150 tmpfs_dir_getdents(struct tmpfs_node *node, struct uio *uio, int maxcookies, 1151 u_long *cookies, int *ncookies) 1152 { 1153 struct tmpfs_dir_cursor dc; 1154 struct tmpfs_dirent *de; 1155 off_t off; 1156 int error; 1157 1158 TMPFS_VALIDATE_DIR(node); 1159 1160 off = 0; 1161 1162 /* 1163 * Lookup the node from the current offset. The starting offset of 1164 * 0 will lookup both '.' and '..', and then the first real entry, 1165 * or EOF if there are none. Then find all entries for the dir that 1166 * fit into the buffer. Once no more entries are found (de == NULL), 1167 * the offset is set to TMPFS_DIRCOOKIE_EOF, which will cause the next 1168 * call to return 0. 1169 */ 1170 switch (uio->uio_offset) { 1171 case TMPFS_DIRCOOKIE_DOT: 1172 error = tmpfs_dir_getdotdent(node, uio); 1173 if (error != 0) 1174 return (error); 1175 uio->uio_offset = TMPFS_DIRCOOKIE_DOTDOT; 1176 if (cookies != NULL) 1177 cookies[(*ncookies)++] = off = uio->uio_offset; 1178 /* FALLTHROUGH */ 1179 case TMPFS_DIRCOOKIE_DOTDOT: 1180 error = tmpfs_dir_getdotdotdent(node, uio); 1181 if (error != 0) 1182 return (error); 1183 de = tmpfs_dir_first(node, &dc); 1184 uio->uio_offset = tmpfs_dirent_cookie(de); 1185 if (cookies != NULL) 1186 cookies[(*ncookies)++] = off = uio->uio_offset; 1187 /* EOF. */ 1188 if (de == NULL) 1189 return (0); 1190 break; 1191 case TMPFS_DIRCOOKIE_EOF: 1192 return (0); 1193 default: 1194 de = tmpfs_dir_lookup_cookie(node, uio->uio_offset, &dc); 1195 if (de == NULL) 1196 return (EINVAL); 1197 if (cookies != NULL) 1198 off = tmpfs_dirent_cookie(de); 1199 } 1200 1201 /* Read as much entries as possible; i.e., until we reach the end of 1202 * the directory or we exhaust uio space. */ 1203 do { 1204 struct dirent d; 1205 1206 /* Create a dirent structure representing the current 1207 * tmpfs_node and fill it. */ 1208 if (de->td_node == NULL) { 1209 d.d_fileno = 1; 1210 d.d_type = DT_WHT; 1211 } else { 1212 d.d_fileno = de->td_node->tn_id; 1213 switch (de->td_node->tn_type) { 1214 case VBLK: 1215 d.d_type = DT_BLK; 1216 break; 1217 1218 case VCHR: 1219 d.d_type = DT_CHR; 1220 break; 1221 1222 case VDIR: 1223 d.d_type = DT_DIR; 1224 break; 1225 1226 case VFIFO: 1227 d.d_type = DT_FIFO; 1228 break; 1229 1230 case VLNK: 1231 d.d_type = DT_LNK; 1232 break; 1233 1234 case VREG: 1235 d.d_type = DT_REG; 1236 break; 1237 1238 case VSOCK: 1239 d.d_type = DT_SOCK; 1240 break; 1241 1242 default: 1243 panic("tmpfs_dir_getdents: type %p %d", 1244 de->td_node, (int)de->td_node->tn_type); 1245 } 1246 } 1247 d.d_namlen = de->td_namelen; 1248 MPASS(de->td_namelen < sizeof(d.d_name)); 1249 (void)memcpy(d.d_name, de->ud.td_name, de->td_namelen); 1250 d.d_name[de->td_namelen] = '\0'; 1251 d.d_reclen = GENERIC_DIRSIZ(&d); 1252 1253 /* Stop reading if the directory entry we are treating is 1254 * bigger than the amount of data that can be returned. */ 1255 if (d.d_reclen > uio->uio_resid) { 1256 error = EJUSTRETURN; 1257 break; 1258 } 1259 1260 /* Copy the new dirent structure into the output buffer and 1261 * advance pointers. */ 1262 error = uiomove(&d, d.d_reclen, uio); 1263 if (error == 0) { 1264 de = tmpfs_dir_next(node, &dc); 1265 if (cookies != NULL) { 1266 off = tmpfs_dirent_cookie(de); 1267 MPASS(*ncookies < maxcookies); 1268 cookies[(*ncookies)++] = off; 1269 } 1270 } 1271 } while (error == 0 && uio->uio_resid > 0 && de != NULL); 1272 1273 /* Skip setting off when using cookies as it is already done above. */ 1274 if (cookies == NULL) 1275 off = tmpfs_dirent_cookie(de); 1276 1277 /* Update the offset and cache. */ 1278 uio->uio_offset = off; 1279 node->tn_dir.tn_readdir_lastn = off; 1280 node->tn_dir.tn_readdir_lastp = de; 1281 1282 node->tn_status |= TMPFS_NODE_ACCESSED; 1283 return error; 1284 } 1285 1286 int 1287 tmpfs_dir_whiteout_add(struct vnode *dvp, struct componentname *cnp) 1288 { 1289 struct tmpfs_dirent *de; 1290 int error; 1291 1292 error = tmpfs_alloc_dirent(VFS_TO_TMPFS(dvp->v_mount), NULL, 1293 cnp->cn_nameptr, cnp->cn_namelen, &de); 1294 if (error != 0) 1295 return (error); 1296 tmpfs_dir_attach(dvp, de); 1297 return (0); 1298 } 1299 1300 void 1301 tmpfs_dir_whiteout_remove(struct vnode *dvp, struct componentname *cnp) 1302 { 1303 struct tmpfs_dirent *de; 1304 1305 de = tmpfs_dir_lookup(VP_TO_TMPFS_DIR(dvp), NULL, cnp); 1306 MPASS(de != NULL && de->td_node == NULL); 1307 tmpfs_dir_detach(dvp, de); 1308 tmpfs_free_dirent(VFS_TO_TMPFS(dvp->v_mount), de); 1309 } 1310 1311 /* 1312 * Resizes the aobj associated with the regular file pointed to by 'vp' to the 1313 * size 'newsize'. 'vp' must point to a vnode that represents a regular file. 1314 * 'newsize' must be positive. 1315 * 1316 * Returns zero on success or an appropriate error code on failure. 1317 */ 1318 int 1319 tmpfs_reg_resize(struct vnode *vp, off_t newsize, boolean_t ignerr) 1320 { 1321 struct tmpfs_mount *tmp; 1322 struct tmpfs_node *node; 1323 vm_object_t uobj; 1324 vm_page_t m; 1325 vm_pindex_t idx, newpages, oldpages; 1326 off_t oldsize; 1327 int base, rv; 1328 1329 MPASS(vp->v_type == VREG); 1330 MPASS(newsize >= 0); 1331 1332 node = VP_TO_TMPFS_NODE(vp); 1333 uobj = node->tn_reg.tn_aobj; 1334 tmp = VFS_TO_TMPFS(vp->v_mount); 1335 1336 /* 1337 * Convert the old and new sizes to the number of pages needed to 1338 * store them. It may happen that we do not need to do anything 1339 * because the last allocated page can accommodate the change on 1340 * its own. 1341 */ 1342 oldsize = node->tn_size; 1343 oldpages = OFF_TO_IDX(oldsize + PAGE_MASK); 1344 MPASS(oldpages == uobj->size); 1345 newpages = OFF_TO_IDX(newsize + PAGE_MASK); 1346 if (newpages > oldpages && 1347 tmpfs_pages_check_avail(tmp, newpages - oldpages) == 0) 1348 return (ENOSPC); 1349 1350 VM_OBJECT_WLOCK(uobj); 1351 if (newsize < oldsize) { 1352 /* 1353 * Zero the truncated part of the last page. 1354 */ 1355 base = newsize & PAGE_MASK; 1356 if (base != 0) { 1357 idx = OFF_TO_IDX(newsize); 1358 retry: 1359 m = vm_page_lookup(uobj, idx); 1360 if (m != NULL) { 1361 if (vm_page_sleep_if_busy(m, "tmfssz")) 1362 goto retry; 1363 MPASS(m->valid == VM_PAGE_BITS_ALL); 1364 } else if (vm_pager_has_page(uobj, idx, NULL, NULL)) { 1365 m = vm_page_alloc(uobj, idx, VM_ALLOC_NORMAL); 1366 if (m == NULL) { 1367 VM_OBJECT_WUNLOCK(uobj); 1368 VM_WAIT; 1369 VM_OBJECT_WLOCK(uobj); 1370 goto retry; 1371 } else if (m->valid != VM_PAGE_BITS_ALL) 1372 rv = vm_pager_get_pages(uobj, &m, 1, 0); 1373 else 1374 /* A cached page was reactivated. */ 1375 rv = VM_PAGER_OK; 1376 vm_page_lock(m); 1377 if (rv == VM_PAGER_OK) { 1378 vm_page_deactivate(m); 1379 vm_page_unlock(m); 1380 vm_page_xunbusy(m); 1381 } else { 1382 vm_page_free(m); 1383 vm_page_unlock(m); 1384 if (ignerr) 1385 m = NULL; 1386 else { 1387 VM_OBJECT_WUNLOCK(uobj); 1388 return (EIO); 1389 } 1390 } 1391 } 1392 if (m != NULL) { 1393 pmap_zero_page_area(m, base, PAGE_SIZE - base); 1394 vm_page_dirty(m); 1395 vm_pager_page_unswapped(m); 1396 } 1397 } 1398 1399 /* 1400 * Release any swap space and free any whole pages. 1401 */ 1402 if (newpages < oldpages) { 1403 swap_pager_freespace(uobj, newpages, oldpages - 1404 newpages); 1405 vm_object_page_remove(uobj, newpages, 0, 0); 1406 } 1407 } 1408 uobj->size = newpages; 1409 VM_OBJECT_WUNLOCK(uobj); 1410 1411 TMPFS_LOCK(tmp); 1412 tmp->tm_pages_used += (newpages - oldpages); 1413 TMPFS_UNLOCK(tmp); 1414 1415 node->tn_size = newsize; 1416 return (0); 1417 } 1418 1419 void 1420 tmpfs_check_mtime(struct vnode *vp) 1421 { 1422 struct tmpfs_node *node; 1423 struct vm_object *obj; 1424 1425 ASSERT_VOP_ELOCKED(vp, "check_mtime"); 1426 if (vp->v_type != VREG) 1427 return; 1428 obj = vp->v_object; 1429 KASSERT((obj->flags & (OBJ_TMPFS_NODE | OBJ_TMPFS)) == 1430 (OBJ_TMPFS_NODE | OBJ_TMPFS), ("non-tmpfs obj")); 1431 /* unlocked read */ 1432 if ((obj->flags & OBJ_TMPFS_DIRTY) != 0) { 1433 VM_OBJECT_WLOCK(obj); 1434 if ((obj->flags & OBJ_TMPFS_DIRTY) != 0) { 1435 obj->flags &= ~OBJ_TMPFS_DIRTY; 1436 node = VP_TO_TMPFS_NODE(vp); 1437 node->tn_status |= TMPFS_NODE_MODIFIED | 1438 TMPFS_NODE_CHANGED; 1439 } 1440 VM_OBJECT_WUNLOCK(obj); 1441 } 1442 } 1443 1444 /* 1445 * Change flags of the given vnode. 1446 * Caller should execute tmpfs_update on vp after a successful execution. 1447 * The vnode must be locked on entry and remain locked on exit. 1448 */ 1449 int 1450 tmpfs_chflags(struct vnode *vp, u_long flags, struct ucred *cred, 1451 struct thread *p) 1452 { 1453 int error; 1454 struct tmpfs_node *node; 1455 1456 MPASS(VOP_ISLOCKED(vp)); 1457 1458 node = VP_TO_TMPFS_NODE(vp); 1459 1460 if ((flags & ~(SF_APPEND | SF_ARCHIVED | SF_IMMUTABLE | SF_NOUNLINK | 1461 UF_APPEND | UF_ARCHIVE | UF_HIDDEN | UF_IMMUTABLE | UF_NODUMP | 1462 UF_NOUNLINK | UF_OFFLINE | UF_OPAQUE | UF_READONLY | UF_REPARSE | 1463 UF_SPARSE | UF_SYSTEM)) != 0) 1464 return (EOPNOTSUPP); 1465 1466 /* Disallow this operation if the file system is mounted read-only. */ 1467 if (vp->v_mount->mnt_flag & MNT_RDONLY) 1468 return EROFS; 1469 1470 /* 1471 * Callers may only modify the file flags on objects they 1472 * have VADMIN rights for. 1473 */ 1474 if ((error = VOP_ACCESS(vp, VADMIN, cred, p))) 1475 return (error); 1476 /* 1477 * Unprivileged processes are not permitted to unset system 1478 * flags, or modify flags if any system flags are set. 1479 */ 1480 if (!priv_check_cred(cred, PRIV_VFS_SYSFLAGS, 0)) { 1481 if (node->tn_flags & 1482 (SF_NOUNLINK | SF_IMMUTABLE | SF_APPEND)) { 1483 error = securelevel_gt(cred, 0); 1484 if (error) 1485 return (error); 1486 } 1487 } else { 1488 if (node->tn_flags & 1489 (SF_NOUNLINK | SF_IMMUTABLE | SF_APPEND) || 1490 ((flags ^ node->tn_flags) & SF_SETTABLE)) 1491 return (EPERM); 1492 } 1493 node->tn_flags = flags; 1494 node->tn_status |= TMPFS_NODE_CHANGED; 1495 1496 MPASS(VOP_ISLOCKED(vp)); 1497 1498 return 0; 1499 } 1500 1501 /* 1502 * Change access mode on the given vnode. 1503 * Caller should execute tmpfs_update on vp after a successful execution. 1504 * The vnode must be locked on entry and remain locked on exit. 1505 */ 1506 int 1507 tmpfs_chmod(struct vnode *vp, mode_t mode, struct ucred *cred, struct thread *p) 1508 { 1509 int error; 1510 struct tmpfs_node *node; 1511 1512 MPASS(VOP_ISLOCKED(vp)); 1513 1514 node = VP_TO_TMPFS_NODE(vp); 1515 1516 /* Disallow this operation if the file system is mounted read-only. */ 1517 if (vp->v_mount->mnt_flag & MNT_RDONLY) 1518 return EROFS; 1519 1520 /* Immutable or append-only files cannot be modified, either. */ 1521 if (node->tn_flags & (IMMUTABLE | APPEND)) 1522 return EPERM; 1523 1524 /* 1525 * To modify the permissions on a file, must possess VADMIN 1526 * for that file. 1527 */ 1528 if ((error = VOP_ACCESS(vp, VADMIN, cred, p))) 1529 return (error); 1530 1531 /* 1532 * Privileged processes may set the sticky bit on non-directories, 1533 * as well as set the setgid bit on a file with a group that the 1534 * process is not a member of. 1535 */ 1536 if (vp->v_type != VDIR && (mode & S_ISTXT)) { 1537 if (priv_check_cred(cred, PRIV_VFS_STICKYFILE, 0)) 1538 return (EFTYPE); 1539 } 1540 if (!groupmember(node->tn_gid, cred) && (mode & S_ISGID)) { 1541 error = priv_check_cred(cred, PRIV_VFS_SETGID, 0); 1542 if (error) 1543 return (error); 1544 } 1545 1546 1547 node->tn_mode &= ~ALLPERMS; 1548 node->tn_mode |= mode & ALLPERMS; 1549 1550 node->tn_status |= TMPFS_NODE_CHANGED; 1551 1552 MPASS(VOP_ISLOCKED(vp)); 1553 1554 return 0; 1555 } 1556 1557 /* 1558 * Change ownership of the given vnode. At least one of uid or gid must 1559 * be different than VNOVAL. If one is set to that value, the attribute 1560 * is unchanged. 1561 * Caller should execute tmpfs_update on vp after a successful execution. 1562 * The vnode must be locked on entry and remain locked on exit. 1563 */ 1564 int 1565 tmpfs_chown(struct vnode *vp, uid_t uid, gid_t gid, struct ucred *cred, 1566 struct thread *p) 1567 { 1568 int error; 1569 struct tmpfs_node *node; 1570 uid_t ouid; 1571 gid_t ogid; 1572 1573 MPASS(VOP_ISLOCKED(vp)); 1574 1575 node = VP_TO_TMPFS_NODE(vp); 1576 1577 /* Assign default values if they are unknown. */ 1578 MPASS(uid != VNOVAL || gid != VNOVAL); 1579 if (uid == VNOVAL) 1580 uid = node->tn_uid; 1581 if (gid == VNOVAL) 1582 gid = node->tn_gid; 1583 MPASS(uid != VNOVAL && gid != VNOVAL); 1584 1585 /* Disallow this operation if the file system is mounted read-only. */ 1586 if (vp->v_mount->mnt_flag & MNT_RDONLY) 1587 return EROFS; 1588 1589 /* Immutable or append-only files cannot be modified, either. */ 1590 if (node->tn_flags & (IMMUTABLE | APPEND)) 1591 return EPERM; 1592 1593 /* 1594 * To modify the ownership of a file, must possess VADMIN for that 1595 * file. 1596 */ 1597 if ((error = VOP_ACCESS(vp, VADMIN, cred, p))) 1598 return (error); 1599 1600 /* 1601 * To change the owner of a file, or change the group of a file to a 1602 * group of which we are not a member, the caller must have 1603 * privilege. 1604 */ 1605 if ((uid != node->tn_uid || 1606 (gid != node->tn_gid && !groupmember(gid, cred))) && 1607 (error = priv_check_cred(cred, PRIV_VFS_CHOWN, 0))) 1608 return (error); 1609 1610 ogid = node->tn_gid; 1611 ouid = node->tn_uid; 1612 1613 node->tn_uid = uid; 1614 node->tn_gid = gid; 1615 1616 node->tn_status |= TMPFS_NODE_CHANGED; 1617 1618 if ((node->tn_mode & (S_ISUID | S_ISGID)) && (ouid != uid || ogid != gid)) { 1619 if (priv_check_cred(cred, PRIV_VFS_RETAINSUGID, 0)) 1620 node->tn_mode &= ~(S_ISUID | S_ISGID); 1621 } 1622 1623 MPASS(VOP_ISLOCKED(vp)); 1624 1625 return 0; 1626 } 1627 1628 /* 1629 * Change size of the given vnode. 1630 * Caller should execute tmpfs_update on vp after a successful execution. 1631 * The vnode must be locked on entry and remain locked on exit. 1632 */ 1633 int 1634 tmpfs_chsize(struct vnode *vp, u_quad_t size, struct ucred *cred, 1635 struct thread *p) 1636 { 1637 int error; 1638 struct tmpfs_node *node; 1639 1640 MPASS(VOP_ISLOCKED(vp)); 1641 1642 node = VP_TO_TMPFS_NODE(vp); 1643 1644 /* Decide whether this is a valid operation based on the file type. */ 1645 error = 0; 1646 switch (vp->v_type) { 1647 case VDIR: 1648 return EISDIR; 1649 1650 case VREG: 1651 if (vp->v_mount->mnt_flag & MNT_RDONLY) 1652 return EROFS; 1653 break; 1654 1655 case VBLK: 1656 /* FALLTHROUGH */ 1657 case VCHR: 1658 /* FALLTHROUGH */ 1659 case VFIFO: 1660 /* Allow modifications of special files even if in the file 1661 * system is mounted read-only (we are not modifying the 1662 * files themselves, but the objects they represent). */ 1663 return 0; 1664 1665 default: 1666 /* Anything else is unsupported. */ 1667 return EOPNOTSUPP; 1668 } 1669 1670 /* Immutable or append-only files cannot be modified, either. */ 1671 if (node->tn_flags & (IMMUTABLE | APPEND)) 1672 return EPERM; 1673 1674 error = tmpfs_truncate(vp, size); 1675 /* tmpfs_truncate will raise the NOTE_EXTEND and NOTE_ATTRIB kevents 1676 * for us, as will update tn_status; no need to do that here. */ 1677 1678 MPASS(VOP_ISLOCKED(vp)); 1679 1680 return error; 1681 } 1682 1683 /* 1684 * Change access and modification times of the given vnode. 1685 * Caller should execute tmpfs_update on vp after a successful execution. 1686 * The vnode must be locked on entry and remain locked on exit. 1687 */ 1688 int 1689 tmpfs_chtimes(struct vnode *vp, struct vattr *vap, 1690 struct ucred *cred, struct thread *l) 1691 { 1692 int error; 1693 struct tmpfs_node *node; 1694 1695 MPASS(VOP_ISLOCKED(vp)); 1696 1697 node = VP_TO_TMPFS_NODE(vp); 1698 1699 /* Disallow this operation if the file system is mounted read-only. */ 1700 if (vp->v_mount->mnt_flag & MNT_RDONLY) 1701 return EROFS; 1702 1703 /* Immutable or append-only files cannot be modified, either. */ 1704 if (node->tn_flags & (IMMUTABLE | APPEND)) 1705 return EPERM; 1706 1707 error = vn_utimes_perm(vp, vap, cred, l); 1708 if (error != 0) 1709 return (error); 1710 1711 if (vap->va_atime.tv_sec != VNOVAL && vap->va_atime.tv_nsec != VNOVAL) 1712 node->tn_status |= TMPFS_NODE_ACCESSED; 1713 1714 if (vap->va_mtime.tv_sec != VNOVAL && vap->va_mtime.tv_nsec != VNOVAL) 1715 node->tn_status |= TMPFS_NODE_MODIFIED; 1716 1717 if (vap->va_birthtime.tv_nsec != VNOVAL && 1718 vap->va_birthtime.tv_nsec != VNOVAL) 1719 node->tn_status |= TMPFS_NODE_MODIFIED; 1720 1721 tmpfs_itimes(vp, &vap->va_atime, &vap->va_mtime); 1722 1723 if (vap->va_birthtime.tv_nsec != VNOVAL && 1724 vap->va_birthtime.tv_nsec != VNOVAL) 1725 node->tn_birthtime = vap->va_birthtime; 1726 MPASS(VOP_ISLOCKED(vp)); 1727 1728 return 0; 1729 } 1730 1731 /* Sync timestamps */ 1732 void 1733 tmpfs_itimes(struct vnode *vp, const struct timespec *acc, 1734 const struct timespec *mod) 1735 { 1736 struct tmpfs_node *node; 1737 struct timespec now; 1738 1739 node = VP_TO_TMPFS_NODE(vp); 1740 1741 if ((node->tn_status & (TMPFS_NODE_ACCESSED | TMPFS_NODE_MODIFIED | 1742 TMPFS_NODE_CHANGED)) == 0) 1743 return; 1744 1745 vfs_timestamp(&now); 1746 if (node->tn_status & TMPFS_NODE_ACCESSED) { 1747 if (acc == NULL) 1748 acc = &now; 1749 node->tn_atime = *acc; 1750 } 1751 if (node->tn_status & TMPFS_NODE_MODIFIED) { 1752 if (mod == NULL) 1753 mod = &now; 1754 node->tn_mtime = *mod; 1755 } 1756 if (node->tn_status & TMPFS_NODE_CHANGED) { 1757 node->tn_ctime = now; 1758 } 1759 node->tn_status &= 1760 ~(TMPFS_NODE_ACCESSED | TMPFS_NODE_MODIFIED | TMPFS_NODE_CHANGED); 1761 } 1762 1763 void 1764 tmpfs_update(struct vnode *vp) 1765 { 1766 1767 tmpfs_itimes(vp, NULL, NULL); 1768 } 1769 1770 int 1771 tmpfs_truncate(struct vnode *vp, off_t length) 1772 { 1773 int error; 1774 struct tmpfs_node *node; 1775 1776 node = VP_TO_TMPFS_NODE(vp); 1777 1778 if (length < 0) { 1779 error = EINVAL; 1780 goto out; 1781 } 1782 1783 if (node->tn_size == length) { 1784 error = 0; 1785 goto out; 1786 } 1787 1788 if (length > VFS_TO_TMPFS(vp->v_mount)->tm_maxfilesize) 1789 return (EFBIG); 1790 1791 error = tmpfs_reg_resize(vp, length, FALSE); 1792 if (error == 0) { 1793 node->tn_status |= TMPFS_NODE_CHANGED | TMPFS_NODE_MODIFIED; 1794 } 1795 1796 out: 1797 tmpfs_update(vp); 1798 1799 return error; 1800 } 1801 1802 static __inline int 1803 tmpfs_dirtree_cmp(struct tmpfs_dirent *a, struct tmpfs_dirent *b) 1804 { 1805 if (a->td_hash > b->td_hash) 1806 return (1); 1807 else if (a->td_hash < b->td_hash) 1808 return (-1); 1809 return (0); 1810 } 1811 1812 RB_GENERATE_STATIC(tmpfs_dir, tmpfs_dirent, uh.td_entries, tmpfs_dirtree_cmp); 1813