xref: /netbsd-src/sys/ufs/lfs/lfs_vnops.c (revision b7b7574d3bf8eeb51a1fa3977b59142ec6434a55)
1 /*	$NetBSD: lfs_vnops.c,v 1.268 2014/05/17 07:10:27 dholland 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 1.213 2013/06/08 05:47:02 kardel 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 1.268 2014/05/17 07:10:27 dholland 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 #include <sys/fstrans.h>
151 
152 #include <miscfs/fifofs/fifo.h>
153 #include <miscfs/genfs/genfs.h>
154 #include <miscfs/specfs/specdev.h>
155 
156 #include <ufs/lfs/ulfs_inode.h>
157 #include <ufs/lfs/ulfsmount.h>
158 #include <ufs/lfs/ulfs_bswap.h>
159 #include <ufs/lfs/ulfs_extern.h>
160 
161 #include <uvm/uvm.h>
162 #include <uvm/uvm_pmap.h>
163 #include <uvm/uvm_stat.h>
164 #include <uvm/uvm_pager.h>
165 
166 #include <ufs/lfs/lfs.h>
167 #include <ufs/lfs/lfs_kernel.h>
168 #include <ufs/lfs/lfs_extern.h>
169 
170 extern pid_t lfs_writer_daemon;
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 /*
181  * A virgin directory (no blushing please).
182  */
183 static const struct lfs_dirtemplate mastertemplate = {
184 	0,	12,			LFS_DT_DIR,	1,	".",
185 	0,	LFS_DIRBLKSIZ - 12,	LFS_DT_DIR,	2,	".."
186 };
187 
188 /* Global vfs data structures for lfs. */
189 int (**lfs_vnodeop_p)(void *);
190 const struct vnodeopv_entry_desc lfs_vnodeop_entries[] = {
191 	{ &vop_default_desc, vn_default_error },
192 	{ &vop_lookup_desc, ulfs_lookup },		/* lookup */
193 	{ &vop_create_desc, lfs_create },		/* create */
194 	{ &vop_whiteout_desc, ulfs_whiteout },		/* whiteout */
195 	{ &vop_mknod_desc, lfs_mknod },			/* mknod */
196 	{ &vop_open_desc, ulfs_open },			/* open */
197 	{ &vop_close_desc, lfs_close },			/* close */
198 	{ &vop_access_desc, ulfs_access },		/* access */
199 	{ &vop_getattr_desc, lfs_getattr },		/* getattr */
200 	{ &vop_setattr_desc, lfs_setattr },		/* setattr */
201 	{ &vop_read_desc, lfs_read },			/* read */
202 	{ &vop_write_desc, lfs_write },			/* write */
203 	{ &vop_ioctl_desc, ulfs_ioctl },		/* ioctl */
204 	{ &vop_fcntl_desc, lfs_fcntl },			/* fcntl */
205 	{ &vop_poll_desc, ulfs_poll },			/* poll */
206 	{ &vop_kqfilter_desc, genfs_kqfilter },		/* kqfilter */
207 	{ &vop_revoke_desc, ulfs_revoke },		/* revoke */
208 	{ &vop_mmap_desc, lfs_mmap },			/* mmap */
209 	{ &vop_fsync_desc, lfs_fsync },			/* fsync */
210 	{ &vop_seek_desc, ulfs_seek },			/* seek */
211 	{ &vop_remove_desc, lfs_remove },		/* remove */
212 	{ &vop_link_desc, lfs_link },			/* link */
213 	{ &vop_rename_desc, lfs_rename },		/* rename */
214 	{ &vop_mkdir_desc, lfs_mkdir },			/* mkdir */
215 	{ &vop_rmdir_desc, lfs_rmdir },			/* rmdir */
216 	{ &vop_symlink_desc, lfs_symlink },		/* symlink */
217 	{ &vop_readdir_desc, ulfs_readdir },		/* readdir */
218 	{ &vop_readlink_desc, ulfs_readlink },		/* readlink */
219 	{ &vop_abortop_desc, ulfs_abortop },		/* abortop */
220 	{ &vop_inactive_desc, lfs_inactive },		/* inactive */
221 	{ &vop_reclaim_desc, lfs_reclaim },		/* reclaim */
222 	{ &vop_lock_desc, ulfs_lock },			/* lock */
223 	{ &vop_unlock_desc, ulfs_unlock },		/* unlock */
224 	{ &vop_bmap_desc, ulfs_bmap },			/* bmap */
225 	{ &vop_strategy_desc, lfs_strategy },		/* strategy */
226 	{ &vop_print_desc, ulfs_print },		/* print */
227 	{ &vop_islocked_desc, ulfs_islocked },		/* islocked */
228 	{ &vop_pathconf_desc, ulfs_pathconf },		/* pathconf */
229 	{ &vop_advlock_desc, ulfs_advlock },		/* advlock */
230 	{ &vop_bwrite_desc, lfs_bwrite },		/* bwrite */
231 	{ &vop_getpages_desc, lfs_getpages },		/* getpages */
232 	{ &vop_putpages_desc, lfs_putpages },		/* putpages */
233 	{ &vop_openextattr_desc, lfs_openextattr },	/* openextattr */
234 	{ &vop_closeextattr_desc, lfs_closeextattr },	/* closeextattr */
235 	{ &vop_getextattr_desc, lfs_getextattr },	/* getextattr */
236 	{ &vop_setextattr_desc, lfs_setextattr },	/* setextattr */
237 	{ &vop_listextattr_desc, lfs_listextattr },	/* listextattr */
238 	{ &vop_deleteextattr_desc, lfs_deleteextattr },	/* deleteextattr */
239 	{ NULL, NULL }
240 };
241 const struct vnodeopv_desc lfs_vnodeop_opv_desc =
242 	{ &lfs_vnodeop_p, lfs_vnodeop_entries };
243 
244 int (**lfs_specop_p)(void *);
245 const struct vnodeopv_entry_desc lfs_specop_entries[] = {
246 	{ &vop_default_desc, vn_default_error },
247 	{ &vop_lookup_desc, spec_lookup },		/* lookup */
248 	{ &vop_create_desc, spec_create },		/* create */
249 	{ &vop_mknod_desc, spec_mknod },		/* mknod */
250 	{ &vop_open_desc, spec_open },			/* open */
251 	{ &vop_close_desc, lfsspec_close },		/* close */
252 	{ &vop_access_desc, ulfs_access },		/* access */
253 	{ &vop_getattr_desc, lfs_getattr },		/* getattr */
254 	{ &vop_setattr_desc, lfs_setattr },		/* setattr */
255 	{ &vop_read_desc, ulfsspec_read },		/* read */
256 	{ &vop_write_desc, ulfsspec_write },		/* write */
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_ioctl_desc, vn_fifo_bypass },		/* ioctl */
312 	{ &vop_fcntl_desc, ulfs_fcntl },		/* fcntl */
313 	{ &vop_poll_desc, vn_fifo_bypass },		/* poll */
314 	{ &vop_kqfilter_desc, vn_fifo_bypass },		/* kqfilter */
315 	{ &vop_revoke_desc, vn_fifo_bypass },		/* revoke */
316 	{ &vop_mmap_desc, vn_fifo_bypass },		/* mmap */
317 	{ &vop_fsync_desc, vn_fifo_bypass },		/* fsync */
318 	{ &vop_seek_desc, vn_fifo_bypass },		/* seek */
319 	{ &vop_remove_desc, vn_fifo_bypass },		/* remove */
320 	{ &vop_link_desc, vn_fifo_bypass },		/* link */
321 	{ &vop_rename_desc, vn_fifo_bypass },		/* rename */
322 	{ &vop_mkdir_desc, vn_fifo_bypass },		/* mkdir */
323 	{ &vop_rmdir_desc, vn_fifo_bypass },		/* rmdir */
324 	{ &vop_symlink_desc, vn_fifo_bypass },		/* symlink */
325 	{ &vop_readdir_desc, vn_fifo_bypass },		/* readdir */
326 	{ &vop_readlink_desc, vn_fifo_bypass },		/* readlink */
327 	{ &vop_abortop_desc, vn_fifo_bypass },		/* abortop */
328 	{ &vop_inactive_desc, lfs_inactive },		/* inactive */
329 	{ &vop_reclaim_desc, lfs_reclaim },		/* reclaim */
330 	{ &vop_lock_desc, ulfs_lock },			/* lock */
331 	{ &vop_unlock_desc, ulfs_unlock },		/* unlock */
332 	{ &vop_bmap_desc, vn_fifo_bypass },		/* bmap */
333 	{ &vop_strategy_desc, vn_fifo_bypass },		/* strategy */
334 	{ &vop_print_desc, ulfs_print },		/* print */
335 	{ &vop_islocked_desc, ulfs_islocked },		/* islocked */
336 	{ &vop_pathconf_desc, vn_fifo_bypass },		/* pathconf */
337 	{ &vop_advlock_desc, vn_fifo_bypass },		/* advlock */
338 	{ &vop_bwrite_desc, lfs_bwrite },		/* bwrite */
339 	{ &vop_putpages_desc, vn_fifo_bypass },		/* putpages */
340 	{ &vop_openextattr_desc, lfs_openextattr },	/* openextattr */
341 	{ &vop_closeextattr_desc, lfs_closeextattr },	/* closeextattr */
342 	{ &vop_getextattr_desc, lfs_getextattr },	/* getextattr */
343 	{ &vop_setextattr_desc, lfs_setextattr },	/* setextattr */
344 	{ &vop_listextattr_desc, lfs_listextattr },	/* listextattr */
345 	{ &vop_deleteextattr_desc, lfs_deleteextattr },	/* deleteextattr */
346 	{ NULL, NULL }
347 };
348 const struct vnodeopv_desc lfs_fifoop_opv_desc =
349 	{ &lfs_fifoop_p, lfs_fifoop_entries };
350 
351 #define	LFS_READWRITE
352 #include <ufs/lfs/ulfs_readwrite.c>
353 #undef	LFS_READWRITE
354 
355 /*
356  * Synch an open file.
357  */
358 /* ARGSUSED */
359 int
360 lfs_fsync(void *v)
361 {
362 	struct vop_fsync_args /* {
363 		struct vnode *a_vp;
364 		kauth_cred_t a_cred;
365 		int a_flags;
366 		off_t offlo;
367 		off_t offhi;
368 	} */ *ap = v;
369 	struct vnode *vp = ap->a_vp;
370 	int error, wait;
371 	struct inode *ip = VTOI(vp);
372 	struct lfs *fs = ip->i_lfs;
373 
374 	/* If we're mounted read-only, don't try to sync. */
375 	if (fs->lfs_ronly)
376 		return 0;
377 
378 	/* If a removed vnode is being cleaned, no need to sync here. */
379 	if ((ap->a_flags & FSYNC_RECLAIM) != 0 && ip->i_mode == 0)
380 		return 0;
381 
382 	/*
383 	 * Trickle sync simply adds this vnode to the pager list, as if
384 	 * the pagedaemon had requested a pageout.
385 	 */
386 	if (ap->a_flags & FSYNC_LAZY) {
387 		if (lfs_ignore_lazy_sync == 0) {
388 			mutex_enter(&lfs_lock);
389 			if (!(ip->i_flags & IN_PAGING)) {
390 				ip->i_flags |= IN_PAGING;
391 				TAILQ_INSERT_TAIL(&fs->lfs_pchainhd, ip,
392 						  i_lfs_pchain);
393 			}
394 			wakeup(&lfs_writer_daemon);
395 			mutex_exit(&lfs_lock);
396 		}
397 		return 0;
398 	}
399 
400 	/*
401 	 * If a vnode is bring cleaned, flush it out before we try to
402 	 * reuse it.  This prevents the cleaner from writing files twice
403 	 * in the same partial segment, causing an accounting underflow.
404 	 */
405 	if (ap->a_flags & FSYNC_RECLAIM && ip->i_flags & IN_CLEANING) {
406 		lfs_vflush(vp);
407 	}
408 
409 	wait = (ap->a_flags & FSYNC_WAIT);
410 	do {
411 		mutex_enter(vp->v_interlock);
412 		error = VOP_PUTPAGES(vp, trunc_page(ap->a_offlo),
413 				     round_page(ap->a_offhi),
414 				     PGO_CLEANIT | (wait ? PGO_SYNCIO : 0));
415 		if (error == EAGAIN) {
416 			mutex_enter(&lfs_lock);
417 			mtsleep(&fs->lfs_avail, PCATCH | PUSER, "lfs_fsync",
418 				hz / 100 + 1, &lfs_lock);
419 			mutex_exit(&lfs_lock);
420 		}
421 	} while (error == EAGAIN);
422 	if (error)
423 		return error;
424 
425 	if ((ap->a_flags & FSYNC_DATAONLY) == 0)
426 		error = lfs_update(vp, NULL, NULL, wait ? UPDATE_WAIT : 0);
427 
428 	if (error == 0 && ap->a_flags & FSYNC_CACHE) {
429 		int l = 0;
430 		error = VOP_IOCTL(ip->i_devvp, DIOCCACHESYNC, &l, FWRITE,
431 				  curlwp->l_cred);
432 	}
433 	if (wait && !VPISEMPTY(vp))
434 		LFS_SET_UINO(ip, IN_MODIFIED);
435 
436 	return error;
437 }
438 
439 /*
440  * Take IN_ADIROP off, then call ulfs_inactive.
441  */
442 int
443 lfs_inactive(void *v)
444 {
445 	struct vop_inactive_args /* {
446 		struct vnode *a_vp;
447 	} */ *ap = v;
448 
449 	lfs_unmark_vnode(ap->a_vp);
450 
451 	/*
452 	 * The Ifile is only ever inactivated on unmount.
453 	 * Streamline this process by not giving it more dirty blocks.
454 	 */
455 	if (VTOI(ap->a_vp)->i_number == LFS_IFILE_INUM) {
456 		mutex_enter(&lfs_lock);
457 		LFS_CLR_UINO(VTOI(ap->a_vp), IN_ALLMOD);
458 		mutex_exit(&lfs_lock);
459 		VOP_UNLOCK(ap->a_vp);
460 		return 0;
461 	}
462 
463 #ifdef DEBUG
464 	/*
465 	 * This might happen on unmount.
466 	 * XXX If it happens at any other time, it should be a panic.
467 	 */
468 	if (ap->a_vp->v_uflag & VU_DIROP) {
469 		struct inode *ip = VTOI(ap->a_vp);
470 		printf("lfs_inactive: inactivating VU_DIROP? ino = %d\n", (int)ip->i_number);
471 	}
472 #endif /* DIAGNOSTIC */
473 
474 	return ulfs_inactive(v);
475 }
476 
477 int
478 lfs_set_dirop(struct vnode *dvp, struct vnode *vp)
479 {
480 	struct lfs *fs;
481 	int error;
482 
483 	KASSERT(VOP_ISLOCKED(dvp));
484 	KASSERT(vp == NULL || VOP_ISLOCKED(vp));
485 
486 	fs = VTOI(dvp)->i_lfs;
487 
488 	ASSERT_NO_SEGLOCK(fs);
489 	/*
490 	 * LFS_NRESERVE calculates direct and indirect blocks as well
491 	 * as an inode block; an overestimate in most cases.
492 	 */
493 	if ((error = lfs_reserve(fs, dvp, vp, LFS_NRESERVE(fs))) != 0)
494 		return (error);
495 
496     restart:
497 	mutex_enter(&lfs_lock);
498 	if (fs->lfs_dirops == 0) {
499 		mutex_exit(&lfs_lock);
500 		lfs_check(dvp, LFS_UNUSED_LBN, 0);
501 		mutex_enter(&lfs_lock);
502 	}
503 	while (fs->lfs_writer) {
504 		error = mtsleep(&fs->lfs_dirops, (PRIBIO + 1) | PCATCH,
505 		    "lfs_sdirop", 0, &lfs_lock);
506 		if (error == EINTR) {
507 			mutex_exit(&lfs_lock);
508 			goto unreserve;
509 		}
510 	}
511 	if (lfs_dirvcount > LFS_MAX_DIROP && fs->lfs_dirops == 0) {
512 		wakeup(&lfs_writer_daemon);
513 		mutex_exit(&lfs_lock);
514 		preempt();
515 		goto restart;
516 	}
517 
518 	if (lfs_dirvcount > LFS_MAX_DIROP) {
519 		DLOG((DLOG_DIROP, "lfs_set_dirop: sleeping with dirops=%d, "
520 		      "dirvcount=%d\n", fs->lfs_dirops, lfs_dirvcount));
521 		if ((error = mtsleep(&lfs_dirvcount,
522 		    PCATCH | PUSER | PNORELOCK, "lfs_maxdirop", 0,
523 		    &lfs_lock)) != 0) {
524 			goto unreserve;
525 		}
526 		goto restart;
527 	}
528 
529 	++fs->lfs_dirops;
530 	/* fs->lfs_doifile = 1; */ /* XXX why? --ks */
531 	mutex_exit(&lfs_lock);
532 
533 	/* Hold a reference so SET_ENDOP will be happy */
534 	vref(dvp);
535 	if (vp) {
536 		vref(vp);
537 		MARK_VNODE(vp);
538 	}
539 
540 	MARK_VNODE(dvp);
541 	return 0;
542 
543   unreserve:
544 	lfs_reserve(fs, dvp, vp, -LFS_NRESERVE(fs));
545 	return error;
546 }
547 
548 /*
549  * Opposite of lfs_set_dirop... mostly. For now at least must call
550  * UNMARK_VNODE(dvp) explicitly first. (XXX: clean that up)
551  */
552 void
553 lfs_unset_dirop(struct lfs *fs, struct vnode *dvp, const char *str)
554 {
555 	mutex_enter(&lfs_lock);
556 	--fs->lfs_dirops;
557 	if (!fs->lfs_dirops) {
558 		if (fs->lfs_nadirop) {
559 			panic("lfs_unset_dirop: %s: no dirops but "
560 			      " nadirop=%d", str,
561 			      fs->lfs_nadirop);
562 		}
563 		wakeup(&fs->lfs_writer);
564 		mutex_exit(&lfs_lock);
565 		lfs_check(dvp, LFS_UNUSED_LBN, 0);
566 	} else {
567 		mutex_exit(&lfs_lock);
568 	}
569 	lfs_reserve(fs, dvp, NULL, -LFS_NRESERVE(fs));
570 }
571 
572 void
573 lfs_mark_vnode(struct vnode *vp)
574 {
575 	struct inode *ip = VTOI(vp);
576 	struct lfs *fs = ip->i_lfs;
577 
578 	mutex_enter(&lfs_lock);
579 	if (!(ip->i_flag & IN_ADIROP)) {
580 		if (!(vp->v_uflag & VU_DIROP)) {
581 			mutex_exit(&lfs_lock);
582 			mutex_enter(vp->v_interlock);
583 			if (lfs_vref(vp) != 0)
584 				panic("lfs_mark_vnode: could not vref");
585 			mutex_enter(&lfs_lock);
586 			++lfs_dirvcount;
587 			++fs->lfs_dirvcount;
588 			TAILQ_INSERT_TAIL(&fs->lfs_dchainhd, ip, i_lfs_dchain);
589 			vp->v_uflag |= VU_DIROP;
590 		}
591 		++fs->lfs_nadirop;
592 		ip->i_flag &= ~IN_CDIROP;
593 		ip->i_flag |= IN_ADIROP;
594 	} else
595 		KASSERT(vp->v_uflag & VU_DIROP);
596 	mutex_exit(&lfs_lock);
597 }
598 
599 void
600 lfs_unmark_vnode(struct vnode *vp)
601 {
602 	struct inode *ip = VTOI(vp);
603 
604 	mutex_enter(&lfs_lock);
605 	if (ip && (ip->i_flag & IN_ADIROP)) {
606 		KASSERT(vp->v_uflag & VU_DIROP);
607 		--ip->i_lfs->lfs_nadirop;
608 		ip->i_flag &= ~IN_ADIROP;
609 	}
610 	mutex_exit(&lfs_lock);
611 }
612 
613 int
614 lfs_symlink(void *v)
615 {
616 	struct vop_symlink_v3_args /* {
617 		struct vnode *a_dvp;
618 		struct vnode **a_vpp;
619 		struct componentname *a_cnp;
620 		struct vattr *a_vap;
621 		char *a_target;
622 	} */ *ap = v;
623 	struct lfs *fs;
624 	struct vnode *dvp, **vpp;
625 	struct inode *ip;
626 	struct ulfs_lookup_results *ulr;
627 	ssize_t len; /* XXX should be size_t */
628 	int error;
629 
630 	dvp = ap->a_dvp;
631 	vpp = ap->a_vpp;
632 
633 	KASSERT(vpp != NULL);
634 	KASSERT(*vpp == NULL);
635 
636 	/* XXX should handle this material another way */
637 	ulr = &VTOI(ap->a_dvp)->i_crap;
638 	ULFS_CHECK_CRAPCOUNTER(VTOI(ap->a_dvp));
639 
640 	fs = VFSTOULFS(dvp->v_mount)->um_lfs;
641 	ASSERT_NO_SEGLOCK(fs);
642 	if (fs->lfs_ronly) {
643 		return EROFS;
644 	}
645 
646 	/*
647 	 * Get a new vnode *before* adjusting the dirop count, to
648 	 * avoid a deadlock in getnewvnode(), if we have a stacked
649 	 * filesystem mounted on top of us.
650 	 *
651 	 * NB: this means we have to destroy the new vnode on error.
652 	 */
653 
654 	error = getnewvnode(VT_LFS, dvp->v_mount, lfs_vnodeop_p, NULL, vpp);
655 	if (error) {
656 		DLOG((DLOG_ALLOC, "lfs_mkdir: dvp %p error %d\n", dvp, error));
657 		return error;
658 	}
659 	KASSERT(*vpp != NULL);
660 
661 	error = lfs_set_dirop(dvp, NULL);
662 	if (error) {
663 		ungetnewvnode(*vpp);
664 		*vpp = NULL;
665 		return error;
666 	}
667 
668 	fstrans_start(dvp->v_mount, FSTRANS_SHARED);
669 	error = ulfs_makeinode(LFS_IFLNK | ap->a_vap->va_mode, dvp, ulr,
670 			      vpp, ap->a_cnp);
671 	if (error) {
672 		goto out;
673 	}
674 
675 	VN_KNOTE(ap->a_dvp, NOTE_WRITE);
676 	ip = VTOI(*vpp);
677 
678 	len = strlen(ap->a_target);
679 	if (len < ip->i_lfs->um_maxsymlinklen) {
680 		memcpy((char *)SHORTLINK(ip), ap->a_target, len);
681 		ip->i_size = len;
682 		DIP_ASSIGN(ip, size, len);
683 		uvm_vnp_setsize(*vpp, ip->i_size);
684 		ip->i_flag |= IN_CHANGE | IN_UPDATE;
685 		if ((*vpp)->v_mount->mnt_flag & MNT_RELATIME)
686 			ip->i_flag |= IN_ACCESS;
687 	} else {
688 		error = vn_rdwr(UIO_WRITE, *vpp, ap->a_target, len, (off_t)0,
689 		    UIO_SYSSPACE, IO_NODELOCKED | IO_JOURNALLOCKED,
690 		    ap->a_cnp->cn_cred, NULL, NULL);
691 	}
692 
693 	VOP_UNLOCK(*vpp);
694 	if (error)
695 		vrele(*vpp);
696 
697 out:
698 	fstrans_done(dvp->v_mount);
699 
700 	UNMARK_VNODE(dvp);
701 	/* XXX: is it even possible for the symlink to get MARK'd? */
702 	UNMARK_VNODE(*vpp);
703 	if (!((*vpp)->v_uflag & VU_DIROP)) {
704 		KASSERT(error != 0);
705 		ungetnewvnode(*vpp);
706 		*vpp = NULL;
707 	}
708 	else {
709 		KASSERT(error == 0);
710 	}
711 	lfs_unset_dirop(fs, dvp, "symlink");
712 
713 	vrele(dvp);
714 	return (error);
715 }
716 
717 int
718 lfs_mknod(void *v)
719 {
720 	struct vop_mknod_v3_args	/* {
721 		struct vnode *a_dvp;
722 		struct vnode **a_vpp;
723 		struct componentname *a_cnp;
724 		struct vattr *a_vap;
725 	} */ *ap = v;
726 	struct lfs *fs;
727 	struct vnode *dvp, **vpp;
728 	struct vattr *vap;
729 	struct inode *ip;
730 	int error;
731 	struct mount	*mp;
732 	ino_t		ino;
733 	struct ulfs_lookup_results *ulr;
734 
735 	dvp = ap->a_dvp;
736 	vpp = ap->a_vpp;
737 	vap = ap->a_vap;
738 
739 	KASSERT(vpp != NULL);
740 	KASSERT(*vpp == NULL);
741 
742 	/* XXX should handle this material another way */
743 	ulr = &VTOI(dvp)->i_crap;
744 	ULFS_CHECK_CRAPCOUNTER(VTOI(dvp));
745 
746 	fs = VFSTOULFS(dvp->v_mount)->um_lfs;
747 	ASSERT_NO_SEGLOCK(fs);
748 	if (fs->lfs_ronly) {
749 		return EROFS;
750 	}
751 
752 	/*
753 	 * Get a new vnode *before* adjusting the dirop count, to
754 	 * avoid a deadlock in getnewvnode(), if we have a stacked
755 	 * filesystem mounted on top of us.
756 	 *
757 	 * NB: this means we have to destroy the new vnode on error.
758 	 */
759 
760 	error = getnewvnode(VT_LFS, dvp->v_mount, lfs_vnodeop_p, NULL, vpp);
761 	if (error) {
762 		DLOG((DLOG_ALLOC, "lfs_mknod: dvp %p error %d\n", dvp, error));
763 		return error;
764 	}
765 	KASSERT(*vpp != NULL);
766 
767 	error = lfs_set_dirop(dvp, NULL);
768 	if (error) {
769 		ungetnewvnode(*vpp);
770 		*vpp = NULL;
771 		return error;
772 	}
773 
774 	fstrans_start(ap->a_dvp->v_mount, FSTRANS_SHARED);
775 	error = ulfs_makeinode(MAKEIMODE(vap->va_type, vap->va_mode),
776 			      dvp, ulr, vpp, ap->a_cnp);
777 
778 	/* Either way we're done with the dirop at this point */
779 	UNMARK_VNODE(dvp);
780 	UNMARK_VNODE(*vpp);
781 	if (!((*vpp)->v_uflag & VU_DIROP)) {
782 		KASSERT(error != 0);
783 		ungetnewvnode(*vpp);
784 		*vpp = NULL;
785 	}
786 	else {
787 		KASSERT(error == 0);
788 	}
789 	lfs_unset_dirop(fs, dvp, "mknod");
790 	/*
791 	 * XXX this is where this used to be (though inside some evil
792 	 * macros) but it clearly should be moved further down.
793 	 * - dholland 20140515
794 	 */
795 	vrele(dvp);
796 
797 	if (error) {
798 		fstrans_done(ap->a_dvp->v_mount);
799 		*vpp = NULL;
800 		return (error);
801 	}
802 
803 	VN_KNOTE(dvp, NOTE_WRITE);
804 	ip = VTOI(*vpp);
805 	mp  = (*vpp)->v_mount;
806 	ino = ip->i_number;
807 	ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
808 	if (vap->va_rdev != VNOVAL) {
809 		struct ulfsmount *ump = ip->i_ump;
810 		KASSERT(fs == ip->i_lfs);
811 		/*
812 		 * Want to be able to use this to make badblock
813 		 * inodes, so don't truncate the dev number.
814 		 */
815 		if (ump->um_fstype == ULFS1)
816 			ip->i_ffs1_rdev = ulfs_rw32(vap->va_rdev,
817 			    ULFS_MPNEEDSWAP(fs));
818 		else
819 			ip->i_ffs2_rdev = ulfs_rw64(vap->va_rdev,
820 			    ULFS_MPNEEDSWAP(fs));
821 	}
822 
823 	/*
824 	 * Call fsync to write the vnode so that we don't have to deal with
825 	 * flushing it when it's marked VU_DIROP or reclaiming.
826 	 *
827 	 * XXX KS - If we can't flush we also can't call vgone(), so must
828 	 * return.  But, that leaves this vnode in limbo, also not good.
829 	 * Can this ever happen (barring hardware failure)?
830 	 */
831 	if ((error = VOP_FSYNC(*vpp, NOCRED, FSYNC_WAIT, 0, 0)) != 0) {
832 		panic("lfs_mknod: couldn't fsync (ino %llu)",
833 		      (unsigned long long)ino);
834 		/* return (error); */
835 	}
836 	/*
837 	 * Remove vnode so that it will be reloaded by VFS_VGET and
838 	 * checked to see if it is an alias of an existing entry in
839 	 * the inode cache.
840 	 */
841 	/* Used to be vput, but that causes us to call VOP_INACTIVE twice. */
842 
843 	(*vpp)->v_type = VNON;
844 	VOP_UNLOCK(*vpp);
845 	vgone(*vpp);
846 	error = VFS_VGET(mp, ino, vpp);
847 
848 	fstrans_done(ap->a_dvp->v_mount);
849 	if (error != 0) {
850 		*vpp = NULL;
851 		return (error);
852 	}
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(vpp != NULL);
880 	KASSERT(*vpp == NULL);
881 
882 	/* XXX should handle this material another way */
883 	ulr = &VTOI(dvp)->i_crap;
884 	ULFS_CHECK_CRAPCOUNTER(VTOI(dvp));
885 
886 	fs = VFSTOULFS(dvp->v_mount)->um_lfs;
887 	ASSERT_NO_SEGLOCK(fs);
888 	if (fs->lfs_ronly) {
889 		return EROFS;
890 	}
891 
892 	/*
893 	 * Get a new vnode *before* adjusting the dirop count, to
894 	 * avoid a deadlock in getnewvnode(), if we have a stacked
895 	 * filesystem mounted on top of us.
896 	 *
897 	 * NB: this means we have to destroy the new vnode on error.
898 	 */
899 
900 	error = getnewvnode(VT_LFS, dvp->v_mount, lfs_vnodeop_p, NULL, vpp);
901 	if (error) {
902 		DLOG((DLOG_ALLOC, "lfs_create: dvp %p error %d\n", dvp,error));
903 		return error;
904 	}
905 	error = lfs_set_dirop(dvp, NULL);
906 	if (error) {
907 		ungetnewvnode(*vpp);
908 		*vpp = NULL;
909 		return error;
910 	}
911 
912 	fstrans_start(dvp->v_mount, FSTRANS_SHARED);
913 	error = ulfs_makeinode(MAKEIMODE(vap->va_type, vap->va_mode),
914 			  dvp, ulr, vpp, ap->a_cnp);
915 	if (error) {
916 		fstrans_done(dvp->v_mount);
917 		goto out;
918 	}
919 	fstrans_done(dvp->v_mount);
920 	VN_KNOTE(dvp, NOTE_WRITE);
921 	VOP_UNLOCK(*vpp);
922 
923 out:
924 
925 	UNMARK_VNODE(dvp);
926 	UNMARK_VNODE(*vpp);
927 	if (!((*vpp)->v_uflag & VU_DIROP)) {
928 		KASSERT(error != 0);
929 		ungetnewvnode(*vpp);
930 		*vpp = NULL;
931 	}
932 	else {
933 		KASSERT(error == 0);
934 	}
935 	lfs_unset_dirop(fs, dvp, "create");
936 
937 	vrele(dvp);
938 	return (error);
939 }
940 
941 int
942 lfs_mkdir(void *v)
943 {
944 	struct vop_mkdir_v3_args	/* {
945 		struct vnode *a_dvp;
946 		struct vnode **a_vpp;
947 		struct componentname *a_cnp;
948 		struct vattr *a_vap;
949 	} */ *ap = v;
950 	struct lfs *fs;
951 	struct vnode *dvp, *tvp, **vpp;
952 	struct inode *dp, *ip;
953 	struct componentname *cnp;
954 	struct vattr *vap;
955 	struct ulfs_lookup_results *ulr;
956 	struct buf *bp;
957 	struct lfs_dirtemplate dirtemplate;
958 	struct lfs_direct *newdir;
959 	int dirblksiz;
960 	int dmode;
961 	int error;
962 
963 	dvp = ap->a_dvp;
964 	tvp = NULL;
965 	vpp = ap->a_vpp;
966 	cnp = ap->a_cnp;
967 	vap = ap->a_vap;
968 
969 	dp = VTOI(dvp);
970 	ip = NULL;
971 
972 	KASSERT(vpp != NULL);
973 	KASSERT(*vpp == NULL);
974 
975 	/* XXX should handle this material another way */
976 	ulr = &dp->i_crap;
977 	ULFS_CHECK_CRAPCOUNTER(dp);
978 
979 	fs = VFSTOULFS(dvp->v_mount)->um_lfs;
980 	ASSERT_NO_SEGLOCK(fs);
981 	if (fs->lfs_ronly) {
982 		return EROFS;
983 	}
984 	dirblksiz = fs->um_dirblksiz;
985 
986 	/*
987 	 * Get a new vnode *before* adjusting the dirop count, to
988 	 * avoid a deadlock in getnewvnode(), if we have a stacked
989 	 * filesystem mounted on top of us.
990 	 *
991 	 * NB: this means we have to destroy the new vnode on error.
992 	 */
993 
994 	error = getnewvnode(VT_LFS, dvp->v_mount, lfs_vnodeop_p, NULL, vpp);
995 	if (error) {
996 		DLOG((DLOG_ALLOC, "lfs_mkdir: dvp %p error %d\n", dvp, error));
997 		return error;
998 	}
999 	error = lfs_set_dirop(dvp, NULL);
1000 	if (error) {
1001 		ungetnewvnode(*vpp);
1002 		*vpp = NULL;
1003 		return error;
1004 	}
1005 
1006 	fstrans_start(dvp->v_mount, FSTRANS_SHARED);
1007 
1008 	if ((nlink_t)dp->i_nlink >= LINK_MAX) {
1009 		error = EMLINK;
1010 		goto out;
1011 	}
1012 
1013 	dmode = vap->va_mode & ACCESSPERMS;
1014 	dmode |= LFS_IFDIR;
1015 	/*
1016 	 * Must simulate part of ulfs_makeinode here to acquire the inode,
1017 	 * but not have it entered in the parent directory. The entry is
1018 	 * made later after writing "." and ".." entries.
1019 	 */
1020 	if ((error = lfs_valloc(dvp, dmode, cnp->cn_cred, vpp)) != 0)
1021 		goto out;
1022 
1023 	tvp = *vpp;
1024 	ip = VTOI(tvp);
1025 
1026 	ip->i_uid = kauth_cred_geteuid(cnp->cn_cred);
1027 	DIP_ASSIGN(ip, uid, ip->i_uid);
1028 	ip->i_gid = dp->i_gid;
1029 	DIP_ASSIGN(ip, gid, ip->i_gid);
1030 #if defined(LFS_QUOTA) || defined(LFS_QUOTA2)
1031 	if ((error = lfs_chkiq(ip, 1, cnp->cn_cred, 0))) {
1032 		lfs_vfree(tvp, ip->i_number, dmode);
1033 		fstrans_done(dvp->v_mount);
1034 		vput(tvp);
1035 		goto out2;
1036 	}
1037 #endif
1038 	ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
1039 	ip->i_mode = dmode;
1040 	DIP_ASSIGN(ip, mode, dmode);
1041 	tvp->v_type = VDIR;	/* Rest init'd in getnewvnode(). */
1042 	ip->i_nlink = 2;
1043 	DIP_ASSIGN(ip, nlink, 2);
1044 	if (cnp->cn_flags & ISWHITEOUT) {
1045 		ip->i_flags |= UF_OPAQUE;
1046 		DIP_ASSIGN(ip, flags, ip->i_flags);
1047 	}
1048 
1049 	/*
1050 	 * Bump link count in parent directory to reflect work done below.
1051 	 */
1052 	dp->i_nlink++;
1053 	DIP_ASSIGN(dp, nlink, dp->i_nlink);
1054 	dp->i_flag |= IN_CHANGE;
1055 	if ((error = lfs_update(dvp, NULL, NULL, UPDATE_DIROP)) != 0)
1056 		goto bad;
1057 
1058 	/*
1059 	 * Initialize directory with "." and ".." from static template.
1060 	 */
1061 	dirtemplate = mastertemplate;
1062 	dirtemplate.dotdot_reclen = dirblksiz - dirtemplate.dot_reclen;
1063 	dirtemplate.dot_ino = ulfs_rw32(ip->i_number, ULFS_MPNEEDSWAP(fs));
1064 	dirtemplate.dotdot_ino = ulfs_rw32(dp->i_number, ULFS_MPNEEDSWAP(fs));
1065 	dirtemplate.dot_reclen = ulfs_rw16(dirtemplate.dot_reclen,
1066 	    ULFS_MPNEEDSWAP(fs));
1067 	dirtemplate.dotdot_reclen = ulfs_rw16(dirtemplate.dotdot_reclen,
1068 	    ULFS_MPNEEDSWAP(fs));
1069 	if (fs->um_maxsymlinklen <= 0) {
1070 #if BYTE_ORDER == LITTLE_ENDIAN
1071 		if (ULFS_MPNEEDSWAP(fs) == 0)
1072 #else
1073 		if (ULFS_MPNEEDSWAP(fs) != 0)
1074 #endif
1075 		{
1076 			dirtemplate.dot_type = dirtemplate.dot_namlen;
1077 			dirtemplate.dotdot_type = dirtemplate.dotdot_namlen;
1078 			dirtemplate.dot_namlen = dirtemplate.dotdot_namlen = 0;
1079 		} else
1080 			dirtemplate.dot_type = dirtemplate.dotdot_type = 0;
1081 	}
1082 	if ((error = lfs_balloc(tvp, (off_t)0, dirblksiz, cnp->cn_cred,
1083 	    B_CLRBUF, &bp)) != 0)
1084 		goto bad;
1085 	ip->i_size = dirblksiz;
1086 	DIP_ASSIGN(ip, size, dirblksiz);
1087 	ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
1088 	uvm_vnp_setsize(tvp, ip->i_size);
1089 	memcpy((void *)bp->b_data, (void *)&dirtemplate, sizeof dirtemplate);
1090 
1091 	/*
1092 	 * Directory set up; now install its entry in the parent directory.
1093 	 */
1094 	if ((error = VOP_BWRITE(bp->b_vp, bp)) != 0)
1095 		goto bad;
1096 	if ((error = lfs_update(tvp, NULL, NULL, UPDATE_DIROP)) != 0) {
1097 		goto bad;
1098 	}
1099 	newdir = pool_cache_get(ulfs_direct_cache, PR_WAITOK);
1100 	ulfs_makedirentry(ip, cnp, newdir);
1101 	error = ulfs_direnter(dvp, ulr, tvp, newdir, cnp, bp);
1102 	pool_cache_put(ulfs_direct_cache, newdir);
1103  bad:
1104 	if (error == 0) {
1105 		VN_KNOTE(dvp, NOTE_WRITE | NOTE_LINK);
1106 		VOP_UNLOCK(tvp);
1107 	} else {
1108 		dp->i_nlink--;
1109 		DIP_ASSIGN(dp, nlink, dp->i_nlink);
1110 		dp->i_flag |= IN_CHANGE;
1111 		/*
1112 		 * No need to do an explicit lfs_truncate here, vrele will
1113 		 * do this for us because we set the link count to 0.
1114 		 */
1115 		ip->i_nlink = 0;
1116 		DIP_ASSIGN(ip, nlink, 0);
1117 		ip->i_flag |= IN_CHANGE;
1118 		/* If IN_ADIROP, account for it */
1119 		lfs_unmark_vnode(tvp);
1120 		vput(tvp);
1121 	}
1122 
1123 out:
1124 	fstrans_done(dvp->v_mount);
1125 #if defined(LFS_QUOTA) || defined(LFS_QUOTA2)
1126 out2:
1127 #endif
1128 
1129 	UNMARK_VNODE(dvp);
1130 	UNMARK_VNODE(*vpp);
1131 	if (!((*vpp)->v_uflag & VU_DIROP)) {
1132 		KASSERT(error != 0);
1133 		ungetnewvnode(*vpp);
1134 		*vpp = NULL;
1135 	}
1136 	else {
1137 		KASSERT(error == 0);
1138 	}
1139 	lfs_unset_dirop(fs, dvp, "mkdir");
1140 
1141 	vrele(dvp);
1142 	return (error);
1143 }
1144 
1145 int
1146 lfs_remove(void *v)
1147 {
1148 	struct vop_remove_args	/* {
1149 		struct vnode *a_dvp;
1150 		struct vnode *a_vp;
1151 		struct componentname *a_cnp;
1152 	} */ *ap = v;
1153 	struct vnode *dvp, *vp;
1154 	struct inode *ip;
1155 	int error;
1156 
1157 	dvp = ap->a_dvp;
1158 	vp = ap->a_vp;
1159 	ip = VTOI(vp);
1160 	if ((error = lfs_set_dirop(dvp, vp)) != 0) {
1161 		if (dvp == vp)
1162 			vrele(vp);
1163 		else
1164 			vput(vp);
1165 		vput(dvp);
1166 		return error;
1167 	}
1168 	error = ulfs_remove(ap);
1169 	if (ip->i_nlink == 0)
1170 		lfs_orphan(ip->i_lfs, ip->i_number);
1171 
1172 	UNMARK_VNODE(dvp);
1173 	if (ap->a_vp) {
1174 		UNMARK_VNODE(ap->a_vp);
1175 	}
1176 	lfs_unset_dirop(ip->i_lfs, dvp, "remove");
1177 	vrele(dvp);
1178 	if (ap->a_vp) {
1179 		vrele(ap->a_vp);
1180 	}
1181 
1182 	return (error);
1183 }
1184 
1185 int
1186 lfs_rmdir(void *v)
1187 {
1188 	struct vop_rmdir_args	/* {
1189 		struct vnodeop_desc *a_desc;
1190 		struct vnode *a_dvp;
1191 		struct vnode *a_vp;
1192 		struct componentname *a_cnp;
1193 	} */ *ap = v;
1194 	struct vnode *vp;
1195 	struct inode *ip;
1196 	int error;
1197 
1198 	vp = ap->a_vp;
1199 	ip = VTOI(vp);
1200 	if ((error = lfs_set_dirop(ap->a_dvp, ap->a_vp)) != 0) {
1201 		if (ap->a_dvp == vp)
1202 			vrele(ap->a_dvp);
1203 		else
1204 			vput(ap->a_dvp);
1205 		vput(vp);
1206 		return error;
1207 	}
1208 	error = ulfs_rmdir(ap);
1209 	if (ip->i_nlink == 0)
1210 		lfs_orphan(ip->i_lfs, ip->i_number);
1211 
1212 	UNMARK_VNODE(ap->a_dvp);
1213 	if (ap->a_vp) {
1214 		UNMARK_VNODE(ap->a_vp);
1215 	}
1216 	lfs_unset_dirop(ip->i_lfs, ap->a_dvp, "rmdir");
1217 	vrele(ap->a_dvp);
1218 	if (ap->a_vp) {
1219 		vrele(ap->a_vp);
1220 	}
1221 
1222 	return (error);
1223 }
1224 
1225 int
1226 lfs_link(void *v)
1227 {
1228 	struct vop_link_args	/* {
1229 		struct vnode *a_dvp;
1230 		struct vnode *a_vp;
1231 		struct componentname *a_cnp;
1232 	} */ *ap = v;
1233 	struct lfs *fs;
1234 	struct vnode *dvp;
1235 	int error;
1236 
1237 	dvp = ap->a_dvp;
1238 
1239 	fs = VFSTOULFS(dvp->v_mount)->um_lfs;
1240 	ASSERT_NO_SEGLOCK(fs);
1241 	if (fs->lfs_ronly) {
1242 		return EROFS;
1243 	}
1244 
1245 	error = lfs_set_dirop(dvp, NULL);
1246 	if (error) {
1247 		/*
1248 		 * XXX dholland 20140515 this was here before but must
1249 		 * be wrong.
1250 		 */
1251 		vput(dvp);
1252 
1253 		return error;
1254 	}
1255 
1256 	error = ulfs_link(ap);
1257 
1258 	UNMARK_VNODE(dvp);
1259 	lfs_unset_dirop(fs, dvp, "link");
1260 	vrele(dvp);
1261 
1262 	return (error);
1263 }
1264 
1265 /* XXX hack to avoid calling ITIMES in getattr */
1266 int
1267 lfs_getattr(void *v)
1268 {
1269 	struct vop_getattr_args /* {
1270 		struct vnode *a_vp;
1271 		struct vattr *a_vap;
1272 		kauth_cred_t a_cred;
1273 	} */ *ap = v;
1274 	struct vnode *vp = ap->a_vp;
1275 	struct inode *ip = VTOI(vp);
1276 	struct vattr *vap = ap->a_vap;
1277 	struct lfs *fs = ip->i_lfs;
1278 
1279 	fstrans_start(vp->v_mount, FSTRANS_SHARED);
1280 	/*
1281 	 * Copy from inode table
1282 	 */
1283 	vap->va_fsid = ip->i_dev;
1284 	vap->va_fileid = ip->i_number;
1285 	vap->va_mode = ip->i_mode & ~LFS_IFMT;
1286 	vap->va_nlink = ip->i_nlink;
1287 	vap->va_uid = ip->i_uid;
1288 	vap->va_gid = ip->i_gid;
1289 	vap->va_rdev = (dev_t)ip->i_ffs1_rdev;
1290 	vap->va_size = vp->v_size;
1291 	vap->va_atime.tv_sec = ip->i_ffs1_atime;
1292 	vap->va_atime.tv_nsec = ip->i_ffs1_atimensec;
1293 	vap->va_mtime.tv_sec = ip->i_ffs1_mtime;
1294 	vap->va_mtime.tv_nsec = ip->i_ffs1_mtimensec;
1295 	vap->va_ctime.tv_sec = ip->i_ffs1_ctime;
1296 	vap->va_ctime.tv_nsec = ip->i_ffs1_ctimensec;
1297 	vap->va_flags = ip->i_flags;
1298 	vap->va_gen = ip->i_gen;
1299 	/* this doesn't belong here */
1300 	if (vp->v_type == VBLK)
1301 		vap->va_blocksize = BLKDEV_IOSIZE;
1302 	else if (vp->v_type == VCHR)
1303 		vap->va_blocksize = MAXBSIZE;
1304 	else
1305 		vap->va_blocksize = vp->v_mount->mnt_stat.f_iosize;
1306 	vap->va_bytes = lfs_fsbtob(fs, (u_quad_t)ip->i_lfs_effnblks);
1307 	vap->va_type = vp->v_type;
1308 	vap->va_filerev = ip->i_modrev;
1309 	fstrans_done(vp->v_mount);
1310 	return (0);
1311 }
1312 
1313 /*
1314  * Check to make sure the inode blocks won't choke the buffer
1315  * cache, then call ulfs_setattr as usual.
1316  */
1317 int
1318 lfs_setattr(void *v)
1319 {
1320 	struct vop_setattr_args /* {
1321 		struct vnode *a_vp;
1322 		struct vattr *a_vap;
1323 		kauth_cred_t a_cred;
1324 	} */ *ap = v;
1325 	struct vnode *vp = ap->a_vp;
1326 
1327 	lfs_check(vp, LFS_UNUSED_LBN, 0);
1328 	return ulfs_setattr(v);
1329 }
1330 
1331 /*
1332  * Release the block we hold on lfs_newseg wrapping.  Called on file close,
1333  * or explicitly from LFCNWRAPGO.  Called with the interlock held.
1334  */
1335 static int
1336 lfs_wrapgo(struct lfs *fs, struct inode *ip, int waitfor)
1337 {
1338 	if (fs->lfs_stoplwp != curlwp)
1339 		return EBUSY;
1340 
1341 	fs->lfs_stoplwp = NULL;
1342 	cv_signal(&fs->lfs_stopcv);
1343 
1344 	KASSERT(fs->lfs_nowrap > 0);
1345 	if (fs->lfs_nowrap <= 0) {
1346 		return 0;
1347 	}
1348 
1349 	if (--fs->lfs_nowrap == 0) {
1350 		log(LOG_NOTICE, "%s: re-enabled log wrap\n", fs->lfs_fsmnt);
1351 		wakeup(&fs->lfs_wrappass);
1352 		lfs_wakeup_cleaner(fs);
1353 	}
1354 	if (waitfor) {
1355 		mtsleep(&fs->lfs_nextseg, PCATCH | PUSER, "segment",
1356 		    0, &lfs_lock);
1357 	}
1358 
1359 	return 0;
1360 }
1361 
1362 /*
1363  * Close called.
1364  *
1365  * Update the times on the inode.
1366  */
1367 /* ARGSUSED */
1368 int
1369 lfs_close(void *v)
1370 {
1371 	struct vop_close_args /* {
1372 		struct vnode *a_vp;
1373 		int  a_fflag;
1374 		kauth_cred_t a_cred;
1375 	} */ *ap = v;
1376 	struct vnode *vp = ap->a_vp;
1377 	struct inode *ip = VTOI(vp);
1378 	struct lfs *fs = ip->i_lfs;
1379 
1380 	if ((ip->i_number == ULFS_ROOTINO || ip->i_number == LFS_IFILE_INUM) &&
1381 	    fs->lfs_stoplwp == curlwp) {
1382 		mutex_enter(&lfs_lock);
1383 		log(LOG_NOTICE, "lfs_close: releasing log wrap control\n");
1384 		lfs_wrapgo(fs, ip, 0);
1385 		mutex_exit(&lfs_lock);
1386 	}
1387 
1388 	if (vp == ip->i_lfs->lfs_ivnode &&
1389 	    vp->v_mount->mnt_iflag & IMNT_UNMOUNT)
1390 		return 0;
1391 
1392 	fstrans_start(vp->v_mount, FSTRANS_SHARED);
1393 	if (vp->v_usecount > 1 && vp != ip->i_lfs->lfs_ivnode) {
1394 		LFS_ITIMES(ip, NULL, NULL, NULL);
1395 	}
1396 	fstrans_done(vp->v_mount);
1397 	return (0);
1398 }
1399 
1400 /*
1401  * Close wrapper for special devices.
1402  *
1403  * Update the times on the inode then do device close.
1404  */
1405 int
1406 lfsspec_close(void *v)
1407 {
1408 	struct vop_close_args /* {
1409 		struct vnode	*a_vp;
1410 		int		a_fflag;
1411 		kauth_cred_t	a_cred;
1412 	} */ *ap = v;
1413 	struct vnode	*vp;
1414 	struct inode	*ip;
1415 
1416 	vp = ap->a_vp;
1417 	ip = VTOI(vp);
1418 	if (vp->v_usecount > 1) {
1419 		LFS_ITIMES(ip, NULL, NULL, NULL);
1420 	}
1421 	return (VOCALL (spec_vnodeop_p, VOFFSET(vop_close), ap));
1422 }
1423 
1424 /*
1425  * Close wrapper for fifo's.
1426  *
1427  * Update the times on the inode then do device close.
1428  */
1429 int
1430 lfsfifo_close(void *v)
1431 {
1432 	struct vop_close_args /* {
1433 		struct vnode	*a_vp;
1434 		int		a_fflag;
1435 		kauth_cred_	a_cred;
1436 	} */ *ap = v;
1437 	struct vnode	*vp;
1438 	struct inode	*ip;
1439 
1440 	vp = ap->a_vp;
1441 	ip = VTOI(vp);
1442 	if (ap->a_vp->v_usecount > 1) {
1443 		LFS_ITIMES(ip, NULL, NULL, NULL);
1444 	}
1445 	return (VOCALL (fifo_vnodeop_p, VOFFSET(vop_close), ap));
1446 }
1447 
1448 /*
1449  * Reclaim an inode so that it can be used for other purposes.
1450  */
1451 
1452 int
1453 lfs_reclaim(void *v)
1454 {
1455 	struct vop_reclaim_args /* {
1456 		struct vnode *a_vp;
1457 	} */ *ap = v;
1458 	struct vnode *vp = ap->a_vp;
1459 	struct inode *ip = VTOI(vp);
1460 	struct lfs *fs = ip->i_lfs;
1461 	int error;
1462 
1463 	/*
1464 	 * The inode must be freed and updated before being removed
1465 	 * from its hash chain.  Other threads trying to gain a hold
1466 	 * or lock on the inode will be stalled.
1467 	 */
1468 	if (ip->i_nlink <= 0 && (vp->v_mount->mnt_flag & MNT_RDONLY) == 0)
1469 		lfs_vfree(vp, ip->i_number, ip->i_omode);
1470 
1471 	mutex_enter(&lfs_lock);
1472 	LFS_CLR_UINO(ip, IN_ALLMOD);
1473 	mutex_exit(&lfs_lock);
1474 	if ((error = ulfs_reclaim(vp)))
1475 		return (error);
1476 
1477 	/*
1478 	 * Take us off the paging and/or dirop queues if we were on them.
1479 	 * We shouldn't be on them.
1480 	 */
1481 	mutex_enter(&lfs_lock);
1482 	if (ip->i_flags & IN_PAGING) {
1483 		log(LOG_WARNING, "%s: reclaimed vnode is IN_PAGING\n",
1484 		    fs->lfs_fsmnt);
1485 		ip->i_flags &= ~IN_PAGING;
1486 		TAILQ_REMOVE(&fs->lfs_pchainhd, ip, i_lfs_pchain);
1487 	}
1488 	if (vp->v_uflag & VU_DIROP) {
1489 		panic("reclaimed vnode is VU_DIROP");
1490 		vp->v_uflag &= ~VU_DIROP;
1491 		TAILQ_REMOVE(&fs->lfs_dchainhd, ip, i_lfs_dchain);
1492 	}
1493 	mutex_exit(&lfs_lock);
1494 
1495 	pool_put(&lfs_dinode_pool, ip->i_din.ffs1_din);
1496 	lfs_deregister_all(vp);
1497 	pool_put(&lfs_inoext_pool, ip->inode_ext.lfs);
1498 	ip->inode_ext.lfs = NULL;
1499 	genfs_node_destroy(vp);
1500 	pool_put(&lfs_inode_pool, vp->v_data);
1501 	vp->v_data = NULL;
1502 	return (0);
1503 }
1504 
1505 /*
1506  * Read a block from a storage device.
1507  *
1508  * Calculate the logical to physical mapping if not done already,
1509  * then call the device strategy routine.
1510  *
1511  * In order to avoid reading blocks that are in the process of being
1512  * written by the cleaner---and hence are not mutexed by the normal
1513  * buffer cache / page cache mechanisms---check for collisions before
1514  * reading.
1515  *
1516  * We inline ulfs_strategy to make sure that the VOP_BMAP occurs *before*
1517  * the active cleaner test.
1518  *
1519  * XXX This code assumes that lfs_markv makes synchronous checkpoints.
1520  */
1521 int
1522 lfs_strategy(void *v)
1523 {
1524 	struct vop_strategy_args /* {
1525 		struct vnode *a_vp;
1526 		struct buf *a_bp;
1527 	} */ *ap = v;
1528 	struct buf	*bp;
1529 	struct lfs	*fs;
1530 	struct vnode	*vp;
1531 	struct inode	*ip;
1532 	daddr_t		tbn;
1533 #define MAXLOOP 25
1534 	int		i, sn, error, slept, loopcount;
1535 
1536 	bp = ap->a_bp;
1537 	vp = ap->a_vp;
1538 	ip = VTOI(vp);
1539 	fs = ip->i_lfs;
1540 
1541 	/* lfs uses its strategy routine only for read */
1542 	KASSERT(bp->b_flags & B_READ);
1543 
1544 	if (vp->v_type == VBLK || vp->v_type == VCHR)
1545 		panic("lfs_strategy: spec");
1546 	KASSERT(bp->b_bcount != 0);
1547 	if (bp->b_blkno == bp->b_lblkno) {
1548 		error = VOP_BMAP(vp, bp->b_lblkno, NULL, &bp->b_blkno,
1549 				 NULL);
1550 		if (error) {
1551 			bp->b_error = error;
1552 			bp->b_resid = bp->b_bcount;
1553 			biodone(bp);
1554 			return (error);
1555 		}
1556 		if ((long)bp->b_blkno == -1) /* no valid data */
1557 			clrbuf(bp);
1558 	}
1559 	if ((long)bp->b_blkno < 0) { /* block is not on disk */
1560 		bp->b_resid = bp->b_bcount;
1561 		biodone(bp);
1562 		return (0);
1563 	}
1564 
1565 	slept = 1;
1566 	loopcount = 0;
1567 	mutex_enter(&lfs_lock);
1568 	while (slept && fs->lfs_seglock) {
1569 		mutex_exit(&lfs_lock);
1570 		/*
1571 		 * Look through list of intervals.
1572 		 * There will only be intervals to look through
1573 		 * if the cleaner holds the seglock.
1574 		 * Since the cleaner is synchronous, we can trust
1575 		 * the list of intervals to be current.
1576 		 */
1577 		tbn = LFS_DBTOFSB(fs, bp->b_blkno);
1578 		sn = lfs_dtosn(fs, tbn);
1579 		slept = 0;
1580 		for (i = 0; i < fs->lfs_cleanind; i++) {
1581 			if (sn == lfs_dtosn(fs, fs->lfs_cleanint[i]) &&
1582 			    tbn >= fs->lfs_cleanint[i]) {
1583 				DLOG((DLOG_CLEAN,
1584 				      "lfs_strategy: ino %d lbn %" PRId64
1585 				      " ind %d sn %d fsb %" PRIx32
1586 				      " given sn %d fsb %" PRIx64 "\n",
1587 				      ip->i_number, bp->b_lblkno, i,
1588 				      lfs_dtosn(fs, fs->lfs_cleanint[i]),
1589 				      fs->lfs_cleanint[i], sn, tbn));
1590 				DLOG((DLOG_CLEAN,
1591 				      "lfs_strategy: sleeping on ino %d lbn %"
1592 				      PRId64 "\n", ip->i_number, bp->b_lblkno));
1593 				mutex_enter(&lfs_lock);
1594 				if (LFS_SEGLOCK_HELD(fs) && fs->lfs_iocount) {
1595 					/*
1596 					 * Cleaner can't wait for itself.
1597 					 * Instead, wait for the blocks
1598 					 * to be written to disk.
1599 					 * XXX we need pribio in the test
1600 					 * XXX here.
1601 					 */
1602  					mtsleep(&fs->lfs_iocount,
1603  						(PRIBIO + 1) | PNORELOCK,
1604 						"clean2", hz/10 + 1,
1605  						&lfs_lock);
1606 					slept = 1;
1607 					++loopcount;
1608 					break;
1609 				} else if (fs->lfs_seglock) {
1610 					mtsleep(&fs->lfs_seglock,
1611 						(PRIBIO + 1) | PNORELOCK,
1612 						"clean1", 0,
1613 						&lfs_lock);
1614 					slept = 1;
1615 					break;
1616 				}
1617 				mutex_exit(&lfs_lock);
1618 			}
1619 		}
1620 		mutex_enter(&lfs_lock);
1621 		if (loopcount > MAXLOOP) {
1622 			printf("lfs_strategy: breaking out of clean2 loop\n");
1623 			break;
1624 		}
1625 	}
1626 	mutex_exit(&lfs_lock);
1627 
1628 	vp = ip->i_devvp;
1629 	return VOP_STRATEGY(vp, bp);
1630 }
1631 
1632 /*
1633  * Inline lfs_segwrite/lfs_writevnodes, but just for dirops.
1634  * Technically this is a checkpoint (the on-disk state is valid)
1635  * even though we are leaving out all the file data.
1636  */
1637 int
1638 lfs_flush_dirops(struct lfs *fs)
1639 {
1640 	struct inode *ip, *nip;
1641 	struct vnode *vp;
1642 	extern int lfs_dostats;
1643 	struct segment *sp;
1644 	int flags = 0;
1645 	int error = 0;
1646 
1647 	ASSERT_MAYBE_SEGLOCK(fs);
1648 	KASSERT(fs->lfs_nadirop == 0);
1649 
1650 	if (fs->lfs_ronly)
1651 		return EROFS;
1652 
1653 	mutex_enter(&lfs_lock);
1654 	if (TAILQ_FIRST(&fs->lfs_dchainhd) == NULL) {
1655 		mutex_exit(&lfs_lock);
1656 		return 0;
1657 	} else
1658 		mutex_exit(&lfs_lock);
1659 
1660 	if (lfs_dostats)
1661 		++lfs_stats.flush_invoked;
1662 
1663 	lfs_imtime(fs);
1664 	lfs_seglock(fs, flags);
1665 	sp = fs->lfs_sp;
1666 
1667 	/*
1668 	 * lfs_writevnodes, optimized to get dirops out of the way.
1669 	 * Only write dirops, and don't flush files' pages, only
1670 	 * blocks from the directories.
1671 	 *
1672 	 * We don't need to vref these files because they are
1673 	 * dirops and so hold an extra reference until the
1674 	 * segunlock clears them of that status.
1675 	 *
1676 	 * We don't need to check for IN_ADIROP because we know that
1677 	 * no dirops are active.
1678 	 *
1679 	 */
1680 	mutex_enter(&lfs_lock);
1681 	for (ip = TAILQ_FIRST(&fs->lfs_dchainhd); ip != NULL; ip = nip) {
1682 		nip = TAILQ_NEXT(ip, i_lfs_dchain);
1683 		mutex_exit(&lfs_lock);
1684 		vp = ITOV(ip);
1685 		mutex_enter(vp->v_interlock);
1686 
1687 		KASSERT((ip->i_flag & IN_ADIROP) == 0);
1688 		KASSERT(vp->v_uflag & VU_DIROP);
1689 		KASSERT(vdead_check(vp, VDEAD_NOWAIT) == 0);
1690 
1691 		/*
1692 		 * All writes to directories come from dirops; all
1693 		 * writes to files' direct blocks go through the page
1694 		 * cache, which we're not touching.  Reads to files
1695 		 * and/or directories will not be affected by writing
1696 		 * directory blocks inodes and file inodes.  So we don't
1697 		 * really need to lock.
1698 		 */
1699 		if (vdead_check(vp, VDEAD_NOWAIT) != 0) {
1700 			mutex_exit(vp->v_interlock);
1701 			mutex_enter(&lfs_lock);
1702 			continue;
1703 		}
1704 		mutex_exit(vp->v_interlock);
1705 		/* XXX see below
1706 		 * waslocked = VOP_ISLOCKED(vp);
1707 		 */
1708 		if (vp->v_type != VREG &&
1709 		    ((ip->i_flag & IN_ALLMOD) || !VPISEMPTY(vp))) {
1710 			error = lfs_writefile(fs, sp, vp);
1711 			if (!VPISEMPTY(vp) && !WRITEINPROG(vp) &&
1712 			    !(ip->i_flag & IN_ALLMOD)) {
1713 			    	mutex_enter(&lfs_lock);
1714 				LFS_SET_UINO(ip, IN_MODIFIED);
1715 			    	mutex_exit(&lfs_lock);
1716 			}
1717 			if (error && (sp->seg_flags & SEGM_SINGLE)) {
1718 				mutex_enter(&lfs_lock);
1719 				error = EAGAIN;
1720 				break;
1721 			}
1722 		}
1723 		KDASSERT(ip->i_number != LFS_IFILE_INUM);
1724 		error = lfs_writeinode(fs, sp, ip);
1725 		mutex_enter(&lfs_lock);
1726 		if (error && (sp->seg_flags & SEGM_SINGLE)) {
1727 			error = EAGAIN;
1728 			break;
1729 		}
1730 
1731 		/*
1732 		 * We might need to update these inodes again,
1733 		 * for example, if they have data blocks to write.
1734 		 * Make sure that after this flush, they are still
1735 		 * marked IN_MODIFIED so that we don't forget to
1736 		 * write them.
1737 		 */
1738 		/* XXX only for non-directories? --KS */
1739 		LFS_SET_UINO(ip, IN_MODIFIED);
1740 	}
1741 	mutex_exit(&lfs_lock);
1742 	/* We've written all the dirops there are */
1743 	((SEGSUM *)(sp->segsum))->ss_flags &= ~(SS_CONT);
1744 	lfs_finalize_fs_seguse(fs);
1745 	(void) lfs_writeseg(fs, sp);
1746 	lfs_segunlock(fs);
1747 
1748 	return error;
1749 }
1750 
1751 /*
1752  * Flush all vnodes for which the pagedaemon has requested pageouts.
1753  * Skip over any files that are marked VU_DIROP (since lfs_flush_dirop()
1754  * has just run, this would be an error).  If we have to skip a vnode
1755  * for any reason, just skip it; if we have to wait for the cleaner,
1756  * abort.  The writer daemon will call us again later.
1757  */
1758 int
1759 lfs_flush_pchain(struct lfs *fs)
1760 {
1761 	struct inode *ip, *nip;
1762 	struct vnode *vp;
1763 	extern int lfs_dostats;
1764 	struct segment *sp;
1765 	int error, error2;
1766 
1767 	ASSERT_NO_SEGLOCK(fs);
1768 
1769 	if (fs->lfs_ronly)
1770 		return EROFS;
1771 
1772 	mutex_enter(&lfs_lock);
1773 	if (TAILQ_FIRST(&fs->lfs_pchainhd) == NULL) {
1774 		mutex_exit(&lfs_lock);
1775 		return 0;
1776 	} else
1777 		mutex_exit(&lfs_lock);
1778 
1779 	/* Get dirops out of the way */
1780 	if ((error = lfs_flush_dirops(fs)) != 0)
1781 		return error;
1782 
1783 	if (lfs_dostats)
1784 		++lfs_stats.flush_invoked;
1785 
1786 	/*
1787 	 * Inline lfs_segwrite/lfs_writevnodes, but just for pageouts.
1788 	 */
1789 	lfs_imtime(fs);
1790 	lfs_seglock(fs, 0);
1791 	sp = fs->lfs_sp;
1792 
1793 	/*
1794 	 * lfs_writevnodes, optimized to clear pageout requests.
1795 	 * Only write non-dirop files that are in the pageout queue.
1796 	 * We're very conservative about what we write; we want to be
1797 	 * fast and async.
1798 	 */
1799 	mutex_enter(&lfs_lock);
1800     top:
1801 	for (ip = TAILQ_FIRST(&fs->lfs_pchainhd); ip != NULL; ip = nip) {
1802 		nip = TAILQ_NEXT(ip, i_lfs_pchain);
1803 		vp = ITOV(ip);
1804 
1805 		if (!(ip->i_flags & IN_PAGING))
1806 			goto top;
1807 
1808 		mutex_enter(vp->v_interlock);
1809 		if (vdead_check(vp, VDEAD_NOWAIT) != 0 ||
1810 		    (vp->v_uflag & VU_DIROP) != 0) {
1811 			mutex_exit(vp->v_interlock);
1812 			continue;
1813 		}
1814 		if (vp->v_type != VREG) {
1815 			mutex_exit(vp->v_interlock);
1816 			continue;
1817 		}
1818 		if (lfs_vref(vp))
1819 			continue;
1820 		mutex_exit(&lfs_lock);
1821 
1822 		if (vn_lock(vp, LK_EXCLUSIVE | LK_NOWAIT | LK_RETRY) != 0) {
1823 			lfs_vunref(vp);
1824 			mutex_enter(&lfs_lock);
1825 			continue;
1826 		}
1827 
1828 		error = lfs_writefile(fs, sp, vp);
1829 		if (!VPISEMPTY(vp) && !WRITEINPROG(vp) &&
1830 		    !(ip->i_flag & IN_ALLMOD)) {
1831 		    	mutex_enter(&lfs_lock);
1832 			LFS_SET_UINO(ip, IN_MODIFIED);
1833 		    	mutex_exit(&lfs_lock);
1834 		}
1835 		KDASSERT(ip->i_number != LFS_IFILE_INUM);
1836 		error2 = lfs_writeinode(fs, sp, ip);
1837 
1838 		VOP_UNLOCK(vp);
1839 		lfs_vunref(vp);
1840 
1841 		if (error == EAGAIN || error2 == EAGAIN) {
1842 			lfs_writeseg(fs, sp);
1843 			mutex_enter(&lfs_lock);
1844 			break;
1845 		}
1846 		mutex_enter(&lfs_lock);
1847 	}
1848 	mutex_exit(&lfs_lock);
1849 	(void) lfs_writeseg(fs, sp);
1850 	lfs_segunlock(fs);
1851 
1852 	return 0;
1853 }
1854 
1855 /*
1856  * Provide a fcntl interface to sys_lfs_{segwait,bmapv,markv}.
1857  */
1858 int
1859 lfs_fcntl(void *v)
1860 {
1861 	struct vop_fcntl_args /* {
1862 		struct vnode *a_vp;
1863 		u_int a_command;
1864 		void * a_data;
1865 		int  a_fflag;
1866 		kauth_cred_t a_cred;
1867 	} */ *ap = v;
1868 	struct timeval tv;
1869 	struct timeval *tvp;
1870 	BLOCK_INFO *blkiov;
1871 	CLEANERINFO *cip;
1872 	SEGUSE *sup;
1873 	int blkcnt, error;
1874 	size_t fh_size;
1875 	struct lfs_fcntl_markv blkvp;
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 		/* FALLSTHROUGH */
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 		/* FALLSTHROUGH */
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 			wakeup(&fs->lfs_sleepers);
1935 		mutex_exit(&lfs_lock);
1936 		return error;
1937 
1938 	    case LFCNBMAPV:
1939 	    case LFCNMARKV:
1940 		blkvp = *(struct lfs_fcntl_markv *)ap->a_data;
1941 
1942 		blkcnt = blkvp.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 		if ((error = copyin(blkvp.blkiov, blkiov,
1947 		     blkcnt * sizeof(BLOCK_INFO))) != 0) {
1948 			lfs_free(fs, blkiov, LFS_NB_BLKIOV);
1949 			return error;
1950 		}
1951 
1952 		mutex_enter(&lfs_lock);
1953 		++fs->lfs_sleepers;
1954 		mutex_exit(&lfs_lock);
1955 		if (ap->a_command == LFCNBMAPV)
1956 			error = lfs_bmapv(l->l_proc, fsidp, blkiov, blkcnt);
1957 		else /* LFCNMARKV */
1958 			error = lfs_markv(l->l_proc, fsidp, blkiov, blkcnt);
1959 		if (error == 0)
1960 			error = copyout(blkiov, blkvp.blkiov,
1961 					blkcnt * sizeof(BLOCK_INFO));
1962 		mutex_enter(&lfs_lock);
1963 		if (--fs->lfs_sleepers == 0)
1964 			wakeup(&fs->lfs_sleepers);
1965 		mutex_exit(&lfs_lock);
1966 		lfs_free(fs, blkiov, LFS_NB_BLKIOV);
1967 		return error;
1968 
1969 	    case LFCNRECLAIM:
1970 		/*
1971 		 * Flush dirops and write Ifile, allowing empty segments
1972 		 * to be immediately reclaimed.
1973 		 */
1974 		lfs_writer_enter(fs, "pndirop");
1975 		off = fs->lfs_offset;
1976 		lfs_seglock(fs, SEGM_FORCE_CKP | SEGM_CKP);
1977 		lfs_flush_dirops(fs);
1978 		LFS_CLEANERINFO(cip, fs, bp);
1979 		oclean = cip->clean;
1980 		LFS_SYNC_CLEANERINFO(cip, fs, bp, 1);
1981 		lfs_segwrite(ap->a_vp->v_mount, SEGM_FORCE_CKP);
1982 		fs->lfs_sp->seg_flags |= SEGM_PROT;
1983 		lfs_segunlock(fs);
1984 		lfs_writer_leave(fs);
1985 
1986 #ifdef DEBUG
1987 		LFS_CLEANERINFO(cip, fs, bp);
1988 		DLOG((DLOG_CLEAN, "lfs_fcntl: reclaim wrote %" PRId64
1989 		      " blocks, cleaned %" PRId32 " segments (activesb %d)\n",
1990 		      fs->lfs_offset - off, cip->clean - oclean,
1991 		      fs->lfs_activesb));
1992 		LFS_SYNC_CLEANERINFO(cip, fs, bp, 0);
1993 #else
1994 		__USE(oclean);
1995 		__USE(off);
1996 #endif
1997 
1998 		return 0;
1999 
2000 	    case LFCNIFILEFH_COMPAT:
2001 		/* Return the filehandle of the Ifile */
2002 		if ((error = kauth_authorize_system(l->l_cred,
2003 		    KAUTH_SYSTEM_FILEHANDLE, 0, NULL, NULL, NULL)) != 0)
2004 			return (error);
2005 		fhp = (struct fhandle *)ap->a_data;
2006 		fhp->fh_fsid = *fsidp;
2007 		fh_size = 16;	/* former VFS_MAXFIDSIZ */
2008 		return lfs_vptofh(fs->lfs_ivnode, &(fhp->fh_fid), &fh_size);
2009 
2010 	    case LFCNIFILEFH_COMPAT2:
2011 	    case LFCNIFILEFH:
2012 		/* Return the filehandle of the Ifile */
2013 		fhp = (struct fhandle *)ap->a_data;
2014 		fhp->fh_fsid = *fsidp;
2015 		fh_size = sizeof(struct lfs_fhandle) -
2016 		    offsetof(fhandle_t, fh_fid);
2017 		return lfs_vptofh(fs->lfs_ivnode, &(fhp->fh_fid), &fh_size);
2018 
2019 	    case LFCNREWIND:
2020 		/* Move lfs_offset to the lowest-numbered segment */
2021 		return lfs_rewind(fs, *(int *)ap->a_data);
2022 
2023 	    case LFCNINVAL:
2024 		/* Mark a segment SEGUSE_INVAL */
2025 		LFS_SEGENTRY(sup, fs, *(int *)ap->a_data, bp);
2026 		if (sup->su_nbytes > 0) {
2027 			brelse(bp, 0);
2028 			lfs_unset_inval_all(fs);
2029 			return EBUSY;
2030 		}
2031 		sup->su_flags |= SEGUSE_INVAL;
2032 		VOP_BWRITE(bp->b_vp, bp);
2033 		return 0;
2034 
2035 	    case LFCNRESIZE:
2036 		/* Resize the filesystem */
2037 		return lfs_resize_fs(fs, *(int *)ap->a_data);
2038 
2039 	    case LFCNWRAPSTOP:
2040 	    case LFCNWRAPSTOP_COMPAT:
2041 		/*
2042 		 * Hold lfs_newseg at segment 0; if requested, sleep until
2043 		 * the filesystem wraps around.  To support external agents
2044 		 * (dump, fsck-based regression test) that need to look at
2045 		 * a snapshot of the filesystem, without necessarily
2046 		 * requiring that all fs activity stops.
2047 		 */
2048 		if (fs->lfs_stoplwp == curlwp)
2049 			return EALREADY;
2050 
2051 		mutex_enter(&lfs_lock);
2052 		while (fs->lfs_stoplwp != NULL)
2053 			cv_wait(&fs->lfs_stopcv, &lfs_lock);
2054 		fs->lfs_stoplwp = curlwp;
2055 		if (fs->lfs_nowrap == 0)
2056 			log(LOG_NOTICE, "%s: disabled log wrap\n", fs->lfs_fsmnt);
2057 		++fs->lfs_nowrap;
2058 		if (*(int *)ap->a_data == 1
2059 		    || ap->a_command == LFCNWRAPSTOP_COMPAT) {
2060 			log(LOG_NOTICE, "LFCNSTOPWRAP waiting for log wrap\n");
2061 			error = mtsleep(&fs->lfs_nowrap, PCATCH | PUSER,
2062 				"segwrap", 0, &lfs_lock);
2063 			log(LOG_NOTICE, "LFCNSTOPWRAP done waiting\n");
2064 			if (error) {
2065 				lfs_wrapgo(fs, VTOI(ap->a_vp), 0);
2066 			}
2067 		}
2068 		mutex_exit(&lfs_lock);
2069 		return 0;
2070 
2071 	    case LFCNWRAPGO:
2072 	    case LFCNWRAPGO_COMPAT:
2073 		/*
2074 		 * Having done its work, the agent wakes up the writer.
2075 		 * If the argument is 1, it sleeps until a new segment
2076 		 * is selected.
2077 		 */
2078 		mutex_enter(&lfs_lock);
2079 		error = lfs_wrapgo(fs, VTOI(ap->a_vp),
2080 				   ap->a_command == LFCNWRAPGO_COMPAT ? 1 :
2081 				    *((int *)ap->a_data));
2082 		mutex_exit(&lfs_lock);
2083 		return error;
2084 
2085 	    case LFCNWRAPPASS:
2086 		if ((VTOI(ap->a_vp)->i_lfs_iflags & LFSI_WRAPWAIT))
2087 			return EALREADY;
2088 		mutex_enter(&lfs_lock);
2089 		if (fs->lfs_stoplwp != curlwp) {
2090 			mutex_exit(&lfs_lock);
2091 			return EALREADY;
2092 		}
2093 		if (fs->lfs_nowrap == 0) {
2094 			mutex_exit(&lfs_lock);
2095 			return EBUSY;
2096 		}
2097 		fs->lfs_wrappass = 1;
2098 		wakeup(&fs->lfs_wrappass);
2099 		/* Wait for the log to wrap, if asked */
2100 		if (*(int *)ap->a_data) {
2101 			mutex_enter(ap->a_vp->v_interlock);
2102 			if (lfs_vref(ap->a_vp) != 0)
2103 				panic("LFCNWRAPPASS: lfs_vref failed");
2104 			VTOI(ap->a_vp)->i_lfs_iflags |= LFSI_WRAPWAIT;
2105 			log(LOG_NOTICE, "LFCNPASS waiting for log wrap\n");
2106 			error = mtsleep(&fs->lfs_nowrap, PCATCH | PUSER,
2107 				"segwrap", 0, &lfs_lock);
2108 			log(LOG_NOTICE, "LFCNPASS done waiting\n");
2109 			VTOI(ap->a_vp)->i_lfs_iflags &= ~LFSI_WRAPWAIT;
2110 			lfs_vunref(ap->a_vp);
2111 		}
2112 		mutex_exit(&lfs_lock);
2113 		return error;
2114 
2115 	    case LFCNWRAPSTATUS:
2116 		mutex_enter(&lfs_lock);
2117 		*(int *)ap->a_data = fs->lfs_wrapstatus;
2118 		mutex_exit(&lfs_lock);
2119 		return 0;
2120 
2121 	    default:
2122 		return ulfs_fcntl(v);
2123 	}
2124 	return 0;
2125 }
2126 
2127 /*
2128  * Return the last logical file offset that should be written for this file
2129  * if we're doing a write that ends at "size".	If writing, we need to know
2130  * about sizes on disk, i.e. fragments if there are any; if reading, we need
2131  * to know about entire blocks.
2132  */
2133 void
2134 lfs_gop_size(struct vnode *vp, off_t size, off_t *eobp, int flags)
2135 {
2136 	struct inode *ip = VTOI(vp);
2137 	struct lfs *fs = ip->i_lfs;
2138 	daddr_t olbn, nlbn;
2139 
2140 	olbn = lfs_lblkno(fs, ip->i_size);
2141 	nlbn = lfs_lblkno(fs, size);
2142 	if (!(flags & GOP_SIZE_MEM) && nlbn < ULFS_NDADDR && olbn <= nlbn) {
2143 		*eobp = lfs_fragroundup(fs, size);
2144 	} else {
2145 		*eobp = lfs_blkroundup(fs, size);
2146 	}
2147 }
2148 
2149 #ifdef DEBUG
2150 void lfs_dump_vop(void *);
2151 
2152 void
2153 lfs_dump_vop(void *v)
2154 {
2155 	struct vop_putpages_args /* {
2156 		struct vnode *a_vp;
2157 		voff_t a_offlo;
2158 		voff_t a_offhi;
2159 		int a_flags;
2160 	} */ *ap = v;
2161 
2162 #ifdef DDB
2163 	vfs_vnode_print(ap->a_vp, 0, printf);
2164 #endif
2165 	lfs_dump_dinode(VTOI(ap->a_vp)->i_din.ffs1_din);
2166 }
2167 #endif
2168 
2169 int
2170 lfs_mmap(void *v)
2171 {
2172 	struct vop_mmap_args /* {
2173 		const struct vnodeop_desc *a_desc;
2174 		struct vnode *a_vp;
2175 		vm_prot_t a_prot;
2176 		kauth_cred_t a_cred;
2177 	} */ *ap = v;
2178 
2179 	if (VTOI(ap->a_vp)->i_number == LFS_IFILE_INUM)
2180 		return EOPNOTSUPP;
2181 	return ulfs_mmap(v);
2182 }
2183 
2184 static int
2185 lfs_openextattr(void *v)
2186 {
2187 	struct vop_openextattr_args /* {
2188 		struct vnode *a_vp;
2189 		kauth_cred_t a_cred;
2190 		struct proc *a_p;
2191 	} */ *ap = v;
2192 	struct inode *ip = VTOI(ap->a_vp);
2193 	struct ulfsmount *ump = ip->i_ump;
2194 	//struct lfs *fs = ip->i_lfs;
2195 
2196 	/* Not supported for ULFS1 file systems. */
2197 	if (ump->um_fstype == ULFS1)
2198 		return (EOPNOTSUPP);
2199 
2200 	/* XXX Not implemented for ULFS2 file systems. */
2201 	return (EOPNOTSUPP);
2202 }
2203 
2204 static int
2205 lfs_closeextattr(void *v)
2206 {
2207 	struct vop_closeextattr_args /* {
2208 		struct vnode *a_vp;
2209 		int a_commit;
2210 		kauth_cred_t a_cred;
2211 		struct proc *a_p;
2212 	} */ *ap = v;
2213 	struct inode *ip = VTOI(ap->a_vp);
2214 	struct ulfsmount *ump = ip->i_ump;
2215 	//struct lfs *fs = ip->i_lfs;
2216 
2217 	/* Not supported for ULFS1 file systems. */
2218 	if (ump->um_fstype == ULFS1)
2219 		return (EOPNOTSUPP);
2220 
2221 	/* XXX Not implemented for ULFS2 file systems. */
2222 	return (EOPNOTSUPP);
2223 }
2224 
2225 static int
2226 lfs_getextattr(void *v)
2227 {
2228 	struct vop_getextattr_args /* {
2229 		struct vnode *a_vp;
2230 		int a_attrnamespace;
2231 		const char *a_name;
2232 		struct uio *a_uio;
2233 		size_t *a_size;
2234 		kauth_cred_t a_cred;
2235 		struct proc *a_p;
2236 	} */ *ap = v;
2237 	struct vnode *vp = ap->a_vp;
2238 	struct inode *ip = VTOI(vp);
2239 	struct ulfsmount *ump = ip->i_ump;
2240 	//struct lfs *fs = ip->i_lfs;
2241 	int error;
2242 
2243 	if (ump->um_fstype == ULFS1) {
2244 #ifdef LFS_EXTATTR
2245 		fstrans_start(vp->v_mount, FSTRANS_SHARED);
2246 		error = ulfs_getextattr(ap);
2247 		fstrans_done(vp->v_mount);
2248 #else
2249 		error = EOPNOTSUPP;
2250 #endif
2251 		return error;
2252 	}
2253 
2254 	/* XXX Not implemented for ULFS2 file systems. */
2255 	return (EOPNOTSUPP);
2256 }
2257 
2258 static int
2259 lfs_setextattr(void *v)
2260 {
2261 	struct vop_setextattr_args /* {
2262 		struct vnode *a_vp;
2263 		int a_attrnamespace;
2264 		const char *a_name;
2265 		struct uio *a_uio;
2266 		kauth_cred_t a_cred;
2267 		struct proc *a_p;
2268 	} */ *ap = v;
2269 	struct vnode *vp = ap->a_vp;
2270 	struct inode *ip = VTOI(vp);
2271 	struct ulfsmount *ump = ip->i_ump;
2272 	//struct lfs *fs = ip->i_lfs;
2273 	int error;
2274 
2275 	if (ump->um_fstype == ULFS1) {
2276 #ifdef LFS_EXTATTR
2277 		fstrans_start(vp->v_mount, FSTRANS_SHARED);
2278 		error = ulfs_setextattr(ap);
2279 		fstrans_done(vp->v_mount);
2280 #else
2281 		error = EOPNOTSUPP;
2282 #endif
2283 		return error;
2284 	}
2285 
2286 	/* XXX Not implemented for ULFS2 file systems. */
2287 	return (EOPNOTSUPP);
2288 }
2289 
2290 static int
2291 lfs_listextattr(void *v)
2292 {
2293 	struct vop_listextattr_args /* {
2294 		struct vnode *a_vp;
2295 		int a_attrnamespace;
2296 		struct uio *a_uio;
2297 		size_t *a_size;
2298 		kauth_cred_t a_cred;
2299 		struct proc *a_p;
2300 	} */ *ap = v;
2301 	struct vnode *vp = ap->a_vp;
2302 	struct inode *ip = VTOI(vp);
2303 	struct ulfsmount *ump = ip->i_ump;
2304 	//struct lfs *fs = ip->i_lfs;
2305 	int error;
2306 
2307 	if (ump->um_fstype == ULFS1) {
2308 #ifdef LFS_EXTATTR
2309 		fstrans_start(vp->v_mount, FSTRANS_SHARED);
2310 		error = ulfs_listextattr(ap);
2311 		fstrans_done(vp->v_mount);
2312 #else
2313 		error = EOPNOTSUPP;
2314 #endif
2315 		return error;
2316 	}
2317 
2318 	/* XXX Not implemented for ULFS2 file systems. */
2319 	return (EOPNOTSUPP);
2320 }
2321 
2322 static int
2323 lfs_deleteextattr(void *v)
2324 {
2325 	struct vop_deleteextattr_args /* {
2326 		struct vnode *a_vp;
2327 		int a_attrnamespace;
2328 		kauth_cred_t a_cred;
2329 		struct proc *a_p;
2330 	} */ *ap = v;
2331 	struct vnode *vp = ap->a_vp;
2332 	struct inode *ip = VTOI(vp);
2333 	struct ulfsmount *ump = ip->i_ump;
2334 	//struct fs *fs = ip->i_lfs;
2335 	int error;
2336 
2337 	if (ump->um_fstype == ULFS1) {
2338 #ifdef LFS_EXTATTR
2339 		fstrans_start(vp->v_mount, FSTRANS_SHARED);
2340 		error = ulfs_deleteextattr(ap);
2341 		fstrans_done(vp->v_mount);
2342 #else
2343 		error = EOPNOTSUPP;
2344 #endif
2345 		return error;
2346 	}
2347 
2348 	/* XXX Not implemented for ULFS2 file systems. */
2349 	return (EOPNOTSUPP);
2350 }
2351