xref: /freebsd/sys/ufs/ffs/ffs_vnops.c (revision a1a4f1a0d87b594d3f17a97dc0127eec1417e6f6)
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_prot.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, passes, skipmeta;
128 	daddr_t lbn;
129 
130 
131 	if (vp->v_type == VBLK) {
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 (ap->a_waitfor == MNT_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 		 * First time through on a synchronous call,
158 		 * or if it's already scheduled, skip to the next
159 		 * buffer
160 		 */
161 		if ((bp->b_flags & B_SCANNED) ||
162 		    ((skipmeta == 1) && (bp->b_lblkno < 0)) ||
163 		    BUF_LOCK(bp, LK_EXCLUSIVE | LK_NOWAIT))
164 			continue;
165 		if ((bp->b_flags & B_DELWRI) == 0)
166 			panic("ffs_fsync: not dirty");
167 		/*
168 		 * If data is outstanding to another vnode, or we were
169 		 * asked to wait for everything, or it's not a file or BDEV,
170 		 * start the IO on this buffer immediatly.
171 		 */
172 		bp->b_flags |= B_SCANNED;
173 		if (((bp->b_vp != vp) || (ap->a_waitfor == MNT_WAIT)) ||
174 		    ((vp->v_type != VREG) && (vp->v_type != VBLK))) {
175 
176 			/*
177 			 * On our final pass through, do all I/O synchronously
178 			 * so that we can find out if our flush is failing
179 			 * because of write errors.
180 			 */
181 			if (passes > 0 || (ap->a_waitfor != MNT_WAIT)) {
182 				if ((bp->b_flags & B_CLUSTEROK) &&
183 				    ap->a_waitfor != MNT_WAIT) {
184 					BUF_UNLOCK(bp);
185 					(void) vfs_bio_awrite(bp);
186 				} else {
187 					bremfree(bp);
188 					splx(s);
189 					(void) bawrite(bp);
190 					s = splbio();
191 				}
192 			} else {
193 				bremfree(bp);
194 				splx(s);
195 				if ((error = bwrite(bp)) != 0)
196 					return (error);
197 				s = splbio();
198 			}
199 		} else if ((vp->v_type == VREG) && (bp->b_lblkno >= lbn)) {
200 			/*
201 			 * If the buffer is for data that has been truncated
202 			 * off the file, then throw it away.
203 			 */
204 			bremfree(bp);
205 			bp->b_flags |= B_INVAL | B_NOCACHE;
206 			splx(s);
207 			brelse(bp);
208 			s = splbio();
209 		} else {
210 			BUF_UNLOCK(bp);
211 			vfs_bio_awrite(bp);
212 		}
213 		/*
214 		 * Since we may have slept during the I/O, we need
215 		 * to start from a known point.
216 		 */
217 		nbp = TAILQ_FIRST(&vp->v_dirtyblkhd);
218 	}
219 	/*
220 	 * If we were asked to do this synchronously, then go back for
221 	 * another pass, this time doing the metadata.
222 	 */
223 	if (skipmeta) {
224 		skipmeta = 0;
225 		goto loop;
226 	}
227 
228 	if (ap->a_waitfor == MNT_WAIT) {
229 		while (vp->v_numoutput) {
230 			vp->v_flag |= VBWAIT;
231 			(void) tsleep((caddr_t)&vp->v_numoutput,
232 					PRIBIO + 4, "ffsfsn", 0);
233   		}
234 
235 		/*
236 		 * Ensure that any filesystem metatdata associated
237 		 * with the vnode has been written.
238 		 */
239 		splx(s);
240 		if ((error = softdep_sync_metadata(ap)) != 0)
241 			return (error);
242 		s = splbio();
243 
244 		if (!TAILQ_EMPTY(&vp->v_dirtyblkhd)) {
245 			/*
246 			 * Block devices associated with filesystems may
247 			 * have new I/O requests posted for them even if
248 			 * the vnode is locked, so no amount of trying will
249 			 * get them clean. Thus we give block devices a
250 			 * good effort, then just give up. For all other file
251 			 * types, go around and try again until it is clean.
252 			 */
253 			if (passes > 0) {
254 				passes -= 1;
255 				goto loop;
256 			}
257 #ifdef DIAGNOSTIC
258 			if (vp->v_type != VBLK)
259 				vprint("ffs_fsync: dirty", vp);
260 #endif
261 		}
262 	}
263 	splx(s);
264 	return (UFS_UPDATE(vp, ap->a_waitfor == MNT_WAIT));
265 }
266