xref: /freebsd/sys/ufs/ffs/ffs_vnops.c (revision daf1cffce2e07931f27c6c6998652e90df6ba87e)
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 <sys/param.h>
38 #include <sys/systm.h>
39 #include <sys/resourcevar.h>
40 #include <sys/signalvar.h>
41 #include <sys/kernel.h>
42 #include <sys/stat.h>
43 #include <sys/buf.h>
44 #include <sys/proc.h>
45 #include <sys/mount.h>
46 #include <sys/vnode.h>
47 #include <sys/conf.h>
48 
49 #include <machine/limits.h>
50 
51 #include <vm/vm.h>
52 #include <vm/vm_page.h>
53 #include <vm/vm_object.h>
54 #include <vm/vm_extern.h>
55 
56 #include <ufs/ufs/quota.h>
57 #include <ufs/ufs/inode.h>
58 #include <ufs/ufs/ufsmount.h>
59 #include <ufs/ufs/ufs_extern.h>
60 
61 #include <ufs/ffs/fs.h>
62 #include <ufs/ffs/ffs_extern.h>
63 
64 static int	ffs_fsync __P((struct vop_fsync_args *));
65 static int	ffs_getpages __P((struct vop_getpages_args *));
66 static int	ffs_putpages __P((struct vop_putpages_args *));
67 static int	ffs_read __P((struct vop_read_args *));
68 static int	ffs_write __P((struct vop_write_args *));
69 
70 /* Global vfs data structures for ufs. */
71 vop_t **ffs_vnodeop_p;
72 static struct vnodeopv_entry_desc ffs_vnodeop_entries[] = {
73 	{ &vop_default_desc,		(vop_t *) ufs_vnoperate },
74 	{ &vop_fsync_desc,		(vop_t *) ffs_fsync },
75 	{ &vop_getpages_desc,		(vop_t *) ffs_getpages },
76 	{ &vop_putpages_desc,		(vop_t *) ffs_putpages },
77 	{ &vop_read_desc,		(vop_t *) ffs_read },
78 	{ &vop_balloc_desc,		(vop_t *) ffs_balloc },
79 	{ &vop_reallocblks_desc,	(vop_t *) ffs_reallocblks },
80 	{ &vop_write_desc,		(vop_t *) ffs_write },
81 	{ NULL, NULL }
82 };
83 static struct vnodeopv_desc ffs_vnodeop_opv_desc =
84 	{ &ffs_vnodeop_p, ffs_vnodeop_entries };
85 
86 vop_t **ffs_specop_p;
87 static struct vnodeopv_entry_desc ffs_specop_entries[] = {
88 	{ &vop_default_desc,		(vop_t *) ufs_vnoperatespec },
89 	{ &vop_fsync_desc,		(vop_t *) ffs_fsync },
90 	{ NULL, NULL }
91 };
92 static struct vnodeopv_desc ffs_specop_opv_desc =
93 	{ &ffs_specop_p, ffs_specop_entries };
94 
95 vop_t **ffs_fifoop_p;
96 static struct vnodeopv_entry_desc ffs_fifoop_entries[] = {
97 	{ &vop_default_desc,		(vop_t *) ufs_vnoperatefifo },
98 	{ &vop_fsync_desc,		(vop_t *) ffs_fsync },
99 	{ NULL, NULL }
100 };
101 static struct vnodeopv_desc ffs_fifoop_opv_desc =
102 	{ &ffs_fifoop_p, ffs_fifoop_entries };
103 
104 VNODEOP_SET(ffs_vnodeop_opv_desc);
105 VNODEOP_SET(ffs_specop_opv_desc);
106 VNODEOP_SET(ffs_fifoop_opv_desc);
107 
108 #include <ufs/ufs/ufs_readwrite.c>
109 
110 /*
111  * Synch an open file.
112  */
113 /* ARGSUSED */
114 static int
115 ffs_fsync(ap)
116 	struct vop_fsync_args /* {
117 		struct vnode *a_vp;
118 		struct ucred *a_cred;
119 		int a_waitfor;
120 		struct proc *a_p;
121 	} */ *ap;
122 {
123 	struct vnode *vp = ap->a_vp;
124 	struct buf *bp;
125 	struct buf *nbp;
126 	int s, error, wait, passes, skipmeta;
127 	daddr_t lbn;
128 
129 	wait = (ap->a_waitfor == MNT_WAIT);
130 	if (vn_isdisk(vp, NULL)) {
131 		lbn = INT_MAX;
132 		if (vp->v_specmountpoint != NULL &&
133 		    (vp->v_specmountpoint->mnt_flag & MNT_SOFTDEP))
134 			softdep_fsync_mountdev(vp);
135 	} else {
136 		struct inode *ip;
137 		ip = VTOI(vp);
138 		lbn = lblkno(ip->i_fs, (ip->i_size + ip->i_fs->fs_bsize - 1));
139 	}
140 
141 	/*
142 	 * Flush all dirty buffers associated with a vnode.
143 	 */
144 	passes = NIADDR + 1;
145 	skipmeta = 0;
146 	if (wait)
147 		skipmeta = 1;
148 	s = splbio();
149 loop:
150 	for (bp = TAILQ_FIRST(&vp->v_dirtyblkhd); bp;
151 	     bp = TAILQ_NEXT(bp, b_vnbufs))
152 		bp->b_flags &= ~B_SCANNED;
153 	for (bp = TAILQ_FIRST(&vp->v_dirtyblkhd); bp; bp = nbp) {
154 		nbp = TAILQ_NEXT(bp, b_vnbufs);
155 		/*
156 		 * Reasons to skip this buffer: it has already been considered
157 		 * on this pass, this pass is the first time through on a
158 		 * synchronous flush request and the buffer being considered
159 		 * is metadata, the buffer has dependencies that will cause
160 		 * it to be redirtied and it has not already been deferred,
161 		 * or it is already being written.
162 		 */
163 		if ((bp->b_flags & B_SCANNED) != 0)
164 			continue;
165 		bp->b_flags |= B_SCANNED;
166 		if ((skipmeta == 1 && bp->b_lblkno < 0))
167 			continue;
168 		if (!wait && LIST_FIRST(&bp->b_dep) != NULL &&
169 		    (bp->b_flags & B_DEFERRED) == 0 &&
170 		    bioops.io_countdeps && (*bioops.io_countdeps)(bp, 0)) {
171 			bp->b_flags |= B_DEFERRED;
172 			continue;
173 		}
174 		if (BUF_LOCK(bp, LK_EXCLUSIVE | LK_NOWAIT))
175 			continue;
176 		if ((bp->b_flags & B_DELWRI) == 0)
177 			panic("ffs_fsync: not dirty");
178 		if (vp != bp->b_vp)
179 			panic("ffs_fsync: vp != vp->b_vp");
180 		/*
181 		 * If this is a synchronous flush request, or it is not a
182 		 * file or device, start the write on this buffer immediatly.
183 		 */
184 		if (wait || (vp->v_type != VREG && vp->v_type != VBLK)) {
185 
186 			/*
187 			 * On our final pass through, do all I/O synchronously
188 			 * so that we can find out if our flush is failing
189 			 * because of write errors.
190 			 */
191 			if (passes > 0 || !wait) {
192 				if ((bp->b_flags & B_CLUSTEROK) && !wait) {
193 					BUF_UNLOCK(bp);
194 					(void) vfs_bio_awrite(bp);
195 				} else {
196 					bremfree(bp);
197 					splx(s);
198 					(void) bawrite(bp);
199 					s = splbio();
200 				}
201 			} else {
202 				bremfree(bp);
203 				splx(s);
204 				if ((error = bwrite(bp)) != 0)
205 					return (error);
206 				s = splbio();
207 			}
208 		} else if ((vp->v_type == VREG) && (bp->b_lblkno >= lbn)) {
209 			/*
210 			 * If the buffer is for data that has been truncated
211 			 * off the file, then throw it away.
212 			 */
213 			bremfree(bp);
214 			bp->b_flags |= B_INVAL | B_NOCACHE;
215 			splx(s);
216 			brelse(bp);
217 			s = splbio();
218 		} else {
219 			BUF_UNLOCK(bp);
220 			vfs_bio_awrite(bp);
221 		}
222 		/*
223 		 * Since we may have slept during the I/O, we need
224 		 * to start from a known point.
225 		 */
226 		nbp = TAILQ_FIRST(&vp->v_dirtyblkhd);
227 	}
228 	/*
229 	 * If we were asked to do this synchronously, then go back for
230 	 * another pass, this time doing the metadata.
231 	 */
232 	if (skipmeta) {
233 		skipmeta = 0;
234 		goto loop;
235 	}
236 
237 	if (wait) {
238 		while (vp->v_numoutput) {
239 			vp->v_flag |= VBWAIT;
240 			(void) tsleep((caddr_t)&vp->v_numoutput,
241 					PRIBIO + 4, "ffsfsn", 0);
242   		}
243 
244 		/*
245 		 * Ensure that any filesystem metatdata associated
246 		 * with the vnode has been written.
247 		 */
248 		splx(s);
249 		if ((error = softdep_sync_metadata(ap)) != 0)
250 			return (error);
251 		s = splbio();
252 
253 		if (!TAILQ_EMPTY(&vp->v_dirtyblkhd)) {
254 			/*
255 			 * Block devices associated with filesystems may
256 			 * have new I/O requests posted for them even if
257 			 * the vnode is locked, so no amount of trying will
258 			 * get them clean. Thus we give block devices a
259 			 * good effort, then just give up. For all other file
260 			 * types, go around and try again until it is clean.
261 			 */
262 			if (passes > 0) {
263 				passes -= 1;
264 				goto loop;
265 			}
266 #ifdef DIAGNOSTIC
267 			if (!vn_isdisk(vp, NULL))
268 				vprint("ffs_fsync: dirty", vp);
269 #endif
270 		}
271 	}
272 	splx(s);
273 	return (UFS_UPDATE(vp, wait));
274 }
275