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" | |
39 | ||
1da177e4 LT |
40 | |
41 | /* | |
42 | * Allocation group level functions. | |
43 | */ | |
75de2a91 DC |
44 | static inline int |
45 | xfs_ialloc_cluster_alignment( | |
46 | xfs_alloc_arg_t *args) | |
47 | { | |
48 | if (xfs_sb_version_hasalign(&args->mp->m_sb) && | |
49 | args->mp->m_sb.sb_inoalignmt >= | |
50 | XFS_B_TO_FSBT(args->mp, XFS_INODE_CLUSTER_SIZE(args->mp))) | |
51 | return args->mp->m_sb.sb_inoalignmt; | |
52 | return 1; | |
53 | } | |
1da177e4 | 54 | |
fe033cc8 | 55 | /* |
21875505 | 56 | * Lookup a record by ino in the btree given by cur. |
fe033cc8 | 57 | */ |
81e25176 | 58 | int /* error */ |
21875505 | 59 | xfs_inobt_lookup( |
fe033cc8 CH |
60 | struct xfs_btree_cur *cur, /* btree cursor */ |
61 | xfs_agino_t ino, /* starting inode of chunk */ | |
21875505 | 62 | xfs_lookup_t dir, /* <=, >=, == */ |
fe033cc8 CH |
63 | int *stat) /* success/failure */ |
64 | { | |
65 | cur->bc_rec.i.ir_startino = ino; | |
21875505 CH |
66 | cur->bc_rec.i.ir_freecount = 0; |
67 | cur->bc_rec.i.ir_free = 0; | |
68 | return xfs_btree_lookup(cur, dir, stat); | |
fe033cc8 CH |
69 | } |
70 | ||
278d0ca1 | 71 | /* |
afabc24a | 72 | * Update the record referred to by cur to the value given. |
278d0ca1 CH |
73 | * This either works (return 0) or gets an EFSCORRUPTED error. |
74 | */ | |
75 | STATIC int /* error */ | |
76 | xfs_inobt_update( | |
77 | struct xfs_btree_cur *cur, /* btree cursor */ | |
afabc24a | 78 | xfs_inobt_rec_incore_t *irec) /* btree record */ |
278d0ca1 CH |
79 | { |
80 | union xfs_btree_rec rec; | |
81 | ||
afabc24a CH |
82 | rec.inobt.ir_startino = cpu_to_be32(irec->ir_startino); |
83 | rec.inobt.ir_freecount = cpu_to_be32(irec->ir_freecount); | |
84 | rec.inobt.ir_free = cpu_to_be64(irec->ir_free); | |
278d0ca1 CH |
85 | return xfs_btree_update(cur, &rec); |
86 | } | |
87 | ||
8cc938fe CH |
88 | /* |
89 | * Get the data from the pointed-to record. | |
90 | */ | |
91 | int /* error */ | |
92 | xfs_inobt_get_rec( | |
93 | struct xfs_btree_cur *cur, /* btree cursor */ | |
2e287a73 | 94 | xfs_inobt_rec_incore_t *irec, /* btree record */ |
8cc938fe CH |
95 | int *stat) /* output: success/failure */ |
96 | { | |
97 | union xfs_btree_rec *rec; | |
98 | int error; | |
99 | ||
100 | error = xfs_btree_get_rec(cur, &rec, stat); | |
101 | if (!error && *stat == 1) { | |
2e287a73 CH |
102 | irec->ir_startino = be32_to_cpu(rec->inobt.ir_startino); |
103 | irec->ir_freecount = be32_to_cpu(rec->inobt.ir_freecount); | |
104 | irec->ir_free = be64_to_cpu(rec->inobt.ir_free); | |
8cc938fe CH |
105 | } |
106 | return error; | |
107 | } | |
108 | ||
0b48db80 DC |
109 | /* |
110 | * Verify that the number of free inodes in the AGI is correct. | |
111 | */ | |
112 | #ifdef DEBUG | |
113 | STATIC int | |
114 | xfs_check_agi_freecount( | |
115 | struct xfs_btree_cur *cur, | |
116 | struct xfs_agi *agi) | |
117 | { | |
118 | if (cur->bc_nlevels == 1) { | |
119 | xfs_inobt_rec_incore_t rec; | |
120 | int freecount = 0; | |
121 | int error; | |
122 | int i; | |
123 | ||
21875505 | 124 | error = xfs_inobt_lookup(cur, 0, XFS_LOOKUP_GE, &i); |
0b48db80 DC |
125 | if (error) |
126 | return error; | |
127 | ||
128 | do { | |
129 | error = xfs_inobt_get_rec(cur, &rec, &i); | |
130 | if (error) | |
131 | return error; | |
132 | ||
133 | if (i) { | |
134 | freecount += rec.ir_freecount; | |
135 | error = xfs_btree_increment(cur, 0, &i); | |
136 | if (error) | |
137 | return error; | |
138 | } | |
139 | } while (i == 1); | |
140 | ||
141 | if (!XFS_FORCED_SHUTDOWN(cur->bc_mp)) | |
142 | ASSERT(freecount == be32_to_cpu(agi->agi_freecount)); | |
143 | } | |
144 | return 0; | |
145 | } | |
146 | #else | |
147 | #define xfs_check_agi_freecount(cur, agi) 0 | |
148 | #endif | |
149 | ||
85c0b2ab DC |
150 | /* |
151 | * Initialise a new set of inodes. | |
152 | */ | |
2a30f36d | 153 | STATIC int |
85c0b2ab DC |
154 | xfs_ialloc_inode_init( |
155 | struct xfs_mount *mp, | |
156 | struct xfs_trans *tp, | |
157 | xfs_agnumber_t agno, | |
158 | xfs_agblock_t agbno, | |
159 | xfs_agblock_t length, | |
160 | unsigned int gen) | |
161 | { | |
162 | struct xfs_buf *fbuf; | |
163 | struct xfs_dinode *free; | |
164 | int blks_per_cluster, nbufs, ninodes; | |
165 | int version; | |
166 | int i, j; | |
167 | xfs_daddr_t d; | |
168 | ||
169 | /* | |
170 | * Loop over the new block(s), filling in the inodes. | |
171 | * For small block sizes, manipulate the inodes in buffers | |
172 | * which are multiples of the blocks size. | |
173 | */ | |
174 | if (mp->m_sb.sb_blocksize >= XFS_INODE_CLUSTER_SIZE(mp)) { | |
175 | blks_per_cluster = 1; | |
176 | nbufs = length; | |
177 | ninodes = mp->m_sb.sb_inopblock; | |
178 | } else { | |
179 | blks_per_cluster = XFS_INODE_CLUSTER_SIZE(mp) / | |
180 | mp->m_sb.sb_blocksize; | |
181 | nbufs = length / blks_per_cluster; | |
182 | ninodes = blks_per_cluster * mp->m_sb.sb_inopblock; | |
183 | } | |
184 | ||
185 | /* | |
186 | * Figure out what version number to use in the inodes we create. | |
187 | * If the superblock version has caught up to the one that supports | |
188 | * the new inode format, then use the new inode version. Otherwise | |
189 | * use the old version so that old kernels will continue to be | |
190 | * able to use the file system. | |
191 | */ | |
192 | if (xfs_sb_version_hasnlink(&mp->m_sb)) | |
193 | version = 2; | |
194 | else | |
195 | version = 1; | |
196 | ||
197 | for (j = 0; j < nbufs; j++) { | |
198 | /* | |
199 | * Get the block. | |
200 | */ | |
201 | d = XFS_AGB_TO_DADDR(mp, agno, agbno + (j * blks_per_cluster)); | |
202 | fbuf = xfs_trans_get_buf(tp, mp->m_ddev_targp, d, | |
a8acad70 | 203 | mp->m_bsize * blks_per_cluster, 0); |
2a30f36d CS |
204 | if (!fbuf) |
205 | return ENOMEM; | |
85c0b2ab DC |
206 | /* |
207 | * Initialize all inodes in this buffer and then log them. | |
208 | * | |
209 | * XXX: It would be much better if we had just one transaction | |
210 | * to log a whole cluster of inodes instead of all the | |
211 | * individual transactions causing a lot of log traffic. | |
212 | */ | |
b0f539de | 213 | fbuf->b_pre_io = xfs_inode_buf_write_verify; |
1a1a3e97 | 214 | xfs_buf_zero(fbuf, 0, ninodes << mp->m_sb.sb_inodelog); |
85c0b2ab DC |
215 | for (i = 0; i < ninodes; i++) { |
216 | int ioffset = i << mp->m_sb.sb_inodelog; | |
217 | uint isize = sizeof(struct xfs_dinode); | |
218 | ||
219 | free = xfs_make_iptr(mp, fbuf, i); | |
220 | free->di_magic = cpu_to_be16(XFS_DINODE_MAGIC); | |
221 | free->di_version = version; | |
222 | free->di_gen = cpu_to_be32(gen); | |
223 | free->di_next_unlinked = cpu_to_be32(NULLAGINO); | |
224 | xfs_trans_log_buf(tp, fbuf, ioffset, ioffset + isize - 1); | |
225 | } | |
226 | xfs_trans_inode_alloc_buf(tp, fbuf); | |
227 | } | |
2a30f36d | 228 | return 0; |
85c0b2ab DC |
229 | } |
230 | ||
1da177e4 LT |
231 | /* |
232 | * Allocate new inodes in the allocation group specified by agbp. | |
233 | * Return 0 for success, else error code. | |
234 | */ | |
235 | STATIC int /* error code or 0 */ | |
236 | xfs_ialloc_ag_alloc( | |
237 | xfs_trans_t *tp, /* transaction pointer */ | |
238 | xfs_buf_t *agbp, /* alloc group buffer */ | |
239 | int *alloc) | |
240 | { | |
241 | xfs_agi_t *agi; /* allocation group header */ | |
242 | xfs_alloc_arg_t args; /* allocation argument structure */ | |
1da177e4 | 243 | xfs_btree_cur_t *cur; /* inode btree cursor */ |
92821e2b | 244 | xfs_agnumber_t agno; |
1da177e4 | 245 | int error; |
85c0b2ab | 246 | int i; |
1da177e4 LT |
247 | xfs_agino_t newino; /* new first inode's number */ |
248 | xfs_agino_t newlen; /* new number of inodes */ | |
1da177e4 | 249 | xfs_agino_t thisino; /* current inode number, for loop */ |
3ccb8b5f | 250 | int isaligned = 0; /* inode allocation at stripe unit */ |
1da177e4 | 251 | /* boundary */ |
44b56e0a | 252 | struct xfs_perag *pag; |
1da177e4 | 253 | |
a0041684 | 254 | memset(&args, 0, sizeof(args)); |
1da177e4 LT |
255 | args.tp = tp; |
256 | args.mp = tp->t_mountp; | |
257 | ||
258 | /* | |
259 | * Locking will ensure that we don't have two callers in here | |
260 | * at one time. | |
261 | */ | |
262 | newlen = XFS_IALLOC_INODES(args.mp); | |
263 | if (args.mp->m_maxicount && | |
264 | args.mp->m_sb.sb_icount + newlen > args.mp->m_maxicount) | |
265 | return XFS_ERROR(ENOSPC); | |
266 | args.minlen = args.maxlen = XFS_IALLOC_BLOCKS(args.mp); | |
267 | /* | |
3ccb8b5f GO |
268 | * First try to allocate inodes contiguous with the last-allocated |
269 | * chunk of inodes. If the filesystem is striped, this will fill | |
270 | * an entire stripe unit with inodes. | |
271 | */ | |
1da177e4 | 272 | agi = XFS_BUF_TO_AGI(agbp); |
3ccb8b5f | 273 | newino = be32_to_cpu(agi->agi_newino); |
85c0b2ab | 274 | agno = be32_to_cpu(agi->agi_seqno); |
019ff2d5 NS |
275 | args.agbno = XFS_AGINO_TO_AGBNO(args.mp, newino) + |
276 | XFS_IALLOC_BLOCKS(args.mp); | |
277 | if (likely(newino != NULLAGINO && | |
278 | (args.agbno < be32_to_cpu(agi->agi_length)))) { | |
85c0b2ab | 279 | args.fsbno = XFS_AGB_TO_FSB(args.mp, agno, args.agbno); |
3ccb8b5f GO |
280 | args.type = XFS_ALLOCTYPE_THIS_BNO; |
281 | args.mod = args.total = args.wasdel = args.isfl = | |
282 | args.userdata = args.minalignslop = 0; | |
283 | args.prod = 1; | |
75de2a91 | 284 | |
3ccb8b5f | 285 | /* |
75de2a91 DC |
286 | * We need to take into account alignment here to ensure that |
287 | * we don't modify the free list if we fail to have an exact | |
288 | * block. If we don't have an exact match, and every oher | |
289 | * attempt allocation attempt fails, we'll end up cancelling | |
290 | * a dirty transaction and shutting down. | |
291 | * | |
292 | * For an exact allocation, alignment must be 1, | |
293 | * however we need to take cluster alignment into account when | |
294 | * fixing up the freelist. Use the minalignslop field to | |
295 | * indicate that extra blocks might be required for alignment, | |
296 | * but not to use them in the actual exact allocation. | |
3ccb8b5f | 297 | */ |
75de2a91 DC |
298 | args.alignment = 1; |
299 | args.minalignslop = xfs_ialloc_cluster_alignment(&args) - 1; | |
300 | ||
301 | /* Allow space for the inode btree to split. */ | |
0d87e656 | 302 | args.minleft = args.mp->m_in_maxlevels - 1; |
3ccb8b5f GO |
303 | if ((error = xfs_alloc_vextent(&args))) |
304 | return error; | |
305 | } else | |
306 | args.fsbno = NULLFSBLOCK; | |
1da177e4 | 307 | |
3ccb8b5f GO |
308 | if (unlikely(args.fsbno == NULLFSBLOCK)) { |
309 | /* | |
310 | * Set the alignment for the allocation. | |
311 | * If stripe alignment is turned on then align at stripe unit | |
312 | * boundary. | |
019ff2d5 NS |
313 | * If the cluster size is smaller than a filesystem block |
314 | * then we're doing I/O for inodes in filesystem block size | |
3ccb8b5f GO |
315 | * pieces, so don't need alignment anyway. |
316 | */ | |
317 | isaligned = 0; | |
318 | if (args.mp->m_sinoalign) { | |
319 | ASSERT(!(args.mp->m_flags & XFS_MOUNT_NOALIGN)); | |
320 | args.alignment = args.mp->m_dalign; | |
321 | isaligned = 1; | |
75de2a91 DC |
322 | } else |
323 | args.alignment = xfs_ialloc_cluster_alignment(&args); | |
3ccb8b5f GO |
324 | /* |
325 | * Need to figure out where to allocate the inode blocks. | |
326 | * Ideally they should be spaced out through the a.g. | |
327 | * For now, just allocate blocks up front. | |
328 | */ | |
329 | args.agbno = be32_to_cpu(agi->agi_root); | |
85c0b2ab | 330 | args.fsbno = XFS_AGB_TO_FSB(args.mp, agno, args.agbno); |
3ccb8b5f GO |
331 | /* |
332 | * Allocate a fixed-size extent of inodes. | |
333 | */ | |
334 | args.type = XFS_ALLOCTYPE_NEAR_BNO; | |
335 | args.mod = args.total = args.wasdel = args.isfl = | |
336 | args.userdata = args.minalignslop = 0; | |
337 | args.prod = 1; | |
338 | /* | |
339 | * Allow space for the inode btree to split. | |
340 | */ | |
0d87e656 | 341 | args.minleft = args.mp->m_in_maxlevels - 1; |
3ccb8b5f GO |
342 | if ((error = xfs_alloc_vextent(&args))) |
343 | return error; | |
344 | } | |
019ff2d5 | 345 | |
1da177e4 LT |
346 | /* |
347 | * If stripe alignment is turned on, then try again with cluster | |
348 | * alignment. | |
349 | */ | |
350 | if (isaligned && args.fsbno == NULLFSBLOCK) { | |
351 | args.type = XFS_ALLOCTYPE_NEAR_BNO; | |
16259e7d | 352 | args.agbno = be32_to_cpu(agi->agi_root); |
85c0b2ab | 353 | args.fsbno = XFS_AGB_TO_FSB(args.mp, agno, args.agbno); |
75de2a91 | 354 | args.alignment = xfs_ialloc_cluster_alignment(&args); |
1da177e4 LT |
355 | if ((error = xfs_alloc_vextent(&args))) |
356 | return error; | |
357 | } | |
358 | ||
359 | if (args.fsbno == NULLFSBLOCK) { | |
360 | *alloc = 0; | |
361 | return 0; | |
362 | } | |
363 | ASSERT(args.len == args.minlen); | |
1da177e4 | 364 | |
359346a9 | 365 | /* |
85c0b2ab DC |
366 | * Stamp and write the inode buffers. |
367 | * | |
359346a9 DC |
368 | * Seed the new inode cluster with a random generation number. This |
369 | * prevents short-term reuse of generation numbers if a chunk is | |
370 | * freed and then immediately reallocated. We use random numbers | |
371 | * rather than a linear progression to prevent the next generation | |
372 | * number from being easily guessable. | |
373 | */ | |
2a30f36d CS |
374 | error = xfs_ialloc_inode_init(args.mp, tp, agno, args.agbno, |
375 | args.len, random32()); | |
d42f08f6 | 376 | |
2a30f36d CS |
377 | if (error) |
378 | return error; | |
85c0b2ab DC |
379 | /* |
380 | * Convert the results. | |
381 | */ | |
382 | newino = XFS_OFFBNO_TO_AGINO(args.mp, args.agbno, 0); | |
413d57c9 MS |
383 | be32_add_cpu(&agi->agi_count, newlen); |
384 | be32_add_cpu(&agi->agi_freecount, newlen); | |
44b56e0a DC |
385 | pag = xfs_perag_get(args.mp, agno); |
386 | pag->pagi_freecount += newlen; | |
387 | xfs_perag_put(pag); | |
16259e7d | 388 | agi->agi_newino = cpu_to_be32(newino); |
85c0b2ab | 389 | |
1da177e4 LT |
390 | /* |
391 | * Insert records describing the new inode chunk into the btree. | |
392 | */ | |
561f7d17 | 393 | cur = xfs_inobt_init_cursor(args.mp, tp, agbp, agno); |
1da177e4 LT |
394 | for (thisino = newino; |
395 | thisino < newino + newlen; | |
396 | thisino += XFS_INODES_PER_CHUNK) { | |
21875505 CH |
397 | cur->bc_rec.i.ir_startino = thisino; |
398 | cur->bc_rec.i.ir_freecount = XFS_INODES_PER_CHUNK; | |
399 | cur->bc_rec.i.ir_free = XFS_INOBT_ALL_FREE; | |
400 | error = xfs_btree_lookup(cur, XFS_LOOKUP_EQ, &i); | |
401 | if (error) { | |
1da177e4 LT |
402 | xfs_btree_del_cursor(cur, XFS_BTREE_ERROR); |
403 | return error; | |
404 | } | |
405 | ASSERT(i == 0); | |
21875505 CH |
406 | error = xfs_btree_insert(cur, &i); |
407 | if (error) { | |
1da177e4 LT |
408 | xfs_btree_del_cursor(cur, XFS_BTREE_ERROR); |
409 | return error; | |
410 | } | |
411 | ASSERT(i == 1); | |
412 | } | |
413 | xfs_btree_del_cursor(cur, XFS_BTREE_NOERROR); | |
414 | /* | |
415 | * Log allocation group header fields | |
416 | */ | |
417 | xfs_ialloc_log_agi(tp, agbp, | |
418 | XFS_AGI_COUNT | XFS_AGI_FREECOUNT | XFS_AGI_NEWINO); | |
419 | /* | |
420 | * Modify/log superblock values for inode count and inode free count. | |
421 | */ | |
422 | xfs_trans_mod_sb(tp, XFS_TRANS_SB_ICOUNT, (long)newlen); | |
423 | xfs_trans_mod_sb(tp, XFS_TRANS_SB_IFREE, (long)newlen); | |
424 | *alloc = 1; | |
425 | return 0; | |
426 | } | |
427 | ||
b8f82a4a | 428 | STATIC xfs_agnumber_t |
1da177e4 LT |
429 | xfs_ialloc_next_ag( |
430 | xfs_mount_t *mp) | |
431 | { | |
432 | xfs_agnumber_t agno; | |
433 | ||
434 | spin_lock(&mp->m_agirotor_lock); | |
435 | agno = mp->m_agirotor; | |
8aea3ff4 | 436 | if (++mp->m_agirotor >= mp->m_maxagi) |
1da177e4 LT |
437 | mp->m_agirotor = 0; |
438 | spin_unlock(&mp->m_agirotor_lock); | |
439 | ||
440 | return agno; | |
441 | } | |
442 | ||
443 | /* | |
444 | * Select an allocation group to look for a free inode in, based on the parent | |
445 | * inode and then mode. Return the allocation group buffer. | |
446 | */ | |
55d6af64 | 447 | STATIC xfs_agnumber_t |
1da177e4 LT |
448 | xfs_ialloc_ag_select( |
449 | xfs_trans_t *tp, /* transaction pointer */ | |
450 | xfs_ino_t parent, /* parent directory inode number */ | |
576b1d67 | 451 | umode_t mode, /* bits set to indicate file type */ |
1da177e4 LT |
452 | int okalloc) /* ok to allocate more space */ |
453 | { | |
1da177e4 LT |
454 | xfs_agnumber_t agcount; /* number of ag's in the filesystem */ |
455 | xfs_agnumber_t agno; /* current ag number */ | |
456 | int flags; /* alloc buffer locking flags */ | |
457 | xfs_extlen_t ineed; /* blocks needed for inode allocation */ | |
458 | xfs_extlen_t longest = 0; /* longest extent available */ | |
459 | xfs_mount_t *mp; /* mount point structure */ | |
460 | int needspace; /* file mode implies space allocated */ | |
461 | xfs_perag_t *pag; /* per allocation group data */ | |
462 | xfs_agnumber_t pagno; /* parent (starting) ag number */ | |
55d6af64 | 463 | int error; |
1da177e4 LT |
464 | |
465 | /* | |
466 | * Files of these types need at least one block if length > 0 | |
467 | * (and they won't fit in the inode, but that's hard to figure out). | |
468 | */ | |
469 | needspace = S_ISDIR(mode) || S_ISREG(mode) || S_ISLNK(mode); | |
470 | mp = tp->t_mountp; | |
471 | agcount = mp->m_maxagi; | |
472 | if (S_ISDIR(mode)) | |
473 | pagno = xfs_ialloc_next_ag(mp); | |
474 | else { | |
475 | pagno = XFS_INO_TO_AGNO(mp, parent); | |
476 | if (pagno >= agcount) | |
477 | pagno = 0; | |
478 | } | |
55d6af64 | 479 | |
1da177e4 | 480 | ASSERT(pagno < agcount); |
55d6af64 | 481 | |
1da177e4 LT |
482 | /* |
483 | * Loop through allocation groups, looking for one with a little | |
484 | * free space in it. Note we don't look for free inodes, exactly. | |
485 | * Instead, we include whether there is a need to allocate inodes | |
486 | * to mean that blocks must be allocated for them, | |
487 | * if none are currently free. | |
488 | */ | |
489 | agno = pagno; | |
490 | flags = XFS_ALLOC_FLAG_TRYLOCK; | |
1da177e4 | 491 | for (;;) { |
44b56e0a | 492 | pag = xfs_perag_get(mp, agno); |
55d6af64 CH |
493 | if (!pag->pagi_inodeok) { |
494 | xfs_ialloc_next_ag(mp); | |
495 | goto nextag; | |
496 | } | |
497 | ||
1da177e4 | 498 | if (!pag->pagi_init) { |
55d6af64 CH |
499 | error = xfs_ialloc_pagi_init(mp, tp, agno); |
500 | if (error) | |
1da177e4 | 501 | goto nextag; |
55d6af64 | 502 | } |
1da177e4 | 503 | |
55d6af64 CH |
504 | if (pag->pagi_freecount) { |
505 | xfs_perag_put(pag); | |
506 | return agno; | |
1da177e4 LT |
507 | } |
508 | ||
55d6af64 CH |
509 | if (!okalloc) |
510 | goto nextag; | |
511 | ||
512 | if (!pag->pagf_init) { | |
513 | error = xfs_alloc_pagf_init(mp, tp, agno, flags); | |
514 | if (error) | |
1da177e4 | 515 | goto nextag; |
1da177e4 | 516 | } |
55d6af64 CH |
517 | |
518 | /* | |
519 | * Is there enough free space for the file plus a block of | |
520 | * inodes? (if we need to allocate some)? | |
521 | */ | |
522 | ineed = XFS_IALLOC_BLOCKS(mp); | |
523 | longest = pag->pagf_longest; | |
524 | if (!longest) | |
525 | longest = pag->pagf_flcount > 0; | |
526 | ||
527 | if (pag->pagf_freeblks >= needspace + ineed && | |
528 | longest >= ineed) { | |
529 | xfs_perag_put(pag); | |
530 | return agno; | |
1da177e4 | 531 | } |
1da177e4 | 532 | nextag: |
44b56e0a | 533 | xfs_perag_put(pag); |
1da177e4 LT |
534 | /* |
535 | * No point in iterating over the rest, if we're shutting | |
536 | * down. | |
537 | */ | |
1c1c6ebc | 538 | if (XFS_FORCED_SHUTDOWN(mp)) |
55d6af64 | 539 | return NULLAGNUMBER; |
1da177e4 LT |
540 | agno++; |
541 | if (agno >= agcount) | |
542 | agno = 0; | |
543 | if (agno == pagno) { | |
1c1c6ebc | 544 | if (flags == 0) |
55d6af64 | 545 | return NULLAGNUMBER; |
1da177e4 LT |
546 | flags = 0; |
547 | } | |
548 | } | |
549 | } | |
550 | ||
4254b0bb CH |
551 | /* |
552 | * Try to retrieve the next record to the left/right from the current one. | |
553 | */ | |
554 | STATIC int | |
555 | xfs_ialloc_next_rec( | |
556 | struct xfs_btree_cur *cur, | |
557 | xfs_inobt_rec_incore_t *rec, | |
558 | int *done, | |
559 | int left) | |
560 | { | |
561 | int error; | |
562 | int i; | |
563 | ||
564 | if (left) | |
565 | error = xfs_btree_decrement(cur, 0, &i); | |
566 | else | |
567 | error = xfs_btree_increment(cur, 0, &i); | |
568 | ||
569 | if (error) | |
570 | return error; | |
571 | *done = !i; | |
572 | if (i) { | |
573 | error = xfs_inobt_get_rec(cur, rec, &i); | |
574 | if (error) | |
575 | return error; | |
576 | XFS_WANT_CORRUPTED_RETURN(i == 1); | |
577 | } | |
578 | ||
579 | return 0; | |
580 | } | |
581 | ||
bd169565 DC |
582 | STATIC int |
583 | xfs_ialloc_get_rec( | |
584 | struct xfs_btree_cur *cur, | |
585 | xfs_agino_t agino, | |
586 | xfs_inobt_rec_incore_t *rec, | |
587 | int *done, | |
588 | int left) | |
589 | { | |
590 | int error; | |
591 | int i; | |
592 | ||
593 | error = xfs_inobt_lookup(cur, agino, XFS_LOOKUP_EQ, &i); | |
594 | if (error) | |
595 | return error; | |
596 | *done = !i; | |
597 | if (i) { | |
598 | error = xfs_inobt_get_rec(cur, rec, &i); | |
599 | if (error) | |
600 | return error; | |
601 | XFS_WANT_CORRUPTED_RETURN(i == 1); | |
602 | } | |
603 | ||
604 | return 0; | |
605 | } | |
0b48db80 | 606 | |
1da177e4 | 607 | /* |
f2ecc5e4 | 608 | * Allocate an inode. |
1da177e4 | 609 | * |
f2ecc5e4 CH |
610 | * The caller selected an AG for us, and made sure that free inodes are |
611 | * available. | |
1da177e4 | 612 | */ |
f2ecc5e4 CH |
613 | STATIC int |
614 | xfs_dialloc_ag( | |
615 | struct xfs_trans *tp, | |
616 | struct xfs_buf *agbp, | |
617 | xfs_ino_t parent, | |
618 | xfs_ino_t *inop) | |
1da177e4 | 619 | { |
f2ecc5e4 CH |
620 | struct xfs_mount *mp = tp->t_mountp; |
621 | struct xfs_agi *agi = XFS_BUF_TO_AGI(agbp); | |
622 | xfs_agnumber_t agno = be32_to_cpu(agi->agi_seqno); | |
623 | xfs_agnumber_t pagno = XFS_INO_TO_AGNO(mp, parent); | |
624 | xfs_agino_t pagino = XFS_INO_TO_AGINO(mp, parent); | |
625 | struct xfs_perag *pag; | |
626 | struct xfs_btree_cur *cur, *tcur; | |
627 | struct xfs_inobt_rec_incore rec, trec; | |
628 | xfs_ino_t ino; | |
629 | int error; | |
630 | int offset; | |
631 | int i, j; | |
1da177e4 | 632 | |
44b56e0a | 633 | pag = xfs_perag_get(mp, agno); |
bd169565 | 634 | |
4bb61069 CH |
635 | ASSERT(pag->pagi_init); |
636 | ASSERT(pag->pagi_inodeok); | |
637 | ASSERT(pag->pagi_freecount > 0); | |
638 | ||
bd169565 | 639 | restart_pagno: |
f2ecc5e4 | 640 | cur = xfs_inobt_init_cursor(mp, tp, agbp, agno); |
1da177e4 LT |
641 | /* |
642 | * If pagino is 0 (this is the root inode allocation) use newino. | |
643 | * This must work because we've just allocated some. | |
644 | */ | |
645 | if (!pagino) | |
16259e7d | 646 | pagino = be32_to_cpu(agi->agi_newino); |
1da177e4 | 647 | |
0b48db80 DC |
648 | error = xfs_check_agi_freecount(cur, agi); |
649 | if (error) | |
650 | goto error0; | |
1da177e4 | 651 | |
1da177e4 | 652 | /* |
4254b0bb | 653 | * If in the same AG as the parent, try to get near the parent. |
1da177e4 LT |
654 | */ |
655 | if (pagno == agno) { | |
4254b0bb CH |
656 | int doneleft; /* done, to the left */ |
657 | int doneright; /* done, to the right */ | |
bd169565 | 658 | int searchdistance = 10; |
4254b0bb | 659 | |
21875505 | 660 | error = xfs_inobt_lookup(cur, pagino, XFS_LOOKUP_LE, &i); |
4254b0bb | 661 | if (error) |
1da177e4 | 662 | goto error0; |
4254b0bb CH |
663 | XFS_WANT_CORRUPTED_GOTO(i == 1, error0); |
664 | ||
665 | error = xfs_inobt_get_rec(cur, &rec, &j); | |
666 | if (error) | |
667 | goto error0; | |
668 | XFS_WANT_CORRUPTED_GOTO(i == 1, error0); | |
669 | ||
670 | if (rec.ir_freecount > 0) { | |
1da177e4 LT |
671 | /* |
672 | * Found a free inode in the same chunk | |
4254b0bb | 673 | * as the parent, done. |
1da177e4 | 674 | */ |
4254b0bb | 675 | goto alloc_inode; |
1da177e4 | 676 | } |
4254b0bb CH |
677 | |
678 | ||
1da177e4 | 679 | /* |
4254b0bb | 680 | * In the same AG as parent, but parent's chunk is full. |
1da177e4 | 681 | */ |
1da177e4 | 682 | |
4254b0bb CH |
683 | /* duplicate the cursor, search left & right simultaneously */ |
684 | error = xfs_btree_dup_cursor(cur, &tcur); | |
685 | if (error) | |
686 | goto error0; | |
687 | ||
bd169565 DC |
688 | /* |
689 | * Skip to last blocks looked up if same parent inode. | |
690 | */ | |
691 | if (pagino != NULLAGINO && | |
692 | pag->pagl_pagino == pagino && | |
693 | pag->pagl_leftrec != NULLAGINO && | |
694 | pag->pagl_rightrec != NULLAGINO) { | |
695 | error = xfs_ialloc_get_rec(tcur, pag->pagl_leftrec, | |
696 | &trec, &doneleft, 1); | |
697 | if (error) | |
698 | goto error1; | |
4254b0bb | 699 | |
bd169565 DC |
700 | error = xfs_ialloc_get_rec(cur, pag->pagl_rightrec, |
701 | &rec, &doneright, 0); | |
702 | if (error) | |
703 | goto error1; | |
704 | } else { | |
705 | /* search left with tcur, back up 1 record */ | |
706 | error = xfs_ialloc_next_rec(tcur, &trec, &doneleft, 1); | |
707 | if (error) | |
708 | goto error1; | |
709 | ||
710 | /* search right with cur, go forward 1 record. */ | |
711 | error = xfs_ialloc_next_rec(cur, &rec, &doneright, 0); | |
712 | if (error) | |
713 | goto error1; | |
714 | } | |
4254b0bb CH |
715 | |
716 | /* | |
717 | * Loop until we find an inode chunk with a free inode. | |
718 | */ | |
719 | while (!doneleft || !doneright) { | |
720 | int useleft; /* using left inode chunk this time */ | |
721 | ||
bd169565 DC |
722 | if (!--searchdistance) { |
723 | /* | |
724 | * Not in range - save last search | |
725 | * location and allocate a new inode | |
726 | */ | |
3b826386 | 727 | xfs_btree_del_cursor(tcur, XFS_BTREE_NOERROR); |
bd169565 DC |
728 | pag->pagl_leftrec = trec.ir_startino; |
729 | pag->pagl_rightrec = rec.ir_startino; | |
730 | pag->pagl_pagino = pagino; | |
731 | goto newino; | |
732 | } | |
733 | ||
4254b0bb CH |
734 | /* figure out the closer block if both are valid. */ |
735 | if (!doneleft && !doneright) { | |
736 | useleft = pagino - | |
737 | (trec.ir_startino + XFS_INODES_PER_CHUNK - 1) < | |
738 | rec.ir_startino - pagino; | |
739 | } else { | |
740 | useleft = !doneleft; | |
1da177e4 | 741 | } |
4254b0bb CH |
742 | |
743 | /* free inodes to the left? */ | |
744 | if (useleft && trec.ir_freecount) { | |
745 | rec = trec; | |
746 | xfs_btree_del_cursor(cur, XFS_BTREE_NOERROR); | |
747 | cur = tcur; | |
bd169565 DC |
748 | |
749 | pag->pagl_leftrec = trec.ir_startino; | |
750 | pag->pagl_rightrec = rec.ir_startino; | |
751 | pag->pagl_pagino = pagino; | |
4254b0bb | 752 | goto alloc_inode; |
1da177e4 | 753 | } |
1da177e4 | 754 | |
4254b0bb CH |
755 | /* free inodes to the right? */ |
756 | if (!useleft && rec.ir_freecount) { | |
757 | xfs_btree_del_cursor(tcur, XFS_BTREE_NOERROR); | |
bd169565 DC |
758 | |
759 | pag->pagl_leftrec = trec.ir_startino; | |
760 | pag->pagl_rightrec = rec.ir_startino; | |
761 | pag->pagl_pagino = pagino; | |
4254b0bb | 762 | goto alloc_inode; |
1da177e4 | 763 | } |
4254b0bb CH |
764 | |
765 | /* get next record to check */ | |
766 | if (useleft) { | |
767 | error = xfs_ialloc_next_rec(tcur, &trec, | |
768 | &doneleft, 1); | |
769 | } else { | |
770 | error = xfs_ialloc_next_rec(cur, &rec, | |
771 | &doneright, 0); | |
772 | } | |
773 | if (error) | |
774 | goto error1; | |
1da177e4 | 775 | } |
bd169565 DC |
776 | |
777 | /* | |
778 | * We've reached the end of the btree. because | |
779 | * we are only searching a small chunk of the | |
780 | * btree each search, there is obviously free | |
781 | * inodes closer to the parent inode than we | |
782 | * are now. restart the search again. | |
783 | */ | |
784 | pag->pagl_pagino = NULLAGINO; | |
785 | pag->pagl_leftrec = NULLAGINO; | |
786 | pag->pagl_rightrec = NULLAGINO; | |
787 | xfs_btree_del_cursor(tcur, XFS_BTREE_NOERROR); | |
788 | xfs_btree_del_cursor(cur, XFS_BTREE_NOERROR); | |
789 | goto restart_pagno; | |
1da177e4 | 790 | } |
4254b0bb | 791 | |
1da177e4 | 792 | /* |
4254b0bb | 793 | * In a different AG from the parent. |
1da177e4 LT |
794 | * See if the most recently allocated block has any free. |
795 | */ | |
bd169565 | 796 | newino: |
69ef921b | 797 | if (agi->agi_newino != cpu_to_be32(NULLAGINO)) { |
21875505 CH |
798 | error = xfs_inobt_lookup(cur, be32_to_cpu(agi->agi_newino), |
799 | XFS_LOOKUP_EQ, &i); | |
4254b0bb | 800 | if (error) |
1da177e4 | 801 | goto error0; |
4254b0bb CH |
802 | |
803 | if (i == 1) { | |
804 | error = xfs_inobt_get_rec(cur, &rec, &j); | |
805 | if (error) | |
806 | goto error0; | |
807 | ||
808 | if (j == 1 && rec.ir_freecount > 0) { | |
809 | /* | |
810 | * The last chunk allocated in the group | |
811 | * still has a free inode. | |
812 | */ | |
813 | goto alloc_inode; | |
814 | } | |
1da177e4 | 815 | } |
bd169565 | 816 | } |
4254b0bb | 817 | |
bd169565 DC |
818 | /* |
819 | * None left in the last group, search the whole AG | |
820 | */ | |
821 | error = xfs_inobt_lookup(cur, 0, XFS_LOOKUP_GE, &i); | |
822 | if (error) | |
823 | goto error0; | |
824 | XFS_WANT_CORRUPTED_GOTO(i == 1, error0); | |
825 | ||
826 | for (;;) { | |
827 | error = xfs_inobt_get_rec(cur, &rec, &i); | |
828 | if (error) | |
829 | goto error0; | |
830 | XFS_WANT_CORRUPTED_GOTO(i == 1, error0); | |
831 | if (rec.ir_freecount > 0) | |
832 | break; | |
833 | error = xfs_btree_increment(cur, 0, &i); | |
4254b0bb CH |
834 | if (error) |
835 | goto error0; | |
836 | XFS_WANT_CORRUPTED_GOTO(i == 1, error0); | |
1da177e4 | 837 | } |
4254b0bb CH |
838 | |
839 | alloc_inode: | |
824c3131 | 840 | offset = xfs_lowbit64(rec.ir_free); |
1da177e4 LT |
841 | ASSERT(offset >= 0); |
842 | ASSERT(offset < XFS_INODES_PER_CHUNK); | |
843 | ASSERT((XFS_AGINO_TO_OFFSET(mp, rec.ir_startino) % | |
844 | XFS_INODES_PER_CHUNK) == 0); | |
845 | ino = XFS_AGINO_TO_INO(mp, agno, rec.ir_startino + offset); | |
0d87e656 | 846 | rec.ir_free &= ~XFS_INOBT_MASK(offset); |
1da177e4 | 847 | rec.ir_freecount--; |
afabc24a CH |
848 | error = xfs_inobt_update(cur, &rec); |
849 | if (error) | |
1da177e4 | 850 | goto error0; |
413d57c9 | 851 | be32_add_cpu(&agi->agi_freecount, -1); |
1da177e4 | 852 | xfs_ialloc_log_agi(tp, agbp, XFS_AGI_FREECOUNT); |
44b56e0a | 853 | pag->pagi_freecount--; |
1da177e4 | 854 | |
0b48db80 DC |
855 | error = xfs_check_agi_freecount(cur, agi); |
856 | if (error) | |
857 | goto error0; | |
858 | ||
1da177e4 LT |
859 | xfs_btree_del_cursor(cur, XFS_BTREE_NOERROR); |
860 | xfs_trans_mod_sb(tp, XFS_TRANS_SB_IFREE, -1); | |
44b56e0a | 861 | xfs_perag_put(pag); |
1da177e4 LT |
862 | *inop = ino; |
863 | return 0; | |
864 | error1: | |
865 | xfs_btree_del_cursor(tcur, XFS_BTREE_ERROR); | |
866 | error0: | |
867 | xfs_btree_del_cursor(cur, XFS_BTREE_ERROR); | |
44b56e0a | 868 | xfs_perag_put(pag); |
1da177e4 LT |
869 | return error; |
870 | } | |
871 | ||
f2ecc5e4 CH |
872 | /* |
873 | * Allocate an inode on disk. | |
874 | * | |
875 | * Mode is used to tell whether the new inode will need space, and whether it | |
876 | * is a directory. | |
877 | * | |
878 | * This function is designed to be called twice if it has to do an allocation | |
879 | * to make more free inodes. On the first call, *IO_agbp should be set to NULL. | |
880 | * If an inode is available without having to performn an allocation, an inode | |
cd856db6 CM |
881 | * number is returned. In this case, *IO_agbp is set to NULL. If an allocation |
882 | * needs to be done, xfs_dialloc returns the current AGI buffer in *IO_agbp. | |
883 | * The caller should then commit the current transaction, allocate a | |
f2ecc5e4 CH |
884 | * new transaction, and call xfs_dialloc() again, passing in the previous value |
885 | * of *IO_agbp. IO_agbp should be held across the transactions. Since the AGI | |
886 | * buffer is locked across the two calls, the second call is guaranteed to have | |
887 | * a free inode available. | |
888 | * | |
889 | * Once we successfully pick an inode its number is returned and the on-disk | |
890 | * data structures are updated. The inode itself is not read in, since doing so | |
891 | * would break ordering constraints with xfs_reclaim. | |
892 | */ | |
893 | int | |
894 | xfs_dialloc( | |
895 | struct xfs_trans *tp, | |
896 | xfs_ino_t parent, | |
897 | umode_t mode, | |
898 | int okalloc, | |
899 | struct xfs_buf **IO_agbp, | |
f2ecc5e4 CH |
900 | xfs_ino_t *inop) |
901 | { | |
55d6af64 | 902 | struct xfs_mount *mp = tp->t_mountp; |
f2ecc5e4 CH |
903 | struct xfs_buf *agbp; |
904 | xfs_agnumber_t agno; | |
f2ecc5e4 CH |
905 | int error; |
906 | int ialloced; | |
907 | int noroom = 0; | |
be60fe54 | 908 | xfs_agnumber_t start_agno; |
f2ecc5e4 CH |
909 | struct xfs_perag *pag; |
910 | ||
4bb61069 | 911 | if (*IO_agbp) { |
f2ecc5e4 | 912 | /* |
4bb61069 CH |
913 | * If the caller passes in a pointer to the AGI buffer, |
914 | * continue where we left off before. In this case, we | |
f2ecc5e4 CH |
915 | * know that the allocation group has free inodes. |
916 | */ | |
917 | agbp = *IO_agbp; | |
4bb61069 | 918 | goto out_alloc; |
f2ecc5e4 | 919 | } |
4bb61069 CH |
920 | |
921 | /* | |
922 | * We do not have an agbp, so select an initial allocation | |
923 | * group for inode allocation. | |
924 | */ | |
be60fe54 CH |
925 | start_agno = xfs_ialloc_ag_select(tp, parent, mode, okalloc); |
926 | if (start_agno == NULLAGNUMBER) { | |
4bb61069 CH |
927 | *inop = NULLFSINO; |
928 | return 0; | |
929 | } | |
55d6af64 | 930 | |
f2ecc5e4 CH |
931 | /* |
932 | * If we have already hit the ceiling of inode blocks then clear | |
933 | * okalloc so we scan all available agi structures for a free | |
934 | * inode. | |
935 | */ | |
f2ecc5e4 CH |
936 | if (mp->m_maxicount && |
937 | mp->m_sb.sb_icount + XFS_IALLOC_INODES(mp) > mp->m_maxicount) { | |
938 | noroom = 1; | |
939 | okalloc = 0; | |
940 | } | |
941 | ||
942 | /* | |
943 | * Loop until we find an allocation group that either has free inodes | |
944 | * or in which we can allocate some inodes. Iterate through the | |
945 | * allocation groups upward, wrapping at the end. | |
946 | */ | |
be60fe54 CH |
947 | agno = start_agno; |
948 | for (;;) { | |
949 | pag = xfs_perag_get(mp, agno); | |
950 | if (!pag->pagi_inodeok) { | |
951 | xfs_ialloc_next_ag(mp); | |
952 | goto nextag; | |
953 | } | |
954 | ||
955 | if (!pag->pagi_init) { | |
956 | error = xfs_ialloc_pagi_init(mp, tp, agno); | |
957 | if (error) | |
958 | goto out_error; | |
f2ecc5e4 | 959 | } |
be60fe54 | 960 | |
f2ecc5e4 | 961 | /* |
be60fe54 | 962 | * Do a first racy fast path check if this AG is usable. |
f2ecc5e4 | 963 | */ |
be60fe54 CH |
964 | if (!pag->pagi_freecount && !okalloc) |
965 | goto nextag; | |
966 | ||
c4982110 CH |
967 | /* |
968 | * Then read in the AGI buffer and recheck with the AGI buffer | |
969 | * lock held. | |
970 | */ | |
be60fe54 CH |
971 | error = xfs_ialloc_read_agi(mp, tp, agno, &agbp); |
972 | if (error) | |
973 | goto out_error; | |
974 | ||
be60fe54 CH |
975 | if (pag->pagi_freecount) { |
976 | xfs_perag_put(pag); | |
977 | goto out_alloc; | |
978 | } | |
979 | ||
c4982110 CH |
980 | if (!okalloc) |
981 | goto nextag_relse_buffer; | |
982 | ||
be60fe54 CH |
983 | |
984 | error = xfs_ialloc_ag_alloc(tp, agbp, &ialloced); | |
985 | if (error) { | |
986 | xfs_trans_brelse(tp, agbp); | |
987 | ||
988 | if (error != ENOSPC) | |
989 | goto out_error; | |
990 | ||
991 | xfs_perag_put(pag); | |
f2ecc5e4 | 992 | *inop = NULLFSINO; |
be60fe54 | 993 | return 0; |
f2ecc5e4 | 994 | } |
be60fe54 CH |
995 | |
996 | if (ialloced) { | |
997 | /* | |
998 | * We successfully allocated some inodes, return | |
999 | * the current context to the caller so that it | |
1000 | * can commit the current transaction and call | |
1001 | * us again where we left off. | |
1002 | */ | |
1003 | ASSERT(pag->pagi_freecount > 0); | |
f2ecc5e4 | 1004 | xfs_perag_put(pag); |
be60fe54 CH |
1005 | |
1006 | *IO_agbp = agbp; | |
1007 | *inop = NULLFSINO; | |
1008 | return 0; | |
f2ecc5e4 | 1009 | } |
be60fe54 | 1010 | |
c4982110 CH |
1011 | nextag_relse_buffer: |
1012 | xfs_trans_brelse(tp, agbp); | |
be60fe54 | 1013 | nextag: |
f2ecc5e4 | 1014 | xfs_perag_put(pag); |
be60fe54 CH |
1015 | if (++agno == mp->m_sb.sb_agcount) |
1016 | agno = 0; | |
1017 | if (agno == start_agno) { | |
1018 | *inop = NULLFSINO; | |
1019 | return noroom ? ENOSPC : 0; | |
1020 | } | |
f2ecc5e4 CH |
1021 | } |
1022 | ||
4bb61069 | 1023 | out_alloc: |
f2ecc5e4 CH |
1024 | *IO_agbp = NULL; |
1025 | return xfs_dialloc_ag(tp, agbp, parent, inop); | |
be60fe54 CH |
1026 | out_error: |
1027 | xfs_perag_put(pag); | |
1028 | return XFS_ERROR(error); | |
f2ecc5e4 CH |
1029 | } |
1030 | ||
1da177e4 LT |
1031 | /* |
1032 | * Free disk inode. Carefully avoids touching the incore inode, all | |
1033 | * manipulations incore are the caller's responsibility. | |
1034 | * The on-disk inode is not changed by this operation, only the | |
1035 | * btree (free inode mask) is changed. | |
1036 | */ | |
1037 | int | |
1038 | xfs_difree( | |
1039 | xfs_trans_t *tp, /* transaction pointer */ | |
1040 | xfs_ino_t inode, /* inode to be freed */ | |
1041 | xfs_bmap_free_t *flist, /* extents to free */ | |
1042 | int *delete, /* set if inode cluster was deleted */ | |
1043 | xfs_ino_t *first_ino) /* first inode in deleted cluster */ | |
1044 | { | |
1045 | /* REFERENCED */ | |
1046 | xfs_agblock_t agbno; /* block number containing inode */ | |
1047 | xfs_buf_t *agbp; /* buffer containing allocation group header */ | |
1048 | xfs_agino_t agino; /* inode number relative to allocation group */ | |
1049 | xfs_agnumber_t agno; /* allocation group number */ | |
1050 | xfs_agi_t *agi; /* allocation group header */ | |
1051 | xfs_btree_cur_t *cur; /* inode btree cursor */ | |
1052 | int error; /* error return value */ | |
1053 | int i; /* result code */ | |
1054 | int ilen; /* inodes in an inode cluster */ | |
1055 | xfs_mount_t *mp; /* mount structure for filesystem */ | |
1056 | int off; /* offset of inode in inode chunk */ | |
61a25848 | 1057 | xfs_inobt_rec_incore_t rec; /* btree record */ |
44b56e0a | 1058 | struct xfs_perag *pag; |
1da177e4 LT |
1059 | |
1060 | mp = tp->t_mountp; | |
1061 | ||
1062 | /* | |
1063 | * Break up inode number into its components. | |
1064 | */ | |
1065 | agno = XFS_INO_TO_AGNO(mp, inode); | |
1066 | if (agno >= mp->m_sb.sb_agcount) { | |
0b932ccc DC |
1067 | xfs_warn(mp, "%s: agno >= mp->m_sb.sb_agcount (%d >= %d).", |
1068 | __func__, agno, mp->m_sb.sb_agcount); | |
1da177e4 LT |
1069 | ASSERT(0); |
1070 | return XFS_ERROR(EINVAL); | |
1071 | } | |
1072 | agino = XFS_INO_TO_AGINO(mp, inode); | |
1073 | if (inode != XFS_AGINO_TO_INO(mp, agno, agino)) { | |
0b932ccc DC |
1074 | xfs_warn(mp, "%s: inode != XFS_AGINO_TO_INO() (%llu != %llu).", |
1075 | __func__, (unsigned long long)inode, | |
1076 | (unsigned long long)XFS_AGINO_TO_INO(mp, agno, agino)); | |
1da177e4 LT |
1077 | ASSERT(0); |
1078 | return XFS_ERROR(EINVAL); | |
1079 | } | |
1080 | agbno = XFS_AGINO_TO_AGBNO(mp, agino); | |
1081 | if (agbno >= mp->m_sb.sb_agblocks) { | |
0b932ccc DC |
1082 | xfs_warn(mp, "%s: agbno >= mp->m_sb.sb_agblocks (%d >= %d).", |
1083 | __func__, agbno, mp->m_sb.sb_agblocks); | |
1da177e4 LT |
1084 | ASSERT(0); |
1085 | return XFS_ERROR(EINVAL); | |
1086 | } | |
1087 | /* | |
1088 | * Get the allocation group header. | |
1089 | */ | |
1da177e4 | 1090 | error = xfs_ialloc_read_agi(mp, tp, agno, &agbp); |
1da177e4 | 1091 | if (error) { |
0b932ccc DC |
1092 | xfs_warn(mp, "%s: xfs_ialloc_read_agi() returned error %d.", |
1093 | __func__, error); | |
1da177e4 LT |
1094 | return error; |
1095 | } | |
1096 | agi = XFS_BUF_TO_AGI(agbp); | |
69ef921b | 1097 | ASSERT(agi->agi_magicnum == cpu_to_be32(XFS_AGI_MAGIC)); |
16259e7d | 1098 | ASSERT(agbno < be32_to_cpu(agi->agi_length)); |
1da177e4 LT |
1099 | /* |
1100 | * Initialize the cursor. | |
1101 | */ | |
561f7d17 | 1102 | cur = xfs_inobt_init_cursor(mp, tp, agbp, agno); |
1da177e4 | 1103 | |
0b48db80 DC |
1104 | error = xfs_check_agi_freecount(cur, agi); |
1105 | if (error) | |
1106 | goto error0; | |
1107 | ||
1da177e4 LT |
1108 | /* |
1109 | * Look for the entry describing this inode. | |
1110 | */ | |
21875505 | 1111 | if ((error = xfs_inobt_lookup(cur, agino, XFS_LOOKUP_LE, &i))) { |
0b932ccc DC |
1112 | xfs_warn(mp, "%s: xfs_inobt_lookup() returned error %d.", |
1113 | __func__, error); | |
1da177e4 LT |
1114 | goto error0; |
1115 | } | |
1116 | XFS_WANT_CORRUPTED_GOTO(i == 1, error0); | |
2e287a73 CH |
1117 | error = xfs_inobt_get_rec(cur, &rec, &i); |
1118 | if (error) { | |
0b932ccc DC |
1119 | xfs_warn(mp, "%s: xfs_inobt_get_rec() returned error %d.", |
1120 | __func__, error); | |
1da177e4 LT |
1121 | goto error0; |
1122 | } | |
1123 | XFS_WANT_CORRUPTED_GOTO(i == 1, error0); | |
1124 | /* | |
1125 | * Get the offset in the inode chunk. | |
1126 | */ | |
1127 | off = agino - rec.ir_startino; | |
1128 | ASSERT(off >= 0 && off < XFS_INODES_PER_CHUNK); | |
0d87e656 | 1129 | ASSERT(!(rec.ir_free & XFS_INOBT_MASK(off))); |
1da177e4 LT |
1130 | /* |
1131 | * Mark the inode free & increment the count. | |
1132 | */ | |
0d87e656 | 1133 | rec.ir_free |= XFS_INOBT_MASK(off); |
1da177e4 LT |
1134 | rec.ir_freecount++; |
1135 | ||
1136 | /* | |
c41564b5 | 1137 | * When an inode cluster is free, it becomes eligible for removal |
1da177e4 | 1138 | */ |
1bd960ee | 1139 | if (!(mp->m_flags & XFS_MOUNT_IKEEP) && |
1da177e4 LT |
1140 | (rec.ir_freecount == XFS_IALLOC_INODES(mp))) { |
1141 | ||
1142 | *delete = 1; | |
1143 | *first_ino = XFS_AGINO_TO_INO(mp, agno, rec.ir_startino); | |
1144 | ||
1145 | /* | |
1146 | * Remove the inode cluster from the AGI B+Tree, adjust the | |
1147 | * AGI and Superblock inode counts, and mark the disk space | |
1148 | * to be freed when the transaction is committed. | |
1149 | */ | |
1150 | ilen = XFS_IALLOC_INODES(mp); | |
413d57c9 MS |
1151 | be32_add_cpu(&agi->agi_count, -ilen); |
1152 | be32_add_cpu(&agi->agi_freecount, -(ilen - 1)); | |
1da177e4 | 1153 | xfs_ialloc_log_agi(tp, agbp, XFS_AGI_COUNT | XFS_AGI_FREECOUNT); |
44b56e0a DC |
1154 | pag = xfs_perag_get(mp, agno); |
1155 | pag->pagi_freecount -= ilen - 1; | |
1156 | xfs_perag_put(pag); | |
1da177e4 LT |
1157 | xfs_trans_mod_sb(tp, XFS_TRANS_SB_ICOUNT, -ilen); |
1158 | xfs_trans_mod_sb(tp, XFS_TRANS_SB_IFREE, -(ilen - 1)); | |
1159 | ||
91cca5df | 1160 | if ((error = xfs_btree_delete(cur, &i))) { |
0b932ccc DC |
1161 | xfs_warn(mp, "%s: xfs_btree_delete returned error %d.", |
1162 | __func__, error); | |
1da177e4 LT |
1163 | goto error0; |
1164 | } | |
1165 | ||
1166 | xfs_bmap_add_free(XFS_AGB_TO_FSB(mp, | |
1167 | agno, XFS_INO_TO_AGBNO(mp,rec.ir_startino)), | |
1168 | XFS_IALLOC_BLOCKS(mp), flist, mp); | |
1169 | } else { | |
1170 | *delete = 0; | |
1171 | ||
afabc24a CH |
1172 | error = xfs_inobt_update(cur, &rec); |
1173 | if (error) { | |
0b932ccc DC |
1174 | xfs_warn(mp, "%s: xfs_inobt_update returned error %d.", |
1175 | __func__, error); | |
1da177e4 LT |
1176 | goto error0; |
1177 | } | |
afabc24a | 1178 | |
1da177e4 LT |
1179 | /* |
1180 | * Change the inode free counts and log the ag/sb changes. | |
1181 | */ | |
413d57c9 | 1182 | be32_add_cpu(&agi->agi_freecount, 1); |
1da177e4 | 1183 | xfs_ialloc_log_agi(tp, agbp, XFS_AGI_FREECOUNT); |
44b56e0a DC |
1184 | pag = xfs_perag_get(mp, agno); |
1185 | pag->pagi_freecount++; | |
1186 | xfs_perag_put(pag); | |
1da177e4 LT |
1187 | xfs_trans_mod_sb(tp, XFS_TRANS_SB_IFREE, 1); |
1188 | } | |
1189 | ||
0b48db80 DC |
1190 | error = xfs_check_agi_freecount(cur, agi); |
1191 | if (error) | |
1192 | goto error0; | |
1da177e4 | 1193 | |
1da177e4 LT |
1194 | xfs_btree_del_cursor(cur, XFS_BTREE_NOERROR); |
1195 | return 0; | |
1196 | ||
1197 | error0: | |
1198 | xfs_btree_del_cursor(cur, XFS_BTREE_ERROR); | |
1199 | return error; | |
1200 | } | |
1201 | ||
7124fe0a DC |
1202 | STATIC int |
1203 | xfs_imap_lookup( | |
1204 | struct xfs_mount *mp, | |
1205 | struct xfs_trans *tp, | |
1206 | xfs_agnumber_t agno, | |
1207 | xfs_agino_t agino, | |
1208 | xfs_agblock_t agbno, | |
1209 | xfs_agblock_t *chunk_agbno, | |
1210 | xfs_agblock_t *offset_agbno, | |
1211 | int flags) | |
1212 | { | |
1213 | struct xfs_inobt_rec_incore rec; | |
1214 | struct xfs_btree_cur *cur; | |
1215 | struct xfs_buf *agbp; | |
7124fe0a DC |
1216 | int error; |
1217 | int i; | |
1218 | ||
1219 | error = xfs_ialloc_read_agi(mp, tp, agno, &agbp); | |
1220 | if (error) { | |
53487786 DC |
1221 | xfs_alert(mp, |
1222 | "%s: xfs_ialloc_read_agi() returned error %d, agno %d", | |
1223 | __func__, error, agno); | |
7124fe0a DC |
1224 | return error; |
1225 | } | |
1226 | ||
1227 | /* | |
4536f2ad DC |
1228 | * Lookup the inode record for the given agino. If the record cannot be |
1229 | * found, then it's an invalid inode number and we should abort. Once | |
1230 | * we have a record, we need to ensure it contains the inode number | |
1231 | * we are looking up. | |
7124fe0a DC |
1232 | */ |
1233 | cur = xfs_inobt_init_cursor(mp, tp, agbp, agno); | |
4536f2ad | 1234 | error = xfs_inobt_lookup(cur, agino, XFS_LOOKUP_LE, &i); |
7124fe0a DC |
1235 | if (!error) { |
1236 | if (i) | |
1237 | error = xfs_inobt_get_rec(cur, &rec, &i); | |
1238 | if (!error && i == 0) | |
1239 | error = EINVAL; | |
1240 | } | |
1241 | ||
1242 | xfs_trans_brelse(tp, agbp); | |
1243 | xfs_btree_del_cursor(cur, XFS_BTREE_NOERROR); | |
1244 | if (error) | |
1245 | return error; | |
1246 | ||
4536f2ad DC |
1247 | /* check that the returned record contains the required inode */ |
1248 | if (rec.ir_startino > agino || | |
1249 | rec.ir_startino + XFS_IALLOC_INODES(mp) <= agino) | |
1250 | return EINVAL; | |
1251 | ||
7124fe0a | 1252 | /* for untrusted inodes check it is allocated first */ |
1920779e | 1253 | if ((flags & XFS_IGET_UNTRUSTED) && |
7124fe0a DC |
1254 | (rec.ir_free & XFS_INOBT_MASK(agino - rec.ir_startino))) |
1255 | return EINVAL; | |
1256 | ||
1257 | *chunk_agbno = XFS_AGINO_TO_AGBNO(mp, rec.ir_startino); | |
1258 | *offset_agbno = agbno - *chunk_agbno; | |
1259 | return 0; | |
1260 | } | |
1261 | ||
1da177e4 | 1262 | /* |
94e1b69d | 1263 | * Return the location of the inode in imap, for mapping it into a buffer. |
1da177e4 | 1264 | */ |
1da177e4 | 1265 | int |
94e1b69d CH |
1266 | xfs_imap( |
1267 | xfs_mount_t *mp, /* file system mount structure */ | |
1268 | xfs_trans_t *tp, /* transaction pointer */ | |
1da177e4 | 1269 | xfs_ino_t ino, /* inode to locate */ |
94e1b69d CH |
1270 | struct xfs_imap *imap, /* location map structure */ |
1271 | uint flags) /* flags for inode btree lookup */ | |
1da177e4 LT |
1272 | { |
1273 | xfs_agblock_t agbno; /* block number of inode in the alloc group */ | |
1da177e4 LT |
1274 | xfs_agino_t agino; /* inode number within alloc group */ |
1275 | xfs_agnumber_t agno; /* allocation group number */ | |
1276 | int blks_per_cluster; /* num blocks per inode cluster */ | |
1277 | xfs_agblock_t chunk_agbno; /* first block in inode chunk */ | |
1da177e4 | 1278 | xfs_agblock_t cluster_agbno; /* first block in inode cluster */ |
1da177e4 | 1279 | int error; /* error code */ |
1da177e4 LT |
1280 | int offset; /* index of inode in its buffer */ |
1281 | int offset_agbno; /* blks from chunk start to inode */ | |
1282 | ||
1283 | ASSERT(ino != NULLFSINO); | |
94e1b69d | 1284 | |
1da177e4 LT |
1285 | /* |
1286 | * Split up the inode number into its parts. | |
1287 | */ | |
1288 | agno = XFS_INO_TO_AGNO(mp, ino); | |
1289 | agino = XFS_INO_TO_AGINO(mp, ino); | |
1290 | agbno = XFS_AGINO_TO_AGBNO(mp, agino); | |
1291 | if (agno >= mp->m_sb.sb_agcount || agbno >= mp->m_sb.sb_agblocks || | |
1292 | ino != XFS_AGINO_TO_INO(mp, agno, agino)) { | |
1293 | #ifdef DEBUG | |
1920779e DC |
1294 | /* |
1295 | * Don't output diagnostic information for untrusted inodes | |
1296 | * as they can be invalid without implying corruption. | |
1297 | */ | |
1298 | if (flags & XFS_IGET_UNTRUSTED) | |
4d1a2ed3 | 1299 | return XFS_ERROR(EINVAL); |
1da177e4 | 1300 | if (agno >= mp->m_sb.sb_agcount) { |
53487786 DC |
1301 | xfs_alert(mp, |
1302 | "%s: agno (%d) >= mp->m_sb.sb_agcount (%d)", | |
1303 | __func__, agno, mp->m_sb.sb_agcount); | |
1da177e4 LT |
1304 | } |
1305 | if (agbno >= mp->m_sb.sb_agblocks) { | |
53487786 DC |
1306 | xfs_alert(mp, |
1307 | "%s: agbno (0x%llx) >= mp->m_sb.sb_agblocks (0x%lx)", | |
1308 | __func__, (unsigned long long)agbno, | |
1309 | (unsigned long)mp->m_sb.sb_agblocks); | |
1da177e4 LT |
1310 | } |
1311 | if (ino != XFS_AGINO_TO_INO(mp, agno, agino)) { | |
53487786 DC |
1312 | xfs_alert(mp, |
1313 | "%s: ino (0x%llx) != XFS_AGINO_TO_INO() (0x%llx)", | |
1314 | __func__, ino, | |
1315 | XFS_AGINO_TO_INO(mp, agno, agino)); | |
1da177e4 | 1316 | } |
745b1f47 | 1317 | xfs_stack_trace(); |
1da177e4 LT |
1318 | #endif /* DEBUG */ |
1319 | return XFS_ERROR(EINVAL); | |
1320 | } | |
94e1b69d | 1321 | |
7124fe0a DC |
1322 | blks_per_cluster = XFS_INODE_CLUSTER_SIZE(mp) >> mp->m_sb.sb_blocklog; |
1323 | ||
1324 | /* | |
1325 | * For bulkstat and handle lookups, we have an untrusted inode number | |
1326 | * that we have to verify is valid. We cannot do this just by reading | |
1327 | * the inode buffer as it may have been unlinked and removed leaving | |
1328 | * inodes in stale state on disk. Hence we have to do a btree lookup | |
1329 | * in all cases where an untrusted inode number is passed. | |
1330 | */ | |
1920779e | 1331 | if (flags & XFS_IGET_UNTRUSTED) { |
7124fe0a DC |
1332 | error = xfs_imap_lookup(mp, tp, agno, agino, agbno, |
1333 | &chunk_agbno, &offset_agbno, flags); | |
1334 | if (error) | |
1335 | return error; | |
1336 | goto out_map; | |
1337 | } | |
1338 | ||
94e1b69d CH |
1339 | /* |
1340 | * If the inode cluster size is the same as the blocksize or | |
1341 | * smaller we get to the buffer by simple arithmetics. | |
1342 | */ | |
1343 | if (XFS_INODE_CLUSTER_SIZE(mp) <= mp->m_sb.sb_blocksize) { | |
1da177e4 LT |
1344 | offset = XFS_INO_TO_OFFSET(mp, ino); |
1345 | ASSERT(offset < mp->m_sb.sb_inopblock); | |
94e1b69d CH |
1346 | |
1347 | imap->im_blkno = XFS_AGB_TO_DADDR(mp, agno, agbno); | |
1348 | imap->im_len = XFS_FSB_TO_BB(mp, 1); | |
1349 | imap->im_boffset = (ushort)(offset << mp->m_sb.sb_inodelog); | |
1da177e4 LT |
1350 | return 0; |
1351 | } | |
94e1b69d | 1352 | |
94e1b69d CH |
1353 | /* |
1354 | * If the inode chunks are aligned then use simple maths to | |
1355 | * find the location. Otherwise we have to do a btree | |
1356 | * lookup to find the location. | |
1357 | */ | |
1da177e4 LT |
1358 | if (mp->m_inoalign_mask) { |
1359 | offset_agbno = agbno & mp->m_inoalign_mask; | |
1360 | chunk_agbno = agbno - offset_agbno; | |
1361 | } else { | |
7124fe0a DC |
1362 | error = xfs_imap_lookup(mp, tp, agno, agino, agbno, |
1363 | &chunk_agbno, &offset_agbno, flags); | |
1da177e4 LT |
1364 | if (error) |
1365 | return error; | |
1da177e4 | 1366 | } |
94e1b69d | 1367 | |
7124fe0a | 1368 | out_map: |
1da177e4 LT |
1369 | ASSERT(agbno >= chunk_agbno); |
1370 | cluster_agbno = chunk_agbno + | |
1371 | ((offset_agbno / blks_per_cluster) * blks_per_cluster); | |
1372 | offset = ((agbno - cluster_agbno) * mp->m_sb.sb_inopblock) + | |
1373 | XFS_INO_TO_OFFSET(mp, ino); | |
94e1b69d CH |
1374 | |
1375 | imap->im_blkno = XFS_AGB_TO_DADDR(mp, agno, cluster_agbno); | |
1376 | imap->im_len = XFS_FSB_TO_BB(mp, blks_per_cluster); | |
1377 | imap->im_boffset = (ushort)(offset << mp->m_sb.sb_inodelog); | |
1378 | ||
1379 | /* | |
1380 | * If the inode number maps to a block outside the bounds | |
1381 | * of the file system then return NULL rather than calling | |
1382 | * read_buf and panicing when we get an error from the | |
1383 | * driver. | |
1384 | */ | |
1385 | if ((imap->im_blkno + imap->im_len) > | |
1386 | XFS_FSB_TO_BB(mp, mp->m_sb.sb_dblocks)) { | |
53487786 DC |
1387 | xfs_alert(mp, |
1388 | "%s: (im_blkno (0x%llx) + im_len (0x%llx)) > sb_dblocks (0x%llx)", | |
1389 | __func__, (unsigned long long) imap->im_blkno, | |
94e1b69d CH |
1390 | (unsigned long long) imap->im_len, |
1391 | XFS_FSB_TO_BB(mp, mp->m_sb.sb_dblocks)); | |
1392 | return XFS_ERROR(EINVAL); | |
1393 | } | |
1da177e4 | 1394 | return 0; |
1da177e4 LT |
1395 | } |
1396 | ||
1397 | /* | |
1398 | * Compute and fill in value of m_in_maxlevels. | |
1399 | */ | |
1400 | void | |
1401 | xfs_ialloc_compute_maxlevels( | |
1402 | xfs_mount_t *mp) /* file system mount structure */ | |
1403 | { | |
1404 | int level; | |
1405 | uint maxblocks; | |
1406 | uint maxleafents; | |
1407 | int minleafrecs; | |
1408 | int minnoderecs; | |
1409 | ||
1410 | maxleafents = (1LL << XFS_INO_AGINO_BITS(mp)) >> | |
1411 | XFS_INODES_PER_CHUNK_LOG; | |
1412 | minleafrecs = mp->m_alloc_mnr[0]; | |
1413 | minnoderecs = mp->m_alloc_mnr[1]; | |
1414 | maxblocks = (maxleafents + minleafrecs - 1) / minleafrecs; | |
1415 | for (level = 1; maxblocks > 1; level++) | |
1416 | maxblocks = (maxblocks + minnoderecs - 1) / minnoderecs; | |
1417 | mp->m_in_maxlevels = level; | |
1418 | } | |
1419 | ||
1420 | /* | |
1421 | * Log specified fields for the ag hdr (inode section) | |
1422 | */ | |
1423 | void | |
1424 | xfs_ialloc_log_agi( | |
1425 | xfs_trans_t *tp, /* transaction pointer */ | |
1426 | xfs_buf_t *bp, /* allocation group header buffer */ | |
1427 | int fields) /* bitmask of fields to log */ | |
1428 | { | |
1429 | int first; /* first byte number */ | |
1430 | int last; /* last byte number */ | |
1431 | static const short offsets[] = { /* field starting offsets */ | |
1432 | /* keep in sync with bit definitions */ | |
1433 | offsetof(xfs_agi_t, agi_magicnum), | |
1434 | offsetof(xfs_agi_t, agi_versionnum), | |
1435 | offsetof(xfs_agi_t, agi_seqno), | |
1436 | offsetof(xfs_agi_t, agi_length), | |
1437 | offsetof(xfs_agi_t, agi_count), | |
1438 | offsetof(xfs_agi_t, agi_root), | |
1439 | offsetof(xfs_agi_t, agi_level), | |
1440 | offsetof(xfs_agi_t, agi_freecount), | |
1441 | offsetof(xfs_agi_t, agi_newino), | |
1442 | offsetof(xfs_agi_t, agi_dirino), | |
1443 | offsetof(xfs_agi_t, agi_unlinked), | |
1444 | sizeof(xfs_agi_t) | |
1445 | }; | |
1446 | #ifdef DEBUG | |
1447 | xfs_agi_t *agi; /* allocation group header */ | |
1448 | ||
1449 | agi = XFS_BUF_TO_AGI(bp); | |
69ef921b | 1450 | ASSERT(agi->agi_magicnum == cpu_to_be32(XFS_AGI_MAGIC)); |
1da177e4 LT |
1451 | #endif |
1452 | /* | |
1453 | * Compute byte offsets for the first and last fields. | |
1454 | */ | |
1455 | xfs_btree_offsets(fields, offsets, XFS_AGI_NUM_BITS, &first, &last); | |
1456 | /* | |
1457 | * Log the allocation group inode header buffer. | |
1458 | */ | |
1459 | xfs_trans_log_buf(tp, bp, first, last); | |
1460 | } | |
1461 | ||
5e1be0fb CH |
1462 | #ifdef DEBUG |
1463 | STATIC void | |
1464 | xfs_check_agi_unlinked( | |
1465 | struct xfs_agi *agi) | |
1466 | { | |
1467 | int i; | |
1468 | ||
1469 | for (i = 0; i < XFS_AGI_UNLINKED_BUCKETS; i++) | |
1470 | ASSERT(agi->agi_unlinked[i]); | |
1471 | } | |
1472 | #else | |
1473 | #define xfs_check_agi_unlinked(agi) | |
1474 | #endif | |
1475 | ||
3702ce6e | 1476 | static void |
612cfbfe | 1477 | xfs_agi_verify( |
3702ce6e DC |
1478 | struct xfs_buf *bp) |
1479 | { | |
1480 | struct xfs_mount *mp = bp->b_target->bt_mount; | |
1481 | struct xfs_agi *agi = XFS_BUF_TO_AGI(bp); | |
1482 | int agi_ok; | |
1483 | ||
1484 | /* | |
1485 | * Validate the magic number of the agi block. | |
1486 | */ | |
1487 | agi_ok = agi->agi_magicnum == cpu_to_be32(XFS_AGI_MAGIC) && | |
1488 | XFS_AGI_GOOD_VERSION(be32_to_cpu(agi->agi_versionnum)); | |
1489 | ||
1490 | /* | |
1491 | * during growfs operations, the perag is not fully initialised, | |
1492 | * so we can't use it for any useful checking. growfs ensures we can't | |
1493 | * use it by using uncached buffers that don't have the perag attached | |
1494 | * so we can detect and avoid this problem. | |
1495 | */ | |
1496 | if (bp->b_pag) | |
1497 | agi_ok = agi_ok && be32_to_cpu(agi->agi_seqno) == | |
1498 | bp->b_pag->pag_agno; | |
1499 | ||
1500 | if (unlikely(XFS_TEST_ERROR(!agi_ok, mp, XFS_ERRTAG_IALLOC_READ_AGI, | |
1501 | XFS_RANDOM_IALLOC_READ_AGI))) { | |
1502 | XFS_CORRUPTION_ERROR(__func__, XFS_ERRLEVEL_LOW, mp, agi); | |
1503 | xfs_buf_ioerror(bp, EFSCORRUPTED); | |
1504 | } | |
1505 | xfs_check_agi_unlinked(agi); | |
612cfbfe DC |
1506 | } |
1507 | ||
b0f539de | 1508 | void |
612cfbfe DC |
1509 | xfs_agi_write_verify( |
1510 | struct xfs_buf *bp) | |
1511 | { | |
1512 | xfs_agi_verify(bp); | |
1513 | } | |
1514 | ||
b0f539de | 1515 | static void |
612cfbfe DC |
1516 | xfs_agi_read_verify( |
1517 | struct xfs_buf *bp) | |
1518 | { | |
1519 | xfs_agi_verify(bp); | |
1520 | bp->b_pre_io = xfs_agi_write_verify; | |
3702ce6e DC |
1521 | bp->b_iodone = NULL; |
1522 | xfs_buf_ioend(bp, 0); | |
1523 | } | |
1524 | ||
1da177e4 LT |
1525 | /* |
1526 | * Read in the allocation group header (inode allocation section) | |
1527 | */ | |
1528 | int | |
5e1be0fb CH |
1529 | xfs_read_agi( |
1530 | struct xfs_mount *mp, /* file system mount structure */ | |
1531 | struct xfs_trans *tp, /* transaction pointer */ | |
1532 | xfs_agnumber_t agno, /* allocation group number */ | |
1533 | struct xfs_buf **bpp) /* allocation group hdr buf */ | |
1da177e4 | 1534 | { |
5e1be0fb | 1535 | int error; |
1da177e4 LT |
1536 | |
1537 | ASSERT(agno != NULLAGNUMBER); | |
5e1be0fb CH |
1538 | |
1539 | error = xfs_trans_read_buf(mp, tp, mp->m_ddev_targp, | |
1da177e4 | 1540 | XFS_AG_DADDR(mp, agno, XFS_AGI_DADDR(mp)), |
3702ce6e | 1541 | XFS_FSS_TO_BB(mp, 1), 0, bpp, xfs_agi_read_verify); |
1da177e4 LT |
1542 | if (error) |
1543 | return error; | |
5e1be0fb | 1544 | |
5a52c2a5 | 1545 | ASSERT(!xfs_buf_geterror(*bpp)); |
38f23232 | 1546 | xfs_buf_set_ref(*bpp, XFS_AGI_REF); |
5e1be0fb CH |
1547 | return 0; |
1548 | } | |
1549 | ||
1550 | int | |
1551 | xfs_ialloc_read_agi( | |
1552 | struct xfs_mount *mp, /* file system mount structure */ | |
1553 | struct xfs_trans *tp, /* transaction pointer */ | |
1554 | xfs_agnumber_t agno, /* allocation group number */ | |
1555 | struct xfs_buf **bpp) /* allocation group hdr buf */ | |
1556 | { | |
1557 | struct xfs_agi *agi; /* allocation group header */ | |
1558 | struct xfs_perag *pag; /* per allocation group data */ | |
1559 | int error; | |
1560 | ||
1561 | error = xfs_read_agi(mp, tp, agno, bpp); | |
1562 | if (error) | |
1563 | return error; | |
1564 | ||
1565 | agi = XFS_BUF_TO_AGI(*bpp); | |
44b56e0a | 1566 | pag = xfs_perag_get(mp, agno); |
1da177e4 | 1567 | if (!pag->pagi_init) { |
16259e7d | 1568 | pag->pagi_freecount = be32_to_cpu(agi->agi_freecount); |
92821e2b | 1569 | pag->pagi_count = be32_to_cpu(agi->agi_count); |
1da177e4 | 1570 | pag->pagi_init = 1; |
1da177e4 | 1571 | } |
1da177e4 | 1572 | |
5e1be0fb CH |
1573 | /* |
1574 | * It's possible for these to be out of sync if | |
1575 | * we are in the middle of a forced shutdown. | |
1576 | */ | |
1577 | ASSERT(pag->pagi_freecount == be32_to_cpu(agi->agi_freecount) || | |
1578 | XFS_FORCED_SHUTDOWN(mp)); | |
44b56e0a | 1579 | xfs_perag_put(pag); |
1da177e4 LT |
1580 | return 0; |
1581 | } | |
92821e2b DC |
1582 | |
1583 | /* | |
1584 | * Read in the agi to initialise the per-ag data in the mount structure | |
1585 | */ | |
1586 | int | |
1587 | xfs_ialloc_pagi_init( | |
1588 | xfs_mount_t *mp, /* file system mount structure */ | |
1589 | xfs_trans_t *tp, /* transaction pointer */ | |
1590 | xfs_agnumber_t agno) /* allocation group number */ | |
1591 | { | |
1592 | xfs_buf_t *bp = NULL; | |
1593 | int error; | |
1594 | ||
1595 | error = xfs_ialloc_read_agi(mp, tp, agno, &bp); | |
1596 | if (error) | |
1597 | return error; | |
1598 | if (bp) | |
1599 | xfs_trans_brelse(tp, bp); | |
1600 | return 0; | |
1601 | } |