Make an existing vnode lock assertion a bit more verbose.
[dragonfly.git] / sys / kern / vfs_subr.c
CommitLineData
984263bc
MD
1/*
2 * Copyright (c) 1989, 1993
3 * The Regents of the University of California. All rights reserved.
4 * (c) UNIX System Laboratories, Inc.
5 * All or some portions of this file are derived from material licensed
6 * to the University of California by American Telephone and Telegraph
7 * Co. or Unix System Laboratories, Inc. and are reproduced herein with
8 * the permission of UNIX System Laboratories, Inc.
9 *
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
12 * are met:
13 * 1. Redistributions of source code must retain the above copyright
14 * notice, this list of conditions and the following disclaimer.
15 * 2. Redistributions in binary form must reproduce the above copyright
16 * notice, this list of conditions and the following disclaimer in the
17 * documentation and/or other materials provided with the distribution.
18 * 3. All advertising materials mentioning features or use of this software
19 * must display the following acknowledgement:
20 * This product includes software developed by the University of
21 * California, Berkeley and its contributors.
22 * 4. Neither the name of the University nor the names of its contributors
23 * may be used to endorse or promote products derived from this software
24 * without specific prior written permission.
25 *
26 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
27 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
28 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
29 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
30 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
31 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
32 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
33 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
34 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
35 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
36 * SUCH DAMAGE.
37 *
38 * @(#)vfs_subr.c 8.31 (Berkeley) 5/26/95
39 * $FreeBSD: src/sys/kern/vfs_subr.c,v 1.249.2.30 2003/04/04 20:35:57 tegge Exp $
5fd012e0 40 * $DragonFly: src/sys/kern/vfs_subr.c,v 1.43 2004/10/12 19:20:46 dillon Exp $
984263bc
MD
41 */
42
43/*
44 * External virtual filesystem routines
45 */
46#include "opt_ddb.h"
47
48#include <sys/param.h>
49#include <sys/systm.h>
50#include <sys/buf.h>
51#include <sys/conf.h>
52#include <sys/dirent.h>
53#include <sys/domain.h>
54#include <sys/eventhandler.h>
55#include <sys/fcntl.h>
56#include <sys/kernel.h>
57#include <sys/kthread.h>
58#include <sys/malloc.h>
59#include <sys/mbuf.h>
60#include <sys/mount.h>
984263bc 61#include <sys/proc.h>
dadab5e9 62#include <sys/namei.h>
984263bc
MD
63#include <sys/reboot.h>
64#include <sys/socket.h>
65#include <sys/stat.h>
66#include <sys/sysctl.h>
67#include <sys/syslog.h>
68#include <sys/vmmeter.h>
69#include <sys/vnode.h>
70
71#include <machine/limits.h>
72
73#include <vm/vm.h>
74#include <vm/vm_object.h>
75#include <vm/vm_extern.h>
6ef943a3 76#include <vm/vm_kern.h>
984263bc
MD
77#include <vm/pmap.h>
78#include <vm/vm_map.h>
79#include <vm/vm_page.h>
80#include <vm/vm_pager.h>
81#include <vm/vnode_pager.h>
82#include <vm/vm_zone.h>
83
3020e3be 84#include <sys/buf2.h>
f5865223 85#include <sys/thread2.h>
3020e3be 86
984263bc
MD
87static MALLOC_DEFINE(M_NETADDR, "Export Host", "Export host address structure");
88
5fd012e0 89int numvnodes;
984263bc
MD
90SYSCTL_INT(_debug, OID_AUTO, numvnodes, CTLFLAG_RD, &numvnodes, 0, "");
91
92enum vtype iftovt_tab[16] = {
93 VNON, VFIFO, VCHR, VNON, VDIR, VNON, VBLK, VNON,
94 VREG, VNON, VLNK, VNON, VSOCK, VNON, VNON, VBAD,
95};
96int vttoif_tab[9] = {
97 0, S_IFREG, S_IFDIR, S_IFBLK, S_IFCHR, S_IFLNK,
98 S_IFSOCK, S_IFIFO, S_IFMT,
99};
100
984263bc 101static int reassignbufcalls;
dd98570a
MD
102SYSCTL_INT(_vfs, OID_AUTO, reassignbufcalls, CTLFLAG_RW,
103 &reassignbufcalls, 0, "");
984263bc 104static int reassignbufloops;
dd98570a
MD
105SYSCTL_INT(_vfs, OID_AUTO, reassignbufloops, CTLFLAG_RW,
106 &reassignbufloops, 0, "");
984263bc 107static int reassignbufsortgood;
dd98570a
MD
108SYSCTL_INT(_vfs, OID_AUTO, reassignbufsortgood, CTLFLAG_RW,
109 &reassignbufsortgood, 0, "");
984263bc 110static int reassignbufsortbad;
dd98570a
MD
111SYSCTL_INT(_vfs, OID_AUTO, reassignbufsortbad, CTLFLAG_RW,
112 &reassignbufsortbad, 0, "");
984263bc 113static int reassignbufmethod = 1;
dd98570a
MD
114SYSCTL_INT(_vfs, OID_AUTO, reassignbufmethod, CTLFLAG_RW,
115 &reassignbufmethod, 0, "");
984263bc
MD
116
117#ifdef ENABLE_VFS_IOOPT
118int vfs_ioopt = 0;
119SYSCTL_INT(_vfs, OID_AUTO, ioopt, CTLFLAG_RW, &vfs_ioopt, 0, "");
120#endif
121
984263bc 122int nfs_mount_type = -1;
8a8d5d85 123static struct lwkt_token spechash_token;
984263bc 124struct nfs_public nfs_pub; /* publicly exported FS */
984263bc
MD
125
126int desiredvnodes;
127SYSCTL_INT(_kern, KERN_MAXVNODES, maxvnodes, CTLFLAG_RW,
dd98570a 128 &desiredvnodes, 0, "Maximum number of vnodes");
984263bc 129
402ed7e1
RG
130static void vfs_free_addrlist (struct netexport *nep);
131static int vfs_free_netcred (struct radix_node *rn, void *w);
132static int vfs_hang_addrlist (struct mount *mp, struct netexport *nep,
133 struct export_args *argp);
984263bc 134
e4c9c0c8 135extern int dev_ref_debug;
0961aa92 136extern struct vnodeopv_entry_desc spec_vnodeop_entries[];
e4c9c0c8 137
41a01a4d 138/*
5fd012e0
MD
139 * Return 0 if the vnode is already on the free list or cannot be placed
140 * on the free list. Return 1 if the vnode can be placed on the free list.
41a01a4d 141 */
5fd012e0
MD
142static __inline int
143vshouldfree(struct vnode *vp, int usecount)
41a01a4d 144{
5fd012e0
MD
145 if (vp->v_flag & VFREE)
146 return (0); /* already free */
147 if (vp->v_holdcnt != 0 || vp->v_usecount != usecount)
148 return (0); /* other holderse */
149 if (vp->v_object &&
150 (vp->v_object->ref_count || vp->v_object->resident_page_count)) {
151 return (0);
152 }
153 return (1);
41a01a4d 154}
5fd012e0 155
984263bc 156/*
5fd012e0
MD
157 * Initialize the vnode management data structures.
158 *
159 * Called from vfsinit()
984263bc
MD
160 */
161void
5fd012e0 162vfs_subr_init(void)
984263bc 163{
6ef943a3
MD
164 /*
165 * Desired vnodes is a result of the physical page count
166 * and the size of kernel's heap. It scales in proportion
167 * to the amount of available physical memory. This can
168 * cause trouble on 64-bit and large memory platforms.
169 */
170 /* desiredvnodes = maxproc + vmstats.v_page_count / 4; */
171 desiredvnodes =
172 min(maxproc + vmstats.v_page_count /4,
173 2 * (VM_MAX_KERNEL_ADDRESS - VM_MIN_KERNEL_ADDRESS) /
174 (5 * (sizeof(struct vm_object) + sizeof(struct vnode))));
175
41a01a4d 176 lwkt_token_init(&spechash_token);
984263bc
MD
177}
178
179/*
180 * Knob to control the precision of file timestamps:
181 *
182 * 0 = seconds only; nanoseconds zeroed.
183 * 1 = seconds and nanoseconds, accurate within 1/HZ.
184 * 2 = seconds and nanoseconds, truncated to microseconds.
185 * >=3 = seconds and nanoseconds, maximum precision.
186 */
187enum { TSP_SEC, TSP_HZ, TSP_USEC, TSP_NSEC };
188
189static int timestamp_precision = TSP_SEC;
190SYSCTL_INT(_vfs, OID_AUTO, timestamp_precision, CTLFLAG_RW,
dd98570a 191 &timestamp_precision, 0, "");
984263bc
MD
192
193/*
194 * Get a current timestamp.
195 */
196void
dd98570a 197vfs_timestamp(struct timespec *tsp)
984263bc
MD
198{
199 struct timeval tv;
200
201 switch (timestamp_precision) {
202 case TSP_SEC:
203 tsp->tv_sec = time_second;
204 tsp->tv_nsec = 0;
205 break;
206 case TSP_HZ:
207 getnanotime(tsp);
208 break;
209 case TSP_USEC:
210 microtime(&tv);
211 TIMEVAL_TO_TIMESPEC(&tv, tsp);
212 break;
213 case TSP_NSEC:
214 default:
215 nanotime(tsp);
216 break;
217 }
218}
219
220/*
221 * Set vnode attributes to VNOVAL
222 */
223void
dd98570a 224vattr_null(struct vattr *vap)
984263bc 225{
984263bc
MD
226 vap->va_type = VNON;
227 vap->va_size = VNOVAL;
228 vap->va_bytes = VNOVAL;
229 vap->va_mode = VNOVAL;
230 vap->va_nlink = VNOVAL;
231 vap->va_uid = VNOVAL;
232 vap->va_gid = VNOVAL;
233 vap->va_fsid = VNOVAL;
234 vap->va_fileid = VNOVAL;
235 vap->va_blocksize = VNOVAL;
236 vap->va_rdev = VNOVAL;
237 vap->va_atime.tv_sec = VNOVAL;
238 vap->va_atime.tv_nsec = VNOVAL;
239 vap->va_mtime.tv_sec = VNOVAL;
240 vap->va_mtime.tv_nsec = VNOVAL;
241 vap->va_ctime.tv_sec = VNOVAL;
242 vap->va_ctime.tv_nsec = VNOVAL;
243 vap->va_flags = VNOVAL;
244 vap->va_gen = VNOVAL;
245 vap->va_vaflags = 0;
246}
247
984263bc
MD
248/*
249 * Update outstanding I/O count and do wakeup if requested.
250 */
251void
dd98570a 252vwakeup(struct buf *bp)
984263bc 253{
1fd87d54 254 struct vnode *vp;
984263bc
MD
255
256 bp->b_flags &= ~B_WRITEINPROG;
257 if ((vp = bp->b_vp)) {
258 vp->v_numoutput--;
259 if (vp->v_numoutput < 0)
260 panic("vwakeup: neg numoutput");
261 if ((vp->v_numoutput == 0) && (vp->v_flag & VBWAIT)) {
262 vp->v_flag &= ~VBWAIT;
263 wakeup((caddr_t) &vp->v_numoutput);
264 }
265 }
266}
267
268/*
269 * Flush out and invalidate all buffers associated with a vnode.
5fd012e0
MD
270 *
271 * vp must be locked.
984263bc
MD
272 */
273int
3b568787
MD
274vinvalbuf(struct vnode *vp, int flags, struct thread *td,
275 int slpflag, int slptimeo)
984263bc 276{
1fd87d54 277 struct buf *bp;
984263bc
MD
278 struct buf *nbp, *blist;
279 int s, error;
280 vm_object_t object;
281
282 if (flags & V_SAVE) {
283 s = splbio();
284 while (vp->v_numoutput) {
285 vp->v_flag |= VBWAIT;
286 error = tsleep((caddr_t)&vp->v_numoutput,
377d4740 287 slpflag, "vinvlbuf", slptimeo);
984263bc
MD
288 if (error) {
289 splx(s);
290 return (error);
291 }
292 }
293 if (!TAILQ_EMPTY(&vp->v_dirtyblkhd)) {
294 splx(s);
3b568787 295 if ((error = VOP_FSYNC(vp, MNT_WAIT, td)) != 0)
984263bc
MD
296 return (error);
297 s = splbio();
298 if (vp->v_numoutput > 0 ||
299 !TAILQ_EMPTY(&vp->v_dirtyblkhd))
300 panic("vinvalbuf: dirty bufs");
301 }
302 splx(s);
303 }
304 s = splbio();
305 for (;;) {
306 blist = TAILQ_FIRST(&vp->v_cleanblkhd);
307 if (!blist)
308 blist = TAILQ_FIRST(&vp->v_dirtyblkhd);
309 if (!blist)
310 break;
311
312 for (bp = blist; bp; bp = nbp) {
313 nbp = TAILQ_NEXT(bp, b_vnbufs);
314 if (BUF_LOCK(bp, LK_EXCLUSIVE | LK_NOWAIT)) {
315 error = BUF_TIMELOCK(bp,
316 LK_EXCLUSIVE | LK_SLEEPFAIL,
317 "vinvalbuf", slpflag, slptimeo);
318 if (error == ENOLCK)
319 break;
320 splx(s);
321 return (error);
322 }
323 /*
324 * XXX Since there are no node locks for NFS, I
325 * believe there is a slight chance that a delayed
326 * write will occur while sleeping just above, so
327 * check for it. Note that vfs_bio_awrite expects
328 * buffers to reside on a queue, while VOP_BWRITE and
329 * brelse do not.
330 */
331 if (((bp->b_flags & (B_DELWRI | B_INVAL)) == B_DELWRI) &&
332 (flags & V_SAVE)) {
333
334 if (bp->b_vp == vp) {
335 if (bp->b_flags & B_CLUSTEROK) {
336 BUF_UNLOCK(bp);
337 vfs_bio_awrite(bp);
338 } else {
339 bremfree(bp);
340 bp->b_flags |= B_ASYNC;
341 VOP_BWRITE(bp->b_vp, bp);
342 }
343 } else {
344 bremfree(bp);
345 (void) VOP_BWRITE(bp->b_vp, bp);
346 }
347 break;
348 }
349 bremfree(bp);
350 bp->b_flags |= (B_INVAL | B_NOCACHE | B_RELBUF);
351 bp->b_flags &= ~B_ASYNC;
352 brelse(bp);
353 }
354 }
355
356 /*
357 * Wait for I/O to complete. XXX needs cleaning up. The vnode can
358 * have write I/O in-progress but if there is a VM object then the
359 * VM object can also have read-I/O in-progress.
360 */
361 do {
362 while (vp->v_numoutput > 0) {
363 vp->v_flag |= VBWAIT;
377d4740 364 tsleep(&vp->v_numoutput, 0, "vnvlbv", 0);
984263bc
MD
365 }
366 if (VOP_GETVOBJECT(vp, &object) == 0) {
367 while (object->paging_in_progress)
368 vm_object_pip_sleep(object, "vnvlbx");
369 }
370 } while (vp->v_numoutput > 0);
371
372 splx(s);
373
374 /*
375 * Destroy the copy in the VM cache, too.
376 */
984263bc
MD
377 if (VOP_GETVOBJECT(vp, &object) == 0) {
378 vm_object_page_remove(object, 0, 0,
379 (flags & V_SAVE) ? TRUE : FALSE);
380 }
984263bc
MD
381
382 if (!TAILQ_EMPTY(&vp->v_dirtyblkhd) || !TAILQ_EMPTY(&vp->v_cleanblkhd))
383 panic("vinvalbuf: flush failed");
384 return (0);
385}
386
387/*
388 * Truncate a file's buffer and pages to a specified length. This
389 * is in lieu of the old vinvalbuf mechanism, which performed unneeded
390 * sync activity.
5fd012e0
MD
391 *
392 * The vnode must be locked.
984263bc
MD
393 */
394int
3b568787 395vtruncbuf(struct vnode *vp, struct thread *td, off_t length, int blksize)
984263bc 396{
dadab5e9 397 struct buf *bp;
984263bc
MD
398 struct buf *nbp;
399 int s, anyfreed;
400 int trunclbn;
401
402 /*
403 * Round up to the *next* lbn.
404 */
405 trunclbn = (length + blksize - 1) / blksize;
406
407 s = splbio();
408restart:
409 anyfreed = 1;
410 for (;anyfreed;) {
411 anyfreed = 0;
412 for (bp = TAILQ_FIRST(&vp->v_cleanblkhd); bp; bp = nbp) {
413 nbp = TAILQ_NEXT(bp, b_vnbufs);
414 if (bp->b_lblkno >= trunclbn) {
415 if (BUF_LOCK(bp, LK_EXCLUSIVE | LK_NOWAIT)) {
416 BUF_LOCK(bp, LK_EXCLUSIVE|LK_SLEEPFAIL);
417 goto restart;
418 } else {
419 bremfree(bp);
420 bp->b_flags |= (B_INVAL | B_RELBUF);
421 bp->b_flags &= ~B_ASYNC;
422 brelse(bp);
423 anyfreed = 1;
424 }
425 if (nbp &&
426 (((nbp->b_xflags & BX_VNCLEAN) == 0) ||
427 (nbp->b_vp != vp) ||
428 (nbp->b_flags & B_DELWRI))) {
429 goto restart;
430 }
431 }
432 }
433
434 for (bp = TAILQ_FIRST(&vp->v_dirtyblkhd); bp; bp = nbp) {
435 nbp = TAILQ_NEXT(bp, b_vnbufs);
436 if (bp->b_lblkno >= trunclbn) {
437 if (BUF_LOCK(bp, LK_EXCLUSIVE | LK_NOWAIT)) {
438 BUF_LOCK(bp, LK_EXCLUSIVE|LK_SLEEPFAIL);
439 goto restart;
440 } else {
441 bremfree(bp);
442 bp->b_flags |= (B_INVAL | B_RELBUF);
443 bp->b_flags &= ~B_ASYNC;
444 brelse(bp);
445 anyfreed = 1;
446 }
447 if (nbp &&
448 (((nbp->b_xflags & BX_VNDIRTY) == 0) ||
449 (nbp->b_vp != vp) ||
450 (nbp->b_flags & B_DELWRI) == 0)) {
451 goto restart;
452 }
453 }
454 }
455 }
456
457 if (length > 0) {
458restartsync:
459 for (bp = TAILQ_FIRST(&vp->v_dirtyblkhd); bp; bp = nbp) {
460 nbp = TAILQ_NEXT(bp, b_vnbufs);
461 if ((bp->b_flags & B_DELWRI) && (bp->b_lblkno < 0)) {
462 if (BUF_LOCK(bp, LK_EXCLUSIVE | LK_NOWAIT)) {
463 BUF_LOCK(bp, LK_EXCLUSIVE|LK_SLEEPFAIL);
464 goto restart;
465 } else {
466 bremfree(bp);
467 if (bp->b_vp == vp) {
468 bp->b_flags |= B_ASYNC;
469 } else {
470 bp->b_flags &= ~B_ASYNC;
471 }
472 VOP_BWRITE(bp->b_vp, bp);
473 }
474 goto restartsync;
475 }
476
477 }
478 }
479
480 while (vp->v_numoutput > 0) {
481 vp->v_flag |= VBWAIT;
377d4740 482 tsleep(&vp->v_numoutput, 0, "vbtrunc", 0);
984263bc
MD
483 }
484
485 splx(s);
486
487 vnode_pager_setsize(vp, length);
488
489 return (0);
490}
491
492/*
493 * Associate a buffer with a vnode.
494 */
495void
dd98570a 496bgetvp(struct vnode *vp, struct buf *bp)
984263bc 497{
984263bc
MD
498 KASSERT(bp->b_vp == NULL, ("bgetvp: not free"));
499
500 vhold(vp);
501 bp->b_vp = vp;
502 bp->b_dev = vn_todev(vp);
503 /*
504 * Insert onto list for new vnode.
505 */
5fd012e0 506 crit_enter();
984263bc
MD
507 bp->b_xflags |= BX_VNCLEAN;
508 bp->b_xflags &= ~BX_VNDIRTY;
509 TAILQ_INSERT_TAIL(&vp->v_cleanblkhd, bp, b_vnbufs);
5fd012e0 510 crit_exit();
984263bc
MD
511}
512
513/*
514 * Disassociate a buffer from a vnode.
515 */
516void
dd98570a 517brelvp(struct buf *bp)
984263bc
MD
518{
519 struct vnode *vp;
520 struct buflists *listheadp;
984263bc
MD
521
522 KASSERT(bp->b_vp != NULL, ("brelvp: NULL"));
523
524 /*
525 * Delete from old vnode list, if on one.
526 */
527 vp = bp->b_vp;
5fd012e0 528 crit_enter();
984263bc
MD
529 if (bp->b_xflags & (BX_VNDIRTY | BX_VNCLEAN)) {
530 if (bp->b_xflags & BX_VNDIRTY)
531 listheadp = &vp->v_dirtyblkhd;
532 else
533 listheadp = &vp->v_cleanblkhd;
534 TAILQ_REMOVE(listheadp, bp, b_vnbufs);
535 bp->b_xflags &= ~(BX_VNDIRTY | BX_VNCLEAN);
536 }
537 if ((vp->v_flag & VONWORKLST) && TAILQ_EMPTY(&vp->v_dirtyblkhd)) {
538 vp->v_flag &= ~VONWORKLST;
539 LIST_REMOVE(vp, v_synclist);
540 }
f5865223 541 crit_exit();
5fd012e0
MD
542 bp->b_vp = NULL;
543 vdrop(vp);
984263bc
MD
544}
545
546/*
547 * Associate a p-buffer with a vnode.
548 *
549 * Also sets B_PAGING flag to indicate that vnode is not fully associated
550 * with the buffer. i.e. the bp has not been linked into the vnode or
551 * ref-counted.
552 */
553void
dd98570a 554pbgetvp(struct vnode *vp, struct buf *bp)
984263bc 555{
984263bc
MD
556 KASSERT(bp->b_vp == NULL, ("pbgetvp: not free"));
557
558 bp->b_vp = vp;
559 bp->b_flags |= B_PAGING;
560 bp->b_dev = vn_todev(vp);
561}
562
563/*
564 * Disassociate a p-buffer from a vnode.
565 */
566void
dd98570a 567pbrelvp(struct buf *bp)
984263bc 568{
984263bc
MD
569 KASSERT(bp->b_vp != NULL, ("pbrelvp: NULL"));
570
571 /* XXX REMOVE ME */
572 if (TAILQ_NEXT(bp, b_vnbufs) != NULL) {
573 panic(
574 "relpbuf(): b_vp was probably reassignbuf()d %p %x",
575 bp,
576 (int)bp->b_flags
577 );
578 }
579 bp->b_vp = (struct vnode *) 0;
580 bp->b_flags &= ~B_PAGING;
581}
582
583void
dd98570a 584pbreassignbuf(struct buf *bp, struct vnode *newvp)
984263bc
MD
585{
586 if ((bp->b_flags & B_PAGING) == 0) {
587 panic(
588 "pbreassignbuf() on non phys bp %p",
589 bp
590 );
591 }
592 bp->b_vp = newvp;
593}
594
595/*
596 * Reassign a buffer from one vnode to another.
597 * Used to assign file specific control information
598 * (indirect blocks) to the vnode to which they belong.
599 */
600void
dd98570a 601reassignbuf(struct buf *bp, struct vnode *newvp)
984263bc
MD
602{
603 struct buflists *listheadp;
604 int delay;
984263bc
MD
605
606 if (newvp == NULL) {
607 printf("reassignbuf: NULL");
608 return;
609 }
610 ++reassignbufcalls;
611
612 /*
613 * B_PAGING flagged buffers cannot be reassigned because their vp
614 * is not fully linked in.
615 */
616 if (bp->b_flags & B_PAGING)
617 panic("cannot reassign paging buffer");
618
5fd012e0 619 crit_enter();
984263bc
MD
620 /*
621 * Delete from old vnode list, if on one.
622 */
623 if (bp->b_xflags & (BX_VNDIRTY | BX_VNCLEAN)) {
624 if (bp->b_xflags & BX_VNDIRTY)
625 listheadp = &bp->b_vp->v_dirtyblkhd;
626 else
627 listheadp = &bp->b_vp->v_cleanblkhd;
628 TAILQ_REMOVE(listheadp, bp, b_vnbufs);
629 bp->b_xflags &= ~(BX_VNDIRTY | BX_VNCLEAN);
630 if (bp->b_vp != newvp) {
631 vdrop(bp->b_vp);
632 bp->b_vp = NULL; /* for clarification */
633 }
634 }
635 /*
636 * If dirty, put on list of dirty buffers; otherwise insert onto list
637 * of clean buffers.
638 */
639 if (bp->b_flags & B_DELWRI) {
640 struct buf *tbp;
641
642 listheadp = &newvp->v_dirtyblkhd;
643 if ((newvp->v_flag & VONWORKLST) == 0) {
644 switch (newvp->v_type) {
645 case VDIR:
646 delay = dirdelay;
647 break;
648 case VCHR:
649 case VBLK:
e4c9c0c8
MD
650 if (newvp->v_rdev &&
651 newvp->v_rdev->si_mountpoint != NULL) {
984263bc
MD
652 delay = metadelay;
653 break;
654 }
655 /* fall through */
656 default:
657 delay = filedelay;
658 }
659 vn_syncer_add_to_worklist(newvp, delay);
660 }
661 bp->b_xflags |= BX_VNDIRTY;
662 tbp = TAILQ_FIRST(listheadp);
663 if (tbp == NULL ||
664 bp->b_lblkno == 0 ||
665 (bp->b_lblkno > 0 && tbp->b_lblkno < 0) ||
666 (bp->b_lblkno > 0 && bp->b_lblkno < tbp->b_lblkno)) {
667 TAILQ_INSERT_HEAD(listheadp, bp, b_vnbufs);
668 ++reassignbufsortgood;
669 } else if (bp->b_lblkno < 0) {
670 TAILQ_INSERT_TAIL(listheadp, bp, b_vnbufs);
671 ++reassignbufsortgood;
672 } else if (reassignbufmethod == 1) {
673 /*
674 * New sorting algorithm, only handle sequential case,
675 * otherwise append to end (but before metadata)
676 */
677 if ((tbp = gbincore(newvp, bp->b_lblkno - 1)) != NULL &&
678 (tbp->b_xflags & BX_VNDIRTY)) {
679 /*
680 * Found the best place to insert the buffer
681 */
682 TAILQ_INSERT_AFTER(listheadp, tbp, bp, b_vnbufs);
683 ++reassignbufsortgood;
684 } else {
685 /*
686 * Missed, append to end, but before meta-data.
687 * We know that the head buffer in the list is
688 * not meta-data due to prior conditionals.
689 *
690 * Indirect effects: NFS second stage write
691 * tends to wind up here, giving maximum
692 * distance between the unstable write and the
693 * commit rpc.
694 */
695 tbp = TAILQ_LAST(listheadp, buflists);
696 while (tbp && tbp->b_lblkno < 0)
697 tbp = TAILQ_PREV(tbp, buflists, b_vnbufs);
698 TAILQ_INSERT_AFTER(listheadp, tbp, bp, b_vnbufs);
699 ++reassignbufsortbad;
700 }
701 } else {
702 /*
703 * Old sorting algorithm, scan queue and insert
704 */
705 struct buf *ttbp;
706 while ((ttbp = TAILQ_NEXT(tbp, b_vnbufs)) &&
707 (ttbp->b_lblkno < bp->b_lblkno)) {
708 ++reassignbufloops;
709 tbp = ttbp;
710 }
711 TAILQ_INSERT_AFTER(listheadp, tbp, bp, b_vnbufs);
712 }
713 } else {
714 bp->b_xflags |= BX_VNCLEAN;
715 TAILQ_INSERT_TAIL(&newvp->v_cleanblkhd, bp, b_vnbufs);
716 if ((newvp->v_flag & VONWORKLST) &&
717 TAILQ_EMPTY(&newvp->v_dirtyblkhd)) {
718 newvp->v_flag &= ~VONWORKLST;
719 LIST_REMOVE(newvp, v_synclist);
720 }
721 }
722 if (bp->b_vp != newvp) {
723 bp->b_vp = newvp;
724 vhold(bp->b_vp);
725 }
5fd012e0 726 crit_exit();
984263bc
MD
727}
728
729/*
730 * Create a vnode for a block device.
731 * Used for mounting the root file system.
732 */
733int
e4c9c0c8 734bdevvp(dev_t dev, struct vnode **vpp)
984263bc 735{
1fd87d54 736 struct vnode *vp;
984263bc
MD
737 struct vnode *nvp;
738 int error;
739
740 if (dev == NODEV) {
741 *vpp = NULLVP;
742 return (ENXIO);
743 }
3446c007 744 error = getnewvnode(VT_NON, NULL, spec_vnode_vops, &nvp, 0, 0);
984263bc
MD
745 if (error) {
746 *vpp = NULLVP;
747 return (error);
748 }
749 vp = nvp;
e4c9c0c8
MD
750 vp->v_type = VCHR;
751 vp->v_udev = dev->si_udev;
5fd012e0 752 vx_unlock(vp);
984263bc
MD
753 *vpp = vp;
754 return (0);
5fd012e0 755}
41a01a4d 756
984263bc 757int
5fd012e0 758v_associate_rdev(struct vnode *vp, dev_t dev)
984263bc 759{
5fd012e0 760 lwkt_tokref ilock;
984263bc 761
5fd012e0
MD
762 if (dev == NULL || dev == NODEV)
763 return(ENXIO);
764 if (dev_is_good(dev) == 0)
765 return(ENXIO);
766 KKASSERT(vp->v_rdev == NULL);
767 if (dev_ref_debug)
768 printf("Z1");
769 vp->v_rdev = reference_dev(dev);
770 lwkt_gettoken(&ilock, &spechash_token);
771 SLIST_INSERT_HEAD(&dev->si_hlist, vp, v_specnext);
772 lwkt_reltoken(&ilock);
773 return(0);
774}
984263bc 775
5fd012e0
MD
776void
777v_release_rdev(struct vnode *vp)
778{
779 lwkt_tokref ilock;
780 dev_t dev;
984263bc 781
5fd012e0
MD
782 if ((dev = vp->v_rdev) != NULL) {
783 lwkt_gettoken(&ilock, &spechash_token);
784 SLIST_REMOVE(&dev->si_hlist, vp, vnode, v_specnext);
785 if (dev_ref_debug && vp->v_opencount != 0) {
786 printf("releasing rdev with non-0 "
787 "v_opencount(%d) (revoked?)\n",
788 vp->v_opencount);
41a01a4d 789 }
5fd012e0
MD
790 vp->v_rdev = NULL;
791 vp->v_opencount = 0;
792 release_dev(dev);
793 lwkt_reltoken(&ilock);
984263bc 794 }
984263bc
MD
795}
796
797/*
5fd012e0
MD
798 * Add a vnode to the alias list hung off the dev_t. We only associate
799 * the device number with the vnode. The actual device is not associated
800 * until the vnode is opened (usually in spec_open()), and will be
801 * disassociated on last close.
984263bc 802 */
5fd012e0
MD
803void
804addaliasu(struct vnode *nvp, udev_t nvp_udev)
984263bc 805{
5fd012e0
MD
806 if (nvp->v_type != VBLK && nvp->v_type != VCHR)
807 panic("addaliasu on non-special vnode");
808 nvp->v_udev = nvp_udev;
984263bc
MD
809}
810
811/*
5fd012e0
MD
812 * Disassociate a vnode from its underlying filesystem.
813 *
814 * The vnode must be VX locked and refd
815 *
816 * If there are v_usecount references to the vnode other then ours we have
817 * to VOP_CLOSE the vnode before we can deactivate and reclaim it.
984263bc 818 */
5fd012e0
MD
819void
820vclean(struct vnode *vp, int flags, struct thread *td)
984263bc
MD
821{
822 int active;
823
824 /*
5fd012e0 825 * If the vnode has already been reclaimed we have nothing to do.
984263bc 826 */
5fd012e0
MD
827 if (vp->v_flag & VRECLAIMED)
828 return;
829 vp->v_flag |= VRECLAIMED;
984263bc
MD
830
831 /*
5fd012e0 832 * Scrap the vfs cache
984263bc 833 */
5fd012e0 834 cache_inval_vp(vp, CINV_SELF);
41a01a4d 835
984263bc 836 /*
5fd012e0
MD
837 * Check to see if the vnode is in use. If so we have to reference it
838 * before we clean it out so that its count cannot fall to zero and
839 * generate a race against ourselves to recycle it.
984263bc 840 */
5fd012e0 841 active = (vp->v_usecount > 1);
984263bc
MD
842
843 /*
5fd012e0
MD
844 * Clean out any buffers associated with the vnode and destroy its
845 * object, if it has one.
984263bc 846 */
3b568787 847 vinvalbuf(vp, V_SAVE, td, 0, 0);
984263bc
MD
848 VOP_DESTROYVOBJECT(vp);
849
850 /*
851 * If purging an active vnode, it must be closed and
5fd012e0
MD
852 * deactivated before being reclaimed. XXX
853 *
854 * Note that neither of these routines unlocks the vnode.
984263bc
MD
855 */
856 if (active) {
857 if (flags & DOCLOSE)
3b568787 858 VOP_CLOSE(vp, FNONBLOCK, td);
5fd012e0
MD
859 }
860
861 /*
862 * If the vnode has not be deactivated, deactivated it.
863 */
864 if ((vp->v_flag & VINACTIVE) == 0) {
865 vp->v_flag |= VINACTIVE;
dadab5e9 866 VOP_INACTIVE(vp, td);
984263bc 867 }
5fd012e0 868
984263bc
MD
869 /*
870 * Reclaim the vnode.
871 */
dadab5e9 872 if (VOP_RECLAIM(vp, td))
984263bc
MD
873 panic("vclean: cannot reclaim");
874
984263bc
MD
875 /*
876 * Done with purge, notify sleepers of the grim news.
877 */
0961aa92 878 vp->v_ops = dead_vnode_vops;
984263bc
MD
879 vn_pollgone(vp);
880 vp->v_tag = VT_NON;
984263bc
MD
881}
882
883/*
884 * Eliminate all activity associated with the requested vnode
885 * and with all vnodes aliased to the requested vnode.
dd98570a 886 *
5fd012e0
MD
887 * The vnode must be referenced and vx_lock()'d
888 *
dd98570a 889 * revoke { struct vnode *a_vp, int a_flags }
984263bc
MD
890 */
891int
2d3e977e 892vop_stdrevoke(struct vop_revoke_args *ap)
984263bc
MD
893{
894 struct vnode *vp, *vq;
41a01a4d 895 lwkt_tokref ilock;
984263bc
MD
896 dev_t dev;
897
898 KASSERT((ap->a_flags & REVOKEALL) != 0, ("vop_revoke"));
899
900 vp = ap->a_vp;
5fd012e0 901
984263bc 902 /*
5fd012e0 903 * If the vnode is already dead don't try to revoke it
984263bc 904 */
5fd012e0 905 if (vp->v_flag & VRECLAIMED)
984263bc 906 return (0);
e4c9c0c8
MD
907
908 /*
909 * If the vnode has a device association, scrap all vnodes associated
910 * with the device. Don't let the device disappear on us while we
911 * are scrapping the vnodes.
5fd012e0
MD
912 *
913 * The passed vp will probably show up in the list, do not VX lock
914 * it twice!
e4c9c0c8
MD
915 */
916 if (vp->v_type != VCHR && vp->v_type != VBLK)
917 return(0);
918 if ((dev = vp->v_rdev) == NULL) {
919 if ((dev = udev2dev(vp->v_udev, vp->v_type == VBLK)) == NODEV)
920 return(0);
921 }
922 reference_dev(dev);
5fd012e0
MD
923 lwkt_gettoken(&ilock, &spechash_token);
924 while ((vq = SLIST_FIRST(&dev->si_hlist)) != NULL) {
925 if (vp == vq || vx_get(vq) == 0) {
926 if (vq == SLIST_FIRST(&dev->si_hlist))
927 vgone(vq);
928 if (vp != vq)
929 vx_put(vq);
930 }
984263bc 931 }
5fd012e0 932 lwkt_reltoken(&ilock);
e4c9c0c8 933 release_dev(dev);
984263bc
MD
934 return (0);
935}
936
937/*
938 * Recycle an unused vnode to the front of the free list.
5fd012e0
MD
939 *
940 * Returns 1 if we were successfully able to recycle the vnode,
941 * 0 otherwise.
984263bc
MD
942 */
943int
5fd012e0 944vrecycle(struct vnode *vp, struct thread *td)
984263bc 945{
5fd012e0
MD
946 if (vp->v_usecount == 1) {
947 vgone(vp);
984263bc
MD
948 return (1);
949 }
984263bc
MD
950 return (0);
951}
952
953/*
5fd012e0
MD
954 * Eliminate all activity associated with a vnode in preparation for reuse.
955 *
956 * The vnode must be VX locked and will remain VX locked on return. This
957 * routine may be called with the vnode in any state, as long as it is
958 * VX locked. The vnode will be cleaned out and marked VRECLAIMED but will
959 * not actually be reused until all existing refs and holds go away.
960 *
961 * NOTE: This routine may be called on a vnode which has not yet been
962 * already been deactivated (VOP_INACTIVE), or on a vnode which has
963 * already been reclaimed.
964 *
965 * This routine is not responsible for placing us back on the freelist.
966 * Instead, it happens automatically when the caller releases the VX lock
967 * (assuming there aren't any other references).
984263bc
MD
968 */
969void
dadab5e9 970vgone(struct vnode *vp)
984263bc 971{
984263bc 972 /*
5fd012e0
MD
973 * assert that the VX lock is held. This is an absolute requirement
974 * now for vgone() to be called.
984263bc 975 */
5fd012e0 976 KKASSERT(vp->v_lock.lk_exclusivecount == 1);
984263bc
MD
977
978 /*
5fd012e0
MD
979 * Clean out the filesystem specific data and set the VRECLAIMED
980 * bit. Also deactivate the vnode if necessary.
984263bc 981 */
5fd012e0 982 vclean(vp, DOCLOSE, curthread);
984263bc
MD
983
984 /*
985 * Delete from old mount point vnode list, if on one.
986 */
987 if (vp->v_mount != NULL)
5fd012e0 988 insmntque(vp, NULL);
dd98570a 989
984263bc
MD
990 /*
991 * If special device, remove it from special device alias list
1fbb5fc0
MD
992 * if it is on one. This should normally only occur if a vnode is
993 * being revoked as the device should otherwise have been released
994 * naturally.
984263bc
MD
995 */
996 if ((vp->v_type == VBLK || vp->v_type == VCHR) && vp->v_rdev != NULL) {
e4c9c0c8 997 v_release_rdev(vp);
984263bc
MD
998 }
999
1000 /*
5fd012e0 1001 * Set us to VBAD
984263bc 1002 */
984263bc 1003 vp->v_type = VBAD;
984263bc
MD
1004}
1005
1006/*
1007 * Lookup a vnode by device number.
1008 */
1009int
dd98570a 1010vfinddev(dev_t dev, enum vtype type, struct vnode **vpp)
984263bc 1011{
41a01a4d 1012 lwkt_tokref ilock;
984263bc
MD
1013 struct vnode *vp;
1014
41a01a4d 1015 lwkt_gettoken(&ilock, &spechash_token);
984263bc
MD
1016 SLIST_FOREACH(vp, &dev->si_hlist, v_specnext) {
1017 if (type == vp->v_type) {
1018 *vpp = vp;
41a01a4d 1019 lwkt_reltoken(&ilock);
984263bc
MD
1020 return (1);
1021 }
1022 }
41a01a4d 1023 lwkt_reltoken(&ilock);
984263bc
MD
1024 return (0);
1025}
1026
1027/*
e4c9c0c8
MD
1028 * Calculate the total number of references to a special device. This
1029 * routine may only be called for VBLK and VCHR vnodes since v_rdev is
1030 * an overloaded field. Since udev2dev can now return NODEV, we have
1031 * to check for a NULL v_rdev.
984263bc
MD
1032 */
1033int
e4c9c0c8 1034count_dev(dev_t dev)
984263bc 1035{
41a01a4d 1036 lwkt_tokref ilock;
e4c9c0c8
MD
1037 struct vnode *vp;
1038 int count = 0;
984263bc 1039
e4c9c0c8
MD
1040 if (SLIST_FIRST(&dev->si_hlist)) {
1041 lwkt_gettoken(&ilock, &spechash_token);
1042 SLIST_FOREACH(vp, &dev->si_hlist, v_specnext) {
1043 count += vp->v_usecount;
1044 }
1045 lwkt_reltoken(&ilock);
1046 }
1047 return(count);
984263bc
MD
1048}
1049
984263bc 1050int
e4c9c0c8 1051count_udev(udev_t udev)
984263bc 1052{
e4c9c0c8 1053 dev_t dev;
984263bc 1054
e4c9c0c8
MD
1055 if ((dev = udev2dev(udev, 0)) == NODEV)
1056 return(0);
1057 return(count_dev(dev));
1058}
1059
1060int
1061vcount(struct vnode *vp)
1062{
1063 if (vp->v_rdev == NULL)
1064 return(0);
1065 return(count_dev(vp->v_rdev));
984263bc
MD
1066}
1067
1068/*
1069 * Print out a description of a vnode.
1070 */
1071static char *typename[] =
1072{"VNON", "VREG", "VDIR", "VBLK", "VCHR", "VLNK", "VSOCK", "VFIFO", "VBAD"};
1073
1074void
dd98570a 1075vprint(char *label, struct vnode *vp)
984263bc
MD
1076{
1077 char buf[96];
1078
1079 if (label != NULL)
1080 printf("%s: %p: ", label, (void *)vp);
1081 else
1082 printf("%p: ", (void *)vp);
1083 printf("type %s, usecount %d, writecount %d, refcount %d,",
1084 typename[vp->v_type], vp->v_usecount, vp->v_writecount,
1085 vp->v_holdcnt);
1086 buf[0] = '\0';
1087 if (vp->v_flag & VROOT)
1088 strcat(buf, "|VROOT");
1089 if (vp->v_flag & VTEXT)
1090 strcat(buf, "|VTEXT");
1091 if (vp->v_flag & VSYSTEM)
1092 strcat(buf, "|VSYSTEM");
984263bc
MD
1093 if (vp->v_flag & VBWAIT)
1094 strcat(buf, "|VBWAIT");
984263bc
MD
1095 if (vp->v_flag & VFREE)
1096 strcat(buf, "|VFREE");
1097 if (vp->v_flag & VOBJBUF)
1098 strcat(buf, "|VOBJBUF");
1099 if (buf[0] != '\0')
1100 printf(" flags (%s)", &buf[1]);
1101 if (vp->v_data == NULL) {
1102 printf("\n");
1103 } else {
1104 printf("\n\t");
1105 VOP_PRINT(vp);
1106 }
1107}
1108
1109#ifdef DDB
1110#include <ddb/ddb.h>
1111/*
1112 * List all of the locked vnodes in the system.
1113 * Called when debugging the kernel.
1114 */
1115DB_SHOW_COMMAND(lockedvnodes, lockedvnodes)
1116{
dadab5e9 1117 struct thread *td = curthread; /* XXX */
41a01a4d 1118 lwkt_tokref ilock;
984263bc
MD
1119 struct mount *mp, *nmp;
1120 struct vnode *vp;
1121
1122 printf("Locked vnodes\n");
41a01a4d 1123 lwkt_gettoken(&ilock, &mountlist_token);
984263bc 1124 for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
41a01a4d 1125 if (vfs_busy(mp, LK_NOWAIT, &ilock, td)) {
984263bc
MD
1126 nmp = TAILQ_NEXT(mp, mnt_list);
1127 continue;
1128 }
1129 TAILQ_FOREACH(vp, &mp->mnt_nvnodelist, v_nmntvnodes) {
1130 if (VOP_ISLOCKED(vp, NULL))
1131 vprint((char *)0, vp);
1132 }
41a01a4d 1133 lwkt_gettokref(&ilock);
984263bc 1134 nmp = TAILQ_NEXT(mp, mnt_list);
dadab5e9 1135 vfs_unbusy(mp, td);
984263bc 1136 }
41a01a4d 1137 lwkt_reltoken(&ilock);
984263bc
MD
1138}
1139#endif
1140
1141/*
1142 * Top level filesystem related information gathering.
1143 */
402ed7e1 1144static int sysctl_ovfs_conf (SYSCTL_HANDLER_ARGS);
984263bc
MD
1145
1146static int
1147vfs_sysctl(SYSCTL_HANDLER_ARGS)
1148{
1149 int *name = (int *)arg1 - 1; /* XXX */
1150 u_int namelen = arg2 + 1; /* XXX */
1151 struct vfsconf *vfsp;
1152
1153#if 1 || defined(COMPAT_PRELITE2)
1154 /* Resolve ambiguity between VFS_VFSCONF and VFS_GENERIC. */
1155 if (namelen == 1)
1156 return (sysctl_ovfs_conf(oidp, arg1, arg2, req));
1157#endif
1158
1159#ifdef notyet
1160 /* all sysctl names at this level are at least name and field */
1161 if (namelen < 2)
1162 return (ENOTDIR); /* overloaded */
1163 if (name[0] != VFS_GENERIC) {
1164 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next)
1165 if (vfsp->vfc_typenum == name[0])
1166 break;
1167 if (vfsp == NULL)
1168 return (EOPNOTSUPP);
1169 return ((*vfsp->vfc_vfsops->vfs_sysctl)(&name[1], namelen - 1,
1170 oldp, oldlenp, newp, newlen, p));
1171 }
1172#endif
1173 switch (name[1]) {
1174 case VFS_MAXTYPENUM:
1175 if (namelen != 2)
1176 return (ENOTDIR);
1177 return (SYSCTL_OUT(req, &maxvfsconf, sizeof(int)));
1178 case VFS_CONF:
1179 if (namelen != 3)
1180 return (ENOTDIR); /* overloaded */
1181 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next)
1182 if (vfsp->vfc_typenum == name[2])
1183 break;
1184 if (vfsp == NULL)
1185 return (EOPNOTSUPP);
1186 return (SYSCTL_OUT(req, vfsp, sizeof *vfsp));
1187 }
1188 return (EOPNOTSUPP);
1189}
1190
1191SYSCTL_NODE(_vfs, VFS_GENERIC, generic, CTLFLAG_RD, vfs_sysctl,
1192 "Generic filesystem");
1193
1194#if 1 || defined(COMPAT_PRELITE2)
1195
1196static int
1197sysctl_ovfs_conf(SYSCTL_HANDLER_ARGS)
1198{
1199 int error;
1200 struct vfsconf *vfsp;
1201 struct ovfsconf ovfs;
1202
1203 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) {
1204 ovfs.vfc_vfsops = vfsp->vfc_vfsops; /* XXX used as flag */
1205 strcpy(ovfs.vfc_name, vfsp->vfc_name);
1206 ovfs.vfc_index = vfsp->vfc_typenum;
1207 ovfs.vfc_refcount = vfsp->vfc_refcount;
1208 ovfs.vfc_flags = vfsp->vfc_flags;
1209 error = SYSCTL_OUT(req, &ovfs, sizeof ovfs);
1210 if (error)
1211 return error;
1212 }
1213 return 0;
1214}
1215
1216#endif /* 1 || COMPAT_PRELITE2 */
1217
1218#if 0
1219#define KINFO_VNODESLOP 10
1220/*
1221 * Dump vnode list (via sysctl).
1222 * Copyout address of vnode followed by vnode.
1223 */
1224/* ARGSUSED */
1225static int
1226sysctl_vnode(SYSCTL_HANDLER_ARGS)
1227{
1228 struct proc *p = curproc; /* XXX */
1229 struct mount *mp, *nmp;
1230 struct vnode *nvp, *vp;
41a01a4d
MD
1231 lwkt_tokref ilock;
1232 lwkt_tokref jlock;
984263bc
MD
1233 int error;
1234
1235#define VPTRSZ sizeof (struct vnode *)
1236#define VNODESZ sizeof (struct vnode)
1237
1238 req->lock = 0;
1239 if (!req->oldptr) /* Make an estimate */
1240 return (SYSCTL_OUT(req, 0,
1241 (numvnodes + KINFO_VNODESLOP) * (VPTRSZ + VNODESZ)));
1242
41a01a4d 1243 lwkt_gettoken(&ilock, &mountlist_token);
984263bc 1244 for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
41a01a4d 1245 if (vfs_busy(mp, LK_NOWAIT, &ilock, p)) {
984263bc
MD
1246 nmp = TAILQ_NEXT(mp, mnt_list);
1247 continue;
1248 }
41a01a4d 1249 lwkt_gettoken(&jlock, &mntvnode_token);
984263bc 1250again:
984263bc
MD
1251 for (vp = TAILQ_FIRST(&mp->mnt_nvnodelist);
1252 vp != NULL;
1253 vp = nvp) {
1254 /*
1255 * Check that the vp is still associated with
1256 * this filesystem. RACE: could have been
1257 * recycled onto the same filesystem.
1258 */
41a01a4d 1259 if (vp->v_mount != mp)
984263bc 1260 goto again;
984263bc 1261 nvp = TAILQ_NEXT(vp, v_nmntvnodes);
984263bc 1262 if ((error = SYSCTL_OUT(req, &vp, VPTRSZ)) ||
41a01a4d
MD
1263 (error = SYSCTL_OUT(req, vp, VNODESZ))) {
1264 lwkt_reltoken(&jlock);
984263bc 1265 return (error);
41a01a4d 1266 }
984263bc 1267 }
41a01a4d
MD
1268 lwkt_reltoken(&jlock);
1269 lwkt_gettokref(&ilock);
1270 nmp = TAILQ_NEXT(mp, mnt_list); /* ZZZ */
984263bc
MD
1271 vfs_unbusy(mp, p);
1272 }
41a01a4d 1273 lwkt_reltoken(&ilock);
984263bc
MD
1274
1275 return (0);
1276}
1277#endif
1278
1279/*
1280 * XXX
1281 * Exporting the vnode list on large systems causes them to crash.
1282 * Exporting the vnode list on medium systems causes sysctl to coredump.
1283 */
1284#if 0
1285SYSCTL_PROC(_kern, KERN_VNODE, vnode, CTLTYPE_OPAQUE|CTLFLAG_RD,
1286 0, 0, sysctl_vnode, "S,vnode", "");
1287#endif
1288
1289/*
1290 * Check to see if a filesystem is mounted on a block device.
1291 */
1292int
e4c9c0c8 1293vfs_mountedon(struct vnode *vp)
984263bc 1294{
e4c9c0c8 1295 dev_t dev;
984263bc 1296
e4c9c0c8
MD
1297 if ((dev = vp->v_rdev) == NULL)
1298 dev = udev2dev(vp->v_udev, (vp->v_type == VBLK));
1299 if (dev != NODEV && dev->si_mountpoint)
984263bc
MD
1300 return (EBUSY);
1301 return (0);
1302}
1303
1304/*
1305 * Unmount all filesystems. The list is traversed in reverse order
1306 * of mounting to avoid dependencies.
1307 */
1308void
dd98570a 1309vfs_unmountall(void)
984263bc
MD
1310{
1311 struct mount *mp;
dadab5e9 1312 struct thread *td = curthread;
984263bc
MD
1313 int error;
1314
dadab5e9
MD
1315 if (td->td_proc == NULL)
1316 td = initproc->p_thread; /* XXX XXX use proc0 instead? */
1317
984263bc
MD
1318 /*
1319 * Since this only runs when rebooting, it is not interlocked.
1320 */
1321 while(!TAILQ_EMPTY(&mountlist)) {
1322 mp = TAILQ_LAST(&mountlist, mntlist);
dadab5e9 1323 error = dounmount(mp, MNT_FORCE, td);
984263bc
MD
1324 if (error) {
1325 TAILQ_REMOVE(&mountlist, mp, mnt_list);
1326 printf("unmount of %s failed (",
1327 mp->mnt_stat.f_mntonname);
1328 if (error == EBUSY)
1329 printf("BUSY)\n");
1330 else
1331 printf("%d)\n", error);
1332 } else {
1333 /* The unmount has removed mp from the mountlist */
1334 }
1335 }
1336}
1337
1338/*
1339 * Build hash lists of net addresses and hang them off the mount point.
1340 * Called by ufs_mount() to set up the lists of export addresses.
1341 */
1342static int
dd98570a
MD
1343vfs_hang_addrlist(struct mount *mp, struct netexport *nep,
1344 struct export_args *argp)
984263bc 1345{
1fd87d54
RG
1346 struct netcred *np;
1347 struct radix_node_head *rnh;
1348 int i;
984263bc
MD
1349 struct radix_node *rn;
1350 struct sockaddr *saddr, *smask = 0;
1351 struct domain *dom;
1352 int error;
1353
1354 if (argp->ex_addrlen == 0) {
1355 if (mp->mnt_flag & MNT_DEFEXPORTED)
1356 return (EPERM);
1357 np = &nep->ne_defexported;
1358 np->netc_exflags = argp->ex_flags;
1359 np->netc_anon = argp->ex_anon;
1360 np->netc_anon.cr_ref = 1;
1361 mp->mnt_flag |= MNT_DEFEXPORTED;
1362 return (0);
1363 }
1364
0260ddf9
MD
1365 if (argp->ex_addrlen < 0 || argp->ex_addrlen > MLEN)
1366 return (EINVAL);
1367 if (argp->ex_masklen < 0 || argp->ex_masklen > MLEN)
984263bc
MD
1368 return (EINVAL);
1369
1370 i = sizeof(struct netcred) + argp->ex_addrlen + argp->ex_masklen;
1371 np = (struct netcred *) malloc(i, M_NETADDR, M_WAITOK);
1372 bzero((caddr_t) np, i);
1373 saddr = (struct sockaddr *) (np + 1);
1374 if ((error = copyin(argp->ex_addr, (caddr_t) saddr, argp->ex_addrlen)))
1375 goto out;
1376 if (saddr->sa_len > argp->ex_addrlen)
1377 saddr->sa_len = argp->ex_addrlen;
1378 if (argp->ex_masklen) {
dd98570a
MD
1379 smask = (struct sockaddr *)((caddr_t)saddr + argp->ex_addrlen);
1380 error = copyin(argp->ex_mask, (caddr_t)smask, argp->ex_masklen);
984263bc
MD
1381 if (error)
1382 goto out;
1383 if (smask->sa_len > argp->ex_masklen)
1384 smask->sa_len = argp->ex_masklen;
1385 }
1386 i = saddr->sa_family;
1387 if ((rnh = nep->ne_rtable[i]) == 0) {
1388 /*
1389 * Seems silly to initialize every AF when most are not used,
1390 * do so on demand here
1391 */
1392 for (dom = domains; dom; dom = dom->dom_next)
1393 if (dom->dom_family == i && dom->dom_rtattach) {
1394 dom->dom_rtattach((void **) &nep->ne_rtable[i],
1395 dom->dom_rtoffset);
1396 break;
1397 }
1398 if ((rnh = nep->ne_rtable[i]) == 0) {
1399 error = ENOBUFS;
1400 goto out;
1401 }
1402 }
1403 rn = (*rnh->rnh_addaddr) ((caddr_t) saddr, (caddr_t) smask, rnh,
1404 np->netc_rnodes);
1405 if (rn == 0 || np != (struct netcred *) rn) { /* already exists */
1406 error = EPERM;
1407 goto out;
1408 }
1409 np->netc_exflags = argp->ex_flags;
1410 np->netc_anon = argp->ex_anon;
1411 np->netc_anon.cr_ref = 1;
1412 return (0);
1413out:
1414 free(np, M_NETADDR);
1415 return (error);
1416}
1417
1418/* ARGSUSED */
1419static int
dd98570a 1420vfs_free_netcred(struct radix_node *rn, void *w)
984263bc 1421{
1fd87d54 1422 struct radix_node_head *rnh = (struct radix_node_head *) w;
984263bc
MD
1423
1424 (*rnh->rnh_deladdr) (rn->rn_key, rn->rn_mask, rnh);
1425 free((caddr_t) rn, M_NETADDR);
1426 return (0);
1427}
1428
1429/*
1430 * Free the net address hash lists that are hanging off the mount points.
1431 */
1432static void
dd98570a 1433vfs_free_addrlist(struct netexport *nep)
984263bc 1434{
1fd87d54
RG
1435 int i;
1436 struct radix_node_head *rnh;
984263bc
MD
1437
1438 for (i = 0; i <= AF_MAX; i++)
1439 if ((rnh = nep->ne_rtable[i])) {
1440 (*rnh->rnh_walktree) (rnh, vfs_free_netcred,
1441 (caddr_t) rnh);
1442 free((caddr_t) rnh, M_RTABLE);
1443 nep->ne_rtable[i] = 0;
1444 }
1445}
1446
1447int
dd98570a 1448vfs_export(struct mount *mp, struct netexport *nep, struct export_args *argp)
984263bc
MD
1449{
1450 int error;
1451
1452 if (argp->ex_flags & MNT_DELEXPORT) {
1453 if (mp->mnt_flag & MNT_EXPUBLIC) {
1454 vfs_setpublicfs(NULL, NULL, NULL);
1455 mp->mnt_flag &= ~MNT_EXPUBLIC;
1456 }
1457 vfs_free_addrlist(nep);
1458 mp->mnt_flag &= ~(MNT_EXPORTED | MNT_DEFEXPORTED);
1459 }
1460 if (argp->ex_flags & MNT_EXPORTED) {
1461 if (argp->ex_flags & MNT_EXPUBLIC) {
1462 if ((error = vfs_setpublicfs(mp, nep, argp)) != 0)
1463 return (error);
1464 mp->mnt_flag |= MNT_EXPUBLIC;
1465 }
1466 if ((error = vfs_hang_addrlist(mp, nep, argp)))
1467 return (error);
1468 mp->mnt_flag |= MNT_EXPORTED;
1469 }
1470 return (0);
1471}
1472
1473
1474/*
1475 * Set the publicly exported filesystem (WebNFS). Currently, only
1476 * one public filesystem is possible in the spec (RFC 2054 and 2055)
1477 */
1478int
dd98570a
MD
1479vfs_setpublicfs(struct mount *mp, struct netexport *nep,
1480 struct export_args *argp)
984263bc
MD
1481{
1482 int error;
1483 struct vnode *rvp;
1484 char *cp;
1485
1486 /*
1487 * mp == NULL -> invalidate the current info, the FS is
1488 * no longer exported. May be called from either vfs_export
1489 * or unmount, so check if it hasn't already been done.
1490 */
1491 if (mp == NULL) {
1492 if (nfs_pub.np_valid) {
1493 nfs_pub.np_valid = 0;
1494 if (nfs_pub.np_index != NULL) {
1495 FREE(nfs_pub.np_index, M_TEMP);
1496 nfs_pub.np_index = NULL;
1497 }
1498 }
1499 return (0);
1500 }
1501
1502 /*
1503 * Only one allowed at a time.
1504 */
1505 if (nfs_pub.np_valid != 0 && mp != nfs_pub.np_mount)
1506 return (EBUSY);
1507
1508 /*
1509 * Get real filehandle for root of exported FS.
1510 */
1511 bzero((caddr_t)&nfs_pub.np_handle, sizeof(nfs_pub.np_handle));
1512 nfs_pub.np_handle.fh_fsid = mp->mnt_stat.f_fsid;
1513
1514 if ((error = VFS_ROOT(mp, &rvp)))
1515 return (error);
1516
1517 if ((error = VFS_VPTOFH(rvp, &nfs_pub.np_handle.fh_fid)))
1518 return (error);
1519
1520 vput(rvp);
1521
1522 /*
1523 * If an indexfile was specified, pull it in.
1524 */
1525 if (argp->ex_indexfile != NULL) {
1526 MALLOC(nfs_pub.np_index, char *, MAXNAMLEN + 1, M_TEMP,
1527 M_WAITOK);
1528 error = copyinstr(argp->ex_indexfile, nfs_pub.np_index,
1529 MAXNAMLEN, (size_t *)0);
1530 if (!error) {
1531 /*
1532 * Check for illegal filenames.
1533 */
1534 for (cp = nfs_pub.np_index; *cp; cp++) {
1535 if (*cp == '/') {
1536 error = EINVAL;
1537 break;
1538 }
1539 }
1540 }
1541 if (error) {
1542 FREE(nfs_pub.np_index, M_TEMP);
1543 return (error);
1544 }
1545 }
1546
1547 nfs_pub.np_mount = mp;
1548 nfs_pub.np_valid = 1;
1549 return (0);
1550}
1551
1552struct netcred *
dd98570a
MD
1553vfs_export_lookup(struct mount *mp, struct netexport *nep,
1554 struct sockaddr *nam)
984263bc 1555{
1fd87d54
RG
1556 struct netcred *np;
1557 struct radix_node_head *rnh;
984263bc
MD
1558 struct sockaddr *saddr;
1559
1560 np = NULL;
1561 if (mp->mnt_flag & MNT_EXPORTED) {
1562 /*
1563 * Lookup in the export list first.
1564 */
1565 if (nam != NULL) {
1566 saddr = nam;
1567 rnh = nep->ne_rtable[saddr->sa_family];
1568 if (rnh != NULL) {
1569 np = (struct netcred *)
1570 (*rnh->rnh_matchaddr)((caddr_t)saddr,
1571 rnh);
1572 if (np && np->netc_rnodes->rn_flags & RNF_ROOT)
1573 np = NULL;
1574 }
1575 }
1576 /*
1577 * If no address match, use the default if it exists.
1578 */
1579 if (np == NULL && mp->mnt_flag & MNT_DEFEXPORTED)
1580 np = &nep->ne_defexported;
1581 }
1582 return (np);
1583}
1584
1585/*
41a01a4d
MD
1586 * perform msync on all vnodes under a mount point. The mount point must
1587 * be locked. This code is also responsible for lazy-freeing unreferenced
1588 * vnodes whos VM objects no longer contain pages.
1589 *
1590 * NOTE: MNT_WAIT still skips vnodes in the VXLOCK state.
984263bc 1591 */
41a01a4d 1592static int vfs_msync_scan1(struct mount *mp, struct vnode *vp, void *data);
5fd012e0 1593static int vfs_msync_scan2(struct mount *mp, struct vnode *vp, void *data);
41a01a4d 1594
984263bc
MD
1595void
1596vfs_msync(struct mount *mp, int flags)
1597{
5fd012e0
MD
1598 vmntvnodescan(mp, VMSC_REFVP, vfs_msync_scan1, vfs_msync_scan2,
1599 (void *)flags);
41a01a4d 1600}
984263bc 1601
41a01a4d
MD
1602/*
1603 * scan1 is a fast pre-check. There could be hundreds of thousands of
1604 * vnodes, we cannot afford to do anything heavy weight until we have a
1605 * fairly good indication that there is work to do.
1606 */
1607static
1608int
1609vfs_msync_scan1(struct mount *mp, struct vnode *vp, void *data)
1610{
1611 int flags = (int)data;
984263bc 1612
5fd012e0
MD
1613 if ((vp->v_flag & VRECLAIMED) == 0) {
1614 if (vshouldfree(vp, 0))
1615 return(0); /* call scan2 */
41a01a4d
MD
1616 if ((mp->mnt_flag & MNT_RDONLY) == 0 &&
1617 (vp->v_flag & VOBJDIRTY) &&
984263bc 1618 (flags == MNT_WAIT || VOP_ISLOCKED(vp, NULL) == 0)) {
5fd012e0 1619 return(0); /* call scan2 */
41a01a4d
MD
1620 }
1621 }
5fd012e0
MD
1622
1623 /*
1624 * do not call scan2, continue the loop
1625 */
41a01a4d
MD
1626 return(-1);
1627}
1628
1629static
1630int
5fd012e0 1631vfs_msync_scan2(struct mount *mp, struct vnode *vp, void *data)
41a01a4d
MD
1632{
1633 vm_object_t obj;
41a01a4d
MD
1634 int flags = (int)data;
1635
5fd012e0 1636 if (vp->v_flag & VRECLAIMED)
41a01a4d
MD
1637 return(0);
1638
1639 if ((mp->mnt_flag & MNT_RDONLY) == 0 &&
1640 (vp->v_flag & VOBJDIRTY) &&
1641 (flags == MNT_WAIT || VOP_ISLOCKED(vp, NULL) == 0)) {
5fd012e0
MD
1642 if (VOP_GETVOBJECT(vp, &obj) == 0) {
1643 vm_object_page_clean(obj, 0, 0,
1644 flags == MNT_WAIT ? OBJPC_SYNC : OBJPC_NOSYNC);
984263bc
MD
1645 }
1646 }
41a01a4d 1647 return(0);
984263bc
MD
1648}
1649
1650/*
1651 * Create the VM object needed for VMIO and mmap support. This
1652 * is done for all VREG files in the system. Some filesystems might
1653 * afford the additional metadata buffering capability of the
1654 * VMIO code by making the device node be VMIO mode also.
1655 *
1656 * vp must be locked when vfs_object_create is called.
1657 */
1658int
3b568787 1659vfs_object_create(struct vnode *vp, struct thread *td)
984263bc 1660{
3b568787 1661 return (VOP_CREATEVOBJECT(vp, td));
984263bc
MD
1662}
1663
984263bc
MD
1664/*
1665 * Record a process's interest in events which might happen to
1666 * a vnode. Because poll uses the historic select-style interface
1667 * internally, this routine serves as both the ``check for any
1668 * pending events'' and the ``record my interest in future events''
1669 * functions. (These are done together, while the lock is held,
1670 * to avoid race conditions.)
1671 */
1672int
dadab5e9 1673vn_pollrecord(struct vnode *vp, struct thread *td, int events)
984263bc 1674{
41a01a4d
MD
1675 lwkt_tokref ilock;
1676
1677 lwkt_gettoken(&ilock, &vp->v_pollinfo.vpi_token);
984263bc
MD
1678 if (vp->v_pollinfo.vpi_revents & events) {
1679 /*
1680 * This leaves events we are not interested
1681 * in available for the other process which
1682 * which presumably had requested them
1683 * (otherwise they would never have been
1684 * recorded).
1685 */
1686 events &= vp->v_pollinfo.vpi_revents;
1687 vp->v_pollinfo.vpi_revents &= ~events;
1688
41a01a4d 1689 lwkt_reltoken(&ilock);
984263bc
MD
1690 return events;
1691 }
1692 vp->v_pollinfo.vpi_events |= events;
dadab5e9 1693 selrecord(td, &vp->v_pollinfo.vpi_selinfo);
41a01a4d 1694 lwkt_reltoken(&ilock);
984263bc
MD
1695 return 0;
1696}
1697
1698/*
1699 * Note the occurrence of an event. If the VN_POLLEVENT macro is used,
1700 * it is possible for us to miss an event due to race conditions, but
1701 * that condition is expected to be rare, so for the moment it is the
1702 * preferred interface.
1703 */
1704void
dd98570a 1705vn_pollevent(struct vnode *vp, int events)
984263bc 1706{
41a01a4d
MD
1707 lwkt_tokref ilock;
1708
1709 lwkt_gettoken(&ilock, &vp->v_pollinfo.vpi_token);
984263bc
MD
1710 if (vp->v_pollinfo.vpi_events & events) {
1711 /*
1712 * We clear vpi_events so that we don't
1713 * call selwakeup() twice if two events are
1714 * posted before the polling process(es) is
1715 * awakened. This also ensures that we take at
1716 * most one selwakeup() if the polling process
1717 * is no longer interested. However, it does
1718 * mean that only one event can be noticed at
1719 * a time. (Perhaps we should only clear those
1720 * event bits which we note?) XXX
1721 */
1722 vp->v_pollinfo.vpi_events = 0; /* &= ~events ??? */
1723 vp->v_pollinfo.vpi_revents |= events;
1724 selwakeup(&vp->v_pollinfo.vpi_selinfo);
1725 }
41a01a4d 1726 lwkt_reltoken(&ilock);
984263bc
MD
1727}
1728
1729/*
1730 * Wake up anyone polling on vp because it is being revoked.
1731 * This depends on dead_poll() returning POLLHUP for correct
1732 * behavior.
1733 */
1734void
dd98570a 1735vn_pollgone(struct vnode *vp)
984263bc 1736{
41a01a4d
MD
1737 lwkt_tokref ilock;
1738
1739 lwkt_gettoken(&ilock, &vp->v_pollinfo.vpi_token);
984263bc
MD
1740 if (vp->v_pollinfo.vpi_events) {
1741 vp->v_pollinfo.vpi_events = 0;
1742 selwakeup(&vp->v_pollinfo.vpi_selinfo);
1743 }
41a01a4d 1744 lwkt_reltoken(&ilock);
984263bc
MD
1745}
1746
984263bc 1747/*
e4c9c0c8
MD
1748 * extract the dev_t from a VBLK or VCHR. The vnode must have been opened
1749 * (or v_rdev might be NULL).
984263bc
MD
1750 */
1751dev_t
e4c9c0c8 1752vn_todev(struct vnode *vp)
984263bc
MD
1753{
1754 if (vp->v_type != VBLK && vp->v_type != VCHR)
1755 return (NODEV);
e4c9c0c8 1756 KKASSERT(vp->v_rdev != NULL);
984263bc
MD
1757 return (vp->v_rdev);
1758}
1759
1760/*
e4c9c0c8
MD
1761 * Check if vnode represents a disk device. The vnode does not need to be
1762 * opened.
984263bc
MD
1763 */
1764int
e4c9c0c8 1765vn_isdisk(struct vnode *vp, int *errp)
984263bc 1766{
e4c9c0c8
MD
1767 dev_t dev;
1768
984263bc
MD
1769 if (vp->v_type != VBLK && vp->v_type != VCHR) {
1770 if (errp != NULL)
1771 *errp = ENOTBLK;
1772 return (0);
1773 }
e4c9c0c8
MD
1774
1775 if ((dev = vp->v_rdev) == NULL)
1776 dev = udev2dev(vp->v_udev, (vp->v_type == VBLK));
1777 if (dev == NULL || dev == NODEV) {
984263bc
MD
1778 if (errp != NULL)
1779 *errp = ENXIO;
1780 return (0);
1781 }
e4c9c0c8 1782 if (dev_is_good(dev) == 0) {
984263bc
MD
1783 if (errp != NULL)
1784 *errp = ENXIO;
1785 return (0);
1786 }
e4c9c0c8 1787 if ((dev_dflags(dev) & D_DISK) == 0) {
984263bc
MD
1788 if (errp != NULL)
1789 *errp = ENOTBLK;
1790 return (0);
1791 }
1792 if (errp != NULL)
1793 *errp = 0;
1794 return (1);
1795}
1796
1797void
dd98570a 1798NDFREE(struct nameidata *ndp, const uint flags)
984263bc
MD
1799{
1800 if (!(flags & NDF_NO_FREE_PNBUF) &&
2b69e610 1801 (ndp->ni_cnd.cn_flags & CNP_HASBUF)) {
984263bc 1802 zfree(namei_zone, ndp->ni_cnd.cn_pnbuf);
2b69e610 1803 ndp->ni_cnd.cn_flags &= ~CNP_HASBUF;
984263bc
MD
1804 }
1805 if (!(flags & NDF_NO_DVP_UNLOCK) &&
2b69e610 1806 (ndp->ni_cnd.cn_flags & CNP_LOCKPARENT) &&
bc0c094e 1807 ndp->ni_dvp != ndp->ni_vp) {
5fd012e0 1808 VOP_UNLOCK(ndp->ni_dvp, 0, ndp->ni_cnd.cn_td);
bc0c094e 1809 }
984263bc 1810 if (!(flags & NDF_NO_DVP_RELE) &&
2b69e610 1811 (ndp->ni_cnd.cn_flags & (CNP_LOCKPARENT|CNP_WANTPARENT))) {
984263bc
MD
1812 vrele(ndp->ni_dvp);
1813 ndp->ni_dvp = NULL;
1814 }
1815 if (!(flags & NDF_NO_VP_UNLOCK) &&
bc0c094e 1816 (ndp->ni_cnd.cn_flags & CNP_LOCKLEAF) && ndp->ni_vp) {
5fd012e0 1817 VOP_UNLOCK(ndp->ni_vp, 0, ndp->ni_cnd.cn_td);
bc0c094e 1818 }
984263bc
MD
1819 if (!(flags & NDF_NO_VP_RELE) &&
1820 ndp->ni_vp) {
1821 vrele(ndp->ni_vp);
1822 ndp->ni_vp = NULL;
1823 }
1824 if (!(flags & NDF_NO_STARTDIR_RELE) &&
2b69e610 1825 (ndp->ni_cnd.cn_flags & CNP_SAVESTART)) {
984263bc
MD
1826 vrele(ndp->ni_startdir);
1827 ndp->ni_startdir = NULL;
1828 }
1829}
bc0c094e 1830
4f322a84
EN
1831#ifdef DEBUG_VFS_LOCKS
1832
1833void
1834assert_vop_locked(struct vnode *vp, const char *str)
1835{
4f322a84
EN
1836 if (vp && IS_LOCKING_VFS(vp) && !VOP_ISLOCKED(vp, NULL)) {
1837 panic("%s: %p is not locked shared but should be", str, vp);
1838 }
1839}
1840
1841void
1842assert_vop_unlocked(struct vnode *vp, const char *str)
1843{
4f322a84
EN
1844 if (vp && IS_LOCKING_VFS(vp)) {
1845 if (VOP_ISLOCKED(vp, curthread) == LK_EXCLUSIVE) {
1846 panic("%s: %p is locked but should not be", str, vp);
1847 }
1848 }
1849}
1850
1851#endif