]>
Commit | Line | Data |
---|---|---|
1da177e4 | 1 | /* |
7b718769 NS |
2 | * Copyright (c) 2000-2002,2005 Silicon Graphics, Inc. |
3 | * All Rights Reserved. | |
1da177e4 | 4 | * |
7b718769 NS |
5 | * This program is free software; you can redistribute it and/or |
6 | * modify it under the terms of the GNU General Public License as | |
1da177e4 LT |
7 | * published by the Free Software Foundation. |
8 | * | |
7b718769 NS |
9 | * This program is distributed in the hope that it would be useful, |
10 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
11 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
12 | * GNU General Public License for more details. | |
1da177e4 | 13 | * |
7b718769 NS |
14 | * You should have received a copy of the GNU General Public License |
15 | * along with this program; if not, write the Free Software Foundation, | |
16 | * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA | |
1da177e4 | 17 | */ |
1da177e4 | 18 | #include "xfs.h" |
a844f451 | 19 | #include "xfs_fs.h" |
1da177e4 | 20 | #include "xfs_types.h" |
a844f451 | 21 | #include "xfs_bit.h" |
1da177e4 | 22 | #include "xfs_log.h" |
a844f451 | 23 | #include "xfs_inum.h" |
1da177e4 LT |
24 | #include "xfs_trans.h" |
25 | #include "xfs_sb.h" | |
26 | #include "xfs_ag.h" | |
1da177e4 | 27 | #include "xfs_mount.h" |
1da177e4 | 28 | #include "xfs_bmap_btree.h" |
a844f451 | 29 | #include "xfs_alloc_btree.h" |
1da177e4 | 30 | #include "xfs_ialloc_btree.h" |
1da177e4 LT |
31 | #include "xfs_dinode.h" |
32 | #include "xfs_inode.h" | |
a844f451 NS |
33 | #include "xfs_btree.h" |
34 | #include "xfs_ialloc.h" | |
1da177e4 | 35 | #include "xfs_alloc.h" |
1da177e4 LT |
36 | #include "xfs_rtalloc.h" |
37 | #include "xfs_error.h" | |
38 | #include "xfs_bmap.h" | |
983d09ff DC |
39 | #include "xfs_cksum.h" |
40 | #include "xfs_buf_item.h" | |
1da177e4 | 41 | |
1da177e4 LT |
42 | |
43 | /* | |
44 | * Allocation group level functions. | |
45 | */ | |
75de2a91 DC |
46 | static inline int |
47 | xfs_ialloc_cluster_alignment( | |
48 | xfs_alloc_arg_t *args) | |
49 | { | |
50 | if (xfs_sb_version_hasalign(&args->mp->m_sb) && | |
51 | args->mp->m_sb.sb_inoalignmt >= | |
52 | XFS_B_TO_FSBT(args->mp, XFS_INODE_CLUSTER_SIZE(args->mp))) | |
53 | return args->mp->m_sb.sb_inoalignmt; | |
54 | return 1; | |
55 | } | |
1da177e4 | 56 | |
fe033cc8 | 57 | /* |
21875505 | 58 | * Lookup a record by ino in the btree given by cur. |
fe033cc8 | 59 | */ |
81e25176 | 60 | int /* error */ |
21875505 | 61 | xfs_inobt_lookup( |
fe033cc8 CH |
62 | struct xfs_btree_cur *cur, /* btree cursor */ |
63 | xfs_agino_t ino, /* starting inode of chunk */ | |
21875505 | 64 | xfs_lookup_t dir, /* <=, >=, == */ |
fe033cc8 CH |
65 | int *stat) /* success/failure */ |
66 | { | |
67 | cur->bc_rec.i.ir_startino = ino; | |
21875505 CH |
68 | cur->bc_rec.i.ir_freecount = 0; |
69 | cur->bc_rec.i.ir_free = 0; | |
70 | return xfs_btree_lookup(cur, dir, stat); | |
fe033cc8 CH |
71 | } |
72 | ||
278d0ca1 | 73 | /* |
afabc24a | 74 | * Update the record referred to by cur to the value given. |
278d0ca1 CH |
75 | * This either works (return 0) or gets an EFSCORRUPTED error. |
76 | */ | |
77 | STATIC int /* error */ | |
78 | xfs_inobt_update( | |
79 | struct xfs_btree_cur *cur, /* btree cursor */ | |
afabc24a | 80 | xfs_inobt_rec_incore_t *irec) /* btree record */ |
278d0ca1 CH |
81 | { |
82 | union xfs_btree_rec rec; | |
83 | ||
afabc24a CH |
84 | rec.inobt.ir_startino = cpu_to_be32(irec->ir_startino); |
85 | rec.inobt.ir_freecount = cpu_to_be32(irec->ir_freecount); | |
86 | rec.inobt.ir_free = cpu_to_be64(irec->ir_free); | |
278d0ca1 CH |
87 | return xfs_btree_update(cur, &rec); |
88 | } | |
89 | ||
8cc938fe CH |
90 | /* |
91 | * Get the data from the pointed-to record. | |
92 | */ | |
93 | int /* error */ | |
94 | xfs_inobt_get_rec( | |
95 | struct xfs_btree_cur *cur, /* btree cursor */ | |
2e287a73 | 96 | xfs_inobt_rec_incore_t *irec, /* btree record */ |
8cc938fe CH |
97 | int *stat) /* output: success/failure */ |
98 | { | |
99 | union xfs_btree_rec *rec; | |
100 | int error; | |
101 | ||
102 | error = xfs_btree_get_rec(cur, &rec, stat); | |
103 | if (!error && *stat == 1) { | |
2e287a73 CH |
104 | irec->ir_startino = be32_to_cpu(rec->inobt.ir_startino); |
105 | irec->ir_freecount = be32_to_cpu(rec->inobt.ir_freecount); | |
106 | irec->ir_free = be64_to_cpu(rec->inobt.ir_free); | |
8cc938fe CH |
107 | } |
108 | return error; | |
109 | } | |
110 | ||
0b48db80 DC |
111 | /* |
112 | * Verify that the number of free inodes in the AGI is correct. | |
113 | */ | |
114 | #ifdef DEBUG | |
115 | STATIC int | |
116 | xfs_check_agi_freecount( | |
117 | struct xfs_btree_cur *cur, | |
118 | struct xfs_agi *agi) | |
119 | { | |
120 | if (cur->bc_nlevels == 1) { | |
121 | xfs_inobt_rec_incore_t rec; | |
122 | int freecount = 0; | |
123 | int error; | |
124 | int i; | |
125 | ||
21875505 | 126 | error = xfs_inobt_lookup(cur, 0, XFS_LOOKUP_GE, &i); |
0b48db80 DC |
127 | if (error) |
128 | return error; | |
129 | ||
130 | do { | |
131 | error = xfs_inobt_get_rec(cur, &rec, &i); | |
132 | if (error) | |
133 | return error; | |
134 | ||
135 | if (i) { | |
136 | freecount += rec.ir_freecount; | |
137 | error = xfs_btree_increment(cur, 0, &i); | |
138 | if (error) | |
139 | return error; | |
140 | } | |
141 | } while (i == 1); | |
142 | ||
143 | if (!XFS_FORCED_SHUTDOWN(cur->bc_mp)) | |
144 | ASSERT(freecount == be32_to_cpu(agi->agi_freecount)); | |
145 | } | |
146 | return 0; | |
147 | } | |
148 | #else | |
149 | #define xfs_check_agi_freecount(cur, agi) 0 | |
150 | #endif | |
151 | ||
85c0b2ab DC |
152 | /* |
153 | * Initialise a new set of inodes. | |
154 | */ | |
2a30f36d | 155 | STATIC int |
85c0b2ab DC |
156 | xfs_ialloc_inode_init( |
157 | struct xfs_mount *mp, | |
158 | struct xfs_trans *tp, | |
159 | xfs_agnumber_t agno, | |
160 | xfs_agblock_t agbno, | |
161 | xfs_agblock_t length, | |
162 | unsigned int gen) | |
163 | { | |
164 | struct xfs_buf *fbuf; | |
165 | struct xfs_dinode *free; | |
166 | int blks_per_cluster, nbufs, ninodes; | |
167 | int version; | |
168 | int i, j; | |
169 | xfs_daddr_t d; | |
93848a99 | 170 | xfs_ino_t ino = 0; |
85c0b2ab DC |
171 | |
172 | /* | |
173 | * Loop over the new block(s), filling in the inodes. | |
174 | * For small block sizes, manipulate the inodes in buffers | |
175 | * which are multiples of the blocks size. | |
176 | */ | |
177 | if (mp->m_sb.sb_blocksize >= XFS_INODE_CLUSTER_SIZE(mp)) { | |
178 | blks_per_cluster = 1; | |
179 | nbufs = length; | |
180 | ninodes = mp->m_sb.sb_inopblock; | |
181 | } else { | |
182 | blks_per_cluster = XFS_INODE_CLUSTER_SIZE(mp) / | |
183 | mp->m_sb.sb_blocksize; | |
184 | nbufs = length / blks_per_cluster; | |
185 | ninodes = blks_per_cluster * mp->m_sb.sb_inopblock; | |
186 | } | |
187 | ||
188 | /* | |
93848a99 CH |
189 | * Figure out what version number to use in the inodes we create. If |
190 | * the superblock version has caught up to the one that supports the new | |
191 | * inode format, then use the new inode version. Otherwise use the old | |
192 | * version so that old kernels will continue to be able to use the file | |
193 | * system. | |
194 | * | |
195 | * For v3 inodes, we also need to write the inode number into the inode, | |
196 | * so calculate the first inode number of the chunk here as | |
197 | * XFS_OFFBNO_TO_AGINO() only works within a filesystem block, not | |
198 | * across multiple filesystem blocks (such as a cluster) and so cannot | |
199 | * be used in the cluster buffer loop below. | |
200 | * | |
201 | * Further, because we are writing the inode directly into the buffer | |
202 | * and calculating a CRC on the entire inode, we have ot log the entire | |
203 | * inode so that the entire range the CRC covers is present in the log. | |
204 | * That means for v3 inode we log the entire buffer rather than just the | |
205 | * inode cores. | |
85c0b2ab | 206 | */ |
93848a99 CH |
207 | if (xfs_sb_version_hascrc(&mp->m_sb)) { |
208 | version = 3; | |
209 | ino = XFS_AGINO_TO_INO(mp, agno, | |
210 | XFS_OFFBNO_TO_AGINO(mp, agbno, 0)); | |
211 | } else if (xfs_sb_version_hasnlink(&mp->m_sb)) | |
85c0b2ab DC |
212 | version = 2; |
213 | else | |
214 | version = 1; | |
215 | ||
216 | for (j = 0; j < nbufs; j++) { | |
217 | /* | |
218 | * Get the block. | |
219 | */ | |
220 | d = XFS_AGB_TO_DADDR(mp, agno, agbno + (j * blks_per_cluster)); | |
221 | fbuf = xfs_trans_get_buf(tp, mp->m_ddev_targp, d, | |
7c4cebe8 DC |
222 | mp->m_bsize * blks_per_cluster, |
223 | XBF_UNMAPPED); | |
2a30f36d CS |
224 | if (!fbuf) |
225 | return ENOMEM; | |
85c0b2ab DC |
226 | /* |
227 | * Initialize all inodes in this buffer and then log them. | |
228 | * | |
229 | * XXX: It would be much better if we had just one transaction | |
230 | * to log a whole cluster of inodes instead of all the | |
231 | * individual transactions causing a lot of log traffic. | |
232 | */ | |
1813dd64 | 233 | fbuf->b_ops = &xfs_inode_buf_ops; |
93848a99 | 234 | xfs_buf_zero(fbuf, 0, BBTOB(fbuf->b_length)); |
85c0b2ab DC |
235 | for (i = 0; i < ninodes; i++) { |
236 | int ioffset = i << mp->m_sb.sb_inodelog; | |
93848a99 | 237 | uint isize = xfs_dinode_size(version); |
85c0b2ab DC |
238 | |
239 | free = xfs_make_iptr(mp, fbuf, i); | |
240 | free->di_magic = cpu_to_be16(XFS_DINODE_MAGIC); | |
241 | free->di_version = version; | |
242 | free->di_gen = cpu_to_be32(gen); | |
243 | free->di_next_unlinked = cpu_to_be32(NULLAGINO); | |
93848a99 CH |
244 | |
245 | if (version == 3) { | |
246 | free->di_ino = cpu_to_be64(ino); | |
247 | ino++; | |
248 | uuid_copy(&free->di_uuid, &mp->m_sb.sb_uuid); | |
249 | xfs_dinode_calc_crc(mp, free); | |
250 | } else { | |
251 | /* just log the inode core */ | |
252 | xfs_trans_log_buf(tp, fbuf, ioffset, | |
253 | ioffset + isize - 1); | |
254 | } | |
255 | } | |
256 | if (version == 3) { | |
257 | /* need to log the entire buffer */ | |
258 | xfs_trans_log_buf(tp, fbuf, 0, | |
259 | BBTOB(fbuf->b_length) - 1); | |
85c0b2ab DC |
260 | } |
261 | xfs_trans_inode_alloc_buf(tp, fbuf); | |
262 | } | |
2a30f36d | 263 | return 0; |
85c0b2ab DC |
264 | } |
265 | ||
1da177e4 LT |
266 | /* |
267 | * Allocate new inodes in the allocation group specified by agbp. | |
268 | * Return 0 for success, else error code. | |
269 | */ | |
270 | STATIC int /* error code or 0 */ | |
271 | xfs_ialloc_ag_alloc( | |
272 | xfs_trans_t *tp, /* transaction pointer */ | |
273 | xfs_buf_t *agbp, /* alloc group buffer */ | |
274 | int *alloc) | |
275 | { | |
276 | xfs_agi_t *agi; /* allocation group header */ | |
277 | xfs_alloc_arg_t args; /* allocation argument structure */ | |
1da177e4 | 278 | xfs_btree_cur_t *cur; /* inode btree cursor */ |
92821e2b | 279 | xfs_agnumber_t agno; |
1da177e4 | 280 | int error; |
85c0b2ab | 281 | int i; |
1da177e4 LT |
282 | xfs_agino_t newino; /* new first inode's number */ |
283 | xfs_agino_t newlen; /* new number of inodes */ | |
1da177e4 | 284 | xfs_agino_t thisino; /* current inode number, for loop */ |
3ccb8b5f | 285 | int isaligned = 0; /* inode allocation at stripe unit */ |
1da177e4 | 286 | /* boundary */ |
44b56e0a | 287 | struct xfs_perag *pag; |
1da177e4 | 288 | |
a0041684 | 289 | memset(&args, 0, sizeof(args)); |
1da177e4 LT |
290 | args.tp = tp; |
291 | args.mp = tp->t_mountp; | |
292 | ||
293 | /* | |
294 | * Locking will ensure that we don't have two callers in here | |
295 | * at one time. | |
296 | */ | |
297 | newlen = XFS_IALLOC_INODES(args.mp); | |
298 | if (args.mp->m_maxicount && | |
299 | args.mp->m_sb.sb_icount + newlen > args.mp->m_maxicount) | |
300 | return XFS_ERROR(ENOSPC); | |
301 | args.minlen = args.maxlen = XFS_IALLOC_BLOCKS(args.mp); | |
302 | /* | |
3ccb8b5f GO |
303 | * First try to allocate inodes contiguous with the last-allocated |
304 | * chunk of inodes. If the filesystem is striped, this will fill | |
305 | * an entire stripe unit with inodes. | |
306 | */ | |
1da177e4 | 307 | agi = XFS_BUF_TO_AGI(agbp); |
3ccb8b5f | 308 | newino = be32_to_cpu(agi->agi_newino); |
85c0b2ab | 309 | agno = be32_to_cpu(agi->agi_seqno); |
019ff2d5 NS |
310 | args.agbno = XFS_AGINO_TO_AGBNO(args.mp, newino) + |
311 | XFS_IALLOC_BLOCKS(args.mp); | |
312 | if (likely(newino != NULLAGINO && | |
313 | (args.agbno < be32_to_cpu(agi->agi_length)))) { | |
85c0b2ab | 314 | args.fsbno = XFS_AGB_TO_FSB(args.mp, agno, args.agbno); |
3ccb8b5f | 315 | args.type = XFS_ALLOCTYPE_THIS_BNO; |
3ccb8b5f | 316 | args.prod = 1; |
75de2a91 | 317 | |
3ccb8b5f | 318 | /* |
75de2a91 DC |
319 | * We need to take into account alignment here to ensure that |
320 | * we don't modify the free list if we fail to have an exact | |
321 | * block. If we don't have an exact match, and every oher | |
322 | * attempt allocation attempt fails, we'll end up cancelling | |
323 | * a dirty transaction and shutting down. | |
324 | * | |
325 | * For an exact allocation, alignment must be 1, | |
326 | * however we need to take cluster alignment into account when | |
327 | * fixing up the freelist. Use the minalignslop field to | |
328 | * indicate that extra blocks might be required for alignment, | |
329 | * but not to use them in the actual exact allocation. | |
3ccb8b5f | 330 | */ |
75de2a91 DC |
331 | args.alignment = 1; |
332 | args.minalignslop = xfs_ialloc_cluster_alignment(&args) - 1; | |
333 | ||
334 | /* Allow space for the inode btree to split. */ | |
0d87e656 | 335 | args.minleft = args.mp->m_in_maxlevels - 1; |
3ccb8b5f GO |
336 | if ((error = xfs_alloc_vextent(&args))) |
337 | return error; | |
338 | } else | |
339 | args.fsbno = NULLFSBLOCK; | |
1da177e4 | 340 | |
3ccb8b5f GO |
341 | if (unlikely(args.fsbno == NULLFSBLOCK)) { |
342 | /* | |
343 | * Set the alignment for the allocation. | |
344 | * If stripe alignment is turned on then align at stripe unit | |
345 | * boundary. | |
019ff2d5 NS |
346 | * If the cluster size is smaller than a filesystem block |
347 | * then we're doing I/O for inodes in filesystem block size | |
3ccb8b5f GO |
348 | * pieces, so don't need alignment anyway. |
349 | */ | |
350 | isaligned = 0; | |
351 | if (args.mp->m_sinoalign) { | |
352 | ASSERT(!(args.mp->m_flags & XFS_MOUNT_NOALIGN)); | |
353 | args.alignment = args.mp->m_dalign; | |
354 | isaligned = 1; | |
75de2a91 DC |
355 | } else |
356 | args.alignment = xfs_ialloc_cluster_alignment(&args); | |
3ccb8b5f GO |
357 | /* |
358 | * Need to figure out where to allocate the inode blocks. | |
359 | * Ideally they should be spaced out through the a.g. | |
360 | * For now, just allocate blocks up front. | |
361 | */ | |
362 | args.agbno = be32_to_cpu(agi->agi_root); | |
85c0b2ab | 363 | args.fsbno = XFS_AGB_TO_FSB(args.mp, agno, args.agbno); |
3ccb8b5f GO |
364 | /* |
365 | * Allocate a fixed-size extent of inodes. | |
366 | */ | |
367 | args.type = XFS_ALLOCTYPE_NEAR_BNO; | |
3ccb8b5f GO |
368 | args.prod = 1; |
369 | /* | |
370 | * Allow space for the inode btree to split. | |
371 | */ | |
0d87e656 | 372 | args.minleft = args.mp->m_in_maxlevels - 1; |
3ccb8b5f GO |
373 | if ((error = xfs_alloc_vextent(&args))) |
374 | return error; | |
375 | } | |
019ff2d5 | 376 | |
1da177e4 LT |
377 | /* |
378 | * If stripe alignment is turned on, then try again with cluster | |
379 | * alignment. | |
380 | */ | |
381 | if (isaligned && args.fsbno == NULLFSBLOCK) { | |
382 | args.type = XFS_ALLOCTYPE_NEAR_BNO; | |
16259e7d | 383 | args.agbno = be32_to_cpu(agi->agi_root); |
85c0b2ab | 384 | args.fsbno = XFS_AGB_TO_FSB(args.mp, agno, args.agbno); |
75de2a91 | 385 | args.alignment = xfs_ialloc_cluster_alignment(&args); |
1da177e4 LT |
386 | if ((error = xfs_alloc_vextent(&args))) |
387 | return error; | |
388 | } | |
389 | ||
390 | if (args.fsbno == NULLFSBLOCK) { | |
391 | *alloc = 0; | |
392 | return 0; | |
393 | } | |
394 | ASSERT(args.len == args.minlen); | |
1da177e4 | 395 | |
359346a9 | 396 | /* |
85c0b2ab DC |
397 | * Stamp and write the inode buffers. |
398 | * | |
359346a9 DC |
399 | * Seed the new inode cluster with a random generation number. This |
400 | * prevents short-term reuse of generation numbers if a chunk is | |
401 | * freed and then immediately reallocated. We use random numbers | |
402 | * rather than a linear progression to prevent the next generation | |
403 | * number from being easily guessable. | |
404 | */ | |
2a30f36d | 405 | error = xfs_ialloc_inode_init(args.mp, tp, agno, args.agbno, |
ecb3403d | 406 | args.len, prandom_u32()); |
d42f08f6 | 407 | |
2a30f36d CS |
408 | if (error) |
409 | return error; | |
85c0b2ab DC |
410 | /* |
411 | * Convert the results. | |
412 | */ | |
413 | newino = XFS_OFFBNO_TO_AGINO(args.mp, args.agbno, 0); | |
413d57c9 MS |
414 | be32_add_cpu(&agi->agi_count, newlen); |
415 | be32_add_cpu(&agi->agi_freecount, newlen); | |
44b56e0a DC |
416 | pag = xfs_perag_get(args.mp, agno); |
417 | pag->pagi_freecount += newlen; | |
418 | xfs_perag_put(pag); | |
16259e7d | 419 | agi->agi_newino = cpu_to_be32(newino); |
85c0b2ab | 420 | |
1da177e4 LT |
421 | /* |
422 | * Insert records describing the new inode chunk into the btree. | |
423 | */ | |
561f7d17 | 424 | cur = xfs_inobt_init_cursor(args.mp, tp, agbp, agno); |
1da177e4 LT |
425 | for (thisino = newino; |
426 | thisino < newino + newlen; | |
427 | thisino += XFS_INODES_PER_CHUNK) { | |
21875505 CH |
428 | cur->bc_rec.i.ir_startino = thisino; |
429 | cur->bc_rec.i.ir_freecount = XFS_INODES_PER_CHUNK; | |
430 | cur->bc_rec.i.ir_free = XFS_INOBT_ALL_FREE; | |
431 | error = xfs_btree_lookup(cur, XFS_LOOKUP_EQ, &i); | |
432 | if (error) { | |
1da177e4 LT |
433 | xfs_btree_del_cursor(cur, XFS_BTREE_ERROR); |
434 | return error; | |
435 | } | |
436 | ASSERT(i == 0); | |
21875505 CH |
437 | error = xfs_btree_insert(cur, &i); |
438 | if (error) { | |
1da177e4 LT |
439 | xfs_btree_del_cursor(cur, XFS_BTREE_ERROR); |
440 | return error; | |
441 | } | |
442 | ASSERT(i == 1); | |
443 | } | |
444 | xfs_btree_del_cursor(cur, XFS_BTREE_NOERROR); | |
445 | /* | |
446 | * Log allocation group header fields | |
447 | */ | |
448 | xfs_ialloc_log_agi(tp, agbp, | |
449 | XFS_AGI_COUNT | XFS_AGI_FREECOUNT | XFS_AGI_NEWINO); | |
450 | /* | |
451 | * Modify/log superblock values for inode count and inode free count. | |
452 | */ | |
453 | xfs_trans_mod_sb(tp, XFS_TRANS_SB_ICOUNT, (long)newlen); | |
454 | xfs_trans_mod_sb(tp, XFS_TRANS_SB_IFREE, (long)newlen); | |
455 | *alloc = 1; | |
456 | return 0; | |
457 | } | |
458 | ||
b8f82a4a | 459 | STATIC xfs_agnumber_t |
1da177e4 LT |
460 | xfs_ialloc_next_ag( |
461 | xfs_mount_t *mp) | |
462 | { | |
463 | xfs_agnumber_t agno; | |
464 | ||
465 | spin_lock(&mp->m_agirotor_lock); | |
466 | agno = mp->m_agirotor; | |
8aea3ff4 | 467 | if (++mp->m_agirotor >= mp->m_maxagi) |
1da177e4 LT |
468 | mp->m_agirotor = 0; |
469 | spin_unlock(&mp->m_agirotor_lock); | |
470 | ||
471 | return agno; | |
472 | } | |
473 | ||
474 | /* | |
475 | * Select an allocation group to look for a free inode in, based on the parent | |
476 | * inode and then mode. Return the allocation group buffer. | |
477 | */ | |
55d6af64 | 478 | STATIC xfs_agnumber_t |
1da177e4 LT |
479 | xfs_ialloc_ag_select( |
480 | xfs_trans_t *tp, /* transaction pointer */ | |
481 | xfs_ino_t parent, /* parent directory inode number */ | |
576b1d67 | 482 | umode_t mode, /* bits set to indicate file type */ |
1da177e4 LT |
483 | int okalloc) /* ok to allocate more space */ |
484 | { | |
1da177e4 LT |
485 | xfs_agnumber_t agcount; /* number of ag's in the filesystem */ |
486 | xfs_agnumber_t agno; /* current ag number */ | |
487 | int flags; /* alloc buffer locking flags */ | |
488 | xfs_extlen_t ineed; /* blocks needed for inode allocation */ | |
489 | xfs_extlen_t longest = 0; /* longest extent available */ | |
490 | xfs_mount_t *mp; /* mount point structure */ | |
491 | int needspace; /* file mode implies space allocated */ | |
492 | xfs_perag_t *pag; /* per allocation group data */ | |
493 | xfs_agnumber_t pagno; /* parent (starting) ag number */ | |
55d6af64 | 494 | int error; |
1da177e4 LT |
495 | |
496 | /* | |
497 | * Files of these types need at least one block if length > 0 | |
498 | * (and they won't fit in the inode, but that's hard to figure out). | |
499 | */ | |
500 | needspace = S_ISDIR(mode) || S_ISREG(mode) || S_ISLNK(mode); | |
501 | mp = tp->t_mountp; | |
502 | agcount = mp->m_maxagi; | |
503 | if (S_ISDIR(mode)) | |
504 | pagno = xfs_ialloc_next_ag(mp); | |
505 | else { | |
506 | pagno = XFS_INO_TO_AGNO(mp, parent); | |
507 | if (pagno >= agcount) | |
508 | pagno = 0; | |
509 | } | |
55d6af64 | 510 | |
1da177e4 | 511 | ASSERT(pagno < agcount); |
55d6af64 | 512 | |
1da177e4 LT |
513 | /* |
514 | * Loop through allocation groups, looking for one with a little | |
515 | * free space in it. Note we don't look for free inodes, exactly. | |
516 | * Instead, we include whether there is a need to allocate inodes | |
517 | * to mean that blocks must be allocated for them, | |
518 | * if none are currently free. | |
519 | */ | |
520 | agno = pagno; | |
521 | flags = XFS_ALLOC_FLAG_TRYLOCK; | |
1da177e4 | 522 | for (;;) { |
44b56e0a | 523 | pag = xfs_perag_get(mp, agno); |
55d6af64 CH |
524 | if (!pag->pagi_inodeok) { |
525 | xfs_ialloc_next_ag(mp); | |
526 | goto nextag; | |
527 | } | |
528 | ||
1da177e4 | 529 | if (!pag->pagi_init) { |
55d6af64 CH |
530 | error = xfs_ialloc_pagi_init(mp, tp, agno); |
531 | if (error) | |
1da177e4 | 532 | goto nextag; |
55d6af64 | 533 | } |
1da177e4 | 534 | |
55d6af64 CH |
535 | if (pag->pagi_freecount) { |
536 | xfs_perag_put(pag); | |
537 | return agno; | |
1da177e4 LT |
538 | } |
539 | ||
55d6af64 CH |
540 | if (!okalloc) |
541 | goto nextag; | |
542 | ||
543 | if (!pag->pagf_init) { | |
544 | error = xfs_alloc_pagf_init(mp, tp, agno, flags); | |
545 | if (error) | |
1da177e4 | 546 | goto nextag; |
1da177e4 | 547 | } |
55d6af64 CH |
548 | |
549 | /* | |
550 | * Is there enough free space for the file plus a block of | |
551 | * inodes? (if we need to allocate some)? | |
552 | */ | |
553 | ineed = XFS_IALLOC_BLOCKS(mp); | |
554 | longest = pag->pagf_longest; | |
555 | if (!longest) | |
556 | longest = pag->pagf_flcount > 0; | |
557 | ||
558 | if (pag->pagf_freeblks >= needspace + ineed && | |
559 | longest >= ineed) { | |
560 | xfs_perag_put(pag); | |
561 | return agno; | |
1da177e4 | 562 | } |
1da177e4 | 563 | nextag: |
44b56e0a | 564 | xfs_perag_put(pag); |
1da177e4 LT |
565 | /* |
566 | * No point in iterating over the rest, if we're shutting | |
567 | * down. | |
568 | */ | |
1c1c6ebc | 569 | if (XFS_FORCED_SHUTDOWN(mp)) |
55d6af64 | 570 | return NULLAGNUMBER; |
1da177e4 LT |
571 | agno++; |
572 | if (agno >= agcount) | |
573 | agno = 0; | |
574 | if (agno == pagno) { | |
1c1c6ebc | 575 | if (flags == 0) |
55d6af64 | 576 | return NULLAGNUMBER; |
1da177e4 LT |
577 | flags = 0; |
578 | } | |
579 | } | |
580 | } | |
581 | ||
4254b0bb CH |
582 | /* |
583 | * Try to retrieve the next record to the left/right from the current one. | |
584 | */ | |
585 | STATIC int | |
586 | xfs_ialloc_next_rec( | |
587 | struct xfs_btree_cur *cur, | |
588 | xfs_inobt_rec_incore_t *rec, | |
589 | int *done, | |
590 | int left) | |
591 | { | |
592 | int error; | |
593 | int i; | |
594 | ||
595 | if (left) | |
596 | error = xfs_btree_decrement(cur, 0, &i); | |
597 | else | |
598 | error = xfs_btree_increment(cur, 0, &i); | |
599 | ||
600 | if (error) | |
601 | return error; | |
602 | *done = !i; | |
603 | if (i) { | |
604 | error = xfs_inobt_get_rec(cur, rec, &i); | |
605 | if (error) | |
606 | return error; | |
607 | XFS_WANT_CORRUPTED_RETURN(i == 1); | |
608 | } | |
609 | ||
610 | return 0; | |
611 | } | |
612 | ||
bd169565 DC |
613 | STATIC int |
614 | xfs_ialloc_get_rec( | |
615 | struct xfs_btree_cur *cur, | |
616 | xfs_agino_t agino, | |
617 | xfs_inobt_rec_incore_t *rec, | |
43df2ee6 | 618 | int *done) |
bd169565 DC |
619 | { |
620 | int error; | |
621 | int i; | |
622 | ||
623 | error = xfs_inobt_lookup(cur, agino, XFS_LOOKUP_EQ, &i); | |
624 | if (error) | |
625 | return error; | |
626 | *done = !i; | |
627 | if (i) { | |
628 | error = xfs_inobt_get_rec(cur, rec, &i); | |
629 | if (error) | |
630 | return error; | |
631 | XFS_WANT_CORRUPTED_RETURN(i == 1); | |
632 | } | |
633 | ||
634 | return 0; | |
635 | } | |
0b48db80 | 636 | |
1da177e4 | 637 | /* |
f2ecc5e4 | 638 | * Allocate an inode. |
1da177e4 | 639 | * |
f2ecc5e4 CH |
640 | * The caller selected an AG for us, and made sure that free inodes are |
641 | * available. | |
1da177e4 | 642 | */ |
f2ecc5e4 CH |
643 | STATIC int |
644 | xfs_dialloc_ag( | |
645 | struct xfs_trans *tp, | |
646 | struct xfs_buf *agbp, | |
647 | xfs_ino_t parent, | |
648 | xfs_ino_t *inop) | |
1da177e4 | 649 | { |
f2ecc5e4 CH |
650 | struct xfs_mount *mp = tp->t_mountp; |
651 | struct xfs_agi *agi = XFS_BUF_TO_AGI(agbp); | |
652 | xfs_agnumber_t agno = be32_to_cpu(agi->agi_seqno); | |
653 | xfs_agnumber_t pagno = XFS_INO_TO_AGNO(mp, parent); | |
654 | xfs_agino_t pagino = XFS_INO_TO_AGINO(mp, parent); | |
655 | struct xfs_perag *pag; | |
656 | struct xfs_btree_cur *cur, *tcur; | |
657 | struct xfs_inobt_rec_incore rec, trec; | |
658 | xfs_ino_t ino; | |
659 | int error; | |
660 | int offset; | |
661 | int i, j; | |
1da177e4 | 662 | |
44b56e0a | 663 | pag = xfs_perag_get(mp, agno); |
bd169565 | 664 | |
4bb61069 CH |
665 | ASSERT(pag->pagi_init); |
666 | ASSERT(pag->pagi_inodeok); | |
667 | ASSERT(pag->pagi_freecount > 0); | |
668 | ||
bd169565 | 669 | restart_pagno: |
f2ecc5e4 | 670 | cur = xfs_inobt_init_cursor(mp, tp, agbp, agno); |
1da177e4 LT |
671 | /* |
672 | * If pagino is 0 (this is the root inode allocation) use newino. | |
673 | * This must work because we've just allocated some. | |
674 | */ | |
675 | if (!pagino) | |
16259e7d | 676 | pagino = be32_to_cpu(agi->agi_newino); |
1da177e4 | 677 | |
0b48db80 DC |
678 | error = xfs_check_agi_freecount(cur, agi); |
679 | if (error) | |
680 | goto error0; | |
1da177e4 | 681 | |
1da177e4 | 682 | /* |
4254b0bb | 683 | * If in the same AG as the parent, try to get near the parent. |
1da177e4 LT |
684 | */ |
685 | if (pagno == agno) { | |
4254b0bb CH |
686 | int doneleft; /* done, to the left */ |
687 | int doneright; /* done, to the right */ | |
bd169565 | 688 | int searchdistance = 10; |
4254b0bb | 689 | |
21875505 | 690 | error = xfs_inobt_lookup(cur, pagino, XFS_LOOKUP_LE, &i); |
4254b0bb | 691 | if (error) |
1da177e4 | 692 | goto error0; |
4254b0bb CH |
693 | XFS_WANT_CORRUPTED_GOTO(i == 1, error0); |
694 | ||
695 | error = xfs_inobt_get_rec(cur, &rec, &j); | |
696 | if (error) | |
697 | goto error0; | |
698 | XFS_WANT_CORRUPTED_GOTO(i == 1, error0); | |
699 | ||
700 | if (rec.ir_freecount > 0) { | |
1da177e4 LT |
701 | /* |
702 | * Found a free inode in the same chunk | |
4254b0bb | 703 | * as the parent, done. |
1da177e4 | 704 | */ |
4254b0bb | 705 | goto alloc_inode; |
1da177e4 | 706 | } |
4254b0bb CH |
707 | |
708 | ||
1da177e4 | 709 | /* |
4254b0bb | 710 | * In the same AG as parent, but parent's chunk is full. |
1da177e4 | 711 | */ |
1da177e4 | 712 | |
4254b0bb CH |
713 | /* duplicate the cursor, search left & right simultaneously */ |
714 | error = xfs_btree_dup_cursor(cur, &tcur); | |
715 | if (error) | |
716 | goto error0; | |
717 | ||
bd169565 DC |
718 | /* |
719 | * Skip to last blocks looked up if same parent inode. | |
720 | */ | |
721 | if (pagino != NULLAGINO && | |
722 | pag->pagl_pagino == pagino && | |
723 | pag->pagl_leftrec != NULLAGINO && | |
724 | pag->pagl_rightrec != NULLAGINO) { | |
725 | error = xfs_ialloc_get_rec(tcur, pag->pagl_leftrec, | |
43df2ee6 | 726 | &trec, &doneleft); |
bd169565 DC |
727 | if (error) |
728 | goto error1; | |
4254b0bb | 729 | |
bd169565 | 730 | error = xfs_ialloc_get_rec(cur, pag->pagl_rightrec, |
43df2ee6 | 731 | &rec, &doneright); |
bd169565 DC |
732 | if (error) |
733 | goto error1; | |
734 | } else { | |
735 | /* search left with tcur, back up 1 record */ | |
736 | error = xfs_ialloc_next_rec(tcur, &trec, &doneleft, 1); | |
737 | if (error) | |
738 | goto error1; | |
739 | ||
740 | /* search right with cur, go forward 1 record. */ | |
741 | error = xfs_ialloc_next_rec(cur, &rec, &doneright, 0); | |
742 | if (error) | |
743 | goto error1; | |
744 | } | |
4254b0bb CH |
745 | |
746 | /* | |
747 | * Loop until we find an inode chunk with a free inode. | |
748 | */ | |
749 | while (!doneleft || !doneright) { | |
750 | int useleft; /* using left inode chunk this time */ | |
751 | ||
bd169565 DC |
752 | if (!--searchdistance) { |
753 | /* | |
754 | * Not in range - save last search | |
755 | * location and allocate a new inode | |
756 | */ | |
3b826386 | 757 | xfs_btree_del_cursor(tcur, XFS_BTREE_NOERROR); |
bd169565 DC |
758 | pag->pagl_leftrec = trec.ir_startino; |
759 | pag->pagl_rightrec = rec.ir_startino; | |
760 | pag->pagl_pagino = pagino; | |
761 | goto newino; | |
762 | } | |
763 | ||
4254b0bb CH |
764 | /* figure out the closer block if both are valid. */ |
765 | if (!doneleft && !doneright) { | |
766 | useleft = pagino - | |
767 | (trec.ir_startino + XFS_INODES_PER_CHUNK - 1) < | |
768 | rec.ir_startino - pagino; | |
769 | } else { | |
770 | useleft = !doneleft; | |
1da177e4 | 771 | } |
4254b0bb CH |
772 | |
773 | /* free inodes to the left? */ | |
774 | if (useleft && trec.ir_freecount) { | |
775 | rec = trec; | |
776 | xfs_btree_del_cursor(cur, XFS_BTREE_NOERROR); | |
777 | cur = tcur; | |
bd169565 DC |
778 | |
779 | pag->pagl_leftrec = trec.ir_startino; | |
780 | pag->pagl_rightrec = rec.ir_startino; | |
781 | pag->pagl_pagino = pagino; | |
4254b0bb | 782 | goto alloc_inode; |
1da177e4 | 783 | } |
1da177e4 | 784 | |
4254b0bb CH |
785 | /* free inodes to the right? */ |
786 | if (!useleft && rec.ir_freecount) { | |
787 | xfs_btree_del_cursor(tcur, XFS_BTREE_NOERROR); | |
bd169565 DC |
788 | |
789 | pag->pagl_leftrec = trec.ir_startino; | |
790 | pag->pagl_rightrec = rec.ir_startino; | |
791 | pag->pagl_pagino = pagino; | |
4254b0bb | 792 | goto alloc_inode; |
1da177e4 | 793 | } |
4254b0bb CH |
794 | |
795 | /* get next record to check */ | |
796 | if (useleft) { | |
797 | error = xfs_ialloc_next_rec(tcur, &trec, | |
798 | &doneleft, 1); | |
799 | } else { | |
800 | error = xfs_ialloc_next_rec(cur, &rec, | |
801 | &doneright, 0); | |
802 | } | |
803 | if (error) | |
804 | goto error1; | |
1da177e4 | 805 | } |
bd169565 DC |
806 | |
807 | /* | |
808 | * We've reached the end of the btree. because | |
809 | * we are only searching a small chunk of the | |
810 | * btree each search, there is obviously free | |
811 | * inodes closer to the parent inode than we | |
812 | * are now. restart the search again. | |
813 | */ | |
814 | pag->pagl_pagino = NULLAGINO; | |
815 | pag->pagl_leftrec = NULLAGINO; | |
816 | pag->pagl_rightrec = NULLAGINO; | |
817 | xfs_btree_del_cursor(tcur, XFS_BTREE_NOERROR); | |
818 | xfs_btree_del_cursor(cur, XFS_BTREE_NOERROR); | |
819 | goto restart_pagno; | |
1da177e4 | 820 | } |
4254b0bb | 821 | |
1da177e4 | 822 | /* |
4254b0bb | 823 | * In a different AG from the parent. |
1da177e4 LT |
824 | * See if the most recently allocated block has any free. |
825 | */ | |
bd169565 | 826 | newino: |
69ef921b | 827 | if (agi->agi_newino != cpu_to_be32(NULLAGINO)) { |
21875505 CH |
828 | error = xfs_inobt_lookup(cur, be32_to_cpu(agi->agi_newino), |
829 | XFS_LOOKUP_EQ, &i); | |
4254b0bb | 830 | if (error) |
1da177e4 | 831 | goto error0; |
4254b0bb CH |
832 | |
833 | if (i == 1) { | |
834 | error = xfs_inobt_get_rec(cur, &rec, &j); | |
835 | if (error) | |
836 | goto error0; | |
837 | ||
838 | if (j == 1 && rec.ir_freecount > 0) { | |
839 | /* | |
840 | * The last chunk allocated in the group | |
841 | * still has a free inode. | |
842 | */ | |
843 | goto alloc_inode; | |
844 | } | |
1da177e4 | 845 | } |
bd169565 | 846 | } |
4254b0bb | 847 | |
bd169565 DC |
848 | /* |
849 | * None left in the last group, search the whole AG | |
850 | */ | |
851 | error = xfs_inobt_lookup(cur, 0, XFS_LOOKUP_GE, &i); | |
852 | if (error) | |
853 | goto error0; | |
854 | XFS_WANT_CORRUPTED_GOTO(i == 1, error0); | |
855 | ||
856 | for (;;) { | |
857 | error = xfs_inobt_get_rec(cur, &rec, &i); | |
858 | if (error) | |
859 | goto error0; | |
860 | XFS_WANT_CORRUPTED_GOTO(i == 1, error0); | |
861 | if (rec.ir_freecount > 0) | |
862 | break; | |
863 | error = xfs_btree_increment(cur, 0, &i); | |
4254b0bb CH |
864 | if (error) |
865 | goto error0; | |
866 | XFS_WANT_CORRUPTED_GOTO(i == 1, error0); | |
1da177e4 | 867 | } |
4254b0bb CH |
868 | |
869 | alloc_inode: | |
824c3131 | 870 | offset = xfs_lowbit64(rec.ir_free); |
1da177e4 LT |
871 | ASSERT(offset >= 0); |
872 | ASSERT(offset < XFS_INODES_PER_CHUNK); | |
873 | ASSERT((XFS_AGINO_TO_OFFSET(mp, rec.ir_startino) % | |
874 | XFS_INODES_PER_CHUNK) == 0); | |
875 | ino = XFS_AGINO_TO_INO(mp, agno, rec.ir_startino + offset); | |
0d87e656 | 876 | rec.ir_free &= ~XFS_INOBT_MASK(offset); |
1da177e4 | 877 | rec.ir_freecount--; |
afabc24a CH |
878 | error = xfs_inobt_update(cur, &rec); |
879 | if (error) | |
1da177e4 | 880 | goto error0; |
413d57c9 | 881 | be32_add_cpu(&agi->agi_freecount, -1); |
1da177e4 | 882 | xfs_ialloc_log_agi(tp, agbp, XFS_AGI_FREECOUNT); |
44b56e0a | 883 | pag->pagi_freecount--; |
1da177e4 | 884 | |
0b48db80 DC |
885 | error = xfs_check_agi_freecount(cur, agi); |
886 | if (error) | |
887 | goto error0; | |
888 | ||
1da177e4 LT |
889 | xfs_btree_del_cursor(cur, XFS_BTREE_NOERROR); |
890 | xfs_trans_mod_sb(tp, XFS_TRANS_SB_IFREE, -1); | |
44b56e0a | 891 | xfs_perag_put(pag); |
1da177e4 LT |
892 | *inop = ino; |
893 | return 0; | |
894 | error1: | |
895 | xfs_btree_del_cursor(tcur, XFS_BTREE_ERROR); | |
896 | error0: | |
897 | xfs_btree_del_cursor(cur, XFS_BTREE_ERROR); | |
44b56e0a | 898 | xfs_perag_put(pag); |
1da177e4 LT |
899 | return error; |
900 | } | |
901 | ||
f2ecc5e4 CH |
902 | /* |
903 | * Allocate an inode on disk. | |
904 | * | |
905 | * Mode is used to tell whether the new inode will need space, and whether it | |
906 | * is a directory. | |
907 | * | |
908 | * This function is designed to be called twice if it has to do an allocation | |
909 | * to make more free inodes. On the first call, *IO_agbp should be set to NULL. | |
910 | * If an inode is available without having to performn an allocation, an inode | |
cd856db6 CM |
911 | * number is returned. In this case, *IO_agbp is set to NULL. If an allocation |
912 | * needs to be done, xfs_dialloc returns the current AGI buffer in *IO_agbp. | |
913 | * The caller should then commit the current transaction, allocate a | |
f2ecc5e4 CH |
914 | * new transaction, and call xfs_dialloc() again, passing in the previous value |
915 | * of *IO_agbp. IO_agbp should be held across the transactions. Since the AGI | |
916 | * buffer is locked across the two calls, the second call is guaranteed to have | |
917 | * a free inode available. | |
918 | * | |
919 | * Once we successfully pick an inode its number is returned and the on-disk | |
920 | * data structures are updated. The inode itself is not read in, since doing so | |
921 | * would break ordering constraints with xfs_reclaim. | |
922 | */ | |
923 | int | |
924 | xfs_dialloc( | |
925 | struct xfs_trans *tp, | |
926 | xfs_ino_t parent, | |
927 | umode_t mode, | |
928 | int okalloc, | |
929 | struct xfs_buf **IO_agbp, | |
f2ecc5e4 CH |
930 | xfs_ino_t *inop) |
931 | { | |
55d6af64 | 932 | struct xfs_mount *mp = tp->t_mountp; |
f2ecc5e4 CH |
933 | struct xfs_buf *agbp; |
934 | xfs_agnumber_t agno; | |
f2ecc5e4 CH |
935 | int error; |
936 | int ialloced; | |
937 | int noroom = 0; | |
be60fe54 | 938 | xfs_agnumber_t start_agno; |
f2ecc5e4 CH |
939 | struct xfs_perag *pag; |
940 | ||
4bb61069 | 941 | if (*IO_agbp) { |
f2ecc5e4 | 942 | /* |
4bb61069 CH |
943 | * If the caller passes in a pointer to the AGI buffer, |
944 | * continue where we left off before. In this case, we | |
f2ecc5e4 CH |
945 | * know that the allocation group has free inodes. |
946 | */ | |
947 | agbp = *IO_agbp; | |
4bb61069 | 948 | goto out_alloc; |
f2ecc5e4 | 949 | } |
4bb61069 CH |
950 | |
951 | /* | |
952 | * We do not have an agbp, so select an initial allocation | |
953 | * group for inode allocation. | |
954 | */ | |
be60fe54 CH |
955 | start_agno = xfs_ialloc_ag_select(tp, parent, mode, okalloc); |
956 | if (start_agno == NULLAGNUMBER) { | |
4bb61069 CH |
957 | *inop = NULLFSINO; |
958 | return 0; | |
959 | } | |
55d6af64 | 960 | |
f2ecc5e4 CH |
961 | /* |
962 | * If we have already hit the ceiling of inode blocks then clear | |
963 | * okalloc so we scan all available agi structures for a free | |
964 | * inode. | |
965 | */ | |
f2ecc5e4 CH |
966 | if (mp->m_maxicount && |
967 | mp->m_sb.sb_icount + XFS_IALLOC_INODES(mp) > mp->m_maxicount) { | |
968 | noroom = 1; | |
969 | okalloc = 0; | |
970 | } | |
971 | ||
972 | /* | |
973 | * Loop until we find an allocation group that either has free inodes | |
974 | * or in which we can allocate some inodes. Iterate through the | |
975 | * allocation groups upward, wrapping at the end. | |
976 | */ | |
be60fe54 CH |
977 | agno = start_agno; |
978 | for (;;) { | |
979 | pag = xfs_perag_get(mp, agno); | |
980 | if (!pag->pagi_inodeok) { | |
981 | xfs_ialloc_next_ag(mp); | |
982 | goto nextag; | |
983 | } | |
984 | ||
985 | if (!pag->pagi_init) { | |
986 | error = xfs_ialloc_pagi_init(mp, tp, agno); | |
987 | if (error) | |
988 | goto out_error; | |
f2ecc5e4 | 989 | } |
be60fe54 | 990 | |
f2ecc5e4 | 991 | /* |
be60fe54 | 992 | * Do a first racy fast path check if this AG is usable. |
f2ecc5e4 | 993 | */ |
be60fe54 CH |
994 | if (!pag->pagi_freecount && !okalloc) |
995 | goto nextag; | |
996 | ||
c4982110 CH |
997 | /* |
998 | * Then read in the AGI buffer and recheck with the AGI buffer | |
999 | * lock held. | |
1000 | */ | |
be60fe54 CH |
1001 | error = xfs_ialloc_read_agi(mp, tp, agno, &agbp); |
1002 | if (error) | |
1003 | goto out_error; | |
1004 | ||
be60fe54 CH |
1005 | if (pag->pagi_freecount) { |
1006 | xfs_perag_put(pag); | |
1007 | goto out_alloc; | |
1008 | } | |
1009 | ||
c4982110 CH |
1010 | if (!okalloc) |
1011 | goto nextag_relse_buffer; | |
1012 | ||
be60fe54 CH |
1013 | |
1014 | error = xfs_ialloc_ag_alloc(tp, agbp, &ialloced); | |
1015 | if (error) { | |
1016 | xfs_trans_brelse(tp, agbp); | |
1017 | ||
1018 | if (error != ENOSPC) | |
1019 | goto out_error; | |
1020 | ||
1021 | xfs_perag_put(pag); | |
f2ecc5e4 | 1022 | *inop = NULLFSINO; |
be60fe54 | 1023 | return 0; |
f2ecc5e4 | 1024 | } |
be60fe54 CH |
1025 | |
1026 | if (ialloced) { | |
1027 | /* | |
1028 | * We successfully allocated some inodes, return | |
1029 | * the current context to the caller so that it | |
1030 | * can commit the current transaction and call | |
1031 | * us again where we left off. | |
1032 | */ | |
1033 | ASSERT(pag->pagi_freecount > 0); | |
f2ecc5e4 | 1034 | xfs_perag_put(pag); |
be60fe54 CH |
1035 | |
1036 | *IO_agbp = agbp; | |
1037 | *inop = NULLFSINO; | |
1038 | return 0; | |
f2ecc5e4 | 1039 | } |
be60fe54 | 1040 | |
c4982110 CH |
1041 | nextag_relse_buffer: |
1042 | xfs_trans_brelse(tp, agbp); | |
be60fe54 | 1043 | nextag: |
f2ecc5e4 | 1044 | xfs_perag_put(pag); |
be60fe54 CH |
1045 | if (++agno == mp->m_sb.sb_agcount) |
1046 | agno = 0; | |
1047 | if (agno == start_agno) { | |
1048 | *inop = NULLFSINO; | |
1049 | return noroom ? ENOSPC : 0; | |
1050 | } | |
f2ecc5e4 CH |
1051 | } |
1052 | ||
4bb61069 | 1053 | out_alloc: |
f2ecc5e4 CH |
1054 | *IO_agbp = NULL; |
1055 | return xfs_dialloc_ag(tp, agbp, parent, inop); | |
be60fe54 CH |
1056 | out_error: |
1057 | xfs_perag_put(pag); | |
1058 | return XFS_ERROR(error); | |
f2ecc5e4 CH |
1059 | } |
1060 | ||
1da177e4 LT |
1061 | /* |
1062 | * Free disk inode. Carefully avoids touching the incore inode, all | |
1063 | * manipulations incore are the caller's responsibility. | |
1064 | * The on-disk inode is not changed by this operation, only the | |
1065 | * btree (free inode mask) is changed. | |
1066 | */ | |
1067 | int | |
1068 | xfs_difree( | |
1069 | xfs_trans_t *tp, /* transaction pointer */ | |
1070 | xfs_ino_t inode, /* inode to be freed */ | |
1071 | xfs_bmap_free_t *flist, /* extents to free */ | |
1072 | int *delete, /* set if inode cluster was deleted */ | |
1073 | xfs_ino_t *first_ino) /* first inode in deleted cluster */ | |
1074 | { | |
1075 | /* REFERENCED */ | |
1076 | xfs_agblock_t agbno; /* block number containing inode */ | |
1077 | xfs_buf_t *agbp; /* buffer containing allocation group header */ | |
1078 | xfs_agino_t agino; /* inode number relative to allocation group */ | |
1079 | xfs_agnumber_t agno; /* allocation group number */ | |
1080 | xfs_agi_t *agi; /* allocation group header */ | |
1081 | xfs_btree_cur_t *cur; /* inode btree cursor */ | |
1082 | int error; /* error return value */ | |
1083 | int i; /* result code */ | |
1084 | int ilen; /* inodes in an inode cluster */ | |
1085 | xfs_mount_t *mp; /* mount structure for filesystem */ | |
1086 | int off; /* offset of inode in inode chunk */ | |
61a25848 | 1087 | xfs_inobt_rec_incore_t rec; /* btree record */ |
44b56e0a | 1088 | struct xfs_perag *pag; |
1da177e4 LT |
1089 | |
1090 | mp = tp->t_mountp; | |
1091 | ||
1092 | /* | |
1093 | * Break up inode number into its components. | |
1094 | */ | |
1095 | agno = XFS_INO_TO_AGNO(mp, inode); | |
1096 | if (agno >= mp->m_sb.sb_agcount) { | |
0b932ccc DC |
1097 | xfs_warn(mp, "%s: agno >= mp->m_sb.sb_agcount (%d >= %d).", |
1098 | __func__, agno, mp->m_sb.sb_agcount); | |
1da177e4 LT |
1099 | ASSERT(0); |
1100 | return XFS_ERROR(EINVAL); | |
1101 | } | |
1102 | agino = XFS_INO_TO_AGINO(mp, inode); | |
1103 | if (inode != XFS_AGINO_TO_INO(mp, agno, agino)) { | |
0b932ccc DC |
1104 | xfs_warn(mp, "%s: inode != XFS_AGINO_TO_INO() (%llu != %llu).", |
1105 | __func__, (unsigned long long)inode, | |
1106 | (unsigned long long)XFS_AGINO_TO_INO(mp, agno, agino)); | |
1da177e4 LT |
1107 | ASSERT(0); |
1108 | return XFS_ERROR(EINVAL); | |
1109 | } | |
1110 | agbno = XFS_AGINO_TO_AGBNO(mp, agino); | |
1111 | if (agbno >= mp->m_sb.sb_agblocks) { | |
0b932ccc DC |
1112 | xfs_warn(mp, "%s: agbno >= mp->m_sb.sb_agblocks (%d >= %d).", |
1113 | __func__, agbno, mp->m_sb.sb_agblocks); | |
1da177e4 LT |
1114 | ASSERT(0); |
1115 | return XFS_ERROR(EINVAL); | |
1116 | } | |
1117 | /* | |
1118 | * Get the allocation group header. | |
1119 | */ | |
1da177e4 | 1120 | error = xfs_ialloc_read_agi(mp, tp, agno, &agbp); |
1da177e4 | 1121 | if (error) { |
0b932ccc DC |
1122 | xfs_warn(mp, "%s: xfs_ialloc_read_agi() returned error %d.", |
1123 | __func__, error); | |
1da177e4 LT |
1124 | return error; |
1125 | } | |
1126 | agi = XFS_BUF_TO_AGI(agbp); | |
69ef921b | 1127 | ASSERT(agi->agi_magicnum == cpu_to_be32(XFS_AGI_MAGIC)); |
16259e7d | 1128 | ASSERT(agbno < be32_to_cpu(agi->agi_length)); |
1da177e4 LT |
1129 | /* |
1130 | * Initialize the cursor. | |
1131 | */ | |
561f7d17 | 1132 | cur = xfs_inobt_init_cursor(mp, tp, agbp, agno); |
1da177e4 | 1133 | |
0b48db80 DC |
1134 | error = xfs_check_agi_freecount(cur, agi); |
1135 | if (error) | |
1136 | goto error0; | |
1137 | ||
1da177e4 LT |
1138 | /* |
1139 | * Look for the entry describing this inode. | |
1140 | */ | |
21875505 | 1141 | if ((error = xfs_inobt_lookup(cur, agino, XFS_LOOKUP_LE, &i))) { |
0b932ccc DC |
1142 | xfs_warn(mp, "%s: xfs_inobt_lookup() returned error %d.", |
1143 | __func__, error); | |
1da177e4 LT |
1144 | goto error0; |
1145 | } | |
1146 | XFS_WANT_CORRUPTED_GOTO(i == 1, error0); | |
2e287a73 CH |
1147 | error = xfs_inobt_get_rec(cur, &rec, &i); |
1148 | if (error) { | |
0b932ccc DC |
1149 | xfs_warn(mp, "%s: xfs_inobt_get_rec() returned error %d.", |
1150 | __func__, error); | |
1da177e4 LT |
1151 | goto error0; |
1152 | } | |
1153 | XFS_WANT_CORRUPTED_GOTO(i == 1, error0); | |
1154 | /* | |
1155 | * Get the offset in the inode chunk. | |
1156 | */ | |
1157 | off = agino - rec.ir_startino; | |
1158 | ASSERT(off >= 0 && off < XFS_INODES_PER_CHUNK); | |
0d87e656 | 1159 | ASSERT(!(rec.ir_free & XFS_INOBT_MASK(off))); |
1da177e4 LT |
1160 | /* |
1161 | * Mark the inode free & increment the count. | |
1162 | */ | |
0d87e656 | 1163 | rec.ir_free |= XFS_INOBT_MASK(off); |
1da177e4 LT |
1164 | rec.ir_freecount++; |
1165 | ||
1166 | /* | |
c41564b5 | 1167 | * When an inode cluster is free, it becomes eligible for removal |
1da177e4 | 1168 | */ |
1bd960ee | 1169 | if (!(mp->m_flags & XFS_MOUNT_IKEEP) && |
1da177e4 LT |
1170 | (rec.ir_freecount == XFS_IALLOC_INODES(mp))) { |
1171 | ||
1172 | *delete = 1; | |
1173 | *first_ino = XFS_AGINO_TO_INO(mp, agno, rec.ir_startino); | |
1174 | ||
1175 | /* | |
1176 | * Remove the inode cluster from the AGI B+Tree, adjust the | |
1177 | * AGI and Superblock inode counts, and mark the disk space | |
1178 | * to be freed when the transaction is committed. | |
1179 | */ | |
1180 | ilen = XFS_IALLOC_INODES(mp); | |
413d57c9 MS |
1181 | be32_add_cpu(&agi->agi_count, -ilen); |
1182 | be32_add_cpu(&agi->agi_freecount, -(ilen - 1)); | |
1da177e4 | 1183 | xfs_ialloc_log_agi(tp, agbp, XFS_AGI_COUNT | XFS_AGI_FREECOUNT); |
44b56e0a DC |
1184 | pag = xfs_perag_get(mp, agno); |
1185 | pag->pagi_freecount -= ilen - 1; | |
1186 | xfs_perag_put(pag); | |
1da177e4 LT |
1187 | xfs_trans_mod_sb(tp, XFS_TRANS_SB_ICOUNT, -ilen); |
1188 | xfs_trans_mod_sb(tp, XFS_TRANS_SB_IFREE, -(ilen - 1)); | |
1189 | ||
91cca5df | 1190 | if ((error = xfs_btree_delete(cur, &i))) { |
0b932ccc DC |
1191 | xfs_warn(mp, "%s: xfs_btree_delete returned error %d.", |
1192 | __func__, error); | |
1da177e4 LT |
1193 | goto error0; |
1194 | } | |
1195 | ||
1196 | xfs_bmap_add_free(XFS_AGB_TO_FSB(mp, | |
1197 | agno, XFS_INO_TO_AGBNO(mp,rec.ir_startino)), | |
1198 | XFS_IALLOC_BLOCKS(mp), flist, mp); | |
1199 | } else { | |
1200 | *delete = 0; | |
1201 | ||
afabc24a CH |
1202 | error = xfs_inobt_update(cur, &rec); |
1203 | if (error) { | |
0b932ccc DC |
1204 | xfs_warn(mp, "%s: xfs_inobt_update returned error %d.", |
1205 | __func__, error); | |
1da177e4 LT |
1206 | goto error0; |
1207 | } | |
afabc24a | 1208 | |
1da177e4 LT |
1209 | /* |
1210 | * Change the inode free counts and log the ag/sb changes. | |
1211 | */ | |
413d57c9 | 1212 | be32_add_cpu(&agi->agi_freecount, 1); |
1da177e4 | 1213 | xfs_ialloc_log_agi(tp, agbp, XFS_AGI_FREECOUNT); |
44b56e0a DC |
1214 | pag = xfs_perag_get(mp, agno); |
1215 | pag->pagi_freecount++; | |
1216 | xfs_perag_put(pag); | |
1da177e4 LT |
1217 | xfs_trans_mod_sb(tp, XFS_TRANS_SB_IFREE, 1); |
1218 | } | |
1219 | ||
0b48db80 DC |
1220 | error = xfs_check_agi_freecount(cur, agi); |
1221 | if (error) | |
1222 | goto error0; | |
1da177e4 | 1223 | |
1da177e4 LT |
1224 | xfs_btree_del_cursor(cur, XFS_BTREE_NOERROR); |
1225 | return 0; | |
1226 | ||
1227 | error0: | |
1228 | xfs_btree_del_cursor(cur, XFS_BTREE_ERROR); | |
1229 | return error; | |
1230 | } | |
1231 | ||
7124fe0a DC |
1232 | STATIC int |
1233 | xfs_imap_lookup( | |
1234 | struct xfs_mount *mp, | |
1235 | struct xfs_trans *tp, | |
1236 | xfs_agnumber_t agno, | |
1237 | xfs_agino_t agino, | |
1238 | xfs_agblock_t agbno, | |
1239 | xfs_agblock_t *chunk_agbno, | |
1240 | xfs_agblock_t *offset_agbno, | |
1241 | int flags) | |
1242 | { | |
1243 | struct xfs_inobt_rec_incore rec; | |
1244 | struct xfs_btree_cur *cur; | |
1245 | struct xfs_buf *agbp; | |
7124fe0a DC |
1246 | int error; |
1247 | int i; | |
1248 | ||
1249 | error = xfs_ialloc_read_agi(mp, tp, agno, &agbp); | |
1250 | if (error) { | |
53487786 DC |
1251 | xfs_alert(mp, |
1252 | "%s: xfs_ialloc_read_agi() returned error %d, agno %d", | |
1253 | __func__, error, agno); | |
7124fe0a DC |
1254 | return error; |
1255 | } | |
1256 | ||
1257 | /* | |
4536f2ad DC |
1258 | * Lookup the inode record for the given agino. If the record cannot be |
1259 | * found, then it's an invalid inode number and we should abort. Once | |
1260 | * we have a record, we need to ensure it contains the inode number | |
1261 | * we are looking up. | |
7124fe0a DC |
1262 | */ |
1263 | cur = xfs_inobt_init_cursor(mp, tp, agbp, agno); | |
4536f2ad | 1264 | error = xfs_inobt_lookup(cur, agino, XFS_LOOKUP_LE, &i); |
7124fe0a DC |
1265 | if (!error) { |
1266 | if (i) | |
1267 | error = xfs_inobt_get_rec(cur, &rec, &i); | |
1268 | if (!error && i == 0) | |
1269 | error = EINVAL; | |
1270 | } | |
1271 | ||
1272 | xfs_trans_brelse(tp, agbp); | |
1273 | xfs_btree_del_cursor(cur, XFS_BTREE_NOERROR); | |
1274 | if (error) | |
1275 | return error; | |
1276 | ||
4536f2ad DC |
1277 | /* check that the returned record contains the required inode */ |
1278 | if (rec.ir_startino > agino || | |
1279 | rec.ir_startino + XFS_IALLOC_INODES(mp) <= agino) | |
1280 | return EINVAL; | |
1281 | ||
7124fe0a | 1282 | /* for untrusted inodes check it is allocated first */ |
1920779e | 1283 | if ((flags & XFS_IGET_UNTRUSTED) && |
7124fe0a DC |
1284 | (rec.ir_free & XFS_INOBT_MASK(agino - rec.ir_startino))) |
1285 | return EINVAL; | |
1286 | ||
1287 | *chunk_agbno = XFS_AGINO_TO_AGBNO(mp, rec.ir_startino); | |
1288 | *offset_agbno = agbno - *chunk_agbno; | |
1289 | return 0; | |
1290 | } | |
1291 | ||
1da177e4 | 1292 | /* |
94e1b69d | 1293 | * Return the location of the inode in imap, for mapping it into a buffer. |
1da177e4 | 1294 | */ |
1da177e4 | 1295 | int |
94e1b69d CH |
1296 | xfs_imap( |
1297 | xfs_mount_t *mp, /* file system mount structure */ | |
1298 | xfs_trans_t *tp, /* transaction pointer */ | |
1da177e4 | 1299 | xfs_ino_t ino, /* inode to locate */ |
94e1b69d CH |
1300 | struct xfs_imap *imap, /* location map structure */ |
1301 | uint flags) /* flags for inode btree lookup */ | |
1da177e4 LT |
1302 | { |
1303 | xfs_agblock_t agbno; /* block number of inode in the alloc group */ | |
1da177e4 LT |
1304 | xfs_agino_t agino; /* inode number within alloc group */ |
1305 | xfs_agnumber_t agno; /* allocation group number */ | |
1306 | int blks_per_cluster; /* num blocks per inode cluster */ | |
1307 | xfs_agblock_t chunk_agbno; /* first block in inode chunk */ | |
1da177e4 | 1308 | xfs_agblock_t cluster_agbno; /* first block in inode cluster */ |
1da177e4 | 1309 | int error; /* error code */ |
1da177e4 LT |
1310 | int offset; /* index of inode in its buffer */ |
1311 | int offset_agbno; /* blks from chunk start to inode */ | |
1312 | ||
1313 | ASSERT(ino != NULLFSINO); | |
94e1b69d | 1314 | |
1da177e4 LT |
1315 | /* |
1316 | * Split up the inode number into its parts. | |
1317 | */ | |
1318 | agno = XFS_INO_TO_AGNO(mp, ino); | |
1319 | agino = XFS_INO_TO_AGINO(mp, ino); | |
1320 | agbno = XFS_AGINO_TO_AGBNO(mp, agino); | |
1321 | if (agno >= mp->m_sb.sb_agcount || agbno >= mp->m_sb.sb_agblocks || | |
1322 | ino != XFS_AGINO_TO_INO(mp, agno, agino)) { | |
1323 | #ifdef DEBUG | |
1920779e DC |
1324 | /* |
1325 | * Don't output diagnostic information for untrusted inodes | |
1326 | * as they can be invalid without implying corruption. | |
1327 | */ | |
1328 | if (flags & XFS_IGET_UNTRUSTED) | |
4d1a2ed3 | 1329 | return XFS_ERROR(EINVAL); |
1da177e4 | 1330 | if (agno >= mp->m_sb.sb_agcount) { |
53487786 DC |
1331 | xfs_alert(mp, |
1332 | "%s: agno (%d) >= mp->m_sb.sb_agcount (%d)", | |
1333 | __func__, agno, mp->m_sb.sb_agcount); | |
1da177e4 LT |
1334 | } |
1335 | if (agbno >= mp->m_sb.sb_agblocks) { | |
53487786 DC |
1336 | xfs_alert(mp, |
1337 | "%s: agbno (0x%llx) >= mp->m_sb.sb_agblocks (0x%lx)", | |
1338 | __func__, (unsigned long long)agbno, | |
1339 | (unsigned long)mp->m_sb.sb_agblocks); | |
1da177e4 LT |
1340 | } |
1341 | if (ino != XFS_AGINO_TO_INO(mp, agno, agino)) { | |
53487786 DC |
1342 | xfs_alert(mp, |
1343 | "%s: ino (0x%llx) != XFS_AGINO_TO_INO() (0x%llx)", | |
1344 | __func__, ino, | |
1345 | XFS_AGINO_TO_INO(mp, agno, agino)); | |
1da177e4 | 1346 | } |
745b1f47 | 1347 | xfs_stack_trace(); |
1da177e4 LT |
1348 | #endif /* DEBUG */ |
1349 | return XFS_ERROR(EINVAL); | |
1350 | } | |
94e1b69d | 1351 | |
7124fe0a DC |
1352 | blks_per_cluster = XFS_INODE_CLUSTER_SIZE(mp) >> mp->m_sb.sb_blocklog; |
1353 | ||
1354 | /* | |
1355 | * For bulkstat and handle lookups, we have an untrusted inode number | |
1356 | * that we have to verify is valid. We cannot do this just by reading | |
1357 | * the inode buffer as it may have been unlinked and removed leaving | |
1358 | * inodes in stale state on disk. Hence we have to do a btree lookup | |
1359 | * in all cases where an untrusted inode number is passed. | |
1360 | */ | |
1920779e | 1361 | if (flags & XFS_IGET_UNTRUSTED) { |
7124fe0a DC |
1362 | error = xfs_imap_lookup(mp, tp, agno, agino, agbno, |
1363 | &chunk_agbno, &offset_agbno, flags); | |
1364 | if (error) | |
1365 | return error; | |
1366 | goto out_map; | |
1367 | } | |
1368 | ||
94e1b69d CH |
1369 | /* |
1370 | * If the inode cluster size is the same as the blocksize or | |
1371 | * smaller we get to the buffer by simple arithmetics. | |
1372 | */ | |
1373 | if (XFS_INODE_CLUSTER_SIZE(mp) <= mp->m_sb.sb_blocksize) { | |
1da177e4 LT |
1374 | offset = XFS_INO_TO_OFFSET(mp, ino); |
1375 | ASSERT(offset < mp->m_sb.sb_inopblock); | |
94e1b69d CH |
1376 | |
1377 | imap->im_blkno = XFS_AGB_TO_DADDR(mp, agno, agbno); | |
1378 | imap->im_len = XFS_FSB_TO_BB(mp, 1); | |
1379 | imap->im_boffset = (ushort)(offset << mp->m_sb.sb_inodelog); | |
1da177e4 LT |
1380 | return 0; |
1381 | } | |
94e1b69d | 1382 | |
94e1b69d CH |
1383 | /* |
1384 | * If the inode chunks are aligned then use simple maths to | |
1385 | * find the location. Otherwise we have to do a btree | |
1386 | * lookup to find the location. | |
1387 | */ | |
1da177e4 LT |
1388 | if (mp->m_inoalign_mask) { |
1389 | offset_agbno = agbno & mp->m_inoalign_mask; | |
1390 | chunk_agbno = agbno - offset_agbno; | |
1391 | } else { | |
7124fe0a DC |
1392 | error = xfs_imap_lookup(mp, tp, agno, agino, agbno, |
1393 | &chunk_agbno, &offset_agbno, flags); | |
1da177e4 LT |
1394 | if (error) |
1395 | return error; | |
1da177e4 | 1396 | } |
94e1b69d | 1397 | |
7124fe0a | 1398 | out_map: |
1da177e4 LT |
1399 | ASSERT(agbno >= chunk_agbno); |
1400 | cluster_agbno = chunk_agbno + | |
1401 | ((offset_agbno / blks_per_cluster) * blks_per_cluster); | |
1402 | offset = ((agbno - cluster_agbno) * mp->m_sb.sb_inopblock) + | |
1403 | XFS_INO_TO_OFFSET(mp, ino); | |
94e1b69d CH |
1404 | |
1405 | imap->im_blkno = XFS_AGB_TO_DADDR(mp, agno, cluster_agbno); | |
1406 | imap->im_len = XFS_FSB_TO_BB(mp, blks_per_cluster); | |
1407 | imap->im_boffset = (ushort)(offset << mp->m_sb.sb_inodelog); | |
1408 | ||
1409 | /* | |
1410 | * If the inode number maps to a block outside the bounds | |
1411 | * of the file system then return NULL rather than calling | |
1412 | * read_buf and panicing when we get an error from the | |
1413 | * driver. | |
1414 | */ | |
1415 | if ((imap->im_blkno + imap->im_len) > | |
1416 | XFS_FSB_TO_BB(mp, mp->m_sb.sb_dblocks)) { | |
53487786 DC |
1417 | xfs_alert(mp, |
1418 | "%s: (im_blkno (0x%llx) + im_len (0x%llx)) > sb_dblocks (0x%llx)", | |
1419 | __func__, (unsigned long long) imap->im_blkno, | |
94e1b69d CH |
1420 | (unsigned long long) imap->im_len, |
1421 | XFS_FSB_TO_BB(mp, mp->m_sb.sb_dblocks)); | |
1422 | return XFS_ERROR(EINVAL); | |
1423 | } | |
1da177e4 | 1424 | return 0; |
1da177e4 LT |
1425 | } |
1426 | ||
1427 | /* | |
1428 | * Compute and fill in value of m_in_maxlevels. | |
1429 | */ | |
1430 | void | |
1431 | xfs_ialloc_compute_maxlevels( | |
1432 | xfs_mount_t *mp) /* file system mount structure */ | |
1433 | { | |
1434 | int level; | |
1435 | uint maxblocks; | |
1436 | uint maxleafents; | |
1437 | int minleafrecs; | |
1438 | int minnoderecs; | |
1439 | ||
1440 | maxleafents = (1LL << XFS_INO_AGINO_BITS(mp)) >> | |
1441 | XFS_INODES_PER_CHUNK_LOG; | |
1442 | minleafrecs = mp->m_alloc_mnr[0]; | |
1443 | minnoderecs = mp->m_alloc_mnr[1]; | |
1444 | maxblocks = (maxleafents + minleafrecs - 1) / minleafrecs; | |
1445 | for (level = 1; maxblocks > 1; level++) | |
1446 | maxblocks = (maxblocks + minnoderecs - 1) / minnoderecs; | |
1447 | mp->m_in_maxlevels = level; | |
1448 | } | |
1449 | ||
1450 | /* | |
1451 | * Log specified fields for the ag hdr (inode section) | |
1452 | */ | |
1453 | void | |
1454 | xfs_ialloc_log_agi( | |
1455 | xfs_trans_t *tp, /* transaction pointer */ | |
1456 | xfs_buf_t *bp, /* allocation group header buffer */ | |
1457 | int fields) /* bitmask of fields to log */ | |
1458 | { | |
1459 | int first; /* first byte number */ | |
1460 | int last; /* last byte number */ | |
1461 | static const short offsets[] = { /* field starting offsets */ | |
1462 | /* keep in sync with bit definitions */ | |
1463 | offsetof(xfs_agi_t, agi_magicnum), | |
1464 | offsetof(xfs_agi_t, agi_versionnum), | |
1465 | offsetof(xfs_agi_t, agi_seqno), | |
1466 | offsetof(xfs_agi_t, agi_length), | |
1467 | offsetof(xfs_agi_t, agi_count), | |
1468 | offsetof(xfs_agi_t, agi_root), | |
1469 | offsetof(xfs_agi_t, agi_level), | |
1470 | offsetof(xfs_agi_t, agi_freecount), | |
1471 | offsetof(xfs_agi_t, agi_newino), | |
1472 | offsetof(xfs_agi_t, agi_dirino), | |
1473 | offsetof(xfs_agi_t, agi_unlinked), | |
1474 | sizeof(xfs_agi_t) | |
1475 | }; | |
1476 | #ifdef DEBUG | |
1477 | xfs_agi_t *agi; /* allocation group header */ | |
1478 | ||
1479 | agi = XFS_BUF_TO_AGI(bp); | |
69ef921b | 1480 | ASSERT(agi->agi_magicnum == cpu_to_be32(XFS_AGI_MAGIC)); |
1da177e4 LT |
1481 | #endif |
1482 | /* | |
1483 | * Compute byte offsets for the first and last fields. | |
1484 | */ | |
1485 | xfs_btree_offsets(fields, offsets, XFS_AGI_NUM_BITS, &first, &last); | |
1486 | /* | |
1487 | * Log the allocation group inode header buffer. | |
1488 | */ | |
61fe135c | 1489 | xfs_trans_buf_set_type(tp, bp, XFS_BLFT_AGI_BUF); |
1da177e4 LT |
1490 | xfs_trans_log_buf(tp, bp, first, last); |
1491 | } | |
1492 | ||
5e1be0fb CH |
1493 | #ifdef DEBUG |
1494 | STATIC void | |
1495 | xfs_check_agi_unlinked( | |
1496 | struct xfs_agi *agi) | |
1497 | { | |
1498 | int i; | |
1499 | ||
1500 | for (i = 0; i < XFS_AGI_UNLINKED_BUCKETS; i++) | |
1501 | ASSERT(agi->agi_unlinked[i]); | |
1502 | } | |
1503 | #else | |
1504 | #define xfs_check_agi_unlinked(agi) | |
1505 | #endif | |
1506 | ||
983d09ff | 1507 | static bool |
612cfbfe | 1508 | xfs_agi_verify( |
3702ce6e DC |
1509 | struct xfs_buf *bp) |
1510 | { | |
1511 | struct xfs_mount *mp = bp->b_target->bt_mount; | |
1512 | struct xfs_agi *agi = XFS_BUF_TO_AGI(bp); | |
3702ce6e | 1513 | |
983d09ff DC |
1514 | if (xfs_sb_version_hascrc(&mp->m_sb) && |
1515 | !uuid_equal(&agi->agi_uuid, &mp->m_sb.sb_uuid)) | |
1516 | return false; | |
3702ce6e DC |
1517 | /* |
1518 | * Validate the magic number of the agi block. | |
1519 | */ | |
983d09ff DC |
1520 | if (agi->agi_magicnum != cpu_to_be32(XFS_AGI_MAGIC)) |
1521 | return false; | |
1522 | if (!XFS_AGI_GOOD_VERSION(be32_to_cpu(agi->agi_versionnum))) | |
1523 | return false; | |
3702ce6e DC |
1524 | |
1525 | /* | |
1526 | * during growfs operations, the perag is not fully initialised, | |
1527 | * so we can't use it for any useful checking. growfs ensures we can't | |
1528 | * use it by using uncached buffers that don't have the perag attached | |
1529 | * so we can detect and avoid this problem. | |
1530 | */ | |
983d09ff DC |
1531 | if (bp->b_pag && be32_to_cpu(agi->agi_seqno) != bp->b_pag->pag_agno) |
1532 | return false; | |
3702ce6e | 1533 | |
3702ce6e | 1534 | xfs_check_agi_unlinked(agi); |
983d09ff | 1535 | return true; |
612cfbfe DC |
1536 | } |
1537 | ||
1813dd64 DC |
1538 | static void |
1539 | xfs_agi_read_verify( | |
612cfbfe DC |
1540 | struct xfs_buf *bp) |
1541 | { | |
983d09ff DC |
1542 | struct xfs_mount *mp = bp->b_target->bt_mount; |
1543 | int agi_ok = 1; | |
1544 | ||
1545 | if (xfs_sb_version_hascrc(&mp->m_sb)) | |
1546 | agi_ok = xfs_verify_cksum(bp->b_addr, BBTOB(bp->b_length), | |
1547 | offsetof(struct xfs_agi, agi_crc)); | |
1548 | agi_ok = agi_ok && xfs_agi_verify(bp); | |
1549 | ||
1550 | if (unlikely(XFS_TEST_ERROR(!agi_ok, mp, XFS_ERRTAG_IALLOC_READ_AGI, | |
1551 | XFS_RANDOM_IALLOC_READ_AGI))) { | |
1552 | XFS_CORRUPTION_ERROR(__func__, XFS_ERRLEVEL_LOW, mp, bp->b_addr); | |
1553 | xfs_buf_ioerror(bp, EFSCORRUPTED); | |
1554 | } | |
612cfbfe DC |
1555 | } |
1556 | ||
b0f539de | 1557 | static void |
1813dd64 | 1558 | xfs_agi_write_verify( |
612cfbfe DC |
1559 | struct xfs_buf *bp) |
1560 | { | |
983d09ff DC |
1561 | struct xfs_mount *mp = bp->b_target->bt_mount; |
1562 | struct xfs_buf_log_item *bip = bp->b_fspriv; | |
1563 | ||
1564 | if (!xfs_agi_verify(bp)) { | |
1565 | XFS_CORRUPTION_ERROR(__func__, XFS_ERRLEVEL_LOW, mp, bp->b_addr); | |
1566 | xfs_buf_ioerror(bp, EFSCORRUPTED); | |
1567 | return; | |
1568 | } | |
1569 | ||
1570 | if (!xfs_sb_version_hascrc(&mp->m_sb)) | |
1571 | return; | |
1572 | ||
1573 | if (bip) | |
1574 | XFS_BUF_TO_AGI(bp)->agi_lsn = cpu_to_be64(bip->bli_item.li_lsn); | |
1575 | xfs_update_cksum(bp->b_addr, BBTOB(bp->b_length), | |
1576 | offsetof(struct xfs_agi, agi_crc)); | |
3702ce6e DC |
1577 | } |
1578 | ||
1813dd64 DC |
1579 | const struct xfs_buf_ops xfs_agi_buf_ops = { |
1580 | .verify_read = xfs_agi_read_verify, | |
1581 | .verify_write = xfs_agi_write_verify, | |
1582 | }; | |
1583 | ||
1da177e4 LT |
1584 | /* |
1585 | * Read in the allocation group header (inode allocation section) | |
1586 | */ | |
1587 | int | |
5e1be0fb CH |
1588 | xfs_read_agi( |
1589 | struct xfs_mount *mp, /* file system mount structure */ | |
1590 | struct xfs_trans *tp, /* transaction pointer */ | |
1591 | xfs_agnumber_t agno, /* allocation group number */ | |
1592 | struct xfs_buf **bpp) /* allocation group hdr buf */ | |
1da177e4 | 1593 | { |
5e1be0fb | 1594 | int error; |
1da177e4 LT |
1595 | |
1596 | ASSERT(agno != NULLAGNUMBER); | |
5e1be0fb CH |
1597 | |
1598 | error = xfs_trans_read_buf(mp, tp, mp->m_ddev_targp, | |
1da177e4 | 1599 | XFS_AG_DADDR(mp, agno, XFS_AGI_DADDR(mp)), |
1813dd64 | 1600 | XFS_FSS_TO_BB(mp, 1), 0, bpp, &xfs_agi_buf_ops); |
1da177e4 LT |
1601 | if (error) |
1602 | return error; | |
5e1be0fb | 1603 | |
5a52c2a5 | 1604 | ASSERT(!xfs_buf_geterror(*bpp)); |
38f23232 | 1605 | xfs_buf_set_ref(*bpp, XFS_AGI_REF); |
5e1be0fb CH |
1606 | return 0; |
1607 | } | |
1608 | ||
1609 | int | |
1610 | xfs_ialloc_read_agi( | |
1611 | struct xfs_mount *mp, /* file system mount structure */ | |
1612 | struct xfs_trans *tp, /* transaction pointer */ | |
1613 | xfs_agnumber_t agno, /* allocation group number */ | |
1614 | struct xfs_buf **bpp) /* allocation group hdr buf */ | |
1615 | { | |
1616 | struct xfs_agi *agi; /* allocation group header */ | |
1617 | struct xfs_perag *pag; /* per allocation group data */ | |
1618 | int error; | |
1619 | ||
1620 | error = xfs_read_agi(mp, tp, agno, bpp); | |
1621 | if (error) | |
1622 | return error; | |
1623 | ||
1624 | agi = XFS_BUF_TO_AGI(*bpp); | |
44b56e0a | 1625 | pag = xfs_perag_get(mp, agno); |
1da177e4 | 1626 | if (!pag->pagi_init) { |
16259e7d | 1627 | pag->pagi_freecount = be32_to_cpu(agi->agi_freecount); |
92821e2b | 1628 | pag->pagi_count = be32_to_cpu(agi->agi_count); |
1da177e4 | 1629 | pag->pagi_init = 1; |
1da177e4 | 1630 | } |
1da177e4 | 1631 | |
5e1be0fb CH |
1632 | /* |
1633 | * It's possible for these to be out of sync if | |
1634 | * we are in the middle of a forced shutdown. | |
1635 | */ | |
1636 | ASSERT(pag->pagi_freecount == be32_to_cpu(agi->agi_freecount) || | |
1637 | XFS_FORCED_SHUTDOWN(mp)); | |
44b56e0a | 1638 | xfs_perag_put(pag); |
1da177e4 LT |
1639 | return 0; |
1640 | } | |
92821e2b DC |
1641 | |
1642 | /* | |
1643 | * Read in the agi to initialise the per-ag data in the mount structure | |
1644 | */ | |
1645 | int | |
1646 | xfs_ialloc_pagi_init( | |
1647 | xfs_mount_t *mp, /* file system mount structure */ | |
1648 | xfs_trans_t *tp, /* transaction pointer */ | |
1649 | xfs_agnumber_t agno) /* allocation group number */ | |
1650 | { | |
1651 | xfs_buf_t *bp = NULL; | |
1652 | int error; | |
1653 | ||
1654 | error = xfs_ialloc_read_agi(mp, tp, agno, &bp); | |
1655 | if (error) | |
1656 | return error; | |
1657 | if (bp) | |
1658 | xfs_trans_brelse(tp, bp); | |
1659 | return 0; | |
1660 | } |