1 /* $NetBSD: vfs_subr.c,v 1.419 2011/02/17 19:04:03 matt Exp $ */ 2 3 /*- 4 * Copyright (c) 1997, 1998, 2004, 2005, 2007, 2008 The NetBSD Foundation, Inc. 5 * All rights reserved. 6 * 7 * This code is derived from software contributed to The NetBSD Foundation 8 * by Jason R. Thorpe of the Numerical Aerospace Simulation Facility, 9 * NASA Ames Research Center, by Charles M. Hannum, and by Andrew Doran. 10 * 11 * Redistribution and use in source and binary forms, with or without 12 * modification, are permitted provided that the following conditions 13 * are met: 14 * 1. Redistributions of source code must retain the above copyright 15 * notice, this list of conditions and the following disclaimer. 16 * 2. Redistributions in binary form must reproduce the above copyright 17 * notice, this list of conditions and the following disclaimer in the 18 * documentation and/or other materials provided with the distribution. 19 * 20 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 21 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 22 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 23 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 24 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 25 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 26 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 27 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 28 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 29 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 30 * POSSIBILITY OF SUCH DAMAGE. 31 */ 32 33 /* 34 * Copyright (c) 1989, 1993 35 * The Regents of the University of California. All rights reserved. 36 * (c) UNIX System Laboratories, Inc. 37 * All or some portions of this file are derived from material licensed 38 * to the University of California by American Telephone and Telegraph 39 * Co. or Unix System Laboratories, Inc. and are reproduced herein with 40 * the permission of UNIX System Laboratories, Inc. 41 * 42 * Redistribution and use in source and binary forms, with or without 43 * modification, are permitted provided that the following conditions 44 * are met: 45 * 1. Redistributions of source code must retain the above copyright 46 * notice, this list of conditions and the following disclaimer. 47 * 2. Redistributions in binary form must reproduce the above copyright 48 * notice, this list of conditions and the following disclaimer in the 49 * documentation and/or other materials provided with the distribution. 50 * 3. Neither the name of the University nor the names of its contributors 51 * may be used to endorse or promote products derived from this software 52 * without specific prior written permission. 53 * 54 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 55 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 56 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 57 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 58 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 59 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 60 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 61 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 62 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 63 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 64 * SUCH DAMAGE. 65 * 66 * @(#)vfs_subr.c 8.13 (Berkeley) 4/18/94 67 */ 68 69 /* 70 * Note on v_usecount and locking: 71 * 72 * At nearly all points it is known that v_usecount could be zero, the 73 * vnode interlock will be held. 74 * 75 * To change v_usecount away from zero, the interlock must be held. To 76 * change from a non-zero value to zero, again the interlock must be 77 * held. 78 * 79 * There's a flag bit, VC_XLOCK, embedded in v_usecount. 80 * To raise v_usecount, if the VC_XLOCK bit is set in it, the interlock 81 * must be held. 82 * To modify the VC_XLOCK bit, the interlock must be held. 83 * We always keep the usecount (v_usecount & VC_MASK) non-zero while the 84 * VC_XLOCK bit is set. 85 * 86 * Unless the VC_XLOCK bit is set, changing the usecount from a non-zero 87 * value to a non-zero value can safely be done using atomic operations, 88 * without the interlock held. 89 * Even if the VC_XLOCK bit is set, decreasing the usecount to a non-zero 90 * value can be done using atomic operations, without the interlock held. 91 */ 92 93 #include <sys/cdefs.h> 94 __KERNEL_RCSID(0, "$NetBSD: vfs_subr.c,v 1.419 2011/02/17 19:04:03 matt Exp $"); 95 96 #include "opt_ddb.h" 97 #include "opt_compat_netbsd.h" 98 #include "opt_compat_43.h" 99 100 #include <sys/param.h> 101 #include <sys/systm.h> 102 #include <sys/conf.h> 103 #include <sys/dirent.h> 104 #include <sys/proc.h> 105 #include <sys/kernel.h> 106 #include <sys/mount.h> 107 #include <sys/fcntl.h> 108 #include <sys/vnode.h> 109 #include <sys/stat.h> 110 #include <sys/namei.h> 111 #include <sys/ucred.h> 112 #include <sys/buf.h> 113 #include <sys/errno.h> 114 #include <sys/kmem.h> 115 #include <sys/syscallargs.h> 116 #include <sys/device.h> 117 #include <sys/filedesc.h> 118 #include <sys/kauth.h> 119 #include <sys/atomic.h> 120 #include <sys/kthread.h> 121 #include <sys/wapbl.h> 122 #include <sys/module.h> 123 124 #include <miscfs/genfs/genfs.h> 125 #include <miscfs/specfs/specdev.h> 126 #include <miscfs/syncfs/syncfs.h> 127 128 #include <uvm/uvm.h> /* extern struct uvm uvm */ 129 #include <uvm/uvm_readahead.h> 130 #include <uvm/uvm_ddb.h> 131 132 #include <sys/sysctl.h> 133 134 const enum vtype iftovt_tab[16] = { 135 VNON, VFIFO, VCHR, VNON, VDIR, VNON, VBLK, VNON, 136 VREG, VNON, VLNK, VNON, VSOCK, VNON, VNON, VBAD, 137 }; 138 const int vttoif_tab[9] = { 139 0, S_IFREG, S_IFDIR, S_IFBLK, S_IFCHR, S_IFLNK, 140 S_IFSOCK, S_IFIFO, S_IFMT, 141 }; 142 143 /* 144 * Insq/Remq for the vnode usage lists. 145 */ 146 #define bufinsvn(bp, dp) LIST_INSERT_HEAD(dp, bp, b_vnbufs) 147 #define bufremvn(bp) { \ 148 LIST_REMOVE(bp, b_vnbufs); \ 149 (bp)->b_vnbufs.le_next = NOLIST; \ 150 } 151 152 int doforce = 1; /* 1 => permit forcible unmounting */ 153 int prtactive = 0; /* 1 => print out reclaim of active vnodes */ 154 155 static vnodelst_t vnode_free_list = TAILQ_HEAD_INITIALIZER(vnode_free_list); 156 static vnodelst_t vnode_hold_list = TAILQ_HEAD_INITIALIZER(vnode_hold_list); 157 static vnodelst_t vrele_list = TAILQ_HEAD_INITIALIZER(vrele_list); 158 159 struct mntlist mountlist = /* mounted filesystem list */ 160 CIRCLEQ_HEAD_INITIALIZER(mountlist); 161 162 u_int numvnodes; 163 static specificdata_domain_t mount_specificdata_domain; 164 165 static int vrele_pending; 166 static int vrele_gen; 167 static kmutex_t vrele_lock; 168 static kcondvar_t vrele_cv; 169 static lwp_t *vrele_lwp; 170 171 static uint64_t mountgen = 0; 172 static kmutex_t mountgen_lock; 173 174 kmutex_t mountlist_lock; 175 kmutex_t mntid_lock; 176 kmutex_t mntvnode_lock; 177 kmutex_t vnode_free_list_lock; 178 kmutex_t vfs_list_lock; 179 180 static pool_cache_t vnode_cache; 181 182 /* 183 * These define the root filesystem and device. 184 */ 185 struct vnode *rootvnode; 186 struct device *root_device; /* root device */ 187 188 /* 189 * Local declarations. 190 */ 191 192 static void vrele_thread(void *); 193 static void insmntque(vnode_t *, struct mount *); 194 static int getdevvp(dev_t, vnode_t **, enum vtype); 195 static vnode_t *getcleanvnode(void); 196 void vpanic(vnode_t *, const char *); 197 static void vfs_shutdown1(struct lwp *); 198 199 #ifdef DEBUG 200 void printlockedvnodes(void); 201 #endif 202 203 #ifdef DIAGNOSTIC 204 void 205 vpanic(vnode_t *vp, const char *msg) 206 { 207 208 vprint(NULL, vp); 209 panic("%s\n", msg); 210 } 211 #else 212 #define vpanic(vp, msg) /* nothing */ 213 #endif 214 215 void 216 vn_init1(void) 217 { 218 219 vnode_cache = pool_cache_init(sizeof(struct vnode), 0, 0, 0, "vnodepl", 220 NULL, IPL_NONE, NULL, NULL, NULL); 221 KASSERT(vnode_cache != NULL); 222 223 /* Create deferred release thread. */ 224 mutex_init(&vrele_lock, MUTEX_DEFAULT, IPL_NONE); 225 cv_init(&vrele_cv, "vrele"); 226 if (kthread_create(PRI_VM, KTHREAD_MPSAFE, NULL, vrele_thread, 227 NULL, &vrele_lwp, "vrele")) 228 panic("fork vrele"); 229 } 230 231 /* 232 * Initialize the vnode management data structures. 233 */ 234 void 235 vntblinit(void) 236 { 237 238 mutex_init(&mountgen_lock, MUTEX_DEFAULT, IPL_NONE); 239 mutex_init(&mountlist_lock, MUTEX_DEFAULT, IPL_NONE); 240 mutex_init(&mntid_lock, MUTEX_DEFAULT, IPL_NONE); 241 mutex_init(&mntvnode_lock, MUTEX_DEFAULT, IPL_NONE); 242 mutex_init(&vnode_free_list_lock, MUTEX_DEFAULT, IPL_NONE); 243 mutex_init(&vfs_list_lock, MUTEX_DEFAULT, IPL_NONE); 244 245 mount_specificdata_domain = specificdata_domain_create(); 246 247 /* Initialize the filesystem syncer. */ 248 vn_initialize_syncerd(); 249 vn_init1(); 250 } 251 252 int 253 vfs_drainvnodes(long target, struct lwp *l) 254 { 255 256 while (numvnodes > target) { 257 vnode_t *vp; 258 259 mutex_enter(&vnode_free_list_lock); 260 vp = getcleanvnode(); 261 if (vp == NULL) 262 return EBUSY; /* give up */ 263 ungetnewvnode(vp); 264 } 265 266 return 0; 267 } 268 269 /* 270 * Lookup a mount point by filesystem identifier. 271 * 272 * XXX Needs to add a reference to the mount point. 273 */ 274 struct mount * 275 vfs_getvfs(fsid_t *fsid) 276 { 277 struct mount *mp; 278 279 mutex_enter(&mountlist_lock); 280 CIRCLEQ_FOREACH(mp, &mountlist, mnt_list) { 281 if (mp->mnt_stat.f_fsidx.__fsid_val[0] == fsid->__fsid_val[0] && 282 mp->mnt_stat.f_fsidx.__fsid_val[1] == fsid->__fsid_val[1]) { 283 mutex_exit(&mountlist_lock); 284 return (mp); 285 } 286 } 287 mutex_exit(&mountlist_lock); 288 return ((struct mount *)0); 289 } 290 291 /* 292 * Drop a reference to a mount structure, freeing if the last reference. 293 */ 294 void 295 vfs_destroy(struct mount *mp) 296 { 297 298 if (__predict_true((int)atomic_dec_uint_nv(&mp->mnt_refcnt) > 0)) { 299 return; 300 } 301 302 /* 303 * Nothing else has visibility of the mount: we can now 304 * free the data structures. 305 */ 306 KASSERT(mp->mnt_refcnt == 0); 307 specificdata_fini(mount_specificdata_domain, &mp->mnt_specdataref); 308 rw_destroy(&mp->mnt_unmounting); 309 mutex_destroy(&mp->mnt_updating); 310 mutex_destroy(&mp->mnt_renamelock); 311 if (mp->mnt_op != NULL) { 312 vfs_delref(mp->mnt_op); 313 } 314 kmem_free(mp, sizeof(*mp)); 315 } 316 317 /* 318 * grab a vnode from freelist and clean it. 319 */ 320 vnode_t * 321 getcleanvnode(void) 322 { 323 vnode_t *vp; 324 vnodelst_t *listhd; 325 326 KASSERT(mutex_owned(&vnode_free_list_lock)); 327 328 retry: 329 listhd = &vnode_free_list; 330 try_nextlist: 331 TAILQ_FOREACH(vp, listhd, v_freelist) { 332 /* 333 * It's safe to test v_usecount and v_iflag 334 * without holding the interlock here, since 335 * these vnodes should never appear on the 336 * lists. 337 */ 338 if (vp->v_usecount != 0) { 339 vpanic(vp, "free vnode isn't"); 340 } 341 if ((vp->v_iflag & VI_CLEAN) != 0) { 342 vpanic(vp, "clean vnode on freelist"); 343 } 344 if (vp->v_freelisthd != listhd) { 345 printf("vnode sez %p, listhd %p\n", vp->v_freelisthd, listhd); 346 vpanic(vp, "list head mismatch"); 347 } 348 if (!mutex_tryenter(&vp->v_interlock)) 349 continue; 350 if ((vp->v_iflag & VI_XLOCK) == 0) 351 break; 352 mutex_exit(&vp->v_interlock); 353 } 354 355 if (vp == NULL) { 356 if (listhd == &vnode_free_list) { 357 listhd = &vnode_hold_list; 358 goto try_nextlist; 359 } 360 mutex_exit(&vnode_free_list_lock); 361 return NULL; 362 } 363 364 /* Remove it from the freelist. */ 365 TAILQ_REMOVE(listhd, vp, v_freelist); 366 vp->v_freelisthd = NULL; 367 mutex_exit(&vnode_free_list_lock); 368 369 KASSERT(vp->v_usecount == 0); 370 371 /* 372 * The vnode is still associated with a file system, so we must 373 * clean it out before reusing it. We need to add a reference 374 * before doing this. If the vnode gains another reference while 375 * being cleaned out then we lose - retry. 376 */ 377 atomic_add_int(&vp->v_usecount, 1 + VC_XLOCK); 378 vclean(vp, DOCLOSE); 379 KASSERT(vp->v_usecount >= 1 + VC_XLOCK); 380 atomic_add_int(&vp->v_usecount, -VC_XLOCK); 381 if (vp->v_usecount == 1) { 382 /* We're about to dirty it. */ 383 vp->v_iflag &= ~VI_CLEAN; 384 mutex_exit(&vp->v_interlock); 385 if (vp->v_type == VBLK || vp->v_type == VCHR) { 386 spec_node_destroy(vp); 387 } 388 vp->v_type = VNON; 389 } else { 390 /* 391 * Don't return to freelist - the holder of the last 392 * reference will destroy it. 393 */ 394 vrelel(vp, 0); /* releases vp->v_interlock */ 395 mutex_enter(&vnode_free_list_lock); 396 goto retry; 397 } 398 399 if (vp->v_data != NULL || vp->v_uobj.uo_npages != 0 || 400 !TAILQ_EMPTY(&vp->v_uobj.memq)) { 401 vpanic(vp, "cleaned vnode isn't"); 402 } 403 if (vp->v_numoutput != 0) { 404 vpanic(vp, "clean vnode has pending I/O's"); 405 } 406 if ((vp->v_iflag & VI_ONWORKLST) != 0) { 407 vpanic(vp, "clean vnode on syncer list"); 408 } 409 410 return vp; 411 } 412 413 /* 414 * Mark a mount point as busy, and gain a new reference to it. Used to 415 * prevent the file system from being unmounted during critical sections. 416 * 417 * => The caller must hold a pre-existing reference to the mount. 418 * => Will fail if the file system is being unmounted, or is unmounted. 419 */ 420 int 421 vfs_busy(struct mount *mp, struct mount **nextp) 422 { 423 424 KASSERT(mp->mnt_refcnt > 0); 425 426 if (__predict_false(!rw_tryenter(&mp->mnt_unmounting, RW_READER))) { 427 if (nextp != NULL) { 428 KASSERT(mutex_owned(&mountlist_lock)); 429 *nextp = CIRCLEQ_NEXT(mp, mnt_list); 430 } 431 return EBUSY; 432 } 433 if (__predict_false((mp->mnt_iflag & IMNT_GONE) != 0)) { 434 rw_exit(&mp->mnt_unmounting); 435 if (nextp != NULL) { 436 KASSERT(mutex_owned(&mountlist_lock)); 437 *nextp = CIRCLEQ_NEXT(mp, mnt_list); 438 } 439 return ENOENT; 440 } 441 if (nextp != NULL) { 442 mutex_exit(&mountlist_lock); 443 } 444 atomic_inc_uint(&mp->mnt_refcnt); 445 return 0; 446 } 447 448 /* 449 * Unbusy a busy filesystem. 450 * 451 * => If keepref is true, preserve reference added by vfs_busy(). 452 * => If nextp != NULL, acquire mountlist_lock. 453 */ 454 void 455 vfs_unbusy(struct mount *mp, bool keepref, struct mount **nextp) 456 { 457 458 KASSERT(mp->mnt_refcnt > 0); 459 460 if (nextp != NULL) { 461 mutex_enter(&mountlist_lock); 462 } 463 rw_exit(&mp->mnt_unmounting); 464 if (!keepref) { 465 vfs_destroy(mp); 466 } 467 if (nextp != NULL) { 468 KASSERT(mutex_owned(&mountlist_lock)); 469 *nextp = CIRCLEQ_NEXT(mp, mnt_list); 470 } 471 } 472 473 struct mount * 474 vfs_mountalloc(struct vfsops *vfsops, struct vnode *vp) 475 { 476 int error; 477 struct mount *mp; 478 479 mp = kmem_zalloc(sizeof(*mp), KM_SLEEP); 480 if (mp == NULL) 481 return NULL; 482 483 mp->mnt_op = vfsops; 484 mp->mnt_refcnt = 1; 485 TAILQ_INIT(&mp->mnt_vnodelist); 486 rw_init(&mp->mnt_unmounting); 487 mutex_init(&mp->mnt_renamelock, MUTEX_DEFAULT, IPL_NONE); 488 mutex_init(&mp->mnt_updating, MUTEX_DEFAULT, IPL_NONE); 489 error = vfs_busy(mp, NULL); 490 KASSERT(error == 0); 491 mp->mnt_vnodecovered = vp; 492 mount_initspecific(mp); 493 494 mutex_enter(&mountgen_lock); 495 mp->mnt_gen = mountgen++; 496 mutex_exit(&mountgen_lock); 497 498 return mp; 499 } 500 501 /* 502 * Lookup a filesystem type, and if found allocate and initialize 503 * a mount structure for it. 504 * 505 * Devname is usually updated by mount(8) after booting. 506 */ 507 int 508 vfs_rootmountalloc(const char *fstypename, const char *devname, 509 struct mount **mpp) 510 { 511 struct vfsops *vfsp = NULL; 512 struct mount *mp; 513 514 mutex_enter(&vfs_list_lock); 515 LIST_FOREACH(vfsp, &vfs_list, vfs_list) 516 if (!strncmp(vfsp->vfs_name, fstypename, 517 sizeof(mp->mnt_stat.f_fstypename))) 518 break; 519 if (vfsp == NULL) { 520 mutex_exit(&vfs_list_lock); 521 return (ENODEV); 522 } 523 vfsp->vfs_refcount++; 524 mutex_exit(&vfs_list_lock); 525 526 if ((mp = vfs_mountalloc(vfsp, NULL)) == NULL) 527 return ENOMEM; 528 mp->mnt_flag = MNT_RDONLY; 529 (void)strlcpy(mp->mnt_stat.f_fstypename, vfsp->vfs_name, 530 sizeof(mp->mnt_stat.f_fstypename)); 531 mp->mnt_stat.f_mntonname[0] = '/'; 532 mp->mnt_stat.f_mntonname[1] = '\0'; 533 mp->mnt_stat.f_mntfromname[sizeof(mp->mnt_stat.f_mntfromname) - 1] = 534 '\0'; 535 (void)copystr(devname, mp->mnt_stat.f_mntfromname, 536 sizeof(mp->mnt_stat.f_mntfromname) - 1, 0); 537 *mpp = mp; 538 return (0); 539 } 540 541 /* 542 * Routines having to do with the management of the vnode table. 543 */ 544 extern int (**dead_vnodeop_p)(void *); 545 546 /* 547 * Return the next vnode from the free list. 548 */ 549 int 550 getnewvnode(enum vtagtype tag, struct mount *mp, int (**vops)(void *), 551 vnode_t **vpp) 552 { 553 struct uvm_object *uobj; 554 static int toggle; 555 vnode_t *vp; 556 int error = 0, tryalloc; 557 558 try_again: 559 if (mp != NULL) { 560 /* 561 * Mark filesystem busy while we're creating a 562 * vnode. If unmount is in progress, this will 563 * fail. 564 */ 565 error = vfs_busy(mp, NULL); 566 if (error) 567 return error; 568 } 569 570 /* 571 * We must choose whether to allocate a new vnode or recycle an 572 * existing one. The criterion for allocating a new one is that 573 * the total number of vnodes is less than the number desired or 574 * there are no vnodes on either free list. Generally we only 575 * want to recycle vnodes that have no buffers associated with 576 * them, so we look first on the vnode_free_list. If it is empty, 577 * we next consider vnodes with referencing buffers on the 578 * vnode_hold_list. The toggle ensures that half the time we 579 * will use a buffer from the vnode_hold_list, and half the time 580 * we will allocate a new one unless the list has grown to twice 581 * the desired size. We are reticent to recycle vnodes from the 582 * vnode_hold_list because we will lose the identity of all its 583 * referencing buffers. 584 */ 585 586 vp = NULL; 587 588 mutex_enter(&vnode_free_list_lock); 589 590 toggle ^= 1; 591 if (numvnodes > 2 * desiredvnodes) 592 toggle = 0; 593 594 tryalloc = numvnodes < desiredvnodes || 595 (TAILQ_FIRST(&vnode_free_list) == NULL && 596 (TAILQ_FIRST(&vnode_hold_list) == NULL || toggle)); 597 598 if (tryalloc) { 599 numvnodes++; 600 mutex_exit(&vnode_free_list_lock); 601 if ((vp = vnalloc(NULL)) == NULL) { 602 mutex_enter(&vnode_free_list_lock); 603 numvnodes--; 604 } else 605 vp->v_usecount = 1; 606 } 607 608 if (vp == NULL) { 609 vp = getcleanvnode(); 610 if (vp == NULL) { 611 if (mp != NULL) { 612 vfs_unbusy(mp, false, NULL); 613 } 614 if (tryalloc) { 615 printf("WARNING: unable to allocate new " 616 "vnode, retrying...\n"); 617 kpause("newvn", false, hz, NULL); 618 goto try_again; 619 } 620 tablefull("vnode", "increase kern.maxvnodes or NVNODE"); 621 *vpp = 0; 622 return (ENFILE); 623 } 624 vp->v_iflag = 0; 625 vp->v_vflag = 0; 626 vp->v_uflag = 0; 627 vp->v_socket = NULL; 628 } 629 630 KASSERT(vp->v_usecount == 1); 631 KASSERT(vp->v_freelisthd == NULL); 632 KASSERT(LIST_EMPTY(&vp->v_nclist)); 633 KASSERT(LIST_EMPTY(&vp->v_dnclist)); 634 635 vp->v_type = VNON; 636 vp->v_tag = tag; 637 vp->v_op = vops; 638 insmntque(vp, mp); 639 *vpp = vp; 640 vp->v_data = NULL; 641 642 /* 643 * initialize uvm_object within vnode. 644 */ 645 646 uobj = &vp->v_uobj; 647 KASSERT(uobj->pgops == &uvm_vnodeops); 648 KASSERT(uobj->uo_npages == 0); 649 KASSERT(TAILQ_FIRST(&uobj->memq) == NULL); 650 vp->v_size = vp->v_writesize = VSIZENOTSET; 651 652 if (mp != NULL) { 653 if ((mp->mnt_iflag & IMNT_MPSAFE) != 0) 654 vp->v_vflag |= VV_MPSAFE; 655 vfs_unbusy(mp, true, NULL); 656 } 657 658 return (0); 659 } 660 661 /* 662 * This is really just the reverse of getnewvnode(). Needed for 663 * VFS_VGET functions who may need to push back a vnode in case 664 * of a locking race. 665 */ 666 void 667 ungetnewvnode(vnode_t *vp) 668 { 669 670 KASSERT(vp->v_usecount == 1); 671 KASSERT(vp->v_data == NULL); 672 KASSERT(vp->v_freelisthd == NULL); 673 674 mutex_enter(&vp->v_interlock); 675 vp->v_iflag |= VI_CLEAN; 676 vrelel(vp, 0); 677 } 678 679 /* 680 * Allocate a new, uninitialized vnode. If 'mp' is non-NULL, this is a 681 * marker vnode and we are prepared to wait for the allocation. 682 */ 683 vnode_t * 684 vnalloc(struct mount *mp) 685 { 686 vnode_t *vp; 687 688 vp = pool_cache_get(vnode_cache, (mp != NULL ? PR_WAITOK : PR_NOWAIT)); 689 if (vp == NULL) { 690 return NULL; 691 } 692 693 memset(vp, 0, sizeof(*vp)); 694 UVM_OBJ_INIT(&vp->v_uobj, &uvm_vnodeops, 0); 695 cv_init(&vp->v_cv, "vnode"); 696 /* 697 * done by memset() above. 698 * LIST_INIT(&vp->v_nclist); 699 * LIST_INIT(&vp->v_dnclist); 700 */ 701 702 if (mp != NULL) { 703 vp->v_mount = mp; 704 vp->v_type = VBAD; 705 vp->v_iflag = VI_MARKER; 706 } else { 707 rw_init(&vp->v_lock); 708 } 709 710 return vp; 711 } 712 713 /* 714 * Free an unused, unreferenced vnode. 715 */ 716 void 717 vnfree(vnode_t *vp) 718 { 719 720 KASSERT(vp->v_usecount == 0); 721 722 if ((vp->v_iflag & VI_MARKER) == 0) { 723 rw_destroy(&vp->v_lock); 724 mutex_enter(&vnode_free_list_lock); 725 numvnodes--; 726 mutex_exit(&vnode_free_list_lock); 727 } 728 729 UVM_OBJ_DESTROY(&vp->v_uobj); 730 cv_destroy(&vp->v_cv); 731 pool_cache_put(vnode_cache, vp); 732 } 733 734 /* 735 * Remove a vnode from its freelist. 736 */ 737 static inline void 738 vremfree(vnode_t *vp) 739 { 740 741 KASSERT(mutex_owned(&vp->v_interlock)); 742 KASSERT(vp->v_usecount == 0); 743 744 /* 745 * Note that the reference count must not change until 746 * the vnode is removed. 747 */ 748 mutex_enter(&vnode_free_list_lock); 749 if (vp->v_holdcnt > 0) { 750 KASSERT(vp->v_freelisthd == &vnode_hold_list); 751 } else { 752 KASSERT(vp->v_freelisthd == &vnode_free_list); 753 } 754 TAILQ_REMOVE(vp->v_freelisthd, vp, v_freelist); 755 vp->v_freelisthd = NULL; 756 mutex_exit(&vnode_free_list_lock); 757 } 758 759 /* 760 * Move a vnode from one mount queue to another. 761 */ 762 static void 763 insmntque(vnode_t *vp, struct mount *mp) 764 { 765 struct mount *omp; 766 767 #ifdef DIAGNOSTIC 768 if ((mp != NULL) && 769 (mp->mnt_iflag & IMNT_UNMOUNT) && 770 vp->v_tag != VT_VFS) { 771 panic("insmntque into dying filesystem"); 772 } 773 #endif 774 775 mutex_enter(&mntvnode_lock); 776 /* 777 * Delete from old mount point vnode list, if on one. 778 */ 779 if ((omp = vp->v_mount) != NULL) 780 TAILQ_REMOVE(&vp->v_mount->mnt_vnodelist, vp, v_mntvnodes); 781 /* 782 * Insert into list of vnodes for the new mount point, if 783 * available. The caller must take a reference on the mount 784 * structure and donate to the vnode. 785 */ 786 if ((vp->v_mount = mp) != NULL) 787 TAILQ_INSERT_TAIL(&mp->mnt_vnodelist, vp, v_mntvnodes); 788 mutex_exit(&mntvnode_lock); 789 790 if (omp != NULL) { 791 /* Release reference to old mount. */ 792 vfs_destroy(omp); 793 } 794 } 795 796 /* 797 * Wait for a vnode (typically with VI_XLOCK set) to be cleaned or 798 * recycled. 799 */ 800 void 801 vwait(vnode_t *vp, int flags) 802 { 803 804 KASSERT(mutex_owned(&vp->v_interlock)); 805 KASSERT(vp->v_usecount != 0); 806 807 while ((vp->v_iflag & flags) != 0) 808 cv_wait(&vp->v_cv, &vp->v_interlock); 809 } 810 811 /* 812 * Insert a marker vnode into a mount's vnode list, after the 813 * specified vnode. mntvnode_lock must be held. 814 */ 815 void 816 vmark(vnode_t *mvp, vnode_t *vp) 817 { 818 struct mount *mp; 819 820 mp = mvp->v_mount; 821 822 KASSERT(mutex_owned(&mntvnode_lock)); 823 KASSERT((mvp->v_iflag & VI_MARKER) != 0); 824 KASSERT(vp->v_mount == mp); 825 826 TAILQ_INSERT_AFTER(&mp->mnt_vnodelist, vp, mvp, v_mntvnodes); 827 } 828 829 /* 830 * Remove a marker vnode from a mount's vnode list, and return 831 * a pointer to the next vnode in the list. mntvnode_lock must 832 * be held. 833 */ 834 vnode_t * 835 vunmark(vnode_t *mvp) 836 { 837 vnode_t *vp; 838 struct mount *mp; 839 840 mp = mvp->v_mount; 841 842 KASSERT(mutex_owned(&mntvnode_lock)); 843 KASSERT((mvp->v_iflag & VI_MARKER) != 0); 844 845 vp = TAILQ_NEXT(mvp, v_mntvnodes); 846 TAILQ_REMOVE(&mp->mnt_vnodelist, mvp, v_mntvnodes); 847 848 KASSERT(vp == NULL || vp->v_mount == mp); 849 850 return vp; 851 } 852 853 /* 854 * Update outstanding I/O count and do wakeup if requested. 855 */ 856 void 857 vwakeup(struct buf *bp) 858 { 859 struct vnode *vp; 860 861 if ((vp = bp->b_vp) == NULL) 862 return; 863 864 KASSERT(bp->b_objlock == &vp->v_interlock); 865 KASSERT(mutex_owned(bp->b_objlock)); 866 867 if (--vp->v_numoutput < 0) 868 panic("vwakeup: neg numoutput, vp %p", vp); 869 if (vp->v_numoutput == 0) 870 cv_broadcast(&vp->v_cv); 871 } 872 873 /* 874 * Flush out and invalidate all buffers associated with a vnode. 875 * Called with the underlying vnode locked, which should prevent new dirty 876 * buffers from being queued. 877 */ 878 int 879 vinvalbuf(struct vnode *vp, int flags, kauth_cred_t cred, struct lwp *l, 880 bool catch, int slptimeo) 881 { 882 struct buf *bp, *nbp; 883 int error; 884 int flushflags = PGO_ALLPAGES | PGO_FREE | PGO_SYNCIO | 885 (flags & V_SAVE ? PGO_CLEANIT | PGO_RECLAIM : 0); 886 887 /* XXXUBC this doesn't look at flags or slp* */ 888 mutex_enter(&vp->v_interlock); 889 error = VOP_PUTPAGES(vp, 0, 0, flushflags); 890 if (error) { 891 return error; 892 } 893 894 if (flags & V_SAVE) { 895 error = VOP_FSYNC(vp, cred, FSYNC_WAIT|FSYNC_RECLAIM, 0, 0); 896 if (error) 897 return (error); 898 KASSERT(LIST_EMPTY(&vp->v_dirtyblkhd)); 899 } 900 901 mutex_enter(&bufcache_lock); 902 restart: 903 for (bp = LIST_FIRST(&vp->v_dirtyblkhd); bp; bp = nbp) { 904 nbp = LIST_NEXT(bp, b_vnbufs); 905 error = bbusy(bp, catch, slptimeo, NULL); 906 if (error != 0) { 907 if (error == EPASSTHROUGH) 908 goto restart; 909 mutex_exit(&bufcache_lock); 910 return (error); 911 } 912 brelsel(bp, BC_INVAL | BC_VFLUSH); 913 } 914 915 for (bp = LIST_FIRST(&vp->v_cleanblkhd); bp; bp = nbp) { 916 nbp = LIST_NEXT(bp, b_vnbufs); 917 error = bbusy(bp, catch, slptimeo, NULL); 918 if (error != 0) { 919 if (error == EPASSTHROUGH) 920 goto restart; 921 mutex_exit(&bufcache_lock); 922 return (error); 923 } 924 /* 925 * XXX Since there are no node locks for NFS, I believe 926 * there is a slight chance that a delayed write will 927 * occur while sleeping just above, so check for it. 928 */ 929 if ((bp->b_oflags & BO_DELWRI) && (flags & V_SAVE)) { 930 #ifdef DEBUG 931 printf("buffer still DELWRI\n"); 932 #endif 933 bp->b_cflags |= BC_BUSY | BC_VFLUSH; 934 mutex_exit(&bufcache_lock); 935 VOP_BWRITE(bp); 936 mutex_enter(&bufcache_lock); 937 goto restart; 938 } 939 brelsel(bp, BC_INVAL | BC_VFLUSH); 940 } 941 942 #ifdef DIAGNOSTIC 943 if (!LIST_EMPTY(&vp->v_cleanblkhd) || !LIST_EMPTY(&vp->v_dirtyblkhd)) 944 panic("vinvalbuf: flush failed, vp %p", vp); 945 #endif 946 947 mutex_exit(&bufcache_lock); 948 949 return (0); 950 } 951 952 /* 953 * Destroy any in core blocks past the truncation length. 954 * Called with the underlying vnode locked, which should prevent new dirty 955 * buffers from being queued. 956 */ 957 int 958 vtruncbuf(struct vnode *vp, daddr_t lbn, bool catch, int slptimeo) 959 { 960 struct buf *bp, *nbp; 961 int error; 962 voff_t off; 963 964 off = round_page((voff_t)lbn << vp->v_mount->mnt_fs_bshift); 965 mutex_enter(&vp->v_interlock); 966 error = VOP_PUTPAGES(vp, off, 0, PGO_FREE | PGO_SYNCIO); 967 if (error) { 968 return error; 969 } 970 971 mutex_enter(&bufcache_lock); 972 restart: 973 for (bp = LIST_FIRST(&vp->v_dirtyblkhd); bp; bp = nbp) { 974 nbp = LIST_NEXT(bp, b_vnbufs); 975 if (bp->b_lblkno < lbn) 976 continue; 977 error = bbusy(bp, catch, slptimeo, NULL); 978 if (error != 0) { 979 if (error == EPASSTHROUGH) 980 goto restart; 981 mutex_exit(&bufcache_lock); 982 return (error); 983 } 984 brelsel(bp, BC_INVAL | BC_VFLUSH); 985 } 986 987 for (bp = LIST_FIRST(&vp->v_cleanblkhd); bp; bp = nbp) { 988 nbp = LIST_NEXT(bp, b_vnbufs); 989 if (bp->b_lblkno < lbn) 990 continue; 991 error = bbusy(bp, catch, slptimeo, NULL); 992 if (error != 0) { 993 if (error == EPASSTHROUGH) 994 goto restart; 995 mutex_exit(&bufcache_lock); 996 return (error); 997 } 998 brelsel(bp, BC_INVAL | BC_VFLUSH); 999 } 1000 mutex_exit(&bufcache_lock); 1001 1002 return (0); 1003 } 1004 1005 /* 1006 * Flush all dirty buffers from a vnode. 1007 * Called with the underlying vnode locked, which should prevent new dirty 1008 * buffers from being queued. 1009 */ 1010 void 1011 vflushbuf(struct vnode *vp, int sync) 1012 { 1013 struct buf *bp, *nbp; 1014 int flags = PGO_CLEANIT | PGO_ALLPAGES | (sync ? PGO_SYNCIO : 0); 1015 bool dirty; 1016 1017 mutex_enter(&vp->v_interlock); 1018 (void) VOP_PUTPAGES(vp, 0, 0, flags); 1019 1020 loop: 1021 mutex_enter(&bufcache_lock); 1022 for (bp = LIST_FIRST(&vp->v_dirtyblkhd); bp; bp = nbp) { 1023 nbp = LIST_NEXT(bp, b_vnbufs); 1024 if ((bp->b_cflags & BC_BUSY)) 1025 continue; 1026 if ((bp->b_oflags & BO_DELWRI) == 0) 1027 panic("vflushbuf: not dirty, bp %p", bp); 1028 bp->b_cflags |= BC_BUSY | BC_VFLUSH; 1029 mutex_exit(&bufcache_lock); 1030 /* 1031 * Wait for I/O associated with indirect blocks to complete, 1032 * since there is no way to quickly wait for them below. 1033 */ 1034 if (bp->b_vp == vp || sync == 0) 1035 (void) bawrite(bp); 1036 else 1037 (void) bwrite(bp); 1038 goto loop; 1039 } 1040 mutex_exit(&bufcache_lock); 1041 1042 if (sync == 0) 1043 return; 1044 1045 mutex_enter(&vp->v_interlock); 1046 while (vp->v_numoutput != 0) 1047 cv_wait(&vp->v_cv, &vp->v_interlock); 1048 dirty = !LIST_EMPTY(&vp->v_dirtyblkhd); 1049 mutex_exit(&vp->v_interlock); 1050 1051 if (dirty) { 1052 vprint("vflushbuf: dirty", vp); 1053 goto loop; 1054 } 1055 } 1056 1057 /* 1058 * Create a vnode for a block device. 1059 * Used for root filesystem and swap areas. 1060 * Also used for memory file system special devices. 1061 */ 1062 int 1063 bdevvp(dev_t dev, vnode_t **vpp) 1064 { 1065 1066 return (getdevvp(dev, vpp, VBLK)); 1067 } 1068 1069 /* 1070 * Create a vnode for a character device. 1071 * Used for kernfs and some console handling. 1072 */ 1073 int 1074 cdevvp(dev_t dev, vnode_t **vpp) 1075 { 1076 1077 return (getdevvp(dev, vpp, VCHR)); 1078 } 1079 1080 /* 1081 * Associate a buffer with a vnode. There must already be a hold on 1082 * the vnode. 1083 */ 1084 void 1085 bgetvp(struct vnode *vp, struct buf *bp) 1086 { 1087 1088 KASSERT(bp->b_vp == NULL); 1089 KASSERT(bp->b_objlock == &buffer_lock); 1090 KASSERT(mutex_owned(&vp->v_interlock)); 1091 KASSERT(mutex_owned(&bufcache_lock)); 1092 KASSERT((bp->b_cflags & BC_BUSY) != 0); 1093 KASSERT(!cv_has_waiters(&bp->b_done)); 1094 1095 vholdl(vp); 1096 bp->b_vp = vp; 1097 if (vp->v_type == VBLK || vp->v_type == VCHR) 1098 bp->b_dev = vp->v_rdev; 1099 else 1100 bp->b_dev = NODEV; 1101 1102 /* 1103 * Insert onto list for new vnode. 1104 */ 1105 bufinsvn(bp, &vp->v_cleanblkhd); 1106 bp->b_objlock = &vp->v_interlock; 1107 } 1108 1109 /* 1110 * Disassociate a buffer from a vnode. 1111 */ 1112 void 1113 brelvp(struct buf *bp) 1114 { 1115 struct vnode *vp = bp->b_vp; 1116 1117 KASSERT(vp != NULL); 1118 KASSERT(bp->b_objlock == &vp->v_interlock); 1119 KASSERT(mutex_owned(&vp->v_interlock)); 1120 KASSERT(mutex_owned(&bufcache_lock)); 1121 KASSERT((bp->b_cflags & BC_BUSY) != 0); 1122 KASSERT(!cv_has_waiters(&bp->b_done)); 1123 1124 /* 1125 * Delete from old vnode list, if on one. 1126 */ 1127 if (LIST_NEXT(bp, b_vnbufs) != NOLIST) 1128 bufremvn(bp); 1129 1130 if (vp->v_uobj.uo_npages == 0 && (vp->v_iflag & VI_ONWORKLST) && 1131 LIST_FIRST(&vp->v_dirtyblkhd) == NULL) { 1132 vp->v_iflag &= ~VI_WRMAPDIRTY; 1133 vn_syncer_remove_from_worklist(vp); 1134 } 1135 1136 bp->b_objlock = &buffer_lock; 1137 bp->b_vp = NULL; 1138 holdrelel(vp); 1139 } 1140 1141 /* 1142 * Reassign a buffer from one vnode list to another. 1143 * The list reassignment must be within the same vnode. 1144 * Used to assign file specific control information 1145 * (indirect blocks) to the list to which they belong. 1146 */ 1147 void 1148 reassignbuf(struct buf *bp, struct vnode *vp) 1149 { 1150 struct buflists *listheadp; 1151 int delayx; 1152 1153 KASSERT(mutex_owned(&bufcache_lock)); 1154 KASSERT(bp->b_objlock == &vp->v_interlock); 1155 KASSERT(mutex_owned(&vp->v_interlock)); 1156 KASSERT((bp->b_cflags & BC_BUSY) != 0); 1157 1158 /* 1159 * Delete from old vnode list, if on one. 1160 */ 1161 if (LIST_NEXT(bp, b_vnbufs) != NOLIST) 1162 bufremvn(bp); 1163 1164 /* 1165 * If dirty, put on list of dirty buffers; 1166 * otherwise insert onto list of clean buffers. 1167 */ 1168 if ((bp->b_oflags & BO_DELWRI) == 0) { 1169 listheadp = &vp->v_cleanblkhd; 1170 if (vp->v_uobj.uo_npages == 0 && 1171 (vp->v_iflag & VI_ONWORKLST) && 1172 LIST_FIRST(&vp->v_dirtyblkhd) == NULL) { 1173 vp->v_iflag &= ~VI_WRMAPDIRTY; 1174 vn_syncer_remove_from_worklist(vp); 1175 } 1176 } else { 1177 listheadp = &vp->v_dirtyblkhd; 1178 if ((vp->v_iflag & VI_ONWORKLST) == 0) { 1179 switch (vp->v_type) { 1180 case VDIR: 1181 delayx = dirdelay; 1182 break; 1183 case VBLK: 1184 if (vp->v_specmountpoint != NULL) { 1185 delayx = metadelay; 1186 break; 1187 } 1188 /* fall through */ 1189 default: 1190 delayx = filedelay; 1191 break; 1192 } 1193 if (!vp->v_mount || 1194 (vp->v_mount->mnt_flag & MNT_ASYNC) == 0) 1195 vn_syncer_add_to_worklist(vp, delayx); 1196 } 1197 } 1198 bufinsvn(bp, listheadp); 1199 } 1200 1201 /* 1202 * Create a vnode for a device. 1203 * Used by bdevvp (block device) for root file system etc., 1204 * and by cdevvp (character device) for console and kernfs. 1205 */ 1206 static int 1207 getdevvp(dev_t dev, vnode_t **vpp, enum vtype type) 1208 { 1209 vnode_t *vp; 1210 vnode_t *nvp; 1211 int error; 1212 1213 if (dev == NODEV) { 1214 *vpp = NULL; 1215 return (0); 1216 } 1217 error = getnewvnode(VT_NON, NULL, spec_vnodeop_p, &nvp); 1218 if (error) { 1219 *vpp = NULL; 1220 return (error); 1221 } 1222 vp = nvp; 1223 vp->v_type = type; 1224 vp->v_vflag |= VV_MPSAFE; 1225 uvm_vnp_setsize(vp, 0); 1226 spec_node_init(vp, dev); 1227 *vpp = vp; 1228 return (0); 1229 } 1230 1231 /* 1232 * Try to gain a reference to a vnode, without acquiring its interlock. 1233 * The caller must hold a lock that will prevent the vnode from being 1234 * recycled or freed. 1235 */ 1236 bool 1237 vtryget(vnode_t *vp) 1238 { 1239 u_int use, next; 1240 1241 /* 1242 * If the vnode is being freed, don't make life any harder 1243 * for vclean() by adding another reference without waiting. 1244 * This is not strictly necessary, but we'll do it anyway. 1245 */ 1246 if (__predict_false((vp->v_iflag & VI_XLOCK) != 0)) { 1247 return false; 1248 } 1249 for (use = vp->v_usecount;; use = next) { 1250 if (use == 0 || __predict_false((use & VC_XLOCK) != 0)) { 1251 /* Need interlock held if first reference. */ 1252 return false; 1253 } 1254 next = atomic_cas_uint(&vp->v_usecount, use, use + 1); 1255 if (__predict_true(next == use)) { 1256 return true; 1257 } 1258 } 1259 } 1260 1261 /* 1262 * Grab a particular vnode from the free list, increment its 1263 * reference count and lock it. If the vnode lock bit is set the 1264 * vnode is being eliminated in vgone. In that case, we can not 1265 * grab the vnode, so the process is awakened when the transition is 1266 * completed, and an error returned to indicate that the vnode is no 1267 * longer usable (possibly having been changed to a new file system type). 1268 * Called with v_interlock held. 1269 */ 1270 int 1271 vget(vnode_t *vp, int flags) 1272 { 1273 int error = 0; 1274 1275 KASSERT((vp->v_iflag & VI_MARKER) == 0); 1276 KASSERT(mutex_owned(&vp->v_interlock)); 1277 KASSERT((flags & ~(LK_SHARED|LK_EXCLUSIVE|LK_NOWAIT)) == 0); 1278 1279 /* 1280 * Before adding a reference, we must remove the vnode 1281 * from its freelist. 1282 */ 1283 if (vp->v_usecount == 0) { 1284 vremfree(vp); 1285 vp->v_usecount = 1; 1286 } else { 1287 atomic_inc_uint(&vp->v_usecount); 1288 } 1289 1290 /* 1291 * If the vnode is in the process of being cleaned out for 1292 * another use, we wait for the cleaning to finish and then 1293 * return failure. Cleaning is determined by checking if 1294 * the VI_XLOCK flag is set. 1295 */ 1296 if ((vp->v_iflag & VI_XLOCK) != 0) { 1297 if ((flags & LK_NOWAIT) != 0) { 1298 vrelel(vp, 0); 1299 return EBUSY; 1300 } 1301 vwait(vp, VI_XLOCK); 1302 vrelel(vp, 0); 1303 return ENOENT; 1304 } 1305 1306 /* 1307 * Ok, we got it in good shape. Just locking left. 1308 */ 1309 KASSERT((vp->v_iflag & VI_CLEAN) == 0); 1310 mutex_exit(&vp->v_interlock); 1311 if (flags & (LK_EXCLUSIVE | LK_SHARED)) { 1312 error = vn_lock(vp, flags); 1313 if (error != 0) { 1314 vrele(vp); 1315 } 1316 } 1317 return error; 1318 } 1319 1320 /* 1321 * vput(), just unlock and vrele() 1322 */ 1323 void 1324 vput(vnode_t *vp) 1325 { 1326 1327 KASSERT((vp->v_iflag & VI_MARKER) == 0); 1328 1329 VOP_UNLOCK(vp); 1330 vrele(vp); 1331 } 1332 1333 /* 1334 * Try to drop reference on a vnode. Abort if we are releasing the 1335 * last reference. Note: this _must_ succeed if not the last reference. 1336 */ 1337 static inline bool 1338 vtryrele(vnode_t *vp) 1339 { 1340 u_int use, next; 1341 1342 for (use = vp->v_usecount;; use = next) { 1343 if (use == 1) { 1344 return false; 1345 } 1346 KASSERT((use & VC_MASK) > 1); 1347 next = atomic_cas_uint(&vp->v_usecount, use, use - 1); 1348 if (__predict_true(next == use)) { 1349 return true; 1350 } 1351 } 1352 } 1353 1354 /* 1355 * Vnode release. If reference count drops to zero, call inactive 1356 * routine and either return to freelist or free to the pool. 1357 */ 1358 void 1359 vrelel(vnode_t *vp, int flags) 1360 { 1361 bool recycle, defer; 1362 int error; 1363 1364 KASSERT(mutex_owned(&vp->v_interlock)); 1365 KASSERT((vp->v_iflag & VI_MARKER) == 0); 1366 KASSERT(vp->v_freelisthd == NULL); 1367 1368 if (__predict_false(vp->v_op == dead_vnodeop_p && 1369 (vp->v_iflag & (VI_CLEAN|VI_XLOCK)) == 0)) { 1370 vpanic(vp, "dead but not clean"); 1371 } 1372 1373 /* 1374 * If not the last reference, just drop the reference count 1375 * and unlock. 1376 */ 1377 if (vtryrele(vp)) { 1378 vp->v_iflag |= VI_INACTREDO; 1379 mutex_exit(&vp->v_interlock); 1380 return; 1381 } 1382 if (vp->v_usecount <= 0 || vp->v_writecount != 0) { 1383 vpanic(vp, "vrelel: bad ref count"); 1384 } 1385 1386 KASSERT((vp->v_iflag & VI_XLOCK) == 0); 1387 1388 /* 1389 * If not clean, deactivate the vnode, but preserve 1390 * our reference across the call to VOP_INACTIVE(). 1391 */ 1392 retry: 1393 if ((vp->v_iflag & VI_CLEAN) == 0) { 1394 recycle = false; 1395 vp->v_iflag |= VI_INACTNOW; 1396 1397 /* 1398 * XXX This ugly block can be largely eliminated if 1399 * locking is pushed down into the file systems. 1400 * 1401 * Defer vnode release to vrele_thread if caller 1402 * requests it explicitly. 1403 */ 1404 if ((curlwp == uvm.pagedaemon_lwp) || 1405 (flags & VRELEL_ASYNC_RELE) != 0) { 1406 /* The pagedaemon can't wait around; defer. */ 1407 defer = true; 1408 } else if (curlwp == vrele_lwp) { 1409 /* We have to try harder. */ 1410 vp->v_iflag &= ~VI_INACTREDO; 1411 mutex_exit(&vp->v_interlock); 1412 error = vn_lock(vp, LK_EXCLUSIVE); 1413 if (error != 0) { 1414 /* XXX */ 1415 vpanic(vp, "vrele: unable to lock %p"); 1416 } 1417 defer = false; 1418 } else if ((vp->v_iflag & VI_LAYER) != 0) { 1419 /* 1420 * Acquiring the stack's lock in vclean() even 1421 * for an honest vput/vrele is dangerous because 1422 * our caller may hold other vnode locks; defer. 1423 */ 1424 defer = true; 1425 } else { 1426 /* If we can't acquire the lock, then defer. */ 1427 vp->v_iflag &= ~VI_INACTREDO; 1428 mutex_exit(&vp->v_interlock); 1429 error = vn_lock(vp, LK_EXCLUSIVE | LK_NOWAIT); 1430 if (error != 0) { 1431 defer = true; 1432 mutex_enter(&vp->v_interlock); 1433 } else { 1434 defer = false; 1435 } 1436 } 1437 1438 if (defer) { 1439 /* 1440 * Defer reclaim to the kthread; it's not safe to 1441 * clean it here. We donate it our last reference. 1442 */ 1443 KASSERT(mutex_owned(&vp->v_interlock)); 1444 KASSERT((vp->v_iflag & VI_INACTPEND) == 0); 1445 vp->v_iflag &= ~VI_INACTNOW; 1446 vp->v_iflag |= VI_INACTPEND; 1447 mutex_enter(&vrele_lock); 1448 TAILQ_INSERT_TAIL(&vrele_list, vp, v_freelist); 1449 if (++vrele_pending > (desiredvnodes >> 8)) 1450 cv_signal(&vrele_cv); 1451 mutex_exit(&vrele_lock); 1452 mutex_exit(&vp->v_interlock); 1453 return; 1454 } 1455 1456 #ifdef DIAGNOSTIC 1457 if ((vp->v_type == VBLK || vp->v_type == VCHR) && 1458 vp->v_specnode != NULL && vp->v_specnode->sn_opencnt != 0) { 1459 vprint("vrelel: missing VOP_CLOSE()", vp); 1460 } 1461 #endif 1462 1463 /* 1464 * The vnode can gain another reference while being 1465 * deactivated. If VOP_INACTIVE() indicates that 1466 * the described file has been deleted, then recycle 1467 * the vnode irrespective of additional references. 1468 * Another thread may be waiting to re-use the on-disk 1469 * inode. 1470 * 1471 * Note that VOP_INACTIVE() will drop the vnode lock. 1472 */ 1473 VOP_INACTIVE(vp, &recycle); 1474 mutex_enter(&vp->v_interlock); 1475 vp->v_iflag &= ~VI_INACTNOW; 1476 if (!recycle) { 1477 if (vtryrele(vp)) { 1478 mutex_exit(&vp->v_interlock); 1479 return; 1480 } 1481 1482 /* 1483 * If we grew another reference while 1484 * VOP_INACTIVE() was underway, retry. 1485 */ 1486 if ((vp->v_iflag & VI_INACTREDO) != 0) { 1487 goto retry; 1488 } 1489 } 1490 1491 /* Take care of space accounting. */ 1492 if (vp->v_iflag & VI_EXECMAP) { 1493 atomic_add_int(&uvmexp.execpages, 1494 -vp->v_uobj.uo_npages); 1495 atomic_add_int(&uvmexp.filepages, 1496 vp->v_uobj.uo_npages); 1497 } 1498 vp->v_iflag &= ~(VI_TEXT|VI_EXECMAP|VI_WRMAP); 1499 vp->v_vflag &= ~VV_MAPPED; 1500 1501 /* 1502 * Recycle the vnode if the file is now unused (unlinked), 1503 * otherwise just free it. 1504 */ 1505 if (recycle) { 1506 vclean(vp, DOCLOSE); 1507 } 1508 KASSERT(vp->v_usecount > 0); 1509 } 1510 1511 if (atomic_dec_uint_nv(&vp->v_usecount) != 0) { 1512 /* Gained another reference while being reclaimed. */ 1513 mutex_exit(&vp->v_interlock); 1514 return; 1515 } 1516 1517 if ((vp->v_iflag & VI_CLEAN) != 0) { 1518 /* 1519 * It's clean so destroy it. It isn't referenced 1520 * anywhere since it has been reclaimed. 1521 */ 1522 KASSERT(vp->v_holdcnt == 0); 1523 KASSERT(vp->v_writecount == 0); 1524 mutex_exit(&vp->v_interlock); 1525 insmntque(vp, NULL); 1526 if (vp->v_type == VBLK || vp->v_type == VCHR) { 1527 spec_node_destroy(vp); 1528 } 1529 vnfree(vp); 1530 } else { 1531 /* 1532 * Otherwise, put it back onto the freelist. It 1533 * can't be destroyed while still associated with 1534 * a file system. 1535 */ 1536 mutex_enter(&vnode_free_list_lock); 1537 if (vp->v_holdcnt > 0) { 1538 vp->v_freelisthd = &vnode_hold_list; 1539 } else { 1540 vp->v_freelisthd = &vnode_free_list; 1541 } 1542 TAILQ_INSERT_TAIL(vp->v_freelisthd, vp, v_freelist); 1543 mutex_exit(&vnode_free_list_lock); 1544 mutex_exit(&vp->v_interlock); 1545 } 1546 } 1547 1548 void 1549 vrele(vnode_t *vp) 1550 { 1551 1552 KASSERT((vp->v_iflag & VI_MARKER) == 0); 1553 1554 if ((vp->v_iflag & VI_INACTNOW) == 0 && vtryrele(vp)) { 1555 return; 1556 } 1557 mutex_enter(&vp->v_interlock); 1558 vrelel(vp, 0); 1559 } 1560 1561 /* 1562 * Asynchronous vnode release, vnode is released in different context. 1563 */ 1564 void 1565 vrele_async(vnode_t *vp) 1566 { 1567 1568 KASSERT((vp->v_iflag & VI_MARKER) == 0); 1569 1570 if ((vp->v_iflag & VI_INACTNOW) == 0 && vtryrele(vp)) { 1571 return; 1572 } 1573 1574 mutex_enter(&vp->v_interlock); 1575 vrelel(vp, VRELEL_ASYNC_RELE); 1576 } 1577 1578 static void 1579 vrele_thread(void *cookie) 1580 { 1581 vnode_t *vp; 1582 1583 for (;;) { 1584 mutex_enter(&vrele_lock); 1585 while (TAILQ_EMPTY(&vrele_list)) { 1586 vrele_gen++; 1587 cv_broadcast(&vrele_cv); 1588 cv_timedwait(&vrele_cv, &vrele_lock, hz); 1589 } 1590 vp = TAILQ_FIRST(&vrele_list); 1591 TAILQ_REMOVE(&vrele_list, vp, v_freelist); 1592 vrele_pending--; 1593 mutex_exit(&vrele_lock); 1594 1595 /* 1596 * If not the last reference, then ignore the vnode 1597 * and look for more work. 1598 */ 1599 mutex_enter(&vp->v_interlock); 1600 KASSERT((vp->v_iflag & VI_INACTPEND) != 0); 1601 vp->v_iflag &= ~VI_INACTPEND; 1602 vrelel(vp, 0); 1603 } 1604 } 1605 1606 /* 1607 * Page or buffer structure gets a reference. 1608 * Called with v_interlock held. 1609 */ 1610 void 1611 vholdl(vnode_t *vp) 1612 { 1613 1614 KASSERT(mutex_owned(&vp->v_interlock)); 1615 KASSERT((vp->v_iflag & VI_MARKER) == 0); 1616 1617 if (vp->v_holdcnt++ == 0 && vp->v_usecount == 0) { 1618 mutex_enter(&vnode_free_list_lock); 1619 KASSERT(vp->v_freelisthd == &vnode_free_list); 1620 TAILQ_REMOVE(vp->v_freelisthd, vp, v_freelist); 1621 vp->v_freelisthd = &vnode_hold_list; 1622 TAILQ_INSERT_TAIL(vp->v_freelisthd, vp, v_freelist); 1623 mutex_exit(&vnode_free_list_lock); 1624 } 1625 } 1626 1627 /* 1628 * Page or buffer structure frees a reference. 1629 * Called with v_interlock held. 1630 */ 1631 void 1632 holdrelel(vnode_t *vp) 1633 { 1634 1635 KASSERT(mutex_owned(&vp->v_interlock)); 1636 KASSERT((vp->v_iflag & VI_MARKER) == 0); 1637 1638 if (vp->v_holdcnt <= 0) { 1639 vpanic(vp, "holdrelel: holdcnt vp %p"); 1640 } 1641 1642 vp->v_holdcnt--; 1643 if (vp->v_holdcnt == 0 && vp->v_usecount == 0) { 1644 mutex_enter(&vnode_free_list_lock); 1645 KASSERT(vp->v_freelisthd == &vnode_hold_list); 1646 TAILQ_REMOVE(vp->v_freelisthd, vp, v_freelist); 1647 vp->v_freelisthd = &vnode_free_list; 1648 TAILQ_INSERT_TAIL(vp->v_freelisthd, vp, v_freelist); 1649 mutex_exit(&vnode_free_list_lock); 1650 } 1651 } 1652 1653 /* 1654 * Vnode reference, where a reference is already held by some other 1655 * object (for example, a file structure). 1656 */ 1657 void 1658 vref(vnode_t *vp) 1659 { 1660 1661 KASSERT((vp->v_iflag & VI_MARKER) == 0); 1662 KASSERT(vp->v_usecount != 0); 1663 1664 atomic_inc_uint(&vp->v_usecount); 1665 } 1666 1667 /* 1668 * Remove any vnodes in the vnode table belonging to mount point mp. 1669 * 1670 * If FORCECLOSE is not specified, there should not be any active ones, 1671 * return error if any are found (nb: this is a user error, not a 1672 * system error). If FORCECLOSE is specified, detach any active vnodes 1673 * that are found. 1674 * 1675 * If WRITECLOSE is set, only flush out regular file vnodes open for 1676 * writing. 1677 * 1678 * SKIPSYSTEM causes any vnodes marked V_SYSTEM to be skipped. 1679 */ 1680 #ifdef DEBUG 1681 int busyprt = 0; /* print out busy vnodes */ 1682 struct ctldebug debug1 = { "busyprt", &busyprt }; 1683 #endif 1684 1685 static vnode_t * 1686 vflushnext(vnode_t *mvp, int *when) 1687 { 1688 1689 if (hardclock_ticks > *when) { 1690 mutex_exit(&mntvnode_lock); 1691 yield(); 1692 mutex_enter(&mntvnode_lock); 1693 *when = hardclock_ticks + hz / 10; 1694 } 1695 1696 return vunmark(mvp); 1697 } 1698 1699 int 1700 vflush(struct mount *mp, vnode_t *skipvp, int flags) 1701 { 1702 vnode_t *vp, *mvp; 1703 int busy = 0, when = 0, gen; 1704 1705 /* 1706 * First, flush out any vnode references from vrele_list. 1707 */ 1708 mutex_enter(&vrele_lock); 1709 gen = vrele_gen; 1710 while (vrele_pending && gen == vrele_gen) { 1711 cv_broadcast(&vrele_cv); 1712 cv_wait(&vrele_cv, &vrele_lock); 1713 } 1714 mutex_exit(&vrele_lock); 1715 1716 /* Allocate a marker vnode. */ 1717 if ((mvp = vnalloc(mp)) == NULL) 1718 return (ENOMEM); 1719 1720 /* 1721 * NOTE: not using the TAILQ_FOREACH here since in this loop vgone() 1722 * and vclean() are called 1723 */ 1724 mutex_enter(&mntvnode_lock); 1725 for (vp = TAILQ_FIRST(&mp->mnt_vnodelist); vp != NULL; 1726 vp = vflushnext(mvp, &when)) { 1727 vmark(mvp, vp); 1728 if (vp->v_mount != mp || vismarker(vp)) 1729 continue; 1730 /* 1731 * Skip over a selected vnode. 1732 */ 1733 if (vp == skipvp) 1734 continue; 1735 mutex_enter(&vp->v_interlock); 1736 /* 1737 * Ignore clean but still referenced vnodes. 1738 */ 1739 if ((vp->v_iflag & VI_CLEAN) != 0) { 1740 mutex_exit(&vp->v_interlock); 1741 continue; 1742 } 1743 /* 1744 * Skip over a vnodes marked VSYSTEM. 1745 */ 1746 if ((flags & SKIPSYSTEM) && (vp->v_vflag & VV_SYSTEM)) { 1747 mutex_exit(&vp->v_interlock); 1748 continue; 1749 } 1750 /* 1751 * If WRITECLOSE is set, only flush out regular file 1752 * vnodes open for writing. 1753 */ 1754 if ((flags & WRITECLOSE) && 1755 (vp->v_writecount == 0 || vp->v_type != VREG)) { 1756 mutex_exit(&vp->v_interlock); 1757 continue; 1758 } 1759 /* 1760 * With v_usecount == 0, all we need to do is clear 1761 * out the vnode data structures and we are done. 1762 */ 1763 if (vp->v_usecount == 0) { 1764 mutex_exit(&mntvnode_lock); 1765 vremfree(vp); 1766 vp->v_usecount = 1; 1767 vclean(vp, DOCLOSE); 1768 vrelel(vp, 0); 1769 mutex_enter(&mntvnode_lock); 1770 continue; 1771 } 1772 /* 1773 * If FORCECLOSE is set, forcibly close the vnode. 1774 * For block or character devices, revert to an 1775 * anonymous device. For all other files, just 1776 * kill them. 1777 */ 1778 if (flags & FORCECLOSE) { 1779 mutex_exit(&mntvnode_lock); 1780 atomic_inc_uint(&vp->v_usecount); 1781 if (vp->v_type != VBLK && vp->v_type != VCHR) { 1782 vclean(vp, DOCLOSE); 1783 vrelel(vp, 0); 1784 } else { 1785 vclean(vp, 0); 1786 vp->v_op = spec_vnodeop_p; /* XXXSMP */ 1787 mutex_exit(&vp->v_interlock); 1788 /* 1789 * The vnode isn't clean, but still resides 1790 * on the mount list. Remove it. XXX This 1791 * is a bit dodgy. 1792 */ 1793 insmntque(vp, NULL); 1794 vrele(vp); 1795 } 1796 mutex_enter(&mntvnode_lock); 1797 continue; 1798 } 1799 #ifdef DEBUG 1800 if (busyprt) 1801 vprint("vflush: busy vnode", vp); 1802 #endif 1803 mutex_exit(&vp->v_interlock); 1804 busy++; 1805 } 1806 mutex_exit(&mntvnode_lock); 1807 vnfree(mvp); 1808 if (busy) 1809 return (EBUSY); 1810 return (0); 1811 } 1812 1813 /* 1814 * Disassociate the underlying file system from a vnode. 1815 * 1816 * Must be called with the interlock held, and will return with it held. 1817 */ 1818 void 1819 vclean(vnode_t *vp, int flags) 1820 { 1821 lwp_t *l = curlwp; 1822 bool recycle, active; 1823 int error; 1824 1825 KASSERT(mutex_owned(&vp->v_interlock)); 1826 KASSERT((vp->v_iflag & VI_MARKER) == 0); 1827 KASSERT(vp->v_usecount != 0); 1828 1829 /* If cleaning is already in progress wait until done and return. */ 1830 if (vp->v_iflag & VI_XLOCK) { 1831 vwait(vp, VI_XLOCK); 1832 return; 1833 } 1834 1835 /* If already clean, nothing to do. */ 1836 if ((vp->v_iflag & VI_CLEAN) != 0) { 1837 return; 1838 } 1839 1840 /* 1841 * Prevent the vnode from being recycled or brought into use 1842 * while we clean it out. 1843 */ 1844 vp->v_iflag |= VI_XLOCK; 1845 if (vp->v_iflag & VI_EXECMAP) { 1846 atomic_add_int(&uvmexp.execpages, -vp->v_uobj.uo_npages); 1847 atomic_add_int(&uvmexp.filepages, vp->v_uobj.uo_npages); 1848 } 1849 vp->v_iflag &= ~(VI_TEXT|VI_EXECMAP); 1850 active = (vp->v_usecount & VC_MASK) > 1; 1851 1852 /* XXXAD should not lock vnode under layer */ 1853 mutex_exit(&vp->v_interlock); 1854 VOP_LOCK(vp, LK_EXCLUSIVE); 1855 1856 /* 1857 * Clean out any cached data associated with the vnode. 1858 * If purging an active vnode, it must be closed and 1859 * deactivated before being reclaimed. Note that the 1860 * VOP_INACTIVE will unlock the vnode. 1861 */ 1862 if (flags & DOCLOSE) { 1863 error = vinvalbuf(vp, V_SAVE, NOCRED, l, 0, 0); 1864 if (error != 0) { 1865 /* XXX, fix vn_start_write's grab of mp and use that. */ 1866 1867 if (wapbl_vphaswapbl(vp)) 1868 WAPBL_DISCARD(wapbl_vptomp(vp)); 1869 error = vinvalbuf(vp, 0, NOCRED, l, 0, 0); 1870 } 1871 KASSERT(error == 0); 1872 KASSERT((vp->v_iflag & VI_ONWORKLST) == 0); 1873 if (active && (vp->v_type == VBLK || vp->v_type == VCHR)) { 1874 spec_node_revoke(vp); 1875 } 1876 } 1877 if (active) { 1878 VOP_INACTIVE(vp, &recycle); 1879 } else { 1880 /* 1881 * Any other processes trying to obtain this lock must first 1882 * wait for VI_XLOCK to clear, then call the new lock operation. 1883 */ 1884 VOP_UNLOCK(vp); 1885 } 1886 1887 /* Disassociate the underlying file system from the vnode. */ 1888 if (VOP_RECLAIM(vp)) { 1889 vpanic(vp, "vclean: cannot reclaim"); 1890 } 1891 1892 KASSERT(vp->v_uobj.uo_npages == 0); 1893 if (vp->v_type == VREG && vp->v_ractx != NULL) { 1894 uvm_ra_freectx(vp->v_ractx); 1895 vp->v_ractx = NULL; 1896 } 1897 cache_purge(vp); 1898 1899 /* Done with purge, notify sleepers of the grim news. */ 1900 mutex_enter(&vp->v_interlock); 1901 vp->v_op = dead_vnodeop_p; 1902 vp->v_tag = VT_NON; 1903 KNOTE(&vp->v_klist, NOTE_REVOKE); 1904 vp->v_iflag &= ~VI_XLOCK; 1905 vp->v_vflag &= ~VV_LOCKSWORK; 1906 if ((flags & DOCLOSE) != 0) { 1907 vp->v_iflag |= VI_CLEAN; 1908 } 1909 cv_broadcast(&vp->v_cv); 1910 1911 KASSERT((vp->v_iflag & VI_ONWORKLST) == 0); 1912 } 1913 1914 /* 1915 * Recycle an unused vnode to the front of the free list. 1916 * Release the passed interlock if the vnode will be recycled. 1917 */ 1918 int 1919 vrecycle(vnode_t *vp, kmutex_t *inter_lkp, struct lwp *l) 1920 { 1921 1922 KASSERT((vp->v_iflag & VI_MARKER) == 0); 1923 1924 mutex_enter(&vp->v_interlock); 1925 if (vp->v_usecount != 0) { 1926 mutex_exit(&vp->v_interlock); 1927 return (0); 1928 } 1929 if (inter_lkp) 1930 mutex_exit(inter_lkp); 1931 vremfree(vp); 1932 vp->v_usecount = 1; 1933 vclean(vp, DOCLOSE); 1934 vrelel(vp, 0); 1935 return (1); 1936 } 1937 1938 /* 1939 * Eliminate all activity associated with a vnode in preparation for 1940 * reuse. Drops a reference from the vnode. 1941 */ 1942 void 1943 vgone(vnode_t *vp) 1944 { 1945 1946 mutex_enter(&vp->v_interlock); 1947 vclean(vp, DOCLOSE); 1948 vrelel(vp, 0); 1949 } 1950 1951 /* 1952 * Lookup a vnode by device number and return it referenced. 1953 */ 1954 int 1955 vfinddev(dev_t dev, enum vtype type, vnode_t **vpp) 1956 { 1957 vnode_t *vp; 1958 1959 mutex_enter(&device_lock); 1960 for (vp = specfs_hash[SPECHASH(dev)]; vp; vp = vp->v_specnext) { 1961 if (dev == vp->v_rdev && type == vp->v_type) 1962 break; 1963 } 1964 if (vp == NULL) { 1965 mutex_exit(&device_lock); 1966 return 0; 1967 } 1968 mutex_enter(&vp->v_interlock); 1969 mutex_exit(&device_lock); 1970 if (vget(vp, 0) != 0) 1971 return 0; 1972 *vpp = vp; 1973 return 1; 1974 } 1975 1976 /* 1977 * Revoke all the vnodes corresponding to the specified minor number 1978 * range (endpoints inclusive) of the specified major. 1979 */ 1980 void 1981 vdevgone(int maj, int minl, int minh, enum vtype type) 1982 { 1983 vnode_t *vp, **vpp; 1984 dev_t dev; 1985 int mn; 1986 1987 vp = NULL; /* XXX gcc */ 1988 1989 mutex_enter(&device_lock); 1990 for (mn = minl; mn <= minh; mn++) { 1991 dev = makedev(maj, mn); 1992 vpp = &specfs_hash[SPECHASH(dev)]; 1993 for (vp = *vpp; vp != NULL;) { 1994 mutex_enter(&vp->v_interlock); 1995 if ((vp->v_iflag & VI_CLEAN) != 0 || 1996 dev != vp->v_rdev || type != vp->v_type) { 1997 mutex_exit(&vp->v_interlock); 1998 vp = vp->v_specnext; 1999 continue; 2000 } 2001 mutex_exit(&device_lock); 2002 if (vget(vp, 0) == 0) { 2003 VOP_REVOKE(vp, REVOKEALL); 2004 vrele(vp); 2005 } 2006 mutex_enter(&device_lock); 2007 vp = *vpp; 2008 } 2009 } 2010 mutex_exit(&device_lock); 2011 } 2012 2013 /* 2014 * Eliminate all activity associated with the requested vnode 2015 * and with all vnodes aliased to the requested vnode. 2016 */ 2017 void 2018 vrevoke(vnode_t *vp) 2019 { 2020 vnode_t *vq, **vpp; 2021 enum vtype type; 2022 dev_t dev; 2023 2024 KASSERT(vp->v_usecount > 0); 2025 2026 mutex_enter(&vp->v_interlock); 2027 if ((vp->v_iflag & VI_CLEAN) != 0) { 2028 mutex_exit(&vp->v_interlock); 2029 return; 2030 } else if (vp->v_type != VBLK && vp->v_type != VCHR) { 2031 atomic_inc_uint(&vp->v_usecount); 2032 vclean(vp, DOCLOSE); 2033 vrelel(vp, 0); 2034 return; 2035 } else { 2036 dev = vp->v_rdev; 2037 type = vp->v_type; 2038 mutex_exit(&vp->v_interlock); 2039 } 2040 2041 vpp = &specfs_hash[SPECHASH(dev)]; 2042 mutex_enter(&device_lock); 2043 for (vq = *vpp; vq != NULL;) { 2044 /* If clean or being cleaned, then ignore it. */ 2045 mutex_enter(&vq->v_interlock); 2046 if ((vq->v_iflag & (VI_CLEAN | VI_XLOCK)) != 0 || 2047 vq->v_rdev != dev || vq->v_type != type) { 2048 mutex_exit(&vq->v_interlock); 2049 vq = vq->v_specnext; 2050 continue; 2051 } 2052 mutex_exit(&device_lock); 2053 if (vq->v_usecount == 0) { 2054 vremfree(vq); 2055 vq->v_usecount = 1; 2056 } else { 2057 atomic_inc_uint(&vq->v_usecount); 2058 } 2059 vclean(vq, DOCLOSE); 2060 vrelel(vq, 0); 2061 mutex_enter(&device_lock); 2062 vq = *vpp; 2063 } 2064 mutex_exit(&device_lock); 2065 } 2066 2067 /* 2068 * sysctl helper routine to return list of supported fstypes 2069 */ 2070 int 2071 sysctl_vfs_generic_fstypes(SYSCTLFN_ARGS) 2072 { 2073 char bf[sizeof(((struct statvfs *)NULL)->f_fstypename)]; 2074 char *where = oldp; 2075 struct vfsops *v; 2076 size_t needed, left, slen; 2077 int error, first; 2078 2079 if (newp != NULL) 2080 return (EPERM); 2081 if (namelen != 0) 2082 return (EINVAL); 2083 2084 first = 1; 2085 error = 0; 2086 needed = 0; 2087 left = *oldlenp; 2088 2089 sysctl_unlock(); 2090 mutex_enter(&vfs_list_lock); 2091 LIST_FOREACH(v, &vfs_list, vfs_list) { 2092 if (where == NULL) 2093 needed += strlen(v->vfs_name) + 1; 2094 else { 2095 memset(bf, 0, sizeof(bf)); 2096 if (first) { 2097 strncpy(bf, v->vfs_name, sizeof(bf)); 2098 first = 0; 2099 } else { 2100 bf[0] = ' '; 2101 strncpy(bf + 1, v->vfs_name, sizeof(bf) - 1); 2102 } 2103 bf[sizeof(bf)-1] = '\0'; 2104 slen = strlen(bf); 2105 if (left < slen + 1) 2106 break; 2107 v->vfs_refcount++; 2108 mutex_exit(&vfs_list_lock); 2109 /* +1 to copy out the trailing NUL byte */ 2110 error = copyout(bf, where, slen + 1); 2111 mutex_enter(&vfs_list_lock); 2112 v->vfs_refcount--; 2113 if (error) 2114 break; 2115 where += slen; 2116 needed += slen; 2117 left -= slen; 2118 } 2119 } 2120 mutex_exit(&vfs_list_lock); 2121 sysctl_relock(); 2122 *oldlenp = needed; 2123 return (error); 2124 } 2125 2126 2127 int kinfo_vdebug = 1; 2128 int kinfo_vgetfailed; 2129 #define KINFO_VNODESLOP 10 2130 /* 2131 * Dump vnode list (via sysctl). 2132 * Copyout address of vnode followed by vnode. 2133 */ 2134 /* ARGSUSED */ 2135 int 2136 sysctl_kern_vnode(SYSCTLFN_ARGS) 2137 { 2138 char *where = oldp; 2139 size_t *sizep = oldlenp; 2140 struct mount *mp, *nmp; 2141 vnode_t *vp, *mvp, vbuf; 2142 char *bp = where; 2143 char *ewhere; 2144 int error; 2145 2146 if (namelen != 0) 2147 return (EOPNOTSUPP); 2148 if (newp != NULL) 2149 return (EPERM); 2150 2151 #define VPTRSZ sizeof(vnode_t *) 2152 #define VNODESZ sizeof(vnode_t) 2153 if (where == NULL) { 2154 *sizep = (numvnodes + KINFO_VNODESLOP) * (VPTRSZ + VNODESZ); 2155 return (0); 2156 } 2157 ewhere = where + *sizep; 2158 2159 sysctl_unlock(); 2160 mutex_enter(&mountlist_lock); 2161 for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist; 2162 mp = nmp) { 2163 if (vfs_busy(mp, &nmp)) { 2164 continue; 2165 } 2166 /* Allocate a marker vnode. */ 2167 mvp = vnalloc(mp); 2168 /* Should never fail for mp != NULL */ 2169 KASSERT(mvp != NULL); 2170 mutex_enter(&mntvnode_lock); 2171 for (vp = TAILQ_FIRST(&mp->mnt_vnodelist); vp; 2172 vp = vunmark(mvp)) { 2173 vmark(mvp, vp); 2174 /* 2175 * Check that the vp is still associated with 2176 * this filesystem. RACE: could have been 2177 * recycled onto the same filesystem. 2178 */ 2179 if (vp->v_mount != mp || vismarker(vp)) 2180 continue; 2181 if (bp + VPTRSZ + VNODESZ > ewhere) { 2182 (void)vunmark(mvp); 2183 mutex_exit(&mntvnode_lock); 2184 vnfree(mvp); 2185 vfs_unbusy(mp, false, NULL); 2186 sysctl_relock(); 2187 *sizep = bp - where; 2188 return (ENOMEM); 2189 } 2190 memcpy(&vbuf, vp, VNODESZ); 2191 mutex_exit(&mntvnode_lock); 2192 if ((error = copyout(&vp, bp, VPTRSZ)) || 2193 (error = copyout(&vbuf, bp + VPTRSZ, VNODESZ))) { 2194 mutex_enter(&mntvnode_lock); 2195 (void)vunmark(mvp); 2196 mutex_exit(&mntvnode_lock); 2197 vnfree(mvp); 2198 vfs_unbusy(mp, false, NULL); 2199 sysctl_relock(); 2200 return (error); 2201 } 2202 bp += VPTRSZ + VNODESZ; 2203 mutex_enter(&mntvnode_lock); 2204 } 2205 mutex_exit(&mntvnode_lock); 2206 vnfree(mvp); 2207 vfs_unbusy(mp, false, &nmp); 2208 } 2209 mutex_exit(&mountlist_lock); 2210 sysctl_relock(); 2211 2212 *sizep = bp - where; 2213 return (0); 2214 } 2215 2216 /* 2217 * Remove clean vnodes from a mountpoint's vnode list. 2218 */ 2219 void 2220 vfs_scrubvnlist(struct mount *mp) 2221 { 2222 vnode_t *vp, *nvp; 2223 2224 retry: 2225 mutex_enter(&mntvnode_lock); 2226 for (vp = TAILQ_FIRST(&mp->mnt_vnodelist); vp; vp = nvp) { 2227 nvp = TAILQ_NEXT(vp, v_mntvnodes); 2228 mutex_enter(&vp->v_interlock); 2229 if ((vp->v_iflag & VI_CLEAN) != 0) { 2230 TAILQ_REMOVE(&mp->mnt_vnodelist, vp, v_mntvnodes); 2231 vp->v_mount = NULL; 2232 mutex_exit(&mntvnode_lock); 2233 mutex_exit(&vp->v_interlock); 2234 vfs_destroy(mp); 2235 goto retry; 2236 } 2237 mutex_exit(&vp->v_interlock); 2238 } 2239 mutex_exit(&mntvnode_lock); 2240 } 2241 2242 /* 2243 * Check to see if a filesystem is mounted on a block device. 2244 */ 2245 int 2246 vfs_mountedon(vnode_t *vp) 2247 { 2248 vnode_t *vq; 2249 int error = 0; 2250 2251 if (vp->v_type != VBLK) 2252 return ENOTBLK; 2253 if (vp->v_specmountpoint != NULL) 2254 return (EBUSY); 2255 mutex_enter(&device_lock); 2256 for (vq = specfs_hash[SPECHASH(vp->v_rdev)]; vq != NULL; 2257 vq = vq->v_specnext) { 2258 if (vq->v_rdev != vp->v_rdev || vq->v_type != vp->v_type) 2259 continue; 2260 if (vq->v_specmountpoint != NULL) { 2261 error = EBUSY; 2262 break; 2263 } 2264 } 2265 mutex_exit(&device_lock); 2266 return (error); 2267 } 2268 2269 /* 2270 * Unmount all file systems. 2271 * We traverse the list in reverse order under the assumption that doing so 2272 * will avoid needing to worry about dependencies. 2273 */ 2274 bool 2275 vfs_unmountall(struct lwp *l) 2276 { 2277 2278 printf("unmounting file systems..."); 2279 return vfs_unmountall1(l, true, true); 2280 } 2281 2282 static void 2283 vfs_unmount_print(struct mount *mp, const char *pfx) 2284 { 2285 2286 aprint_verbose("%sunmounted %s on %s type %s\n", pfx, 2287 mp->mnt_stat.f_mntfromname, mp->mnt_stat.f_mntonname, 2288 mp->mnt_stat.f_fstypename); 2289 } 2290 2291 bool 2292 vfs_unmount_forceone(struct lwp *l) 2293 { 2294 struct mount *mp, *nmp; 2295 int error; 2296 2297 nmp = NULL; 2298 2299 CIRCLEQ_FOREACH_REVERSE(mp, &mountlist, mnt_list) { 2300 if (nmp == NULL || mp->mnt_gen > nmp->mnt_gen) { 2301 nmp = mp; 2302 } 2303 } 2304 if (nmp == NULL) { 2305 return false; 2306 } 2307 2308 #ifdef DEBUG 2309 printf("\nforcefully unmounting %s (%s)...", 2310 nmp->mnt_stat.f_mntonname, nmp->mnt_stat.f_mntfromname); 2311 #endif 2312 atomic_inc_uint(&nmp->mnt_refcnt); 2313 if ((error = dounmount(nmp, MNT_FORCE, l)) == 0) { 2314 vfs_unmount_print(nmp, "forcefully "); 2315 return true; 2316 } else { 2317 vfs_destroy(nmp); 2318 } 2319 2320 #ifdef DEBUG 2321 printf("forceful unmount of %s failed with error %d\n", 2322 nmp->mnt_stat.f_mntonname, error); 2323 #endif 2324 2325 return false; 2326 } 2327 2328 bool 2329 vfs_unmountall1(struct lwp *l, bool force, bool verbose) 2330 { 2331 struct mount *mp, *nmp; 2332 bool any_error = false, progress = false; 2333 int error; 2334 2335 for (mp = CIRCLEQ_LAST(&mountlist); 2336 mp != (void *)&mountlist; 2337 mp = nmp) { 2338 nmp = CIRCLEQ_PREV(mp, mnt_list); 2339 #ifdef DEBUG 2340 printf("\nunmounting %p %s (%s)...", 2341 (void *)mp, mp->mnt_stat.f_mntonname, 2342 mp->mnt_stat.f_mntfromname); 2343 #endif 2344 atomic_inc_uint(&mp->mnt_refcnt); 2345 if ((error = dounmount(mp, force ? MNT_FORCE : 0, l)) == 0) { 2346 vfs_unmount_print(mp, ""); 2347 progress = true; 2348 } else { 2349 vfs_destroy(mp); 2350 if (verbose) { 2351 printf("unmount of %s failed with error %d\n", 2352 mp->mnt_stat.f_mntonname, error); 2353 } 2354 any_error = true; 2355 } 2356 } 2357 if (verbose) { 2358 printf(" done\n"); 2359 } 2360 if (any_error && verbose) { 2361 printf("WARNING: some file systems would not unmount\n"); 2362 } 2363 return progress; 2364 } 2365 2366 /* 2367 * Sync and unmount file systems before shutting down. 2368 */ 2369 void 2370 vfs_shutdown(void) 2371 { 2372 2373 /* XXX we're certainly not running in lwp0's context! */ 2374 vfs_shutdown1(curlwp); 2375 } 2376 2377 void 2378 vfs_sync_all(struct lwp *l) 2379 { 2380 printf("syncing disks... "); 2381 2382 /* remove user processes from run queue */ 2383 suspendsched(); 2384 (void) spl0(); 2385 2386 /* avoid coming back this way again if we panic. */ 2387 doing_shutdown = 1; 2388 2389 sys_sync(l, NULL, NULL); 2390 2391 /* Wait for sync to finish. */ 2392 if (buf_syncwait() != 0) { 2393 #if defined(DDB) && defined(DEBUG_HALT_BUSY) 2394 Debugger(); 2395 #endif 2396 printf("giving up\n"); 2397 return; 2398 } else 2399 printf("done\n"); 2400 } 2401 2402 static void 2403 vfs_shutdown1(struct lwp *l) 2404 { 2405 2406 vfs_sync_all(l); 2407 2408 /* 2409 * If we've panic'd, don't make the situation potentially 2410 * worse by unmounting the file systems. 2411 */ 2412 if (panicstr != NULL) 2413 return; 2414 2415 /* Release inodes held by texts before update. */ 2416 #ifdef notdef 2417 vnshutdown(); 2418 #endif 2419 /* Unmount file systems. */ 2420 vfs_unmountall(l); 2421 } 2422 2423 /* 2424 * Print a list of supported file system types (used by vfs_mountroot) 2425 */ 2426 static void 2427 vfs_print_fstypes(void) 2428 { 2429 struct vfsops *v; 2430 int cnt = 0; 2431 2432 mutex_enter(&vfs_list_lock); 2433 LIST_FOREACH(v, &vfs_list, vfs_list) 2434 ++cnt; 2435 mutex_exit(&vfs_list_lock); 2436 2437 if (cnt == 0) { 2438 printf("WARNING: No file system modules have been loaded.\n"); 2439 return; 2440 } 2441 2442 printf("Supported file systems:"); 2443 mutex_enter(&vfs_list_lock); 2444 LIST_FOREACH(v, &vfs_list, vfs_list) { 2445 printf(" %s", v->vfs_name); 2446 } 2447 mutex_exit(&vfs_list_lock); 2448 printf("\n"); 2449 } 2450 2451 /* 2452 * Mount the root file system. If the operator didn't specify a 2453 * file system to use, try all possible file systems until one 2454 * succeeds. 2455 */ 2456 int 2457 vfs_mountroot(void) 2458 { 2459 struct vfsops *v; 2460 int error = ENODEV; 2461 2462 if (root_device == NULL) 2463 panic("vfs_mountroot: root device unknown"); 2464 2465 switch (device_class(root_device)) { 2466 case DV_IFNET: 2467 if (rootdev != NODEV) 2468 panic("vfs_mountroot: rootdev set for DV_IFNET " 2469 "(0x%llx -> %llu,%llu)", 2470 (unsigned long long)rootdev, 2471 (unsigned long long)major(rootdev), 2472 (unsigned long long)minor(rootdev)); 2473 break; 2474 2475 case DV_DISK: 2476 if (rootdev == NODEV) 2477 panic("vfs_mountroot: rootdev not set for DV_DISK"); 2478 if (bdevvp(rootdev, &rootvp)) 2479 panic("vfs_mountroot: can't get vnode for rootdev"); 2480 error = VOP_OPEN(rootvp, FREAD, FSCRED); 2481 if (error) { 2482 printf("vfs_mountroot: can't open root device\n"); 2483 return (error); 2484 } 2485 break; 2486 2487 case DV_VIRTUAL: 2488 break; 2489 2490 default: 2491 printf("%s: inappropriate for root file system\n", 2492 device_xname(root_device)); 2493 return (ENODEV); 2494 } 2495 2496 /* 2497 * If user specified a root fs type, use it. Make sure the 2498 * specified type exists and has a mount_root() 2499 */ 2500 if (strcmp(rootfstype, ROOT_FSTYPE_ANY) != 0) { 2501 v = vfs_getopsbyname(rootfstype); 2502 error = EFTYPE; 2503 if (v != NULL) { 2504 if (v->vfs_mountroot != NULL) { 2505 error = (v->vfs_mountroot)(); 2506 } 2507 v->vfs_refcount--; 2508 } 2509 goto done; 2510 } 2511 2512 /* 2513 * Try each file system currently configured into the kernel. 2514 */ 2515 mutex_enter(&vfs_list_lock); 2516 LIST_FOREACH(v, &vfs_list, vfs_list) { 2517 if (v->vfs_mountroot == NULL) 2518 continue; 2519 #ifdef DEBUG 2520 aprint_normal("mountroot: trying %s...\n", v->vfs_name); 2521 #endif 2522 v->vfs_refcount++; 2523 mutex_exit(&vfs_list_lock); 2524 error = (*v->vfs_mountroot)(); 2525 mutex_enter(&vfs_list_lock); 2526 v->vfs_refcount--; 2527 if (!error) { 2528 aprint_normal("root file system type: %s\n", 2529 v->vfs_name); 2530 break; 2531 } 2532 } 2533 mutex_exit(&vfs_list_lock); 2534 2535 if (v == NULL) { 2536 vfs_print_fstypes(); 2537 printf("no file system for %s", device_xname(root_device)); 2538 if (device_class(root_device) == DV_DISK) 2539 printf(" (dev 0x%llx)", (unsigned long long)rootdev); 2540 printf("\n"); 2541 error = EFTYPE; 2542 } 2543 2544 done: 2545 if (error && device_class(root_device) == DV_DISK) { 2546 VOP_CLOSE(rootvp, FREAD, FSCRED); 2547 vrele(rootvp); 2548 } 2549 if (error == 0) { 2550 extern struct cwdinfo cwdi0; 2551 2552 CIRCLEQ_FIRST(&mountlist)->mnt_flag |= MNT_ROOTFS; 2553 CIRCLEQ_FIRST(&mountlist)->mnt_op->vfs_refcount++; 2554 2555 /* 2556 * Get the vnode for '/'. Set cwdi0.cwdi_cdir to 2557 * reference it. 2558 */ 2559 error = VFS_ROOT(CIRCLEQ_FIRST(&mountlist), &rootvnode); 2560 if (error) 2561 panic("cannot find root vnode, error=%d", error); 2562 cwdi0.cwdi_cdir = rootvnode; 2563 vref(cwdi0.cwdi_cdir); 2564 VOP_UNLOCK(rootvnode); 2565 cwdi0.cwdi_rdir = NULL; 2566 2567 /* 2568 * Now that root is mounted, we can fixup initproc's CWD 2569 * info. All other processes are kthreads, which merely 2570 * share proc0's CWD info. 2571 */ 2572 initproc->p_cwdi->cwdi_cdir = rootvnode; 2573 vref(initproc->p_cwdi->cwdi_cdir); 2574 initproc->p_cwdi->cwdi_rdir = NULL; 2575 /* 2576 * Enable loading of modules from the filesystem 2577 */ 2578 module_load_vfs_init(); 2579 2580 } 2581 return (error); 2582 } 2583 2584 /* 2585 * Get a new unique fsid 2586 */ 2587 void 2588 vfs_getnewfsid(struct mount *mp) 2589 { 2590 static u_short xxxfs_mntid; 2591 fsid_t tfsid; 2592 int mtype; 2593 2594 mutex_enter(&mntid_lock); 2595 mtype = makefstype(mp->mnt_op->vfs_name); 2596 mp->mnt_stat.f_fsidx.__fsid_val[0] = makedev(mtype, 0); 2597 mp->mnt_stat.f_fsidx.__fsid_val[1] = mtype; 2598 mp->mnt_stat.f_fsid = mp->mnt_stat.f_fsidx.__fsid_val[0]; 2599 if (xxxfs_mntid == 0) 2600 ++xxxfs_mntid; 2601 tfsid.__fsid_val[0] = makedev(mtype & 0xff, xxxfs_mntid); 2602 tfsid.__fsid_val[1] = mtype; 2603 if (!CIRCLEQ_EMPTY(&mountlist)) { 2604 while (vfs_getvfs(&tfsid)) { 2605 tfsid.__fsid_val[0]++; 2606 xxxfs_mntid++; 2607 } 2608 } 2609 mp->mnt_stat.f_fsidx.__fsid_val[0] = tfsid.__fsid_val[0]; 2610 mp->mnt_stat.f_fsid = mp->mnt_stat.f_fsidx.__fsid_val[0]; 2611 mutex_exit(&mntid_lock); 2612 } 2613 2614 /* 2615 * Make a 'unique' number from a mount type name. 2616 */ 2617 long 2618 makefstype(const char *type) 2619 { 2620 long rv; 2621 2622 for (rv = 0; *type; type++) { 2623 rv <<= 2; 2624 rv ^= *type; 2625 } 2626 return rv; 2627 } 2628 2629 /* 2630 * Set vnode attributes to VNOVAL 2631 */ 2632 void 2633 vattr_null(struct vattr *vap) 2634 { 2635 2636 memset(vap, 0, sizeof(*vap)); 2637 2638 vap->va_type = VNON; 2639 2640 /* 2641 * Assign individually so that it is safe even if size and 2642 * sign of each member are varied. 2643 */ 2644 vap->va_mode = VNOVAL; 2645 vap->va_nlink = VNOVAL; 2646 vap->va_uid = VNOVAL; 2647 vap->va_gid = VNOVAL; 2648 vap->va_fsid = VNOVAL; 2649 vap->va_fileid = VNOVAL; 2650 vap->va_size = VNOVAL; 2651 vap->va_blocksize = VNOVAL; 2652 vap->va_atime.tv_sec = 2653 vap->va_mtime.tv_sec = 2654 vap->va_ctime.tv_sec = 2655 vap->va_birthtime.tv_sec = VNOVAL; 2656 vap->va_atime.tv_nsec = 2657 vap->va_mtime.tv_nsec = 2658 vap->va_ctime.tv_nsec = 2659 vap->va_birthtime.tv_nsec = VNOVAL; 2660 vap->va_gen = VNOVAL; 2661 vap->va_flags = VNOVAL; 2662 vap->va_rdev = VNOVAL; 2663 vap->va_bytes = VNOVAL; 2664 } 2665 2666 #define ARRAY_SIZE(arr) (sizeof(arr) / sizeof(arr[0])) 2667 #define ARRAY_PRINT(idx, arr) \ 2668 ((unsigned int)(idx) < ARRAY_SIZE(arr) ? (arr)[(idx)] : "UNKNOWN") 2669 2670 const char * const vnode_tags[] = { VNODE_TAGS }; 2671 const char * const vnode_types[] = { VNODE_TYPES }; 2672 const char vnode_flagbits[] = VNODE_FLAGBITS; 2673 2674 /* 2675 * Print out a description of a vnode. 2676 */ 2677 void 2678 vprint(const char *label, struct vnode *vp) 2679 { 2680 char bf[96]; 2681 int flag; 2682 2683 flag = vp->v_iflag | vp->v_vflag | vp->v_uflag; 2684 snprintb(bf, sizeof(bf), vnode_flagbits, flag); 2685 2686 if (label != NULL) 2687 printf("%s: ", label); 2688 printf("vnode @ %p, flags (%s)\n\ttag %s(%d), type %s(%d), " 2689 "usecount %d, writecount %d, holdcount %d\n" 2690 "\tfreelisthd %p, mount %p, data %p lock %p\n", 2691 vp, bf, ARRAY_PRINT(vp->v_tag, vnode_tags), vp->v_tag, 2692 ARRAY_PRINT(vp->v_type, vnode_types), vp->v_type, 2693 vp->v_usecount, vp->v_writecount, vp->v_holdcnt, 2694 vp->v_freelisthd, vp->v_mount, vp->v_data, &vp->v_lock); 2695 if (vp->v_data != NULL) { 2696 printf("\t"); 2697 VOP_PRINT(vp); 2698 } 2699 } 2700 2701 #ifdef DEBUG 2702 /* 2703 * List all of the locked vnodes in the system. 2704 * Called when debugging the kernel. 2705 */ 2706 void 2707 printlockedvnodes(void) 2708 { 2709 struct mount *mp, *nmp; 2710 struct vnode *vp; 2711 2712 printf("Locked vnodes\n"); 2713 mutex_enter(&mountlist_lock); 2714 for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist; 2715 mp = nmp) { 2716 if (vfs_busy(mp, &nmp)) { 2717 continue; 2718 } 2719 TAILQ_FOREACH(vp, &mp->mnt_vnodelist, v_mntvnodes) { 2720 if (VOP_ISLOCKED(vp)) 2721 vprint(NULL, vp); 2722 } 2723 mutex_enter(&mountlist_lock); 2724 vfs_unbusy(mp, false, &nmp); 2725 } 2726 mutex_exit(&mountlist_lock); 2727 } 2728 #endif 2729 2730 /* Deprecated. Kept for KPI compatibility. */ 2731 int 2732 vaccess(enum vtype type, mode_t file_mode, uid_t uid, gid_t gid, 2733 mode_t acc_mode, kauth_cred_t cred) 2734 { 2735 2736 #ifdef DIAGNOSTIC 2737 printf("vaccess: deprecated interface used.\n"); 2738 #endif /* DIAGNOSTIC */ 2739 2740 return genfs_can_access(type, file_mode, uid, gid, acc_mode, cred); 2741 } 2742 2743 /* 2744 * Given a file system name, look up the vfsops for that 2745 * file system, or return NULL if file system isn't present 2746 * in the kernel. 2747 */ 2748 struct vfsops * 2749 vfs_getopsbyname(const char *name) 2750 { 2751 struct vfsops *v; 2752 2753 mutex_enter(&vfs_list_lock); 2754 LIST_FOREACH(v, &vfs_list, vfs_list) { 2755 if (strcmp(v->vfs_name, name) == 0) 2756 break; 2757 } 2758 if (v != NULL) 2759 v->vfs_refcount++; 2760 mutex_exit(&vfs_list_lock); 2761 2762 return (v); 2763 } 2764 2765 void 2766 copy_statvfs_info(struct statvfs *sbp, const struct mount *mp) 2767 { 2768 const struct statvfs *mbp; 2769 2770 if (sbp == (mbp = &mp->mnt_stat)) 2771 return; 2772 2773 (void)memcpy(&sbp->f_fsidx, &mbp->f_fsidx, sizeof(sbp->f_fsidx)); 2774 sbp->f_fsid = mbp->f_fsid; 2775 sbp->f_owner = mbp->f_owner; 2776 sbp->f_flag = mbp->f_flag; 2777 sbp->f_syncwrites = mbp->f_syncwrites; 2778 sbp->f_asyncwrites = mbp->f_asyncwrites; 2779 sbp->f_syncreads = mbp->f_syncreads; 2780 sbp->f_asyncreads = mbp->f_asyncreads; 2781 (void)memcpy(sbp->f_spare, mbp->f_spare, sizeof(mbp->f_spare)); 2782 (void)memcpy(sbp->f_fstypename, mbp->f_fstypename, 2783 sizeof(sbp->f_fstypename)); 2784 (void)memcpy(sbp->f_mntonname, mbp->f_mntonname, 2785 sizeof(sbp->f_mntonname)); 2786 (void)memcpy(sbp->f_mntfromname, mp->mnt_stat.f_mntfromname, 2787 sizeof(sbp->f_mntfromname)); 2788 sbp->f_namemax = mbp->f_namemax; 2789 } 2790 2791 int 2792 set_statvfs_info(const char *onp, int ukon, const char *fromp, int ukfrom, 2793 const char *vfsname, struct mount *mp, struct lwp *l) 2794 { 2795 int error; 2796 size_t size; 2797 struct statvfs *sfs = &mp->mnt_stat; 2798 int (*fun)(const void *, void *, size_t, size_t *); 2799 2800 (void)strlcpy(mp->mnt_stat.f_fstypename, vfsname, 2801 sizeof(mp->mnt_stat.f_fstypename)); 2802 2803 if (onp) { 2804 struct cwdinfo *cwdi = l->l_proc->p_cwdi; 2805 fun = (ukon == UIO_SYSSPACE) ? copystr : copyinstr; 2806 if (cwdi->cwdi_rdir != NULL) { 2807 size_t len; 2808 char *bp; 2809 char *path = PNBUF_GET(); 2810 2811 bp = path + MAXPATHLEN; 2812 *--bp = '\0'; 2813 rw_enter(&cwdi->cwdi_lock, RW_READER); 2814 error = getcwd_common(cwdi->cwdi_rdir, rootvnode, &bp, 2815 path, MAXPATHLEN / 2, 0, l); 2816 rw_exit(&cwdi->cwdi_lock); 2817 if (error) { 2818 PNBUF_PUT(path); 2819 return error; 2820 } 2821 2822 len = strlen(bp); 2823 if (len > sizeof(sfs->f_mntonname) - 1) 2824 len = sizeof(sfs->f_mntonname) - 1; 2825 (void)strncpy(sfs->f_mntonname, bp, len); 2826 PNBUF_PUT(path); 2827 2828 if (len < sizeof(sfs->f_mntonname) - 1) { 2829 error = (*fun)(onp, &sfs->f_mntonname[len], 2830 sizeof(sfs->f_mntonname) - len - 1, &size); 2831 if (error) 2832 return error; 2833 size += len; 2834 } else { 2835 size = len; 2836 } 2837 } else { 2838 error = (*fun)(onp, &sfs->f_mntonname, 2839 sizeof(sfs->f_mntonname) - 1, &size); 2840 if (error) 2841 return error; 2842 } 2843 (void)memset(sfs->f_mntonname + size, 0, 2844 sizeof(sfs->f_mntonname) - size); 2845 } 2846 2847 if (fromp) { 2848 fun = (ukfrom == UIO_SYSSPACE) ? copystr : copyinstr; 2849 error = (*fun)(fromp, sfs->f_mntfromname, 2850 sizeof(sfs->f_mntfromname) - 1, &size); 2851 if (error) 2852 return error; 2853 (void)memset(sfs->f_mntfromname + size, 0, 2854 sizeof(sfs->f_mntfromname) - size); 2855 } 2856 return 0; 2857 } 2858 2859 void 2860 vfs_timestamp(struct timespec *ts) 2861 { 2862 2863 nanotime(ts); 2864 } 2865 2866 time_t rootfstime; /* recorded root fs time, if known */ 2867 void 2868 setrootfstime(time_t t) 2869 { 2870 rootfstime = t; 2871 } 2872 2873 static const uint8_t vttodt_tab[9] = { 2874 DT_UNKNOWN, /* VNON */ 2875 DT_REG, /* VREG */ 2876 DT_DIR, /* VDIR */ 2877 DT_BLK, /* VBLK */ 2878 DT_CHR, /* VCHR */ 2879 DT_LNK, /* VLNK */ 2880 DT_SOCK, /* VSUCK */ 2881 DT_FIFO, /* VFIFO */ 2882 DT_UNKNOWN /* VBAD */ 2883 }; 2884 2885 uint8_t 2886 vtype2dt(enum vtype vt) 2887 { 2888 2889 CTASSERT(VBAD == __arraycount(vttodt_tab) - 1); 2890 return vttodt_tab[vt]; 2891 } 2892 2893 /* 2894 * mount_specific_key_create -- 2895 * Create a key for subsystem mount-specific data. 2896 */ 2897 int 2898 mount_specific_key_create(specificdata_key_t *keyp, specificdata_dtor_t dtor) 2899 { 2900 2901 return (specificdata_key_create(mount_specificdata_domain, keyp, dtor)); 2902 } 2903 2904 /* 2905 * mount_specific_key_delete -- 2906 * Delete a key for subsystem mount-specific data. 2907 */ 2908 void 2909 mount_specific_key_delete(specificdata_key_t key) 2910 { 2911 2912 specificdata_key_delete(mount_specificdata_domain, key); 2913 } 2914 2915 /* 2916 * mount_initspecific -- 2917 * Initialize a mount's specificdata container. 2918 */ 2919 void 2920 mount_initspecific(struct mount *mp) 2921 { 2922 int error; 2923 2924 error = specificdata_init(mount_specificdata_domain, 2925 &mp->mnt_specdataref); 2926 KASSERT(error == 0); 2927 } 2928 2929 /* 2930 * mount_finispecific -- 2931 * Finalize a mount's specificdata container. 2932 */ 2933 void 2934 mount_finispecific(struct mount *mp) 2935 { 2936 2937 specificdata_fini(mount_specificdata_domain, &mp->mnt_specdataref); 2938 } 2939 2940 /* 2941 * mount_getspecific -- 2942 * Return mount-specific data corresponding to the specified key. 2943 */ 2944 void * 2945 mount_getspecific(struct mount *mp, specificdata_key_t key) 2946 { 2947 2948 return (specificdata_getspecific(mount_specificdata_domain, 2949 &mp->mnt_specdataref, key)); 2950 } 2951 2952 /* 2953 * mount_setspecific -- 2954 * Set mount-specific data corresponding to the specified key. 2955 */ 2956 void 2957 mount_setspecific(struct mount *mp, specificdata_key_t key, void *data) 2958 { 2959 2960 specificdata_setspecific(mount_specificdata_domain, 2961 &mp->mnt_specdataref, key, data); 2962 } 2963 2964 int 2965 VFS_MOUNT(struct mount *mp, const char *a, void *b, size_t *c) 2966 { 2967 int error; 2968 2969 KERNEL_LOCK(1, NULL); 2970 error = (*(mp->mnt_op->vfs_mount))(mp, a, b, c); 2971 KERNEL_UNLOCK_ONE(NULL); 2972 2973 return error; 2974 } 2975 2976 int 2977 VFS_START(struct mount *mp, int a) 2978 { 2979 int error; 2980 2981 if ((mp->mnt_iflag & IMNT_MPSAFE) == 0) { 2982 KERNEL_LOCK(1, NULL); 2983 } 2984 error = (*(mp->mnt_op->vfs_start))(mp, a); 2985 if ((mp->mnt_iflag & IMNT_MPSAFE) == 0) { 2986 KERNEL_UNLOCK_ONE(NULL); 2987 } 2988 2989 return error; 2990 } 2991 2992 int 2993 VFS_UNMOUNT(struct mount *mp, int a) 2994 { 2995 int error; 2996 2997 KERNEL_LOCK(1, NULL); 2998 error = (*(mp->mnt_op->vfs_unmount))(mp, a); 2999 KERNEL_UNLOCK_ONE(NULL); 3000 3001 return error; 3002 } 3003 3004 int 3005 VFS_ROOT(struct mount *mp, struct vnode **a) 3006 { 3007 int error; 3008 3009 if ((mp->mnt_iflag & IMNT_MPSAFE) == 0) { 3010 KERNEL_LOCK(1, NULL); 3011 } 3012 error = (*(mp->mnt_op->vfs_root))(mp, a); 3013 if ((mp->mnt_iflag & IMNT_MPSAFE) == 0) { 3014 KERNEL_UNLOCK_ONE(NULL); 3015 } 3016 3017 return error; 3018 } 3019 3020 int 3021 VFS_QUOTACTL(struct mount *mp, int a, uid_t b, void *c) 3022 { 3023 int error; 3024 3025 if ((mp->mnt_iflag & IMNT_MPSAFE) == 0) { 3026 KERNEL_LOCK(1, NULL); 3027 } 3028 error = (*(mp->mnt_op->vfs_quotactl))(mp, a, b, c); 3029 if ((mp->mnt_iflag & IMNT_MPSAFE) == 0) { 3030 KERNEL_UNLOCK_ONE(NULL); 3031 } 3032 3033 return error; 3034 } 3035 3036 int 3037 VFS_STATVFS(struct mount *mp, struct statvfs *a) 3038 { 3039 int error; 3040 3041 if ((mp->mnt_iflag & IMNT_MPSAFE) == 0) { 3042 KERNEL_LOCK(1, NULL); 3043 } 3044 error = (*(mp->mnt_op->vfs_statvfs))(mp, a); 3045 if ((mp->mnt_iflag & IMNT_MPSAFE) == 0) { 3046 KERNEL_UNLOCK_ONE(NULL); 3047 } 3048 3049 return error; 3050 } 3051 3052 int 3053 VFS_SYNC(struct mount *mp, int a, struct kauth_cred *b) 3054 { 3055 int error; 3056 3057 if ((mp->mnt_iflag & IMNT_MPSAFE) == 0) { 3058 KERNEL_LOCK(1, NULL); 3059 } 3060 error = (*(mp->mnt_op->vfs_sync))(mp, a, b); 3061 if ((mp->mnt_iflag & IMNT_MPSAFE) == 0) { 3062 KERNEL_UNLOCK_ONE(NULL); 3063 } 3064 3065 return error; 3066 } 3067 3068 int 3069 VFS_FHTOVP(struct mount *mp, struct fid *a, struct vnode **b) 3070 { 3071 int error; 3072 3073 if ((mp->mnt_iflag & IMNT_MPSAFE) == 0) { 3074 KERNEL_LOCK(1, NULL); 3075 } 3076 error = (*(mp->mnt_op->vfs_fhtovp))(mp, a, b); 3077 if ((mp->mnt_iflag & IMNT_MPSAFE) == 0) { 3078 KERNEL_UNLOCK_ONE(NULL); 3079 } 3080 3081 return error; 3082 } 3083 3084 int 3085 VFS_VPTOFH(struct vnode *vp, struct fid *a, size_t *b) 3086 { 3087 int error; 3088 3089 if ((vp->v_vflag & VV_MPSAFE) == 0) { 3090 KERNEL_LOCK(1, NULL); 3091 } 3092 error = (*(vp->v_mount->mnt_op->vfs_vptofh))(vp, a, b); 3093 if ((vp->v_vflag & VV_MPSAFE) == 0) { 3094 KERNEL_UNLOCK_ONE(NULL); 3095 } 3096 3097 return error; 3098 } 3099 3100 int 3101 VFS_SNAPSHOT(struct mount *mp, struct vnode *a, struct timespec *b) 3102 { 3103 int error; 3104 3105 if ((mp->mnt_iflag & IMNT_MPSAFE) == 0) { 3106 KERNEL_LOCK(1, NULL); 3107 } 3108 error = (*(mp->mnt_op->vfs_snapshot))(mp, a, b); 3109 if ((mp->mnt_iflag & IMNT_MPSAFE) == 0) { 3110 KERNEL_UNLOCK_ONE(NULL); 3111 } 3112 3113 return error; 3114 } 3115 3116 int 3117 VFS_EXTATTRCTL(struct mount *mp, int a, struct vnode *b, int c, const char *d) 3118 { 3119 int error; 3120 3121 KERNEL_LOCK(1, NULL); /* XXXSMP check ffs */ 3122 error = (*(mp->mnt_op->vfs_extattrctl))(mp, a, b, c, d); 3123 KERNEL_UNLOCK_ONE(NULL); /* XXX */ 3124 3125 return error; 3126 } 3127 3128 int 3129 VFS_SUSPENDCTL(struct mount *mp, int a) 3130 { 3131 int error; 3132 3133 if ((mp->mnt_iflag & IMNT_MPSAFE) == 0) { 3134 KERNEL_LOCK(1, NULL); 3135 } 3136 error = (*(mp->mnt_op->vfs_suspendctl))(mp, a); 3137 if ((mp->mnt_iflag & IMNT_MPSAFE) == 0) { 3138 KERNEL_UNLOCK_ONE(NULL); 3139 } 3140 3141 return error; 3142 } 3143 3144 #if defined(DDB) || defined(DEBUGPRINT) 3145 static const char buf_flagbits[] = BUF_FLAGBITS; 3146 3147 void 3148 vfs_buf_print(struct buf *bp, int full, void (*pr)(const char *, ...)) 3149 { 3150 char bf[1024]; 3151 3152 (*pr)(" vp %p lblkno 0x%"PRIx64" blkno 0x%"PRIx64" rawblkno 0x%" 3153 PRIx64 " dev 0x%x\n", 3154 bp->b_vp, bp->b_lblkno, bp->b_blkno, bp->b_rawblkno, bp->b_dev); 3155 3156 snprintb(bf, sizeof(bf), 3157 buf_flagbits, bp->b_flags | bp->b_oflags | bp->b_cflags); 3158 (*pr)(" error %d flags 0x%s\n", bp->b_error, bf); 3159 3160 (*pr)(" bufsize 0x%lx bcount 0x%lx resid 0x%lx\n", 3161 bp->b_bufsize, bp->b_bcount, bp->b_resid); 3162 (*pr)(" data %p saveaddr %p\n", 3163 bp->b_data, bp->b_saveaddr); 3164 (*pr)(" iodone %p objlock %p\n", bp->b_iodone, bp->b_objlock); 3165 } 3166 3167 3168 void 3169 vfs_vnode_print(struct vnode *vp, int full, void (*pr)(const char *, ...)) 3170 { 3171 char bf[256]; 3172 3173 uvm_object_printit(&vp->v_uobj, full, pr); 3174 snprintb(bf, sizeof(bf), 3175 vnode_flagbits, vp->v_iflag | vp->v_vflag | vp->v_uflag); 3176 (*pr)("\nVNODE flags %s\n", bf); 3177 (*pr)("mp %p numoutput %d size 0x%llx writesize 0x%llx\n", 3178 vp->v_mount, vp->v_numoutput, vp->v_size, vp->v_writesize); 3179 3180 (*pr)("data %p writecount %ld holdcnt %ld\n", 3181 vp->v_data, vp->v_writecount, vp->v_holdcnt); 3182 3183 (*pr)("tag %s(%d) type %s(%d) mount %p typedata %p\n", 3184 ARRAY_PRINT(vp->v_tag, vnode_tags), vp->v_tag, 3185 ARRAY_PRINT(vp->v_type, vnode_types), vp->v_type, 3186 vp->v_mount, vp->v_mountedhere); 3187 3188 (*pr)("v_lock %p\n", &vp->v_lock); 3189 3190 if (full) { 3191 struct buf *bp; 3192 3193 (*pr)("clean bufs:\n"); 3194 LIST_FOREACH(bp, &vp->v_cleanblkhd, b_vnbufs) { 3195 (*pr)(" bp %p\n", bp); 3196 vfs_buf_print(bp, full, pr); 3197 } 3198 3199 (*pr)("dirty bufs:\n"); 3200 LIST_FOREACH(bp, &vp->v_dirtyblkhd, b_vnbufs) { 3201 (*pr)(" bp %p\n", bp); 3202 vfs_buf_print(bp, full, pr); 3203 } 3204 } 3205 } 3206 3207 void 3208 vfs_mount_print(struct mount *mp, int full, void (*pr)(const char *, ...)) 3209 { 3210 char sbuf[256]; 3211 3212 (*pr)("vnodecovered = %p syncer = %p data = %p\n", 3213 mp->mnt_vnodecovered,mp->mnt_syncer,mp->mnt_data); 3214 3215 (*pr)("fs_bshift %d dev_bshift = %d\n", 3216 mp->mnt_fs_bshift,mp->mnt_dev_bshift); 3217 3218 snprintb(sbuf, sizeof(sbuf), __MNT_FLAG_BITS, mp->mnt_flag); 3219 (*pr)("flag = %s\n", sbuf); 3220 3221 snprintb(sbuf, sizeof(sbuf), __IMNT_FLAG_BITS, mp->mnt_iflag); 3222 (*pr)("iflag = %s\n", sbuf); 3223 3224 (*pr)("refcnt = %d unmounting @ %p updating @ %p\n", mp->mnt_refcnt, 3225 &mp->mnt_unmounting, &mp->mnt_updating); 3226 3227 (*pr)("statvfs cache:\n"); 3228 (*pr)("\tbsize = %lu\n",mp->mnt_stat.f_bsize); 3229 (*pr)("\tfrsize = %lu\n",mp->mnt_stat.f_frsize); 3230 (*pr)("\tiosize = %lu\n",mp->mnt_stat.f_iosize); 3231 3232 (*pr)("\tblocks = %"PRIu64"\n",mp->mnt_stat.f_blocks); 3233 (*pr)("\tbfree = %"PRIu64"\n",mp->mnt_stat.f_bfree); 3234 (*pr)("\tbavail = %"PRIu64"\n",mp->mnt_stat.f_bavail); 3235 (*pr)("\tbresvd = %"PRIu64"\n",mp->mnt_stat.f_bresvd); 3236 3237 (*pr)("\tfiles = %"PRIu64"\n",mp->mnt_stat.f_files); 3238 (*pr)("\tffree = %"PRIu64"\n",mp->mnt_stat.f_ffree); 3239 (*pr)("\tfavail = %"PRIu64"\n",mp->mnt_stat.f_favail); 3240 (*pr)("\tfresvd = %"PRIu64"\n",mp->mnt_stat.f_fresvd); 3241 3242 (*pr)("\tf_fsidx = { 0x%"PRIx32", 0x%"PRIx32" }\n", 3243 mp->mnt_stat.f_fsidx.__fsid_val[0], 3244 mp->mnt_stat.f_fsidx.__fsid_val[1]); 3245 3246 (*pr)("\towner = %"PRIu32"\n",mp->mnt_stat.f_owner); 3247 (*pr)("\tnamemax = %lu\n",mp->mnt_stat.f_namemax); 3248 3249 snprintb(sbuf, sizeof(sbuf), __MNT_FLAG_BITS, mp->mnt_stat.f_flag); 3250 3251 (*pr)("\tflag = %s\n",sbuf); 3252 (*pr)("\tsyncwrites = %" PRIu64 "\n",mp->mnt_stat.f_syncwrites); 3253 (*pr)("\tasyncwrites = %" PRIu64 "\n",mp->mnt_stat.f_asyncwrites); 3254 (*pr)("\tsyncreads = %" PRIu64 "\n",mp->mnt_stat.f_syncreads); 3255 (*pr)("\tasyncreads = %" PRIu64 "\n",mp->mnt_stat.f_asyncreads); 3256 (*pr)("\tfstypename = %s\n",mp->mnt_stat.f_fstypename); 3257 (*pr)("\tmntonname = %s\n",mp->mnt_stat.f_mntonname); 3258 (*pr)("\tmntfromname = %s\n",mp->mnt_stat.f_mntfromname); 3259 3260 { 3261 int cnt = 0; 3262 struct vnode *vp; 3263 (*pr)("locked vnodes ="); 3264 TAILQ_FOREACH(vp, &mp->mnt_vnodelist, v_mntvnodes) { 3265 if (VOP_ISLOCKED(vp)) { 3266 if ((++cnt % 6) == 0) { 3267 (*pr)(" %p,\n\t", vp); 3268 } else { 3269 (*pr)(" %p,", vp); 3270 } 3271 } 3272 } 3273 (*pr)("\n"); 3274 } 3275 3276 if (full) { 3277 int cnt = 0; 3278 struct vnode *vp; 3279 (*pr)("all vnodes ="); 3280 TAILQ_FOREACH(vp, &mp->mnt_vnodelist, v_mntvnodes) { 3281 if (!TAILQ_NEXT(vp, v_mntvnodes)) { 3282 (*pr)(" %p", vp); 3283 } else if ((++cnt % 6) == 0) { 3284 (*pr)(" %p,\n\t", vp); 3285 } else { 3286 (*pr)(" %p,", vp); 3287 } 3288 } 3289 (*pr)("\n", vp); 3290 } 3291 } 3292 #endif /* DDB || DEBUGPRINT */ 3293 3294 /* 3295 * Check if a device pointed to by vp is mounted. 3296 * 3297 * Returns: 3298 * EINVAL if it's not a disk 3299 * EBUSY if it's a disk and mounted 3300 * 0 if it's a disk and not mounted 3301 */ 3302 int 3303 rawdev_mounted(struct vnode *vp, struct vnode **bvpp) 3304 { 3305 struct vnode *bvp; 3306 dev_t dev; 3307 int d_type; 3308 3309 bvp = NULL; 3310 dev = vp->v_rdev; 3311 d_type = D_OTHER; 3312 3313 if (iskmemvp(vp)) 3314 return EINVAL; 3315 3316 switch (vp->v_type) { 3317 case VCHR: { 3318 const struct cdevsw *cdev; 3319 3320 cdev = cdevsw_lookup(dev); 3321 if (cdev != NULL) { 3322 dev_t blkdev; 3323 3324 blkdev = devsw_chr2blk(dev); 3325 if (blkdev != NODEV) { 3326 if (vfinddev(blkdev, VBLK, &bvp) != 0) { 3327 d_type = (cdev->d_flag & D_TYPEMASK); 3328 /* XXX: what if bvp disappears? */ 3329 vrele(bvp); 3330 } 3331 } 3332 } 3333 3334 break; 3335 } 3336 3337 case VBLK: { 3338 const struct bdevsw *bdev; 3339 3340 bdev = bdevsw_lookup(dev); 3341 if (bdev != NULL) 3342 d_type = (bdev->d_flag & D_TYPEMASK); 3343 3344 bvp = vp; 3345 3346 break; 3347 } 3348 3349 default: 3350 break; 3351 } 3352 3353 if (d_type != D_DISK) 3354 return EINVAL; 3355 3356 if (bvpp != NULL) 3357 *bvpp = bvp; 3358 3359 /* 3360 * XXX: This is bogus. We should be failing the request 3361 * XXX: not only if this specific slice is mounted, but 3362 * XXX: if it's on a disk with any other mounted slice. 3363 */ 3364 if (vfs_mountedon(bvp)) 3365 return EBUSY; 3366 3367 return 0; 3368 } 3369