xref: /freebsd/sys/ufs/ffs/ffs_vnops.c (revision c17d43407fe04133a94055b0dbc7ea8965654a9f)
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(struct vop_fsync_args *);
69 static int	ffs_getpages(struct vop_getpages_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_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 	wait = (ap->a_waitfor == MNT_WAIT);
144 	if (vn_isdisk(vp, NULL)) {
145 		lbn = INT_MAX;
146 		if (vp->v_rdev->si_mountpoint != NULL &&
147 		    (vp->v_rdev->si_mountpoint->mnt_flag & MNT_SOFTDEP))
148 			softdep_fsync_mountdev(vp);
149 	} else {
150 		lbn = lblkno(ip->i_fs, (ip->i_size + ip->i_fs->fs_bsize - 1));
151 	}
152 
153 	/*
154 	 * Flush all dirty buffers associated with a vnode.
155 	 */
156 	passes = NIADDR + 1;
157 	skipmeta = 0;
158 	if (wait)
159 		skipmeta = 1;
160 	s = splbio();
161 loop:
162 	TAILQ_FOREACH(bp, &vp->v_dirtyblkhd, b_vnbufs)
163 		bp->b_flags &= ~B_SCANNED;
164 	for (bp = TAILQ_FIRST(&vp->v_dirtyblkhd); bp; bp = nbp) {
165 		nbp = TAILQ_NEXT(bp, b_vnbufs);
166 		/*
167 		 * Reasons to skip this buffer: it has already been considered
168 		 * on this pass, this pass is the first time through on a
169 		 * synchronous flush request and the buffer being considered
170 		 * is metadata, the buffer has dependencies that will cause
171 		 * it to be redirtied and it has not already been deferred,
172 		 * or it is already being written.
173 		 */
174 		if ((bp->b_flags & B_SCANNED) != 0)
175 			continue;
176 		bp->b_flags |= B_SCANNED;
177 		if ((skipmeta == 1 && bp->b_lblkno < 0))
178 			continue;
179 		if (!wait && LIST_FIRST(&bp->b_dep) != NULL &&
180 		    (bp->b_flags & B_DEFERRED) == 0 &&
181 		    buf_countdeps(bp, 0)) {
182 			bp->b_flags |= B_DEFERRED;
183 			continue;
184 		}
185 		if (BUF_LOCK(bp, LK_EXCLUSIVE | LK_NOWAIT))
186 			continue;
187 		if ((bp->b_flags & B_DELWRI) == 0)
188 			panic("ffs_fsync: not dirty");
189 		if (vp != bp->b_vp)
190 			panic("ffs_fsync: vp != vp->b_vp");
191 		/*
192 		 * If this is a synchronous flush request, or it is not a
193 		 * file or device, start the write on this buffer immediatly.
194 		 */
195 		if (wait || (vp->v_type != VREG && vp->v_type != VBLK)) {
196 
197 			/*
198 			 * On our final pass through, do all I/O synchronously
199 			 * so that we can find out if our flush is failing
200 			 * because of write errors.
201 			 */
202 			if (passes > 0 || !wait) {
203 				if ((bp->b_flags & B_CLUSTEROK) && !wait) {
204 					BUF_UNLOCK(bp);
205 					(void) vfs_bio_awrite(bp);
206 				} else {
207 					bremfree(bp);
208 					splx(s);
209 					(void) bawrite(bp);
210 					s = splbio();
211 				}
212 			} else {
213 				bremfree(bp);
214 				splx(s);
215 				if ((error = bwrite(bp)) != 0)
216 					return (error);
217 				s = splbio();
218 			}
219 		} else if ((vp->v_type == VREG) && (bp->b_lblkno >= lbn)) {
220 			/*
221 			 * If the buffer is for data that has been truncated
222 			 * off the file, then throw it away.
223 			 */
224 			bremfree(bp);
225 			bp->b_flags |= B_INVAL | B_NOCACHE;
226 			splx(s);
227 			brelse(bp);
228 			s = splbio();
229 		} else {
230 			BUF_UNLOCK(bp);
231 			vfs_bio_awrite(bp);
232 		}
233 		/*
234 		 * Since we may have slept during the I/O, we need
235 		 * to start from a known point.
236 		 */
237 		nbp = TAILQ_FIRST(&vp->v_dirtyblkhd);
238 	}
239 	/*
240 	 * If we were asked to do this synchronously, then go back for
241 	 * another pass, this time doing the metadata.
242 	 */
243 	if (skipmeta) {
244 		skipmeta = 0;
245 		goto loop;
246 	}
247 
248 	if (wait) {
249 		while (vp->v_numoutput) {
250 			vp->v_flag |= VBWAIT;
251 			(void) tsleep((caddr_t)&vp->v_numoutput,
252 					PRIBIO + 4, "ffsfsn", 0);
253   		}
254 
255 		/*
256 		 * Ensure that any filesystem metatdata associated
257 		 * with the vnode has been written.
258 		 */
259 		splx(s);
260 		if ((error = softdep_sync_metadata(ap)) != 0)
261 			return (error);
262 		s = splbio();
263 
264 		if (!TAILQ_EMPTY(&vp->v_dirtyblkhd)) {
265 			/*
266 			 * Block devices associated with filesystems may
267 			 * have new I/O requests posted for them even if
268 			 * the vnode is locked, so no amount of trying will
269 			 * get them clean. Thus we give block devices a
270 			 * good effort, then just give up. For all other file
271 			 * types, go around and try again until it is clean.
272 			 */
273 			if (passes > 0) {
274 				passes -= 1;
275 				goto loop;
276 			}
277 #ifdef DIAGNOSTIC
278 			if (!vn_isdisk(vp, NULL))
279 				vprint("ffs_fsync: dirty", vp);
280 #endif
281 		}
282 	}
283 	splx(s);
284 	return (UFS_UPDATE(vp, wait));
285 }
286