xref: /dragonfly/sys/vfs/ufs/ffs_vnops.c (revision 1d1731fa)
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 1.6 2003/08/20 09:56:34 rob 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 <sys/buf2.h>
58 
59 #include "quota.h"
60 #include "inode.h"
61 #include "ufsmount.h"
62 #include "ufs_extern.h"
63 
64 #include "fs.h"
65 #include "ffs_extern.h"
66 
67 static int	ffs_fsync (struct vop_fsync_args *);
68 static int	ffs_getpages (struct vop_getpages_args *);
69 static int	ffs_putpages (struct vop_putpages_args *);
70 static int	ffs_read (struct vop_read_args *);
71 static int	ffs_write (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_putpages_desc,		(vop_t *) ffs_putpages },
80 	{ &vop_read_desc,		(vop_t *) ffs_read },
81 	{ &vop_balloc_desc,		(vop_t *) ffs_balloc },
82 	{ &vop_reallocblks_desc,	(vop_t *) ffs_reallocblks },
83 	{ &vop_write_desc,		(vop_t *) ffs_write },
84 	{ NULL, NULL }
85 };
86 static struct vnodeopv_desc ffs_vnodeop_opv_desc =
87 	{ &ffs_vnodeop_p, ffs_vnodeop_entries };
88 
89 vop_t **ffs_specop_p;
90 static struct vnodeopv_entry_desc ffs_specop_entries[] = {
91 	{ &vop_default_desc,		(vop_t *) ufs_vnoperatespec },
92 	{ &vop_fsync_desc,		(vop_t *) ffs_fsync },
93 	{ NULL, NULL }
94 };
95 static struct vnodeopv_desc ffs_specop_opv_desc =
96 	{ &ffs_specop_p, ffs_specop_entries };
97 
98 vop_t **ffs_fifoop_p;
99 static struct vnodeopv_entry_desc ffs_fifoop_entries[] = {
100 	{ &vop_default_desc,		(vop_t *) ufs_vnoperatefifo },
101 	{ &vop_fsync_desc,		(vop_t *) ffs_fsync },
102 	{ NULL, NULL }
103 };
104 static struct vnodeopv_desc ffs_fifoop_opv_desc =
105 	{ &ffs_fifoop_p, ffs_fifoop_entries };
106 
107 VNODEOP_SET(ffs_vnodeop_opv_desc);
108 VNODEOP_SET(ffs_specop_opv_desc);
109 VNODEOP_SET(ffs_fifoop_opv_desc);
110 
111 #include "ufs_readwrite.c"
112 
113 /*
114  * Synch an open file.
115  */
116 /* ARGSUSED */
117 static int
118 ffs_fsync(ap)
119 	struct vop_fsync_args /* {
120 		struct vnode *a_vp;
121 		struct ucred *a_cred;
122 		int a_waitfor;
123 		struct proc *a_p;
124 	} */ *ap;
125 {
126 	struct vnode *vp = ap->a_vp;
127 	struct buf *bp;
128 	struct buf *nbp;
129 	int s, error, wait, passes, skipmeta;
130 	daddr_t lbn;
131 
132 	wait = (ap->a_waitfor == MNT_WAIT);
133 	if (vn_isdisk(vp, NULL)) {
134 		lbn = INT_MAX;
135 		if (vp->v_specmountpoint != NULL &&
136 		    (vp->v_specmountpoint->mnt_flag & MNT_SOFTDEP))
137 			softdep_fsync_mountdev(vp);
138 	} else {
139 		struct inode *ip;
140 		ip = VTOI(vp);
141 		lbn = lblkno(ip->i_fs, (ip->i_size + ip->i_fs->fs_bsize - 1));
142 	}
143 
144 	/*
145 	 * Flush all dirty buffers associated with a vnode.
146 	 */
147 	passes = NIADDR + 1;
148 	skipmeta = 0;
149 	if (wait)
150 		skipmeta = 1;
151 	s = splbio();
152 loop:
153 	for (bp = TAILQ_FIRST(&vp->v_dirtyblkhd); bp;
154 	     bp = TAILQ_NEXT(bp, b_vnbufs))
155 		bp->b_flags &= ~B_SCANNED;
156 	for (bp = TAILQ_FIRST(&vp->v_dirtyblkhd); bp; bp = nbp) {
157 		nbp = TAILQ_NEXT(bp, b_vnbufs);
158 		/*
159 		 * Reasons to skip this buffer: it has already been considered
160 		 * on this pass, this pass is the first time through on a
161 		 * synchronous flush request and the buffer being considered
162 		 * is metadata, the buffer has dependencies that will cause
163 		 * it to be redirtied and it has not already been deferred,
164 		 * or it is already being written.
165 		 */
166 		if ((bp->b_flags & B_SCANNED) != 0)
167 			continue;
168 		bp->b_flags |= B_SCANNED;
169 		if ((skipmeta == 1 && bp->b_lblkno < 0))
170 			continue;
171 		if (!wait && LIST_FIRST(&bp->b_dep) != NULL &&
172 		    (bp->b_flags & B_DEFERRED) == 0 &&
173 		    bioops.io_countdeps && (*bioops.io_countdeps)(bp, 0)) {
174 			bp->b_flags |= B_DEFERRED;
175 			continue;
176 		}
177 		if (BUF_LOCK(bp, LK_EXCLUSIVE | LK_NOWAIT))
178 			continue;
179 		if ((bp->b_flags & B_DELWRI) == 0)
180 			panic("ffs_fsync: not dirty");
181 		if (vp != bp->b_vp)
182 			panic("ffs_fsync: vp != vp->b_vp");
183 		/*
184 		 * If this is a synchronous flush request, or it is not a
185 		 * file or device, start the write on this buffer immediatly.
186 		 */
187 		if (wait || (vp->v_type != VREG && vp->v_type != VBLK)) {
188 
189 			/*
190 			 * On our final pass through, do all I/O synchronously
191 			 * so that we can find out if our flush is failing
192 			 * because of write errors.
193 			 */
194 			if (passes > 0 || !wait) {
195 				if ((bp->b_flags & B_CLUSTEROK) && !wait) {
196 					BUF_UNLOCK(bp);
197 					(void) vfs_bio_awrite(bp);
198 				} else {
199 					bremfree(bp);
200 					splx(s);
201 					(void) bawrite(bp);
202 					s = splbio();
203 				}
204 			} else {
205 				bremfree(bp);
206 				splx(s);
207 				if ((error = bwrite(bp)) != 0)
208 					return (error);
209 				s = splbio();
210 			}
211 		} else if ((vp->v_type == VREG) && (bp->b_lblkno >= lbn)) {
212 			/*
213 			 * If the buffer is for data that has been truncated
214 			 * off the file, then throw it away.
215 			 */
216 			bremfree(bp);
217 			bp->b_flags |= B_INVAL | B_NOCACHE;
218 			splx(s);
219 			brelse(bp);
220 			s = splbio();
221 		} else {
222 			BUF_UNLOCK(bp);
223 			vfs_bio_awrite(bp);
224 		}
225 		/*
226 		 * Since we may have slept during the I/O, we need
227 		 * to start from a known point.
228 		 */
229 		nbp = TAILQ_FIRST(&vp->v_dirtyblkhd);
230 	}
231 	/*
232 	 * If we were asked to do this synchronously, then go back for
233 	 * another pass, this time doing the metadata.
234 	 */
235 	if (skipmeta) {
236 		skipmeta = 0;
237 		goto loop;
238 	}
239 
240 	if (wait) {
241 		while (vp->v_numoutput) {
242 			vp->v_flag |= VBWAIT;
243 			(void) tsleep((caddr_t)&vp->v_numoutput, 0, "ffsfsn", 0);
244   		}
245 
246 		/*
247 		 * Ensure that any filesystem metatdata associated
248 		 * with the vnode has been written.
249 		 */
250 		splx(s);
251 		if ((error = softdep_sync_metadata(ap)) != 0)
252 			return (error);
253 		s = splbio();
254 
255 		if (!TAILQ_EMPTY(&vp->v_dirtyblkhd)) {
256 			/*
257 			 * Block devices associated with filesystems may
258 			 * have new I/O requests posted for them even if
259 			 * the vnode is locked, so no amount of trying will
260 			 * get them clean. Thus we give block devices a
261 			 * good effort, then just give up. For all other file
262 			 * types, go around and try again until it is clean.
263 			 */
264 			if (passes > 0) {
265 				passes -= 1;
266 				goto loop;
267 			}
268 #ifdef DIAGNOSTIC
269 			if (!vn_isdisk(vp, NULL))
270 				vprint("ffs_fsync: dirty", vp);
271 #endif
272 		}
273 	}
274 	splx(s);
275 	return (UFS_UPDATE(vp, wait));
276 }
277