xref: /dragonfly/sys/vfs/ufs/ffs_vnops.c (revision 1de703da)
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: src/sys/ufs/ffs/ffs_vnops.c,v 1.64 2000/01/10 12:04:25 phk Exp $
35  * $DragonFly: src/sys/vfs/ufs/ffs_vnops.c,v 1.2 2003/06/17 04:28:59 dillon Exp $
36  */
37 
38 #include <sys/param.h>
39 #include <sys/systm.h>
40 #include <sys/resourcevar.h>
41 #include <sys/signalvar.h>
42 #include <sys/kernel.h>
43 #include <sys/stat.h>
44 #include <sys/buf.h>
45 #include <sys/proc.h>
46 #include <sys/mount.h>
47 #include <sys/vnode.h>
48 #include <sys/conf.h>
49 
50 #include <machine/limits.h>
51 
52 #include <vm/vm.h>
53 #include <vm/vm_page.h>
54 #include <vm/vm_object.h>
55 #include <vm/vm_extern.h>
56 
57 #include <ufs/ufs/quota.h>
58 #include <ufs/ufs/inode.h>
59 #include <ufs/ufs/ufsmount.h>
60 #include <ufs/ufs/ufs_extern.h>
61 
62 #include <ufs/ffs/fs.h>
63 #include <ufs/ffs/ffs_extern.h>
64 
65 static int	ffs_fsync __P((struct vop_fsync_args *));
66 static int	ffs_getpages __P((struct vop_getpages_args *));
67 static int	ffs_putpages __P((struct vop_putpages_args *));
68 static int	ffs_read __P((struct vop_read_args *));
69 static int	ffs_write __P((struct vop_write_args *));
70 
71 /* Global vfs data structures for ufs. */
72 vop_t **ffs_vnodeop_p;
73 static struct vnodeopv_entry_desc ffs_vnodeop_entries[] = {
74 	{ &vop_default_desc,		(vop_t *) ufs_vnoperate },
75 	{ &vop_fsync_desc,		(vop_t *) ffs_fsync },
76 	{ &vop_getpages_desc,		(vop_t *) ffs_getpages },
77 	{ &vop_putpages_desc,		(vop_t *) ffs_putpages },
78 	{ &vop_read_desc,		(vop_t *) ffs_read },
79 	{ &vop_balloc_desc,		(vop_t *) ffs_balloc },
80 	{ &vop_reallocblks_desc,	(vop_t *) ffs_reallocblks },
81 	{ &vop_write_desc,		(vop_t *) ffs_write },
82 	{ NULL, NULL }
83 };
84 static struct vnodeopv_desc ffs_vnodeop_opv_desc =
85 	{ &ffs_vnodeop_p, ffs_vnodeop_entries };
86 
87 vop_t **ffs_specop_p;
88 static struct vnodeopv_entry_desc ffs_specop_entries[] = {
89 	{ &vop_default_desc,		(vop_t *) ufs_vnoperatespec },
90 	{ &vop_fsync_desc,		(vop_t *) ffs_fsync },
91 	{ NULL, NULL }
92 };
93 static struct vnodeopv_desc ffs_specop_opv_desc =
94 	{ &ffs_specop_p, ffs_specop_entries };
95 
96 vop_t **ffs_fifoop_p;
97 static struct vnodeopv_entry_desc ffs_fifoop_entries[] = {
98 	{ &vop_default_desc,		(vop_t *) ufs_vnoperatefifo },
99 	{ &vop_fsync_desc,		(vop_t *) ffs_fsync },
100 	{ NULL, NULL }
101 };
102 static struct vnodeopv_desc ffs_fifoop_opv_desc =
103 	{ &ffs_fifoop_p, ffs_fifoop_entries };
104 
105 VNODEOP_SET(ffs_vnodeop_opv_desc);
106 VNODEOP_SET(ffs_specop_opv_desc);
107 VNODEOP_SET(ffs_fifoop_opv_desc);
108 
109 #include <ufs/ufs/ufs_readwrite.c>
110 
111 /*
112  * Synch an open file.
113  */
114 /* ARGSUSED */
115 static int
116 ffs_fsync(ap)
117 	struct vop_fsync_args /* {
118 		struct vnode *a_vp;
119 		struct ucred *a_cred;
120 		int a_waitfor;
121 		struct proc *a_p;
122 	} */ *ap;
123 {
124 	struct vnode *vp = ap->a_vp;
125 	struct buf *bp;
126 	struct buf *nbp;
127 	int s, error, wait, passes, skipmeta;
128 	daddr_t lbn;
129 
130 	wait = (ap->a_waitfor == MNT_WAIT);
131 	if (vn_isdisk(vp, NULL)) {
132 		lbn = INT_MAX;
133 		if (vp->v_specmountpoint != NULL &&
134 		    (vp->v_specmountpoint->mnt_flag & MNT_SOFTDEP))
135 			softdep_fsync_mountdev(vp);
136 	} else {
137 		struct inode *ip;
138 		ip = VTOI(vp);
139 		lbn = lblkno(ip->i_fs, (ip->i_size + ip->i_fs->fs_bsize - 1));
140 	}
141 
142 	/*
143 	 * Flush all dirty buffers associated with a vnode.
144 	 */
145 	passes = NIADDR + 1;
146 	skipmeta = 0;
147 	if (wait)
148 		skipmeta = 1;
149 	s = splbio();
150 loop:
151 	for (bp = TAILQ_FIRST(&vp->v_dirtyblkhd); bp;
152 	     bp = TAILQ_NEXT(bp, b_vnbufs))
153 		bp->b_flags &= ~B_SCANNED;
154 	for (bp = TAILQ_FIRST(&vp->v_dirtyblkhd); bp; bp = nbp) {
155 		nbp = TAILQ_NEXT(bp, b_vnbufs);
156 		/*
157 		 * Reasons to skip this buffer: it has already been considered
158 		 * on this pass, this pass is the first time through on a
159 		 * synchronous flush request and the buffer being considered
160 		 * is metadata, the buffer has dependencies that will cause
161 		 * it to be redirtied and it has not already been deferred,
162 		 * or it is already being written.
163 		 */
164 		if ((bp->b_flags & B_SCANNED) != 0)
165 			continue;
166 		bp->b_flags |= B_SCANNED;
167 		if ((skipmeta == 1 && bp->b_lblkno < 0))
168 			continue;
169 		if (!wait && LIST_FIRST(&bp->b_dep) != NULL &&
170 		    (bp->b_flags & B_DEFERRED) == 0 &&
171 		    bioops.io_countdeps && (*bioops.io_countdeps)(bp, 0)) {
172 			bp->b_flags |= B_DEFERRED;
173 			continue;
174 		}
175 		if (BUF_LOCK(bp, LK_EXCLUSIVE | LK_NOWAIT))
176 			continue;
177 		if ((bp->b_flags & B_DELWRI) == 0)
178 			panic("ffs_fsync: not dirty");
179 		if (vp != bp->b_vp)
180 			panic("ffs_fsync: vp != vp->b_vp");
181 		/*
182 		 * If this is a synchronous flush request, or it is not a
183 		 * file or device, start the write on this buffer immediatly.
184 		 */
185 		if (wait || (vp->v_type != VREG && vp->v_type != VBLK)) {
186 
187 			/*
188 			 * On our final pass through, do all I/O synchronously
189 			 * so that we can find out if our flush is failing
190 			 * because of write errors.
191 			 */
192 			if (passes > 0 || !wait) {
193 				if ((bp->b_flags & B_CLUSTEROK) && !wait) {
194 					BUF_UNLOCK(bp);
195 					(void) vfs_bio_awrite(bp);
196 				} else {
197 					bremfree(bp);
198 					splx(s);
199 					(void) bawrite(bp);
200 					s = splbio();
201 				}
202 			} else {
203 				bremfree(bp);
204 				splx(s);
205 				if ((error = bwrite(bp)) != 0)
206 					return (error);
207 				s = splbio();
208 			}
209 		} else if ((vp->v_type == VREG) && (bp->b_lblkno >= lbn)) {
210 			/*
211 			 * If the buffer is for data that has been truncated
212 			 * off the file, then throw it away.
213 			 */
214 			bremfree(bp);
215 			bp->b_flags |= B_INVAL | B_NOCACHE;
216 			splx(s);
217 			brelse(bp);
218 			s = splbio();
219 		} else {
220 			BUF_UNLOCK(bp);
221 			vfs_bio_awrite(bp);
222 		}
223 		/*
224 		 * Since we may have slept during the I/O, we need
225 		 * to start from a known point.
226 		 */
227 		nbp = TAILQ_FIRST(&vp->v_dirtyblkhd);
228 	}
229 	/*
230 	 * If we were asked to do this synchronously, then go back for
231 	 * another pass, this time doing the metadata.
232 	 */
233 	if (skipmeta) {
234 		skipmeta = 0;
235 		goto loop;
236 	}
237 
238 	if (wait) {
239 		while (vp->v_numoutput) {
240 			vp->v_flag |= VBWAIT;
241 			(void) tsleep((caddr_t)&vp->v_numoutput,
242 					PRIBIO + 4, "ffsfsn", 0);
243   		}
244 
245 		/*
246 		 * Ensure that any filesystem metatdata associated
247 		 * with the vnode has been written.
248 		 */
249 		splx(s);
250 		if ((error = softdep_sync_metadata(ap)) != 0)
251 			return (error);
252 		s = splbio();
253 
254 		if (!TAILQ_EMPTY(&vp->v_dirtyblkhd)) {
255 			/*
256 			 * Block devices associated with filesystems may
257 			 * have new I/O requests posted for them even if
258 			 * the vnode is locked, so no amount of trying will
259 			 * get them clean. Thus we give block devices a
260 			 * good effort, then just give up. For all other file
261 			 * types, go around and try again until it is clean.
262 			 */
263 			if (passes > 0) {
264 				passes -= 1;
265 				goto loop;
266 			}
267 #ifdef DIAGNOSTIC
268 			if (!vn_isdisk(vp, NULL))
269 				vprint("ffs_fsync: dirty", vp);
270 #endif
271 		}
272 	}
273 	splx(s);
274 	return (UFS_UPDATE(vp, wait));
275 }
276