xref: /netbsd-src/sys/ufs/lfs/lfs_vnops.c (revision bdc22b2e01993381dcefeff2bc9b56ca75a4235c)
1 /*	$NetBSD: lfs_vnops.c,v 1.321 2017/08/20 05:37:03 maya Exp $	*/
2 
3 /*-
4  * Copyright (c) 1999, 2000, 2001, 2002, 2003 The NetBSD Foundation, Inc.
5  * All rights reserved.
6  *
7  * This code is derived from software contributed to The NetBSD Foundation
8  * by Konrad E. Schroder <perseant@hhhh.org>.
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  *
19  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
20  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
21  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
22  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
23  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
24  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
25  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
26  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
27  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
28  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
29  * POSSIBILITY OF SUCH DAMAGE.
30  */
31 /*
32  * Copyright (c) 1986, 1989, 1991, 1993, 1995
33  *	The Regents of the University of California.  All rights reserved.
34  *
35  * Redistribution and use in source and binary forms, with or without
36  * modification, are permitted provided that the following conditions
37  * are met:
38  * 1. Redistributions of source code must retain the above copyright
39  *    notice, this list of conditions and the following disclaimer.
40  * 2. Redistributions in binary form must reproduce the above copyright
41  *    notice, this list of conditions and the following disclaimer in the
42  *    documentation and/or other materials provided with the distribution.
43  * 3. Neither the name of the University nor the names of its contributors
44  *    may be used to endorse or promote products derived from this software
45  *    without specific prior written permission.
46  *
47  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
48  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
49  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
50  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
51  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
52  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
53  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
54  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
55  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
56  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
57  * SUCH DAMAGE.
58  *
59  *	@(#)lfs_vnops.c	8.13 (Berkeley) 6/10/95
60  */
61 
62 /*  from NetBSD: ufs_vnops.c,v 1.232 2016/05/19 18:32:03 riastradh Exp  */
63 /*-
64  * Copyright (c) 2008 The NetBSD Foundation, Inc.
65  * All rights reserved.
66  *
67  * This code is derived from software contributed to The NetBSD Foundation
68  * by Wasabi Systems, Inc.
69  *
70  * Redistribution and use in source and binary forms, with or without
71  * modification, are permitted provided that the following conditions
72  * are met:
73  * 1. Redistributions of source code must retain the above copyright
74  *    notice, this list of conditions and the following disclaimer.
75  * 2. Redistributions in binary form must reproduce the above copyright
76  *    notice, this list of conditions and the following disclaimer in the
77  *    documentation and/or other materials provided with the distribution.
78  *
79  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
80  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
81  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
82  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
83  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
84  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
85  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
86  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
87  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
88  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
89  * POSSIBILITY OF SUCH DAMAGE.
90  */
91 /*
92  * Copyright (c) 1982, 1986, 1989, 1993, 1995
93  *	The Regents of the University of California.  All rights reserved.
94  * (c) UNIX System Laboratories, Inc.
95  * All or some portions of this file are derived from material licensed
96  * to the University of California by American Telephone and Telegraph
97  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
98  * the permission of UNIX System Laboratories, Inc.
99  *
100  * Redistribution and use in source and binary forms, with or without
101  * modification, are permitted provided that the following conditions
102  * are met:
103  * 1. Redistributions of source code must retain the above copyright
104  *    notice, this list of conditions and the following disclaimer.
105  * 2. Redistributions in binary form must reproduce the above copyright
106  *    notice, this list of conditions and the following disclaimer in the
107  *    documentation and/or other materials provided with the distribution.
108  * 3. Neither the name of the University nor the names of its contributors
109  *    may be used to endorse or promote products derived from this software
110  *    without specific prior written permission.
111  *
112  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
113  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
114  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
115  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
116  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
117  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
118  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
119  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
120  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
121  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
122  * SUCH DAMAGE.
123  *
124  *	@(#)ufs_vnops.c	8.28 (Berkeley) 7/31/95
125  */
126 
127 #include <sys/cdefs.h>
128 __KERNEL_RCSID(0, "$NetBSD: lfs_vnops.c,v 1.321 2017/08/20 05:37:03 maya Exp $");
129 
130 #ifdef _KERNEL_OPT
131 #include "opt_compat_netbsd.h"
132 #include "opt_uvm_page_trkown.h"
133 #endif
134 
135 #include <sys/param.h>
136 #include <sys/systm.h>
137 #include <sys/namei.h>
138 #include <sys/resourcevar.h>
139 #include <sys/kernel.h>
140 #include <sys/file.h>
141 #include <sys/stat.h>
142 #include <sys/buf.h>
143 #include <sys/proc.h>
144 #include <sys/mount.h>
145 #include <sys/vnode.h>
146 #include <sys/pool.h>
147 #include <sys/signalvar.h>
148 #include <sys/kauth.h>
149 #include <sys/syslog.h>
150 
151 #include <miscfs/fifofs/fifo.h>
152 #include <miscfs/genfs/genfs.h>
153 #include <miscfs/specfs/specdev.h>
154 
155 #include <ufs/lfs/ulfs_inode.h>
156 #include <ufs/lfs/ulfsmount.h>
157 #include <ufs/lfs/ulfs_bswap.h>
158 #include <ufs/lfs/ulfs_extern.h>
159 
160 #include <uvm/uvm.h>
161 #include <uvm/uvm_pmap.h>
162 #include <uvm/uvm_stat.h>
163 #include <uvm/uvm_pager.h>
164 
165 #include <ufs/lfs/lfs.h>
166 #include <ufs/lfs/lfs_accessors.h>
167 #include <ufs/lfs/lfs_kernel.h>
168 #include <ufs/lfs/lfs_extern.h>
169 
170 extern kcondvar_t lfs_writerd_cv;
171 int lfs_ignore_lazy_sync = 1;
172 
173 static int lfs_openextattr(void *v);
174 static int lfs_closeextattr(void *v);
175 static int lfs_getextattr(void *v);
176 static int lfs_setextattr(void *v);
177 static int lfs_listextattr(void *v);
178 static int lfs_deleteextattr(void *v);
179 
180 static int lfs_makeinode(struct vattr *vap, struct vnode *,
181 		      const struct ulfs_lookup_results *,
182 		      struct vnode **, struct componentname *);
183 
184 /* Global vfs data structures for lfs. */
185 int (**lfs_vnodeop_p)(void *);
186 const struct vnodeopv_entry_desc lfs_vnodeop_entries[] = {
187 	{ &vop_default_desc, vn_default_error },
188 	{ &vop_lookup_desc, ulfs_lookup },		/* lookup */
189 	{ &vop_create_desc, lfs_create },		/* create */
190 	{ &vop_whiteout_desc, ulfs_whiteout },		/* whiteout */
191 	{ &vop_mknod_desc, lfs_mknod },			/* mknod */
192 	{ &vop_open_desc, ulfs_open },			/* open */
193 	{ &vop_close_desc, lfs_close },			/* close */
194 	{ &vop_access_desc, ulfs_access },		/* access */
195 	{ &vop_getattr_desc, lfs_getattr },		/* getattr */
196 	{ &vop_setattr_desc, lfs_setattr },		/* setattr */
197 	{ &vop_read_desc, lfs_read },			/* read */
198 	{ &vop_write_desc, lfs_write },			/* write */
199 	{ &vop_fallocate_desc, genfs_eopnotsupp },	/* fallocate */
200 	{ &vop_fdiscard_desc, genfs_eopnotsupp },	/* fdiscard */
201 	{ &vop_ioctl_desc, ulfs_ioctl },		/* ioctl */
202 	{ &vop_fcntl_desc, lfs_fcntl },			/* fcntl */
203 	{ &vop_poll_desc, ulfs_poll },			/* poll */
204 	{ &vop_kqfilter_desc, genfs_kqfilter },		/* kqfilter */
205 	{ &vop_revoke_desc, ulfs_revoke },		/* revoke */
206 	{ &vop_mmap_desc, lfs_mmap },			/* mmap */
207 	{ &vop_fsync_desc, lfs_fsync },			/* fsync */
208 	{ &vop_seek_desc, ulfs_seek },			/* seek */
209 	{ &vop_remove_desc, lfs_remove },		/* remove */
210 	{ &vop_link_desc, lfs_link },			/* link */
211 	{ &vop_rename_desc, lfs_rename },		/* rename */
212 	{ &vop_mkdir_desc, lfs_mkdir },			/* mkdir */
213 	{ &vop_rmdir_desc, lfs_rmdir },			/* rmdir */
214 	{ &vop_symlink_desc, lfs_symlink },		/* symlink */
215 	{ &vop_readdir_desc, ulfs_readdir },		/* readdir */
216 	{ &vop_readlink_desc, ulfs_readlink },		/* readlink */
217 	{ &vop_abortop_desc, ulfs_abortop },		/* abortop */
218 	{ &vop_inactive_desc, lfs_inactive },		/* inactive */
219 	{ &vop_reclaim_desc, lfs_reclaim },		/* reclaim */
220 	{ &vop_lock_desc, ulfs_lock },			/* lock */
221 	{ &vop_unlock_desc, ulfs_unlock },		/* unlock */
222 	{ &vop_bmap_desc, ulfs_bmap },			/* bmap */
223 	{ &vop_strategy_desc, lfs_strategy },		/* strategy */
224 	{ &vop_print_desc, ulfs_print },		/* print */
225 	{ &vop_islocked_desc, ulfs_islocked },		/* islocked */
226 	{ &vop_pathconf_desc, ulfs_pathconf },		/* pathconf */
227 	{ &vop_advlock_desc, ulfs_advlock },		/* advlock */
228 	{ &vop_bwrite_desc, lfs_bwrite },		/* bwrite */
229 	{ &vop_getpages_desc, lfs_getpages },		/* getpages */
230 	{ &vop_putpages_desc, lfs_putpages },		/* putpages */
231 	{ &vop_openextattr_desc, lfs_openextattr },	/* openextattr */
232 	{ &vop_closeextattr_desc, lfs_closeextattr },	/* closeextattr */
233 	{ &vop_getextattr_desc, lfs_getextattr },	/* getextattr */
234 	{ &vop_setextattr_desc, lfs_setextattr },	/* setextattr */
235 	{ &vop_listextattr_desc, lfs_listextattr },	/* listextattr */
236 	{ &vop_deleteextattr_desc, lfs_deleteextattr },	/* deleteextattr */
237 	{ NULL, NULL }
238 };
239 const struct vnodeopv_desc lfs_vnodeop_opv_desc =
240 	{ &lfs_vnodeop_p, lfs_vnodeop_entries };
241 
242 int (**lfs_specop_p)(void *);
243 const struct vnodeopv_entry_desc lfs_specop_entries[] = {
244 	{ &vop_default_desc, vn_default_error },
245 	{ &vop_lookup_desc, spec_lookup },		/* lookup */
246 	{ &vop_create_desc, spec_create },		/* create */
247 	{ &vop_mknod_desc, spec_mknod },		/* mknod */
248 	{ &vop_open_desc, spec_open },			/* open */
249 	{ &vop_close_desc, lfsspec_close },		/* close */
250 	{ &vop_access_desc, ulfs_access },		/* access */
251 	{ &vop_getattr_desc, lfs_getattr },		/* getattr */
252 	{ &vop_setattr_desc, lfs_setattr },		/* setattr */
253 	{ &vop_read_desc, ulfsspec_read },		/* read */
254 	{ &vop_write_desc, ulfsspec_write },		/* write */
255 	{ &vop_fallocate_desc, spec_fallocate },	/* fallocate */
256 	{ &vop_fdiscard_desc, spec_fdiscard },		/* fdiscard */
257 	{ &vop_ioctl_desc, spec_ioctl },		/* ioctl */
258 	{ &vop_fcntl_desc, ulfs_fcntl },		/* fcntl */
259 	{ &vop_poll_desc, spec_poll },			/* poll */
260 	{ &vop_kqfilter_desc, spec_kqfilter },		/* kqfilter */
261 	{ &vop_revoke_desc, spec_revoke },		/* revoke */
262 	{ &vop_mmap_desc, spec_mmap },			/* mmap */
263 	{ &vop_fsync_desc, spec_fsync },		/* fsync */
264 	{ &vop_seek_desc, spec_seek },			/* seek */
265 	{ &vop_remove_desc, spec_remove },		/* remove */
266 	{ &vop_link_desc, spec_link },			/* link */
267 	{ &vop_rename_desc, spec_rename },		/* rename */
268 	{ &vop_mkdir_desc, spec_mkdir },		/* mkdir */
269 	{ &vop_rmdir_desc, spec_rmdir },		/* rmdir */
270 	{ &vop_symlink_desc, spec_symlink },		/* symlink */
271 	{ &vop_readdir_desc, spec_readdir },		/* readdir */
272 	{ &vop_readlink_desc, spec_readlink },		/* readlink */
273 	{ &vop_abortop_desc, spec_abortop },		/* abortop */
274 	{ &vop_inactive_desc, lfs_inactive },		/* inactive */
275 	{ &vop_reclaim_desc, lfs_reclaim },		/* reclaim */
276 	{ &vop_lock_desc, ulfs_lock },			/* lock */
277 	{ &vop_unlock_desc, ulfs_unlock },		/* unlock */
278 	{ &vop_bmap_desc, spec_bmap },			/* bmap */
279 	{ &vop_strategy_desc, spec_strategy },		/* strategy */
280 	{ &vop_print_desc, ulfs_print },		/* print */
281 	{ &vop_islocked_desc, ulfs_islocked },		/* islocked */
282 	{ &vop_pathconf_desc, spec_pathconf },		/* pathconf */
283 	{ &vop_advlock_desc, spec_advlock },		/* advlock */
284 	{ &vop_bwrite_desc, vn_bwrite },		/* bwrite */
285 	{ &vop_getpages_desc, spec_getpages },		/* getpages */
286 	{ &vop_putpages_desc, spec_putpages },		/* putpages */
287 	{ &vop_openextattr_desc, lfs_openextattr },	/* openextattr */
288 	{ &vop_closeextattr_desc, lfs_closeextattr },	/* closeextattr */
289 	{ &vop_getextattr_desc, lfs_getextattr },	/* getextattr */
290 	{ &vop_setextattr_desc, lfs_setextattr },	/* setextattr */
291 	{ &vop_listextattr_desc, lfs_listextattr },	/* listextattr */
292 	{ &vop_deleteextattr_desc, lfs_deleteextattr },	/* deleteextattr */
293 	{ NULL, NULL }
294 };
295 const struct vnodeopv_desc lfs_specop_opv_desc =
296 	{ &lfs_specop_p, lfs_specop_entries };
297 
298 int (**lfs_fifoop_p)(void *);
299 const struct vnodeopv_entry_desc lfs_fifoop_entries[] = {
300 	{ &vop_default_desc, vn_default_error },
301 	{ &vop_lookup_desc, vn_fifo_bypass },		/* lookup */
302 	{ &vop_create_desc, vn_fifo_bypass },		/* create */
303 	{ &vop_mknod_desc, vn_fifo_bypass },		/* mknod */
304 	{ &vop_open_desc, vn_fifo_bypass },		/* open */
305 	{ &vop_close_desc, lfsfifo_close },		/* close */
306 	{ &vop_access_desc, ulfs_access },		/* access */
307 	{ &vop_getattr_desc, lfs_getattr },		/* getattr */
308 	{ &vop_setattr_desc, lfs_setattr },		/* setattr */
309 	{ &vop_read_desc, ulfsfifo_read },		/* read */
310 	{ &vop_write_desc, ulfsfifo_write },		/* write */
311 	{ &vop_fallocate_desc, vn_fifo_bypass },	/* fallocate */
312 	{ &vop_fdiscard_desc, vn_fifo_bypass },		/* fdiscard */
313 	{ &vop_ioctl_desc, vn_fifo_bypass },		/* ioctl */
314 	{ &vop_fcntl_desc, ulfs_fcntl },		/* fcntl */
315 	{ &vop_poll_desc, vn_fifo_bypass },		/* poll */
316 	{ &vop_kqfilter_desc, vn_fifo_bypass },		/* kqfilter */
317 	{ &vop_revoke_desc, vn_fifo_bypass },		/* revoke */
318 	{ &vop_mmap_desc, vn_fifo_bypass },		/* mmap */
319 	{ &vop_fsync_desc, vn_fifo_bypass },		/* fsync */
320 	{ &vop_seek_desc, vn_fifo_bypass },		/* seek */
321 	{ &vop_remove_desc, vn_fifo_bypass },		/* remove */
322 	{ &vop_link_desc, vn_fifo_bypass },		/* link */
323 	{ &vop_rename_desc, vn_fifo_bypass },		/* rename */
324 	{ &vop_mkdir_desc, vn_fifo_bypass },		/* mkdir */
325 	{ &vop_rmdir_desc, vn_fifo_bypass },		/* rmdir */
326 	{ &vop_symlink_desc, vn_fifo_bypass },		/* symlink */
327 	{ &vop_readdir_desc, vn_fifo_bypass },		/* readdir */
328 	{ &vop_readlink_desc, vn_fifo_bypass },		/* readlink */
329 	{ &vop_abortop_desc, vn_fifo_bypass },		/* abortop */
330 	{ &vop_inactive_desc, lfs_inactive },		/* inactive */
331 	{ &vop_reclaim_desc, lfs_reclaim },		/* reclaim */
332 	{ &vop_lock_desc, ulfs_lock },			/* lock */
333 	{ &vop_unlock_desc, ulfs_unlock },		/* unlock */
334 	{ &vop_bmap_desc, vn_fifo_bypass },		/* bmap */
335 	{ &vop_strategy_desc, vn_fifo_bypass },		/* strategy */
336 	{ &vop_print_desc, ulfs_print },		/* print */
337 	{ &vop_islocked_desc, ulfs_islocked },		/* islocked */
338 	{ &vop_pathconf_desc, vn_fifo_bypass },		/* pathconf */
339 	{ &vop_advlock_desc, vn_fifo_bypass },		/* advlock */
340 	{ &vop_bwrite_desc, lfs_bwrite },		/* bwrite */
341 	{ &vop_putpages_desc, vn_fifo_bypass },		/* putpages */
342 	{ &vop_openextattr_desc, lfs_openextattr },	/* openextattr */
343 	{ &vop_closeextattr_desc, lfs_closeextattr },	/* closeextattr */
344 	{ &vop_getextattr_desc, lfs_getextattr },	/* getextattr */
345 	{ &vop_setextattr_desc, lfs_setextattr },	/* setextattr */
346 	{ &vop_listextattr_desc, lfs_listextattr },	/* listextattr */
347 	{ &vop_deleteextattr_desc, lfs_deleteextattr },	/* deleteextattr */
348 	{ NULL, NULL }
349 };
350 const struct vnodeopv_desc lfs_fifoop_opv_desc =
351 	{ &lfs_fifoop_p, lfs_fifoop_entries };
352 
353 #define	LFS_READWRITE
354 #include <ufs/lfs/ulfs_readwrite.c>
355 #undef	LFS_READWRITE
356 
357 /*
358  * Allocate a new inode.
359  */
360 static int
361 lfs_makeinode(struct vattr *vap, struct vnode *dvp,
362 	const struct ulfs_lookup_results *ulr,
363 	struct vnode **vpp, struct componentname *cnp)
364 {
365 	struct inode	*ip;
366 	struct vnode	*tvp;
367 	int		error;
368 
369 	error = vcache_new(dvp->v_mount, dvp, vap, cnp->cn_cred, &tvp);
370 	if (error)
371 		return error;
372 	error = vn_lock(tvp, LK_EXCLUSIVE);
373 	if (error) {
374 		vrele(tvp);
375 		return error;
376 	}
377 	MARK_VNODE(tvp);
378 	*vpp = tvp;
379 	ip = VTOI(tvp);
380 	ip->i_state |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
381 	ip->i_nlink = 1;
382 	DIP_ASSIGN(ip, nlink, 1);
383 
384 	/* Authorize setting SGID if needed. */
385 	if (ip->i_mode & ISGID) {
386 		error = kauth_authorize_vnode(cnp->cn_cred, KAUTH_VNODE_WRITE_SECURITY,
387 		    tvp, NULL, genfs_can_chmod(tvp->v_type, cnp->cn_cred, ip->i_uid,
388 		    ip->i_gid, MAKEIMODE(vap->va_type, vap->va_mode)));
389 		if (error) {
390 			ip->i_mode &= ~ISGID;
391 			DIP_ASSIGN(ip, mode, ip->i_mode);
392 		}
393 	}
394 
395 	if (cnp->cn_flags & ISWHITEOUT) {
396 		ip->i_flags |= UF_OPAQUE;
397 		DIP_ASSIGN(ip, flags, ip->i_flags);
398 	}
399 
400 	/*
401 	 * Make sure inode goes to disk before directory entry.
402 	 */
403 	if ((error = lfs_update(tvp, NULL, NULL, UPDATE_DIROP)) != 0)
404 		goto bad;
405 	error = ulfs_direnter(dvp, ulr, tvp,
406 			      cnp, ip->i_number, LFS_IFTODT(ip->i_mode), NULL);
407 	if (error)
408 		goto bad;
409 	*vpp = tvp;
410 	KASSERT(VOP_ISLOCKED(*vpp) == LK_EXCLUSIVE);
411 	return (0);
412 
413  bad:
414 	/*
415 	 * Write error occurred trying to update the inode
416 	 * or the directory so must deallocate the inode.
417 	 */
418 	ip->i_nlink = 0;
419 	DIP_ASSIGN(ip, nlink, 0);
420 	ip->i_state |= IN_CHANGE;
421 	/* If IN_ADIROP, account for it */
422 	UNMARK_VNODE(tvp);
423 	vput(tvp);
424 	return (error);
425 }
426 
427 /*
428  * Synch an open file.
429  */
430 /* ARGSUSED */
431 int
432 lfs_fsync(void *v)
433 {
434 	struct vop_fsync_args /* {
435 		struct vnode *a_vp;
436 		kauth_cred_t a_cred;
437 		int a_flags;
438 		off_t offlo;
439 		off_t offhi;
440 	} */ *ap = v;
441 	struct vnode *vp = ap->a_vp;
442 	int wait;
443 	struct inode *ip = VTOI(vp);
444 	struct lfs *fs = ip->i_lfs;
445 	int error = 0;
446 
447 	KASSERT(VOP_ISLOCKED(vp) == LK_EXCLUSIVE);
448 
449 	/* If we're mounted read-only, don't try to sync. */
450 	if (fs->lfs_ronly)
451 		goto out;
452 
453 	/* If a removed vnode is being cleaned, no need to sync here. */
454 	if ((ap->a_flags & FSYNC_RECLAIM) != 0 && ip->i_mode == 0)
455 		goto out;
456 
457 	/*
458 	 * Trickle sync simply adds this vnode to the pager list, as if
459 	 * the pagedaemon had requested a pageout.
460 	 */
461 	if (ap->a_flags & FSYNC_LAZY) {
462 		if (lfs_ignore_lazy_sync == 0) {
463 			mutex_enter(&lfs_lock);
464 			if (!(ip->i_state & IN_PAGING)) {
465 				ip->i_state |= IN_PAGING;
466 				TAILQ_INSERT_TAIL(&fs->lfs_pchainhd, ip,
467 						  i_lfs_pchain);
468 			}
469 			cv_broadcast(&lfs_writerd_cv);
470 			mutex_exit(&lfs_lock);
471 		}
472 		goto out;
473 	}
474 
475 	/*
476 	 * If a vnode is being cleaned, flush it out before we try to
477 	 * reuse it.  This prevents the cleaner from writing files twice
478 	 * in the same partial segment, causing an accounting underflow.
479 	 */
480 	if (ap->a_flags & FSYNC_RECLAIM && ip->i_state & IN_CLEANING) {
481 		lfs_vflush(vp);
482 	}
483 
484 	wait = (ap->a_flags & FSYNC_WAIT);
485 	do {
486 		mutex_enter(vp->v_interlock);
487 		error = VOP_PUTPAGES(vp, trunc_page(ap->a_offlo),
488 				     round_page(ap->a_offhi),
489 				     PGO_CLEANIT | (wait ? PGO_SYNCIO : 0));
490 		if (error == EAGAIN) {
491 			mutex_enter(&lfs_lock);
492 			mtsleep(&fs->lfs_availsleep, PCATCH | PUSER,
493 				"lfs_fsync", hz / 100 + 1, &lfs_lock);
494 			mutex_exit(&lfs_lock);
495 		}
496 	} while (error == EAGAIN);
497 	if (error)
498 		goto out;
499 
500 	if ((ap->a_flags & FSYNC_DATAONLY) == 0)
501 		error = lfs_update(vp, NULL, NULL, wait ? UPDATE_WAIT : 0);
502 
503 	if (error == 0 && ap->a_flags & FSYNC_CACHE) {
504 		int l = 0;
505 		error = VOP_IOCTL(ip->i_devvp, DIOCCACHESYNC, &l, FWRITE,
506 				  curlwp->l_cred);
507 	}
508 	if (wait && !VPISEMPTY(vp))
509 		LFS_SET_UINO(ip, IN_MODIFIED);
510 
511 out:
512 	KASSERT(VOP_ISLOCKED(vp) == LK_EXCLUSIVE);
513 	return error;
514 }
515 
516 /*
517  * Take IN_ADIROP off, then call ulfs_inactive.
518  */
519 int
520 lfs_inactive(void *v)
521 {
522 	struct vop_inactive_v2_args /* {
523 		struct vnode *a_vp;
524 		bool *a_recycle;
525 	} */ *ap = v;
526 
527 	KASSERT(VOP_ISLOCKED(ap->a_vp) == LK_EXCLUSIVE);
528 
529 	UNMARK_VNODE(ap->a_vp);
530 
531 	/*
532 	 * The Ifile is only ever inactivated on unmount.
533 	 * Streamline this process by not giving it more dirty blocks.
534 	 */
535 	if (VTOI(ap->a_vp)->i_number == LFS_IFILE_INUM) {
536 		mutex_enter(&lfs_lock);
537 		LFS_CLR_UINO(VTOI(ap->a_vp), IN_ALLMOD);
538 		mutex_exit(&lfs_lock);
539 		return 0;
540 	}
541 
542 #ifdef DEBUG
543 	/*
544 	 * This might happen on unmount.
545 	 * XXX If it happens at any other time, it should be a panic.
546 	 */
547 	if (ap->a_vp->v_uflag & VU_DIROP) {
548 		struct inode *ip = VTOI(ap->a_vp);
549 		printf("lfs_inactive: inactivating VU_DIROP? ino = %llu\n",
550 		    (unsigned long long) ip->i_number);
551 	}
552 #endif /* DIAGNOSTIC */
553 
554 	return ulfs_inactive(v);
555 }
556 
557 int
558 lfs_set_dirop(struct vnode *dvp, struct vnode *vp)
559 {
560 	struct lfs *fs;
561 	int error;
562 
563 	KASSERT(VOP_ISLOCKED(dvp) == LK_EXCLUSIVE);
564 	KASSERT(vp == NULL || VOP_ISLOCKED(vp) == LK_EXCLUSIVE);
565 
566 	fs = VTOI(dvp)->i_lfs;
567 
568 	ASSERT_NO_SEGLOCK(fs);
569 	/*
570 	 * LFS_NRESERVE calculates direct and indirect blocks as well
571 	 * as an inode block; an overestimate in most cases.
572 	 */
573 	if ((error = lfs_reserve(fs, dvp, vp, LFS_NRESERVE(fs))) != 0)
574 		return (error);
575 
576     restart:
577 	mutex_enter(&lfs_lock);
578 	if (fs->lfs_dirops == 0) {
579 		mutex_exit(&lfs_lock);
580 		lfs_check(dvp, LFS_UNUSED_LBN, 0);
581 		mutex_enter(&lfs_lock);
582 	}
583 	while (fs->lfs_writer) {
584 		error = cv_wait_sig(&fs->lfs_diropscv, &lfs_lock);
585 		if (error == EINTR) {
586 			mutex_exit(&lfs_lock);
587 			goto unreserve;
588 		}
589 	}
590 	if (lfs_dirvcount > LFS_MAX_DIROP && fs->lfs_dirops == 0) {
591 		cv_broadcast(&lfs_writerd_cv);
592 		mutex_exit(&lfs_lock);
593 		preempt();
594 		goto restart;
595 	}
596 
597 	if (lfs_dirvcount > LFS_MAX_DIROP) {
598 		DLOG((DLOG_DIROP, "lfs_set_dirop: sleeping with dirops=%d, "
599 		      "dirvcount=%d\n", fs->lfs_dirops, lfs_dirvcount));
600 		if ((error = mtsleep(&lfs_dirvcount,
601 		    PCATCH | PUSER | PNORELOCK, "lfs_maxdirop", 0,
602 		    &lfs_lock)) != 0) {
603 			mutex_exit(&lfs_lock);
604 			goto unreserve;
605 		}
606 		mutex_exit(&lfs_lock);
607 		goto restart;
608 	}
609 
610 	++fs->lfs_dirops;
611 	/* fs->lfs_doifile = 1; */ /* XXX why? --ks */
612 	mutex_exit(&lfs_lock);
613 
614 	/* Hold a reference so SET_ENDOP will be happy */
615 	vref(dvp);
616 	if (vp) {
617 		vref(vp);
618 		MARK_VNODE(vp);
619 	}
620 
621 	MARK_VNODE(dvp);
622 	return 0;
623 
624   unreserve:
625 	lfs_reserve(fs, dvp, vp, -LFS_NRESERVE(fs));
626 	return error;
627 }
628 
629 /*
630  * Opposite of lfs_set_dirop... mostly. For now at least must call
631  * UNMARK_VNODE(dvp) explicitly first. (XXX: clean that up)
632  */
633 void
634 lfs_unset_dirop(struct lfs *fs, struct vnode *dvp, const char *str)
635 {
636 	mutex_enter(&lfs_lock);
637 	--fs->lfs_dirops;
638 	if (!fs->lfs_dirops) {
639 		if (fs->lfs_nadirop) {
640 			panic("lfs_unset_dirop: %s: no dirops but "
641 			      " nadirop=%d", str,
642 			      fs->lfs_nadirop);
643 		}
644 		wakeup(&fs->lfs_writer);
645 		mutex_exit(&lfs_lock);
646 		lfs_check(dvp, LFS_UNUSED_LBN, 0);
647 	} else {
648 		mutex_exit(&lfs_lock);
649 	}
650 	lfs_reserve(fs, dvp, NULL, -LFS_NRESERVE(fs));
651 }
652 
653 void
654 lfs_mark_vnode(struct vnode *vp)
655 {
656 	struct inode *ip = VTOI(vp);
657 	struct lfs *fs = ip->i_lfs;
658 
659 	mutex_enter(&lfs_lock);
660 	if (!(ip->i_state & IN_ADIROP)) {
661 		if (!(vp->v_uflag & VU_DIROP)) {
662 			mutex_exit(&lfs_lock);
663 			vref(vp);
664 			mutex_enter(&lfs_lock);
665 			++lfs_dirvcount;
666 			++fs->lfs_dirvcount;
667 			TAILQ_INSERT_TAIL(&fs->lfs_dchainhd, ip, i_lfs_dchain);
668 			vp->v_uflag |= VU_DIROP;
669 		}
670 		++fs->lfs_nadirop;
671 		ip->i_state &= ~IN_CDIROP;
672 		ip->i_state |= IN_ADIROP;
673 	} else
674 		KASSERT(vp->v_uflag & VU_DIROP);
675 	mutex_exit(&lfs_lock);
676 }
677 
678 void
679 lfs_unmark_vnode(struct vnode *vp)
680 {
681 	struct inode *ip = VTOI(vp);
682 
683 	mutex_enter(&lfs_lock);
684 	if (ip && (ip->i_state & IN_ADIROP)) {
685 		KASSERT(vp->v_uflag & VU_DIROP);
686 		--ip->i_lfs->lfs_nadirop;
687 		ip->i_state &= ~IN_ADIROP;
688 	}
689 	mutex_exit(&lfs_lock);
690 }
691 
692 int
693 lfs_symlink(void *v)
694 {
695 	struct vop_symlink_v3_args /* {
696 		struct vnode *a_dvp;
697 		struct vnode **a_vpp;
698 		struct componentname *a_cnp;
699 		struct vattr *a_vap;
700 		char *a_target;
701 	} */ *ap = v;
702 	struct lfs *fs;
703 	struct vnode *dvp, **vpp;
704 	struct inode *ip;
705 	struct ulfs_lookup_results *ulr;
706 	ssize_t len; /* XXX should be size_t */
707 	int error;
708 
709 	dvp = ap->a_dvp;
710 	vpp = ap->a_vpp;
711 
712 	KASSERT(VOP_ISLOCKED(dvp) == LK_EXCLUSIVE);
713 	KASSERT(vpp != NULL);
714 	KASSERT(*vpp == NULL);
715 	KASSERT(ap->a_vap->va_type == VLNK);
716 
717 	/* XXX should handle this material another way */
718 	ulr = &VTOI(ap->a_dvp)->i_crap;
719 	ULFS_CHECK_CRAPCOUNTER(VTOI(ap->a_dvp));
720 
721 	fs = VFSTOULFS(dvp->v_mount)->um_lfs;
722 	ASSERT_NO_SEGLOCK(fs);
723 	if (fs->lfs_ronly) {
724 		return EROFS;
725 	}
726 
727 	error = lfs_set_dirop(dvp, NULL);
728 	if (error)
729 		return error;
730 
731 	error = lfs_makeinode(ap->a_vap, dvp, ulr, vpp, ap->a_cnp);
732 	if (error) {
733 		goto out;
734 	}
735 	KASSERT(VOP_ISLOCKED(*vpp) == LK_EXCLUSIVE);
736 
737 	VN_KNOTE(ap->a_dvp, NOTE_WRITE);
738 	ip = VTOI(*vpp);
739 
740 	/*
741 	 * This test is off by one. um_maxsymlinklen contains the
742 	 * number of bytes available, and we aren't storing a \0, so
743 	 * the test should properly be <=. However, it cannot be
744 	 * changed as this would break compatibility with existing fs
745 	 * images -- see the way ulfs_readlink() works.
746 	 */
747 	len = strlen(ap->a_target);
748 	if (len < ip->i_lfs->um_maxsymlinklen) {
749 		memcpy((char *)SHORTLINK(ip), ap->a_target, len);
750 		ip->i_size = len;
751 		DIP_ASSIGN(ip, size, len);
752 		uvm_vnp_setsize(*vpp, ip->i_size);
753 		ip->i_state |= IN_CHANGE | IN_UPDATE;
754 		if ((*vpp)->v_mount->mnt_flag & MNT_RELATIME)
755 			ip->i_state |= IN_ACCESS;
756 	} else {
757 		error = ulfs_bufio(UIO_WRITE, *vpp, ap->a_target, len, (off_t)0,
758 		    IO_NODELOCKED | IO_JOURNALLOCKED, ap->a_cnp->cn_cred, NULL,
759 		    NULL);
760 	}
761 
762 	VOP_UNLOCK(*vpp);
763 	if (error)
764 		vrele(*vpp);
765 
766 out:
767 	UNMARK_VNODE(dvp);
768 	/* XXX: is it even possible for the symlink to get MARK'd? */
769 	UNMARK_VNODE(*vpp);
770 	if (error) {
771 		*vpp = NULL;
772 	}
773 	lfs_unset_dirop(fs, dvp, "symlink");
774 
775 	vrele(dvp);
776 	return (error);
777 }
778 
779 int
780 lfs_mknod(void *v)
781 {
782 	struct vop_mknod_v3_args	/* {
783 		struct vnode *a_dvp;
784 		struct vnode **a_vpp;
785 		struct componentname *a_cnp;
786 		struct vattr *a_vap;
787 	} */ *ap = v;
788 	struct lfs *fs;
789 	struct vnode *dvp, **vpp;
790 	struct vattr *vap;
791 	struct inode *ip;
792 	int error;
793 	ino_t		ino;
794 	struct ulfs_lookup_results *ulr;
795 
796 	dvp = ap->a_dvp;
797 	vpp = ap->a_vpp;
798 	vap = ap->a_vap;
799 
800 	KASSERT(VOP_ISLOCKED(dvp) == LK_EXCLUSIVE);
801 	KASSERT(vpp != NULL);
802 	KASSERT(*vpp == NULL);
803 
804 	/* XXX should handle this material another way */
805 	ulr = &VTOI(dvp)->i_crap;
806 	ULFS_CHECK_CRAPCOUNTER(VTOI(dvp));
807 
808 	fs = VFSTOULFS(dvp->v_mount)->um_lfs;
809 	ASSERT_NO_SEGLOCK(fs);
810 	if (fs->lfs_ronly) {
811 		return EROFS;
812 	}
813 
814 	error = lfs_set_dirop(dvp, NULL);
815 	if (error)
816 		return error;
817 
818 	error = lfs_makeinode(vap, dvp, ulr, vpp, ap->a_cnp);
819 
820 	/* Either way we're done with the dirop at this point */
821 	UNMARK_VNODE(dvp);
822 	UNMARK_VNODE(*vpp);
823 	lfs_unset_dirop(fs, dvp, "mknod");
824 
825 	if (error) {
826 		vrele(dvp);
827 		*vpp = NULL;
828 		return (error);
829 	}
830 	KASSERT(VOP_ISLOCKED(*vpp) == LK_EXCLUSIVE);
831 
832 	VN_KNOTE(dvp, NOTE_WRITE);
833 	ip = VTOI(*vpp);
834 	ino = ip->i_number;
835 	ip->i_state |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
836 
837 	/*
838 	 * Call fsync to write the vnode so that we don't have to deal with
839 	 * flushing it when it's marked VU_DIROP or reclaiming.
840 	 *
841 	 * XXX KS - If we can't flush we also can't call vgone(), so must
842 	 * return.  But, that leaves this vnode in limbo, also not good.
843 	 * Can this ever happen (barring hardware failure)?
844 	 */
845 	if ((error = VOP_FSYNC(*vpp, NOCRED, FSYNC_WAIT, 0, 0)) != 0) {
846 		panic("lfs_mknod: couldn't fsync (ino %llu)",
847 		    (unsigned long long) ino);
848 		/* return (error); */
849 	}
850 
851 	vrele(dvp);
852 	KASSERT(error == 0);
853 	VOP_UNLOCK(*vpp);
854 	return (0);
855 }
856 
857 /*
858  * Create a regular file
859  */
860 int
861 lfs_create(void *v)
862 {
863 	struct vop_create_v3_args	/* {
864 		struct vnode *a_dvp;
865 		struct vnode **a_vpp;
866 		struct componentname *a_cnp;
867 		struct vattr *a_vap;
868 	} */ *ap = v;
869 	struct lfs *fs;
870 	struct vnode *dvp, **vpp;
871 	struct vattr *vap;
872 	struct ulfs_lookup_results *ulr;
873 	int error;
874 
875 	dvp = ap->a_dvp;
876 	vpp = ap->a_vpp;
877 	vap = ap->a_vap;
878 
879 	KASSERT(VOP_ISLOCKED(dvp) == LK_EXCLUSIVE);
880 	KASSERT(vpp != NULL);
881 	KASSERT(*vpp == NULL);
882 
883 	/* XXX should handle this material another way */
884 	ulr = &VTOI(dvp)->i_crap;
885 	ULFS_CHECK_CRAPCOUNTER(VTOI(dvp));
886 
887 	fs = VFSTOULFS(dvp->v_mount)->um_lfs;
888 	ASSERT_NO_SEGLOCK(fs);
889 	if (fs->lfs_ronly) {
890 		return EROFS;
891 	}
892 
893 	error = lfs_set_dirop(dvp, NULL);
894 	if (error)
895 		return error;
896 
897 	error = lfs_makeinode(vap, dvp, ulr, vpp, ap->a_cnp);
898 	if (error) {
899 		goto out;
900 	}
901 	KASSERT(VOP_ISLOCKED(*vpp) == LK_EXCLUSIVE);
902 	VN_KNOTE(dvp, NOTE_WRITE);
903 	VOP_UNLOCK(*vpp);
904 
905 out:
906 
907 	UNMARK_VNODE(dvp);
908 	UNMARK_VNODE(*vpp);
909 	if (error) {
910 		*vpp = NULL;
911 	}
912 	lfs_unset_dirop(fs, dvp, "create");
913 
914 	vrele(dvp);
915 	return (error);
916 }
917 
918 int
919 lfs_mkdir(void *v)
920 {
921 	struct vop_mkdir_v3_args	/* {
922 		struct vnode *a_dvp;
923 		struct vnode **a_vpp;
924 		struct componentname *a_cnp;
925 		struct vattr *a_vap;
926 	} */ *ap = v;
927 	struct lfs *fs;
928 	struct vnode *dvp, *tvp, **vpp;
929 	struct inode *dp, *ip;
930 	struct componentname *cnp;
931 	struct vattr *vap;
932 	struct ulfs_lookup_results *ulr;
933 	struct buf *bp;
934 	LFS_DIRHEADER *dirp;
935 	int dirblksiz;
936 	int error;
937 
938 	dvp = ap->a_dvp;
939 	tvp = NULL;
940 	vpp = ap->a_vpp;
941 	cnp = ap->a_cnp;
942 	vap = ap->a_vap;
943 
944 	KASSERT(VOP_ISLOCKED(dvp) == LK_EXCLUSIVE);
945 
946 	dp = VTOI(dvp);
947 	ip = NULL;
948 
949 	KASSERT(vap->va_type == VDIR);
950 	KASSERT(vpp != NULL);
951 	KASSERT(*vpp == NULL);
952 
953 	/* XXX should handle this material another way */
954 	ulr = &dp->i_crap;
955 	ULFS_CHECK_CRAPCOUNTER(dp);
956 
957 	fs = VFSTOULFS(dvp->v_mount)->um_lfs;
958 	ASSERT_NO_SEGLOCK(fs);
959 	if (fs->lfs_ronly) {
960 		return EROFS;
961 	}
962 	dirblksiz = fs->um_dirblksiz;
963 	/* XXX dholland 20150911 I believe this to be true, but... */
964 	//KASSERT(dirblksiz == LFS_DIRBLKSIZ);
965 
966 	error = lfs_set_dirop(dvp, NULL);
967 	if (error)
968 		return error;
969 
970 	if ((nlink_t)dp->i_nlink >= LINK_MAX) {
971 		error = EMLINK;
972 		goto out;
973 	}
974 
975 	/*
976 	 * Must simulate part of lfs_makeinode here to acquire the inode,
977 	 * but not have it entered in the parent directory. The entry is
978 	 * made later after writing "." and ".." entries.
979 	 */
980 	error = vcache_new(dvp->v_mount, dvp, vap, cnp->cn_cred, ap->a_vpp);
981 	if (error)
982 		goto out;
983 
984 	error = vn_lock(*ap->a_vpp, LK_EXCLUSIVE);
985 	if (error) {
986 		vrele(*ap->a_vpp);
987 		*ap->a_vpp = NULL;
988 		goto out;
989 	}
990 
991 	tvp = *ap->a_vpp;
992 	MARK_VNODE(tvp);
993 	ip = VTOI(tvp);
994 	ip->i_state |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
995 	ip->i_nlink = 2;
996 	DIP_ASSIGN(ip, nlink, 2);
997 	if (cnp->cn_flags & ISWHITEOUT) {
998 		ip->i_flags |= UF_OPAQUE;
999 		DIP_ASSIGN(ip, flags, ip->i_flags);
1000 	}
1001 
1002 	/*
1003 	 * Bump link count in parent directory to reflect work done below.
1004 	 */
1005 	dp->i_nlink++;
1006 	DIP_ASSIGN(dp, nlink, dp->i_nlink);
1007 	dp->i_state |= IN_CHANGE;
1008 	if ((error = lfs_update(dvp, NULL, NULL, UPDATE_DIROP)) != 0)
1009 		goto bad;
1010 
1011 	/*
1012 	 * Initialize directory with "." and "..". This used to use a
1013 	 * static template but that adds moving parts for very little
1014 	 * benefit.
1015 	 */
1016 	if ((error = lfs_balloc(tvp, (off_t)0, dirblksiz, cnp->cn_cred,
1017 	    B_CLRBUF, &bp)) != 0)
1018 		goto bad;
1019 	ip->i_size = dirblksiz;
1020 	DIP_ASSIGN(ip, size, dirblksiz);
1021 	ip->i_state |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
1022 	uvm_vnp_setsize(tvp, ip->i_size);
1023 	dirp = bp->b_data;
1024 
1025 	/* . */
1026 	lfs_dir_setino(fs, dirp, ip->i_number);
1027 	lfs_dir_setreclen(fs, dirp, LFS_DIRECTSIZ(fs, 1));
1028 	lfs_dir_settype(fs, dirp, LFS_DT_DIR);
1029 	lfs_dir_setnamlen(fs, dirp, 1);
1030 	lfs_copydirname(fs, lfs_dir_nameptr(fs, dirp), ".", 1,
1031 			LFS_DIRECTSIZ(fs, 1));
1032 	dirp = LFS_NEXTDIR(fs, dirp);
1033 	/* .. */
1034 	lfs_dir_setino(fs, dirp, dp->i_number);
1035 	lfs_dir_setreclen(fs, dirp, dirblksiz - LFS_DIRECTSIZ(fs, 1));
1036 	lfs_dir_settype(fs, dirp, LFS_DT_DIR);
1037 	lfs_dir_setnamlen(fs, dirp, 2);
1038 	lfs_copydirname(fs, lfs_dir_nameptr(fs, dirp), "..", 2,
1039 			dirblksiz - LFS_DIRECTSIZ(fs, 1));
1040 
1041 	/*
1042 	 * Directory set up; now install its entry in the parent directory.
1043 	 */
1044 	if ((error = VOP_BWRITE(bp->b_vp, bp)) != 0)
1045 		goto bad;
1046 	if ((error = lfs_update(tvp, NULL, NULL, UPDATE_DIROP)) != 0) {
1047 		goto bad;
1048 	}
1049 	error = ulfs_direnter(dvp, ulr, tvp,
1050 			      cnp, ip->i_number, LFS_IFTODT(ip->i_mode), bp);
1051  bad:
1052 	if (error == 0) {
1053 		VN_KNOTE(dvp, NOTE_WRITE | NOTE_LINK);
1054 		VOP_UNLOCK(tvp);
1055 	} else {
1056 		dp->i_nlink--;
1057 		DIP_ASSIGN(dp, nlink, dp->i_nlink);
1058 		dp->i_state |= IN_CHANGE;
1059 		/*
1060 		 * No need to do an explicit lfs_truncate here, vrele will
1061 		 * do this for us because we set the link count to 0.
1062 		 */
1063 		ip->i_nlink = 0;
1064 		DIP_ASSIGN(ip, nlink, 0);
1065 		ip->i_state |= IN_CHANGE;
1066 		/* If IN_ADIROP, account for it */
1067 		UNMARK_VNODE(tvp);
1068 		vput(tvp);
1069 	}
1070 
1071 out:
1072 	UNMARK_VNODE(dvp);
1073 	UNMARK_VNODE(*vpp);
1074 	if (error) {
1075 		*vpp = NULL;
1076 	}
1077 	lfs_unset_dirop(fs, dvp, "mkdir");
1078 
1079 	vrele(dvp);
1080 	return (error);
1081 }
1082 
1083 int
1084 lfs_remove(void *v)
1085 {
1086 	struct vop_remove_v2_args /* {
1087 		struct vnode *a_dvp;
1088 		struct vnode *a_vp;
1089 		struct componentname *a_cnp;
1090 	} */ *ap = v;
1091 	struct vnode *dvp, *vp;
1092 	struct inode *ip;
1093 	int error;
1094 
1095 	dvp = ap->a_dvp;
1096 	vp = ap->a_vp;
1097 
1098 	KASSERT(VOP_ISLOCKED(dvp) == LK_EXCLUSIVE);
1099 	KASSERT(VOP_ISLOCKED(vp) == LK_EXCLUSIVE);
1100 
1101 	ip = VTOI(vp);
1102 	if ((error = lfs_set_dirop(dvp, vp)) != 0) {
1103 		if (dvp == vp)
1104 			vrele(vp);
1105 		else
1106 			vput(vp);
1107 		return error;
1108 	}
1109 	error = ulfs_remove(ap);
1110 	if (ip->i_nlink == 0)
1111 		lfs_orphan(ip->i_lfs, ip->i_number);
1112 
1113 	UNMARK_VNODE(dvp);
1114 	if (ap->a_vp) {
1115 		UNMARK_VNODE(ap->a_vp);
1116 	}
1117 	lfs_unset_dirop(ip->i_lfs, dvp, "remove");
1118 	vrele(dvp);
1119 	if (ap->a_vp) {
1120 		vrele(ap->a_vp);
1121 	}
1122 
1123 	return (error);
1124 }
1125 
1126 int
1127 lfs_rmdir(void *v)
1128 {
1129 	struct vop_rmdir_v2_args /* {
1130 		struct vnodeop_desc *a_desc;
1131 		struct vnode *a_dvp;
1132 		struct vnode *a_vp;
1133 		struct componentname *a_cnp;
1134 	} */ *ap = v;
1135 	struct vnode *vp;
1136 	struct inode *ip;
1137 	int error;
1138 
1139 	vp = ap->a_vp;
1140 
1141 	KASSERT(VOP_ISLOCKED(ap->a_dvp) == LK_EXCLUSIVE);
1142 	KASSERT(VOP_ISLOCKED(vp) == LK_EXCLUSIVE);
1143 
1144 	ip = VTOI(vp);
1145 	if ((error = lfs_set_dirop(ap->a_dvp, ap->a_vp)) != 0) {
1146 		if (ap->a_dvp == vp)
1147 			vrele(vp);
1148 		else
1149 			vput(vp);
1150 		return error;
1151 	}
1152 	error = ulfs_rmdir(ap);
1153 	if (ip->i_nlink == 0)
1154 		lfs_orphan(ip->i_lfs, ip->i_number);
1155 
1156 	UNMARK_VNODE(ap->a_dvp);
1157 	if (ap->a_vp) {
1158 		UNMARK_VNODE(ap->a_vp);
1159 	}
1160 	lfs_unset_dirop(ip->i_lfs, ap->a_dvp, "rmdir");
1161 	vrele(ap->a_dvp);
1162 	if (ap->a_vp) {
1163 		vrele(ap->a_vp);
1164 	}
1165 
1166 	return (error);
1167 }
1168 
1169 int
1170 lfs_link(void *v)
1171 {
1172 	struct vop_link_v2_args	/* {
1173 		struct vnode *a_dvp;
1174 		struct vnode *a_vp;
1175 		struct componentname *a_cnp;
1176 	} */ *ap = v;
1177 	struct lfs *fs;
1178 	struct vnode *dvp;
1179 	int error;
1180 
1181 	dvp = ap->a_dvp;
1182 
1183 	KASSERT(VOP_ISLOCKED(dvp) == LK_EXCLUSIVE);
1184 
1185 	fs = VFSTOULFS(dvp->v_mount)->um_lfs;
1186 	ASSERT_NO_SEGLOCK(fs);
1187 	if (fs->lfs_ronly) {
1188 		return EROFS;
1189 	}
1190 
1191 	error = lfs_set_dirop(dvp, NULL);
1192 	if (error) {
1193 		return error;
1194 	}
1195 
1196 	error = ulfs_link(ap);
1197 
1198 	UNMARK_VNODE(dvp);
1199 	lfs_unset_dirop(fs, dvp, "link");
1200 	vrele(dvp);
1201 
1202 	return (error);
1203 }
1204 
1205 /* XXX hack to avoid calling ITIMES in getattr */
1206 int
1207 lfs_getattr(void *v)
1208 {
1209 	struct vop_getattr_args /* {
1210 		struct vnode *a_vp;
1211 		struct vattr *a_vap;
1212 		kauth_cred_t a_cred;
1213 	} */ *ap = v;
1214 	struct vnode *vp = ap->a_vp;
1215 	struct inode *ip;
1216 	struct vattr *vap = ap->a_vap;
1217 	struct lfs *fs;
1218 
1219 	KASSERT(VOP_ISLOCKED(vp));
1220 
1221 	ip = VTOI(vp);
1222 	fs = ip->i_lfs;
1223 
1224 	/*
1225 	 * Copy from inode table
1226 	 */
1227 	vap->va_fsid = ip->i_dev;
1228 	vap->va_fileid = ip->i_number;
1229 	vap->va_mode = ip->i_mode & ~LFS_IFMT;
1230 	vap->va_nlink = ip->i_nlink;
1231 	vap->va_uid = ip->i_uid;
1232 	vap->va_gid = ip->i_gid;
1233 	switch (vp->v_type) {
1234 	    case VBLK:
1235 	    case VCHR:
1236 		vap->va_rdev = (dev_t)lfs_dino_getrdev(fs, ip->i_din);
1237 		break;
1238 	    default:
1239 		vap->va_rdev = NODEV;
1240 		break;
1241 	}
1242 	vap->va_size = vp->v_size;
1243 	vap->va_atime.tv_sec = lfs_dino_getatime(fs, ip->i_din);
1244 	vap->va_atime.tv_nsec = lfs_dino_getatimensec(fs, ip->i_din);
1245 	vap->va_mtime.tv_sec = lfs_dino_getmtime(fs, ip->i_din);
1246 	vap->va_mtime.tv_nsec = lfs_dino_getmtimensec(fs, ip->i_din);
1247 	vap->va_ctime.tv_sec = lfs_dino_getctime(fs, ip->i_din);
1248 	vap->va_ctime.tv_nsec = lfs_dino_getctimensec(fs, ip->i_din);
1249 	vap->va_flags = ip->i_flags;
1250 	vap->va_gen = ip->i_gen;
1251 	/* this doesn't belong here */
1252 	if (vp->v_type == VBLK)
1253 		vap->va_blocksize = BLKDEV_IOSIZE;
1254 	else if (vp->v_type == VCHR)
1255 		vap->va_blocksize = MAXBSIZE;
1256 	else
1257 		vap->va_blocksize = vp->v_mount->mnt_stat.f_iosize;
1258 	vap->va_bytes = lfs_fsbtob(fs, ip->i_lfs_effnblks);
1259 	vap->va_type = vp->v_type;
1260 	vap->va_filerev = ip->i_modrev;
1261 	return (0);
1262 }
1263 
1264 /*
1265  * Check to make sure the inode blocks won't choke the buffer
1266  * cache, then call ulfs_setattr as usual.
1267  */
1268 int
1269 lfs_setattr(void *v)
1270 {
1271 	struct vop_setattr_args /* {
1272 		struct vnode *a_vp;
1273 		struct vattr *a_vap;
1274 		kauth_cred_t a_cred;
1275 	} */ *ap = v;
1276 	struct vnode *vp = ap->a_vp;
1277 
1278 	KASSERT(VOP_ISLOCKED(vp) == LK_EXCLUSIVE);
1279 	lfs_check(vp, LFS_UNUSED_LBN, 0);
1280 	return ulfs_setattr(v);
1281 }
1282 
1283 /*
1284  * Release the block we hold on lfs_newseg wrapping.  Called on file close,
1285  * or explicitly from LFCNWRAPGO.  Called with the interlock held.
1286  */
1287 static int
1288 lfs_wrapgo(struct lfs *fs, struct inode *ip, int waitfor)
1289 {
1290 	if (fs->lfs_stoplwp != curlwp)
1291 		return EBUSY;
1292 
1293 	fs->lfs_stoplwp = NULL;
1294 	cv_signal(&fs->lfs_stopcv);
1295 
1296 	KASSERT(fs->lfs_nowrap > 0);
1297 	if (fs->lfs_nowrap <= 0) {
1298 		return 0;
1299 	}
1300 
1301 	if (--fs->lfs_nowrap == 0) {
1302 		log(LOG_NOTICE, "%s: re-enabled log wrap\n",
1303 		    lfs_sb_getfsmnt(fs));
1304 		wakeup(&fs->lfs_wrappass);
1305 		lfs_wakeup_cleaner(fs);
1306 	}
1307 	if (waitfor) {
1308 		cv_wait_sig(&fs->lfs_nextsegsleep, &lfs_lock);
1309 	}
1310 
1311 	return 0;
1312 }
1313 
1314 /*
1315  * Close called.
1316  *
1317  * Update the times on the inode.
1318  */
1319 /* ARGSUSED */
1320 int
1321 lfs_close(void *v)
1322 {
1323 	struct vop_close_args /* {
1324 		struct vnode *a_vp;
1325 		int  a_fflag;
1326 		kauth_cred_t a_cred;
1327 	} */ *ap = v;
1328 	struct vnode *vp = ap->a_vp;
1329 	struct inode *ip;
1330 	struct lfs *fs;
1331 
1332 	KASSERT(VOP_ISLOCKED(vp) == LK_EXCLUSIVE);
1333 
1334 	ip = VTOI(vp);
1335 	fs = ip->i_lfs;
1336 
1337 	if ((ip->i_number == ULFS_ROOTINO || ip->i_number == LFS_IFILE_INUM) &&
1338 	    fs->lfs_stoplwp == curlwp) {
1339 		mutex_enter(&lfs_lock);
1340 		log(LOG_NOTICE, "lfs_close: releasing log wrap control\n");
1341 		lfs_wrapgo(fs, ip, 0);
1342 		mutex_exit(&lfs_lock);
1343 	}
1344 
1345 	if (vp == ip->i_lfs->lfs_ivnode &&
1346 	    vp->v_mount->mnt_iflag & IMNT_UNMOUNT)
1347 		return 0;
1348 
1349 	if (vp->v_usecount > 1 && vp != ip->i_lfs->lfs_ivnode) {
1350 		LFS_ITIMES(ip, NULL, NULL, NULL);
1351 	}
1352 	return (0);
1353 }
1354 
1355 /*
1356  * Close wrapper for special devices.
1357  *
1358  * Update the times on the inode then do device close.
1359  */
1360 int
1361 lfsspec_close(void *v)
1362 {
1363 	struct vop_close_args /* {
1364 		struct vnode	*a_vp;
1365 		int		a_fflag;
1366 		kauth_cred_t	a_cred;
1367 	} */ *ap = v;
1368 	struct vnode	*vp;
1369 	struct inode	*ip;
1370 
1371 	vp = ap->a_vp;
1372 
1373 	KASSERT(VOP_ISLOCKED(vp) == LK_EXCLUSIVE);
1374 
1375 	ip = VTOI(vp);
1376 	if (vp->v_usecount > 1) {
1377 		LFS_ITIMES(ip, NULL, NULL, NULL);
1378 	}
1379 	return (VOCALL (spec_vnodeop_p, VOFFSET(vop_close), ap));
1380 }
1381 
1382 /*
1383  * Close wrapper for fifo's.
1384  *
1385  * Update the times on the inode then do device close.
1386  */
1387 int
1388 lfsfifo_close(void *v)
1389 {
1390 	struct vop_close_args /* {
1391 		struct vnode	*a_vp;
1392 		int		a_fflag;
1393 		kauth_cred_	a_cred;
1394 	} */ *ap = v;
1395 	struct vnode	*vp;
1396 	struct inode	*ip;
1397 
1398 	vp = ap->a_vp;
1399 
1400 	KASSERT(VOP_ISLOCKED(vp) == LK_EXCLUSIVE);
1401 
1402 	ip = VTOI(vp);
1403 	if (ap->a_vp->v_usecount > 1) {
1404 		LFS_ITIMES(ip, NULL, NULL, NULL);
1405 	}
1406 	return (VOCALL (fifo_vnodeop_p, VOFFSET(vop_close), ap));
1407 }
1408 
1409 /*
1410  * Reclaim an inode so that it can be used for other purposes.
1411  */
1412 
1413 int
1414 lfs_reclaim(void *v)
1415 {
1416 	struct vop_reclaim_v2_args /* {
1417 		struct vnode *a_vp;
1418 	} */ *ap = v;
1419 	struct vnode *vp = ap->a_vp;
1420 	struct inode *ip;
1421 	struct lfs *fs;
1422 	int error;
1423 
1424 	VOP_UNLOCK(vp);
1425 
1426 	ip = VTOI(vp);
1427 	fs = ip->i_lfs;
1428 
1429 	/*
1430 	 * The inode must be freed and updated before being removed
1431 	 * from its hash chain.  Other threads trying to gain a hold
1432 	 * or lock on the inode will be stalled.
1433 	 */
1434 	if (ip->i_nlink <= 0 && (vp->v_mount->mnt_flag & MNT_RDONLY) == 0)
1435 		lfs_vfree(vp, ip->i_number, ip->i_omode);
1436 
1437 	mutex_enter(&lfs_lock);
1438 	LFS_CLR_UINO(ip, IN_ALLMOD);
1439 	mutex_exit(&lfs_lock);
1440 	if ((error = ulfs_reclaim(vp)))
1441 		return (error);
1442 
1443 	/*
1444 	 * Take us off the paging and/or dirop queues if we were on them.
1445 	 * We shouldn't be on them.
1446 	 */
1447 	mutex_enter(&lfs_lock);
1448 	if (ip->i_state & IN_PAGING) {
1449 		log(LOG_WARNING, "%s: reclaimed vnode is IN_PAGING\n",
1450 		    lfs_sb_getfsmnt(fs));
1451 		ip->i_state &= ~IN_PAGING;
1452 		TAILQ_REMOVE(&fs->lfs_pchainhd, ip, i_lfs_pchain);
1453 	}
1454 	if (vp->v_uflag & VU_DIROP)
1455 		panic("reclaimed vnode is VU_DIROP");
1456 	mutex_exit(&lfs_lock);
1457 
1458 	pool_put(&lfs_dinode_pool, ip->i_din);
1459 	lfs_deregister_all(vp);
1460 	pool_put(&lfs_inoext_pool, ip->inode_ext.lfs);
1461 	ip->inode_ext.lfs = NULL;
1462 	genfs_node_destroy(vp);
1463 	pool_put(&lfs_inode_pool, vp->v_data);
1464 	vp->v_data = NULL;
1465 	return (0);
1466 }
1467 
1468 /*
1469  * Read a block from a storage device.
1470  *
1471  * Calculate the logical to physical mapping if not done already,
1472  * then call the device strategy routine.
1473  *
1474  * In order to avoid reading blocks that are in the process of being
1475  * written by the cleaner---and hence are not mutexed by the normal
1476  * buffer cache / page cache mechanisms---check for collisions before
1477  * reading.
1478  *
1479  * We inline ulfs_strategy to make sure that the VOP_BMAP occurs *before*
1480  * the active cleaner test.
1481  *
1482  * XXX This code assumes that lfs_markv makes synchronous checkpoints.
1483  */
1484 int
1485 lfs_strategy(void *v)
1486 {
1487 	struct vop_strategy_args /* {
1488 		struct vnode *a_vp;
1489 		struct buf *a_bp;
1490 	} */ *ap = v;
1491 	struct buf	*bp;
1492 	struct lfs	*fs;
1493 	struct vnode	*vp;
1494 	struct inode	*ip;
1495 	daddr_t		tbn;
1496 #define MAXLOOP 25
1497 	int		i, sn, error, slept, loopcount;
1498 
1499 	bp = ap->a_bp;
1500 	vp = ap->a_vp;
1501 	ip = VTOI(vp);
1502 	fs = ip->i_lfs;
1503 
1504 	/* lfs uses its strategy routine only for read */
1505 	KASSERT(bp->b_flags & B_READ);
1506 
1507 	if (vp->v_type == VBLK || vp->v_type == VCHR)
1508 		panic("lfs_strategy: spec");
1509 	KASSERT(bp->b_bcount != 0);
1510 	if (bp->b_blkno == bp->b_lblkno) {
1511 		error = VOP_BMAP(vp, bp->b_lblkno, NULL, &bp->b_blkno,
1512 				 NULL);
1513 		if (error) {
1514 			bp->b_error = error;
1515 			bp->b_resid = bp->b_bcount;
1516 			biodone(bp);
1517 			return (error);
1518 		}
1519 		if ((long)bp->b_blkno == -1) /* no valid data */
1520 			clrbuf(bp);
1521 	}
1522 	if ((long)bp->b_blkno < 0) { /* block is not on disk */
1523 		bp->b_resid = bp->b_bcount;
1524 		biodone(bp);
1525 		return (0);
1526 	}
1527 
1528 	slept = 1;
1529 	loopcount = 0;
1530 	mutex_enter(&lfs_lock);
1531 	while (slept && fs->lfs_seglock) {
1532 		mutex_exit(&lfs_lock);
1533 		/*
1534 		 * Look through list of intervals.
1535 		 * There will only be intervals to look through
1536 		 * if the cleaner holds the seglock.
1537 		 * Since the cleaner is synchronous, we can trust
1538 		 * the list of intervals to be current.
1539 		 */
1540 		tbn = LFS_DBTOFSB(fs, bp->b_blkno);
1541 		sn = lfs_dtosn(fs, tbn);
1542 		slept = 0;
1543 		for (i = 0; i < fs->lfs_cleanind; i++) {
1544 			if (sn == lfs_dtosn(fs, fs->lfs_cleanint[i]) &&
1545 			    tbn >= fs->lfs_cleanint[i]) {
1546 				DLOG((DLOG_CLEAN,
1547 				      "lfs_strategy: ino %llu lbn %" PRId64
1548 				      " ind %d sn %d fsb %" PRIx64
1549 				      " given sn %d fsb %" PRIx64 "\n",
1550 				      (unsigned long long) ip->i_number,
1551 				      bp->b_lblkno, i,
1552 				      lfs_dtosn(fs, fs->lfs_cleanint[i]),
1553 				      fs->lfs_cleanint[i], sn, tbn));
1554 				DLOG((DLOG_CLEAN,
1555 				      "lfs_strategy: sleeping on ino %llu lbn %"
1556 				      PRId64 "\n",
1557 				      (unsigned long long) ip->i_number,
1558 				      bp->b_lblkno));
1559 				mutex_enter(&lfs_lock);
1560 				if (LFS_SEGLOCK_HELD(fs) && fs->lfs_iocount) {
1561 					/*
1562 					 * Cleaner can't wait for itself.
1563 					 * Instead, wait for the blocks
1564 					 * to be written to disk.
1565 					 * XXX we need pribio in the test
1566 					 * XXX here.
1567 					 */
1568  					mtsleep(&fs->lfs_iocount,
1569  						(PRIBIO + 1) | PNORELOCK,
1570 						"clean2", hz/10 + 1,
1571  						&lfs_lock);
1572 					slept = 1;
1573 					++loopcount;
1574 					break;
1575 				} else if (fs->lfs_seglock) {
1576 					mtsleep(&fs->lfs_seglock,
1577 						(PRIBIO + 1) | PNORELOCK,
1578 						"clean1", 0,
1579 						&lfs_lock);
1580 					slept = 1;
1581 					break;
1582 				}
1583 				mutex_exit(&lfs_lock);
1584 			}
1585 		}
1586 		mutex_enter(&lfs_lock);
1587 		if (loopcount > MAXLOOP) {
1588 			printf("lfs_strategy: breaking out of clean2 loop\n");
1589 			break;
1590 		}
1591 	}
1592 	mutex_exit(&lfs_lock);
1593 
1594 	vp = ip->i_devvp;
1595 	return VOP_STRATEGY(vp, bp);
1596 }
1597 
1598 /*
1599  * Inline lfs_segwrite/lfs_writevnodes, but just for dirops.
1600  * Technically this is a checkpoint (the on-disk state is valid)
1601  * even though we are leaving out all the file data.
1602  */
1603 int
1604 lfs_flush_dirops(struct lfs *fs)
1605 {
1606 	struct inode *ip, *nip;
1607 	struct vnode *vp;
1608 	extern int lfs_dostats; /* XXX this does not belong here */
1609 	struct segment *sp;
1610 	SEGSUM *ssp;
1611 	int flags = 0;
1612 	int error = 0;
1613 
1614 	ASSERT_MAYBE_SEGLOCK(fs);
1615 	KASSERT(fs->lfs_nadirop == 0);
1616 
1617 	if (fs->lfs_ronly)
1618 		return EROFS;
1619 
1620 	mutex_enter(&lfs_lock);
1621 	if (TAILQ_FIRST(&fs->lfs_dchainhd) == NULL) {
1622 		mutex_exit(&lfs_lock);
1623 		return 0;
1624 	} else
1625 		mutex_exit(&lfs_lock);
1626 
1627 	if (lfs_dostats)
1628 		++lfs_stats.flush_invoked;
1629 
1630 	lfs_imtime(fs);
1631 	lfs_seglock(fs, flags);
1632 	sp = fs->lfs_sp;
1633 
1634 	/*
1635 	 * lfs_writevnodes, optimized to get dirops out of the way.
1636 	 * Only write dirops, and don't flush files' pages, only
1637 	 * blocks from the directories.
1638 	 *
1639 	 * We don't need to vref these files because they are
1640 	 * dirops and so hold an extra reference until the
1641 	 * segunlock clears them of that status.
1642 	 *
1643 	 * We don't need to check for IN_ADIROP because we know that
1644 	 * no dirops are active.
1645 	 *
1646 	 */
1647 	mutex_enter(&lfs_lock);
1648 	for (ip = TAILQ_FIRST(&fs->lfs_dchainhd); ip != NULL; ip = nip) {
1649 		nip = TAILQ_NEXT(ip, i_lfs_dchain);
1650 		mutex_exit(&lfs_lock);
1651 		vp = ITOV(ip);
1652 		mutex_enter(vp->v_interlock);
1653 
1654 		KASSERT((ip->i_state & IN_ADIROP) == 0);
1655 		KASSERT(vp->v_uflag & VU_DIROP);
1656 		KASSERT(vdead_check(vp, VDEAD_NOWAIT) == 0);
1657 
1658 		/*
1659 		 * All writes to directories come from dirops; all
1660 		 * writes to files' direct blocks go through the page
1661 		 * cache, which we're not touching.  Reads to files
1662 		 * and/or directories will not be affected by writing
1663 		 * directory blocks inodes and file inodes.  So we don't
1664 		 * really need to lock.
1665 		 */
1666 		if (vdead_check(vp, VDEAD_NOWAIT) != 0) {
1667 			mutex_exit(vp->v_interlock);
1668 			mutex_enter(&lfs_lock);
1669 			continue;
1670 		}
1671 		mutex_exit(vp->v_interlock);
1672 		/* XXX see below
1673 		 * waslocked = VOP_ISLOCKED(vp);
1674 		 */
1675 		if (vp->v_type != VREG &&
1676 		    ((ip->i_state & IN_ALLMOD) || !VPISEMPTY(vp))) {
1677 			error = lfs_writefile(fs, sp, vp);
1678 			if (!VPISEMPTY(vp) && !WRITEINPROG(vp) &&
1679 			    !(ip->i_state & IN_ALLMOD)) {
1680 			    	mutex_enter(&lfs_lock);
1681 				LFS_SET_UINO(ip, IN_MODIFIED);
1682 			    	mutex_exit(&lfs_lock);
1683 			}
1684 			if (error && (sp->seg_flags & SEGM_SINGLE)) {
1685 				mutex_enter(&lfs_lock);
1686 				error = EAGAIN;
1687 				break;
1688 			}
1689 		}
1690 		KDASSERT(ip->i_number != LFS_IFILE_INUM);
1691 		error = lfs_writeinode(fs, sp, ip);
1692 		mutex_enter(&lfs_lock);
1693 		if (error && (sp->seg_flags & SEGM_SINGLE)) {
1694 			error = EAGAIN;
1695 			break;
1696 		}
1697 
1698 		/*
1699 		 * We might need to update these inodes again,
1700 		 * for example, if they have data blocks to write.
1701 		 * Make sure that after this flush, they are still
1702 		 * marked IN_MODIFIED so that we don't forget to
1703 		 * write them.
1704 		 */
1705 		/* XXX only for non-directories? --KS */
1706 		LFS_SET_UINO(ip, IN_MODIFIED);
1707 	}
1708 	mutex_exit(&lfs_lock);
1709 	/* We've written all the dirops there are */
1710 	ssp = (SEGSUM *)sp->segsum;
1711 	lfs_ss_setflags(fs, ssp, lfs_ss_getflags(fs, ssp) & ~(SS_CONT));
1712 	lfs_finalize_fs_seguse(fs);
1713 	(void) lfs_writeseg(fs, sp);
1714 	lfs_segunlock(fs);
1715 
1716 	return error;
1717 }
1718 
1719 /*
1720  * Flush all vnodes for which the pagedaemon has requested pageouts.
1721  * Skip over any files that are marked VU_DIROP (since lfs_flush_dirop()
1722  * has just run, this would be an error).  If we have to skip a vnode
1723  * for any reason, just skip it; if we have to wait for the cleaner,
1724  * abort.  The writer daemon will call us again later.
1725  */
1726 int
1727 lfs_flush_pchain(struct lfs *fs)
1728 {
1729 	struct inode *ip, *nip;
1730 	struct vnode *vp;
1731 	extern int lfs_dostats;
1732 	struct segment *sp;
1733 	int error, error2;
1734 
1735 	ASSERT_NO_SEGLOCK(fs);
1736 
1737 	if (fs->lfs_ronly)
1738 		return EROFS;
1739 
1740 	mutex_enter(&lfs_lock);
1741 	if (TAILQ_FIRST(&fs->lfs_pchainhd) == NULL) {
1742 		mutex_exit(&lfs_lock);
1743 		return 0;
1744 	} else
1745 		mutex_exit(&lfs_lock);
1746 
1747 	/* Get dirops out of the way */
1748 	if ((error = lfs_flush_dirops(fs)) != 0)
1749 		return error;
1750 
1751 	if (lfs_dostats)
1752 		++lfs_stats.flush_invoked;
1753 
1754 	/*
1755 	 * Inline lfs_segwrite/lfs_writevnodes, but just for pageouts.
1756 	 */
1757 	lfs_imtime(fs);
1758 	lfs_seglock(fs, 0);
1759 	sp = fs->lfs_sp;
1760 
1761 	/*
1762 	 * lfs_writevnodes, optimized to clear pageout requests.
1763 	 * Only write non-dirop files that are in the pageout queue.
1764 	 * We're very conservative about what we write; we want to be
1765 	 * fast and async.
1766 	 */
1767 	mutex_enter(&lfs_lock);
1768     top:
1769 	for (ip = TAILQ_FIRST(&fs->lfs_pchainhd); ip != NULL; ip = nip) {
1770 		struct mount *mp = ITOV(ip)->v_mount;
1771 		ino_t ino = ip->i_number;
1772 
1773 		nip = TAILQ_NEXT(ip, i_lfs_pchain);
1774 
1775 		if (!(ip->i_state & IN_PAGING))
1776 			goto top;
1777 
1778 		mutex_exit(&lfs_lock);
1779 		if (vcache_get(mp, &ino, sizeof(ino), &vp) != 0) {
1780 			mutex_enter(&lfs_lock);
1781 			continue;
1782 		};
1783 		if (vn_lock(vp, LK_EXCLUSIVE | LK_NOWAIT) != 0) {
1784 			vrele(vp);
1785 			mutex_enter(&lfs_lock);
1786 			continue;
1787 		}
1788 		ip = VTOI(vp);
1789 		mutex_enter(&lfs_lock);
1790 		if ((vp->v_uflag & VU_DIROP) != 0 || vp->v_type != VREG ||
1791 		    !(ip->i_state & IN_PAGING)) {
1792 			mutex_exit(&lfs_lock);
1793 			vput(vp);
1794 			mutex_enter(&lfs_lock);
1795 			goto top;
1796 		}
1797 		mutex_exit(&lfs_lock);
1798 
1799 		error = lfs_writefile(fs, sp, vp);
1800 		if (!VPISEMPTY(vp) && !WRITEINPROG(vp) &&
1801 		    !(ip->i_state & IN_ALLMOD)) {
1802 		    	mutex_enter(&lfs_lock);
1803 			LFS_SET_UINO(ip, IN_MODIFIED);
1804 		    	mutex_exit(&lfs_lock);
1805 		}
1806 		KDASSERT(ip->i_number != LFS_IFILE_INUM);
1807 		error2 = lfs_writeinode(fs, sp, ip);
1808 
1809 		VOP_UNLOCK(vp);
1810 		vrele(vp);
1811 
1812 		if (error == EAGAIN || error2 == EAGAIN) {
1813 			lfs_writeseg(fs, sp);
1814 			mutex_enter(&lfs_lock);
1815 			break;
1816 		}
1817 		mutex_enter(&lfs_lock);
1818 	}
1819 	mutex_exit(&lfs_lock);
1820 	(void) lfs_writeseg(fs, sp);
1821 	lfs_segunlock(fs);
1822 
1823 	return 0;
1824 }
1825 
1826 /*
1827  * Conversion for compat.
1828  */
1829 static void
1830 block_info_from_70(BLOCK_INFO *bi, const BLOCK_INFO_70 *bi70)
1831 {
1832 	bi->bi_inode = bi70->bi_inode;
1833 	bi->bi_lbn = bi70->bi_lbn;
1834 	bi->bi_daddr = bi70->bi_daddr;
1835 	bi->bi_segcreate = bi70->bi_segcreate;
1836 	bi->bi_version = bi70->bi_version;
1837 	bi->bi_bp = bi70->bi_bp;
1838 	bi->bi_size = bi70->bi_size;
1839 }
1840 
1841 static void
1842 block_info_to_70(BLOCK_INFO_70 *bi70, const BLOCK_INFO *bi)
1843 {
1844 	bi70->bi_inode = bi->bi_inode;
1845 	bi70->bi_lbn = bi->bi_lbn;
1846 	bi70->bi_daddr = bi->bi_daddr;
1847 	bi70->bi_segcreate = bi->bi_segcreate;
1848 	bi70->bi_version = bi->bi_version;
1849 	bi70->bi_bp = bi->bi_bp;
1850 	bi70->bi_size = bi->bi_size;
1851 }
1852 
1853 /*
1854  * Provide a fcntl interface to sys_lfs_{segwait,bmapv,markv}.
1855  */
1856 int
1857 lfs_fcntl(void *v)
1858 {
1859 	struct vop_fcntl_args /* {
1860 		struct vnode *a_vp;
1861 		u_int a_command;
1862 		void * a_data;
1863 		int  a_fflag;
1864 		kauth_cred_t a_cred;
1865 	} */ *ap = v;
1866 	struct timeval tv;
1867 	struct timeval *tvp;
1868 	BLOCK_INFO *blkiov;
1869 	BLOCK_INFO_70 *blkiov70;
1870 	CLEANERINFO *cip;
1871 	SEGUSE *sup;
1872 	int blkcnt, i, error;
1873 	size_t fh_size;
1874 	struct lfs_fcntl_markv blkvp;
1875 	struct lfs_fcntl_markv_70 blkvp70;
1876 	struct lwp *l;
1877 	fsid_t *fsidp;
1878 	struct lfs *fs;
1879 	struct buf *bp;
1880 	fhandle_t *fhp;
1881 	daddr_t off;
1882 	int oclean;
1883 
1884 	/* Only respect LFS fcntls on fs root or Ifile */
1885 	if (VTOI(ap->a_vp)->i_number != ULFS_ROOTINO &&
1886 	    VTOI(ap->a_vp)->i_number != LFS_IFILE_INUM) {
1887 		return ulfs_fcntl(v);
1888 	}
1889 
1890 	/* Avoid locking a draining lock */
1891 	if (ap->a_vp->v_mount->mnt_iflag & IMNT_UNMOUNT) {
1892 		return ESHUTDOWN;
1893 	}
1894 
1895 	/* LFS control and monitoring fcntls are available only to root */
1896 	l = curlwp;
1897 	if (((ap->a_command & 0xff00) >> 8) == 'L' &&
1898 	    (error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_LFS,
1899 	     KAUTH_REQ_SYSTEM_LFS_FCNTL, NULL, NULL, NULL)) != 0)
1900 		return (error);
1901 
1902 	fs = VTOI(ap->a_vp)->i_lfs;
1903 	fsidp = &ap->a_vp->v_mount->mnt_stat.f_fsidx;
1904 
1905 	error = 0;
1906 	switch ((int)ap->a_command) {
1907 	    case LFCNSEGWAITALL_COMPAT_50:
1908 	    case LFCNSEGWAITALL_COMPAT:
1909 		fsidp = NULL;
1910 		/* FALLTHROUGH */
1911 	    case LFCNSEGWAIT_COMPAT_50:
1912 	    case LFCNSEGWAIT_COMPAT:
1913 		{
1914 			struct timeval50 *tvp50
1915 				= (struct timeval50 *)ap->a_data;
1916 			timeval50_to_timeval(tvp50, &tv);
1917 			tvp = &tv;
1918 		}
1919 		goto segwait_common;
1920 	    case LFCNSEGWAITALL:
1921 		fsidp = NULL;
1922 		/* FALLTHROUGH */
1923 	    case LFCNSEGWAIT:
1924 		tvp = (struct timeval *)ap->a_data;
1925 segwait_common:
1926 		mutex_enter(&lfs_lock);
1927 		++fs->lfs_sleepers;
1928 		mutex_exit(&lfs_lock);
1929 
1930 		error = lfs_segwait(fsidp, tvp);
1931 
1932 		mutex_enter(&lfs_lock);
1933 		if (--fs->lfs_sleepers == 0)
1934 			cv_broadcast(&fs->lfs_sleeperscv);
1935 		mutex_exit(&lfs_lock);
1936 		return error;
1937 
1938 	    case LFCNBMAPV_COMPAT_70:
1939 	    case LFCNMARKV_COMPAT_70:
1940 		blkvp70 = *(struct lfs_fcntl_markv_70 *)ap->a_data;
1941 
1942 		blkcnt = blkvp70.blkcnt;
1943 		if ((u_int) blkcnt > LFS_MARKV_MAXBLKCNT)
1944 			return (EINVAL);
1945 		blkiov = lfs_malloc(fs, blkcnt * sizeof(BLOCK_INFO), LFS_NB_BLKIOV);
1946 		blkiov70 = lfs_malloc(fs, sizeof(BLOCK_INFO_70), LFS_NB_BLKIOV);
1947 		for (i = 0; i < blkcnt; i++) {
1948 			error = copyin(&blkvp70.blkiov[i], blkiov70,
1949 				       sizeof(*blkiov70));
1950 			if (error) {
1951 				lfs_free(fs, blkiov70, LFS_NB_BLKIOV);
1952 				lfs_free(fs, blkiov, LFS_NB_BLKIOV);
1953 				return error;
1954 			}
1955 			block_info_from_70(&blkiov[i], blkiov70);
1956 		}
1957 
1958 		mutex_enter(&lfs_lock);
1959 		++fs->lfs_sleepers;
1960 		mutex_exit(&lfs_lock);
1961 		if (ap->a_command == LFCNBMAPV)
1962 			error = lfs_bmapv(l, fsidp, blkiov, blkcnt);
1963 		else /* LFCNMARKV */
1964 			error = lfs_markv(l, fsidp, blkiov, blkcnt);
1965 		if (error == 0) {
1966 			for (i = 0; i < blkcnt; i++) {
1967 				block_info_to_70(blkiov70, &blkiov[i]);
1968 				error = copyout(blkiov70, &blkvp70.blkiov[i],
1969 						sizeof(*blkiov70));
1970 				if (error) {
1971 					break;
1972 				}
1973 			}
1974 		}
1975 		mutex_enter(&lfs_lock);
1976 		if (--fs->lfs_sleepers == 0)
1977 			cv_broadcast(&fs->lfs_sleeperscv);
1978 		mutex_exit(&lfs_lock);
1979 		lfs_free(fs, blkiov, LFS_NB_BLKIOV);
1980 		return error;
1981 
1982 	    case LFCNBMAPV:
1983 	    case LFCNMARKV:
1984 		blkvp = *(struct lfs_fcntl_markv *)ap->a_data;
1985 
1986 		blkcnt = blkvp.blkcnt;
1987 		if ((u_int) blkcnt > LFS_MARKV_MAXBLKCNT)
1988 			return (EINVAL);
1989 		blkiov = lfs_malloc(fs, blkcnt * sizeof(BLOCK_INFO), LFS_NB_BLKIOV);
1990 		if ((error = copyin(blkvp.blkiov, blkiov,
1991 		     blkcnt * sizeof(BLOCK_INFO))) != 0) {
1992 			lfs_free(fs, blkiov, LFS_NB_BLKIOV);
1993 			return error;
1994 		}
1995 
1996 		mutex_enter(&lfs_lock);
1997 		++fs->lfs_sleepers;
1998 		mutex_exit(&lfs_lock);
1999 		if (ap->a_command == LFCNBMAPV)
2000 			error = lfs_bmapv(l, fsidp, blkiov, blkcnt);
2001 		else /* LFCNMARKV */
2002 			error = lfs_markv(l, fsidp, blkiov, blkcnt);
2003 		if (error == 0)
2004 			error = copyout(blkiov, blkvp.blkiov,
2005 					blkcnt * sizeof(BLOCK_INFO));
2006 		mutex_enter(&lfs_lock);
2007 		if (--fs->lfs_sleepers == 0)
2008 			cv_broadcast(&fs->lfs_sleeperscv);
2009 		mutex_exit(&lfs_lock);
2010 		lfs_free(fs, blkiov, LFS_NB_BLKIOV);
2011 		return error;
2012 
2013 	    case LFCNRECLAIM:
2014 		/*
2015 		 * Flush dirops and write Ifile, allowing empty segments
2016 		 * to be immediately reclaimed.
2017 		 */
2018 		lfs_writer_enter(fs, "pndirop");
2019 		off = lfs_sb_getoffset(fs);
2020 		lfs_seglock(fs, SEGM_FORCE_CKP | SEGM_CKP);
2021 		lfs_flush_dirops(fs);
2022 		LFS_CLEANERINFO(cip, fs, bp);
2023 		oclean = lfs_ci_getclean(fs, cip);
2024 		LFS_SYNC_CLEANERINFO(cip, fs, bp, 1);
2025 		lfs_segwrite(ap->a_vp->v_mount, SEGM_FORCE_CKP);
2026 		fs->lfs_sp->seg_flags |= SEGM_PROT;
2027 		lfs_segunlock(fs);
2028 		lfs_writer_leave(fs);
2029 
2030 #ifdef DEBUG
2031 		LFS_CLEANERINFO(cip, fs, bp);
2032 		DLOG((DLOG_CLEAN, "lfs_fcntl: reclaim wrote %" PRId64
2033 		      " blocks, cleaned %" PRId32 " segments (activesb %d)\n",
2034 		      lfs_sb_getoffset(fs) - off,
2035 		      lfs_ci_getclean(fs, cip) - oclean,
2036 		      fs->lfs_activesb));
2037 		LFS_SYNC_CLEANERINFO(cip, fs, bp, 0);
2038 #else
2039 		__USE(oclean);
2040 		__USE(off);
2041 #endif
2042 
2043 		return 0;
2044 
2045 	    case LFCNIFILEFH_COMPAT:
2046 		/* Return the filehandle of the Ifile */
2047 		if ((error = kauth_authorize_system(l->l_cred,
2048 		    KAUTH_SYSTEM_FILEHANDLE, 0, NULL, NULL, NULL)) != 0)
2049 			return (error);
2050 		fhp = (struct fhandle *)ap->a_data;
2051 		fhp->fh_fsid = *fsidp;
2052 		fh_size = 16;	/* former VFS_MAXFIDSIZ */
2053 		return lfs_vptofh(fs->lfs_ivnode, &(fhp->fh_fid), &fh_size);
2054 
2055 	    case LFCNIFILEFH_COMPAT2:
2056 	    case LFCNIFILEFH:
2057 		/* Return the filehandle of the Ifile */
2058 		fhp = (struct fhandle *)ap->a_data;
2059 		fhp->fh_fsid = *fsidp;
2060 		fh_size = sizeof(struct lfs_fhandle) -
2061 		    offsetof(fhandle_t, fh_fid);
2062 		return lfs_vptofh(fs->lfs_ivnode, &(fhp->fh_fid), &fh_size);
2063 
2064 	    case LFCNREWIND:
2065 		/* Move lfs_offset to the lowest-numbered segment */
2066 		return lfs_rewind(fs, *(int *)ap->a_data);
2067 
2068 	    case LFCNINVAL:
2069 		/* Mark a segment SEGUSE_INVAL */
2070 		LFS_SEGENTRY(sup, fs, *(int *)ap->a_data, bp);
2071 		if (sup->su_nbytes > 0) {
2072 			brelse(bp, 0);
2073 			lfs_unset_inval_all(fs);
2074 			return EBUSY;
2075 		}
2076 		sup->su_flags |= SEGUSE_INVAL;
2077 		VOP_BWRITE(bp->b_vp, bp);
2078 		return 0;
2079 
2080 	    case LFCNRESIZE:
2081 		/* Resize the filesystem */
2082 		return lfs_resize_fs(fs, *(int *)ap->a_data);
2083 
2084 	    case LFCNWRAPSTOP:
2085 	    case LFCNWRAPSTOP_COMPAT:
2086 		/*
2087 		 * Hold lfs_newseg at segment 0; if requested, sleep until
2088 		 * the filesystem wraps around.  To support external agents
2089 		 * (dump, fsck-based regression test) that need to look at
2090 		 * a snapshot of the filesystem, without necessarily
2091 		 * requiring that all fs activity stops.
2092 		 */
2093 		if (fs->lfs_stoplwp == curlwp)
2094 			return EALREADY;
2095 
2096 		mutex_enter(&lfs_lock);
2097 		while (fs->lfs_stoplwp != NULL)
2098 			cv_wait(&fs->lfs_stopcv, &lfs_lock);
2099 		fs->lfs_stoplwp = curlwp;
2100 		if (fs->lfs_nowrap == 0)
2101 			log(LOG_NOTICE, "%s: disabled log wrap\n",
2102 			    lfs_sb_getfsmnt(fs));
2103 		++fs->lfs_nowrap;
2104 		if (*(int *)ap->a_data == 1
2105 		    || ap->a_command == LFCNWRAPSTOP_COMPAT) {
2106 			log(LOG_NOTICE, "LFCNSTOPWRAP waiting for log wrap\n");
2107 			error = mtsleep(&fs->lfs_nowrap, PCATCH | PUSER,
2108 				"segwrap", 0, &lfs_lock);
2109 			log(LOG_NOTICE, "LFCNSTOPWRAP done waiting\n");
2110 			if (error) {
2111 				lfs_wrapgo(fs, VTOI(ap->a_vp), 0);
2112 			}
2113 		}
2114 		mutex_exit(&lfs_lock);
2115 		return 0;
2116 
2117 	    case LFCNWRAPGO:
2118 	    case LFCNWRAPGO_COMPAT:
2119 		/*
2120 		 * Having done its work, the agent wakes up the writer.
2121 		 * If the argument is 1, it sleeps until a new segment
2122 		 * is selected.
2123 		 */
2124 		mutex_enter(&lfs_lock);
2125 		error = lfs_wrapgo(fs, VTOI(ap->a_vp),
2126 				   ap->a_command == LFCNWRAPGO_COMPAT ? 1 :
2127 				    *((int *)ap->a_data));
2128 		mutex_exit(&lfs_lock);
2129 		return error;
2130 
2131 	    case LFCNWRAPPASS:
2132 		if ((VTOI(ap->a_vp)->i_lfs_iflags & LFSI_WRAPWAIT))
2133 			return EALREADY;
2134 		mutex_enter(&lfs_lock);
2135 		if (fs->lfs_stoplwp != curlwp) {
2136 			mutex_exit(&lfs_lock);
2137 			return EALREADY;
2138 		}
2139 		if (fs->lfs_nowrap == 0) {
2140 			mutex_exit(&lfs_lock);
2141 			return EBUSY;
2142 		}
2143 		fs->lfs_wrappass = 1;
2144 		wakeup(&fs->lfs_wrappass);
2145 		/* Wait for the log to wrap, if asked */
2146 		if (*(int *)ap->a_data) {
2147 			vref(ap->a_vp);
2148 			VTOI(ap->a_vp)->i_lfs_iflags |= LFSI_WRAPWAIT;
2149 			log(LOG_NOTICE, "LFCNPASS waiting for log wrap\n");
2150 			error = mtsleep(&fs->lfs_nowrap, PCATCH | PUSER,
2151 				"segwrap", 0, &lfs_lock);
2152 			log(LOG_NOTICE, "LFCNPASS done waiting\n");
2153 			VTOI(ap->a_vp)->i_lfs_iflags &= ~LFSI_WRAPWAIT;
2154 			vrele(ap->a_vp);
2155 		}
2156 		mutex_exit(&lfs_lock);
2157 		return error;
2158 
2159 	    case LFCNWRAPSTATUS:
2160 		mutex_enter(&lfs_lock);
2161 		*(int *)ap->a_data = fs->lfs_wrapstatus;
2162 		mutex_exit(&lfs_lock);
2163 		return 0;
2164 
2165 	    default:
2166 		return ulfs_fcntl(v);
2167 	}
2168 	return 0;
2169 }
2170 
2171 /*
2172  * Return the last logical file offset that should be written for this file
2173  * if we're doing a write that ends at "size".	If writing, we need to know
2174  * about sizes on disk, i.e. fragments if there are any; if reading, we need
2175  * to know about entire blocks.
2176  */
2177 void
2178 lfs_gop_size(struct vnode *vp, off_t size, off_t *eobp, int flags)
2179 {
2180 	struct inode *ip = VTOI(vp);
2181 	struct lfs *fs = ip->i_lfs;
2182 	daddr_t olbn, nlbn;
2183 
2184 	olbn = lfs_lblkno(fs, ip->i_size);
2185 	nlbn = lfs_lblkno(fs, size);
2186 	if (!(flags & GOP_SIZE_MEM) && nlbn < ULFS_NDADDR && olbn <= nlbn) {
2187 		*eobp = lfs_fragroundup(fs, size);
2188 	} else {
2189 		*eobp = lfs_blkroundup(fs, size);
2190 	}
2191 }
2192 
2193 #ifdef DEBUG
2194 void lfs_dump_vop(void *);
2195 
2196 void
2197 lfs_dump_vop(void *v)
2198 {
2199 	struct vop_putpages_args /* {
2200 		struct vnode *a_vp;
2201 		voff_t a_offlo;
2202 		voff_t a_offhi;
2203 		int a_flags;
2204 	} */ *ap = v;
2205 
2206 	struct inode *ip = VTOI(ap->a_vp);
2207 	struct lfs *fs = ip->i_lfs;
2208 
2209 #ifdef DDB
2210 	vfs_vnode_print(ap->a_vp, 0, printf);
2211 #endif
2212 	lfs_dump_dinode(fs, ip->i_din);
2213 }
2214 #endif
2215 
2216 int
2217 lfs_mmap(void *v)
2218 {
2219 	struct vop_mmap_args /* {
2220 		const struct vnodeop_desc *a_desc;
2221 		struct vnode *a_vp;
2222 		vm_prot_t a_prot;
2223 		kauth_cred_t a_cred;
2224 	} */ *ap = v;
2225 
2226 	if (VTOI(ap->a_vp)->i_number == LFS_IFILE_INUM)
2227 		return EOPNOTSUPP;
2228 	return ulfs_mmap(v);
2229 }
2230 
2231 static int
2232 lfs_openextattr(void *v)
2233 {
2234 	struct vop_openextattr_args /* {
2235 		struct vnode *a_vp;
2236 		kauth_cred_t a_cred;
2237 		struct proc *a_p;
2238 	} */ *ap = v;
2239 	struct vnode *vp = ap->a_vp;
2240 	struct inode *ip;
2241 	struct ulfsmount *ump;
2242 
2243 	KASSERT(VOP_ISLOCKED(vp) == LK_EXCLUSIVE);
2244 
2245 	ip = VTOI(vp);
2246 	ump = ip->i_ump;
2247 
2248 	/* Not supported for ULFS1 file systems. */
2249 	if (ump->um_fstype == ULFS1)
2250 		return (EOPNOTSUPP);
2251 
2252 	/* XXX Not implemented for ULFS2 file systems. */
2253 	return (EOPNOTSUPP);
2254 }
2255 
2256 static int
2257 lfs_closeextattr(void *v)
2258 {
2259 	struct vop_closeextattr_args /* {
2260 		struct vnode *a_vp;
2261 		int a_commit;
2262 		kauth_cred_t a_cred;
2263 		struct proc *a_p;
2264 	} */ *ap = v;
2265 	struct vnode *vp = ap->a_vp;
2266 	struct inode *ip;
2267 	struct ulfsmount *ump;
2268 
2269 	KASSERT(VOP_ISLOCKED(vp) == LK_EXCLUSIVE);
2270 
2271 	ip = VTOI(vp);
2272 	ump = ip->i_ump;
2273 
2274 	/* Not supported for ULFS1 file systems. */
2275 	if (ump->um_fstype == ULFS1)
2276 		return (EOPNOTSUPP);
2277 
2278 	/* XXX Not implemented for ULFS2 file systems. */
2279 	return (EOPNOTSUPP);
2280 }
2281 
2282 static int
2283 lfs_getextattr(void *v)
2284 {
2285 	struct vop_getextattr_args /* {
2286 		struct vnode *a_vp;
2287 		int a_attrnamespace;
2288 		const char *a_name;
2289 		struct uio *a_uio;
2290 		size_t *a_size;
2291 		kauth_cred_t a_cred;
2292 		struct proc *a_p;
2293 	} */ *ap = v;
2294 	struct vnode *vp = ap->a_vp;
2295 	struct inode *ip;
2296 	struct ulfsmount *ump;
2297 	int error;
2298 
2299 	KASSERT(VOP_ISLOCKED(vp));
2300 
2301 	ip = VTOI(vp);
2302 	ump = ip->i_ump;
2303 
2304 	if (ump->um_fstype == ULFS1) {
2305 #ifdef LFS_EXTATTR
2306 		error = ulfs_getextattr(ap);
2307 #else
2308 		error = EOPNOTSUPP;
2309 #endif
2310 		return error;
2311 	}
2312 
2313 	/* XXX Not implemented for ULFS2 file systems. */
2314 	return (EOPNOTSUPP);
2315 }
2316 
2317 static int
2318 lfs_setextattr(void *v)
2319 {
2320 	struct vop_setextattr_args /* {
2321 		struct vnode *a_vp;
2322 		int a_attrnamespace;
2323 		const char *a_name;
2324 		struct uio *a_uio;
2325 		kauth_cred_t a_cred;
2326 		struct proc *a_p;
2327 	} */ *ap = v;
2328 	struct vnode *vp = ap->a_vp;
2329 	struct inode *ip;
2330 	struct ulfsmount *ump;
2331 	int error;
2332 
2333 	KASSERT(VOP_ISLOCKED(vp) == LK_EXCLUSIVE);
2334 
2335 	ip = VTOI(vp);
2336 	ump = ip->i_ump;
2337 
2338 	if (ump->um_fstype == ULFS1) {
2339 #ifdef LFS_EXTATTR
2340 		error = ulfs_setextattr(ap);
2341 #else
2342 		error = EOPNOTSUPP;
2343 #endif
2344 		return error;
2345 	}
2346 
2347 	/* XXX Not implemented for ULFS2 file systems. */
2348 	return (EOPNOTSUPP);
2349 }
2350 
2351 static int
2352 lfs_listextattr(void *v)
2353 {
2354 	struct vop_listextattr_args /* {
2355 		struct vnode *a_vp;
2356 		int a_attrnamespace;
2357 		struct uio *a_uio;
2358 		size_t *a_size;
2359 		kauth_cred_t a_cred;
2360 		struct proc *a_p;
2361 	} */ *ap = v;
2362 	struct vnode *vp = ap->a_vp;
2363 	struct inode *ip;
2364 	struct ulfsmount *ump;
2365 	int error;
2366 
2367 	KASSERT(VOP_ISLOCKED(vp));
2368 
2369 	ip = VTOI(vp);
2370 	ump = ip->i_ump;
2371 
2372 	if (ump->um_fstype == ULFS1) {
2373 #ifdef LFS_EXTATTR
2374 		error = ulfs_listextattr(ap);
2375 #else
2376 		error = EOPNOTSUPP;
2377 #endif
2378 		return error;
2379 	}
2380 
2381 	/* XXX Not implemented for ULFS2 file systems. */
2382 	return (EOPNOTSUPP);
2383 }
2384 
2385 static int
2386 lfs_deleteextattr(void *v)
2387 {
2388 	struct vop_deleteextattr_args /* {
2389 		struct vnode *a_vp;
2390 		int a_attrnamespace;
2391 		kauth_cred_t a_cred;
2392 		struct proc *a_p;
2393 	} */ *ap = v;
2394 	struct vnode *vp = ap->a_vp;
2395 	struct inode *ip;
2396 	struct ulfsmount *ump;
2397 	int error;
2398 
2399 	KASSERT(VOP_ISLOCKED(vp) == LK_EXCLUSIVE);
2400 
2401 	ip = VTOI(vp);
2402 	ump = ip->i_ump;
2403 
2404 	if (ump->um_fstype == ULFS1) {
2405 #ifdef LFS_EXTATTR
2406 		error = ulfs_deleteextattr(ap);
2407 #else
2408 		error = EOPNOTSUPP;
2409 #endif
2410 		return error;
2411 	}
2412 
2413 	/* XXX Not implemented for ULFS2 file systems. */
2414 	return (EOPNOTSUPP);
2415 }
2416 
2417