xref: /freebsd/sys/ufs/ffs/ffs_balloc.c (revision b601c69bdbe8755d26570261d7fd4c02ee4eff74)
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  * $FreeBSD$
35  */
36 
37 #include <sys/param.h>
38 #include <sys/systm.h>
39 #include <sys/bio.h>
40 #include <sys/buf.h>
41 #include <sys/lock.h>
42 #include <sys/mount.h>
43 #include <sys/vnode.h>
44 
45 #include <ufs/ufs/quota.h>
46 #include <ufs/ufs/inode.h>
47 #include <ufs/ufs/ufs_extern.h>
48 
49 #include <ufs/ffs/fs.h>
50 #include <ufs/ffs/ffs_extern.h>
51 
52 /*
53  * Balloc defines the structure of file system storage
54  * by allocating the physical blocks on a device given
55  * the inode and the logical block number in a file.
56  */
57 int
58 ffs_balloc(ap)
59 	struct vop_balloc_args /* {
60 		struct vnode *a_vp;
61 		ufs_daddr_t a_lbn;
62 		int a_size;
63 		struct ucred *a_cred;
64 		int a_flags;
65 		struct buf *a_bpp;
66 	} */ *ap;
67 {
68 	struct inode *ip;
69 	ufs_daddr_t lbn;
70 	int size;
71 	struct ucred *cred;
72 	int flags;
73 	struct fs *fs;
74 	ufs_daddr_t nb;
75 	struct buf *bp, *nbp;
76 	struct vnode *vp;
77 	struct indir indirs[NIADDR + 2];
78 	ufs_daddr_t newb, *bap, pref;
79 	int deallocated, osize, nsize, num, i, error;
80 	ufs_daddr_t *allocib, *blkp, *allocblk, allociblk[NIADDR + 1];
81 	int unwindidx = -1;
82 	struct proc *p = curproc;	/* XXX */
83 
84 	vp = ap->a_vp;
85 	ip = VTOI(vp);
86 	fs = ip->i_fs;
87 	lbn = lblkno(fs, ap->a_startoffset);
88 	size = blkoff(fs, ap->a_startoffset) + ap->a_size;
89 	if (size > fs->fs_bsize)
90 		panic("ffs_balloc: blk too big");
91 	*ap->a_bpp = NULL;
92 	if (lbn < 0)
93 		return (EFBIG);
94 	cred = ap->a_cred;
95 	flags = ap->a_flags;
96 
97 	/*
98 	 * If the next write will extend the file into a new block,
99 	 * and the file is currently composed of a fragment
100 	 * this fragment has to be extended to be a full block.
101 	 */
102 	nb = lblkno(fs, ip->i_size);
103 	if (nb < NDADDR && nb < lbn) {
104 		osize = blksize(fs, ip, nb);
105 		if (osize < fs->fs_bsize && osize > 0) {
106 			error = ffs_realloccg(ip, nb,
107 				ffs_blkpref(ip, nb, (int)nb, &ip->i_db[0]),
108 				osize, (int)fs->fs_bsize, cred, &bp);
109 			if (error)
110 				return (error);
111 			if (DOINGSOFTDEP(vp))
112 				softdep_setup_allocdirect(ip, nb,
113 				    dbtofsb(fs, bp->b_blkno), ip->i_db[nb],
114 				    fs->fs_bsize, osize, bp);
115 			ip->i_size = smalllblktosize(fs, nb + 1);
116 			ip->i_db[nb] = dbtofsb(fs, bp->b_blkno);
117 			ip->i_flag |= IN_CHANGE | IN_UPDATE;
118 			if (flags & B_SYNC)
119 				bwrite(bp);
120 			else
121 				bawrite(bp);
122 		}
123 	}
124 	/*
125 	 * The first NDADDR blocks are direct blocks
126 	 */
127 	if (lbn < NDADDR) {
128 		if (flags & B_METAONLY)
129 			panic("ffs_balloc: B_METAONLY for direct block");
130 		nb = ip->i_db[lbn];
131 		if (nb != 0 && ip->i_size >= smalllblktosize(fs, lbn + 1)) {
132 			error = bread(vp, lbn, fs->fs_bsize, NOCRED, &bp);
133 			if (error) {
134 				brelse(bp);
135 				return (error);
136 			}
137 			bp->b_blkno = fsbtodb(fs, nb);
138 			*ap->a_bpp = bp;
139 			return (0);
140 		}
141 		if (nb != 0) {
142 			/*
143 			 * Consider need to reallocate a fragment.
144 			 */
145 			osize = fragroundup(fs, blkoff(fs, ip->i_size));
146 			nsize = fragroundup(fs, size);
147 			if (nsize <= osize) {
148 				error = bread(vp, lbn, osize, NOCRED, &bp);
149 				if (error) {
150 					brelse(bp);
151 					return (error);
152 				}
153 				bp->b_blkno = fsbtodb(fs, nb);
154 			} else {
155 				error = ffs_realloccg(ip, lbn,
156 				    ffs_blkpref(ip, lbn, (int)lbn,
157 					&ip->i_db[0]), osize, nsize, cred, &bp);
158 				if (error)
159 					return (error);
160 				if (DOINGSOFTDEP(vp))
161 					softdep_setup_allocdirect(ip, lbn,
162 					    dbtofsb(fs, bp->b_blkno), nb,
163 					    nsize, osize, bp);
164 			}
165 		} else {
166 			if (ip->i_size < smalllblktosize(fs, lbn + 1))
167 				nsize = fragroundup(fs, size);
168 			else
169 				nsize = fs->fs_bsize;
170 			error = ffs_alloc(ip, lbn,
171 			    ffs_blkpref(ip, lbn, (int)lbn, &ip->i_db[0]),
172 			    nsize, cred, &newb);
173 			if (error)
174 				return (error);
175 			bp = getblk(vp, lbn, nsize, 0, 0);
176 			bp->b_blkno = fsbtodb(fs, newb);
177 			if (flags & B_CLRBUF)
178 				vfs_bio_clrbuf(bp);
179 			if (DOINGSOFTDEP(vp))
180 				softdep_setup_allocdirect(ip, lbn, newb, 0,
181 				    nsize, 0, bp);
182 		}
183 		ip->i_db[lbn] = dbtofsb(fs, bp->b_blkno);
184 		ip->i_flag |= IN_CHANGE | IN_UPDATE;
185 		*ap->a_bpp = bp;
186 		return (0);
187 	}
188 	/*
189 	 * Determine the number of levels of indirection.
190 	 */
191 	pref = 0;
192 	if ((error = ufs_getlbns(vp, lbn, indirs, &num)) != 0)
193 		return(error);
194 #ifdef DIAGNOSTIC
195 	if (num < 1)
196 		panic ("ffs_balloc: ufs_bmaparray returned indirect block");
197 #endif
198 	/*
199 	 * Fetch the first indirect block allocating if necessary.
200 	 */
201 	--num;
202 	nb = ip->i_ib[indirs[0].in_off];
203 	allocib = NULL;
204 	allocblk = allociblk;
205 	if (nb == 0) {
206 		pref = ffs_blkpref(ip, lbn, 0, (ufs_daddr_t *)0);
207 	        if ((error = ffs_alloc(ip, lbn, pref, (int)fs->fs_bsize,
208 		    cred, &newb)) != 0)
209 			return (error);
210 		nb = newb;
211 		*allocblk++ = nb;
212 		bp = getblk(vp, indirs[1].in_lbn, fs->fs_bsize, 0, 0);
213 		bp->b_blkno = fsbtodb(fs, nb);
214 		vfs_bio_clrbuf(bp);
215 		if (DOINGSOFTDEP(vp)) {
216 			softdep_setup_allocdirect(ip, NDADDR + indirs[0].in_off,
217 			    newb, 0, fs->fs_bsize, 0, bp);
218 			bdwrite(bp);
219 		} else {
220 			/*
221 			 * Write synchronously so that indirect blocks
222 			 * never point at garbage.
223 			 */
224 			if (DOINGASYNC(vp))
225 				bdwrite(bp);
226 			else if ((error = bwrite(bp)) != 0)
227 				goto fail;
228 		}
229 		allocib = &ip->i_ib[indirs[0].in_off];
230 		*allocib = nb;
231 		ip->i_flag |= IN_CHANGE | IN_UPDATE;
232 	}
233 	/*
234 	 * Fetch through the indirect blocks, allocating as necessary.
235 	 */
236 	for (i = 1;;) {
237 		error = bread(vp,
238 		    indirs[i].in_lbn, (int)fs->fs_bsize, NOCRED, &bp);
239 		if (error) {
240 			brelse(bp);
241 			goto fail;
242 		}
243 		bap = (ufs_daddr_t *)bp->b_data;
244 		nb = bap[indirs[i].in_off];
245 		if (i == num)
246 			break;
247 		i += 1;
248 		if (nb != 0) {
249 			bqrelse(bp);
250 			continue;
251 		}
252 		if (pref == 0)
253 			pref = ffs_blkpref(ip, lbn, 0, (ufs_daddr_t *)0);
254 		if ((error =
255 		    ffs_alloc(ip, lbn, pref, (int)fs->fs_bsize, cred, &newb)) != 0) {
256 			brelse(bp);
257 			goto fail;
258 		}
259 		nb = newb;
260 		*allocblk++ = nb;
261 		nbp = getblk(vp, indirs[i].in_lbn, fs->fs_bsize, 0, 0);
262 		nbp->b_blkno = fsbtodb(fs, nb);
263 		vfs_bio_clrbuf(nbp);
264 		if (DOINGSOFTDEP(vp)) {
265 			softdep_setup_allocindir_meta(nbp, ip, bp,
266 			    indirs[i - 1].in_off, nb);
267 			bdwrite(nbp);
268 		} else {
269 			/*
270 			 * Write synchronously so that indirect blocks
271 			 * never point at garbage.
272 			 */
273 			if ((error = bwrite(nbp)) != 0) {
274 				brelse(bp);
275 				goto fail;
276 			}
277 		}
278 		bap[indirs[i - 1].in_off] = nb;
279 		if (allocib == NULL && unwindidx < 0)
280 			unwindidx = i - 1;
281 		/*
282 		 * If required, write synchronously, otherwise use
283 		 * delayed write.
284 		 */
285 		if (flags & B_SYNC) {
286 			bwrite(bp);
287 		} else {
288 			if (bp->b_bufsize == fs->fs_bsize)
289 				bp->b_flags |= B_CLUSTEROK;
290 			bdwrite(bp);
291 		}
292 	}
293 	/*
294 	 * If asked only for the indirect block, then return it.
295 	 */
296 	if (flags & B_METAONLY) {
297 		*ap->a_bpp = bp;
298 		return (0);
299 	}
300 	/*
301 	 * Get the data block, allocating if necessary.
302 	 */
303 	if (nb == 0) {
304 		pref = ffs_blkpref(ip, lbn, indirs[i].in_off, &bap[0]);
305 		error = ffs_alloc(ip,
306 		    lbn, pref, (int)fs->fs_bsize, cred, &newb);
307 		if (error) {
308 			brelse(bp);
309 			goto fail;
310 		}
311 		nb = newb;
312 		*allocblk++ = nb;
313 		nbp = getblk(vp, lbn, fs->fs_bsize, 0, 0);
314 		nbp->b_blkno = fsbtodb(fs, nb);
315 		if (flags & B_CLRBUF)
316 			vfs_bio_clrbuf(nbp);
317 		if (DOINGSOFTDEP(vp))
318 			softdep_setup_allocindir_page(ip, lbn, bp,
319 			    indirs[i].in_off, nb, 0, nbp);
320 		bap[indirs[i].in_off] = nb;
321 		/*
322 		 * If required, write synchronously, otherwise use
323 		 * delayed write.
324 		 */
325 		if (flags & B_SYNC) {
326 			bwrite(bp);
327 		} else {
328 			if (bp->b_bufsize == fs->fs_bsize)
329 				bp->b_flags |= B_CLUSTEROK;
330 			bdwrite(bp);
331 		}
332 		*ap->a_bpp = nbp;
333 		return (0);
334 	}
335 	brelse(bp);
336 	if (flags & B_CLRBUF) {
337 		error = bread(vp, lbn, (int)fs->fs_bsize, NOCRED, &nbp);
338 		if (error) {
339 			brelse(nbp);
340 			goto fail;
341 		}
342 	} else {
343 		nbp = getblk(vp, lbn, fs->fs_bsize, 0, 0);
344 		nbp->b_blkno = fsbtodb(fs, nb);
345 	}
346 	*ap->a_bpp = nbp;
347 	return (0);
348 fail:
349 	/*
350 	 * If we have failed part way through block allocation, we
351 	 * have to deallocate any indirect blocks that we have allocated.
352 	 * We have to fsync the file before we start to get rid of all
353 	 * of its dependencies so that we do not leave them dangling.
354 	 * We have to sync it at the end so that the soft updates code
355 	 * does not find any untracked changes. Although this is really
356 	 * slow, running out of disk space is not expected to be a common
357 	 * occurence. The error return from fsync is ignored as we already
358 	 * have an error to return to the user.
359 	 */
360 	(void) VOP_FSYNC(vp, cred, MNT_WAIT, p);
361 	for (deallocated = 0, blkp = allociblk; blkp < allocblk; blkp++) {
362 		ffs_blkfree(ip, *blkp, fs->fs_bsize);
363 		deallocated += fs->fs_bsize;
364 	}
365 	if (allocib != NULL) {
366 		*allocib = 0;
367 	} else if (unwindidx >= 0) {
368 		int r;
369 
370 		r = bread(vp, indirs[unwindidx].in_lbn,
371 		    (int)fs->fs_bsize, NOCRED, &bp);
372 		if (r) {
373 			panic("Could not unwind indirect block, error %d", r);
374 			brelse(bp);
375 		} else {
376 			bap = (ufs_daddr_t *)bp->b_data;
377 			bap[indirs[unwindidx].in_off] = 0;
378 			if (flags & B_SYNC) {
379 				bwrite(bp);
380 			} else {
381 				if (bp->b_bufsize == fs->fs_bsize)
382 					bp->b_flags |= B_CLUSTEROK;
383 				bdwrite(bp);
384 			}
385 		}
386 	}
387 	if (deallocated) {
388 #ifdef QUOTA
389 		/*
390 		 * Restore user's disk quota because allocation failed.
391 		 */
392 		(void) chkdq(ip, (long)-btodb(deallocated), cred, FORCE);
393 #endif
394 		ip->i_blocks -= btodb(deallocated);
395 		ip->i_flag |= IN_CHANGE | IN_UPDATE;
396 	}
397 	(void) VOP_FSYNC(vp, cred, MNT_WAIT, p);
398 	return (error);
399 }
400