1 /*-
2 * modified for Lites 1.1
3 *
4 * Aug 1995, Godmar Back (gback@cs.utah.edu)
5 * University of Utah, Department of Computer Science
6 */
7 /*-
8 * SPDX-License-Identifier: BSD-3-Clause
9 *
10 * Copyright (c) 1982, 1986, 1989, 1993
11 * The Regents of the University of California. All rights reserved.
12 *
13 * Redistribution and use in source and binary forms, with or without
14 * modification, are permitted provided that the following conditions
15 * are met:
16 * 1. Redistributions of source code must retain the above copyright
17 * notice, this list of conditions and the following disclaimer.
18 * 2. Redistributions in binary form must reproduce the above copyright
19 * notice, this list of conditions and the following disclaimer in the
20 * documentation and/or other materials provided with the distribution.
21 * 3. Neither the name of the University nor the names of its contributors
22 * may be used to endorse or promote products derived from this software
23 * without specific prior written permission.
24 *
25 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
26 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
27 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
28 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
29 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
30 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
31 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
32 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
33 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
34 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
35 * SUCH DAMAGE.
36 *
37 * @(#)ffs_alloc.c 8.8 (Berkeley) 2/21/94
38 * $FreeBSD$
39 */
40
41 #include <sys/param.h>
42 #include <sys/systm.h>
43 #include <sys/conf.h>
44 #include <sys/vnode.h>
45 #include <sys/stat.h>
46 #include <sys/mount.h>
47 #include <sys/sysctl.h>
48 #include <sys/syslog.h>
49 #include <sys/buf2.h>
50 #include <sys/endian.h>
51 #include <sys/malloc.h>
52 #include <sys/mutex2.h>
53
54 #include <vfs/ext2fs/fs.h>
55 #include <vfs/ext2fs/inode.h>
56 #include <vfs/ext2fs/ext2_mount.h>
57 #include <vfs/ext2fs/ext2fs.h>
58 #include <vfs/ext2fs/ext2_extern.h>
59
60 SDT_PROVIDER_DEFINE(ext2fs);
61 /*
62 * ext2fs trace probe:
63 * arg0: verbosity. Higher numbers give more verbose messages
64 * arg1: Textual message
65 */
66 SDT_PROBE_DEFINE2(ext2fs, , alloc, trace, "int", "char*");
67 SDT_PROBE_DEFINE3(ext2fs, , alloc, ext2_reallocblks_realloc,
68 "ino_t", "e2fs_lbn_t", "e2fs_lbn_t");
69 SDT_PROBE_DEFINE1(ext2fs, , alloc, ext2_reallocblks_bap, "uint32_t");
70 SDT_PROBE_DEFINE1(ext2fs, , alloc, ext2_reallocblks_blkno, "e2fs_daddr_t");
71 SDT_PROBE_DEFINE2(ext2fs, , alloc, ext2_b_bitmap_validate_error, "char*", "int");
72 SDT_PROBE_DEFINE3(ext2fs, , alloc, ext2_nodealloccg_bmap_corrupted,
73 "int", "daddr_t", "char*");
74 SDT_PROBE_DEFINE2(ext2fs, , alloc, ext2_blkfree_bad_block, "ino_t", "e4fs_daddr_t");
75 SDT_PROBE_DEFINE2(ext2fs, , alloc, ext2_vfree_doublefree, "char*", "ino_t");
76
77 static daddr_t ext2_alloccg(struct inode *, int, daddr_t, int);
78 static daddr_t ext2_clusteralloc(struct inode *, int, daddr_t, int);
79 static u_long ext2_dirpref(struct inode *);
80 static e4fs_daddr_t ext2_hashalloc(struct inode *, int, long, int,
81 daddr_t (*)(struct inode *, int, daddr_t,
82 int));
83 static daddr_t ext2_nodealloccg(struct inode *, int, daddr_t, int);
84 static daddr_t ext2_mapsearch(struct m_ext2fs *, char *, daddr_t);
85
86 /*
87 * Allocate a block in the filesystem.
88 *
89 * A preference may be optionally specified. If a preference is given
90 * the following hierarchy is used to allocate a block:
91 * 1) allocate the requested block.
92 * 2) allocate a rotationally optimal block in the same cylinder.
93 * 3) allocate a block in the same cylinder group.
94 * 4) quadratically rehash into other cylinder groups, until an
95 * available block is located.
96 * If no block preference is given the following hierarchy is used
97 * to allocate a block:
98 * 1) allocate a block in the cylinder group that contains the
99 * inode for the file.
100 * 2) quadratically rehash into other cylinder groups, until an
101 * available block is located.
102 */
103 int
ext2_alloc(struct inode * ip,daddr_t lbn,e4fs_daddr_t bpref,int size,struct ucred * cred,e4fs_daddr_t * bnp)104 ext2_alloc(struct inode *ip, daddr_t lbn, e4fs_daddr_t bpref, int size,
105 struct ucred *cred, e4fs_daddr_t *bnp)
106 {
107 struct m_ext2fs *fs;
108 struct ext2mount *ump;
109 e4fs_daddr_t bno;
110 int cg;
111
112 *bnp = 0;
113 fs = ip->i_e2fs;
114 ump = ip->i_ump;
115 mtx_assert(EXT2_MTX(ump), MA_OWNED);
116 #ifdef INVARIANTS
117 if ((u_int)size > fs->e2fs_bsize || blkoff(fs, size) != 0) {
118 printf("bsize = %lu, size = %d, fs = %s\n",
119 (long unsigned int)fs->e2fs_bsize, size, fs->e2fs_fsmnt);
120 panic("ext2_alloc: bad size");
121 }
122 if (cred == NOCRED)
123 panic("ext2_alloc: missing credential");
124 #endif /* INVARIANTS */
125 if (size == fs->e2fs_bsize && fs->e2fs_fbcount == 0)
126 goto nospace;
127 if (cred->cr_uid != 0 &&
128 fs->e2fs_fbcount < fs->e2fs_rbcount)
129 goto nospace;
130 if (bpref >= fs->e2fs_bcount)
131 bpref = 0;
132 if (bpref == 0)
133 cg = ino_to_cg(fs, ip->i_number);
134 else
135 cg = dtog(fs, bpref);
136 bno = (daddr_t)ext2_hashalloc(ip, cg, bpref, fs->e2fs_bsize,
137 ext2_alloccg);
138 if (bno > 0) {
139 /* set next_alloc fields as done in block_getblk */
140 ip->i_next_alloc_block = lbn;
141 ip->i_next_alloc_goal = bno;
142
143 ip->i_blocks += btodb(fs->e2fs_bsize);
144 ip->i_flag |= IN_CHANGE | IN_UPDATE;
145 *bnp = bno;
146 return (0);
147 }
148 nospace:
149 EXT2_UNLOCK(ump);
150 SDT_PROBE2(ext2fs, , alloc, trace, 1, "cannot allocate data block");
151 return (ENOSPC);
152 }
153
154 /*
155 * Allocate EA's block for inode.
156 */
157 e4fs_daddr_t
ext2_alloc_meta(struct inode * ip)158 ext2_alloc_meta(struct inode *ip)
159 {
160 struct m_ext2fs *fs;
161 daddr_t blk;
162
163 fs = ip->i_e2fs;
164
165 EXT2_LOCK(ip->i_ump);
166 blk = ext2_hashalloc(ip, ino_to_cg(fs, ip->i_number), 0, fs->e2fs_bsize,
167 ext2_alloccg);
168 if (0 == blk) {
169 EXT2_UNLOCK(ip->i_ump);
170 SDT_PROBE2(ext2fs, , alloc, trace, 1, "cannot allocate meta block");
171 }
172
173 return (blk);
174 }
175
176 /*
177 * Reallocate a sequence of blocks into a contiguous sequence of blocks.
178 *
179 * The vnode and an array of buffer pointers for a range of sequential
180 * logical blocks to be made contiguous is given. The allocator attempts
181 * to find a range of sequential blocks starting as close as possible to
182 * an fs_rotdelay offset from the end of the allocation for the logical
183 * block immediately preceding the current range. If successful, the
184 * physical block numbers in the buffer pointers and in the inode are
185 * changed to reflect the new allocation. If unsuccessful, the allocation
186 * is left unchanged. The success in doing the reallocation is returned.
187 * Note that the error return is not reflected back to the user. Rather
188 * the previous block allocation will be used.
189 */
190
191 static SYSCTL_NODE(_vfs, OID_AUTO, ext2fs, CTLFLAG_RW, 0, "EXT2FS filesystem");
192
193 static int doasyncfree = 1;
194
195 SYSCTL_INT(_vfs_ext2fs, OID_AUTO, doasyncfree, CTLFLAG_RW, &doasyncfree, 0,
196 "Use asynchronous writes to update block pointers when freeing blocks");
197
198 static int doreallocblks = 0;
199
200 SYSCTL_INT(_vfs_ext2fs, OID_AUTO, doreallocblks, CTLFLAG_RW, &doreallocblks, 0, "");
201
202 int
ext2_reallocblks(struct vop_reallocblks_args * ap)203 ext2_reallocblks(struct vop_reallocblks_args *ap)
204 {
205 struct m_ext2fs *fs;
206 struct inode *ip;
207 struct vnode *vp;
208 struct buf *sbp, *ebp;
209 uint32_t *bap, *sbap, *ebap;
210 struct ext2mount *ump;
211 struct cluster_save *buflist;
212 struct indir start_ap[EXT2_NIADDR + 1], end_ap[EXT2_NIADDR + 1], *idp;
213 e2fs_lbn_t start_lbn, end_lbn;
214 int soff;
215 e2fs_daddr_t newblk, blkno;
216 int i, len, start_lvl, end_lvl, pref, ssize;
217
218 if (doreallocblks == 0)
219 return (ENOSPC);
220
221 vp = ap->a_vp;
222 ip = VTOI(vp);
223 fs = ip->i_e2fs;
224 ump = ip->i_ump;
225
226 if (fs->e2fs_contigsumsize <= 0 || ip->i_flag & IN_E4EXTENTS)
227 return (ENOSPC);
228
229 buflist = ap->a_buflist;
230 len = buflist->bs_nchildren;
231 start_lbn = lblkno(fs, buflist->bs_children[0]->b_loffset);
232 end_lbn = start_lbn + len - 1;
233 #ifdef INVARIANTS
234 for (i = 1; i < len; i++)
235 if (buflist->bs_children[i]->b_loffset != lblktodoff(fs, start_lbn + i))
236 panic("ext2_reallocblks: non-cluster");
237 #endif
238 /*
239 * If the cluster crosses the boundary for the first indirect
240 * block, leave space for the indirect block. Indirect blocks
241 * are initially laid out in a position after the last direct
242 * block. Block reallocation would usually destroy locality by
243 * moving the indirect block out of the way to make room for
244 * data blocks if we didn't compensate here. We should also do
245 * this for other indirect block boundaries, but it is only
246 * important for the first one.
247 */
248 if (start_lbn < EXT2_NDADDR && end_lbn >= EXT2_NDADDR)
249 return (ENOSPC);
250 /*
251 * If the latest allocation is in a new cylinder group, assume that
252 * the filesystem has decided to move and do not force it back to
253 * the previous cylinder group.
254 */
255 if (dtog(fs, dofftofsb(fs, buflist->bs_children[0]->b_bio2.bio_offset)) !=
256 dtog(fs, dofftofsb(fs, buflist->bs_children[len - 1]->b_bio2.bio_offset)))
257 return (ENOSPC);
258 if (ext2_getlbns(vp, start_lbn, start_ap, &start_lvl) ||
259 ext2_getlbns(vp, end_lbn, end_ap, &end_lvl))
260 return (ENOSPC);
261 /*
262 * Get the starting offset and block map for the first block.
263 */
264 if (start_lvl == 0) {
265 sbap = &ip->i_db[0];
266 soff = start_lbn;
267 } else {
268 idp = &start_ap[start_lvl - 1];
269 if (bread(vp, lblktodoff(fs, idp->in_lbn), (int)fs->e2fs_bsize,
270 &sbp)) {
271 brelse(sbp);
272 return (ENOSPC);
273 }
274 sbap = (u_int *)sbp->b_data;
275 soff = idp->in_off;
276 }
277 /*
278 * If the block range spans two block maps, get the second map.
279 */
280 ebap = NULL;
281 if (end_lvl == 0 || (idp = &end_ap[end_lvl - 1])->in_off + 1 >= len) {
282 ssize = len;
283 } else {
284 #ifdef INVARIANTS
285 if (start_ap[start_lvl - 1].in_lbn == idp->in_lbn)
286 panic("ext2_reallocblks: start == end");
287 #endif
288 ssize = len - (idp->in_off + 1);
289 if (bread(vp, lblktodoff(fs, idp->in_lbn), (int)fs->e2fs_bsize,
290 &ebp))
291 goto fail;
292 ebap = (u_int *)ebp->b_data;
293 }
294 /*
295 * Find the preferred location for the cluster.
296 */
297 EXT2_LOCK(ump);
298 pref = ext2_blkpref(ip, start_lbn, soff, sbap, 0);
299 /*
300 * Search the block map looking for an allocation of the desired size.
301 */
302 if ((newblk = (e2fs_daddr_t)ext2_hashalloc(ip, dtog(fs, pref), pref,
303 len, ext2_clusteralloc)) == 0) {
304 EXT2_UNLOCK(ump);
305 goto fail;
306 }
307 /*
308 * We have found a new contiguous block.
309 *
310 * First we have to replace the old block pointers with the new
311 * block pointers in the inode and indirect blocks associated
312 * with the file.
313 */
314 SDT_PROBE3(ext2fs, , alloc, ext2_reallocblks_realloc,
315 ip->i_number, start_lbn, end_lbn);
316 blkno = newblk;
317 for (bap = &sbap[soff], i = 0; i < len; i++, blkno += fs->e2fs_fpb) {
318 if (i == ssize) {
319 bap = ebap;
320 soff = -i;
321 }
322 #ifdef INVARIANTS
323 if (buflist->bs_children[i]->b_bio2.bio_offset !=
324 fsbtodoff(fs, *bap))
325 panic("ext2_reallocblks: alloc mismatch");
326 #endif
327 SDT_PROBE1(ext2fs, , alloc, ext2_reallocblks_bap, *bap);
328 *bap++ = blkno;
329 }
330 /*
331 * Next we must write out the modified inode and indirect blocks.
332 * For strict correctness, the writes should be synchronous since
333 * the old block values may have been written to disk. In practise
334 * they are almost never written, but if we are concerned about
335 * strict correctness, the `doasyncfree' flag should be set to zero.
336 *
337 * The test on `doasyncfree' should be changed to test a flag
338 * that shows whether the associated buffers and inodes have
339 * been written. The flag should be set when the cluster is
340 * started and cleared whenever the buffer or inode is flushed.
341 * We can then check below to see if it is set, and do the
342 * synchronous write only when it has been cleared.
343 */
344 if (sbap != &ip->i_db[0]) {
345 if (doasyncfree)
346 bdwrite(sbp);
347 else
348 bwrite(sbp);
349 } else {
350 ip->i_flag |= IN_CHANGE | IN_UPDATE;
351 if (!doasyncfree)
352 ext2_update(vp, 1);
353 }
354 if (ssize < len) {
355 if (doasyncfree)
356 bdwrite(ebp);
357 else
358 bwrite(ebp);
359 }
360 /*
361 * Last, free the old blocks and assign the new blocks to the buffers.
362 */
363 for (blkno = newblk, i = 0; i < len; i++, blkno += fs->e2fs_fpb) {
364 ext2_blkfree(ip, dofftofsb(fs, buflist->bs_children[i]->b_bio2.bio_offset),
365 fs->e2fs_bsize);
366 buflist->bs_children[i]->b_bio2.bio_offset = fsbtodoff(fs, blkno);
367 SDT_PROBE1(ext2fs, , alloc, ext2_reallocblks_blkno, blkno);
368 }
369
370 return (0);
371
372 fail:
373 if (ssize < len)
374 brelse(ebp);
375 if (sbap != &ip->i_db[0])
376 brelse(sbp);
377 return (ENOSPC);
378 }
379
380 /*
381 * Allocate an inode in the filesystem.
382 *
383 */
384 int
ext2_valloc(struct vnode * pvp,int mode,struct ucred * cred,struct vnode ** vpp)385 ext2_valloc(struct vnode *pvp, int mode, struct ucred *cred, struct vnode **vpp)
386 {
387 struct timespec ts;
388 struct m_ext2fs *fs;
389 struct ext2mount *ump;
390 struct inode *pip;
391 struct inode *ip;
392 struct vnode *vp;
393 ino_t ino, ipref;
394 int error, cg;
395
396 *vpp = NULL;
397 pip = VTOI(pvp);
398 fs = pip->i_e2fs;
399 ump = pip->i_ump;
400
401 EXT2_LOCK(ump);
402 if (fs->e2fs_ficount == 0)
403 goto noinodes;
404 /*
405 * If it is a directory then obtain a cylinder group based on
406 * ext2_dirpref else obtain it using ino_to_cg. The preferred inode is
407 * always the next inode.
408 */
409 if ((mode & IFMT) == IFDIR) {
410 cg = ext2_dirpref(pip);
411 if (fs->e2fs_contigdirs[cg] < 255)
412 fs->e2fs_contigdirs[cg]++;
413 } else {
414 cg = ino_to_cg(fs, pip->i_number);
415 if (fs->e2fs_contigdirs[cg] > 0)
416 fs->e2fs_contigdirs[cg]--;
417 }
418 ipref = cg * fs->e2fs_ipg + 1;
419 ino = (ino_t)ext2_hashalloc(pip, cg, (long)ipref, mode, ext2_nodealloccg);
420 if (ino == 0)
421 goto noinodes;
422 restart:
423 if ((vp = ext2_ihashget(ump->um_dev, ino)) != NULL) {
424 printf("ext2_valloc: vp %p exists for inode %lu\n", vp, ino);
425 return (EEXIST);
426 }
427 if (ext2_alloc_vnode(ump->um_mountp, ino, &vp) == -1)
428 goto restart;
429 ip = VTOI(vp);
430
431 if ((error = ext2_vinit(vp->v_mount, &vp)) != 0) {
432 *vpp = NULL;
433 vp->v_type = VBAD;
434 vx_put(vp);
435 return (error);
436 }
437
438 if (EXT2_HAS_INCOMPAT_FEATURE(fs, EXT2F_INCOMPAT_EXTENTS)
439 && (S_ISREG(mode) || S_ISDIR(mode)))
440 ext4_ext_tree_init(ip);
441 else
442 memset(ip->i_data, 0, sizeof(ip->i_data));
443
444 /*
445 * Set up a new generation number for this inode.
446 * Avoid zero values.
447 */
448 do {
449 ip->i_gen = karc4random();
450 } while (ip->i_gen == 0);
451
452 vfs_timestamp(&ts);
453 ip->i_birthtime = ts.tv_sec;
454 ip->i_birthnsec = ts.tv_nsec;
455
456 /*
457 * Finish inode initialization now that aliasing has been resolved.
458 */
459 vref(ip->i_devvp);
460 /*
461 * Return the locked and refd vnode.
462 */
463 vx_downgrade(vp); /* downgrade VX lock to VN lock */
464 *vpp = vp;
465
466 return (0);
467
468 noinodes:
469 EXT2_UNLOCK(ump);
470 SDT_PROBE2(ext2fs, , alloc, trace, 1, "out of inodes");
471 return (ENOSPC);
472 }
473
474 /*
475 * 64-bit compatible getters and setters for struct ext2_gd from ext2fs.h
476 */
477 uint64_t
e2fs_gd_get_b_bitmap(struct ext2_gd * gd)478 e2fs_gd_get_b_bitmap(struct ext2_gd *gd)
479 {
480
481 return (((uint64_t)(le32toh(gd->ext4bgd_b_bitmap_hi)) << 32) |
482 le32toh(gd->ext2bgd_b_bitmap));
483 }
484
485 uint64_t
e2fs_gd_get_i_bitmap(struct ext2_gd * gd)486 e2fs_gd_get_i_bitmap(struct ext2_gd *gd)
487 {
488
489 return (((uint64_t)(le32toh(gd->ext4bgd_i_bitmap_hi)) << 32) |
490 le32toh(gd->ext2bgd_i_bitmap));
491 }
492
493 uint64_t
e2fs_gd_get_i_tables(struct ext2_gd * gd)494 e2fs_gd_get_i_tables(struct ext2_gd *gd)
495 {
496
497 return (((uint64_t)(le32toh(gd->ext4bgd_i_tables_hi)) << 32) |
498 le32toh(gd->ext2bgd_i_tables));
499 }
500
501 static uint32_t
e2fs_gd_get_nbfree(struct ext2_gd * gd)502 e2fs_gd_get_nbfree(struct ext2_gd *gd)
503 {
504
505 return (((uint32_t)(le16toh(gd->ext4bgd_nbfree_hi)) << 16) |
506 le16toh(gd->ext2bgd_nbfree));
507 }
508
509 static void
e2fs_gd_set_nbfree(struct ext2_gd * gd,uint32_t val)510 e2fs_gd_set_nbfree(struct ext2_gd *gd, uint32_t val)
511 {
512
513 gd->ext2bgd_nbfree = htole16(val & 0xffff);
514 gd->ext4bgd_nbfree_hi = htole16(val >> 16);
515 }
516
517 static uint32_t
e2fs_gd_get_nifree(struct ext2_gd * gd)518 e2fs_gd_get_nifree(struct ext2_gd *gd)
519 {
520
521 return (((uint32_t)(le16toh(gd->ext4bgd_nifree_hi)) << 16) |
522 le16toh(gd->ext2bgd_nifree));
523 }
524
525 static void
e2fs_gd_set_nifree(struct ext2_gd * gd,uint32_t val)526 e2fs_gd_set_nifree(struct ext2_gd *gd, uint32_t val)
527 {
528
529 gd->ext2bgd_nifree = htole16(val & 0xffff);
530 gd->ext4bgd_nifree_hi = htole16(val >> 16);
531 }
532
533 uint32_t
e2fs_gd_get_ndirs(struct ext2_gd * gd)534 e2fs_gd_get_ndirs(struct ext2_gd *gd)
535 {
536
537 return (((uint32_t)(le16toh(gd->ext4bgd_ndirs_hi)) << 16) |
538 le16toh(gd->ext2bgd_ndirs));
539 }
540
541 static void
e2fs_gd_set_ndirs(struct ext2_gd * gd,uint32_t val)542 e2fs_gd_set_ndirs(struct ext2_gd *gd, uint32_t val)
543 {
544
545 gd->ext2bgd_ndirs = htole16(val & 0xffff);
546 gd->ext4bgd_ndirs_hi = htole16(val >> 16);
547 }
548
549 static uint32_t
e2fs_gd_get_i_unused(struct ext2_gd * gd)550 e2fs_gd_get_i_unused(struct ext2_gd *gd)
551 {
552 return ((uint32_t)(le16toh(gd->ext4bgd_i_unused_hi) << 16) |
553 le16toh(gd->ext4bgd_i_unused));
554 }
555
556 static void
e2fs_gd_set_i_unused(struct ext2_gd * gd,uint32_t val)557 e2fs_gd_set_i_unused(struct ext2_gd *gd, uint32_t val)
558 {
559
560 gd->ext4bgd_i_unused = htole16(val & 0xffff);
561 gd->ext4bgd_i_unused_hi = htole16(val >> 16);
562 }
563
564 /*
565 * Find a cylinder to place a directory.
566 *
567 * The policy implemented by this algorithm is to allocate a
568 * directory inode in the same cylinder group as its parent
569 * directory, but also to reserve space for its files inodes
570 * and data. Restrict the number of directories which may be
571 * allocated one after another in the same cylinder group
572 * without intervening allocation of files.
573 *
574 * If we allocate a first level directory then force allocation
575 * in another cylinder group.
576 *
577 */
578 static u_long
ext2_dirpref(struct inode * pip)579 ext2_dirpref(struct inode *pip)
580 {
581 struct m_ext2fs *fs;
582 int cg, prefcg, cgsize;
583 uint64_t avgbfree, minbfree;
584 u_int avgifree, avgndir, curdirsize;
585 u_int minifree, maxndir;
586 u_int mincg, minndir;
587 u_int dirsize, maxcontigdirs;
588
589 mtx_assert(EXT2_MTX(pip->i_ump), MA_OWNED);
590 fs = pip->i_e2fs;
591
592 avgifree = fs->e2fs_ficount / fs->e2fs_gcount;
593 avgbfree = fs->e2fs_fbcount / fs->e2fs_gcount;
594 avgndir = fs->e2fs_total_dir / fs->e2fs_gcount;
595
596 /*
597 * Force allocation in another cg if creating a first level dir.
598 */
599 ASSERT_VOP_LOCKED(ITOV(pip), "ext2fs_dirpref");
600 if (ITOV(pip)->v_flag & VROOT) {
601 prefcg = karc4random() % fs->e2fs_gcount;
602 mincg = prefcg;
603 minndir = fs->e2fs_ipg;
604 for (cg = prefcg; cg < fs->e2fs_gcount; cg++)
605 if (e2fs_gd_get_ndirs(&fs->e2fs_gd[cg]) < minndir &&
606 e2fs_gd_get_nifree(&fs->e2fs_gd[cg]) >= avgifree &&
607 e2fs_gd_get_nbfree(&fs->e2fs_gd[cg]) >= avgbfree) {
608 mincg = cg;
609 minndir = e2fs_gd_get_ndirs(&fs->e2fs_gd[cg]);
610 }
611 for (cg = 0; cg < prefcg; cg++)
612 if (e2fs_gd_get_ndirs(&fs->e2fs_gd[cg]) < minndir &&
613 e2fs_gd_get_nifree(&fs->e2fs_gd[cg]) >= avgifree &&
614 e2fs_gd_get_nbfree(&fs->e2fs_gd[cg]) >= avgbfree) {
615 mincg = cg;
616 minndir = e2fs_gd_get_ndirs(&fs->e2fs_gd[cg]);
617 }
618 return (mincg);
619 }
620 /*
621 * Count various limits which used for
622 * optimal allocation of a directory inode.
623 */
624 maxndir = min(avgndir + fs->e2fs_ipg / 16, fs->e2fs_ipg);
625 minifree = avgifree - avgifree / 4;
626 if (minifree < 1)
627 minifree = 1;
628 minbfree = avgbfree - avgbfree / 4;
629 if (minbfree < 1)
630 minbfree = 1;
631 cgsize = fs->e2fs_fsize * fs->e2fs_fpg;
632 dirsize = AVGDIRSIZE;
633 curdirsize = avgndir ?
634 (cgsize - avgbfree * fs->e2fs_bsize) / avgndir : 0;
635 if (dirsize < curdirsize)
636 dirsize = curdirsize;
637 maxcontigdirs = min((avgbfree * fs->e2fs_bsize) / dirsize, 255);
638 maxcontigdirs = min(maxcontigdirs, fs->e2fs_ipg / AFPDIR);
639 if (maxcontigdirs == 0)
640 maxcontigdirs = 1;
641
642 /*
643 * Limit number of dirs in one cg and reserve space for
644 * regular files, but only if we have no deficit in
645 * inodes or space.
646 */
647 prefcg = ino_to_cg(fs, pip->i_number);
648 for (cg = prefcg; cg < fs->e2fs_gcount; cg++)
649 if (e2fs_gd_get_ndirs(&fs->e2fs_gd[cg]) < maxndir &&
650 e2fs_gd_get_nifree(&fs->e2fs_gd[cg]) >= minifree &&
651 e2fs_gd_get_nbfree(&fs->e2fs_gd[cg]) >= minbfree) {
652 if (fs->e2fs_contigdirs[cg] < maxcontigdirs)
653 return (cg);
654 }
655 for (cg = 0; cg < prefcg; cg++)
656 if (e2fs_gd_get_ndirs(&fs->e2fs_gd[cg]) < maxndir &&
657 e2fs_gd_get_nifree(&fs->e2fs_gd[cg]) >= minifree &&
658 e2fs_gd_get_nbfree(&fs->e2fs_gd[cg]) >= minbfree) {
659 if (fs->e2fs_contigdirs[cg] < maxcontigdirs)
660 return (cg);
661 }
662 /*
663 * This is a backstop when we have deficit in space.
664 */
665 for (cg = prefcg; cg < fs->e2fs_gcount; cg++)
666 if (e2fs_gd_get_nifree(&fs->e2fs_gd[cg]) >= avgifree)
667 return (cg);
668 for (cg = 0; cg < prefcg; cg++)
669 if (e2fs_gd_get_nifree(&fs->e2fs_gd[cg]) >= avgifree)
670 break;
671 return (cg);
672 }
673
674 /*
675 * Select the desired position for the next block in a file.
676 *
677 * we try to mimic what Remy does in inode_getblk/block_getblk
678 *
679 * we note: blocknr == 0 means that we're about to allocate either
680 * a direct block or a pointer block at the first level of indirection
681 * (In other words, stuff that will go in i_db[] or i_ib[])
682 *
683 * blocknr != 0 means that we're allocating a block that is none
684 * of the above. Then, blocknr tells us the number of the block
685 * that will hold the pointer
686 */
687 e4fs_daddr_t
ext2_blkpref(struct inode * ip,e2fs_lbn_t lbn,int indx,e2fs_daddr_t * bap,e2fs_daddr_t blocknr)688 ext2_blkpref(struct inode *ip, e2fs_lbn_t lbn, int indx, e2fs_daddr_t *bap,
689 e2fs_daddr_t blocknr)
690 {
691 struct m_ext2fs *fs;
692 int tmp;
693
694 fs = ip->i_e2fs;
695
696 mtx_assert(EXT2_MTX(ip->i_ump), MA_OWNED);
697
698 /*
699 * If the next block is actually what we thought it is, then set the
700 * goal to what we thought it should be.
701 */
702 if (ip->i_next_alloc_block == lbn && ip->i_next_alloc_goal != 0)
703 return ip->i_next_alloc_goal;
704
705 /*
706 * Now check whether we were provided with an array that basically
707 * tells us previous blocks to which we want to stay close.
708 */
709 if (bap)
710 for (tmp = indx - 1; tmp >= 0; tmp--)
711 if (bap[tmp])
712 return (le32toh(bap[tmp]));
713
714 /*
715 * Else lets fall back to the blocknr or, if there is none, follow
716 * the rule that a block should be allocated near its inode.
717 */
718 return (blocknr ? blocknr :
719 (e2fs_daddr_t)(ip->i_block_group *
720 EXT2_BLOCKS_PER_GROUP(fs)) + le32toh(fs->e2fs->e2fs_first_dblock));
721 }
722
723 /*
724 * Implement the cylinder overflow algorithm.
725 *
726 * The policy implemented by this algorithm is:
727 * 1) allocate the block in its requested cylinder group.
728 * 2) quadratically rehash on the cylinder group number.
729 * 3) brute force search for a free block.
730 */
731 static e4fs_daddr_t
ext2_hashalloc(struct inode * ip,int cg,long pref,int size,daddr_t (* allocator)(struct inode *,int,daddr_t,int))732 ext2_hashalloc(struct inode *ip, int cg, long pref, int size,
733 daddr_t (*allocator) (struct inode *, int, daddr_t, int))
734 {
735 struct m_ext2fs *fs;
736 e4fs_daddr_t result;
737 int i, icg = cg;
738
739 mtx_assert(EXT2_MTX(ip->i_ump), MA_OWNED);
740 fs = ip->i_e2fs;
741 /*
742 * 1: preferred cylinder group
743 */
744 result = (*allocator)(ip, cg, pref, size);
745 if (result)
746 return (result);
747 /*
748 * 2: quadratic rehash
749 */
750 for (i = 1; i < fs->e2fs_gcount; i *= 2) {
751 cg += i;
752 if (cg >= fs->e2fs_gcount)
753 cg -= fs->e2fs_gcount;
754 result = (*allocator)(ip, cg, 0, size);
755 if (result)
756 return (result);
757 }
758 /*
759 * 3: brute force search
760 * Note that we start at i == 2, since 0 was checked initially,
761 * and 1 is always checked in the quadratic rehash.
762 */
763 cg = (icg + 2) % fs->e2fs_gcount;
764 for (i = 2; i < fs->e2fs_gcount; i++) {
765 result = (*allocator)(ip, cg, 0, size);
766 if (result)
767 return (result);
768 cg++;
769 if (cg == fs->e2fs_gcount)
770 cg = 0;
771 }
772 return (0);
773 }
774
775 static uint64_t
ext2_cg_number_gdb_nometa(struct m_ext2fs * fs,int cg)776 ext2_cg_number_gdb_nometa(struct m_ext2fs *fs, int cg)
777 {
778
779 if (!ext2_cg_has_sb(fs, cg))
780 return (0);
781
782 if (EXT2_HAS_INCOMPAT_FEATURE(fs, EXT2F_INCOMPAT_META_BG))
783 return (le32toh(fs->e2fs->e3fs_first_meta_bg));
784
785 return ((fs->e2fs_gcount + EXT2_DESCS_PER_BLOCK(fs) - 1) /
786 EXT2_DESCS_PER_BLOCK(fs));
787 }
788
789 static uint64_t
ext2_cg_number_gdb_meta(struct m_ext2fs * fs,int cg)790 ext2_cg_number_gdb_meta(struct m_ext2fs *fs, int cg)
791 {
792 unsigned long metagroup;
793 int first, last;
794
795 metagroup = cg / EXT2_DESCS_PER_BLOCK(fs);
796 first = metagroup * EXT2_DESCS_PER_BLOCK(fs);
797 last = first + EXT2_DESCS_PER_BLOCK(fs) - 1;
798
799 if (cg == first || cg == first + 1 || cg == last)
800 return (1);
801
802 return (0);
803 }
804
805 uint64_t
ext2_cg_number_gdb(struct m_ext2fs * fs,int cg)806 ext2_cg_number_gdb(struct m_ext2fs *fs, int cg)
807 {
808 unsigned long first_meta_bg, metagroup;
809
810 first_meta_bg = le32toh(fs->e2fs->e3fs_first_meta_bg);
811 metagroup = cg / EXT2_DESCS_PER_BLOCK(fs);
812
813 if (!EXT2_HAS_INCOMPAT_FEATURE(fs, EXT2F_INCOMPAT_META_BG) ||
814 metagroup < first_meta_bg)
815 return (ext2_cg_number_gdb_nometa(fs, cg));
816
817 return ext2_cg_number_gdb_meta(fs, cg);
818 }
819
820 static int
ext2_number_base_meta_blocks(struct m_ext2fs * fs,int cg)821 ext2_number_base_meta_blocks(struct m_ext2fs *fs, int cg)
822 {
823 int number;
824
825 number = ext2_cg_has_sb(fs, cg);
826
827 if (!EXT2_HAS_INCOMPAT_FEATURE(fs, EXT2F_INCOMPAT_META_BG) ||
828 cg < le32toh(fs->e2fs->e3fs_first_meta_bg) *
829 EXT2_DESCS_PER_BLOCK(fs)) {
830 if (number) {
831 number += ext2_cg_number_gdb(fs, cg);
832 number += le16toh(fs->e2fs->e2fs_reserved_ngdb);
833 }
834 } else {
835 number += ext2_cg_number_gdb(fs, cg);
836 }
837
838 return (number);
839 }
840
841 static void
ext2_mark_bitmap_end(int start_bit,int end_bit,char * bitmap)842 ext2_mark_bitmap_end(int start_bit, int end_bit, char *bitmap)
843 {
844 int i;
845
846 if (start_bit >= end_bit)
847 return;
848
849 for (i = start_bit; i < ((start_bit + 7) & ~7UL); i++)
850 setbit(bitmap, i);
851 if (i < end_bit)
852 memset(bitmap + (i >> 3), 0xff, (end_bit - i) >> 3);
853 }
854
855 static int
ext2_get_group_number(struct m_ext2fs * fs,e4fs_daddr_t block)856 ext2_get_group_number(struct m_ext2fs *fs, e4fs_daddr_t block)
857 {
858
859 return ((block - le32toh(fs->e2fs->e2fs_first_dblock)) /
860 fs->e2fs_bsize);
861 }
862
863 static int
ext2_block_in_group(struct m_ext2fs * fs,e4fs_daddr_t block,int cg)864 ext2_block_in_group(struct m_ext2fs *fs, e4fs_daddr_t block, int cg)
865 {
866
867 return ((ext2_get_group_number(fs, block) == cg) ? 1 : 0);
868 }
869
870 static int
ext2_cg_block_bitmap_init(struct m_ext2fs * fs,int cg,struct buf * bp)871 ext2_cg_block_bitmap_init(struct m_ext2fs *fs, int cg, struct buf *bp)
872 {
873 int bit, bit_max, inodes_per_block;
874 uint64_t start, tmp;
875
876 if (!(le16toh(fs->e2fs_gd[cg].ext4bgd_flags) & EXT2_BG_BLOCK_UNINIT))
877 return (0);
878
879 memset(bp->b_data, 0, fs->e2fs_bsize);
880
881 bit_max = ext2_number_base_meta_blocks(fs, cg);
882 if ((bit_max >> 3) >= fs->e2fs_bsize)
883 return (EINVAL);
884
885 for (bit = 0; bit < bit_max; bit++)
886 setbit(bp->b_data, bit);
887
888 start = (uint64_t)cg * fs->e2fs_bpg +
889 le32toh(fs->e2fs->e2fs_first_dblock);
890
891 /* Set bits for block and inode bitmaps, and inode table. */
892 tmp = e2fs_gd_get_b_bitmap(&fs->e2fs_gd[cg]);
893 if (!EXT2_HAS_INCOMPAT_FEATURE(fs, EXT2F_INCOMPAT_FLEX_BG) ||
894 ext2_block_in_group(fs, tmp, cg))
895 setbit(bp->b_data, tmp - start);
896
897 tmp = e2fs_gd_get_i_bitmap(&fs->e2fs_gd[cg]);
898 if (!EXT2_HAS_INCOMPAT_FEATURE(fs, EXT2F_INCOMPAT_FLEX_BG) ||
899 ext2_block_in_group(fs, tmp, cg))
900 setbit(bp->b_data, tmp - start);
901
902 tmp = e2fs_gd_get_i_tables(&fs->e2fs_gd[cg]);
903 inodes_per_block = fs->e2fs_bsize/EXT2_INODE_SIZE(fs);
904 while( tmp < e2fs_gd_get_i_tables(&fs->e2fs_gd[cg]) +
905 fs->e2fs_ipg / inodes_per_block ) {
906 if (!EXT2_HAS_INCOMPAT_FEATURE(fs, EXT2F_INCOMPAT_FLEX_BG) ||
907 ext2_block_in_group(fs, tmp, cg))
908 setbit(bp->b_data, tmp - start);
909 tmp++;
910 }
911
912 /*
913 * Also if the number of blocks within the group is less than
914 * the blocksize * 8 ( which is the size of bitmap ), set rest
915 * of the block bitmap to 1
916 */
917 ext2_mark_bitmap_end(fs->e2fs_bpg, fs->e2fs_bsize * 8,
918 bp->b_data);
919
920 /* Clean the flag */
921 fs->e2fs_gd[cg].ext4bgd_flags = htole16(le16toh(
922 fs->e2fs_gd[cg].ext4bgd_flags) & ~EXT2_BG_BLOCK_UNINIT);
923
924 return (0);
925 }
926
927 static int
ext2_b_bitmap_validate(struct m_ext2fs * fs,struct buf * bp,int cg)928 ext2_b_bitmap_validate(struct m_ext2fs *fs, struct buf *bp, int cg)
929 {
930 struct ext2_gd *gd;
931 uint64_t group_first_block;
932 unsigned int offset, max_bit;
933
934 if (EXT2_HAS_INCOMPAT_FEATURE(fs, EXT2F_INCOMPAT_FLEX_BG)) {
935 /*
936 * It is not possible to check block bitmap in case of this
937 * feature, because the inode and block bitmaps and inode table
938 * blocks may not be in the group at all.
939 * So, skip check in this case.
940 */
941 return (0);
942 }
943
944 gd = &fs->e2fs_gd[cg];
945 max_bit = fs->e2fs_fpg;
946 group_first_block = ((uint64_t)cg) * fs->e2fs_fpg +
947 le32toh(fs->e2fs->e2fs_first_dblock);
948
949 /* Check block bitmap block number */
950 offset = e2fs_gd_get_b_bitmap(gd) - group_first_block;
951 if (offset >= max_bit || !isset(bp->b_data, offset)) {
952 SDT_PROBE2(ext2fs, , alloc, ext2_b_bitmap_validate_error,
953 "bad block bitmap, group", cg);
954 return (EINVAL);
955 }
956
957 /* Check inode bitmap block number */
958 offset = e2fs_gd_get_i_bitmap(gd) - group_first_block;
959 if (offset >= max_bit || !isset(bp->b_data, offset)) {
960 SDT_PROBE2(ext2fs, , alloc, ext2_b_bitmap_validate_error,
961 "bad inode bitmap", cg);
962 return (EINVAL);
963 }
964
965 /* Check inode table */
966 offset = e2fs_gd_get_i_tables(gd) - group_first_block;
967 if (offset >= max_bit || offset + fs->e2fs_itpg >= max_bit) {
968 SDT_PROBE2(ext2fs, , alloc, ext2_b_bitmap_validate_error,
969 "bad inode table, group", cg);
970 return (EINVAL);
971 }
972
973 return (0);
974 }
975
976 /*
977 * Determine whether a block can be allocated.
978 *
979 * Check to see if a block of the appropriate size is available,
980 * and if it is, allocate it.
981 */
982 static daddr_t
ext2_alloccg(struct inode * ip,int cg,daddr_t bpref,int size)983 ext2_alloccg(struct inode *ip, int cg, daddr_t bpref, int size)
984 {
985 struct m_ext2fs *fs;
986 struct buf *bp;
987 struct ext2mount *ump;
988 daddr_t bno, runstart, runlen;
989 int bit, loc, end, error, start;
990 char *bbp;
991 /* XXX ondisk32 */
992 fs = ip->i_e2fs;
993 ump = ip->i_ump;
994 if (e2fs_gd_get_nbfree(&fs->e2fs_gd[cg]) == 0)
995 return (0);
996
997 EXT2_UNLOCK(ump);
998 error = bread(ip->i_devvp, fsbtodoff(fs,
999 e2fs_gd_get_b_bitmap(&fs->e2fs_gd[cg])),
1000 (int)fs->e2fs_bsize, &bp);
1001 if (error)
1002 goto fail;
1003
1004 if (EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_GDT_CSUM) ||
1005 EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_METADATA_CKSUM)) {
1006 error = ext2_cg_block_bitmap_init(fs, cg, bp);
1007 if (error)
1008 goto fail;
1009
1010 ext2_gd_b_bitmap_csum_set(fs, cg, bp);
1011 }
1012 error = ext2_gd_b_bitmap_csum_verify(fs, cg, bp);
1013 if (error)
1014 goto fail;
1015
1016 error = ext2_b_bitmap_validate(fs,bp, cg);
1017 if (error)
1018 goto fail;
1019
1020 /*
1021 * Check, that another thread did not not allocate the last block in
1022 * this group while we were waiting for the buffer.
1023 */
1024 if (e2fs_gd_get_nbfree(&fs->e2fs_gd[cg]) == 0)
1025 goto fail;
1026
1027 bbp = (char *)bp->b_data;
1028
1029 if (dtog(fs, bpref) != cg)
1030 bpref = 0;
1031 if (bpref != 0) {
1032 bpref = dtogd(fs, bpref);
1033 /*
1034 * if the requested block is available, use it
1035 */
1036 if (isclr(bbp, bpref)) {
1037 bno = bpref;
1038 goto gotit;
1039 }
1040 }
1041 /*
1042 * no blocks in the requested cylinder, so take next
1043 * available one in this cylinder group.
1044 * first try to get 8 contigous blocks, then fall back to a single
1045 * block.
1046 */
1047 if (bpref)
1048 start = dtogd(fs, bpref) / NBBY;
1049 else
1050 start = 0;
1051 end = howmany(fs->e2fs_fpg, NBBY) - start;
1052 retry:
1053 runlen = 0;
1054 runstart = 0;
1055 for (loc = start; loc < end; loc++) {
1056 if (bbp[loc] == (char)0xff) {
1057 runlen = 0;
1058 continue;
1059 }
1060
1061 /* Start of a run, find the number of high clear bits. */
1062 if (runlen == 0) {
1063 bit = fls(bbp[loc]);
1064 runlen = NBBY - bit;
1065 runstart = loc * NBBY + bit;
1066 } else if (bbp[loc] == 0) {
1067 /* Continue a run. */
1068 runlen += NBBY;
1069 } else {
1070 /*
1071 * Finish the current run. If it isn't long
1072 * enough, start a new one.
1073 */
1074 bit = ffs(bbp[loc]) - 1;
1075 runlen += bit;
1076 if (runlen >= 8) {
1077 bno = runstart;
1078 goto gotit;
1079 }
1080
1081 /* Run was too short, start a new one. */
1082 bit = fls(bbp[loc]);
1083 runlen = NBBY - bit;
1084 runstart = loc * NBBY + bit;
1085 }
1086
1087 /* If the current run is long enough, use it. */
1088 if (runlen >= 8) {
1089 bno = runstart;
1090 goto gotit;
1091 }
1092 }
1093 if (start != 0) {
1094 end = start;
1095 start = 0;
1096 goto retry;
1097 }
1098 bno = ext2_mapsearch(fs, bbp, bpref);
1099 if (bno < 0)
1100 goto fail;
1101
1102 gotit:
1103 #ifdef INVARIANTS
1104 if (isset(bbp, bno)) {
1105 printf("ext2fs_alloccgblk: cg=%d bno=%jd fs=%s\n",
1106 cg, (intmax_t)bno, fs->e2fs_fsmnt);
1107 panic("ext2fs_alloccg: dup alloc");
1108 }
1109 #endif
1110 setbit(bbp, bno);
1111 EXT2_LOCK(ump);
1112 ext2_clusteracct(fs, bbp, cg, bno, -1);
1113 fs->e2fs_fbcount--;
1114 e2fs_gd_set_nbfree(&fs->e2fs_gd[cg],
1115 e2fs_gd_get_nbfree(&fs->e2fs_gd[cg]) - 1);
1116 fs->e2fs_fmod = 1;
1117 EXT2_UNLOCK(ump);
1118 ext2_gd_b_bitmap_csum_set(fs, cg, bp);
1119 bdwrite(bp);
1120 return (((uint64_t)cg) * fs->e2fs_fpg +
1121 le32toh(fs->e2fs->e2fs_first_dblock) + bno);
1122
1123 fail:
1124 brelse(bp);
1125 EXT2_LOCK(ump);
1126 return (0);
1127 }
1128
1129 /*
1130 * Determine whether a cluster can be allocated.
1131 */
1132 static daddr_t
ext2_clusteralloc(struct inode * ip,int cg,daddr_t bpref,int len)1133 ext2_clusteralloc(struct inode *ip, int cg, daddr_t bpref, int len)
1134 {
1135 struct m_ext2fs *fs;
1136 struct ext2mount *ump;
1137 struct buf *bp;
1138 char *bbp;
1139 int bit, error, got, i, loc, run;
1140 int32_t *lp;
1141 daddr_t bno;
1142
1143 fs = ip->i_e2fs;
1144 ump = ip->i_ump;
1145
1146 if (fs->e2fs_maxcluster[cg] < len)
1147 return (0);
1148
1149 EXT2_UNLOCK(ump);
1150 error = bread(ip->i_devvp,
1151 fsbtodoff(fs, e2fs_gd_get_b_bitmap(&fs->e2fs_gd[cg])),
1152 (int)fs->e2fs_bsize, &bp);
1153 if (error)
1154 goto fail_lock;
1155
1156 bbp = (char *)bp->b_data;
1157 EXT2_LOCK(ump);
1158 /*
1159 * Check to see if a cluster of the needed size (or bigger) is
1160 * available in this cylinder group.
1161 */
1162 lp = &fs->e2fs_clustersum[cg].cs_sum[len];
1163 for (i = len; i <= fs->e2fs_contigsumsize; i++)
1164 if (*lp++ > 0)
1165 break;
1166 if (i > fs->e2fs_contigsumsize) {
1167 /*
1168 * Update the cluster summary information to reflect
1169 * the true maximum-sized cluster so that future cluster
1170 * allocation requests can avoid reading the bitmap only
1171 * to find no cluster.
1172 */
1173 lp = &fs->e2fs_clustersum[cg].cs_sum[len - 1];
1174 for (i = len - 1; i > 0; i--)
1175 if (*lp-- > 0)
1176 break;
1177 fs->e2fs_maxcluster[cg] = i;
1178 goto fail;
1179 }
1180 EXT2_UNLOCK(ump);
1181
1182 /* Search the bitmap to find a big enough cluster like in FFS. */
1183 if (dtog(fs, bpref) != cg)
1184 bpref = 0;
1185 if (bpref != 0)
1186 bpref = dtogd(fs, bpref);
1187 loc = bpref / NBBY;
1188 bit = 1 << (bpref % NBBY);
1189 for (run = 0, got = bpref; got < fs->e2fs_fpg; got++) {
1190 if ((bbp[loc] & bit) != 0)
1191 run = 0;
1192 else {
1193 run++;
1194 if (run == len)
1195 break;
1196 }
1197 if ((got & (NBBY - 1)) != (NBBY - 1))
1198 bit <<= 1;
1199 else {
1200 loc++;
1201 bit = 1;
1202 }
1203 }
1204
1205 if (got >= fs->e2fs_fpg)
1206 goto fail_lock;
1207
1208 /* Allocate the cluster that we found. */
1209 for (i = 1; i < len; i++)
1210 if (!isclr(bbp, got - run + i))
1211 panic("ext2_clusteralloc: map mismatch");
1212
1213 bno = got - run + 1;
1214 if (bno >= fs->e2fs_fpg)
1215 panic("ext2_clusteralloc: allocated out of group");
1216
1217 EXT2_LOCK(ump);
1218 for (i = 0; i < len; i += fs->e2fs_fpb) {
1219 setbit(bbp, bno + i);
1220 ext2_clusteracct(fs, bbp, cg, bno + i, -1);
1221 fs->e2fs_fbcount--;
1222 e2fs_gd_set_nbfree(&fs->e2fs_gd[cg],
1223 e2fs_gd_get_nbfree(&fs->e2fs_gd[cg]) - 1);
1224 }
1225 fs->e2fs_fmod = 1;
1226 EXT2_UNLOCK(ump);
1227
1228 bdwrite(bp);
1229 return (cg * fs->e2fs_fpg + le32toh(fs->e2fs->e2fs_first_dblock)
1230 + bno);
1231
1232 fail_lock:
1233 EXT2_LOCK(ump);
1234 fail:
1235 brelse(bp);
1236 return (0);
1237 }
1238
1239 static int
ext2_zero_inode_table(struct inode * ip,int cg)1240 ext2_zero_inode_table(struct inode *ip, int cg)
1241 {
1242 struct m_ext2fs *fs;
1243 struct buf *bp;
1244 int i, all_blks, used_blks;
1245
1246 fs = ip->i_e2fs;
1247
1248 if (le16toh(fs->e2fs_gd[cg].ext4bgd_flags) & EXT2_BG_INODE_ZEROED)
1249 return (0);
1250
1251 all_blks = le16toh(fs->e2fs->e2fs_inode_size) * fs->e2fs_ipg /
1252 fs->e2fs_bsize;
1253
1254 used_blks = howmany(fs->e2fs_ipg -
1255 e2fs_gd_get_i_unused(&fs->e2fs_gd[cg]),
1256 fs->e2fs_bsize / EXT2_INODE_SIZE(fs));
1257
1258 for (i = 0; i < all_blks - used_blks; i++) {
1259 bp = getblk(ip->i_devvp, fsbtodoff(fs,
1260 e2fs_gd_get_i_tables(&fs->e2fs_gd[cg]) + used_blks + i),
1261 fs->e2fs_bsize, 0, 0);
1262 if (!bp)
1263 return (EIO);
1264
1265 vfs_bio_clrbuf(bp);
1266 bawrite(bp);
1267 }
1268
1269 fs->e2fs_gd[cg].ext4bgd_flags = htole16(le16toh(
1270 fs->e2fs_gd[cg].ext4bgd_flags) | EXT2_BG_INODE_ZEROED);
1271
1272 return (0);
1273 }
1274
1275 static void
ext2_fix_bitmap_tail(unsigned char * bitmap,int first,int last)1276 ext2_fix_bitmap_tail(unsigned char *bitmap, int first, int last)
1277 {
1278 int i;
1279
1280 for (i = first; i <= last; i++)
1281 bitmap[i] = 0xff;
1282 }
1283
1284 /*
1285 * Determine whether an inode can be allocated.
1286 *
1287 * Check to see if an inode is available, and if it is,
1288 * allocate it using tode in the specified cylinder group.
1289 */
1290 static daddr_t
ext2_nodealloccg(struct inode * ip,int cg,daddr_t ipref,int mode)1291 ext2_nodealloccg(struct inode *ip, int cg, daddr_t ipref, int mode)
1292 {
1293 struct m_ext2fs *fs;
1294 struct buf *bp;
1295 struct ext2mount *ump;
1296 int error, start, len, ifree, ibytes;
1297 char *ibp, *loc;
1298
1299 ipref--; /* to avoid a lot of (ipref -1) */
1300 if (ipref == -1)
1301 ipref = 0;
1302 fs = ip->i_e2fs;
1303 ump = ip->i_ump;
1304 if (e2fs_gd_get_nifree(&fs->e2fs_gd[cg]) == 0)
1305 return (0);
1306 EXT2_UNLOCK(ump);
1307 error = bread(ip->i_devvp, fsbtodoff(fs,
1308 e2fs_gd_get_i_bitmap(&fs->e2fs_gd[cg])),
1309 (int)fs->e2fs_bsize, &bp);
1310 if (error) {
1311 brelse(bp);
1312 EXT2_LOCK(ump);
1313 return (0);
1314 }
1315 if (EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_GDT_CSUM) ||
1316 EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_METADATA_CKSUM)) {
1317 if (le16toh(fs->e2fs_gd[cg].ext4bgd_flags) &
1318 EXT2_BG_INODE_UNINIT) {
1319 ibytes = fs->e2fs_ipg / 8;
1320 memset(bp->b_data, 0, ibytes - 1);
1321 ext2_fix_bitmap_tail(bp->b_data, ibytes,
1322 fs->e2fs_bsize - 1);
1323 fs->e2fs_gd[cg].ext4bgd_flags = htole16(le16toh(
1324 fs->e2fs_gd[cg].ext4bgd_flags) &
1325 ~EXT2_BG_INODE_UNINIT);
1326 }
1327 ext2_gd_i_bitmap_csum_set(fs, cg, bp);
1328 error = ext2_zero_inode_table(ip, cg);
1329 if (error) {
1330 brelse(bp);
1331 EXT2_LOCK(ump);
1332 return (0);
1333 }
1334 }
1335 error = ext2_gd_i_bitmap_csum_verify(fs, cg, bp);
1336 if (error) {
1337 brelse(bp);
1338 EXT2_LOCK(ump);
1339 return (0);
1340 }
1341 if (e2fs_gd_get_nifree(&fs->e2fs_gd[cg]) == 0) {
1342 /*
1343 * Another thread allocated the last i-node in this
1344 * group while we were waiting for the buffer.
1345 */
1346 brelse(bp);
1347 EXT2_LOCK(ump);
1348 return (0);
1349 }
1350 ibp = (char *)bp->b_data;
1351 if (ipref) {
1352 ipref %= fs->e2fs_ipg;
1353 if (isclr(ibp, ipref))
1354 goto gotit;
1355 }
1356 start = ipref / NBBY;
1357 len = howmany(fs->e2fs_ipg - ipref, NBBY);
1358 loc = memcchr(&ibp[start], 0xff, len);
1359 if (loc == NULL) {
1360 len = start + 1;
1361 start = 0;
1362 loc = memcchr(&ibp[start], 0xff, len);
1363 if (loc == NULL) {
1364 SDT_PROBE3(ext2fs, , alloc,
1365 ext2_nodealloccg_bmap_corrupted, cg, ipref,
1366 fs->e2fs_fsmnt);
1367 brelse(bp);
1368 EXT2_LOCK(ump);
1369 return (0);
1370 }
1371 }
1372 ipref = (loc - ibp) * NBBY + ffs(~*loc) - 1;
1373 gotit:
1374 setbit(ibp, ipref);
1375 EXT2_LOCK(ump);
1376 e2fs_gd_set_nifree(&fs->e2fs_gd[cg],
1377 e2fs_gd_get_nifree(&fs->e2fs_gd[cg]) - 1);
1378 if (EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_GDT_CSUM) ||
1379 EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_METADATA_CKSUM)) {
1380 ifree = fs->e2fs_ipg - e2fs_gd_get_i_unused(&fs->e2fs_gd[cg]);
1381 if (ipref + 1 > ifree)
1382 e2fs_gd_set_i_unused(&fs->e2fs_gd[cg],
1383 fs->e2fs_ipg - (ipref + 1));
1384 }
1385 fs->e2fs_ficount--;
1386 fs->e2fs_fmod = 1;
1387 if ((mode & IFMT) == IFDIR) {
1388 e2fs_gd_set_ndirs(&fs->e2fs_gd[cg],
1389 e2fs_gd_get_ndirs(&fs->e2fs_gd[cg]) + 1);
1390 fs->e2fs_total_dir++;
1391 }
1392 EXT2_UNLOCK(ump);
1393 ext2_gd_i_bitmap_csum_set(fs, cg, bp);
1394 bdwrite(bp);
1395 return ((uint64_t)cg * fs->e2fs_ipg + ipref + 1);
1396 }
1397
1398 /*
1399 * Free a block or fragment.
1400 *
1401 */
1402 void
ext2_blkfree(struct inode * ip,e4fs_daddr_t bno,long size)1403 ext2_blkfree(struct inode *ip, e4fs_daddr_t bno, long size)
1404 {
1405 struct m_ext2fs *fs;
1406 struct buf *bp;
1407 struct ext2mount *ump;
1408 int cg, error;
1409 char *bbp;
1410
1411 fs = ip->i_e2fs;
1412 ump = ip->i_ump;
1413 cg = dtog(fs, bno);
1414 if (bno >= fs->e2fs_bcount) {
1415 SDT_PROBE2(ext2fs, , alloc, ext2_blkfree_bad_block,
1416 ip->i_number, bno);
1417 return;
1418 }
1419 error = bread(ip->i_devvp,
1420 fsbtodoff(fs, e2fs_gd_get_b_bitmap(&fs->e2fs_gd[cg])),
1421 (int)fs->e2fs_bsize, &bp);
1422 if (error) {
1423 brelse(bp);
1424 return;
1425 }
1426 bbp = (char *)bp->b_data;
1427 bno = dtogd(fs, bno);
1428 if (isclr(bbp, bno)) {
1429 panic("ext2_blkfree: freeing free block %lld, fs=%s",
1430 (long long)bno, fs->e2fs_fsmnt);
1431 }
1432 clrbit(bbp, bno);
1433 EXT2_LOCK(ump);
1434 ext2_clusteracct(fs, bbp, cg, bno, 1);
1435 fs->e2fs_fbcount++;
1436 e2fs_gd_set_nbfree(&fs->e2fs_gd[cg],
1437 e2fs_gd_get_nbfree(&fs->e2fs_gd[cg]) + 1);
1438 fs->e2fs_fmod = 1;
1439 EXT2_UNLOCK(ump);
1440 ext2_gd_b_bitmap_csum_set(fs, cg, bp);
1441 bdwrite(bp);
1442 }
1443
1444 /*
1445 * Free an inode.
1446 *
1447 */
1448 int
ext2_vfree(struct vnode * pvp,ino_t ino,int mode)1449 ext2_vfree(struct vnode *pvp, ino_t ino, int mode)
1450 {
1451 struct m_ext2fs *fs;
1452 struct inode *pip;
1453 struct buf *bp;
1454 struct ext2mount *ump;
1455 int error, cg;
1456 char *ibp;
1457
1458 pip = VTOI(pvp);
1459 fs = pip->i_e2fs;
1460 ump = pip->i_ump;
1461 if ((u_int)ino > fs->e2fs_ipg * fs->e2fs_gcount)
1462 panic("ext2_vfree: range: devvp = %p, ino = %ju, fs = %s",
1463 pip->i_devvp, (uintmax_t)ino, fs->e2fs_fsmnt);
1464
1465 cg = ino_to_cg(fs, ino);
1466 error = bread(pip->i_devvp,
1467 fsbtodoff(fs, e2fs_gd_get_i_bitmap(&fs->e2fs_gd[cg])),
1468 (int)fs->e2fs_bsize, &bp);
1469 if (error) {
1470 brelse(bp);
1471 return (0);
1472 }
1473 ibp = (char *)bp->b_data;
1474 ino = (ino - 1) % fs->e2fs_ipg;
1475 if (isclr(ibp, ino)) {
1476 SDT_PROBE2(ext2fs, , alloc, ext2_vfree_doublefree,
1477 fs->e2fs_fsmnt, ino);
1478 if (fs->e2fs_ronly == 0)
1479 panic("ext2_vfree: freeing free inode");
1480 }
1481 clrbit(ibp, ino);
1482 EXT2_LOCK(ump);
1483 fs->e2fs_ficount++;
1484 e2fs_gd_set_nifree(&fs->e2fs_gd[cg],
1485 e2fs_gd_get_nifree(&fs->e2fs_gd[cg]) + 1);
1486 if ((mode & IFMT) == IFDIR) {
1487 e2fs_gd_set_ndirs(&fs->e2fs_gd[cg],
1488 e2fs_gd_get_ndirs(&fs->e2fs_gd[cg]) - 1);
1489 fs->e2fs_total_dir--;
1490 }
1491 fs->e2fs_fmod = 1;
1492 EXT2_UNLOCK(ump);
1493 ext2_gd_i_bitmap_csum_set(fs, cg, bp);
1494 bdwrite(bp);
1495 return (0);
1496 }
1497
1498 /*
1499 * Find a block in the specified cylinder group.
1500 *
1501 * It is a panic if a request is made to find a block if none are
1502 * available.
1503 */
1504 static daddr_t
ext2_mapsearch(struct m_ext2fs * fs,char * bbp,daddr_t bpref)1505 ext2_mapsearch(struct m_ext2fs *fs, char *bbp, daddr_t bpref)
1506 {
1507 char *loc;
1508 int start, len;
1509
1510 /*
1511 * find the fragment by searching through the free block
1512 * map for an appropriate bit pattern
1513 */
1514 if (bpref)
1515 start = dtogd(fs, bpref) / NBBY;
1516 else
1517 start = 0;
1518 len = howmany(fs->e2fs_fpg, NBBY) - start;
1519 loc = memcchr(&bbp[start], 0xff, len);
1520 if (loc == NULL) {
1521 len = start + 1;
1522 start = 0;
1523 loc = memcchr(&bbp[start], 0xff, len);
1524 if (loc == NULL) {
1525 panic("ext2_mapsearch: map corrupted: start=%d, len=%d,"
1526 "fs=%s", start, len, fs->e2fs_fsmnt);
1527 /* NOTREACHED */
1528 }
1529 }
1530 return ((loc - bbp) * NBBY + ffs(~*loc) - 1);
1531 }
1532
1533 int
ext2_cg_has_sb(struct m_ext2fs * fs,int cg)1534 ext2_cg_has_sb(struct m_ext2fs *fs, int cg)
1535 {
1536 int a3, a5, a7;
1537
1538 if (cg == 0)
1539 return (1);
1540
1541 if (EXT2_HAS_COMPAT_FEATURE(fs, EXT2F_COMPAT_SPARSESUPER2)) {
1542 if (cg == le32toh(fs->e2fs->e4fs_backup_bgs[0]) ||
1543 cg == le32toh(fs->e2fs->e4fs_backup_bgs[1]))
1544 return (1);
1545 return (0);
1546 }
1547
1548 if ((cg <= 1) ||
1549 !EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_SPARSESUPER))
1550 return (1);
1551
1552 if (!(cg & 1))
1553 return (0);
1554
1555 for (a3 = 3, a5 = 5, a7 = 7;
1556 a3 <= cg || a5 <= cg || a7 <= cg;
1557 a3 *= 3, a5 *= 5, a7 *= 7)
1558 if (cg == a3 || cg == a5 || cg == a7)
1559 return (1);
1560 return (0);
1561 }
1562