1 /* $NetBSD: tmpfs_vnops.c,v 1.39 2007/07/23 15:41:01 jmmv Exp $ */ 2 3 /*- 4 * SPDX-License-Identifier: BSD-2-Clause-NetBSD 5 * 6 * Copyright (c) 2005, 2006 The NetBSD Foundation, Inc. 7 * All rights reserved. 8 * 9 * This code is derived from software contributed to The NetBSD Foundation 10 * by Julio M. Merino Vidal, developed as part of Google's Summer of Code 11 * 2005 program. 12 * 13 * Redistribution and use in source and binary forms, with or without 14 * modification, are permitted provided that the following conditions 15 * are met: 16 * 1. Redistributions of source code must retain the above copyright 17 * notice, this list of conditions and the following disclaimer. 18 * 2. Redistributions in binary form must reproduce the above copyright 19 * notice, this list of conditions and the following disclaimer in the 20 * documentation and/or other materials provided with the distribution. 21 * 22 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 23 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 24 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 25 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 26 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 27 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 28 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 29 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 30 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 31 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 32 * POSSIBILITY OF SUCH DAMAGE. 33 */ 34 35 /* 36 * tmpfs vnode interface. 37 */ 38 #include <sys/cdefs.h> 39 __FBSDID("$FreeBSD$"); 40 41 #include <sys/param.h> 42 #include <sys/systm.h> 43 #include <sys/dirent.h> 44 #include <sys/fcntl.h> 45 #include <sys/file.h> 46 #include <sys/limits.h> 47 #include <sys/lockf.h> 48 #include <sys/lock.h> 49 #include <sys/mount.h> 50 #include <sys/namei.h> 51 #include <sys/priv.h> 52 #include <sys/proc.h> 53 #include <sys/rwlock.h> 54 #include <sys/sched.h> 55 #include <sys/stat.h> 56 #include <sys/sysctl.h> 57 #include <sys/unistd.h> 58 #include <sys/vnode.h> 59 #include <sys/smr.h> 60 #include <security/audit/audit.h> 61 #include <security/mac/mac_framework.h> 62 63 #include <vm/vm.h> 64 #include <vm/vm_param.h> 65 #include <vm/vm_object.h> 66 67 #include <fs/tmpfs/tmpfs_vnops.h> 68 #include <fs/tmpfs/tmpfs.h> 69 70 SYSCTL_DECL(_vfs_tmpfs); 71 VFS_SMR_DECLARE; 72 73 static volatile int tmpfs_rename_restarts; 74 SYSCTL_INT(_vfs_tmpfs, OID_AUTO, rename_restarts, CTLFLAG_RD, 75 __DEVOLATILE(int *, &tmpfs_rename_restarts), 0, 76 "Times rename had to restart due to lock contention"); 77 78 static int 79 tmpfs_vn_get_ino_alloc(struct mount *mp, void *arg, int lkflags, 80 struct vnode **rvp) 81 { 82 83 return (tmpfs_alloc_vp(mp, arg, lkflags, rvp)); 84 } 85 86 static int 87 tmpfs_lookup1(struct vnode *dvp, struct vnode **vpp, struct componentname *cnp) 88 { 89 struct tmpfs_dirent *de; 90 struct tmpfs_node *dnode, *pnode; 91 struct tmpfs_mount *tm; 92 int error; 93 94 /* Caller assumes responsibility for ensuring access (VEXEC). */ 95 dnode = VP_TO_TMPFS_DIR(dvp); 96 *vpp = NULLVP; 97 98 /* We cannot be requesting the parent directory of the root node. */ 99 MPASS(IMPLIES(dnode->tn_type == VDIR && 100 dnode->tn_dir.tn_parent == dnode, 101 !(cnp->cn_flags & ISDOTDOT))); 102 103 TMPFS_ASSERT_LOCKED(dnode); 104 if (dnode->tn_dir.tn_parent == NULL) { 105 error = ENOENT; 106 goto out; 107 } 108 if (cnp->cn_flags & ISDOTDOT) { 109 tm = VFS_TO_TMPFS(dvp->v_mount); 110 pnode = dnode->tn_dir.tn_parent; 111 tmpfs_ref_node(pnode); 112 error = vn_vget_ino_gen(dvp, tmpfs_vn_get_ino_alloc, 113 pnode, cnp->cn_lkflags, vpp); 114 tmpfs_free_node(tm, pnode); 115 if (error != 0) 116 goto out; 117 } else if (cnp->cn_namelen == 1 && cnp->cn_nameptr[0] == '.') { 118 VREF(dvp); 119 *vpp = dvp; 120 error = 0; 121 } else { 122 de = tmpfs_dir_lookup(dnode, NULL, cnp); 123 if (de != NULL && de->td_node == NULL) 124 cnp->cn_flags |= ISWHITEOUT; 125 if (de == NULL || de->td_node == NULL) { 126 /* 127 * The entry was not found in the directory. 128 * This is OK if we are creating or renaming an 129 * entry and are working on the last component of 130 * the path name. 131 */ 132 if ((cnp->cn_flags & ISLASTCN) && 133 (cnp->cn_nameiop == CREATE || \ 134 cnp->cn_nameiop == RENAME || 135 (cnp->cn_nameiop == DELETE && 136 cnp->cn_flags & DOWHITEOUT && 137 cnp->cn_flags & ISWHITEOUT))) { 138 error = VOP_ACCESS(dvp, VWRITE, cnp->cn_cred, 139 curthread); 140 if (error != 0) 141 goto out; 142 143 error = EJUSTRETURN; 144 } else 145 error = ENOENT; 146 } else { 147 struct tmpfs_node *tnode; 148 149 /* 150 * The entry was found, so get its associated 151 * tmpfs_node. 152 */ 153 tnode = de->td_node; 154 155 /* 156 * If we are not at the last path component and 157 * found a non-directory or non-link entry (which 158 * may itself be pointing to a directory), raise 159 * an error. 160 */ 161 if ((tnode->tn_type != VDIR && 162 tnode->tn_type != VLNK) && 163 !(cnp->cn_flags & ISLASTCN)) { 164 error = ENOTDIR; 165 goto out; 166 } 167 168 /* 169 * If we are deleting or renaming the entry, keep 170 * track of its tmpfs_dirent so that it can be 171 * easily deleted later. 172 */ 173 if ((cnp->cn_flags & ISLASTCN) && 174 (cnp->cn_nameiop == DELETE || 175 cnp->cn_nameiop == RENAME)) { 176 error = VOP_ACCESS(dvp, VWRITE, cnp->cn_cred, 177 curthread); 178 if (error != 0) 179 goto out; 180 181 /* Allocate a new vnode on the matching entry. */ 182 error = tmpfs_alloc_vp(dvp->v_mount, tnode, 183 cnp->cn_lkflags, vpp); 184 if (error != 0) 185 goto out; 186 187 if ((dnode->tn_mode & S_ISTXT) && 188 VOP_ACCESS(dvp, VADMIN, cnp->cn_cred, 189 curthread) && VOP_ACCESS(*vpp, VADMIN, 190 cnp->cn_cred, curthread)) { 191 error = EPERM; 192 vput(*vpp); 193 *vpp = NULL; 194 goto out; 195 } 196 } else { 197 error = tmpfs_alloc_vp(dvp->v_mount, tnode, 198 cnp->cn_lkflags, vpp); 199 if (error != 0) 200 goto out; 201 } 202 } 203 } 204 205 /* 206 * Store the result of this lookup in the cache. Avoid this if the 207 * request was for creation, as it does not improve timings on 208 * emprical tests. 209 */ 210 if ((cnp->cn_flags & MAKEENTRY) != 0 && tmpfs_use_nc(dvp)) 211 cache_enter(dvp, *vpp, cnp); 212 213 out: 214 /* 215 * If there were no errors, *vpp cannot be null and it must be 216 * locked. 217 */ 218 MPASS(IFF(error == 0, *vpp != NULLVP && VOP_ISLOCKED(*vpp))); 219 220 return (error); 221 } 222 223 static int 224 tmpfs_cached_lookup(struct vop_cachedlookup_args *v) 225 { 226 227 return (tmpfs_lookup1(v->a_dvp, v->a_vpp, v->a_cnp)); 228 } 229 230 static int 231 tmpfs_lookup(struct vop_lookup_args *v) 232 { 233 struct vnode *dvp = v->a_dvp; 234 struct vnode **vpp = v->a_vpp; 235 struct componentname *cnp = v->a_cnp; 236 int error; 237 238 /* Check accessibility of requested node as a first step. */ 239 error = vn_dir_check_exec(dvp, cnp); 240 if (error != 0) 241 return (error); 242 243 return (tmpfs_lookup1(dvp, vpp, cnp)); 244 } 245 246 static int 247 tmpfs_create(struct vop_create_args *v) 248 { 249 struct vnode *dvp = v->a_dvp; 250 struct vnode **vpp = v->a_vpp; 251 struct componentname *cnp = v->a_cnp; 252 struct vattr *vap = v->a_vap; 253 int error; 254 255 MPASS(vap->va_type == VREG || vap->va_type == VSOCK); 256 257 error = tmpfs_alloc_file(dvp, vpp, vap, cnp, NULL); 258 if (error == 0 && (cnp->cn_flags & MAKEENTRY) != 0 && tmpfs_use_nc(dvp)) 259 cache_enter(dvp, *vpp, cnp); 260 return (error); 261 } 262 263 static int 264 tmpfs_mknod(struct vop_mknod_args *v) 265 { 266 struct vnode *dvp = v->a_dvp; 267 struct vnode **vpp = v->a_vpp; 268 struct componentname *cnp = v->a_cnp; 269 struct vattr *vap = v->a_vap; 270 271 if (vap->va_type != VBLK && vap->va_type != VCHR && 272 vap->va_type != VFIFO) 273 return (EINVAL); 274 275 return (tmpfs_alloc_file(dvp, vpp, vap, cnp, NULL)); 276 } 277 278 struct fileops tmpfs_fnops; 279 280 static int 281 tmpfs_open(struct vop_open_args *v) 282 { 283 struct vnode *vp; 284 struct tmpfs_node *node; 285 struct file *fp; 286 int error, mode; 287 288 vp = v->a_vp; 289 mode = v->a_mode; 290 node = VP_TO_TMPFS_NODE(vp); 291 292 /* 293 * The file is still active but all its names have been removed 294 * (e.g. by a "rmdir $(pwd)"). It cannot be opened any more as 295 * it is about to die. 296 */ 297 if (node->tn_links < 1) 298 return (ENOENT); 299 300 /* If the file is marked append-only, deny write requests. */ 301 if (node->tn_flags & APPEND && (mode & (FWRITE | O_APPEND)) == FWRITE) 302 error = EPERM; 303 else { 304 error = 0; 305 /* For regular files, the call below is nop. */ 306 KASSERT(vp->v_type != VREG || (node->tn_reg.tn_aobj->flags & 307 OBJ_DEAD) == 0, ("dead object")); 308 vnode_create_vobject(vp, node->tn_size, v->a_td); 309 } 310 311 fp = v->a_fp; 312 MPASS(fp == NULL || fp->f_data == NULL); 313 if (error == 0 && fp != NULL && vp->v_type == VREG) { 314 tmpfs_ref_node(node); 315 finit_vnode(fp, mode, node, &tmpfs_fnops); 316 } 317 318 return (error); 319 } 320 321 static int 322 tmpfs_close(struct vop_close_args *v) 323 { 324 struct vnode *vp = v->a_vp; 325 326 /* Update node times. */ 327 tmpfs_update(vp); 328 329 return (0); 330 } 331 332 int 333 tmpfs_fo_close(struct file *fp, struct thread *td) 334 { 335 struct tmpfs_node *node; 336 337 node = fp->f_data; 338 if (node != NULL) { 339 MPASS(node->tn_type == VREG); 340 tmpfs_free_node(node->tn_reg.tn_tmp, node); 341 } 342 return (vnops.fo_close(fp, td)); 343 } 344 345 /* 346 * VOP_FPLOOKUP_VEXEC routines are subject to special circumstances, see 347 * the comment above cache_fplookup for details. 348 */ 349 int 350 tmpfs_fplookup_vexec(struct vop_fplookup_vexec_args *v) 351 { 352 struct vnode *vp; 353 struct tmpfs_node *node; 354 struct ucred *cred; 355 mode_t all_x, mode; 356 357 vp = v->a_vp; 358 node = VP_TO_TMPFS_NODE_SMR(vp); 359 if (__predict_false(node == NULL)) 360 return (EAGAIN); 361 362 all_x = S_IXUSR | S_IXGRP | S_IXOTH; 363 mode = atomic_load_short(&node->tn_mode); 364 if (__predict_true((mode & all_x) == all_x)) 365 return (0); 366 367 cred = v->a_cred; 368 return (vaccess_vexec_smr(mode, node->tn_uid, node->tn_gid, cred)); 369 } 370 371 int 372 tmpfs_access(struct vop_access_args *v) 373 { 374 struct vnode *vp = v->a_vp; 375 accmode_t accmode = v->a_accmode; 376 struct ucred *cred = v->a_cred; 377 mode_t all_x = S_IXUSR | S_IXGRP | S_IXOTH; 378 int error; 379 struct tmpfs_node *node; 380 381 MPASS(VOP_ISLOCKED(vp)); 382 383 node = VP_TO_TMPFS_NODE(vp); 384 385 /* 386 * Common case path lookup. 387 */ 388 if (__predict_true(accmode == VEXEC && (node->tn_mode & all_x) == all_x)) 389 return (0); 390 391 switch (vp->v_type) { 392 case VDIR: 393 /* FALLTHROUGH */ 394 case VLNK: 395 /* FALLTHROUGH */ 396 case VREG: 397 if (accmode & VWRITE && vp->v_mount->mnt_flag & MNT_RDONLY) { 398 error = EROFS; 399 goto out; 400 } 401 break; 402 403 case VBLK: 404 /* FALLTHROUGH */ 405 case VCHR: 406 /* FALLTHROUGH */ 407 case VSOCK: 408 /* FALLTHROUGH */ 409 case VFIFO: 410 break; 411 412 default: 413 error = EINVAL; 414 goto out; 415 } 416 417 if (accmode & VWRITE && node->tn_flags & IMMUTABLE) { 418 error = EPERM; 419 goto out; 420 } 421 422 error = vaccess(vp->v_type, node->tn_mode, node->tn_uid, node->tn_gid, 423 accmode, cred); 424 425 out: 426 MPASS(VOP_ISLOCKED(vp)); 427 428 return (error); 429 } 430 431 int 432 tmpfs_stat(struct vop_stat_args *v) 433 { 434 struct vnode *vp = v->a_vp; 435 struct stat *sb = v->a_sb; 436 vm_object_t obj; 437 struct tmpfs_node *node; 438 int error; 439 440 node = VP_TO_TMPFS_NODE(vp); 441 442 tmpfs_update_getattr(vp); 443 444 error = vop_stat_helper_pre(v); 445 if (__predict_false(error)) 446 return (error); 447 448 sb->st_dev = vp->v_mount->mnt_stat.f_fsid.val[0]; 449 sb->st_ino = node->tn_id; 450 sb->st_mode = node->tn_mode | VTTOIF(vp->v_type); 451 sb->st_nlink = node->tn_links; 452 sb->st_uid = node->tn_uid; 453 sb->st_gid = node->tn_gid; 454 sb->st_rdev = (vp->v_type == VBLK || vp->v_type == VCHR) ? 455 node->tn_rdev : NODEV; 456 sb->st_size = node->tn_size; 457 sb->st_atim.tv_sec = node->tn_atime.tv_sec; 458 sb->st_atim.tv_nsec = node->tn_atime.tv_nsec; 459 sb->st_mtim.tv_sec = node->tn_mtime.tv_sec; 460 sb->st_mtim.tv_nsec = node->tn_mtime.tv_nsec; 461 sb->st_ctim.tv_sec = node->tn_ctime.tv_sec; 462 sb->st_ctim.tv_nsec = node->tn_ctime.tv_nsec; 463 sb->st_birthtim.tv_sec = node->tn_birthtime.tv_sec; 464 sb->st_birthtim.tv_nsec = node->tn_birthtime.tv_nsec; 465 sb->st_blksize = PAGE_SIZE; 466 sb->st_flags = node->tn_flags; 467 sb->st_gen = node->tn_gen; 468 if (vp->v_type == VREG) { 469 obj = node->tn_reg.tn_aobj; 470 sb->st_blocks = (u_quad_t)obj->resident_page_count * PAGE_SIZE; 471 } else 472 sb->st_blocks = node->tn_size; 473 sb->st_blocks /= S_BLKSIZE; 474 return (vop_stat_helper_post(v, error)); 475 } 476 477 int 478 tmpfs_getattr(struct vop_getattr_args *v) 479 { 480 struct vnode *vp = v->a_vp; 481 struct vattr *vap = v->a_vap; 482 vm_object_t obj; 483 struct tmpfs_node *node; 484 485 node = VP_TO_TMPFS_NODE(vp); 486 487 tmpfs_update_getattr(vp); 488 489 vap->va_type = vp->v_type; 490 vap->va_mode = node->tn_mode; 491 vap->va_nlink = node->tn_links; 492 vap->va_uid = node->tn_uid; 493 vap->va_gid = node->tn_gid; 494 vap->va_fsid = vp->v_mount->mnt_stat.f_fsid.val[0]; 495 vap->va_fileid = node->tn_id; 496 vap->va_size = node->tn_size; 497 vap->va_blocksize = PAGE_SIZE; 498 vap->va_atime = node->tn_atime; 499 vap->va_mtime = node->tn_mtime; 500 vap->va_ctime = node->tn_ctime; 501 vap->va_birthtime = node->tn_birthtime; 502 vap->va_gen = node->tn_gen; 503 vap->va_flags = node->tn_flags; 504 vap->va_rdev = (vp->v_type == VBLK || vp->v_type == VCHR) ? 505 node->tn_rdev : NODEV; 506 if (vp->v_type == VREG) { 507 obj = node->tn_reg.tn_aobj; 508 vap->va_bytes = (u_quad_t)obj->resident_page_count * PAGE_SIZE; 509 } else 510 vap->va_bytes = node->tn_size; 511 vap->va_filerev = 0; 512 513 return (0); 514 } 515 516 int 517 tmpfs_setattr(struct vop_setattr_args *v) 518 { 519 struct vnode *vp = v->a_vp; 520 struct vattr *vap = v->a_vap; 521 struct ucred *cred = v->a_cred; 522 struct thread *td = curthread; 523 524 int error; 525 526 MPASS(VOP_ISLOCKED(vp)); 527 ASSERT_VOP_IN_SEQC(vp); 528 529 error = 0; 530 531 /* Abort if any unsettable attribute is given. */ 532 if (vap->va_type != VNON || 533 vap->va_nlink != VNOVAL || 534 vap->va_fsid != VNOVAL || 535 vap->va_fileid != VNOVAL || 536 vap->va_blocksize != VNOVAL || 537 vap->va_gen != VNOVAL || 538 vap->va_rdev != VNOVAL || 539 vap->va_bytes != VNOVAL) 540 error = EINVAL; 541 542 if (error == 0 && (vap->va_flags != VNOVAL)) 543 error = tmpfs_chflags(vp, vap->va_flags, cred, td); 544 545 if (error == 0 && (vap->va_size != VNOVAL)) 546 error = tmpfs_chsize(vp, vap->va_size, cred, td); 547 548 if (error == 0 && (vap->va_uid != VNOVAL || vap->va_gid != VNOVAL)) 549 error = tmpfs_chown(vp, vap->va_uid, vap->va_gid, cred, td); 550 551 if (error == 0 && (vap->va_mode != (mode_t)VNOVAL)) 552 error = tmpfs_chmod(vp, vap->va_mode, cred, td); 553 554 if (error == 0 && ((vap->va_atime.tv_sec != VNOVAL && 555 vap->va_atime.tv_nsec != VNOVAL) || 556 (vap->va_mtime.tv_sec != VNOVAL && 557 vap->va_mtime.tv_nsec != VNOVAL) || 558 (vap->va_birthtime.tv_sec != VNOVAL && 559 vap->va_birthtime.tv_nsec != VNOVAL))) 560 error = tmpfs_chtimes(vp, vap, cred, td); 561 562 /* 563 * Update the node times. We give preference to the error codes 564 * generated by this function rather than the ones that may arise 565 * from tmpfs_update. 566 */ 567 tmpfs_update(vp); 568 569 MPASS(VOP_ISLOCKED(vp)); 570 571 return (error); 572 } 573 574 static int 575 tmpfs_read(struct vop_read_args *v) 576 { 577 struct vnode *vp; 578 struct uio *uio; 579 struct tmpfs_node *node; 580 581 vp = v->a_vp; 582 if (vp->v_type != VREG) 583 return (EISDIR); 584 uio = v->a_uio; 585 if (uio->uio_offset < 0) 586 return (EINVAL); 587 node = VP_TO_TMPFS_NODE(vp); 588 tmpfs_set_accessed(VFS_TO_TMPFS(vp->v_mount), node); 589 return (uiomove_object(node->tn_reg.tn_aobj, node->tn_size, uio)); 590 } 591 592 static int 593 tmpfs_read_pgcache(struct vop_read_pgcache_args *v) 594 { 595 struct vnode *vp; 596 struct tmpfs_node *node; 597 vm_object_t object; 598 off_t size; 599 int error; 600 601 vp = v->a_vp; 602 VNPASS((vn_irflag_read(vp) & VIRF_PGREAD) != 0, vp); 603 604 if (v->a_uio->uio_offset < 0) 605 return (EINVAL); 606 607 error = EJUSTRETURN; 608 vfs_smr_enter(); 609 610 node = VP_TO_TMPFS_NODE_SMR(vp); 611 if (node == NULL) 612 goto out_smr; 613 MPASS(node->tn_type == VREG); 614 MPASS(node->tn_refcount >= 1); 615 object = node->tn_reg.tn_aobj; 616 if (object == NULL) 617 goto out_smr; 618 619 MPASS(object->type == tmpfs_pager_type); 620 MPASS((object->flags & (OBJ_ANON | OBJ_DEAD | OBJ_SWAP)) == 621 OBJ_SWAP); 622 if (!VN_IS_DOOMED(vp)) { 623 /* size cannot become shorter due to rangelock. */ 624 size = node->tn_size; 625 tmpfs_set_accessed(node->tn_reg.tn_tmp, node); 626 vfs_smr_exit(); 627 error = uiomove_object(object, size, v->a_uio); 628 return (error); 629 } 630 out_smr: 631 vfs_smr_exit(); 632 return (error); 633 } 634 635 static int 636 tmpfs_write(struct vop_write_args *v) 637 { 638 struct vnode *vp; 639 struct uio *uio; 640 struct tmpfs_node *node; 641 off_t oldsize; 642 ssize_t r; 643 int error, ioflag; 644 mode_t newmode; 645 646 vp = v->a_vp; 647 uio = v->a_uio; 648 ioflag = v->a_ioflag; 649 error = 0; 650 node = VP_TO_TMPFS_NODE(vp); 651 oldsize = node->tn_size; 652 653 if (uio->uio_offset < 0 || vp->v_type != VREG) 654 return (EINVAL); 655 if (uio->uio_resid == 0) 656 return (0); 657 if (ioflag & IO_APPEND) 658 uio->uio_offset = node->tn_size; 659 error = vn_rlimit_fsizex(vp, uio, VFS_TO_TMPFS(vp->v_mount)-> 660 tm_maxfilesize, &r, uio->uio_td); 661 if (error != 0) { 662 vn_rlimit_fsizex_res(uio, r); 663 return (error); 664 } 665 666 if (uio->uio_offset + uio->uio_resid > node->tn_size) { 667 error = tmpfs_reg_resize(vp, uio->uio_offset + uio->uio_resid, 668 FALSE); 669 if (error != 0) 670 goto out; 671 } 672 673 error = uiomove_object(node->tn_reg.tn_aobj, node->tn_size, uio); 674 node->tn_status |= TMPFS_NODE_MODIFIED | TMPFS_NODE_CHANGED; 675 node->tn_accessed = true; 676 if (node->tn_mode & (S_ISUID | S_ISGID)) { 677 if (priv_check_cred(v->a_cred, PRIV_VFS_RETAINSUGID)) { 678 newmode = node->tn_mode & ~(S_ISUID | S_ISGID); 679 vn_seqc_write_begin(vp); 680 atomic_store_short(&node->tn_mode, newmode); 681 vn_seqc_write_end(vp); 682 } 683 } 684 if (error != 0) 685 (void)tmpfs_reg_resize(vp, oldsize, TRUE); 686 687 out: 688 MPASS(IMPLIES(error == 0, uio->uio_resid == 0)); 689 MPASS(IMPLIES(error != 0, oldsize == node->tn_size)); 690 691 vn_rlimit_fsizex_res(uio, r); 692 return (error); 693 } 694 695 static int 696 tmpfs_deallocate(struct vop_deallocate_args *v) 697 { 698 return (tmpfs_reg_punch_hole(v->a_vp, v->a_offset, v->a_len)); 699 } 700 701 static int 702 tmpfs_fsync(struct vop_fsync_args *v) 703 { 704 struct vnode *vp = v->a_vp; 705 706 MPASS(VOP_ISLOCKED(vp)); 707 708 tmpfs_check_mtime(vp); 709 tmpfs_update(vp); 710 711 return (0); 712 } 713 714 static int 715 tmpfs_remove(struct vop_remove_args *v) 716 { 717 struct vnode *dvp = v->a_dvp; 718 struct vnode *vp = v->a_vp; 719 720 int error; 721 struct tmpfs_dirent *de; 722 struct tmpfs_mount *tmp; 723 struct tmpfs_node *dnode; 724 struct tmpfs_node *node; 725 726 MPASS(VOP_ISLOCKED(dvp)); 727 MPASS(VOP_ISLOCKED(vp)); 728 729 if (vp->v_type == VDIR) { 730 error = EISDIR; 731 goto out; 732 } 733 734 dnode = VP_TO_TMPFS_DIR(dvp); 735 node = VP_TO_TMPFS_NODE(vp); 736 tmp = VFS_TO_TMPFS(vp->v_mount); 737 de = tmpfs_dir_lookup(dnode, node, v->a_cnp); 738 MPASS(de != NULL); 739 740 /* Files marked as immutable or append-only cannot be deleted. */ 741 if ((node->tn_flags & (IMMUTABLE | APPEND | NOUNLINK)) || 742 (dnode->tn_flags & APPEND)) { 743 error = EPERM; 744 goto out; 745 } 746 747 /* Remove the entry from the directory; as it is a file, we do not 748 * have to change the number of hard links of the directory. */ 749 tmpfs_dir_detach(dvp, de); 750 if (v->a_cnp->cn_flags & DOWHITEOUT) 751 tmpfs_dir_whiteout_add(dvp, v->a_cnp); 752 753 /* Free the directory entry we just deleted. Note that the node 754 * referred by it will not be removed until the vnode is really 755 * reclaimed. */ 756 tmpfs_free_dirent(tmp, de); 757 758 node->tn_status |= TMPFS_NODE_CHANGED; 759 node->tn_accessed = true; 760 error = 0; 761 762 out: 763 return (error); 764 } 765 766 static int 767 tmpfs_link(struct vop_link_args *v) 768 { 769 struct vnode *dvp = v->a_tdvp; 770 struct vnode *vp = v->a_vp; 771 struct componentname *cnp = v->a_cnp; 772 773 int error; 774 struct tmpfs_dirent *de; 775 struct tmpfs_node *node; 776 777 MPASS(VOP_ISLOCKED(dvp)); 778 MPASS(dvp != vp); /* XXX When can this be false? */ 779 node = VP_TO_TMPFS_NODE(vp); 780 781 /* Ensure that we do not overflow the maximum number of links imposed 782 * by the system. */ 783 MPASS(node->tn_links <= TMPFS_LINK_MAX); 784 if (node->tn_links == TMPFS_LINK_MAX) { 785 error = EMLINK; 786 goto out; 787 } 788 789 /* We cannot create links of files marked immutable or append-only. */ 790 if (node->tn_flags & (IMMUTABLE | APPEND)) { 791 error = EPERM; 792 goto out; 793 } 794 795 /* Allocate a new directory entry to represent the node. */ 796 error = tmpfs_alloc_dirent(VFS_TO_TMPFS(vp->v_mount), node, 797 cnp->cn_nameptr, cnp->cn_namelen, &de); 798 if (error != 0) 799 goto out; 800 801 /* Insert the new directory entry into the appropriate directory. */ 802 if (cnp->cn_flags & ISWHITEOUT) 803 tmpfs_dir_whiteout_remove(dvp, cnp); 804 tmpfs_dir_attach(dvp, de); 805 806 /* vp link count has changed, so update node times. */ 807 node->tn_status |= TMPFS_NODE_CHANGED; 808 tmpfs_update(vp); 809 810 error = 0; 811 812 out: 813 return (error); 814 } 815 816 /* 817 * We acquire all but fdvp locks using non-blocking acquisitions. If we 818 * fail to acquire any lock in the path we will drop all held locks, 819 * acquire the new lock in a blocking fashion, and then release it and 820 * restart the rename. This acquire/release step ensures that we do not 821 * spin on a lock waiting for release. On error release all vnode locks 822 * and decrement references the way tmpfs_rename() would do. 823 */ 824 static int 825 tmpfs_rename_relock(struct vnode *fdvp, struct vnode **fvpp, 826 struct vnode *tdvp, struct vnode **tvpp, 827 struct componentname *fcnp, struct componentname *tcnp) 828 { 829 struct vnode *nvp; 830 struct mount *mp; 831 struct tmpfs_dirent *de; 832 int error, restarts = 0; 833 834 VOP_UNLOCK(tdvp); 835 if (*tvpp != NULL && *tvpp != tdvp) 836 VOP_UNLOCK(*tvpp); 837 mp = fdvp->v_mount; 838 839 relock: 840 restarts += 1; 841 error = vn_lock(fdvp, LK_EXCLUSIVE); 842 if (error) 843 goto releout; 844 if (vn_lock(tdvp, LK_EXCLUSIVE | LK_NOWAIT) != 0) { 845 VOP_UNLOCK(fdvp); 846 error = vn_lock(tdvp, LK_EXCLUSIVE); 847 if (error) 848 goto releout; 849 VOP_UNLOCK(tdvp); 850 goto relock; 851 } 852 /* 853 * Re-resolve fvp to be certain it still exists and fetch the 854 * correct vnode. 855 */ 856 de = tmpfs_dir_lookup(VP_TO_TMPFS_DIR(fdvp), NULL, fcnp); 857 if (de == NULL) { 858 VOP_UNLOCK(fdvp); 859 VOP_UNLOCK(tdvp); 860 if ((fcnp->cn_flags & ISDOTDOT) != 0 || 861 (fcnp->cn_namelen == 1 && fcnp->cn_nameptr[0] == '.')) 862 error = EINVAL; 863 else 864 error = ENOENT; 865 goto releout; 866 } 867 error = tmpfs_alloc_vp(mp, de->td_node, LK_EXCLUSIVE | LK_NOWAIT, &nvp); 868 if (error != 0) { 869 VOP_UNLOCK(fdvp); 870 VOP_UNLOCK(tdvp); 871 if (error != EBUSY) 872 goto releout; 873 error = tmpfs_alloc_vp(mp, de->td_node, LK_EXCLUSIVE, &nvp); 874 if (error != 0) 875 goto releout; 876 VOP_UNLOCK(nvp); 877 /* 878 * Concurrent rename race. 879 */ 880 if (nvp == tdvp) { 881 vrele(nvp); 882 error = EINVAL; 883 goto releout; 884 } 885 vrele(*fvpp); 886 *fvpp = nvp; 887 goto relock; 888 } 889 vrele(*fvpp); 890 *fvpp = nvp; 891 VOP_UNLOCK(*fvpp); 892 /* 893 * Re-resolve tvp and acquire the vnode lock if present. 894 */ 895 de = tmpfs_dir_lookup(VP_TO_TMPFS_DIR(tdvp), NULL, tcnp); 896 /* 897 * If tvp disappeared we just carry on. 898 */ 899 if (de == NULL && *tvpp != NULL) { 900 vrele(*tvpp); 901 *tvpp = NULL; 902 } 903 /* 904 * Get the tvp ino if the lookup succeeded. We may have to restart 905 * if the non-blocking acquire fails. 906 */ 907 if (de != NULL) { 908 nvp = NULL; 909 error = tmpfs_alloc_vp(mp, de->td_node, 910 LK_EXCLUSIVE | LK_NOWAIT, &nvp); 911 if (*tvpp != NULL) 912 vrele(*tvpp); 913 *tvpp = nvp; 914 if (error != 0) { 915 VOP_UNLOCK(fdvp); 916 VOP_UNLOCK(tdvp); 917 if (error != EBUSY) 918 goto releout; 919 error = tmpfs_alloc_vp(mp, de->td_node, LK_EXCLUSIVE, 920 &nvp); 921 if (error != 0) 922 goto releout; 923 VOP_UNLOCK(nvp); 924 /* 925 * fdvp contains fvp, thus tvp (=fdvp) is not empty. 926 */ 927 if (nvp == fdvp) { 928 error = ENOTEMPTY; 929 goto releout; 930 } 931 goto relock; 932 } 933 } 934 tmpfs_rename_restarts += restarts; 935 936 return (0); 937 938 releout: 939 vrele(fdvp); 940 vrele(*fvpp); 941 vrele(tdvp); 942 if (*tvpp != NULL) 943 vrele(*tvpp); 944 tmpfs_rename_restarts += restarts; 945 946 return (error); 947 } 948 949 static int 950 tmpfs_rename(struct vop_rename_args *v) 951 { 952 struct vnode *fdvp = v->a_fdvp; 953 struct vnode *fvp = v->a_fvp; 954 struct componentname *fcnp = v->a_fcnp; 955 struct vnode *tdvp = v->a_tdvp; 956 struct vnode *tvp = v->a_tvp; 957 struct componentname *tcnp = v->a_tcnp; 958 char *newname; 959 struct tmpfs_dirent *de; 960 struct tmpfs_mount *tmp; 961 struct tmpfs_node *fdnode; 962 struct tmpfs_node *fnode; 963 struct tmpfs_node *tnode; 964 struct tmpfs_node *tdnode; 965 int error; 966 bool want_seqc_end; 967 968 MPASS(VOP_ISLOCKED(tdvp)); 969 MPASS(IMPLIES(tvp != NULL, VOP_ISLOCKED(tvp))); 970 971 want_seqc_end = false; 972 973 /* 974 * Disallow cross-device renames. 975 * XXX Why isn't this done by the caller? 976 */ 977 if (fvp->v_mount != tdvp->v_mount || 978 (tvp != NULL && fvp->v_mount != tvp->v_mount)) { 979 error = EXDEV; 980 goto out; 981 } 982 983 /* If source and target are the same file, there is nothing to do. */ 984 if (fvp == tvp) { 985 error = 0; 986 goto out; 987 } 988 989 /* 990 * If we need to move the directory between entries, lock the 991 * source so that we can safely operate on it. 992 */ 993 if (fdvp != tdvp && fdvp != tvp) { 994 if (vn_lock(fdvp, LK_EXCLUSIVE | LK_NOWAIT) != 0) { 995 error = tmpfs_rename_relock(fdvp, &fvp, tdvp, &tvp, 996 fcnp, tcnp); 997 if (error != 0) 998 return (error); 999 ASSERT_VOP_ELOCKED(fdvp, 1000 "tmpfs_rename: fdvp not locked"); 1001 ASSERT_VOP_ELOCKED(tdvp, 1002 "tmpfs_rename: tdvp not locked"); 1003 if (tvp != NULL) 1004 ASSERT_VOP_ELOCKED(tvp, 1005 "tmpfs_rename: tvp not locked"); 1006 if (fvp == tvp) { 1007 error = 0; 1008 goto out_locked; 1009 } 1010 } 1011 } 1012 1013 if (tvp != NULL) 1014 vn_seqc_write_begin(tvp); 1015 vn_seqc_write_begin(tdvp); 1016 vn_seqc_write_begin(fvp); 1017 vn_seqc_write_begin(fdvp); 1018 want_seqc_end = true; 1019 1020 tmp = VFS_TO_TMPFS(tdvp->v_mount); 1021 tdnode = VP_TO_TMPFS_DIR(tdvp); 1022 tnode = (tvp == NULL) ? NULL : VP_TO_TMPFS_NODE(tvp); 1023 fdnode = VP_TO_TMPFS_DIR(fdvp); 1024 fnode = VP_TO_TMPFS_NODE(fvp); 1025 de = tmpfs_dir_lookup(fdnode, fnode, fcnp); 1026 1027 /* 1028 * Entry can disappear before we lock fdvp, 1029 * also avoid manipulating '.' and '..' entries. 1030 */ 1031 if (de == NULL) { 1032 if ((fcnp->cn_flags & ISDOTDOT) != 0 || 1033 (fcnp->cn_namelen == 1 && fcnp->cn_nameptr[0] == '.')) 1034 error = EINVAL; 1035 else 1036 error = ENOENT; 1037 goto out_locked; 1038 } 1039 MPASS(de->td_node == fnode); 1040 1041 /* 1042 * If re-naming a directory to another preexisting directory 1043 * ensure that the target directory is empty so that its 1044 * removal causes no side effects. 1045 * Kern_rename guarantees the destination to be a directory 1046 * if the source is one. 1047 */ 1048 if (tvp != NULL) { 1049 MPASS(tnode != NULL); 1050 1051 if ((tnode->tn_flags & (NOUNLINK | IMMUTABLE | APPEND)) || 1052 (tdnode->tn_flags & (APPEND | IMMUTABLE))) { 1053 error = EPERM; 1054 goto out_locked; 1055 } 1056 1057 if (fnode->tn_type == VDIR && tnode->tn_type == VDIR) { 1058 if (tnode->tn_size > 0) { 1059 error = ENOTEMPTY; 1060 goto out_locked; 1061 } 1062 } else if (fnode->tn_type == VDIR && tnode->tn_type != VDIR) { 1063 error = ENOTDIR; 1064 goto out_locked; 1065 } else if (fnode->tn_type != VDIR && tnode->tn_type == VDIR) { 1066 error = EISDIR; 1067 goto out_locked; 1068 } else { 1069 MPASS(fnode->tn_type != VDIR && 1070 tnode->tn_type != VDIR); 1071 } 1072 } 1073 1074 if ((fnode->tn_flags & (NOUNLINK | IMMUTABLE | APPEND)) 1075 || (fdnode->tn_flags & (APPEND | IMMUTABLE))) { 1076 error = EPERM; 1077 goto out_locked; 1078 } 1079 1080 /* 1081 * Ensure that we have enough memory to hold the new name, if it 1082 * has to be changed. 1083 */ 1084 if (fcnp->cn_namelen != tcnp->cn_namelen || 1085 bcmp(fcnp->cn_nameptr, tcnp->cn_nameptr, fcnp->cn_namelen) != 0) { 1086 newname = malloc(tcnp->cn_namelen, M_TMPFSNAME, M_WAITOK); 1087 } else 1088 newname = NULL; 1089 1090 /* 1091 * If the node is being moved to another directory, we have to do 1092 * the move. 1093 */ 1094 if (fdnode != tdnode) { 1095 /* 1096 * In case we are moving a directory, we have to adjust its 1097 * parent to point to the new parent. 1098 */ 1099 if (de->td_node->tn_type == VDIR) { 1100 struct tmpfs_node *n; 1101 1102 /* 1103 * Ensure the target directory is not a child of the 1104 * directory being moved. Otherwise, we'd end up 1105 * with stale nodes. 1106 */ 1107 n = tdnode; 1108 /* 1109 * TMPFS_LOCK guaranties that no nodes are freed while 1110 * traversing the list. Nodes can only be marked as 1111 * removed: tn_parent == NULL. 1112 */ 1113 TMPFS_LOCK(tmp); 1114 TMPFS_NODE_LOCK(n); 1115 while (n != n->tn_dir.tn_parent) { 1116 struct tmpfs_node *parent; 1117 1118 if (n == fnode) { 1119 TMPFS_NODE_UNLOCK(n); 1120 TMPFS_UNLOCK(tmp); 1121 error = EINVAL; 1122 if (newname != NULL) 1123 free(newname, M_TMPFSNAME); 1124 goto out_locked; 1125 } 1126 parent = n->tn_dir.tn_parent; 1127 TMPFS_NODE_UNLOCK(n); 1128 if (parent == NULL) { 1129 n = NULL; 1130 break; 1131 } 1132 TMPFS_NODE_LOCK(parent); 1133 if (parent->tn_dir.tn_parent == NULL) { 1134 TMPFS_NODE_UNLOCK(parent); 1135 n = NULL; 1136 break; 1137 } 1138 n = parent; 1139 } 1140 TMPFS_UNLOCK(tmp); 1141 if (n == NULL) { 1142 error = EINVAL; 1143 if (newname != NULL) 1144 free(newname, M_TMPFSNAME); 1145 goto out_locked; 1146 } 1147 TMPFS_NODE_UNLOCK(n); 1148 1149 /* Adjust the parent pointer. */ 1150 TMPFS_VALIDATE_DIR(fnode); 1151 TMPFS_NODE_LOCK(de->td_node); 1152 de->td_node->tn_dir.tn_parent = tdnode; 1153 TMPFS_NODE_UNLOCK(de->td_node); 1154 1155 /* 1156 * As a result of changing the target of the '..' 1157 * entry, the link count of the source and target 1158 * directories has to be adjusted. 1159 */ 1160 TMPFS_NODE_LOCK(tdnode); 1161 TMPFS_ASSERT_LOCKED(tdnode); 1162 tdnode->tn_links++; 1163 TMPFS_NODE_UNLOCK(tdnode); 1164 1165 TMPFS_NODE_LOCK(fdnode); 1166 TMPFS_ASSERT_LOCKED(fdnode); 1167 fdnode->tn_links--; 1168 TMPFS_NODE_UNLOCK(fdnode); 1169 } 1170 } 1171 1172 /* 1173 * Do the move: just remove the entry from the source directory 1174 * and insert it into the target one. 1175 */ 1176 tmpfs_dir_detach(fdvp, de); 1177 1178 if (fcnp->cn_flags & DOWHITEOUT) 1179 tmpfs_dir_whiteout_add(fdvp, fcnp); 1180 if (tcnp->cn_flags & ISWHITEOUT) 1181 tmpfs_dir_whiteout_remove(tdvp, tcnp); 1182 1183 /* 1184 * If the name has changed, we need to make it effective by changing 1185 * it in the directory entry. 1186 */ 1187 if (newname != NULL) { 1188 MPASS(tcnp->cn_namelen <= MAXNAMLEN); 1189 1190 free(de->ud.td_name, M_TMPFSNAME); 1191 de->ud.td_name = newname; 1192 tmpfs_dirent_init(de, tcnp->cn_nameptr, tcnp->cn_namelen); 1193 1194 fnode->tn_status |= TMPFS_NODE_CHANGED; 1195 tdnode->tn_status |= TMPFS_NODE_MODIFIED; 1196 } 1197 1198 /* 1199 * If we are overwriting an entry, we have to remove the old one 1200 * from the target directory. 1201 */ 1202 if (tvp != NULL) { 1203 struct tmpfs_dirent *tde; 1204 1205 /* Remove the old entry from the target directory. */ 1206 tde = tmpfs_dir_lookup(tdnode, tnode, tcnp); 1207 tmpfs_dir_detach(tdvp, tde); 1208 1209 /* 1210 * Free the directory entry we just deleted. Note that the 1211 * node referred by it will not be removed until the vnode is 1212 * really reclaimed. 1213 */ 1214 tmpfs_free_dirent(VFS_TO_TMPFS(tvp->v_mount), tde); 1215 } 1216 1217 tmpfs_dir_attach(tdvp, de); 1218 1219 if (tmpfs_use_nc(fvp)) { 1220 cache_vop_rename(fdvp, fvp, tdvp, tvp, fcnp, tcnp); 1221 } 1222 1223 error = 0; 1224 1225 out_locked: 1226 if (fdvp != tdvp && fdvp != tvp) 1227 VOP_UNLOCK(fdvp); 1228 1229 out: 1230 if (want_seqc_end) { 1231 if (tvp != NULL) 1232 vn_seqc_write_end(tvp); 1233 vn_seqc_write_end(tdvp); 1234 vn_seqc_write_end(fvp); 1235 vn_seqc_write_end(fdvp); 1236 } 1237 1238 /* 1239 * Release target nodes. 1240 * XXX: I don't understand when tdvp can be the same as tvp, but 1241 * other code takes care of this... 1242 */ 1243 if (tdvp == tvp) 1244 vrele(tdvp); 1245 else 1246 vput(tdvp); 1247 if (tvp != NULL) 1248 vput(tvp); 1249 1250 /* Release source nodes. */ 1251 vrele(fdvp); 1252 vrele(fvp); 1253 1254 return (error); 1255 } 1256 1257 static int 1258 tmpfs_mkdir(struct vop_mkdir_args *v) 1259 { 1260 struct vnode *dvp = v->a_dvp; 1261 struct vnode **vpp = v->a_vpp; 1262 struct componentname *cnp = v->a_cnp; 1263 struct vattr *vap = v->a_vap; 1264 1265 MPASS(vap->va_type == VDIR); 1266 1267 return (tmpfs_alloc_file(dvp, vpp, vap, cnp, NULL)); 1268 } 1269 1270 static int 1271 tmpfs_rmdir(struct vop_rmdir_args *v) 1272 { 1273 struct vnode *dvp = v->a_dvp; 1274 struct vnode *vp = v->a_vp; 1275 1276 int error; 1277 struct tmpfs_dirent *de; 1278 struct tmpfs_mount *tmp; 1279 struct tmpfs_node *dnode; 1280 struct tmpfs_node *node; 1281 1282 MPASS(VOP_ISLOCKED(dvp)); 1283 MPASS(VOP_ISLOCKED(vp)); 1284 1285 tmp = VFS_TO_TMPFS(dvp->v_mount); 1286 dnode = VP_TO_TMPFS_DIR(dvp); 1287 node = VP_TO_TMPFS_DIR(vp); 1288 1289 /* Directories with more than two entries ('.' and '..') cannot be 1290 * removed. */ 1291 if (node->tn_size > 0) { 1292 error = ENOTEMPTY; 1293 goto out; 1294 } 1295 1296 if ((dnode->tn_flags & APPEND) 1297 || (node->tn_flags & (NOUNLINK | IMMUTABLE | APPEND))) { 1298 error = EPERM; 1299 goto out; 1300 } 1301 1302 /* This invariant holds only if we are not trying to remove "..". 1303 * We checked for that above so this is safe now. */ 1304 MPASS(node->tn_dir.tn_parent == dnode); 1305 1306 /* Get the directory entry associated with node (vp). This was 1307 * filled by tmpfs_lookup while looking up the entry. */ 1308 de = tmpfs_dir_lookup(dnode, node, v->a_cnp); 1309 MPASS(TMPFS_DIRENT_MATCHES(de, 1310 v->a_cnp->cn_nameptr, 1311 v->a_cnp->cn_namelen)); 1312 1313 /* Check flags to see if we are allowed to remove the directory. */ 1314 if ((dnode->tn_flags & APPEND) != 0 || 1315 (node->tn_flags & (NOUNLINK | IMMUTABLE | APPEND)) != 0) { 1316 error = EPERM; 1317 goto out; 1318 } 1319 1320 /* Detach the directory entry from the directory (dnode). */ 1321 tmpfs_dir_detach(dvp, de); 1322 if (v->a_cnp->cn_flags & DOWHITEOUT) 1323 tmpfs_dir_whiteout_add(dvp, v->a_cnp); 1324 1325 /* No vnode should be allocated for this entry from this point */ 1326 TMPFS_NODE_LOCK(node); 1327 node->tn_links--; 1328 node->tn_dir.tn_parent = NULL; 1329 node->tn_status |= TMPFS_NODE_CHANGED | TMPFS_NODE_MODIFIED; 1330 node->tn_accessed = true; 1331 1332 TMPFS_NODE_UNLOCK(node); 1333 1334 TMPFS_NODE_LOCK(dnode); 1335 dnode->tn_links--; 1336 dnode->tn_status |= TMPFS_NODE_CHANGED | TMPFS_NODE_MODIFIED; 1337 dnode->tn_accessed = true; 1338 TMPFS_NODE_UNLOCK(dnode); 1339 1340 if (tmpfs_use_nc(dvp)) { 1341 cache_vop_rmdir(dvp, vp); 1342 } 1343 1344 /* Free the directory entry we just deleted. Note that the node 1345 * referred by it will not be removed until the vnode is really 1346 * reclaimed. */ 1347 tmpfs_free_dirent(tmp, de); 1348 1349 /* Release the deleted vnode (will destroy the node, notify 1350 * interested parties and clean it from the cache). */ 1351 1352 dnode->tn_status |= TMPFS_NODE_CHANGED; 1353 tmpfs_update(dvp); 1354 1355 error = 0; 1356 1357 out: 1358 return (error); 1359 } 1360 1361 static int 1362 tmpfs_symlink(struct vop_symlink_args *v) 1363 { 1364 struct vnode *dvp = v->a_dvp; 1365 struct vnode **vpp = v->a_vpp; 1366 struct componentname *cnp = v->a_cnp; 1367 struct vattr *vap = v->a_vap; 1368 const char *target = v->a_target; 1369 1370 #ifdef notyet /* XXX FreeBSD BUG: kern_symlink is not setting VLNK */ 1371 MPASS(vap->va_type == VLNK); 1372 #else 1373 vap->va_type = VLNK; 1374 #endif 1375 1376 return (tmpfs_alloc_file(dvp, vpp, vap, cnp, target)); 1377 } 1378 1379 static int 1380 tmpfs_readdir(struct vop_readdir_args *va) 1381 { 1382 struct vnode *vp; 1383 struct uio *uio; 1384 struct tmpfs_mount *tm; 1385 struct tmpfs_node *node; 1386 uint64_t **cookies; 1387 int *eofflag, *ncookies; 1388 ssize_t startresid; 1389 int error, maxcookies; 1390 1391 vp = va->a_vp; 1392 uio = va->a_uio; 1393 eofflag = va->a_eofflag; 1394 cookies = va->a_cookies; 1395 ncookies = va->a_ncookies; 1396 1397 /* This operation only makes sense on directory nodes. */ 1398 if (vp->v_type != VDIR) 1399 return (ENOTDIR); 1400 1401 maxcookies = 0; 1402 node = VP_TO_TMPFS_DIR(vp); 1403 tm = VFS_TO_TMPFS(vp->v_mount); 1404 1405 startresid = uio->uio_resid; 1406 1407 /* Allocate cookies for NFS and compat modules. */ 1408 if (cookies != NULL && ncookies != NULL) { 1409 maxcookies = howmany(node->tn_size, 1410 sizeof(struct tmpfs_dirent)) + 2; 1411 *cookies = malloc(maxcookies * sizeof(**cookies), M_TEMP, 1412 M_WAITOK); 1413 *ncookies = 0; 1414 } 1415 1416 if (cookies == NULL) 1417 error = tmpfs_dir_getdents(tm, node, uio, 0, NULL, NULL); 1418 else 1419 error = tmpfs_dir_getdents(tm, node, uio, maxcookies, *cookies, 1420 ncookies); 1421 1422 /* Buffer was filled without hitting EOF. */ 1423 if (error == EJUSTRETURN) 1424 error = (uio->uio_resid != startresid) ? 0 : EINVAL; 1425 1426 if (error != 0 && cookies != NULL && ncookies != NULL) { 1427 free(*cookies, M_TEMP); 1428 *cookies = NULL; 1429 *ncookies = 0; 1430 } 1431 1432 if (eofflag != NULL) 1433 *eofflag = 1434 (error == 0 && uio->uio_offset == TMPFS_DIRCOOKIE_EOF); 1435 1436 return (error); 1437 } 1438 1439 static int 1440 tmpfs_readlink(struct vop_readlink_args *v) 1441 { 1442 struct vnode *vp = v->a_vp; 1443 struct uio *uio = v->a_uio; 1444 1445 int error; 1446 struct tmpfs_node *node; 1447 1448 MPASS(uio->uio_offset == 0); 1449 MPASS(vp->v_type == VLNK); 1450 1451 node = VP_TO_TMPFS_NODE(vp); 1452 1453 error = uiomove(node->tn_link_target, MIN(node->tn_size, uio->uio_resid), 1454 uio); 1455 tmpfs_set_accessed(VFS_TO_TMPFS(vp->v_mount), node); 1456 1457 return (error); 1458 } 1459 1460 /* 1461 * VOP_FPLOOKUP_SYMLINK routines are subject to special circumstances, see 1462 * the comment above cache_fplookup for details. 1463 * 1464 * Check tmpfs_alloc_node for tmpfs-specific synchronisation notes. 1465 */ 1466 static int 1467 tmpfs_fplookup_symlink(struct vop_fplookup_symlink_args *v) 1468 { 1469 struct vnode *vp; 1470 struct tmpfs_node *node; 1471 char *symlink; 1472 1473 vp = v->a_vp; 1474 node = VP_TO_TMPFS_NODE_SMR(vp); 1475 if (__predict_false(node == NULL)) 1476 return (EAGAIN); 1477 if (!atomic_load_char(&node->tn_link_smr)) 1478 return (EAGAIN); 1479 symlink = atomic_load_ptr(&node->tn_link_target); 1480 if (symlink == NULL) 1481 return (EAGAIN); 1482 1483 return (cache_symlink_resolve(v->a_fpl, symlink, node->tn_size)); 1484 } 1485 1486 static int 1487 tmpfs_inactive(struct vop_inactive_args *v) 1488 { 1489 struct vnode *vp; 1490 struct tmpfs_node *node; 1491 1492 vp = v->a_vp; 1493 node = VP_TO_TMPFS_NODE(vp); 1494 if (node->tn_links == 0) 1495 vrecycle(vp); 1496 else 1497 tmpfs_check_mtime(vp); 1498 return (0); 1499 } 1500 1501 static int 1502 tmpfs_need_inactive(struct vop_need_inactive_args *ap) 1503 { 1504 struct vnode *vp; 1505 struct tmpfs_node *node; 1506 struct vm_object *obj; 1507 1508 vp = ap->a_vp; 1509 node = VP_TO_TMPFS_NODE(vp); 1510 if (node->tn_links == 0) 1511 goto need; 1512 if (vp->v_type == VREG) { 1513 obj = vp->v_object; 1514 if (obj->generation != obj->cleangeneration) 1515 goto need; 1516 } 1517 return (0); 1518 need: 1519 return (1); 1520 } 1521 1522 int 1523 tmpfs_reclaim(struct vop_reclaim_args *v) 1524 { 1525 struct vnode *vp; 1526 struct tmpfs_mount *tmp; 1527 struct tmpfs_node *node; 1528 bool unlock; 1529 1530 vp = v->a_vp; 1531 node = VP_TO_TMPFS_NODE(vp); 1532 tmp = VFS_TO_TMPFS(vp->v_mount); 1533 1534 if (vp->v_type == VREG) 1535 tmpfs_destroy_vobject(vp, node->tn_reg.tn_aobj); 1536 vp->v_object = NULL; 1537 1538 TMPFS_LOCK(tmp); 1539 TMPFS_NODE_LOCK(node); 1540 tmpfs_free_vp(vp); 1541 1542 /* 1543 * If the node referenced by this vnode was deleted by the user, 1544 * we must free its associated data structures (now that the vnode 1545 * is being reclaimed). 1546 */ 1547 unlock = true; 1548 if (node->tn_links == 0 && 1549 (node->tn_vpstate & TMPFS_VNODE_ALLOCATING) == 0) { 1550 node->tn_vpstate = TMPFS_VNODE_DOOMED; 1551 unlock = !tmpfs_free_node_locked(tmp, node, true); 1552 } 1553 1554 if (unlock) { 1555 TMPFS_NODE_UNLOCK(node); 1556 TMPFS_UNLOCK(tmp); 1557 } 1558 1559 MPASS(vp->v_data == NULL); 1560 return (0); 1561 } 1562 1563 int 1564 tmpfs_print(struct vop_print_args *v) 1565 { 1566 struct vnode *vp = v->a_vp; 1567 1568 struct tmpfs_node *node; 1569 1570 node = VP_TO_TMPFS_NODE(vp); 1571 1572 printf("tag VT_TMPFS, tmpfs_node %p, flags 0x%lx, links %jd\n", 1573 node, node->tn_flags, (uintmax_t)node->tn_links); 1574 printf("\tmode 0%o, owner %d, group %d, size %jd, status 0x%x\n", 1575 node->tn_mode, node->tn_uid, node->tn_gid, 1576 (intmax_t)node->tn_size, node->tn_status); 1577 1578 if (vp->v_type == VFIFO) 1579 fifo_printinfo(vp); 1580 1581 printf("\n"); 1582 1583 return (0); 1584 } 1585 1586 int 1587 tmpfs_pathconf(struct vop_pathconf_args *v) 1588 { 1589 struct vnode *vp = v->a_vp; 1590 int name = v->a_name; 1591 long *retval = v->a_retval; 1592 1593 int error; 1594 1595 error = 0; 1596 1597 switch (name) { 1598 case _PC_LINK_MAX: 1599 *retval = TMPFS_LINK_MAX; 1600 break; 1601 1602 case _PC_SYMLINK_MAX: 1603 *retval = MAXPATHLEN; 1604 break; 1605 1606 case _PC_NAME_MAX: 1607 *retval = NAME_MAX; 1608 break; 1609 1610 case _PC_PIPE_BUF: 1611 if (vp->v_type == VDIR || vp->v_type == VFIFO) 1612 *retval = PIPE_BUF; 1613 else 1614 error = EINVAL; 1615 break; 1616 1617 case _PC_CHOWN_RESTRICTED: 1618 *retval = 1; 1619 break; 1620 1621 case _PC_NO_TRUNC: 1622 *retval = 1; 1623 break; 1624 1625 case _PC_SYNC_IO: 1626 *retval = 1; 1627 break; 1628 1629 case _PC_FILESIZEBITS: 1630 *retval = 64; 1631 break; 1632 1633 default: 1634 error = vop_stdpathconf(v); 1635 } 1636 1637 return (error); 1638 } 1639 1640 static int 1641 tmpfs_vptofh(struct vop_vptofh_args *ap) 1642 /* 1643 vop_vptofh { 1644 IN struct vnode *a_vp; 1645 IN struct fid *a_fhp; 1646 }; 1647 */ 1648 { 1649 struct tmpfs_fid_data tfd; 1650 struct tmpfs_node *node; 1651 struct fid *fhp; 1652 1653 node = VP_TO_TMPFS_NODE(ap->a_vp); 1654 fhp = ap->a_fhp; 1655 fhp->fid_len = sizeof(tfd); 1656 1657 /* 1658 * Copy into fid_data from the stack to avoid unaligned pointer use. 1659 * See the comment in sys/mount.h on struct fid for details. 1660 */ 1661 tfd.tfd_id = node->tn_id; 1662 tfd.tfd_gen = node->tn_gen; 1663 memcpy(fhp->fid_data, &tfd, fhp->fid_len); 1664 1665 return (0); 1666 } 1667 1668 static int 1669 tmpfs_whiteout(struct vop_whiteout_args *ap) 1670 { 1671 struct vnode *dvp = ap->a_dvp; 1672 struct componentname *cnp = ap->a_cnp; 1673 struct tmpfs_dirent *de; 1674 1675 switch (ap->a_flags) { 1676 case LOOKUP: 1677 return (0); 1678 case CREATE: 1679 de = tmpfs_dir_lookup(VP_TO_TMPFS_DIR(dvp), NULL, cnp); 1680 if (de != NULL) 1681 return (de->td_node == NULL ? 0 : EEXIST); 1682 return (tmpfs_dir_whiteout_add(dvp, cnp)); 1683 case DELETE: 1684 tmpfs_dir_whiteout_remove(dvp, cnp); 1685 return (0); 1686 default: 1687 panic("tmpfs_whiteout: unknown op"); 1688 } 1689 } 1690 1691 static int 1692 tmpfs_vptocnp_dir(struct tmpfs_node *tn, struct tmpfs_node *tnp, 1693 struct tmpfs_dirent **pde) 1694 { 1695 struct tmpfs_dir_cursor dc; 1696 struct tmpfs_dirent *de; 1697 1698 for (de = tmpfs_dir_first(tnp, &dc); de != NULL; 1699 de = tmpfs_dir_next(tnp, &dc)) { 1700 if (de->td_node == tn) { 1701 *pde = de; 1702 return (0); 1703 } 1704 } 1705 return (ENOENT); 1706 } 1707 1708 static int 1709 tmpfs_vptocnp_fill(struct vnode *vp, struct tmpfs_node *tn, 1710 struct tmpfs_node *tnp, char *buf, size_t *buflen, struct vnode **dvp) 1711 { 1712 struct tmpfs_dirent *de; 1713 int error, i; 1714 1715 error = vn_vget_ino_gen(vp, tmpfs_vn_get_ino_alloc, tnp, LK_SHARED, 1716 dvp); 1717 if (error != 0) 1718 return (error); 1719 error = tmpfs_vptocnp_dir(tn, tnp, &de); 1720 if (error == 0) { 1721 i = *buflen; 1722 i -= de->td_namelen; 1723 if (i < 0) { 1724 error = ENOMEM; 1725 } else { 1726 bcopy(de->ud.td_name, buf + i, de->td_namelen); 1727 *buflen = i; 1728 } 1729 } 1730 if (error == 0) { 1731 if (vp != *dvp) 1732 VOP_UNLOCK(*dvp); 1733 } else { 1734 if (vp != *dvp) 1735 vput(*dvp); 1736 else 1737 vrele(vp); 1738 } 1739 return (error); 1740 } 1741 1742 static int 1743 tmpfs_vptocnp(struct vop_vptocnp_args *ap) 1744 { 1745 struct vnode *vp, **dvp; 1746 struct tmpfs_node *tn, *tnp, *tnp1; 1747 struct tmpfs_dirent *de; 1748 struct tmpfs_mount *tm; 1749 char *buf; 1750 size_t *buflen; 1751 int error; 1752 1753 vp = ap->a_vp; 1754 dvp = ap->a_vpp; 1755 buf = ap->a_buf; 1756 buflen = ap->a_buflen; 1757 1758 tm = VFS_TO_TMPFS(vp->v_mount); 1759 tn = VP_TO_TMPFS_NODE(vp); 1760 if (tn->tn_type == VDIR) { 1761 tnp = tn->tn_dir.tn_parent; 1762 if (tnp == NULL) 1763 return (ENOENT); 1764 tmpfs_ref_node(tnp); 1765 error = tmpfs_vptocnp_fill(vp, tn, tn->tn_dir.tn_parent, buf, 1766 buflen, dvp); 1767 tmpfs_free_node(tm, tnp); 1768 return (error); 1769 } 1770 restart: 1771 TMPFS_LOCK(tm); 1772 restart_locked: 1773 LIST_FOREACH_SAFE(tnp, &tm->tm_nodes_used, tn_entries, tnp1) { 1774 if (tnp->tn_type != VDIR) 1775 continue; 1776 TMPFS_NODE_LOCK(tnp); 1777 tmpfs_ref_node(tnp); 1778 1779 /* 1780 * tn_vnode cannot be instantiated while we hold the 1781 * node lock, so the directory cannot be changed while 1782 * we iterate over it. Do this to avoid instantiating 1783 * vnode for directories which cannot point to our 1784 * node. 1785 */ 1786 error = tnp->tn_vnode == NULL ? tmpfs_vptocnp_dir(tn, tnp, 1787 &de) : 0; 1788 1789 if (error == 0) { 1790 TMPFS_NODE_UNLOCK(tnp); 1791 TMPFS_UNLOCK(tm); 1792 error = tmpfs_vptocnp_fill(vp, tn, tnp, buf, buflen, 1793 dvp); 1794 if (error == 0) { 1795 tmpfs_free_node(tm, tnp); 1796 return (0); 1797 } 1798 if (VN_IS_DOOMED(vp)) { 1799 tmpfs_free_node(tm, tnp); 1800 return (ENOENT); 1801 } 1802 TMPFS_LOCK(tm); 1803 TMPFS_NODE_LOCK(tnp); 1804 } 1805 if (tmpfs_free_node_locked(tm, tnp, false)) { 1806 goto restart; 1807 } else { 1808 KASSERT(tnp->tn_refcount > 0, 1809 ("node %p refcount zero", tnp)); 1810 if (tnp->tn_attached) { 1811 tnp1 = LIST_NEXT(tnp, tn_entries); 1812 TMPFS_NODE_UNLOCK(tnp); 1813 } else { 1814 TMPFS_NODE_UNLOCK(tnp); 1815 goto restart_locked; 1816 } 1817 } 1818 } 1819 TMPFS_UNLOCK(tm); 1820 return (ENOENT); 1821 } 1822 1823 /* 1824 * Vnode operations vector used for files stored in a tmpfs file system. 1825 */ 1826 struct vop_vector tmpfs_vnodeop_entries = { 1827 .vop_default = &default_vnodeops, 1828 .vop_lookup = vfs_cache_lookup, 1829 .vop_cachedlookup = tmpfs_cached_lookup, 1830 .vop_create = tmpfs_create, 1831 .vop_mknod = tmpfs_mknod, 1832 .vop_open = tmpfs_open, 1833 .vop_close = tmpfs_close, 1834 .vop_fplookup_vexec = tmpfs_fplookup_vexec, 1835 .vop_fplookup_symlink = tmpfs_fplookup_symlink, 1836 .vop_access = tmpfs_access, 1837 .vop_stat = tmpfs_stat, 1838 .vop_getattr = tmpfs_getattr, 1839 .vop_setattr = tmpfs_setattr, 1840 .vop_read = tmpfs_read, 1841 .vop_read_pgcache = tmpfs_read_pgcache, 1842 .vop_write = tmpfs_write, 1843 .vop_deallocate = tmpfs_deallocate, 1844 .vop_fsync = tmpfs_fsync, 1845 .vop_remove = tmpfs_remove, 1846 .vop_link = tmpfs_link, 1847 .vop_rename = tmpfs_rename, 1848 .vop_mkdir = tmpfs_mkdir, 1849 .vop_rmdir = tmpfs_rmdir, 1850 .vop_symlink = tmpfs_symlink, 1851 .vop_readdir = tmpfs_readdir, 1852 .vop_readlink = tmpfs_readlink, 1853 .vop_inactive = tmpfs_inactive, 1854 .vop_need_inactive = tmpfs_need_inactive, 1855 .vop_reclaim = tmpfs_reclaim, 1856 .vop_print = tmpfs_print, 1857 .vop_pathconf = tmpfs_pathconf, 1858 .vop_vptofh = tmpfs_vptofh, 1859 .vop_whiteout = tmpfs_whiteout, 1860 .vop_bmap = VOP_EOPNOTSUPP, 1861 .vop_vptocnp = tmpfs_vptocnp, 1862 .vop_lock1 = vop_lock, 1863 .vop_unlock = vop_unlock, 1864 .vop_islocked = vop_islocked, 1865 .vop_add_writecount = vop_stdadd_writecount_nomsync, 1866 }; 1867 VFS_VOP_VECTOR_REGISTER(tmpfs_vnodeop_entries); 1868 1869 /* 1870 * Same vector for mounts which do not use namecache. 1871 */ 1872 struct vop_vector tmpfs_vnodeop_nonc_entries = { 1873 .vop_default = &tmpfs_vnodeop_entries, 1874 .vop_lookup = tmpfs_lookup, 1875 }; 1876 VFS_VOP_VECTOR_REGISTER(tmpfs_vnodeop_nonc_entries); 1877