xref: /freebsd/sys/ufs/ffs/ffs_vnops.c (revision 1b6c2589164a3a7b2f62d4c28c2ffa1be860959e)
1 /*
2  * Copyright (c) 1982, 1986, 1989, 1993
3  *	The Regents of the University of California.  All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 3. All advertising materials mentioning features or use of this software
14  *    must display the following acknowledgement:
15  *	This product includes software developed by the University of
16  *	California, Berkeley and its contributors.
17  * 4. Neither the name of the University nor the names of its contributors
18  *    may be used to endorse or promote products derived from this software
19  *    without specific prior written permission.
20  *
21  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31  * SUCH DAMAGE.
32  *
33  *	@(#)ffs_vnops.c	8.15 (Berkeley) 5/14/95
34  * $FreeBSD$
35  */
36 
37 #include "opt_ufs.h"
38 
39 #include <sys/param.h>
40 #include <sys/systm.h>
41 #include <sys/resourcevar.h>
42 #include <sys/signalvar.h>
43 #include <sys/kernel.h>
44 #include <sys/stat.h>
45 #include <sys/bio.h>
46 #include <sys/buf.h>
47 #include <sys/proc.h>
48 #include <sys/mount.h>
49 #include <sys/vnode.h>
50 #include <sys/conf.h>
51 
52 #include <machine/limits.h>
53 
54 #include <vm/vm.h>
55 #include <vm/vm_page.h>
56 #include <vm/vm_object.h>
57 #include <vm/vm_extern.h>
58 
59 #include <ufs/ufs/extattr.h>
60 #include <ufs/ufs/quota.h>
61 #include <ufs/ufs/inode.h>
62 #include <ufs/ufs/ufsmount.h>
63 #include <ufs/ufs/ufs_extern.h>
64 
65 #include <ufs/ffs/fs.h>
66 #include <ufs/ffs/ffs_extern.h>
67 
68 int	ffs_fsync __P((struct vop_fsync_args *));
69 static int	ffs_getpages __P((struct vop_getpages_args *));
70 static int	ffs_read __P((struct vop_read_args *));
71 static int	ffs_write __P((struct vop_write_args *));
72 
73 /* Global vfs data structures for ufs. */
74 vop_t **ffs_vnodeop_p;
75 static struct vnodeopv_entry_desc ffs_vnodeop_entries[] = {
76 	{ &vop_default_desc,		(vop_t *) ufs_vnoperate },
77 	{ &vop_fsync_desc,		(vop_t *) ffs_fsync },
78 	{ &vop_getpages_desc,		(vop_t *) ffs_getpages },
79 	{ &vop_read_desc,		(vop_t *) ffs_read },
80 	{ &vop_reallocblks_desc,	(vop_t *) ffs_reallocblks },
81 	{ &vop_write_desc,		(vop_t *) ffs_write },
82 #ifdef UFS_EXTATTR
83 	{ &vop_getextattr_desc, 	(vop_t *) ufs_vop_getextattr },
84 	{ &vop_setextattr_desc,		(vop_t *) ufs_vop_setextattr },
85 #endif
86 	{ NULL, NULL }
87 };
88 static struct vnodeopv_desc ffs_vnodeop_opv_desc =
89 	{ &ffs_vnodeop_p, ffs_vnodeop_entries };
90 
91 vop_t **ffs_specop_p;
92 static struct vnodeopv_entry_desc ffs_specop_entries[] = {
93 	{ &vop_default_desc,		(vop_t *) ufs_vnoperatespec },
94 	{ &vop_fsync_desc,		(vop_t *) ffs_fsync },
95 #ifdef UFS_EXTATTR
96 	{ &vop_getextattr_desc,		(vop_t *) ufs_vop_getextattr },
97 	{ &vop_setextattr_desc,		(vop_t *) ufs_vop_setextattr },
98 #endif
99 	{ NULL, NULL }
100 };
101 static struct vnodeopv_desc ffs_specop_opv_desc =
102 	{ &ffs_specop_p, ffs_specop_entries };
103 
104 vop_t **ffs_fifoop_p;
105 static struct vnodeopv_entry_desc ffs_fifoop_entries[] = {
106 	{ &vop_default_desc,		(vop_t *) ufs_vnoperatefifo },
107 #ifdef UFS_EXTATTR
108 	{ &vop_getextattr_desc,		(vop_t *) ufs_vop_getextattr },
109 	{ &vop_setextattr_desc,		(vop_t *) ufs_vop_setextattr },
110 #endif
111 	{ &vop_fsync_desc,		(vop_t *) ffs_fsync },
112 	{ NULL, NULL }
113 };
114 static struct vnodeopv_desc ffs_fifoop_opv_desc =
115 	{ &ffs_fifoop_p, ffs_fifoop_entries };
116 
117 VNODEOP_SET(ffs_vnodeop_opv_desc);
118 VNODEOP_SET(ffs_specop_opv_desc);
119 VNODEOP_SET(ffs_fifoop_opv_desc);
120 
121 #include <ufs/ufs/ufs_readwrite.c>
122 
123 /*
124  * Synch an open file.
125  */
126 /* ARGSUSED */
127 int
128 ffs_fsync(ap)
129 	struct vop_fsync_args /* {
130 		struct vnode *a_vp;
131 		struct ucred *a_cred;
132 		int a_waitfor;
133 		struct thread *a_td;
134 	} */ *ap;
135 {
136 	struct vnode *vp = ap->a_vp;
137 	struct inode *ip = VTOI(vp);
138 	struct buf *bp;
139 	struct buf *nbp;
140 	int s, error, wait, passes, skipmeta;
141 	daddr_t lbn;
142 
143 	/*
144 	 * Snapshots have to be unlocked so they do not deadlock
145 	 * checking whether they need to copy their written buffers.
146 	 * We always hold a reference, so they cannot be removed
147 	 * out from underneath us.
148 	 */
149 	if (ip->i_flags & SF_SNAPSHOT)
150 		VOP_UNLOCK(vp, 0, ap->a_td);
151 	wait = (ap->a_waitfor == MNT_WAIT);
152 	if (vn_isdisk(vp, NULL)) {
153 		lbn = INT_MAX;
154 		if (vp->v_rdev->si_mountpoint != NULL &&
155 		    (vp->v_rdev->si_mountpoint->mnt_flag & MNT_SOFTDEP))
156 			softdep_fsync_mountdev(vp);
157 	} else {
158 		lbn = lblkno(ip->i_fs, (ip->i_size + ip->i_fs->fs_bsize - 1));
159 	}
160 
161 	/*
162 	 * Flush all dirty buffers associated with a vnode.
163 	 */
164 	passes = NIADDR + 1;
165 	skipmeta = 0;
166 	if (wait)
167 		skipmeta = 1;
168 	s = splbio();
169 loop:
170 	TAILQ_FOREACH(bp, &vp->v_dirtyblkhd, b_vnbufs)
171 		bp->b_flags &= ~B_SCANNED;
172 	for (bp = TAILQ_FIRST(&vp->v_dirtyblkhd); bp; bp = nbp) {
173 		nbp = TAILQ_NEXT(bp, b_vnbufs);
174 		/*
175 		 * Reasons to skip this buffer: it has already been considered
176 		 * on this pass, this pass is the first time through on a
177 		 * synchronous flush request and the buffer being considered
178 		 * is metadata, the buffer has dependencies that will cause
179 		 * it to be redirtied and it has not already been deferred,
180 		 * or it is already being written.
181 		 */
182 		if ((bp->b_flags & B_SCANNED) != 0)
183 			continue;
184 		bp->b_flags |= B_SCANNED;
185 		if ((skipmeta == 1 && bp->b_lblkno < 0))
186 			continue;
187 		if (!wait && LIST_FIRST(&bp->b_dep) != NULL &&
188 		    (bp->b_flags & B_DEFERRED) == 0 &&
189 		    buf_countdeps(bp, 0)) {
190 			bp->b_flags |= B_DEFERRED;
191 			continue;
192 		}
193 		if (BUF_LOCK(bp, LK_EXCLUSIVE | LK_NOWAIT))
194 			continue;
195 		if ((bp->b_flags & B_DELWRI) == 0)
196 			panic("ffs_fsync: not dirty");
197 		if (vp != bp->b_vp)
198 			panic("ffs_fsync: vp != vp->b_vp");
199 		/*
200 		 * If this is a synchronous flush request, or it is not a
201 		 * file or device, start the write on this buffer immediatly.
202 		 */
203 		if (wait || (vp->v_type != VREG && vp->v_type != VBLK)) {
204 
205 			/*
206 			 * On our final pass through, do all I/O synchronously
207 			 * so that we can find out if our flush is failing
208 			 * because of write errors.
209 			 */
210 			if (passes > 0 || !wait) {
211 				if ((bp->b_flags & B_CLUSTEROK) && !wait) {
212 					BUF_UNLOCK(bp);
213 					(void) vfs_bio_awrite(bp);
214 				} else {
215 					bremfree(bp);
216 					splx(s);
217 					(void) bawrite(bp);
218 					s = splbio();
219 				}
220 			} else {
221 				bremfree(bp);
222 				splx(s);
223 				if ((error = bwrite(bp)) != 0)
224 					return (error);
225 				s = splbio();
226 			}
227 		} else if ((vp->v_type == VREG) && (bp->b_lblkno >= lbn)) {
228 			/*
229 			 * If the buffer is for data that has been truncated
230 			 * off the file, then throw it away.
231 			 */
232 			bremfree(bp);
233 			bp->b_flags |= B_INVAL | B_NOCACHE;
234 			splx(s);
235 			brelse(bp);
236 			s = splbio();
237 		} else {
238 			BUF_UNLOCK(bp);
239 			vfs_bio_awrite(bp);
240 		}
241 		/*
242 		 * Since we may have slept during the I/O, we need
243 		 * to start from a known point.
244 		 */
245 		nbp = TAILQ_FIRST(&vp->v_dirtyblkhd);
246 	}
247 	/*
248 	 * If we were asked to do this synchronously, then go back for
249 	 * another pass, this time doing the metadata.
250 	 */
251 	if (skipmeta) {
252 		skipmeta = 0;
253 		goto loop;
254 	}
255 
256 	if (wait) {
257 		while (vp->v_numoutput) {
258 			vp->v_flag |= VBWAIT;
259 			(void) tsleep((caddr_t)&vp->v_numoutput,
260 					PRIBIO + 4, "ffsfsn", 0);
261   		}
262 
263 		/*
264 		 * Ensure that any filesystem metatdata associated
265 		 * with the vnode has been written.
266 		 */
267 		splx(s);
268 		if ((error = softdep_sync_metadata(ap)) != 0)
269 			return (error);
270 		s = splbio();
271 
272 		if (!TAILQ_EMPTY(&vp->v_dirtyblkhd)) {
273 			/*
274 			 * Block devices associated with filesystems may
275 			 * have new I/O requests posted for them even if
276 			 * the vnode is locked, so no amount of trying will
277 			 * get them clean. Thus we give block devices a
278 			 * good effort, then just give up. For all other file
279 			 * types, go around and try again until it is clean.
280 			 */
281 			if (passes > 0) {
282 				passes -= 1;
283 				goto loop;
284 			}
285 #ifdef DIAGNOSTIC
286 			if (!vn_isdisk(vp, NULL))
287 				vprint("ffs_fsync: dirty", vp);
288 #endif
289 		}
290 	}
291 	splx(s);
292 	error = UFS_UPDATE(vp, wait);
293 	if (ip->i_flags & SF_SNAPSHOT)
294 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, ap->a_td);
295 	return (error);
296 }
297