tmpfs_vnops.c revision 232960
1/* $NetBSD: tmpfs_vnops.c,v 1.39 2007/07/23 15:41:01 jmmv Exp $ */ 2 3/*- 4 * Copyright (c) 2005, 2006 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 * tmpfs vnode interface. 35 */ 36#include <sys/cdefs.h> 37__FBSDID("$FreeBSD: head/sys/fs/tmpfs/tmpfs_vnops.c 232960 2012-03-14 09:15:50Z gleb $"); 38 39#include <sys/param.h> 40#include <sys/fcntl.h> 41#include <sys/lockf.h> 42#include <sys/namei.h> 43#include <sys/priv.h> 44#include <sys/proc.h> 45#include <sys/sched.h> 46#include <sys/sf_buf.h> 47#include <sys/stat.h> 48#include <sys/systm.h> 49#include <sys/sysctl.h> 50#include <sys/unistd.h> 51#include <sys/vnode.h> 52 53#include <vm/vm.h> 54#include <vm/vm_object.h> 55#include <vm/vm_page.h> 56#include <vm/vm_pager.h> 57 58#include <fs/tmpfs/tmpfs_vnops.h> 59#include <fs/tmpfs/tmpfs.h> 60 61SYSCTL_DECL(_vfs_tmpfs); 62 63static volatile int tmpfs_rename_restarts; 64SYSCTL_INT(_vfs_tmpfs, OID_AUTO, rename_restarts, CTLFLAG_RD, 65 __DEVOLATILE(int *, &tmpfs_rename_restarts), 0, 66 "Times rename had to restart due to lock contention"); 67 68/* --------------------------------------------------------------------- */ 69 70static int 71tmpfs_lookup(struct vop_cachedlookup_args *v) 72{ 73 struct vnode *dvp = v->a_dvp; 74 struct vnode **vpp = v->a_vpp; 75 struct componentname *cnp = v->a_cnp; 76 77 int error; 78 struct tmpfs_dirent *de; 79 struct tmpfs_node *dnode; 80 81 dnode = VP_TO_TMPFS_DIR(dvp); 82 *vpp = NULLVP; 83 84 /* Check accessibility of requested node as a first step. */ 85 error = VOP_ACCESS(dvp, VEXEC, cnp->cn_cred, cnp->cn_thread); 86 if (error != 0) 87 goto out; 88 89 /* We cannot be requesting the parent directory of the root node. */ 90 MPASS(IMPLIES(dnode->tn_type == VDIR && 91 dnode->tn_dir.tn_parent == dnode, 92 !(cnp->cn_flags & ISDOTDOT))); 93 94 TMPFS_ASSERT_LOCKED(dnode); 95 if (dnode->tn_dir.tn_parent == NULL) { 96 error = ENOENT; 97 goto out; 98 } 99 if (cnp->cn_flags & ISDOTDOT) { 100 int ltype = 0; 101 102 ltype = VOP_ISLOCKED(dvp); 103 vhold(dvp); 104 VOP_UNLOCK(dvp, 0); 105 /* Allocate a new vnode on the matching entry. */ 106 error = tmpfs_alloc_vp(dvp->v_mount, dnode->tn_dir.tn_parent, 107 cnp->cn_lkflags, vpp); 108 109 vn_lock(dvp, ltype | LK_RETRY); 110 vdrop(dvp); 111 } else if (cnp->cn_namelen == 1 && cnp->cn_nameptr[0] == '.') { 112 VREF(dvp); 113 *vpp = dvp; 114 error = 0; 115 } else { 116 de = tmpfs_dir_lookup(dnode, NULL, cnp); 117 if (de != NULL && de->td_node == NULL) 118 cnp->cn_flags |= ISWHITEOUT; 119 if (de == NULL || de->td_node == NULL) { 120 /* The entry was not found in the directory. 121 * This is OK if we are creating or renaming an 122 * entry and are working on the last component of 123 * the path name. */ 124 if ((cnp->cn_flags & ISLASTCN) && 125 (cnp->cn_nameiop == CREATE || \ 126 cnp->cn_nameiop == RENAME || 127 (cnp->cn_nameiop == DELETE && 128 cnp->cn_flags & DOWHITEOUT && 129 cnp->cn_flags & ISWHITEOUT))) { 130 error = VOP_ACCESS(dvp, VWRITE, cnp->cn_cred, 131 cnp->cn_thread); 132 if (error != 0) 133 goto out; 134 135 /* Keep the component name in the buffer for 136 * future uses. */ 137 cnp->cn_flags |= SAVENAME; 138 139 error = EJUSTRETURN; 140 } else 141 error = ENOENT; 142 } else { 143 struct tmpfs_node *tnode; 144 145 /* The entry was found, so get its associated 146 * tmpfs_node. */ 147 tnode = de->td_node; 148 149 /* If we are not at the last path component and 150 * found a non-directory or non-link entry (which 151 * may itself be pointing to a directory), raise 152 * an error. */ 153 if ((tnode->tn_type != VDIR && 154 tnode->tn_type != VLNK) && 155 !(cnp->cn_flags & ISLASTCN)) { 156 error = ENOTDIR; 157 goto out; 158 } 159 160 /* If we are deleting or renaming the entry, keep 161 * track of its tmpfs_dirent so that it can be 162 * easily deleted later. */ 163 if ((cnp->cn_flags & ISLASTCN) && 164 (cnp->cn_nameiop == DELETE || 165 cnp->cn_nameiop == RENAME)) { 166 error = VOP_ACCESS(dvp, VWRITE, cnp->cn_cred, 167 cnp->cn_thread); 168 if (error != 0) 169 goto out; 170 171 /* Allocate a new vnode on the matching entry. */ 172 error = tmpfs_alloc_vp(dvp->v_mount, tnode, 173 cnp->cn_lkflags, vpp); 174 if (error != 0) 175 goto out; 176 177 if ((dnode->tn_mode & S_ISTXT) && 178 VOP_ACCESS(dvp, VADMIN, cnp->cn_cred, cnp->cn_thread) && 179 VOP_ACCESS(*vpp, VADMIN, cnp->cn_cred, cnp->cn_thread)) { 180 error = EPERM; 181 vput(*vpp); 182 *vpp = NULL; 183 goto out; 184 } 185 cnp->cn_flags |= SAVENAME; 186 } else { 187 error = tmpfs_alloc_vp(dvp->v_mount, tnode, 188 cnp->cn_lkflags, vpp); 189 } 190 } 191 } 192 193 /* Store the result of this lookup in the cache. Avoid this if the 194 * request was for creation, as it does not improve timings on 195 * emprical tests. */ 196 if ((cnp->cn_flags & MAKEENTRY) && cnp->cn_nameiop != CREATE) 197 cache_enter(dvp, *vpp, cnp); 198 199out: 200 /* If there were no errors, *vpp cannot be null and it must be 201 * locked. */ 202 MPASS(IFF(error == 0, *vpp != NULLVP && VOP_ISLOCKED(*vpp))); 203 204 return error; 205} 206 207/* --------------------------------------------------------------------- */ 208 209static int 210tmpfs_create(struct vop_create_args *v) 211{ 212 struct vnode *dvp = v->a_dvp; 213 struct vnode **vpp = v->a_vpp; 214 struct componentname *cnp = v->a_cnp; 215 struct vattr *vap = v->a_vap; 216 217 MPASS(vap->va_type == VREG || vap->va_type == VSOCK); 218 219 return tmpfs_alloc_file(dvp, vpp, vap, cnp, NULL); 220} 221/* --------------------------------------------------------------------- */ 222 223static int 224tmpfs_mknod(struct vop_mknod_args *v) 225{ 226 struct vnode *dvp = v->a_dvp; 227 struct vnode **vpp = v->a_vpp; 228 struct componentname *cnp = v->a_cnp; 229 struct vattr *vap = v->a_vap; 230 231 if (vap->va_type != VBLK && vap->va_type != VCHR && 232 vap->va_type != VFIFO) 233 return EINVAL; 234 235 return tmpfs_alloc_file(dvp, vpp, vap, cnp, NULL); 236} 237 238/* --------------------------------------------------------------------- */ 239 240static int 241tmpfs_open(struct vop_open_args *v) 242{ 243 struct vnode *vp = v->a_vp; 244 int mode = v->a_mode; 245 246 int error; 247 struct tmpfs_node *node; 248 249 MPASS(VOP_ISLOCKED(vp)); 250 251 node = VP_TO_TMPFS_NODE(vp); 252 253 /* The file is still active but all its names have been removed 254 * (e.g. by a "rmdir $(pwd)"). It cannot be opened any more as 255 * it is about to die. */ 256 if (node->tn_links < 1) 257 return (ENOENT); 258 259 /* If the file is marked append-only, deny write requests. */ 260 if (node->tn_flags & APPEND && (mode & (FWRITE | O_APPEND)) == FWRITE) 261 error = EPERM; 262 else { 263 error = 0; 264 vnode_create_vobject(vp, node->tn_size, v->a_td); 265 } 266 267 MPASS(VOP_ISLOCKED(vp)); 268 return error; 269} 270 271/* --------------------------------------------------------------------- */ 272 273static int 274tmpfs_close(struct vop_close_args *v) 275{ 276 struct vnode *vp = v->a_vp; 277 278 MPASS(VOP_ISLOCKED(vp)); 279 280 /* Update node times. */ 281 tmpfs_update(vp); 282 283 return (0); 284} 285 286/* --------------------------------------------------------------------- */ 287 288int 289tmpfs_access(struct vop_access_args *v) 290{ 291 struct vnode *vp = v->a_vp; 292 accmode_t accmode = v->a_accmode; 293 struct ucred *cred = v->a_cred; 294 295 int error; 296 struct tmpfs_node *node; 297 298 MPASS(VOP_ISLOCKED(vp)); 299 300 node = VP_TO_TMPFS_NODE(vp); 301 302 switch (vp->v_type) { 303 case VDIR: 304 /* FALLTHROUGH */ 305 case VLNK: 306 /* FALLTHROUGH */ 307 case VREG: 308 if (accmode & VWRITE && vp->v_mount->mnt_flag & MNT_RDONLY) { 309 error = EROFS; 310 goto out; 311 } 312 break; 313 314 case VBLK: 315 /* FALLTHROUGH */ 316 case VCHR: 317 /* FALLTHROUGH */ 318 case VSOCK: 319 /* FALLTHROUGH */ 320 case VFIFO: 321 break; 322 323 default: 324 error = EINVAL; 325 goto out; 326 } 327 328 if (accmode & VWRITE && node->tn_flags & IMMUTABLE) { 329 error = EPERM; 330 goto out; 331 } 332 333 error = vaccess(vp->v_type, node->tn_mode, node->tn_uid, 334 node->tn_gid, accmode, cred, NULL); 335 336out: 337 MPASS(VOP_ISLOCKED(vp)); 338 339 return error; 340} 341 342/* --------------------------------------------------------------------- */ 343 344int 345tmpfs_getattr(struct vop_getattr_args *v) 346{ 347 struct vnode *vp = v->a_vp; 348 struct vattr *vap = v->a_vap; 349 350 struct tmpfs_node *node; 351 352 node = VP_TO_TMPFS_NODE(vp); 353 354 tmpfs_update(vp); 355 356 vap->va_type = vp->v_type; 357 vap->va_mode = node->tn_mode; 358 vap->va_nlink = node->tn_links; 359 vap->va_uid = node->tn_uid; 360 vap->va_gid = node->tn_gid; 361 vap->va_fsid = vp->v_mount->mnt_stat.f_fsid.val[0]; 362 vap->va_fileid = node->tn_id; 363 vap->va_size = node->tn_size; 364 vap->va_blocksize = PAGE_SIZE; 365 vap->va_atime = node->tn_atime; 366 vap->va_mtime = node->tn_mtime; 367 vap->va_ctime = node->tn_ctime; 368 vap->va_birthtime = node->tn_birthtime; 369 vap->va_gen = node->tn_gen; 370 vap->va_flags = node->tn_flags; 371 vap->va_rdev = (vp->v_type == VBLK || vp->v_type == VCHR) ? 372 node->tn_rdev : NODEV; 373 vap->va_bytes = round_page(node->tn_size); 374 vap->va_filerev = 0; 375 376 return 0; 377} 378 379/* --------------------------------------------------------------------- */ 380 381/* XXX Should this operation be atomic? I think it should, but code in 382 * XXX other places (e.g., ufs) doesn't seem to be... */ 383int 384tmpfs_setattr(struct vop_setattr_args *v) 385{ 386 struct vnode *vp = v->a_vp; 387 struct vattr *vap = v->a_vap; 388 struct ucred *cred = v->a_cred; 389 struct thread *td = curthread; 390 391 int error; 392 393 MPASS(VOP_ISLOCKED(vp)); 394 395 error = 0; 396 397 /* Abort if any unsettable attribute is given. */ 398 if (vap->va_type != VNON || 399 vap->va_nlink != VNOVAL || 400 vap->va_fsid != VNOVAL || 401 vap->va_fileid != VNOVAL || 402 vap->va_blocksize != VNOVAL || 403 vap->va_gen != VNOVAL || 404 vap->va_rdev != VNOVAL || 405 vap->va_bytes != VNOVAL) 406 error = EINVAL; 407 408 if (error == 0 && (vap->va_flags != VNOVAL)) 409 error = tmpfs_chflags(vp, vap->va_flags, cred, td); 410 411 if (error == 0 && (vap->va_size != VNOVAL)) 412 error = tmpfs_chsize(vp, vap->va_size, cred, td); 413 414 if (error == 0 && (vap->va_uid != VNOVAL || vap->va_gid != VNOVAL)) 415 error = tmpfs_chown(vp, vap->va_uid, vap->va_gid, cred, td); 416 417 if (error == 0 && (vap->va_mode != (mode_t)VNOVAL)) 418 error = tmpfs_chmod(vp, vap->va_mode, cred, td); 419 420 if (error == 0 && ((vap->va_atime.tv_sec != VNOVAL && 421 vap->va_atime.tv_nsec != VNOVAL) || 422 (vap->va_mtime.tv_sec != VNOVAL && 423 vap->va_mtime.tv_nsec != VNOVAL) || 424 (vap->va_birthtime.tv_sec != VNOVAL && 425 vap->va_birthtime.tv_nsec != VNOVAL))) 426 error = tmpfs_chtimes(vp, &vap->va_atime, &vap->va_mtime, 427 &vap->va_birthtime, vap->va_vaflags, cred, td); 428 429 /* Update the node times. We give preference to the error codes 430 * generated by this function rather than the ones that may arise 431 * from tmpfs_update. */ 432 tmpfs_update(vp); 433 434 MPASS(VOP_ISLOCKED(vp)); 435 436 return error; 437} 438 439/* --------------------------------------------------------------------- */ 440static int 441tmpfs_nocacheread(vm_object_t tobj, vm_pindex_t idx, 442 vm_offset_t offset, size_t tlen, struct uio *uio) 443{ 444 vm_page_t m; 445 int error, rv; 446 447 VM_OBJECT_LOCK(tobj); 448 m = vm_page_grab(tobj, idx, VM_ALLOC_WIRED | 449 VM_ALLOC_NORMAL | VM_ALLOC_RETRY); 450 if (m->valid != VM_PAGE_BITS_ALL) { 451 if (vm_pager_has_page(tobj, idx, NULL, NULL)) { 452 rv = vm_pager_get_pages(tobj, &m, 1, 0); 453 if (rv != VM_PAGER_OK) { 454 vm_page_lock(m); 455 vm_page_free(m); 456 vm_page_unlock(m); 457 VM_OBJECT_UNLOCK(tobj); 458 return (EIO); 459 } 460 } else 461 vm_page_zero_invalid(m, TRUE); 462 } 463 VM_OBJECT_UNLOCK(tobj); 464 error = uiomove_fromphys(&m, offset, tlen, uio); 465 VM_OBJECT_LOCK(tobj); 466 vm_page_lock(m); 467 vm_page_unwire(m, TRUE); 468 vm_page_unlock(m); 469 vm_page_wakeup(m); 470 VM_OBJECT_UNLOCK(tobj); 471 472 return (error); 473} 474 475static __inline int 476tmpfs_nocacheread_buf(vm_object_t tobj, vm_pindex_t idx, 477 vm_offset_t offset, size_t tlen, void *buf) 478{ 479 struct uio uio; 480 struct iovec iov; 481 482 uio.uio_iovcnt = 1; 483 uio.uio_iov = &iov; 484 iov.iov_base = buf; 485 iov.iov_len = tlen; 486 487 uio.uio_offset = 0; 488 uio.uio_resid = tlen; 489 uio.uio_rw = UIO_READ; 490 uio.uio_segflg = UIO_SYSSPACE; 491 uio.uio_td = curthread; 492 493 return (tmpfs_nocacheread(tobj, idx, offset, tlen, &uio)); 494} 495 496static int 497tmpfs_mappedread(vm_object_t vobj, vm_object_t tobj, size_t len, struct uio *uio) 498{ 499 struct sf_buf *sf; 500 vm_pindex_t idx; 501 vm_page_t m; 502 vm_offset_t offset; 503 off_t addr; 504 size_t tlen; 505 char *ma; 506 int error; 507 508 addr = uio->uio_offset; 509 idx = OFF_TO_IDX(addr); 510 offset = addr & PAGE_MASK; 511 tlen = MIN(PAGE_SIZE - offset, len); 512 513 if ((vobj == NULL) || 514 (vobj->resident_page_count == 0 && vobj->cache == NULL)) 515 goto nocache; 516 517 VM_OBJECT_LOCK(vobj); 518lookupvpg: 519 if (((m = vm_page_lookup(vobj, idx)) != NULL) && 520 vm_page_is_valid(m, offset, tlen)) { 521 if ((m->oflags & VPO_BUSY) != 0) { 522 /* 523 * Reference the page before unlocking and sleeping so 524 * that the page daemon is less likely to reclaim it. 525 */ 526 vm_page_reference(m); 527 vm_page_sleep(m, "tmfsmr"); 528 goto lookupvpg; 529 } 530 vm_page_busy(m); 531 VM_OBJECT_UNLOCK(vobj); 532 error = uiomove_fromphys(&m, offset, tlen, uio); 533 VM_OBJECT_LOCK(vobj); 534 vm_page_wakeup(m); 535 VM_OBJECT_UNLOCK(vobj); 536 return (error); 537 } else if (m != NULL && uio->uio_segflg == UIO_NOCOPY) { 538 KASSERT(offset == 0, 539 ("unexpected offset in tmpfs_mappedread for sendfile")); 540 if ((m->oflags & VPO_BUSY) != 0) { 541 /* 542 * Reference the page before unlocking and sleeping so 543 * that the page daemon is less likely to reclaim it. 544 */ 545 vm_page_reference(m); 546 vm_page_sleep(m, "tmfsmr"); 547 goto lookupvpg; 548 } 549 vm_page_busy(m); 550 VM_OBJECT_UNLOCK(vobj); 551 sched_pin(); 552 sf = sf_buf_alloc(m, SFB_CPUPRIVATE); 553 ma = (char *)sf_buf_kva(sf); 554 error = tmpfs_nocacheread_buf(tobj, idx, 0, tlen, ma); 555 if (error == 0) { 556 if (tlen != PAGE_SIZE) 557 bzero(ma + tlen, PAGE_SIZE - tlen); 558 uio->uio_offset += tlen; 559 uio->uio_resid -= tlen; 560 } 561 sf_buf_free(sf); 562 sched_unpin(); 563 VM_OBJECT_LOCK(vobj); 564 if (error == 0) 565 m->valid = VM_PAGE_BITS_ALL; 566 vm_page_wakeup(m); 567 VM_OBJECT_UNLOCK(vobj); 568 return (error); 569 } 570 VM_OBJECT_UNLOCK(vobj); 571nocache: 572 error = tmpfs_nocacheread(tobj, idx, offset, tlen, uio); 573 574 return (error); 575} 576 577static int 578tmpfs_read(struct vop_read_args *v) 579{ 580 struct vnode *vp = v->a_vp; 581 struct uio *uio = v->a_uio; 582 583 struct tmpfs_node *node; 584 vm_object_t uobj; 585 size_t len; 586 int resid; 587 588 int error = 0; 589 590 node = VP_TO_TMPFS_NODE(vp); 591 592 if (vp->v_type != VREG) { 593 error = EISDIR; 594 goto out; 595 } 596 597 if (uio->uio_offset < 0) { 598 error = EINVAL; 599 goto out; 600 } 601 602 node->tn_status |= TMPFS_NODE_ACCESSED; 603 604 uobj = node->tn_reg.tn_aobj; 605 while ((resid = uio->uio_resid) > 0) { 606 error = 0; 607 if (node->tn_size <= uio->uio_offset) 608 break; 609 len = MIN(node->tn_size - uio->uio_offset, resid); 610 if (len == 0) 611 break; 612 error = tmpfs_mappedread(vp->v_object, uobj, len, uio); 613 if ((error != 0) || (resid == uio->uio_resid)) 614 break; 615 } 616 617out: 618 619 return error; 620} 621 622/* --------------------------------------------------------------------- */ 623 624static int 625tmpfs_mappedwrite(vm_object_t vobj, vm_object_t tobj, size_t len, struct uio *uio) 626{ 627 vm_pindex_t idx; 628 vm_page_t vpg, tpg; 629 vm_offset_t offset; 630 off_t addr; 631 size_t tlen; 632 int error, rv; 633 634 error = 0; 635 636 addr = uio->uio_offset; 637 idx = OFF_TO_IDX(addr); 638 offset = addr & PAGE_MASK; 639 tlen = MIN(PAGE_SIZE - offset, len); 640 641 if ((vobj == NULL) || 642 (vobj->resident_page_count == 0 && vobj->cache == NULL)) { 643 vpg = NULL; 644 goto nocache; 645 } 646 647 VM_OBJECT_LOCK(vobj); 648lookupvpg: 649 if (((vpg = vm_page_lookup(vobj, idx)) != NULL) && 650 vm_page_is_valid(vpg, offset, tlen)) { 651 if ((vpg->oflags & VPO_BUSY) != 0) { 652 /* 653 * Reference the page before unlocking and sleeping so 654 * that the page daemon is less likely to reclaim it. 655 */ 656 vm_page_reference(vpg); 657 vm_page_sleep(vpg, "tmfsmw"); 658 goto lookupvpg; 659 } 660 vm_page_busy(vpg); 661 vm_page_undirty(vpg); 662 VM_OBJECT_UNLOCK(vobj); 663 error = uiomove_fromphys(&vpg, offset, tlen, uio); 664 } else { 665 if (__predict_false(vobj->cache != NULL)) 666 vm_page_cache_free(vobj, idx, idx + 1); 667 VM_OBJECT_UNLOCK(vobj); 668 vpg = NULL; 669 } 670nocache: 671 VM_OBJECT_LOCK(tobj); 672 tpg = vm_page_grab(tobj, idx, VM_ALLOC_WIRED | 673 VM_ALLOC_NORMAL | VM_ALLOC_RETRY); 674 if (tpg->valid != VM_PAGE_BITS_ALL) { 675 if (vm_pager_has_page(tobj, idx, NULL, NULL)) { 676 rv = vm_pager_get_pages(tobj, &tpg, 1, 0); 677 if (rv != VM_PAGER_OK) { 678 vm_page_lock(tpg); 679 vm_page_free(tpg); 680 vm_page_unlock(tpg); 681 error = EIO; 682 goto out; 683 } 684 } else 685 vm_page_zero_invalid(tpg, TRUE); 686 } 687 VM_OBJECT_UNLOCK(tobj); 688 if (vpg == NULL) 689 error = uiomove_fromphys(&tpg, offset, tlen, uio); 690 else { 691 KASSERT(vpg->valid == VM_PAGE_BITS_ALL, ("parts of vpg invalid")); 692 pmap_copy_page(vpg, tpg); 693 } 694 VM_OBJECT_LOCK(tobj); 695 if (error == 0) { 696 KASSERT(tpg->valid == VM_PAGE_BITS_ALL, 697 ("parts of tpg invalid")); 698 vm_page_dirty(tpg); 699 } 700 vm_page_lock(tpg); 701 vm_page_unwire(tpg, TRUE); 702 vm_page_unlock(tpg); 703 vm_page_wakeup(tpg); 704out: 705 VM_OBJECT_UNLOCK(tobj); 706 if (vpg != NULL) { 707 VM_OBJECT_LOCK(vobj); 708 vm_page_wakeup(vpg); 709 VM_OBJECT_UNLOCK(vobj); 710 } 711 712 return (error); 713} 714 715static int 716tmpfs_write(struct vop_write_args *v) 717{ 718 struct vnode *vp = v->a_vp; 719 struct uio *uio = v->a_uio; 720 int ioflag = v->a_ioflag; 721 722 boolean_t extended; 723 int error = 0; 724 off_t oldsize; 725 struct tmpfs_node *node; 726 vm_object_t uobj; 727 size_t len; 728 int resid; 729 730 node = VP_TO_TMPFS_NODE(vp); 731 oldsize = node->tn_size; 732 733 if (uio->uio_offset < 0 || vp->v_type != VREG) { 734 error = EINVAL; 735 goto out; 736 } 737 738 if (uio->uio_resid == 0) { 739 error = 0; 740 goto out; 741 } 742 743 if (ioflag & IO_APPEND) 744 uio->uio_offset = node->tn_size; 745 746 if (uio->uio_offset + uio->uio_resid > 747 VFS_TO_TMPFS(vp->v_mount)->tm_maxfilesize) 748 return (EFBIG); 749 750 if (vn_rlimit_fsize(vp, uio, uio->uio_td)) 751 return (EFBIG); 752 753 extended = uio->uio_offset + uio->uio_resid > node->tn_size; 754 if (extended) { 755 error = tmpfs_reg_resize(vp, uio->uio_offset + uio->uio_resid, 756 FALSE); 757 if (error != 0) 758 goto out; 759 } 760 761 uobj = node->tn_reg.tn_aobj; 762 while ((resid = uio->uio_resid) > 0) { 763 if (node->tn_size <= uio->uio_offset) 764 break; 765 len = MIN(node->tn_size - uio->uio_offset, resid); 766 if (len == 0) 767 break; 768 error = tmpfs_mappedwrite(vp->v_object, uobj, len, uio); 769 if ((error != 0) || (resid == uio->uio_resid)) 770 break; 771 } 772 773 node->tn_status |= TMPFS_NODE_ACCESSED | TMPFS_NODE_MODIFIED | 774 (extended ? TMPFS_NODE_CHANGED : 0); 775 776 if (node->tn_mode & (S_ISUID | S_ISGID)) { 777 if (priv_check_cred(v->a_cred, PRIV_VFS_RETAINSUGID, 0)) 778 node->tn_mode &= ~(S_ISUID | S_ISGID); 779 } 780 781 if (error != 0) 782 (void)tmpfs_reg_resize(vp, oldsize, TRUE); 783 784out: 785 MPASS(IMPLIES(error == 0, uio->uio_resid == 0)); 786 MPASS(IMPLIES(error != 0, oldsize == node->tn_size)); 787 788 return error; 789} 790 791/* --------------------------------------------------------------------- */ 792 793static int 794tmpfs_fsync(struct vop_fsync_args *v) 795{ 796 struct vnode *vp = v->a_vp; 797 798 MPASS(VOP_ISLOCKED(vp)); 799 800 tmpfs_update(vp); 801 802 return 0; 803} 804 805/* --------------------------------------------------------------------- */ 806 807static int 808tmpfs_remove(struct vop_remove_args *v) 809{ 810 struct vnode *dvp = v->a_dvp; 811 struct vnode *vp = v->a_vp; 812 813 int error; 814 struct tmpfs_dirent *de; 815 struct tmpfs_mount *tmp; 816 struct tmpfs_node *dnode; 817 struct tmpfs_node *node; 818 819 MPASS(VOP_ISLOCKED(dvp)); 820 MPASS(VOP_ISLOCKED(vp)); 821 822 if (vp->v_type == VDIR) { 823 error = EISDIR; 824 goto out; 825 } 826 827 dnode = VP_TO_TMPFS_DIR(dvp); 828 node = VP_TO_TMPFS_NODE(vp); 829 tmp = VFS_TO_TMPFS(vp->v_mount); 830 de = tmpfs_dir_lookup(dnode, node, v->a_cnp); 831 MPASS(de != NULL); 832 833 /* Files marked as immutable or append-only cannot be deleted. */ 834 if ((node->tn_flags & (IMMUTABLE | APPEND | NOUNLINK)) || 835 (dnode->tn_flags & APPEND)) { 836 error = EPERM; 837 goto out; 838 } 839 840 /* Remove the entry from the directory; as it is a file, we do not 841 * have to change the number of hard links of the directory. */ 842 tmpfs_dir_detach(dvp, de); 843 if (v->a_cnp->cn_flags & DOWHITEOUT) 844 tmpfs_dir_whiteout_add(dvp, v->a_cnp); 845 846 /* Free the directory entry we just deleted. Note that the node 847 * referred by it will not be removed until the vnode is really 848 * reclaimed. */ 849 tmpfs_free_dirent(tmp, de, TRUE); 850 851 node->tn_status |= TMPFS_NODE_ACCESSED | TMPFS_NODE_CHANGED; 852 error = 0; 853 854out: 855 856 return error; 857} 858 859/* --------------------------------------------------------------------- */ 860 861static int 862tmpfs_link(struct vop_link_args *v) 863{ 864 struct vnode *dvp = v->a_tdvp; 865 struct vnode *vp = v->a_vp; 866 struct componentname *cnp = v->a_cnp; 867 868 int error; 869 struct tmpfs_dirent *de; 870 struct tmpfs_node *node; 871 872 MPASS(VOP_ISLOCKED(dvp)); 873 MPASS(cnp->cn_flags & HASBUF); 874 MPASS(dvp != vp); /* XXX When can this be false? */ 875 876 node = VP_TO_TMPFS_NODE(vp); 877 878 /* XXX: Why aren't the following two tests done by the caller? */ 879 880 /* Hard links of directories are forbidden. */ 881 if (vp->v_type == VDIR) { 882 error = EPERM; 883 goto out; 884 } 885 886 /* Cannot create cross-device links. */ 887 if (dvp->v_mount != vp->v_mount) { 888 error = EXDEV; 889 goto out; 890 } 891 892 /* Ensure that we do not overflow the maximum number of links imposed 893 * by the system. */ 894 MPASS(node->tn_links <= LINK_MAX); 895 if (node->tn_links == LINK_MAX) { 896 error = EMLINK; 897 goto out; 898 } 899 900 /* We cannot create links of files marked immutable or append-only. */ 901 if (node->tn_flags & (IMMUTABLE | APPEND)) { 902 error = EPERM; 903 goto out; 904 } 905 906 /* Allocate a new directory entry to represent the node. */ 907 error = tmpfs_alloc_dirent(VFS_TO_TMPFS(vp->v_mount), node, 908 cnp->cn_nameptr, cnp->cn_namelen, &de); 909 if (error != 0) 910 goto out; 911 912 /* Insert the new directory entry into the appropriate directory. */ 913 if (cnp->cn_flags & ISWHITEOUT) 914 tmpfs_dir_whiteout_remove(dvp, cnp); 915 tmpfs_dir_attach(dvp, de); 916 917 /* vp link count has changed, so update node times. */ 918 node->tn_status |= TMPFS_NODE_CHANGED; 919 tmpfs_update(vp); 920 921 error = 0; 922 923out: 924 return error; 925} 926 927/* --------------------------------------------------------------------- */ 928 929/* 930 * We acquire all but fdvp locks using non-blocking acquisitions. If we 931 * fail to acquire any lock in the path we will drop all held locks, 932 * acquire the new lock in a blocking fashion, and then release it and 933 * restart the rename. This acquire/release step ensures that we do not 934 * spin on a lock waiting for release. On error release all vnode locks 935 * and decrement references the way tmpfs_rename() would do. 936 */ 937static int 938tmpfs_rename_relock(struct vnode *fdvp, struct vnode **fvpp, 939 struct vnode *tdvp, struct vnode **tvpp, 940 struct componentname *fcnp, struct componentname *tcnp) 941{ 942 struct vnode *nvp; 943 struct mount *mp; 944 struct tmpfs_dirent *de; 945 int error, restarts = 0; 946 947 VOP_UNLOCK(tdvp, 0); 948 if (*tvpp != NULL && *tvpp != tdvp) 949 VOP_UNLOCK(*tvpp, 0); 950 mp = fdvp->v_mount; 951 952relock: 953 restarts += 1; 954 error = vn_lock(fdvp, LK_EXCLUSIVE); 955 if (error) 956 goto releout; 957 if (vn_lock(tdvp, LK_EXCLUSIVE | LK_NOWAIT) != 0) { 958 VOP_UNLOCK(fdvp, 0); 959 error = vn_lock(tdvp, LK_EXCLUSIVE); 960 if (error) 961 goto releout; 962 VOP_UNLOCK(tdvp, 0); 963 goto relock; 964 } 965 /* 966 * Re-resolve fvp to be certain it still exists and fetch the 967 * correct vnode. 968 */ 969 de = tmpfs_dir_lookup(VP_TO_TMPFS_DIR(fdvp), NULL, fcnp); 970 if (de == NULL) { 971 VOP_UNLOCK(fdvp, 0); 972 VOP_UNLOCK(tdvp, 0); 973 if ((fcnp->cn_flags & ISDOTDOT) != 0 || 974 (fcnp->cn_namelen == 1 && fcnp->cn_nameptr[0] == '.')) 975 error = EINVAL; 976 else 977 error = ENOENT; 978 goto releout; 979 } 980 error = tmpfs_alloc_vp(mp, de->td_node, LK_EXCLUSIVE | LK_NOWAIT, &nvp); 981 if (error != 0) { 982 VOP_UNLOCK(fdvp, 0); 983 VOP_UNLOCK(tdvp, 0); 984 if (error != EBUSY) 985 goto releout; 986 error = tmpfs_alloc_vp(mp, de->td_node, LK_EXCLUSIVE, &nvp); 987 if (error != 0) 988 goto releout; 989 VOP_UNLOCK(nvp, 0); 990 /* 991 * Concurrent rename race. 992 */ 993 if (nvp == tdvp) { 994 vrele(nvp); 995 error = EINVAL; 996 goto releout; 997 } 998 vrele(*fvpp); 999 *fvpp = nvp; 1000 goto relock; 1001 } 1002 vrele(*fvpp); 1003 *fvpp = nvp; 1004 VOP_UNLOCK(*fvpp, 0); 1005 /* 1006 * Re-resolve tvp and acquire the vnode lock if present. 1007 */ 1008 de = tmpfs_dir_lookup(VP_TO_TMPFS_DIR(tdvp), NULL, tcnp); 1009 /* 1010 * If tvp disappeared we just carry on. 1011 */ 1012 if (de == NULL && *tvpp != NULL) { 1013 vrele(*tvpp); 1014 *tvpp = NULL; 1015 } 1016 /* 1017 * Get the tvp ino if the lookup succeeded. We may have to restart 1018 * if the non-blocking acquire fails. 1019 */ 1020 if (de != NULL) { 1021 nvp = NULL; 1022 error = tmpfs_alloc_vp(mp, de->td_node, 1023 LK_EXCLUSIVE | LK_NOWAIT, &nvp); 1024 if (*tvpp != NULL) 1025 vrele(*tvpp); 1026 *tvpp = nvp; 1027 if (error != 0) { 1028 VOP_UNLOCK(fdvp, 0); 1029 VOP_UNLOCK(tdvp, 0); 1030 if (error != EBUSY) 1031 goto releout; 1032 error = tmpfs_alloc_vp(mp, de->td_node, LK_EXCLUSIVE, 1033 &nvp); 1034 if (error != 0) 1035 goto releout; 1036 VOP_UNLOCK(nvp, 0); 1037 /* 1038 * fdvp contains fvp, thus tvp (=fdvp) is not empty. 1039 */ 1040 if (nvp == fdvp) { 1041 error = ENOTEMPTY; 1042 goto releout; 1043 } 1044 goto relock; 1045 } 1046 } 1047 tmpfs_rename_restarts += restarts; 1048 1049 return (0); 1050 1051releout: 1052 vrele(fdvp); 1053 vrele(*fvpp); 1054 vrele(tdvp); 1055 if (*tvpp != NULL) 1056 vrele(*tvpp); 1057 tmpfs_rename_restarts += restarts; 1058 1059 return (error); 1060} 1061 1062static int 1063tmpfs_rename(struct vop_rename_args *v) 1064{ 1065 struct vnode *fdvp = v->a_fdvp; 1066 struct vnode *fvp = v->a_fvp; 1067 struct componentname *fcnp = v->a_fcnp; 1068 struct vnode *tdvp = v->a_tdvp; 1069 struct vnode *tvp = v->a_tvp; 1070 struct componentname *tcnp = v->a_tcnp; 1071 struct mount *mp = NULL; 1072 1073 char *newname; 1074 int error; 1075 struct tmpfs_dirent *de; 1076 struct tmpfs_mount *tmp; 1077 struct tmpfs_node *fdnode; 1078 struct tmpfs_node *fnode; 1079 struct tmpfs_node *tnode; 1080 struct tmpfs_node *tdnode; 1081 1082 MPASS(VOP_ISLOCKED(tdvp)); 1083 MPASS(IMPLIES(tvp != NULL, VOP_ISLOCKED(tvp))); 1084 MPASS(fcnp->cn_flags & HASBUF); 1085 MPASS(tcnp->cn_flags & HASBUF); 1086 1087 /* Disallow cross-device renames. 1088 * XXX Why isn't this done by the caller? */ 1089 if (fvp->v_mount != tdvp->v_mount || 1090 (tvp != NULL && fvp->v_mount != tvp->v_mount)) { 1091 error = EXDEV; 1092 goto out; 1093 } 1094 1095 /* If source and target are the same file, there is nothing to do. */ 1096 if (fvp == tvp) { 1097 error = 0; 1098 goto out; 1099 } 1100 1101 /* If we need to move the directory between entries, lock the 1102 * source so that we can safely operate on it. */ 1103 if (fdvp != tdvp && fdvp != tvp) { 1104 if (vn_lock(fdvp, LK_EXCLUSIVE | LK_NOWAIT) != 0) { 1105 mp = tdvp->v_mount; 1106 error = vfs_busy(mp, 0); 1107 if (error != 0) { 1108 mp = NULL; 1109 goto out; 1110 } 1111 error = tmpfs_rename_relock(fdvp, &fvp, tdvp, &tvp, 1112 fcnp, tcnp); 1113 if (error != 0) { 1114 vfs_unbusy(mp); 1115 return (error); 1116 } 1117 ASSERT_VOP_ELOCKED(fdvp, 1118 "tmpfs_rename: fdvp not locked"); 1119 ASSERT_VOP_ELOCKED(tdvp, 1120 "tmpfs_rename: tdvp not locked"); 1121 if (tvp != NULL) 1122 ASSERT_VOP_ELOCKED(tvp, 1123 "tmpfs_rename: tvp not locked"); 1124 if (fvp == tvp) { 1125 error = 0; 1126 goto out_locked; 1127 } 1128 } 1129 } 1130 1131 tmp = VFS_TO_TMPFS(tdvp->v_mount); 1132 tdnode = VP_TO_TMPFS_DIR(tdvp); 1133 tnode = (tvp == NULL) ? NULL : VP_TO_TMPFS_NODE(tvp); 1134 fdnode = VP_TO_TMPFS_DIR(fdvp); 1135 fnode = VP_TO_TMPFS_NODE(fvp); 1136 de = tmpfs_dir_lookup(fdnode, fnode, fcnp); 1137 1138 /* Entry can disappear before we lock fdvp, 1139 * also avoid manipulating '.' and '..' entries. */ 1140 if (de == NULL) { 1141 if ((fcnp->cn_flags & ISDOTDOT) != 0 || 1142 (fcnp->cn_namelen == 1 && fcnp->cn_nameptr[0] == '.')) 1143 error = EINVAL; 1144 else 1145 error = ENOENT; 1146 goto out_locked; 1147 } 1148 MPASS(de->td_node == fnode); 1149 1150 /* If re-naming a directory to another preexisting directory 1151 * ensure that the target directory is empty so that its 1152 * removal causes no side effects. 1153 * Kern_rename gurantees the destination to be a directory 1154 * if the source is one. */ 1155 if (tvp != NULL) { 1156 MPASS(tnode != NULL); 1157 1158 if ((tnode->tn_flags & (NOUNLINK | IMMUTABLE | APPEND)) || 1159 (tdnode->tn_flags & (APPEND | IMMUTABLE))) { 1160 error = EPERM; 1161 goto out_locked; 1162 } 1163 1164 if (fnode->tn_type == VDIR && tnode->tn_type == VDIR) { 1165 if (tnode->tn_size > 0) { 1166 error = ENOTEMPTY; 1167 goto out_locked; 1168 } 1169 } else if (fnode->tn_type == VDIR && tnode->tn_type != VDIR) { 1170 error = ENOTDIR; 1171 goto out_locked; 1172 } else if (fnode->tn_type != VDIR && tnode->tn_type == VDIR) { 1173 error = EISDIR; 1174 goto out_locked; 1175 } else { 1176 MPASS(fnode->tn_type != VDIR && 1177 tnode->tn_type != VDIR); 1178 } 1179 } 1180 1181 if ((fnode->tn_flags & (NOUNLINK | IMMUTABLE | APPEND)) 1182 || (fdnode->tn_flags & (APPEND | IMMUTABLE))) { 1183 error = EPERM; 1184 goto out_locked; 1185 } 1186 1187 /* Ensure that we have enough memory to hold the new name, if it 1188 * has to be changed. */ 1189 if (fcnp->cn_namelen != tcnp->cn_namelen || 1190 bcmp(fcnp->cn_nameptr, tcnp->cn_nameptr, fcnp->cn_namelen) != 0) { 1191 newname = malloc(tcnp->cn_namelen, M_TMPFSNAME, M_WAITOK); 1192 } else 1193 newname = NULL; 1194 1195 /* If the node is being moved to another directory, we have to do 1196 * the move. */ 1197 if (fdnode != tdnode) { 1198 /* In case we are moving a directory, we have to adjust its 1199 * parent to point to the new parent. */ 1200 if (de->td_node->tn_type == VDIR) { 1201 struct tmpfs_node *n; 1202 1203 /* Ensure the target directory is not a child of the 1204 * directory being moved. Otherwise, we'd end up 1205 * with stale nodes. */ 1206 n = tdnode; 1207 /* TMPFS_LOCK garanties that no nodes are freed while 1208 * traversing the list. Nodes can only be marked as 1209 * removed: tn_parent == NULL. */ 1210 TMPFS_LOCK(tmp); 1211 TMPFS_NODE_LOCK(n); 1212 while (n != n->tn_dir.tn_parent) { 1213 struct tmpfs_node *parent; 1214 1215 if (n == fnode) { 1216 TMPFS_NODE_UNLOCK(n); 1217 TMPFS_UNLOCK(tmp); 1218 error = EINVAL; 1219 if (newname != NULL) 1220 free(newname, M_TMPFSNAME); 1221 goto out_locked; 1222 } 1223 parent = n->tn_dir.tn_parent; 1224 TMPFS_NODE_UNLOCK(n); 1225 if (parent == NULL) { 1226 n = NULL; 1227 break; 1228 } 1229 TMPFS_NODE_LOCK(parent); 1230 if (parent->tn_dir.tn_parent == NULL) { 1231 TMPFS_NODE_UNLOCK(parent); 1232 n = NULL; 1233 break; 1234 } 1235 n = parent; 1236 } 1237 TMPFS_UNLOCK(tmp); 1238 if (n == NULL) { 1239 error = EINVAL; 1240 if (newname != NULL) 1241 free(newname, M_TMPFSNAME); 1242 goto out_locked; 1243 } 1244 TMPFS_NODE_UNLOCK(n); 1245 1246 /* Adjust the parent pointer. */ 1247 TMPFS_VALIDATE_DIR(fnode); 1248 TMPFS_NODE_LOCK(de->td_node); 1249 de->td_node->tn_dir.tn_parent = tdnode; 1250 TMPFS_NODE_UNLOCK(de->td_node); 1251 1252 /* As a result of changing the target of the '..' 1253 * entry, the link count of the source and target 1254 * directories has to be adjusted. */ 1255 TMPFS_NODE_LOCK(tdnode); 1256 TMPFS_ASSERT_LOCKED(tdnode); 1257 tdnode->tn_links++; 1258 TMPFS_NODE_UNLOCK(tdnode); 1259 1260 TMPFS_NODE_LOCK(fdnode); 1261 TMPFS_ASSERT_LOCKED(fdnode); 1262 fdnode->tn_links--; 1263 TMPFS_NODE_UNLOCK(fdnode); 1264 } 1265 1266 /* Do the move: just remove the entry from the source directory 1267 * and insert it into the target one. */ 1268 tmpfs_dir_detach(fdvp, de); 1269 if (fcnp->cn_flags & DOWHITEOUT) 1270 tmpfs_dir_whiteout_add(fdvp, fcnp); 1271 if (tcnp->cn_flags & ISWHITEOUT) 1272 tmpfs_dir_whiteout_remove(tdvp, tcnp); 1273 tmpfs_dir_attach(tdvp, de); 1274 } 1275 1276 /* If the name has changed, we need to make it effective by changing 1277 * it in the directory entry. */ 1278 if (newname != NULL) { 1279 MPASS(tcnp->cn_namelen <= MAXNAMLEN); 1280 1281 free(de->td_name, M_TMPFSNAME); 1282 de->td_namelen = (uint16_t)tcnp->cn_namelen; 1283 memcpy(newname, tcnp->cn_nameptr, tcnp->cn_namelen); 1284 de->td_name = newname; 1285 1286 fnode->tn_status |= TMPFS_NODE_CHANGED; 1287 tdnode->tn_status |= TMPFS_NODE_MODIFIED; 1288 } 1289 1290 /* If we are overwriting an entry, we have to remove the old one 1291 * from the target directory. */ 1292 if (tvp != NULL) { 1293 /* Remove the old entry from the target directory. */ 1294 de = tmpfs_dir_lookup(tdnode, tnode, tcnp); 1295 tmpfs_dir_detach(tdvp, de); 1296 1297 /* Free the directory entry we just deleted. Note that the 1298 * node referred by it will not be removed until the vnode is 1299 * really reclaimed. */ 1300 tmpfs_free_dirent(VFS_TO_TMPFS(tvp->v_mount), de, TRUE); 1301 } 1302 cache_purge(fvp); 1303 if (tvp != NULL) 1304 cache_purge(tvp); 1305 1306 error = 0; 1307 1308out_locked: 1309 if (fdvp != tdvp && fdvp != tvp) 1310 VOP_UNLOCK(fdvp, 0); 1311 1312out: 1313 /* Release target nodes. */ 1314 /* XXX: I don't understand when tdvp can be the same as tvp, but 1315 * other code takes care of this... */ 1316 if (tdvp == tvp) 1317 vrele(tdvp); 1318 else 1319 vput(tdvp); 1320 if (tvp != NULL) 1321 vput(tvp); 1322 1323 /* Release source nodes. */ 1324 vrele(fdvp); 1325 vrele(fvp); 1326 1327 if (mp != NULL) 1328 vfs_unbusy(mp); 1329 1330 return error; 1331} 1332 1333/* --------------------------------------------------------------------- */ 1334 1335static int 1336tmpfs_mkdir(struct vop_mkdir_args *v) 1337{ 1338 struct vnode *dvp = v->a_dvp; 1339 struct vnode **vpp = v->a_vpp; 1340 struct componentname *cnp = v->a_cnp; 1341 struct vattr *vap = v->a_vap; 1342 1343 MPASS(vap->va_type == VDIR); 1344 1345 return tmpfs_alloc_file(dvp, vpp, vap, cnp, NULL); 1346} 1347 1348/* --------------------------------------------------------------------- */ 1349 1350static int 1351tmpfs_rmdir(struct vop_rmdir_args *v) 1352{ 1353 struct vnode *dvp = v->a_dvp; 1354 struct vnode *vp = v->a_vp; 1355 1356 int error; 1357 struct tmpfs_dirent *de; 1358 struct tmpfs_mount *tmp; 1359 struct tmpfs_node *dnode; 1360 struct tmpfs_node *node; 1361 1362 MPASS(VOP_ISLOCKED(dvp)); 1363 MPASS(VOP_ISLOCKED(vp)); 1364 1365 tmp = VFS_TO_TMPFS(dvp->v_mount); 1366 dnode = VP_TO_TMPFS_DIR(dvp); 1367 node = VP_TO_TMPFS_DIR(vp); 1368 1369 /* Directories with more than two entries ('.' and '..') cannot be 1370 * removed. */ 1371 if (node->tn_size > 0) { 1372 error = ENOTEMPTY; 1373 goto out; 1374 } 1375 1376 if ((dnode->tn_flags & APPEND) 1377 || (node->tn_flags & (NOUNLINK | IMMUTABLE | APPEND))) { 1378 error = EPERM; 1379 goto out; 1380 } 1381 1382 /* This invariant holds only if we are not trying to remove "..". 1383 * We checked for that above so this is safe now. */ 1384 MPASS(node->tn_dir.tn_parent == dnode); 1385 1386 /* Get the directory entry associated with node (vp). This was 1387 * filled by tmpfs_lookup while looking up the entry. */ 1388 de = tmpfs_dir_lookup(dnode, node, v->a_cnp); 1389 MPASS(TMPFS_DIRENT_MATCHES(de, 1390 v->a_cnp->cn_nameptr, 1391 v->a_cnp->cn_namelen)); 1392 1393 /* Check flags to see if we are allowed to remove the directory. */ 1394 if (dnode->tn_flags & APPEND 1395 || node->tn_flags & (NOUNLINK | IMMUTABLE | APPEND)) { 1396 error = EPERM; 1397 goto out; 1398 } 1399 1400 1401 /* Detach the directory entry from the directory (dnode). */ 1402 tmpfs_dir_detach(dvp, de); 1403 if (v->a_cnp->cn_flags & DOWHITEOUT) 1404 tmpfs_dir_whiteout_add(dvp, v->a_cnp); 1405 1406 /* No vnode should be allocated for this entry from this point */ 1407 TMPFS_NODE_LOCK(node); 1408 TMPFS_ASSERT_ELOCKED(node); 1409 node->tn_links--; 1410 node->tn_dir.tn_parent = NULL; 1411 node->tn_status |= TMPFS_NODE_ACCESSED | TMPFS_NODE_CHANGED | \ 1412 TMPFS_NODE_MODIFIED; 1413 1414 TMPFS_NODE_UNLOCK(node); 1415 1416 TMPFS_NODE_LOCK(dnode); 1417 TMPFS_ASSERT_ELOCKED(dnode); 1418 dnode->tn_links--; 1419 dnode->tn_status |= TMPFS_NODE_ACCESSED | \ 1420 TMPFS_NODE_CHANGED | TMPFS_NODE_MODIFIED; 1421 TMPFS_NODE_UNLOCK(dnode); 1422 1423 cache_purge(dvp); 1424 cache_purge(vp); 1425 1426 /* Free the directory entry we just deleted. Note that the node 1427 * referred by it will not be removed until the vnode is really 1428 * reclaimed. */ 1429 tmpfs_free_dirent(tmp, de, TRUE); 1430 1431 /* Release the deleted vnode (will destroy the node, notify 1432 * interested parties and clean it from the cache). */ 1433 1434 dnode->tn_status |= TMPFS_NODE_CHANGED; 1435 tmpfs_update(dvp); 1436 1437 error = 0; 1438 1439out: 1440 return error; 1441} 1442 1443/* --------------------------------------------------------------------- */ 1444 1445static int 1446tmpfs_symlink(struct vop_symlink_args *v) 1447{ 1448 struct vnode *dvp = v->a_dvp; 1449 struct vnode **vpp = v->a_vpp; 1450 struct componentname *cnp = v->a_cnp; 1451 struct vattr *vap = v->a_vap; 1452 char *target = v->a_target; 1453 1454#ifdef notyet /* XXX FreeBSD BUG: kern_symlink is not setting VLNK */ 1455 MPASS(vap->va_type == VLNK); 1456#else 1457 vap->va_type = VLNK; 1458#endif 1459 1460 return tmpfs_alloc_file(dvp, vpp, vap, cnp, target); 1461} 1462 1463/* --------------------------------------------------------------------- */ 1464 1465static int 1466tmpfs_readdir(struct vop_readdir_args *v) 1467{ 1468 struct vnode *vp = v->a_vp; 1469 struct uio *uio = v->a_uio; 1470 int *eofflag = v->a_eofflag; 1471 u_long **cookies = v->a_cookies; 1472 int *ncookies = v->a_ncookies; 1473 1474 int error; 1475 off_t startoff; 1476 off_t cnt = 0; 1477 struct tmpfs_node *node; 1478 1479 /* This operation only makes sense on directory nodes. */ 1480 if (vp->v_type != VDIR) 1481 return ENOTDIR; 1482 1483 node = VP_TO_TMPFS_DIR(vp); 1484 1485 startoff = uio->uio_offset; 1486 1487 if (uio->uio_offset == TMPFS_DIRCOOKIE_DOT) { 1488 error = tmpfs_dir_getdotdent(node, uio); 1489 if (error != 0) 1490 goto outok; 1491 cnt++; 1492 } 1493 1494 if (uio->uio_offset == TMPFS_DIRCOOKIE_DOTDOT) { 1495 error = tmpfs_dir_getdotdotdent(node, uio); 1496 if (error != 0) 1497 goto outok; 1498 cnt++; 1499 } 1500 1501 error = tmpfs_dir_getdents(node, uio, &cnt); 1502 1503outok: 1504 MPASS(error >= -1); 1505 1506 if (error == -1) 1507 error = (cnt != 0) ? 0 : EINVAL; 1508 1509 if (eofflag != NULL) 1510 *eofflag = 1511 (error == 0 && uio->uio_offset == TMPFS_DIRCOOKIE_EOF); 1512 1513 /* Update NFS-related variables. */ 1514 if (error == 0 && cookies != NULL && ncookies != NULL) { 1515 off_t i; 1516 off_t off = startoff; 1517 struct tmpfs_dirent *de = NULL; 1518 1519 *ncookies = cnt; 1520 *cookies = malloc(cnt * sizeof(off_t), M_TEMP, M_WAITOK); 1521 1522 for (i = 0; i < cnt; i++) { 1523 MPASS(off != TMPFS_DIRCOOKIE_EOF); 1524 if (off == TMPFS_DIRCOOKIE_DOT) { 1525 off = TMPFS_DIRCOOKIE_DOTDOT; 1526 } else { 1527 if (off == TMPFS_DIRCOOKIE_DOTDOT) { 1528 de = TAILQ_FIRST(&node->tn_dir.tn_dirhead); 1529 } else if (de != NULL) { 1530 de = TAILQ_NEXT(de, td_entries); 1531 } else { 1532 de = tmpfs_dir_lookupbycookie(node, 1533 off); 1534 MPASS(de != NULL); 1535 de = TAILQ_NEXT(de, td_entries); 1536 } 1537 if (de == NULL) 1538 off = TMPFS_DIRCOOKIE_EOF; 1539 else 1540 off = tmpfs_dircookie(de); 1541 } 1542 1543 (*cookies)[i] = off; 1544 } 1545 MPASS(uio->uio_offset == off); 1546 } 1547 1548 return error; 1549} 1550 1551/* --------------------------------------------------------------------- */ 1552 1553static int 1554tmpfs_readlink(struct vop_readlink_args *v) 1555{ 1556 struct vnode *vp = v->a_vp; 1557 struct uio *uio = v->a_uio; 1558 1559 int error; 1560 struct tmpfs_node *node; 1561 1562 MPASS(uio->uio_offset == 0); 1563 MPASS(vp->v_type == VLNK); 1564 1565 node = VP_TO_TMPFS_NODE(vp); 1566 1567 error = uiomove(node->tn_link, MIN(node->tn_size, uio->uio_resid), 1568 uio); 1569 node->tn_status |= TMPFS_NODE_ACCESSED; 1570 1571 return error; 1572} 1573 1574/* --------------------------------------------------------------------- */ 1575 1576static int 1577tmpfs_inactive(struct vop_inactive_args *v) 1578{ 1579 struct vnode *vp = v->a_vp; 1580 struct thread *l = v->a_td; 1581 1582 struct tmpfs_node *node; 1583 1584 MPASS(VOP_ISLOCKED(vp)); 1585 1586 node = VP_TO_TMPFS_NODE(vp); 1587 1588 if (node->tn_links == 0) 1589 vrecycle(vp, l); 1590 1591 return 0; 1592} 1593 1594/* --------------------------------------------------------------------- */ 1595 1596int 1597tmpfs_reclaim(struct vop_reclaim_args *v) 1598{ 1599 struct vnode *vp = v->a_vp; 1600 1601 struct tmpfs_mount *tmp; 1602 struct tmpfs_node *node; 1603 1604 node = VP_TO_TMPFS_NODE(vp); 1605 tmp = VFS_TO_TMPFS(vp->v_mount); 1606 1607 vnode_destroy_vobject(vp); 1608 cache_purge(vp); 1609 1610 TMPFS_NODE_LOCK(node); 1611 TMPFS_ASSERT_ELOCKED(node); 1612 tmpfs_free_vp(vp); 1613 1614 /* If the node referenced by this vnode was deleted by the user, 1615 * we must free its associated data structures (now that the vnode 1616 * is being reclaimed). */ 1617 if (node->tn_links == 0 && 1618 (node->tn_vpstate & TMPFS_VNODE_ALLOCATING) == 0) { 1619 node->tn_vpstate = TMPFS_VNODE_DOOMED; 1620 TMPFS_NODE_UNLOCK(node); 1621 tmpfs_free_node(tmp, node); 1622 } else 1623 TMPFS_NODE_UNLOCK(node); 1624 1625 MPASS(vp->v_data == NULL); 1626 return 0; 1627} 1628 1629/* --------------------------------------------------------------------- */ 1630 1631static int 1632tmpfs_print(struct vop_print_args *v) 1633{ 1634 struct vnode *vp = v->a_vp; 1635 1636 struct tmpfs_node *node; 1637 1638 node = VP_TO_TMPFS_NODE(vp); 1639 1640 printf("tag VT_TMPFS, tmpfs_node %p, flags 0x%x, links %d\n", 1641 node, node->tn_flags, node->tn_links); 1642 printf("\tmode 0%o, owner %d, group %d, size %jd, status 0x%x\n", 1643 node->tn_mode, node->tn_uid, node->tn_gid, 1644 (intmax_t)node->tn_size, node->tn_status); 1645 1646 if (vp->v_type == VFIFO) 1647 fifo_printinfo(vp); 1648 1649 printf("\n"); 1650 1651 return 0; 1652} 1653 1654/* --------------------------------------------------------------------- */ 1655 1656static int 1657tmpfs_pathconf(struct vop_pathconf_args *v) 1658{ 1659 int name = v->a_name; 1660 register_t *retval = v->a_retval; 1661 1662 int error; 1663 1664 error = 0; 1665 1666 switch (name) { 1667 case _PC_LINK_MAX: 1668 *retval = LINK_MAX; 1669 break; 1670 1671 case _PC_NAME_MAX: 1672 *retval = NAME_MAX; 1673 break; 1674 1675 case _PC_PATH_MAX: 1676 *retval = PATH_MAX; 1677 break; 1678 1679 case _PC_PIPE_BUF: 1680 *retval = PIPE_BUF; 1681 break; 1682 1683 case _PC_CHOWN_RESTRICTED: 1684 *retval = 1; 1685 break; 1686 1687 case _PC_NO_TRUNC: 1688 *retval = 1; 1689 break; 1690 1691 case _PC_SYNC_IO: 1692 *retval = 1; 1693 break; 1694 1695 case _PC_FILESIZEBITS: 1696 *retval = 0; /* XXX Don't know which value should I return. */ 1697 break; 1698 1699 default: 1700 error = EINVAL; 1701 } 1702 1703 return error; 1704} 1705 1706static int 1707tmpfs_vptofh(struct vop_vptofh_args *ap) 1708{ 1709 struct tmpfs_fid *tfhp; 1710 struct tmpfs_node *node; 1711 1712 tfhp = (struct tmpfs_fid *)ap->a_fhp; 1713 node = VP_TO_TMPFS_NODE(ap->a_vp); 1714 1715 tfhp->tf_len = sizeof(struct tmpfs_fid); 1716 tfhp->tf_id = node->tn_id; 1717 tfhp->tf_gen = node->tn_gen; 1718 1719 return (0); 1720} 1721 1722static int 1723tmpfs_whiteout(struct vop_whiteout_args *ap) 1724{ 1725 struct vnode *dvp = ap->a_dvp; 1726 struct componentname *cnp = ap->a_cnp; 1727 struct tmpfs_dirent *de; 1728 1729 switch (ap->a_flags) { 1730 case LOOKUP: 1731 return (0); 1732 case CREATE: 1733 de = tmpfs_dir_lookup(VP_TO_TMPFS_DIR(dvp), NULL, cnp); 1734 if (de != NULL) 1735 return (de->td_node == NULL ? 0 : EEXIST); 1736 return (tmpfs_dir_whiteout_add(dvp, cnp)); 1737 case DELETE: 1738 tmpfs_dir_whiteout_remove(dvp, cnp); 1739 return (0); 1740 default: 1741 panic("tmpfs_whiteout: unknown op"); 1742 } 1743} 1744 1745/* --------------------------------------------------------------------- */ 1746 1747/* 1748 * vnode operations vector used for files stored in a tmpfs file system. 1749 */ 1750struct vop_vector tmpfs_vnodeop_entries = { 1751 .vop_default = &default_vnodeops, 1752 .vop_lookup = vfs_cache_lookup, 1753 .vop_cachedlookup = tmpfs_lookup, 1754 .vop_create = tmpfs_create, 1755 .vop_mknod = tmpfs_mknod, 1756 .vop_open = tmpfs_open, 1757 .vop_close = tmpfs_close, 1758 .vop_access = tmpfs_access, 1759 .vop_getattr = tmpfs_getattr, 1760 .vop_setattr = tmpfs_setattr, 1761 .vop_read = tmpfs_read, 1762 .vop_write = tmpfs_write, 1763 .vop_fsync = tmpfs_fsync, 1764 .vop_remove = tmpfs_remove, 1765 .vop_link = tmpfs_link, 1766 .vop_rename = tmpfs_rename, 1767 .vop_mkdir = tmpfs_mkdir, 1768 .vop_rmdir = tmpfs_rmdir, 1769 .vop_symlink = tmpfs_symlink, 1770 .vop_readdir = tmpfs_readdir, 1771 .vop_readlink = tmpfs_readlink, 1772 .vop_inactive = tmpfs_inactive, 1773 .vop_reclaim = tmpfs_reclaim, 1774 .vop_print = tmpfs_print, 1775 .vop_pathconf = tmpfs_pathconf, 1776 .vop_vptofh = tmpfs_vptofh, 1777 .vop_whiteout = tmpfs_whiteout, 1778 .vop_bmap = VOP_EOPNOTSUPP, 1779}; 1780 1781