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 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 39 #include <sys/param.h> 40 #include <sys/systm.h> 41 #include <sys/dirent.h> 42 #include <sys/extattr.h> 43 #include <sys/fcntl.h> 44 #include <sys/file.h> 45 #include <sys/filio.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/smr.h> 56 #include <sys/stat.h> 57 #include <sys/sysctl.h> 58 #include <sys/unistd.h> 59 #include <sys/vnode.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 #include <vm/vm_page.h> 67 #include <vm/vm_pager.h> 68 #include <vm/swap_pager.h> 69 70 #include <fs/tmpfs/tmpfs_vnops.h> 71 #include <fs/tmpfs/tmpfs.h> 72 73 SYSCTL_DECL(_vfs_tmpfs); 74 VFS_SMR_DECLARE; 75 76 static volatile int tmpfs_rename_restarts; 77 SYSCTL_INT(_vfs_tmpfs, OID_AUTO, rename_restarts, CTLFLAG_RD, 78 __DEVOLATILE(int *, &tmpfs_rename_restarts), 0, 79 "Times rename had to restart due to lock contention"); 80 81 MALLOC_DEFINE(M_TMPFSEA, "tmpfs extattr", "tmpfs extattr structure"); 82 83 static int 84 tmpfs_vn_get_ino_alloc(struct mount *mp, void *arg, int lkflags, 85 struct vnode **rvp) 86 { 87 88 return (tmpfs_alloc_vp(mp, arg, lkflags, rvp)); 89 } 90 91 static int 92 tmpfs_lookup1(struct vnode *dvp, struct vnode **vpp, struct componentname *cnp) 93 { 94 struct tmpfs_dirent *de; 95 struct tmpfs_node *dnode, *pnode; 96 struct tmpfs_mount *tm; 97 int error; 98 99 /* Caller assumes responsibility for ensuring access (VEXEC). */ 100 dnode = VP_TO_TMPFS_DIR(dvp); 101 *vpp = NULLVP; 102 103 /* We cannot be requesting the parent directory of the root node. */ 104 MPASS(IMPLIES(dnode->tn_type == VDIR && 105 dnode->tn_dir.tn_parent == dnode, 106 !(cnp->cn_flags & ISDOTDOT))); 107 108 TMPFS_ASSERT_LOCKED(dnode); 109 if (dnode->tn_dir.tn_parent == NULL) { 110 error = ENOENT; 111 goto out; 112 } 113 if (cnp->cn_flags & ISDOTDOT) { 114 tm = VFS_TO_TMPFS(dvp->v_mount); 115 pnode = dnode->tn_dir.tn_parent; 116 tmpfs_ref_node(pnode); 117 error = vn_vget_ino_gen(dvp, tmpfs_vn_get_ino_alloc, 118 pnode, cnp->cn_lkflags, vpp); 119 tmpfs_free_node(tm, pnode); 120 if (error != 0) 121 goto out; 122 } else if (cnp->cn_namelen == 1 && cnp->cn_nameptr[0] == '.') { 123 VREF(dvp); 124 *vpp = dvp; 125 error = 0; 126 } else { 127 de = tmpfs_dir_lookup(dnode, NULL, cnp); 128 if (de != NULL && de->td_node == NULL) 129 cnp->cn_flags |= ISWHITEOUT; 130 if (de == NULL || de->td_node == NULL) { 131 /* 132 * The entry was not found in the directory. 133 * This is OK if we are creating or renaming an 134 * entry and are working on the last component of 135 * the path name. 136 */ 137 if ((cnp->cn_flags & ISLASTCN) && 138 (cnp->cn_nameiop == CREATE || \ 139 cnp->cn_nameiop == RENAME || 140 (cnp->cn_nameiop == DELETE && 141 cnp->cn_flags & DOWHITEOUT && 142 cnp->cn_flags & ISWHITEOUT))) { 143 error = VOP_ACCESS(dvp, VWRITE, cnp->cn_cred, 144 curthread); 145 if (error != 0) 146 goto out; 147 148 error = EJUSTRETURN; 149 } else 150 error = ENOENT; 151 } else { 152 struct tmpfs_node *tnode; 153 154 /* 155 * The entry was found, so get its associated 156 * tmpfs_node. 157 */ 158 tnode = de->td_node; 159 160 /* 161 * If we are not at the last path component and 162 * found a non-directory or non-link entry (which 163 * may itself be pointing to a directory), raise 164 * an error. 165 */ 166 if ((tnode->tn_type != VDIR && 167 tnode->tn_type != VLNK) && 168 !(cnp->cn_flags & ISLASTCN)) { 169 error = ENOTDIR; 170 goto out; 171 } 172 173 /* 174 * If we are deleting or renaming the entry, keep 175 * track of its tmpfs_dirent so that it can be 176 * easily deleted later. 177 */ 178 if ((cnp->cn_flags & ISLASTCN) && 179 (cnp->cn_nameiop == DELETE || 180 cnp->cn_nameiop == RENAME)) { 181 error = VOP_ACCESS(dvp, VWRITE, cnp->cn_cred, 182 curthread); 183 if (error != 0) 184 goto out; 185 186 /* Allocate a new vnode on the matching entry. */ 187 error = tmpfs_alloc_vp(dvp->v_mount, tnode, 188 cnp->cn_lkflags, vpp); 189 if (error != 0) 190 goto out; 191 192 if ((dnode->tn_mode & S_ISTXT) && 193 VOP_ACCESS(dvp, VADMIN, cnp->cn_cred, 194 curthread) && VOP_ACCESS(*vpp, VADMIN, 195 cnp->cn_cred, curthread)) { 196 error = EPERM; 197 vput(*vpp); 198 *vpp = NULL; 199 goto out; 200 } 201 } else { 202 error = tmpfs_alloc_vp(dvp->v_mount, tnode, 203 cnp->cn_lkflags, vpp); 204 if (error != 0) 205 goto out; 206 } 207 } 208 } 209 210 /* 211 * Store the result of this lookup in the cache. Avoid this if the 212 * request was for creation, as it does not improve timings on 213 * emprical tests. 214 */ 215 if ((cnp->cn_flags & MAKEENTRY) != 0 && tmpfs_use_nc(dvp)) 216 cache_enter(dvp, *vpp, cnp); 217 218 out: 219 #ifdef INVARIANTS 220 /* 221 * If there were no errors, *vpp cannot be null and it must be 222 * locked. 223 */ 224 if (error == 0) { 225 MPASS(*vpp != NULLVP); 226 ASSERT_VOP_LOCKED(*vpp, __func__); 227 } else { 228 MPASS(*vpp == NULL); 229 } 230 #endif 231 232 return (error); 233 } 234 235 static int 236 tmpfs_cached_lookup(struct vop_cachedlookup_args *v) 237 { 238 239 return (tmpfs_lookup1(v->a_dvp, v->a_vpp, v->a_cnp)); 240 } 241 242 static int 243 tmpfs_lookup(struct vop_lookup_args *v) 244 { 245 struct vnode *dvp = v->a_dvp; 246 struct vnode **vpp = v->a_vpp; 247 struct componentname *cnp = v->a_cnp; 248 int error; 249 250 /* Check accessibility of requested node as a first step. */ 251 error = vn_dir_check_exec(dvp, cnp); 252 if (error != 0) 253 return (error); 254 255 return (tmpfs_lookup1(dvp, vpp, cnp)); 256 } 257 258 static int 259 tmpfs_create(struct vop_create_args *v) 260 { 261 struct vnode *dvp = v->a_dvp; 262 struct vnode **vpp = v->a_vpp; 263 struct componentname *cnp = v->a_cnp; 264 struct vattr *vap = v->a_vap; 265 int error; 266 267 MPASS(vap->va_type == VREG || vap->va_type == VSOCK); 268 269 error = tmpfs_alloc_file(dvp, vpp, vap, cnp, NULL); 270 if (error == 0 && (cnp->cn_flags & MAKEENTRY) != 0 && tmpfs_use_nc(dvp)) 271 cache_enter(dvp, *vpp, cnp); 272 return (error); 273 } 274 275 static int 276 tmpfs_mknod(struct vop_mknod_args *v) 277 { 278 struct vnode *dvp = v->a_dvp; 279 struct vnode **vpp = v->a_vpp; 280 struct componentname *cnp = v->a_cnp; 281 struct vattr *vap = v->a_vap; 282 283 if (vap->va_type != VBLK && vap->va_type != VCHR && 284 vap->va_type != VFIFO) 285 return (EINVAL); 286 287 return (tmpfs_alloc_file(dvp, vpp, vap, cnp, NULL)); 288 } 289 290 struct fileops tmpfs_fnops; 291 292 static int 293 tmpfs_open(struct vop_open_args *v) 294 { 295 struct vnode *vp; 296 struct tmpfs_node *node; 297 struct file *fp; 298 int error, mode; 299 300 vp = v->a_vp; 301 mode = v->a_mode; 302 node = VP_TO_TMPFS_NODE(vp); 303 304 /* 305 * The file is still active but all its names have been removed 306 * (e.g. by a "rmdir $(pwd)"). It cannot be opened any more as 307 * it is about to die. 308 */ 309 if (node->tn_links < 1) 310 return (ENOENT); 311 312 /* If the file is marked append-only, deny write requests. */ 313 if (node->tn_flags & APPEND && (mode & (FWRITE | O_APPEND)) == FWRITE) 314 error = EPERM; 315 else { 316 error = 0; 317 /* For regular files, the call below is nop. */ 318 KASSERT(vp->v_type != VREG || (node->tn_reg.tn_aobj->flags & 319 OBJ_DEAD) == 0, ("dead object")); 320 vnode_create_vobject(vp, node->tn_size, v->a_td); 321 } 322 323 fp = v->a_fp; 324 MPASS(fp == NULL || fp->f_data == NULL); 325 if (error == 0 && fp != NULL && vp->v_type == VREG) { 326 tmpfs_ref_node(node); 327 finit_vnode(fp, mode, node, &tmpfs_fnops); 328 } 329 330 return (error); 331 } 332 333 static int 334 tmpfs_close(struct vop_close_args *v) 335 { 336 struct vnode *vp = v->a_vp; 337 338 /* Update node times. */ 339 tmpfs_update(vp); 340 341 return (0); 342 } 343 344 int 345 tmpfs_fo_close(struct file *fp, struct thread *td) 346 { 347 struct tmpfs_node *node; 348 349 node = fp->f_data; 350 if (node != NULL) { 351 MPASS(node->tn_type == VREG); 352 tmpfs_free_node(node->tn_reg.tn_tmp, node); 353 } 354 return (vnops.fo_close(fp, td)); 355 } 356 357 /* 358 * VOP_FPLOOKUP_VEXEC routines are subject to special circumstances, see 359 * the comment above cache_fplookup for details. 360 */ 361 int 362 tmpfs_fplookup_vexec(struct vop_fplookup_vexec_args *v) 363 { 364 struct vnode *vp; 365 struct tmpfs_node *node; 366 struct ucred *cred; 367 mode_t all_x, mode; 368 369 vp = v->a_vp; 370 node = VP_TO_TMPFS_NODE_SMR(vp); 371 if (__predict_false(node == NULL)) 372 return (EAGAIN); 373 374 all_x = S_IXUSR | S_IXGRP | S_IXOTH; 375 mode = atomic_load_short(&node->tn_mode); 376 if (__predict_true((mode & all_x) == all_x)) 377 return (0); 378 379 cred = v->a_cred; 380 return (vaccess_vexec_smr(mode, node->tn_uid, node->tn_gid, cred)); 381 } 382 383 static int 384 tmpfs_access_locked(struct vnode *vp, struct tmpfs_node *node, 385 accmode_t accmode, struct ucred *cred) 386 { 387 #ifdef DEBUG_VFS_LOCKS 388 if (!mtx_owned(TMPFS_NODE_MTX(node))) { 389 ASSERT_VOP_LOCKED(vp, 390 "tmpfs_access_locked needs locked vnode or node"); 391 } 392 #endif 393 394 if ((accmode & VWRITE) != 0 && (node->tn_flags & IMMUTABLE) != 0) 395 return (EPERM); 396 return (vaccess(vp->v_type, node->tn_mode, node->tn_uid, node->tn_gid, 397 accmode, cred)); 398 } 399 400 int 401 tmpfs_access(struct vop_access_args *v) 402 { 403 struct vnode *vp = v->a_vp; 404 struct ucred *cred = v->a_cred; 405 struct tmpfs_node *node = VP_TO_TMPFS_NODE(vp); 406 mode_t all_x = S_IXUSR | S_IXGRP | S_IXOTH; 407 accmode_t accmode = v->a_accmode; 408 409 /* 410 * Common case path lookup. 411 */ 412 if (__predict_true(accmode == VEXEC && 413 (node->tn_mode & all_x) == all_x)) 414 return (0); 415 416 switch (vp->v_type) { 417 case VDIR: 418 /* FALLTHROUGH */ 419 case VLNK: 420 /* FALLTHROUGH */ 421 case VREG: 422 if ((accmode & VWRITE) != 0 && 423 (vp->v_mount->mnt_flag & MNT_RDONLY) != 0) 424 return (EROFS); 425 break; 426 427 case VBLK: 428 /* FALLTHROUGH */ 429 case VCHR: 430 /* FALLTHROUGH */ 431 case VSOCK: 432 /* FALLTHROUGH */ 433 case VFIFO: 434 break; 435 436 default: 437 return (EINVAL); 438 } 439 440 return (tmpfs_access_locked(vp, node, accmode, cred)); 441 } 442 443 int 444 tmpfs_stat(struct vop_stat_args *v) 445 { 446 struct vnode *vp = v->a_vp; 447 struct stat *sb = v->a_sb; 448 struct tmpfs_node *node; 449 int error; 450 451 node = VP_TO_TMPFS_NODE(vp); 452 453 tmpfs_update_getattr(vp); 454 455 error = vop_stat_helper_pre(v); 456 if (__predict_false(error)) 457 return (error); 458 459 sb->st_dev = vp->v_mount->mnt_stat.f_fsid.val[0]; 460 sb->st_ino = node->tn_id; 461 sb->st_mode = node->tn_mode | VTTOIF(vp->v_type); 462 sb->st_nlink = node->tn_links; 463 sb->st_uid = node->tn_uid; 464 sb->st_gid = node->tn_gid; 465 sb->st_rdev = (vp->v_type == VBLK || vp->v_type == VCHR) ? 466 node->tn_rdev : NODEV; 467 sb->st_size = node->tn_size; 468 sb->st_atim.tv_sec = node->tn_atime.tv_sec; 469 sb->st_atim.tv_nsec = node->tn_atime.tv_nsec; 470 sb->st_mtim.tv_sec = node->tn_mtime.tv_sec; 471 sb->st_mtim.tv_nsec = node->tn_mtime.tv_nsec; 472 sb->st_ctim.tv_sec = node->tn_ctime.tv_sec; 473 sb->st_ctim.tv_nsec = node->tn_ctime.tv_nsec; 474 sb->st_birthtim.tv_sec = node->tn_birthtime.tv_sec; 475 sb->st_birthtim.tv_nsec = node->tn_birthtime.tv_nsec; 476 sb->st_blksize = PAGE_SIZE; 477 sb->st_flags = node->tn_flags; 478 sb->st_gen = node->tn_gen; 479 if (vp->v_type == VREG) { 480 #ifdef __ILP32__ 481 vm_object_t obj = node->tn_reg.tn_aobj; 482 483 /* Handle torn read */ 484 VM_OBJECT_RLOCK(obj); 485 #endif 486 sb->st_blocks = ptoa(node->tn_reg.tn_pages); 487 #ifdef __ILP32__ 488 VM_OBJECT_RUNLOCK(obj); 489 #endif 490 } else { 491 sb->st_blocks = node->tn_size; 492 } 493 sb->st_blocks /= S_BLKSIZE; 494 return (vop_stat_helper_post(v, error)); 495 } 496 497 int 498 tmpfs_getattr(struct vop_getattr_args *v) 499 { 500 struct vnode *vp = v->a_vp; 501 struct vattr *vap = v->a_vap; 502 struct tmpfs_node *node; 503 504 node = VP_TO_TMPFS_NODE(vp); 505 506 tmpfs_update_getattr(vp); 507 508 vap->va_type = vp->v_type; 509 vap->va_mode = node->tn_mode; 510 vap->va_nlink = node->tn_links; 511 vap->va_uid = node->tn_uid; 512 vap->va_gid = node->tn_gid; 513 vap->va_fsid = vp->v_mount->mnt_stat.f_fsid.val[0]; 514 vap->va_fileid = node->tn_id; 515 vap->va_size = node->tn_size; 516 vap->va_blocksize = PAGE_SIZE; 517 vap->va_atime = node->tn_atime; 518 vap->va_mtime = node->tn_mtime; 519 vap->va_ctime = node->tn_ctime; 520 vap->va_birthtime = node->tn_birthtime; 521 vap->va_gen = node->tn_gen; 522 vap->va_flags = node->tn_flags; 523 vap->va_rdev = (vp->v_type == VBLK || vp->v_type == VCHR) ? 524 node->tn_rdev : NODEV; 525 if (vp->v_type == VREG) { 526 #ifdef __ILP32__ 527 vm_object_t obj = node->tn_reg.tn_aobj; 528 529 VM_OBJECT_RLOCK(obj); 530 #endif 531 vap->va_bytes = ptoa(node->tn_reg.tn_pages); 532 #ifdef __ILP32__ 533 VM_OBJECT_RUNLOCK(obj); 534 #endif 535 } else { 536 vap->va_bytes = node->tn_size; 537 } 538 vap->va_filerev = 0; 539 540 return (0); 541 } 542 543 int 544 tmpfs_setattr(struct vop_setattr_args *v) 545 { 546 struct vnode *vp = v->a_vp; 547 struct vattr *vap = v->a_vap; 548 struct ucred *cred = v->a_cred; 549 struct thread *td = curthread; 550 551 int error; 552 553 ASSERT_VOP_IN_SEQC(vp); 554 555 error = 0; 556 557 /* Abort if any unsettable attribute is given. */ 558 if (vap->va_type != VNON || 559 vap->va_nlink != VNOVAL || 560 vap->va_fsid != VNOVAL || 561 vap->va_fileid != VNOVAL || 562 vap->va_blocksize != VNOVAL || 563 vap->va_gen != VNOVAL || 564 vap->va_rdev != VNOVAL || 565 vap->va_bytes != VNOVAL) 566 error = EINVAL; 567 568 if (error == 0 && (vap->va_flags != VNOVAL)) 569 error = tmpfs_chflags(vp, vap->va_flags, cred, td); 570 571 if (error == 0 && (vap->va_size != VNOVAL)) 572 error = tmpfs_chsize(vp, vap->va_size, cred, td); 573 574 if (error == 0 && (vap->va_uid != VNOVAL || vap->va_gid != VNOVAL)) 575 error = tmpfs_chown(vp, vap->va_uid, vap->va_gid, cred, td); 576 577 if (error == 0 && (vap->va_mode != (mode_t)VNOVAL)) 578 error = tmpfs_chmod(vp, vap->va_mode, cred, td); 579 580 if (error == 0 && ((vap->va_atime.tv_sec != VNOVAL && 581 vap->va_atime.tv_nsec != VNOVAL) || 582 (vap->va_mtime.tv_sec != VNOVAL && 583 vap->va_mtime.tv_nsec != VNOVAL) || 584 (vap->va_birthtime.tv_sec != VNOVAL && 585 vap->va_birthtime.tv_nsec != VNOVAL))) 586 error = tmpfs_chtimes(vp, vap, cred, td); 587 588 /* 589 * Update the node times. We give preference to the error codes 590 * generated by this function rather than the ones that may arise 591 * from tmpfs_update. 592 */ 593 tmpfs_update(vp); 594 595 return (error); 596 } 597 598 static int 599 tmpfs_read(struct vop_read_args *v) 600 { 601 struct vnode *vp; 602 struct uio *uio; 603 struct tmpfs_node *node; 604 605 vp = v->a_vp; 606 if (vp->v_type != VREG) 607 return (EISDIR); 608 uio = v->a_uio; 609 if (uio->uio_offset < 0) 610 return (EINVAL); 611 node = VP_TO_TMPFS_NODE(vp); 612 tmpfs_set_accessed(VFS_TO_TMPFS(vp->v_mount), node); 613 return (uiomove_object(node->tn_reg.tn_aobj, node->tn_size, uio)); 614 } 615 616 static int 617 tmpfs_read_pgcache(struct vop_read_pgcache_args *v) 618 { 619 struct vnode *vp; 620 struct tmpfs_node *node; 621 vm_object_t object; 622 off_t size; 623 int error; 624 625 vp = v->a_vp; 626 VNPASS((vn_irflag_read(vp) & VIRF_PGREAD) != 0, vp); 627 628 if (v->a_uio->uio_offset < 0) 629 return (EINVAL); 630 631 error = EJUSTRETURN; 632 vfs_smr_enter(); 633 634 node = VP_TO_TMPFS_NODE_SMR(vp); 635 if (node == NULL) 636 goto out_smr; 637 MPASS(node->tn_type == VREG); 638 MPASS(node->tn_refcount >= 1); 639 object = node->tn_reg.tn_aobj; 640 if (object == NULL) 641 goto out_smr; 642 643 MPASS(object->type == tmpfs_pager_type); 644 MPASS((object->flags & (OBJ_ANON | OBJ_DEAD | OBJ_SWAP)) == 645 OBJ_SWAP); 646 if (!VN_IS_DOOMED(vp)) { 647 /* size cannot become shorter due to rangelock. */ 648 size = node->tn_size; 649 tmpfs_set_accessed(node->tn_reg.tn_tmp, node); 650 vfs_smr_exit(); 651 error = uiomove_object(object, size, v->a_uio); 652 return (error); 653 } 654 out_smr: 655 vfs_smr_exit(); 656 return (error); 657 } 658 659 static int 660 tmpfs_write(struct vop_write_args *v) 661 { 662 struct vnode *vp; 663 struct uio *uio; 664 struct tmpfs_node *node; 665 off_t oldsize; 666 ssize_t r; 667 int error, ioflag; 668 mode_t newmode; 669 670 vp = v->a_vp; 671 uio = v->a_uio; 672 ioflag = v->a_ioflag; 673 error = 0; 674 node = VP_TO_TMPFS_NODE(vp); 675 oldsize = node->tn_size; 676 677 if (uio->uio_offset < 0 || vp->v_type != VREG) 678 return (EINVAL); 679 if (uio->uio_resid == 0) 680 return (0); 681 if (ioflag & IO_APPEND) 682 uio->uio_offset = node->tn_size; 683 error = vn_rlimit_fsizex(vp, uio, VFS_TO_TMPFS(vp->v_mount)-> 684 tm_maxfilesize, &r, uio->uio_td); 685 if (error != 0) { 686 vn_rlimit_fsizex_res(uio, r); 687 return (error); 688 } 689 690 if (uio->uio_offset + uio->uio_resid > node->tn_size) { 691 error = tmpfs_reg_resize(vp, uio->uio_offset + uio->uio_resid, 692 FALSE); 693 if (error != 0) 694 goto out; 695 } 696 697 error = uiomove_object(node->tn_reg.tn_aobj, node->tn_size, uio); 698 node->tn_status |= TMPFS_NODE_MODIFIED | TMPFS_NODE_CHANGED; 699 node->tn_accessed = true; 700 if (node->tn_mode & (S_ISUID | S_ISGID)) { 701 if (priv_check_cred(v->a_cred, PRIV_VFS_RETAINSUGID)) { 702 newmode = node->tn_mode & ~(S_ISUID | S_ISGID); 703 vn_seqc_write_begin(vp); 704 atomic_store_short(&node->tn_mode, newmode); 705 vn_seqc_write_end(vp); 706 } 707 } 708 if (error != 0) 709 (void)tmpfs_reg_resize(vp, oldsize, TRUE); 710 711 out: 712 MPASS(IMPLIES(error == 0, uio->uio_resid == 0)); 713 MPASS(IMPLIES(error != 0, oldsize == node->tn_size)); 714 715 vn_rlimit_fsizex_res(uio, r); 716 return (error); 717 } 718 719 static int 720 tmpfs_deallocate(struct vop_deallocate_args *v) 721 { 722 return (tmpfs_reg_punch_hole(v->a_vp, v->a_offset, v->a_len)); 723 } 724 725 static int 726 tmpfs_fsync(struct vop_fsync_args *v) 727 { 728 struct vnode *vp = v->a_vp; 729 730 tmpfs_check_mtime(vp); 731 tmpfs_update(vp); 732 733 return (0); 734 } 735 736 static int 737 tmpfs_remove(struct vop_remove_args *v) 738 { 739 struct vnode *dvp = v->a_dvp; 740 struct vnode *vp = v->a_vp; 741 742 int error; 743 struct tmpfs_dirent *de; 744 struct tmpfs_mount *tmp; 745 struct tmpfs_node *dnode; 746 struct tmpfs_node *node; 747 748 if (vp->v_type == VDIR) { 749 error = EISDIR; 750 goto out; 751 } 752 753 dnode = VP_TO_TMPFS_DIR(dvp); 754 node = VP_TO_TMPFS_NODE(vp); 755 tmp = VFS_TO_TMPFS(vp->v_mount); 756 de = tmpfs_dir_lookup(dnode, node, v->a_cnp); 757 MPASS(de != NULL); 758 759 /* Files marked as immutable or append-only cannot be deleted. */ 760 if ((node->tn_flags & (IMMUTABLE | APPEND | NOUNLINK)) || 761 (dnode->tn_flags & APPEND)) { 762 error = EPERM; 763 goto out; 764 } 765 766 /* Remove the entry from the directory; as it is a file, we do not 767 * have to change the number of hard links of the directory. */ 768 tmpfs_dir_detach(dvp, de); 769 if (v->a_cnp->cn_flags & DOWHITEOUT) 770 tmpfs_dir_whiteout_add(dvp, v->a_cnp); 771 772 /* Free the directory entry we just deleted. Note that the node 773 * referred by it will not be removed until the vnode is really 774 * reclaimed. */ 775 tmpfs_free_dirent(tmp, de); 776 777 node->tn_status |= TMPFS_NODE_CHANGED; 778 node->tn_accessed = true; 779 error = 0; 780 781 out: 782 return (error); 783 } 784 785 static int 786 tmpfs_link(struct vop_link_args *v) 787 { 788 struct vnode *dvp = v->a_tdvp; 789 struct vnode *vp = v->a_vp; 790 struct componentname *cnp = v->a_cnp; 791 792 int error; 793 struct tmpfs_dirent *de; 794 struct tmpfs_node *node; 795 796 MPASS(dvp != vp); /* XXX When can this be false? */ 797 node = VP_TO_TMPFS_NODE(vp); 798 799 /* Ensure that we do not overflow the maximum number of links imposed 800 * by the system. */ 801 MPASS(node->tn_links <= TMPFS_LINK_MAX); 802 if (node->tn_links == TMPFS_LINK_MAX) { 803 error = EMLINK; 804 goto out; 805 } 806 807 /* We cannot create links of files marked immutable or append-only. */ 808 if (node->tn_flags & (IMMUTABLE | APPEND)) { 809 error = EPERM; 810 goto out; 811 } 812 813 /* Allocate a new directory entry to represent the node. */ 814 error = tmpfs_alloc_dirent(VFS_TO_TMPFS(vp->v_mount), node, 815 cnp->cn_nameptr, cnp->cn_namelen, &de); 816 if (error != 0) 817 goto out; 818 819 /* Insert the new directory entry into the appropriate directory. */ 820 if (cnp->cn_flags & ISWHITEOUT) 821 tmpfs_dir_whiteout_remove(dvp, cnp); 822 tmpfs_dir_attach(dvp, de); 823 824 /* vp link count has changed, so update node times. */ 825 node->tn_status |= TMPFS_NODE_CHANGED; 826 tmpfs_update(vp); 827 828 error = 0; 829 830 out: 831 return (error); 832 } 833 834 /* 835 * We acquire all but fdvp locks using non-blocking acquisitions. If we 836 * fail to acquire any lock in the path we will drop all held locks, 837 * acquire the new lock in a blocking fashion, and then release it and 838 * restart the rename. This acquire/release step ensures that we do not 839 * spin on a lock waiting for release. On error release all vnode locks 840 * and decrement references the way tmpfs_rename() would do. 841 */ 842 static int 843 tmpfs_rename_relock(struct vnode *fdvp, struct vnode **fvpp, 844 struct vnode *tdvp, struct vnode **tvpp, 845 struct componentname *fcnp, struct componentname *tcnp) 846 { 847 struct vnode *nvp; 848 struct mount *mp; 849 struct tmpfs_dirent *de; 850 int error, restarts = 0; 851 852 VOP_UNLOCK(tdvp); 853 if (*tvpp != NULL && *tvpp != tdvp) 854 VOP_UNLOCK(*tvpp); 855 mp = fdvp->v_mount; 856 857 relock: 858 restarts += 1; 859 error = vn_lock(fdvp, LK_EXCLUSIVE); 860 if (error) 861 goto releout; 862 if (vn_lock(tdvp, LK_EXCLUSIVE | LK_NOWAIT) != 0) { 863 VOP_UNLOCK(fdvp); 864 error = vn_lock(tdvp, LK_EXCLUSIVE); 865 if (error) 866 goto releout; 867 VOP_UNLOCK(tdvp); 868 goto relock; 869 } 870 /* 871 * Re-resolve fvp to be certain it still exists and fetch the 872 * correct vnode. 873 */ 874 de = tmpfs_dir_lookup(VP_TO_TMPFS_DIR(fdvp), NULL, fcnp); 875 if (de == NULL) { 876 VOP_UNLOCK(fdvp); 877 VOP_UNLOCK(tdvp); 878 if ((fcnp->cn_flags & ISDOTDOT) != 0 || 879 (fcnp->cn_namelen == 1 && fcnp->cn_nameptr[0] == '.')) 880 error = EINVAL; 881 else 882 error = ENOENT; 883 goto releout; 884 } 885 error = tmpfs_alloc_vp(mp, de->td_node, LK_EXCLUSIVE | LK_NOWAIT, &nvp); 886 if (error != 0) { 887 VOP_UNLOCK(fdvp); 888 VOP_UNLOCK(tdvp); 889 if (error != EBUSY) 890 goto releout; 891 error = tmpfs_alloc_vp(mp, de->td_node, LK_EXCLUSIVE, &nvp); 892 if (error != 0) 893 goto releout; 894 VOP_UNLOCK(nvp); 895 /* 896 * Concurrent rename race. 897 */ 898 if (nvp == tdvp) { 899 vrele(nvp); 900 error = EINVAL; 901 goto releout; 902 } 903 vrele(*fvpp); 904 *fvpp = nvp; 905 goto relock; 906 } 907 vrele(*fvpp); 908 *fvpp = nvp; 909 VOP_UNLOCK(*fvpp); 910 /* 911 * Re-resolve tvp and acquire the vnode lock if present. 912 */ 913 de = tmpfs_dir_lookup(VP_TO_TMPFS_DIR(tdvp), NULL, tcnp); 914 /* 915 * If tvp disappeared we just carry on. 916 */ 917 if (de == NULL && *tvpp != NULL) { 918 vrele(*tvpp); 919 *tvpp = NULL; 920 } 921 /* 922 * Get the tvp ino if the lookup succeeded. We may have to restart 923 * if the non-blocking acquire fails. 924 */ 925 if (de != NULL) { 926 nvp = NULL; 927 error = tmpfs_alloc_vp(mp, de->td_node, 928 LK_EXCLUSIVE | LK_NOWAIT, &nvp); 929 if (*tvpp != NULL) 930 vrele(*tvpp); 931 *tvpp = nvp; 932 if (error != 0) { 933 VOP_UNLOCK(fdvp); 934 VOP_UNLOCK(tdvp); 935 if (error != EBUSY) 936 goto releout; 937 error = tmpfs_alloc_vp(mp, de->td_node, LK_EXCLUSIVE, 938 &nvp); 939 if (error != 0) 940 goto releout; 941 VOP_UNLOCK(nvp); 942 /* 943 * fdvp contains fvp, thus tvp (=fdvp) is not empty. 944 */ 945 if (nvp == fdvp) { 946 error = ENOTEMPTY; 947 goto releout; 948 } 949 goto relock; 950 } 951 } 952 tmpfs_rename_restarts += restarts; 953 954 return (0); 955 956 releout: 957 vrele(fdvp); 958 vrele(*fvpp); 959 vrele(tdvp); 960 if (*tvpp != NULL) 961 vrele(*tvpp); 962 tmpfs_rename_restarts += restarts; 963 964 return (error); 965 } 966 967 static int 968 tmpfs_rename(struct vop_rename_args *v) 969 { 970 struct vnode *fdvp = v->a_fdvp; 971 struct vnode *fvp = v->a_fvp; 972 struct componentname *fcnp = v->a_fcnp; 973 struct vnode *tdvp = v->a_tdvp; 974 struct vnode *tvp = v->a_tvp; 975 struct componentname *tcnp = v->a_tcnp; 976 char *newname; 977 struct tmpfs_dirent *de; 978 struct tmpfs_mount *tmp; 979 struct tmpfs_node *fdnode; 980 struct tmpfs_node *fnode; 981 struct tmpfs_node *tnode; 982 struct tmpfs_node *tdnode; 983 int error; 984 bool want_seqc_end; 985 986 want_seqc_end = false; 987 988 /* 989 * Disallow cross-device renames. 990 * XXX Why isn't this done by the caller? 991 */ 992 if (fvp->v_mount != tdvp->v_mount || 993 (tvp != NULL && fvp->v_mount != tvp->v_mount)) { 994 error = EXDEV; 995 goto out; 996 } 997 998 /* If source and target are the same file, there is nothing to do. */ 999 if (fvp == tvp) { 1000 error = 0; 1001 goto out; 1002 } 1003 1004 /* 1005 * If we need to move the directory between entries, lock the 1006 * source so that we can safely operate on it. 1007 */ 1008 if (fdvp != tdvp && fdvp != tvp) { 1009 if (vn_lock(fdvp, LK_EXCLUSIVE | LK_NOWAIT) != 0) { 1010 error = tmpfs_rename_relock(fdvp, &fvp, tdvp, &tvp, 1011 fcnp, tcnp); 1012 if (error != 0) 1013 return (error); 1014 ASSERT_VOP_ELOCKED(fdvp, 1015 "tmpfs_rename: fdvp not locked"); 1016 ASSERT_VOP_ELOCKED(tdvp, 1017 "tmpfs_rename: tdvp not locked"); 1018 if (tvp != NULL) 1019 ASSERT_VOP_ELOCKED(tvp, 1020 "tmpfs_rename: tvp not locked"); 1021 if (fvp == tvp) { 1022 error = 0; 1023 goto out_locked; 1024 } 1025 } 1026 } 1027 1028 /* 1029 * Avoid manipulating '.' and '..' entries. 1030 */ 1031 if ((fcnp->cn_flags & ISDOTDOT) != 0 || 1032 (fcnp->cn_namelen == 1 && fcnp->cn_nameptr[0] == '.')) { 1033 error = EINVAL; 1034 goto out_locked; 1035 } 1036 1037 if (tvp != NULL) 1038 vn_seqc_write_begin(tvp); 1039 vn_seqc_write_begin(tdvp); 1040 vn_seqc_write_begin(fvp); 1041 vn_seqc_write_begin(fdvp); 1042 want_seqc_end = true; 1043 1044 tmp = VFS_TO_TMPFS(tdvp->v_mount); 1045 tdnode = VP_TO_TMPFS_DIR(tdvp); 1046 tnode = (tvp == NULL) ? NULL : VP_TO_TMPFS_NODE(tvp); 1047 fdnode = VP_TO_TMPFS_DIR(fdvp); 1048 fnode = VP_TO_TMPFS_NODE(fvp); 1049 de = tmpfs_dir_lookup(fdnode, fnode, fcnp); 1050 1051 /* 1052 * Entry can disappear before we lock fdvp. 1053 */ 1054 if (de == NULL) { 1055 if ((fcnp->cn_flags & ISDOTDOT) != 0 || 1056 (fcnp->cn_namelen == 1 && fcnp->cn_nameptr[0] == '.')) 1057 error = EINVAL; 1058 else 1059 error = ENOENT; 1060 goto out_locked; 1061 } 1062 MPASS(de->td_node == fnode); 1063 1064 /* 1065 * If re-naming a directory to another preexisting directory 1066 * ensure that the target directory is empty so that its 1067 * removal causes no side effects. 1068 * Kern_rename guarantees the destination to be a directory 1069 * if the source is one. 1070 */ 1071 if (tvp != NULL) { 1072 MPASS(tnode != NULL); 1073 1074 if ((tnode->tn_flags & (NOUNLINK | IMMUTABLE | APPEND)) || 1075 (tdnode->tn_flags & (APPEND | IMMUTABLE))) { 1076 error = EPERM; 1077 goto out_locked; 1078 } 1079 1080 if (fnode->tn_type == VDIR && tnode->tn_type == VDIR) { 1081 if (tnode->tn_size != 0 && 1082 ((tcnp->cn_flags & IGNOREWHITEOUT) == 0 || 1083 tnode->tn_size > tnode->tn_dir.tn_wht_size)) { 1084 error = ENOTEMPTY; 1085 goto out_locked; 1086 } 1087 } else if (fnode->tn_type == VDIR && tnode->tn_type != VDIR) { 1088 error = ENOTDIR; 1089 goto out_locked; 1090 } else if (fnode->tn_type != VDIR && tnode->tn_type == VDIR) { 1091 error = EISDIR; 1092 goto out_locked; 1093 } else { 1094 MPASS(fnode->tn_type != VDIR && 1095 tnode->tn_type != VDIR); 1096 } 1097 } 1098 1099 if ((fnode->tn_flags & (NOUNLINK | IMMUTABLE | APPEND)) 1100 || (fdnode->tn_flags & (APPEND | IMMUTABLE))) { 1101 error = EPERM; 1102 goto out_locked; 1103 } 1104 1105 /* 1106 * Ensure that we have enough memory to hold the new name, if it 1107 * has to be changed. 1108 */ 1109 if (fcnp->cn_namelen != tcnp->cn_namelen || 1110 bcmp(fcnp->cn_nameptr, tcnp->cn_nameptr, fcnp->cn_namelen) != 0) { 1111 newname = malloc(tcnp->cn_namelen, M_TMPFSNAME, M_WAITOK); 1112 } else 1113 newname = NULL; 1114 1115 /* 1116 * If the node is being moved to another directory, we have to do 1117 * the move. 1118 */ 1119 if (fdnode != tdnode) { 1120 /* 1121 * In case we are moving a directory, we have to adjust its 1122 * parent to point to the new parent. 1123 */ 1124 if (de->td_node->tn_type == VDIR) { 1125 struct tmpfs_node *n; 1126 1127 TMPFS_NODE_LOCK(fnode); 1128 error = tmpfs_access_locked(fvp, fnode, VWRITE, 1129 tcnp->cn_cred); 1130 TMPFS_NODE_UNLOCK(fnode); 1131 if (error) { 1132 if (newname != NULL) 1133 free(newname, M_TMPFSNAME); 1134 goto out_locked; 1135 } 1136 1137 /* 1138 * Ensure the target directory is not a child of the 1139 * directory being moved. Otherwise, we'd end up 1140 * with stale nodes. 1141 */ 1142 n = tdnode; 1143 /* 1144 * TMPFS_LOCK guaranties that no nodes are freed while 1145 * traversing the list. Nodes can only be marked as 1146 * removed: tn_parent == NULL. 1147 */ 1148 TMPFS_LOCK(tmp); 1149 TMPFS_NODE_LOCK(n); 1150 while (n != n->tn_dir.tn_parent) { 1151 struct tmpfs_node *parent; 1152 1153 if (n == fnode) { 1154 TMPFS_NODE_UNLOCK(n); 1155 TMPFS_UNLOCK(tmp); 1156 error = EINVAL; 1157 if (newname != NULL) 1158 free(newname, M_TMPFSNAME); 1159 goto out_locked; 1160 } 1161 parent = n->tn_dir.tn_parent; 1162 TMPFS_NODE_UNLOCK(n); 1163 if (parent == NULL) { 1164 n = NULL; 1165 break; 1166 } 1167 TMPFS_NODE_LOCK(parent); 1168 if (parent->tn_dir.tn_parent == NULL) { 1169 TMPFS_NODE_UNLOCK(parent); 1170 n = NULL; 1171 break; 1172 } 1173 n = parent; 1174 } 1175 TMPFS_UNLOCK(tmp); 1176 if (n == NULL) { 1177 error = EINVAL; 1178 if (newname != NULL) 1179 free(newname, M_TMPFSNAME); 1180 goto out_locked; 1181 } 1182 TMPFS_NODE_UNLOCK(n); 1183 1184 /* Adjust the parent pointer. */ 1185 TMPFS_VALIDATE_DIR(fnode); 1186 TMPFS_NODE_LOCK(de->td_node); 1187 de->td_node->tn_dir.tn_parent = tdnode; 1188 TMPFS_NODE_UNLOCK(de->td_node); 1189 1190 /* 1191 * As a result of changing the target of the '..' 1192 * entry, the link count of the source and target 1193 * directories has to be adjusted. 1194 */ 1195 TMPFS_NODE_LOCK(tdnode); 1196 TMPFS_ASSERT_LOCKED(tdnode); 1197 tdnode->tn_links++; 1198 TMPFS_NODE_UNLOCK(tdnode); 1199 1200 TMPFS_NODE_LOCK(fdnode); 1201 TMPFS_ASSERT_LOCKED(fdnode); 1202 fdnode->tn_links--; 1203 TMPFS_NODE_UNLOCK(fdnode); 1204 } 1205 } 1206 1207 /* 1208 * Do the move: just remove the entry from the source directory 1209 * and insert it into the target one. 1210 */ 1211 tmpfs_dir_detach(fdvp, de); 1212 1213 if (fcnp->cn_flags & DOWHITEOUT) 1214 tmpfs_dir_whiteout_add(fdvp, fcnp); 1215 if (tcnp->cn_flags & ISWHITEOUT) 1216 tmpfs_dir_whiteout_remove(tdvp, tcnp); 1217 1218 /* 1219 * If the name has changed, we need to make it effective by changing 1220 * it in the directory entry. 1221 */ 1222 if (newname != NULL) { 1223 MPASS(tcnp->cn_namelen <= MAXNAMLEN); 1224 1225 free(de->ud.td_name, M_TMPFSNAME); 1226 de->ud.td_name = newname; 1227 tmpfs_dirent_init(de, tcnp->cn_nameptr, tcnp->cn_namelen); 1228 1229 fnode->tn_status |= TMPFS_NODE_CHANGED; 1230 tdnode->tn_status |= TMPFS_NODE_MODIFIED; 1231 } 1232 1233 /* 1234 * If we are overwriting an entry, we have to remove the old one 1235 * from the target directory. 1236 */ 1237 if (tvp != NULL) { 1238 struct tmpfs_dirent *tde; 1239 1240 /* Remove the old entry from the target directory. */ 1241 tde = tmpfs_dir_lookup(tdnode, tnode, tcnp); 1242 tmpfs_dir_detach(tdvp, tde); 1243 1244 /* 1245 * If we are overwriting a directory, per the ENOTEMPTY check 1246 * above it must either be empty or contain only whiteout 1247 * entries. In the latter case (which can only happen if 1248 * IGNOREWHITEOUT was passed in tcnp->cn_flags), clear the 1249 * whiteout entries to avoid leaking memory. 1250 */ 1251 if (tnode->tn_type == VDIR && tnode->tn_size > 0) 1252 tmpfs_dir_clear_whiteouts(tvp); 1253 1254 /* Update node's ctime because of possible hardlinks. */ 1255 tnode->tn_status |= TMPFS_NODE_CHANGED; 1256 tmpfs_update(tvp); 1257 1258 /* 1259 * Free the directory entry we just deleted. Note that the 1260 * node referred by it will not be removed until the vnode is 1261 * really reclaimed. 1262 */ 1263 tmpfs_free_dirent(VFS_TO_TMPFS(tvp->v_mount), tde); 1264 } 1265 1266 tmpfs_dir_attach(tdvp, de); 1267 1268 if (tmpfs_use_nc(fvp)) { 1269 cache_vop_rename(fdvp, fvp, tdvp, tvp, fcnp, tcnp); 1270 } 1271 1272 error = 0; 1273 1274 out_locked: 1275 if (fdvp != tdvp && fdvp != tvp) 1276 VOP_UNLOCK(fdvp); 1277 1278 out: 1279 if (want_seqc_end) { 1280 if (tvp != NULL) 1281 vn_seqc_write_end(tvp); 1282 vn_seqc_write_end(tdvp); 1283 vn_seqc_write_end(fvp); 1284 vn_seqc_write_end(fdvp); 1285 } 1286 1287 /* 1288 * Release target nodes. 1289 * XXX: I don't understand when tdvp can be the same as tvp, but 1290 * other code takes care of this... 1291 */ 1292 if (tdvp == tvp) 1293 vrele(tdvp); 1294 else 1295 vput(tdvp); 1296 if (tvp != NULL) 1297 vput(tvp); 1298 1299 /* Release source nodes. */ 1300 vrele(fdvp); 1301 vrele(fvp); 1302 1303 return (error); 1304 } 1305 1306 static int 1307 tmpfs_mkdir(struct vop_mkdir_args *v) 1308 { 1309 struct vnode *dvp = v->a_dvp; 1310 struct vnode **vpp = v->a_vpp; 1311 struct componentname *cnp = v->a_cnp; 1312 struct vattr *vap = v->a_vap; 1313 1314 MPASS(vap->va_type == VDIR); 1315 1316 return (tmpfs_alloc_file(dvp, vpp, vap, cnp, NULL)); 1317 } 1318 1319 static int 1320 tmpfs_rmdir(struct vop_rmdir_args *v) 1321 { 1322 struct vnode *dvp = v->a_dvp; 1323 struct vnode *vp = v->a_vp; 1324 struct componentname *cnp = v->a_cnp; 1325 1326 int error; 1327 struct tmpfs_dirent *de; 1328 struct tmpfs_mount *tmp; 1329 struct tmpfs_node *dnode; 1330 struct tmpfs_node *node; 1331 1332 tmp = VFS_TO_TMPFS(dvp->v_mount); 1333 dnode = VP_TO_TMPFS_DIR(dvp); 1334 node = VP_TO_TMPFS_DIR(vp); 1335 1336 /* 1337 * Directories with more than two non-whiteout entries ('.' and '..') 1338 * cannot be removed. 1339 */ 1340 if (node->tn_size != 0 && 1341 ((cnp->cn_flags & IGNOREWHITEOUT) == 0 || 1342 node->tn_size > node->tn_dir.tn_wht_size)) { 1343 error = ENOTEMPTY; 1344 goto out; 1345 } 1346 1347 if ((dnode->tn_flags & APPEND) 1348 || (node->tn_flags & (NOUNLINK | IMMUTABLE | APPEND))) { 1349 error = EPERM; 1350 goto out; 1351 } 1352 1353 /* This invariant holds only if we are not trying to remove "..". 1354 * We checked for that above so this is safe now. */ 1355 MPASS(node->tn_dir.tn_parent == dnode); 1356 1357 /* Get the directory entry associated with node (vp). This was 1358 * filled by tmpfs_lookup while looking up the entry. */ 1359 de = tmpfs_dir_lookup(dnode, node, cnp); 1360 MPASS(TMPFS_DIRENT_MATCHES(de, 1361 cnp->cn_nameptr, 1362 cnp->cn_namelen)); 1363 1364 /* Check flags to see if we are allowed to remove the directory. */ 1365 if ((dnode->tn_flags & APPEND) != 0 || 1366 (node->tn_flags & (NOUNLINK | IMMUTABLE | APPEND)) != 0) { 1367 error = EPERM; 1368 goto out; 1369 } 1370 1371 /* Detach the directory entry from the directory (dnode). */ 1372 tmpfs_dir_detach(dvp, de); 1373 1374 /* 1375 * If we are removing a directory, per the ENOTEMPTY check above it 1376 * must either be empty or contain only whiteout entries. In the 1377 * latter case (which can only happen if IGNOREWHITEOUT was passed 1378 * in cnp->cn_flags), clear the whiteout entries to avoid leaking 1379 * memory. 1380 */ 1381 if (node->tn_size > 0) 1382 tmpfs_dir_clear_whiteouts(vp); 1383 1384 if (cnp->cn_flags & DOWHITEOUT) 1385 tmpfs_dir_whiteout_add(dvp, cnp); 1386 1387 /* No vnode should be allocated for this entry from this point */ 1388 TMPFS_NODE_LOCK(node); 1389 node->tn_links--; 1390 node->tn_dir.tn_parent = NULL; 1391 node->tn_status |= TMPFS_NODE_CHANGED | TMPFS_NODE_MODIFIED; 1392 node->tn_accessed = true; 1393 1394 TMPFS_NODE_UNLOCK(node); 1395 1396 TMPFS_NODE_LOCK(dnode); 1397 dnode->tn_links--; 1398 dnode->tn_status |= TMPFS_NODE_CHANGED | TMPFS_NODE_MODIFIED; 1399 dnode->tn_accessed = true; 1400 TMPFS_NODE_UNLOCK(dnode); 1401 1402 if (tmpfs_use_nc(dvp)) { 1403 cache_vop_rmdir(dvp, vp); 1404 } 1405 1406 /* Free the directory entry we just deleted. Note that the node 1407 * referred by it will not be removed until the vnode is really 1408 * reclaimed. */ 1409 tmpfs_free_dirent(tmp, de); 1410 1411 /* Release the deleted vnode (will destroy the node, notify 1412 * interested parties and clean it from the cache). */ 1413 1414 dnode->tn_status |= TMPFS_NODE_CHANGED; 1415 tmpfs_update(dvp); 1416 1417 error = 0; 1418 1419 out: 1420 return (error); 1421 } 1422 1423 static int 1424 tmpfs_symlink(struct vop_symlink_args *v) 1425 { 1426 struct vnode *dvp = v->a_dvp; 1427 struct vnode **vpp = v->a_vpp; 1428 struct componentname *cnp = v->a_cnp; 1429 struct vattr *vap = v->a_vap; 1430 const char *target = v->a_target; 1431 1432 #ifdef notyet /* XXX FreeBSD BUG: kern_symlink is not setting VLNK */ 1433 MPASS(vap->va_type == VLNK); 1434 #else 1435 vap->va_type = VLNK; 1436 #endif 1437 1438 return (tmpfs_alloc_file(dvp, vpp, vap, cnp, target)); 1439 } 1440 1441 static int 1442 tmpfs_readdir(struct vop_readdir_args *va) 1443 { 1444 struct vnode *vp; 1445 struct uio *uio; 1446 struct tmpfs_mount *tm; 1447 struct tmpfs_node *node; 1448 uint64_t **cookies; 1449 int *eofflag, *ncookies; 1450 ssize_t startresid; 1451 int error, maxcookies; 1452 1453 vp = va->a_vp; 1454 uio = va->a_uio; 1455 eofflag = va->a_eofflag; 1456 cookies = va->a_cookies; 1457 ncookies = va->a_ncookies; 1458 1459 /* This operation only makes sense on directory nodes. */ 1460 if (vp->v_type != VDIR) 1461 return (ENOTDIR); 1462 1463 maxcookies = 0; 1464 node = VP_TO_TMPFS_DIR(vp); 1465 tm = VFS_TO_TMPFS(vp->v_mount); 1466 1467 startresid = uio->uio_resid; 1468 1469 /* Allocate cookies for NFS and compat modules. */ 1470 if (cookies != NULL && ncookies != NULL) { 1471 maxcookies = howmany(node->tn_size, 1472 sizeof(struct tmpfs_dirent)) + 2; 1473 *cookies = malloc(maxcookies * sizeof(**cookies), M_TEMP, 1474 M_WAITOK); 1475 *ncookies = 0; 1476 } 1477 1478 if (cookies == NULL) 1479 error = tmpfs_dir_getdents(tm, node, uio, 0, NULL, NULL); 1480 else 1481 error = tmpfs_dir_getdents(tm, node, uio, maxcookies, *cookies, 1482 ncookies); 1483 1484 /* Buffer was filled without hitting EOF. */ 1485 if (error == EJUSTRETURN) 1486 error = (uio->uio_resid != startresid) ? 0 : EINVAL; 1487 1488 if (error != 0 && cookies != NULL && ncookies != NULL) { 1489 free(*cookies, M_TEMP); 1490 *cookies = NULL; 1491 *ncookies = 0; 1492 } 1493 1494 if (eofflag != NULL) 1495 *eofflag = 1496 (error == 0 && uio->uio_offset == TMPFS_DIRCOOKIE_EOF); 1497 1498 return (error); 1499 } 1500 1501 static int 1502 tmpfs_readlink(struct vop_readlink_args *v) 1503 { 1504 struct vnode *vp = v->a_vp; 1505 struct uio *uio = v->a_uio; 1506 1507 int error; 1508 struct tmpfs_node *node; 1509 1510 MPASS(uio->uio_offset == 0); 1511 MPASS(vp->v_type == VLNK); 1512 1513 node = VP_TO_TMPFS_NODE(vp); 1514 1515 error = uiomove(node->tn_link_target, MIN(node->tn_size, uio->uio_resid), 1516 uio); 1517 tmpfs_set_accessed(VFS_TO_TMPFS(vp->v_mount), node); 1518 1519 return (error); 1520 } 1521 1522 /* 1523 * VOP_FPLOOKUP_SYMLINK routines are subject to special circumstances, see 1524 * the comment above cache_fplookup for details. 1525 * 1526 * Check tmpfs_alloc_node for tmpfs-specific synchronisation notes. 1527 */ 1528 static int 1529 tmpfs_fplookup_symlink(struct vop_fplookup_symlink_args *v) 1530 { 1531 struct vnode *vp; 1532 struct tmpfs_node *node; 1533 char *symlink; 1534 1535 vp = v->a_vp; 1536 node = VP_TO_TMPFS_NODE_SMR(vp); 1537 if (__predict_false(node == NULL)) 1538 return (EAGAIN); 1539 if (!atomic_load_char(&node->tn_link_smr)) 1540 return (EAGAIN); 1541 symlink = atomic_load_ptr(&node->tn_link_target); 1542 if (symlink == NULL) 1543 return (EAGAIN); 1544 1545 return (cache_symlink_resolve(v->a_fpl, symlink, node->tn_size)); 1546 } 1547 1548 static int 1549 tmpfs_inactive(struct vop_inactive_args *v) 1550 { 1551 struct vnode *vp; 1552 struct tmpfs_node *node; 1553 1554 vp = v->a_vp; 1555 node = VP_TO_TMPFS_NODE(vp); 1556 if (node->tn_links == 0) 1557 vrecycle(vp); 1558 else 1559 tmpfs_check_mtime(vp); 1560 return (0); 1561 } 1562 1563 static int 1564 tmpfs_need_inactive(struct vop_need_inactive_args *ap) 1565 { 1566 struct vnode *vp; 1567 struct tmpfs_node *node; 1568 struct vm_object *obj; 1569 1570 vp = ap->a_vp; 1571 node = VP_TO_TMPFS_NODE(vp); 1572 if (node->tn_links == 0) 1573 goto need; 1574 if (vp->v_type == VREG) { 1575 obj = vp->v_object; 1576 if (obj->generation != obj->cleangeneration) 1577 goto need; 1578 } 1579 return (0); 1580 need: 1581 return (1); 1582 } 1583 1584 int 1585 tmpfs_reclaim(struct vop_reclaim_args *v) 1586 { 1587 struct vnode *vp; 1588 struct tmpfs_mount *tmp; 1589 struct tmpfs_node *node; 1590 bool unlock; 1591 1592 vp = v->a_vp; 1593 node = VP_TO_TMPFS_NODE(vp); 1594 tmp = VFS_TO_TMPFS(vp->v_mount); 1595 1596 if (vp->v_type == VREG) 1597 tmpfs_destroy_vobject(vp, node->tn_reg.tn_aobj); 1598 vp->v_object = NULL; 1599 1600 TMPFS_LOCK(tmp); 1601 TMPFS_NODE_LOCK(node); 1602 tmpfs_free_vp(vp); 1603 1604 /* 1605 * If the node referenced by this vnode was deleted by the user, 1606 * we must free its associated data structures (now that the vnode 1607 * is being reclaimed). 1608 */ 1609 unlock = true; 1610 if (node->tn_links == 0 && 1611 (node->tn_vpstate & TMPFS_VNODE_ALLOCATING) == 0) { 1612 node->tn_vpstate = TMPFS_VNODE_DOOMED; 1613 unlock = !tmpfs_free_node_locked(tmp, node, true); 1614 } 1615 1616 if (unlock) { 1617 TMPFS_NODE_UNLOCK(node); 1618 TMPFS_UNLOCK(tmp); 1619 } 1620 1621 MPASS(vp->v_data == NULL); 1622 return (0); 1623 } 1624 1625 int 1626 tmpfs_print(struct vop_print_args *v) 1627 { 1628 struct vnode *vp = v->a_vp; 1629 1630 struct tmpfs_node *node; 1631 1632 node = VP_TO_TMPFS_NODE(vp); 1633 1634 printf("tag VT_TMPFS, tmpfs_node %p, flags 0x%lx, links %jd\n", 1635 node, node->tn_flags, (uintmax_t)node->tn_links); 1636 printf("\tmode 0%o, owner %d, group %d, size %jd, status 0x%x\n", 1637 node->tn_mode, node->tn_uid, node->tn_gid, 1638 (intmax_t)node->tn_size, node->tn_status); 1639 1640 if (vp->v_type == VFIFO) 1641 fifo_printinfo(vp); 1642 1643 printf("\n"); 1644 1645 return (0); 1646 } 1647 1648 int 1649 tmpfs_pathconf(struct vop_pathconf_args *v) 1650 { 1651 struct vnode *vp = v->a_vp; 1652 int name = v->a_name; 1653 long *retval = v->a_retval; 1654 1655 int error; 1656 1657 error = 0; 1658 1659 switch (name) { 1660 case _PC_LINK_MAX: 1661 *retval = TMPFS_LINK_MAX; 1662 break; 1663 1664 case _PC_SYMLINK_MAX: 1665 *retval = MAXPATHLEN; 1666 break; 1667 1668 case _PC_NAME_MAX: 1669 *retval = NAME_MAX; 1670 break; 1671 1672 case _PC_PIPE_BUF: 1673 if (vp->v_type == VDIR || vp->v_type == VFIFO) 1674 *retval = PIPE_BUF; 1675 else 1676 error = EINVAL; 1677 break; 1678 1679 case _PC_CHOWN_RESTRICTED: 1680 *retval = 1; 1681 break; 1682 1683 case _PC_NO_TRUNC: 1684 *retval = 1; 1685 break; 1686 1687 case _PC_SYNC_IO: 1688 *retval = 1; 1689 break; 1690 1691 case _PC_FILESIZEBITS: 1692 *retval = 64; 1693 break; 1694 1695 case _PC_MIN_HOLE_SIZE: 1696 *retval = PAGE_SIZE; 1697 break; 1698 1699 default: 1700 error = vop_stdpathconf(v); 1701 } 1702 1703 return (error); 1704 } 1705 1706 static int 1707 tmpfs_vptofh(struct vop_vptofh_args *ap) 1708 /* 1709 vop_vptofh { 1710 IN struct vnode *a_vp; 1711 IN struct fid *a_fhp; 1712 }; 1713 */ 1714 { 1715 struct tmpfs_fid_data tfd; 1716 struct tmpfs_node *node; 1717 struct fid *fhp; 1718 1719 node = VP_TO_TMPFS_NODE(ap->a_vp); 1720 fhp = ap->a_fhp; 1721 fhp->fid_len = sizeof(tfd); 1722 1723 /* 1724 * Copy into fid_data from the stack to avoid unaligned pointer use. 1725 * See the comment in sys/mount.h on struct fid for details. 1726 */ 1727 tfd.tfd_id = node->tn_id; 1728 tfd.tfd_gen = node->tn_gen; 1729 memcpy(fhp->fid_data, &tfd, fhp->fid_len); 1730 1731 return (0); 1732 } 1733 1734 static int 1735 tmpfs_whiteout(struct vop_whiteout_args *ap) 1736 { 1737 struct vnode *dvp = ap->a_dvp; 1738 struct componentname *cnp = ap->a_cnp; 1739 struct tmpfs_dirent *de; 1740 1741 switch (ap->a_flags) { 1742 case LOOKUP: 1743 return (0); 1744 case CREATE: 1745 de = tmpfs_dir_lookup(VP_TO_TMPFS_DIR(dvp), NULL, cnp); 1746 if (de != NULL) 1747 return (de->td_node == NULL ? 0 : EEXIST); 1748 return (tmpfs_dir_whiteout_add(dvp, cnp)); 1749 case DELETE: 1750 tmpfs_dir_whiteout_remove(dvp, cnp); 1751 return (0); 1752 default: 1753 panic("tmpfs_whiteout: unknown op"); 1754 } 1755 } 1756 1757 static int 1758 tmpfs_vptocnp_dir(struct tmpfs_node *tn, struct tmpfs_node *tnp, 1759 struct tmpfs_dirent **pde) 1760 { 1761 struct tmpfs_dir_cursor dc; 1762 struct tmpfs_dirent *de; 1763 1764 for (de = tmpfs_dir_first(tnp, &dc); de != NULL; 1765 de = tmpfs_dir_next(tnp, &dc)) { 1766 if (de->td_node == tn) { 1767 *pde = de; 1768 return (0); 1769 } 1770 } 1771 return (ENOENT); 1772 } 1773 1774 static int 1775 tmpfs_vptocnp_fill(struct vnode *vp, struct tmpfs_node *tn, 1776 struct tmpfs_node *tnp, char *buf, size_t *buflen, struct vnode **dvp) 1777 { 1778 struct tmpfs_dirent *de; 1779 int error, i; 1780 1781 error = vn_vget_ino_gen(vp, tmpfs_vn_get_ino_alloc, tnp, LK_SHARED, 1782 dvp); 1783 if (error != 0) 1784 return (error); 1785 error = tmpfs_vptocnp_dir(tn, tnp, &de); 1786 if (error == 0) { 1787 i = *buflen; 1788 i -= de->td_namelen; 1789 if (i < 0) { 1790 error = ENOMEM; 1791 } else { 1792 bcopy(de->ud.td_name, buf + i, de->td_namelen); 1793 *buflen = i; 1794 } 1795 } 1796 if (error == 0) { 1797 if (vp != *dvp) 1798 VOP_UNLOCK(*dvp); 1799 } else { 1800 if (vp != *dvp) 1801 vput(*dvp); 1802 else 1803 vrele(vp); 1804 } 1805 return (error); 1806 } 1807 1808 static int 1809 tmpfs_vptocnp(struct vop_vptocnp_args *ap) 1810 { 1811 struct vnode *vp, **dvp; 1812 struct tmpfs_node *tn, *tnp, *tnp1; 1813 struct tmpfs_dirent *de; 1814 struct tmpfs_mount *tm; 1815 char *buf; 1816 size_t *buflen; 1817 int error; 1818 1819 vp = ap->a_vp; 1820 dvp = ap->a_vpp; 1821 buf = ap->a_buf; 1822 buflen = ap->a_buflen; 1823 1824 tm = VFS_TO_TMPFS(vp->v_mount); 1825 tn = VP_TO_TMPFS_NODE(vp); 1826 if (tn->tn_type == VDIR) { 1827 tnp = tn->tn_dir.tn_parent; 1828 if (tnp == NULL) 1829 return (ENOENT); 1830 tmpfs_ref_node(tnp); 1831 error = tmpfs_vptocnp_fill(vp, tn, tn->tn_dir.tn_parent, buf, 1832 buflen, dvp); 1833 tmpfs_free_node(tm, tnp); 1834 return (error); 1835 } 1836 restart: 1837 TMPFS_LOCK(tm); 1838 restart_locked: 1839 LIST_FOREACH_SAFE(tnp, &tm->tm_nodes_used, tn_entries, tnp1) { 1840 if (tnp->tn_type != VDIR) 1841 continue; 1842 TMPFS_NODE_LOCK(tnp); 1843 tmpfs_ref_node(tnp); 1844 1845 /* 1846 * tn_vnode cannot be instantiated while we hold the 1847 * node lock, so the directory cannot be changed while 1848 * we iterate over it. Do this to avoid instantiating 1849 * vnode for directories which cannot point to our 1850 * node. 1851 */ 1852 error = tnp->tn_vnode == NULL ? tmpfs_vptocnp_dir(tn, tnp, 1853 &de) : 0; 1854 1855 if (error == 0) { 1856 TMPFS_NODE_UNLOCK(tnp); 1857 TMPFS_UNLOCK(tm); 1858 error = tmpfs_vptocnp_fill(vp, tn, tnp, buf, buflen, 1859 dvp); 1860 if (error == 0) { 1861 tmpfs_free_node(tm, tnp); 1862 return (0); 1863 } 1864 if (VN_IS_DOOMED(vp)) { 1865 tmpfs_free_node(tm, tnp); 1866 return (ENOENT); 1867 } 1868 TMPFS_LOCK(tm); 1869 TMPFS_NODE_LOCK(tnp); 1870 } 1871 if (tmpfs_free_node_locked(tm, tnp, false)) { 1872 goto restart; 1873 } else { 1874 KASSERT(tnp->tn_refcount > 0, 1875 ("node %p refcount zero", tnp)); 1876 if (tnp->tn_attached) { 1877 tnp1 = LIST_NEXT(tnp, tn_entries); 1878 TMPFS_NODE_UNLOCK(tnp); 1879 } else { 1880 TMPFS_NODE_UNLOCK(tnp); 1881 goto restart_locked; 1882 } 1883 } 1884 } 1885 TMPFS_UNLOCK(tm); 1886 return (ENOENT); 1887 } 1888 1889 void 1890 tmpfs_extattr_free(struct tmpfs_extattr *ea) 1891 { 1892 free(ea->ea_name, M_TMPFSEA); 1893 free(ea->ea_value, M_TMPFSEA); 1894 free(ea, M_TMPFSEA); 1895 } 1896 1897 static bool 1898 tmpfs_extattr_update_mem(struct tmpfs_mount *tmp, ssize_t size) 1899 { 1900 TMPFS_LOCK(tmp); 1901 if (size > 0 && 1902 !tmpfs_pages_check_avail(tmp, howmany(size, PAGE_SIZE))) { 1903 TMPFS_UNLOCK(tmp); 1904 return (false); 1905 } 1906 if (tmp->tm_ea_memory_inuse + size > tmp->tm_ea_memory_max) { 1907 TMPFS_UNLOCK(tmp); 1908 return (false); 1909 } 1910 tmp->tm_ea_memory_inuse += size; 1911 TMPFS_UNLOCK(tmp); 1912 return (true); 1913 } 1914 1915 static int 1916 tmpfs_deleteextattr(struct vop_deleteextattr_args *ap) 1917 { 1918 struct vnode *vp = ap->a_vp; 1919 struct tmpfs_mount *tmp; 1920 struct tmpfs_node *node; 1921 struct tmpfs_extattr *ea; 1922 size_t namelen; 1923 ssize_t diff; 1924 int error; 1925 1926 node = VP_TO_TMPFS_NODE(vp); 1927 tmp = VFS_TO_TMPFS(vp->v_mount); 1928 if (ap->a_vp->v_type == VCHR || ap->a_vp->v_type == VBLK) 1929 return (EOPNOTSUPP); 1930 error = extattr_check_cred(ap->a_vp, ap->a_attrnamespace, 1931 ap->a_cred, ap->a_td, VWRITE); 1932 if (error != 0) 1933 return (error); 1934 if (ap->a_name == NULL || ap->a_name[0] == '\0') 1935 return (EINVAL); 1936 namelen = strlen(ap->a_name); 1937 if (namelen > EXTATTR_MAXNAMELEN) 1938 return (EINVAL); 1939 1940 LIST_FOREACH(ea, &node->tn_extattrs, ea_extattrs) { 1941 if (ea->ea_namespace == ap->a_attrnamespace && 1942 namelen == ea->ea_namelen && 1943 memcmp(ap->a_name, ea->ea_name, namelen) == 0) 1944 break; 1945 } 1946 1947 if (ea == NULL) 1948 return (ENOATTR); 1949 LIST_REMOVE(ea, ea_extattrs); 1950 diff = -(sizeof(struct tmpfs_extattr) + namelen + ea->ea_size); 1951 tmpfs_extattr_update_mem(tmp, diff); 1952 tmpfs_extattr_free(ea); 1953 return (0); 1954 } 1955 1956 static int 1957 tmpfs_getextattr(struct vop_getextattr_args *ap) 1958 { 1959 struct vnode *vp = ap->a_vp; 1960 struct tmpfs_node *node; 1961 struct tmpfs_extattr *ea; 1962 size_t namelen; 1963 int error; 1964 1965 node = VP_TO_TMPFS_NODE(vp); 1966 if (ap->a_vp->v_type == VCHR || ap->a_vp->v_type == VBLK) 1967 return (EOPNOTSUPP); 1968 error = extattr_check_cred(ap->a_vp, ap->a_attrnamespace, 1969 ap->a_cred, ap->a_td, VREAD); 1970 if (error != 0) 1971 return (error); 1972 if (ap->a_name == NULL || ap->a_name[0] == '\0') 1973 return (EINVAL); 1974 namelen = strlen(ap->a_name); 1975 if (namelen > EXTATTR_MAXNAMELEN) 1976 return (EINVAL); 1977 1978 LIST_FOREACH(ea, &node->tn_extattrs, ea_extattrs) { 1979 if (ea->ea_namespace == ap->a_attrnamespace && 1980 namelen == ea->ea_namelen && 1981 memcmp(ap->a_name, ea->ea_name, namelen) == 0) 1982 break; 1983 } 1984 1985 if (ea == NULL) 1986 return (ENOATTR); 1987 if (ap->a_size != NULL) 1988 *ap->a_size = ea->ea_size; 1989 if (ap->a_uio != NULL && ea->ea_size != 0) 1990 error = uiomove(ea->ea_value, ea->ea_size, ap->a_uio); 1991 return (error); 1992 } 1993 1994 static int 1995 tmpfs_listextattr(struct vop_listextattr_args *ap) 1996 { 1997 struct vnode *vp = ap->a_vp; 1998 struct tmpfs_node *node; 1999 struct tmpfs_extattr *ea; 2000 int error; 2001 2002 node = VP_TO_TMPFS_NODE(vp); 2003 if (ap->a_vp->v_type == VCHR || ap->a_vp->v_type == VBLK) 2004 return (EOPNOTSUPP); 2005 error = extattr_check_cred(ap->a_vp, ap->a_attrnamespace, 2006 ap->a_cred, ap->a_td, VREAD); 2007 if (error != 0) 2008 return (error); 2009 if (ap->a_size != NULL) 2010 *ap->a_size = 0; 2011 2012 LIST_FOREACH(ea, &node->tn_extattrs, ea_extattrs) { 2013 if (ea->ea_namespace != ap->a_attrnamespace) 2014 continue; 2015 if (ap->a_size != NULL) 2016 *ap->a_size += ea->ea_namelen + 1; 2017 if (ap->a_uio != NULL) { 2018 error = uiomove(&ea->ea_namelen, 1, ap->a_uio); 2019 if (error != 0) 2020 break; 2021 error = uiomove(ea->ea_name, ea->ea_namelen, ap->a_uio); 2022 if (error != 0) 2023 break; 2024 } 2025 } 2026 2027 return (error); 2028 } 2029 2030 static int 2031 tmpfs_setextattr(struct vop_setextattr_args *ap) 2032 { 2033 struct vnode *vp = ap->a_vp; 2034 struct tmpfs_mount *tmp; 2035 struct tmpfs_node *node; 2036 struct tmpfs_extattr *ea; 2037 struct tmpfs_extattr *new_ea; 2038 size_t attr_size; 2039 size_t namelen; 2040 ssize_t diff; 2041 int error; 2042 2043 node = VP_TO_TMPFS_NODE(vp); 2044 tmp = VFS_TO_TMPFS(vp->v_mount); 2045 attr_size = ap->a_uio->uio_resid; 2046 diff = 0; 2047 if (ap->a_vp->v_type == VCHR || ap->a_vp->v_type == VBLK) 2048 return (EOPNOTSUPP); 2049 error = extattr_check_cred(ap->a_vp, ap->a_attrnamespace, 2050 ap->a_cred, ap->a_td, VWRITE); 2051 if (error != 0) 2052 return (error); 2053 if (ap->a_name == NULL || ap->a_name[0] == '\0') 2054 return (EINVAL); 2055 namelen = strlen(ap->a_name); 2056 if (namelen > EXTATTR_MAXNAMELEN) 2057 return (EINVAL); 2058 2059 LIST_FOREACH(ea, &node->tn_extattrs, ea_extattrs) { 2060 if (ea->ea_namespace == ap->a_attrnamespace && 2061 namelen == ea->ea_namelen && 2062 memcmp(ap->a_name, ea->ea_name, namelen) == 0) { 2063 diff -= sizeof(struct tmpfs_extattr) + ea->ea_namelen + 2064 ea->ea_size; 2065 break; 2066 } 2067 } 2068 2069 diff += sizeof(struct tmpfs_extattr) + namelen + attr_size; 2070 if (!tmpfs_extattr_update_mem(tmp, diff)) 2071 return (ENOSPC); 2072 new_ea = malloc(sizeof(struct tmpfs_extattr), M_TMPFSEA, M_WAITOK); 2073 new_ea->ea_namespace = ap->a_attrnamespace; 2074 new_ea->ea_name = malloc(namelen, M_TMPFSEA, M_WAITOK); 2075 new_ea->ea_namelen = namelen; 2076 memcpy(new_ea->ea_name, ap->a_name, namelen); 2077 if (attr_size != 0) { 2078 new_ea->ea_value = malloc(attr_size, M_TMPFSEA, M_WAITOK); 2079 new_ea->ea_size = attr_size; 2080 error = uiomove(new_ea->ea_value, attr_size, ap->a_uio); 2081 } else { 2082 new_ea->ea_value = NULL; 2083 new_ea->ea_size = 0; 2084 } 2085 if (error != 0) { 2086 tmpfs_extattr_update_mem(tmp, -diff); 2087 tmpfs_extattr_free(new_ea); 2088 return (error); 2089 } 2090 if (ea != NULL) { 2091 LIST_REMOVE(ea, ea_extattrs); 2092 tmpfs_extattr_free(ea); 2093 } 2094 LIST_INSERT_HEAD(&node->tn_extattrs, new_ea, ea_extattrs); 2095 return (0); 2096 } 2097 2098 static off_t 2099 tmpfs_seek_data_locked(vm_object_t obj, off_t noff) 2100 { 2101 vm_page_t m; 2102 vm_pindex_t p, p_m, p_swp; 2103 2104 p = OFF_TO_IDX(noff); 2105 m = vm_page_find_least(obj, p); 2106 2107 /* 2108 * Microoptimize the most common case for SEEK_DATA, where 2109 * there is no hole and the page is resident. 2110 */ 2111 if (m != NULL && vm_page_any_valid(m) && m->pindex == p) 2112 return (noff); 2113 2114 p_swp = swap_pager_find_least(obj, p); 2115 if (p_swp == p) 2116 return (noff); 2117 2118 p_m = m == NULL ? obj->size : m->pindex; 2119 return (IDX_TO_OFF(MIN(p_m, p_swp))); 2120 } 2121 2122 static off_t 2123 tmpfs_seek_next(off_t noff) 2124 { 2125 return (noff + PAGE_SIZE - (noff & PAGE_MASK)); 2126 } 2127 2128 static int 2129 tmpfs_seek_clamp(struct tmpfs_node *tn, off_t *noff, bool seekdata) 2130 { 2131 if (*noff < tn->tn_size) 2132 return (0); 2133 if (seekdata) 2134 return (ENXIO); 2135 *noff = tn->tn_size; 2136 return (0); 2137 } 2138 2139 static off_t 2140 tmpfs_seek_hole_locked(vm_object_t obj, off_t noff) 2141 { 2142 vm_page_t m; 2143 vm_pindex_t p, p_swp; 2144 2145 for (;; noff = tmpfs_seek_next(noff)) { 2146 /* 2147 * Walk over the largest sequential run of the valid pages. 2148 */ 2149 for (m = vm_page_lookup(obj, OFF_TO_IDX(noff)); 2150 m != NULL && vm_page_any_valid(m); 2151 m = vm_page_next(m), noff = tmpfs_seek_next(noff)) 2152 ; 2153 2154 /* 2155 * Found a hole in the object's page queue. Check if 2156 * there is a hole in the swap at the same place. 2157 */ 2158 p = OFF_TO_IDX(noff); 2159 p_swp = swap_pager_find_least(obj, p); 2160 if (p_swp != p) { 2161 noff = IDX_TO_OFF(p); 2162 break; 2163 } 2164 } 2165 return (noff); 2166 } 2167 2168 static int 2169 tmpfs_seek_datahole(struct vnode *vp, off_t *off, bool seekdata) 2170 { 2171 struct tmpfs_node *tn; 2172 vm_object_t obj; 2173 off_t noff; 2174 int error; 2175 2176 if (vp->v_type != VREG) 2177 return (ENOTTY); 2178 tn = VP_TO_TMPFS_NODE(vp); 2179 noff = *off; 2180 if (noff < 0) 2181 return (ENXIO); 2182 error = tmpfs_seek_clamp(tn, &noff, seekdata); 2183 if (error != 0) 2184 return (error); 2185 obj = tn->tn_reg.tn_aobj; 2186 2187 VM_OBJECT_RLOCK(obj); 2188 noff = seekdata ? tmpfs_seek_data_locked(obj, noff) : 2189 tmpfs_seek_hole_locked(obj, noff); 2190 VM_OBJECT_RUNLOCK(obj); 2191 2192 error = tmpfs_seek_clamp(tn, &noff, seekdata); 2193 if (error == 0) 2194 *off = noff; 2195 return (error); 2196 } 2197 2198 static int 2199 tmpfs_ioctl(struct vop_ioctl_args *ap) 2200 { 2201 struct vnode *vp = ap->a_vp; 2202 int error = 0; 2203 2204 switch (ap->a_command) { 2205 case FIOSEEKDATA: 2206 case FIOSEEKHOLE: 2207 error = vn_lock(vp, LK_SHARED); 2208 if (error != 0) { 2209 error = EBADF; 2210 break; 2211 } 2212 error = tmpfs_seek_datahole(vp, (off_t *)ap->a_data, 2213 ap->a_command == FIOSEEKDATA); 2214 VOP_UNLOCK(vp); 2215 break; 2216 default: 2217 error = ENOTTY; 2218 break; 2219 } 2220 return (error); 2221 } 2222 2223 /* 2224 * Vnode operations vector used for files stored in a tmpfs file system. 2225 */ 2226 struct vop_vector tmpfs_vnodeop_entries = { 2227 .vop_default = &default_vnodeops, 2228 .vop_lookup = vfs_cache_lookup, 2229 .vop_cachedlookup = tmpfs_cached_lookup, 2230 .vop_create = tmpfs_create, 2231 .vop_mknod = tmpfs_mknod, 2232 .vop_open = tmpfs_open, 2233 .vop_close = tmpfs_close, 2234 .vop_fplookup_vexec = tmpfs_fplookup_vexec, 2235 .vop_fplookup_symlink = tmpfs_fplookup_symlink, 2236 .vop_access = tmpfs_access, 2237 .vop_stat = tmpfs_stat, 2238 .vop_getattr = tmpfs_getattr, 2239 .vop_setattr = tmpfs_setattr, 2240 .vop_read = tmpfs_read, 2241 .vop_read_pgcache = tmpfs_read_pgcache, 2242 .vop_write = tmpfs_write, 2243 .vop_deallocate = tmpfs_deallocate, 2244 .vop_fsync = tmpfs_fsync, 2245 .vop_remove = tmpfs_remove, 2246 .vop_link = tmpfs_link, 2247 .vop_rename = tmpfs_rename, 2248 .vop_mkdir = tmpfs_mkdir, 2249 .vop_rmdir = tmpfs_rmdir, 2250 .vop_symlink = tmpfs_symlink, 2251 .vop_readdir = tmpfs_readdir, 2252 .vop_readlink = tmpfs_readlink, 2253 .vop_inactive = tmpfs_inactive, 2254 .vop_need_inactive = tmpfs_need_inactive, 2255 .vop_reclaim = tmpfs_reclaim, 2256 .vop_print = tmpfs_print, 2257 .vop_pathconf = tmpfs_pathconf, 2258 .vop_vptofh = tmpfs_vptofh, 2259 .vop_whiteout = tmpfs_whiteout, 2260 .vop_bmap = VOP_EOPNOTSUPP, 2261 .vop_vptocnp = tmpfs_vptocnp, 2262 .vop_lock1 = vop_lock, 2263 .vop_unlock = vop_unlock, 2264 .vop_islocked = vop_islocked, 2265 .vop_deleteextattr = tmpfs_deleteextattr, 2266 .vop_getextattr = tmpfs_getextattr, 2267 .vop_listextattr = tmpfs_listextattr, 2268 .vop_setextattr = tmpfs_setextattr, 2269 .vop_add_writecount = vop_stdadd_writecount_nomsync, 2270 .vop_ioctl = tmpfs_ioctl, 2271 }; 2272 VFS_VOP_VECTOR_REGISTER(tmpfs_vnodeop_entries); 2273 2274 /* 2275 * Same vector for mounts which do not use namecache. 2276 */ 2277 struct vop_vector tmpfs_vnodeop_nonc_entries = { 2278 .vop_default = &tmpfs_vnodeop_entries, 2279 .vop_lookup = tmpfs_lookup, 2280 }; 2281 VFS_VOP_VECTOR_REGISTER(tmpfs_vnodeop_nonc_entries); 2282