xref: /freebsd/sys/ufs/ffs/ffs_balloc.c (revision 7f3dea244c40159a41ab22da77a434d7c5b5e85a)
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_balloc.c	8.8 (Berkeley) 6/16/95
34  * $Id: ffs_balloc.c,v 1.22 1999/01/28 00:57:54 dillon Exp $
35  */
36 
37 #include <sys/param.h>
38 #include <sys/systm.h>
39 #include <sys/buf.h>
40 #include <sys/lock.h>
41 #include <sys/mount.h>
42 #include <sys/vnode.h>
43 
44 #include <ufs/ufs/quota.h>
45 #include <ufs/ufs/inode.h>
46 #include <ufs/ufs/ufs_extern.h>
47 
48 #include <ufs/ffs/fs.h>
49 #include <ufs/ffs/ffs_extern.h>
50 
51 /*
52  * Balloc defines the structure of file system storage
53  * by allocating the physical blocks on a device given
54  * the inode and the logical block number in a file.
55  */
56 int
57 ffs_balloc(ap)
58 	struct vop_balloc_args /* {
59 		struct inode *a_ip;
60 		ufs_daddr_t a_lbn;
61 		int a_size;
62 		struct ucred *a_cred;
63 		int a_flags;
64 		struct buf *a_bpp;
65 	} */ *ap;
66 {
67 	register struct inode *ip;
68 	register ufs_daddr_t lbn;
69 	int size;
70 	struct ucred *cred;
71 	int flags;
72 	struct fs *fs;
73 	ufs_daddr_t nb;
74 	struct buf *bp, *nbp;
75 	struct vnode *vp;
76 	struct indir indirs[NIADDR + 2];
77 	ufs_daddr_t newb, *bap, pref;
78 	int deallocated, osize, nsize, num, i, error;
79 	ufs_daddr_t *allocib, *blkp, *allocblk, allociblk[NIADDR + 1];
80 
81 	vp = ap->a_vp;
82 	ip = VTOI(vp);
83 	fs = ip->i_fs;
84 	lbn = lblkno(fs, ap->a_startoffset);
85 	size = blkoff(fs, ap->a_startoffset) + ap->a_size;
86 	if (size > fs->fs_bsize)
87 		panic("ffs_balloc: blk too big");
88 	*ap->a_bpp = NULL;
89 	if (lbn < 0)
90 		return (EFBIG);
91 	cred = ap->a_cred;
92 	flags = ap->a_flags;
93 
94 	/*
95 	 * If the next write will extend the file into a new block,
96 	 * and the file is currently composed of a fragment
97 	 * this fragment has to be extended to be a full block.
98 	 */
99 	nb = lblkno(fs, ip->i_size);
100 	if (nb < NDADDR && nb < lbn) {
101 		osize = blksize(fs, ip, nb);
102 		if (osize < fs->fs_bsize && osize > 0) {
103 			error = ffs_realloccg(ip, nb,
104 				ffs_blkpref(ip, nb, (int)nb, &ip->i_db[0]),
105 				osize, (int)fs->fs_bsize, cred, &bp);
106 			if (error)
107 				return (error);
108 			if (DOINGSOFTDEP(vp))
109 				softdep_setup_allocdirect(ip, nb,
110 				    dbtofsb(fs, bp->b_blkno), ip->i_db[nb],
111 				    fs->fs_bsize, osize, bp);
112 			ip->i_size = smalllblktosize(fs, nb + 1);
113 			ip->i_db[nb] = dbtofsb(fs, bp->b_blkno);
114 			ip->i_flag |= IN_CHANGE | IN_UPDATE;
115 			if (flags & B_SYNC)
116 				bwrite(bp);
117 			else
118 				bawrite(bp);
119 		}
120 	}
121 	/*
122 	 * The first NDADDR blocks are direct blocks
123 	 */
124 	if (lbn < NDADDR) {
125 		nb = ip->i_db[lbn];
126 		if (nb != 0 && ip->i_size >= smalllblktosize(fs, lbn + 1)) {
127 			error = bread(vp, lbn, fs->fs_bsize, NOCRED, &bp);
128 			if (error) {
129 				brelse(bp);
130 				return (error);
131 			}
132 			bp->b_blkno = fsbtodb(fs, nb);
133 			*ap->a_bpp = bp;
134 			return (0);
135 		}
136 		if (nb != 0) {
137 			/*
138 			 * Consider need to reallocate a fragment.
139 			 */
140 			osize = fragroundup(fs, blkoff(fs, ip->i_size));
141 			nsize = fragroundup(fs, size);
142 			if (nsize <= osize) {
143 				error = bread(vp, lbn, osize, NOCRED, &bp);
144 				if (error) {
145 					brelse(bp);
146 					return (error);
147 				}
148 				bp->b_blkno = fsbtodb(fs, nb);
149 			} else {
150 				error = ffs_realloccg(ip, lbn,
151 				    ffs_blkpref(ip, lbn, (int)lbn,
152 					&ip->i_db[0]), osize, nsize, cred, &bp);
153 				if (error)
154 					return (error);
155 				if (DOINGSOFTDEP(vp))
156 					softdep_setup_allocdirect(ip, lbn,
157 					    dbtofsb(fs, bp->b_blkno), nb,
158 					    nsize, osize, bp);
159 			}
160 		} else {
161 			if (ip->i_size < smalllblktosize(fs, lbn + 1))
162 				nsize = fragroundup(fs, size);
163 			else
164 				nsize = fs->fs_bsize;
165 			error = ffs_alloc(ip, lbn,
166 			    ffs_blkpref(ip, lbn, (int)lbn, &ip->i_db[0]),
167 			    nsize, cred, &newb);
168 			if (error)
169 				return (error);
170 			bp = getblk(vp, lbn, nsize, 0, 0);
171 			bp->b_blkno = fsbtodb(fs, newb);
172 			if (flags & B_CLRBUF)
173 				vfs_bio_clrbuf(bp);
174 			if (DOINGSOFTDEP(vp))
175 				softdep_setup_allocdirect(ip, lbn, newb, 0,
176 				    nsize, 0, bp);
177 		}
178 		ip->i_db[lbn] = dbtofsb(fs, bp->b_blkno);
179 		ip->i_flag |= IN_CHANGE | IN_UPDATE;
180 		*ap->a_bpp = bp;
181 		return (0);
182 	}
183 	/*
184 	 * Determine the number of levels of indirection.
185 	 */
186 	pref = 0;
187 	if ((error = ufs_getlbns(vp, lbn, indirs, &num)) != 0)
188 		return(error);
189 #ifdef DIAGNOSTIC
190 	if (num < 1)
191 		panic ("ffs_balloc: ufs_bmaparray returned indirect block");
192 #endif
193 	/*
194 	 * Fetch the first indirect block allocating if necessary.
195 	 */
196 	--num;
197 	nb = ip->i_ib[indirs[0].in_off];
198 	allocib = NULL;
199 	allocblk = allociblk;
200 	if (nb == 0) {
201 		pref = ffs_blkpref(ip, lbn, 0, (ufs_daddr_t *)0);
202 	        if ((error = ffs_alloc(ip, lbn, pref, (int)fs->fs_bsize,
203 		    cred, &newb)) != 0)
204 			return (error);
205 		nb = newb;
206 		*allocblk++ = nb;
207 		bp = getblk(vp, indirs[1].in_lbn, fs->fs_bsize, 0, 0);
208 		bp->b_blkno = fsbtodb(fs, nb);
209 		vfs_bio_clrbuf(bp);
210 		if (DOINGSOFTDEP(vp)) {
211 			softdep_setup_allocdirect(ip, NDADDR + indirs[0].in_off,
212 			    newb, 0, fs->fs_bsize, 0, bp);
213 			bdwrite(bp);
214 		} else {
215 			/*
216 			 * Write synchronously so that indirect blocks
217 			 * never point at garbage.
218 			 */
219 			if (DOINGASYNC(vp))
220 				bdwrite(bp);
221 			else if ((error = bwrite(bp)) != 0)
222 				goto fail;
223 		}
224 		allocib = &ip->i_ib[indirs[0].in_off];
225 		*allocib = nb;
226 		ip->i_flag |= IN_CHANGE | IN_UPDATE;
227 	}
228 	/*
229 	 * Fetch through the indirect blocks, allocating as necessary.
230 	 */
231 	for (i = 1;;) {
232 		error = bread(vp,
233 		    indirs[i].in_lbn, (int)fs->fs_bsize, NOCRED, &bp);
234 		if (error) {
235 			brelse(bp);
236 			goto fail;
237 		}
238 		bap = (ufs_daddr_t *)bp->b_data;
239 		nb = bap[indirs[i].in_off];
240 		if (i == num)
241 			break;
242 		i += 1;
243 		if (nb != 0) {
244 			bqrelse(bp);
245 			continue;
246 		}
247 		if (pref == 0)
248 			pref = ffs_blkpref(ip, lbn, 0, (ufs_daddr_t *)0);
249 		if ((error =
250 		    ffs_alloc(ip, lbn, pref, (int)fs->fs_bsize, cred, &newb)) != 0) {
251 			brelse(bp);
252 			goto fail;
253 		}
254 		nb = newb;
255 		*allocblk++ = nb;
256 		nbp = getblk(vp, indirs[i].in_lbn, fs->fs_bsize, 0, 0);
257 		nbp->b_blkno = fsbtodb(fs, nb);
258 		vfs_bio_clrbuf(nbp);
259 		if (DOINGSOFTDEP(vp)) {
260 			softdep_setup_allocindir_meta(nbp, ip, bp,
261 			    indirs[i - 1].in_off, nb);
262 			bdwrite(nbp);
263 		} else {
264 			/*
265 			 * Write synchronously so that indirect blocks
266 			 * never point at garbage.
267 			 */
268 			if ((error = bwrite(nbp)) != 0) {
269 				brelse(bp);
270 				goto fail;
271 			}
272 		}
273 		bap[indirs[i - 1].in_off] = nb;
274 		/*
275 		 * If required, write synchronously, otherwise use
276 		 * delayed write.
277 		 */
278 		if (flags & B_SYNC) {
279 			bwrite(bp);
280 		} else {
281 			if (bp->b_bufsize == fs->fs_bsize)
282 				bp->b_flags |= B_CLUSTEROK;
283 			bdwrite(bp);
284 		}
285 	}
286 	/*
287 	 * Get the data block, allocating if necessary.
288 	 */
289 	if (nb == 0) {
290 		pref = ffs_blkpref(ip, lbn, indirs[i].in_off, &bap[0]);
291 		error = ffs_alloc(ip,
292 		    lbn, pref, (int)fs->fs_bsize, cred, &newb);
293 		if (error) {
294 			brelse(bp);
295 			goto fail;
296 		}
297 		nb = newb;
298 		*allocblk++ = nb;
299 		nbp = getblk(vp, lbn, fs->fs_bsize, 0, 0);
300 		nbp->b_blkno = fsbtodb(fs, nb);
301 		if (flags & B_CLRBUF)
302 			vfs_bio_clrbuf(nbp);
303 		if (DOINGSOFTDEP(vp))
304 			softdep_setup_allocindir_page(ip, lbn, bp,
305 			    indirs[i].in_off, nb, 0, nbp);
306 		bap[indirs[i].in_off] = nb;
307 		/*
308 		 * If required, write synchronously, otherwise use
309 		 * delayed write.
310 		 */
311 		if (flags & B_SYNC) {
312 			bwrite(bp);
313 		} else {
314 			if (bp->b_bufsize == fs->fs_bsize)
315 				bp->b_flags |= B_CLUSTEROK;
316 			bdwrite(bp);
317 		}
318 		*ap->a_bpp = nbp;
319 		return (0);
320 	}
321 	brelse(bp);
322 	if (flags & B_CLRBUF) {
323 		error = bread(vp, lbn, (int)fs->fs_bsize, NOCRED, &nbp);
324 		if (error) {
325 			brelse(nbp);
326 			goto fail;
327 		}
328 	} else {
329 		nbp = getblk(vp, lbn, fs->fs_bsize, 0, 0);
330 		nbp->b_blkno = fsbtodb(fs, nb);
331 	}
332 	*ap->a_bpp = nbp;
333 	return (0);
334 fail:
335 	/*
336 	 * If we have failed part way through block allocation, we
337 	 * have to deallocate any indirect blocks that we have allocated.
338 	 */
339 	for (deallocated = 0, blkp = allociblk; blkp < allocblk; blkp++) {
340 		ffs_blkfree(ip, *blkp, fs->fs_bsize);
341 		deallocated += fs->fs_bsize;
342 	}
343 	if (allocib != NULL)
344 		*allocib = 0;
345 	if (deallocated) {
346 #ifdef QUOTA
347 		/*
348 		 * Restore user's disk quota because allocation failed.
349 		 */
350 		(void) chkdq(ip, (long)-btodb(deallocated), cred, FORCE);
351 #endif
352 		ip->i_blocks -= btodb(deallocated);
353 		ip->i_flag |= IN_CHANGE | IN_UPDATE;
354 	}
355 	return (error);
356 }
357