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 | */ | |
1a1a3e97 | 213 | xfs_buf_zero(fbuf, 0, ninodes << mp->m_sb.sb_inodelog); |
85c0b2ab DC |
214 | for (i = 0; i < ninodes; i++) { |
215 | int ioffset = i << mp->m_sb.sb_inodelog; | |
216 | uint isize = sizeof(struct xfs_dinode); | |
217 | ||
218 | free = xfs_make_iptr(mp, fbuf, i); | |
219 | free->di_magic = cpu_to_be16(XFS_DINODE_MAGIC); | |
220 | free->di_version = version; | |
221 | free->di_gen = cpu_to_be32(gen); | |
222 | free->di_next_unlinked = cpu_to_be32(NULLAGINO); | |
223 | xfs_trans_log_buf(tp, fbuf, ioffset, ioffset + isize - 1); | |
224 | } | |
225 | xfs_trans_inode_alloc_buf(tp, fbuf); | |
226 | } | |
2a30f36d | 227 | return 0; |
85c0b2ab DC |
228 | } |
229 | ||
1da177e4 LT |
230 | /* |
231 | * Allocate new inodes in the allocation group specified by agbp. | |
232 | * Return 0 for success, else error code. | |
233 | */ | |
234 | STATIC int /* error code or 0 */ | |
235 | xfs_ialloc_ag_alloc( | |
236 | xfs_trans_t *tp, /* transaction pointer */ | |
237 | xfs_buf_t *agbp, /* alloc group buffer */ | |
238 | int *alloc) | |
239 | { | |
240 | xfs_agi_t *agi; /* allocation group header */ | |
241 | xfs_alloc_arg_t args; /* allocation argument structure */ | |
1da177e4 | 242 | xfs_btree_cur_t *cur; /* inode btree cursor */ |
92821e2b | 243 | xfs_agnumber_t agno; |
1da177e4 | 244 | int error; |
85c0b2ab | 245 | int i; |
1da177e4 LT |
246 | xfs_agino_t newino; /* new first inode's number */ |
247 | xfs_agino_t newlen; /* new number of inodes */ | |
1da177e4 | 248 | xfs_agino_t thisino; /* current inode number, for loop */ |
3ccb8b5f | 249 | int isaligned = 0; /* inode allocation at stripe unit */ |
1da177e4 | 250 | /* boundary */ |
44b56e0a | 251 | struct xfs_perag *pag; |
1da177e4 LT |
252 | |
253 | args.tp = tp; | |
254 | args.mp = tp->t_mountp; | |
255 | ||
256 | /* | |
257 | * Locking will ensure that we don't have two callers in here | |
258 | * at one time. | |
259 | */ | |
260 | newlen = XFS_IALLOC_INODES(args.mp); | |
261 | if (args.mp->m_maxicount && | |
262 | args.mp->m_sb.sb_icount + newlen > args.mp->m_maxicount) | |
263 | return XFS_ERROR(ENOSPC); | |
264 | args.minlen = args.maxlen = XFS_IALLOC_BLOCKS(args.mp); | |
265 | /* | |
3ccb8b5f GO |
266 | * First try to allocate inodes contiguous with the last-allocated |
267 | * chunk of inodes. If the filesystem is striped, this will fill | |
268 | * an entire stripe unit with inodes. | |
269 | */ | |
1da177e4 | 270 | agi = XFS_BUF_TO_AGI(agbp); |
3ccb8b5f | 271 | newino = be32_to_cpu(agi->agi_newino); |
85c0b2ab | 272 | agno = be32_to_cpu(agi->agi_seqno); |
019ff2d5 NS |
273 | args.agbno = XFS_AGINO_TO_AGBNO(args.mp, newino) + |
274 | XFS_IALLOC_BLOCKS(args.mp); | |
275 | if (likely(newino != NULLAGINO && | |
276 | (args.agbno < be32_to_cpu(agi->agi_length)))) { | |
85c0b2ab | 277 | args.fsbno = XFS_AGB_TO_FSB(args.mp, agno, args.agbno); |
3ccb8b5f GO |
278 | args.type = XFS_ALLOCTYPE_THIS_BNO; |
279 | args.mod = args.total = args.wasdel = args.isfl = | |
280 | args.userdata = args.minalignslop = 0; | |
281 | args.prod = 1; | |
75de2a91 | 282 | |
3ccb8b5f | 283 | /* |
75de2a91 DC |
284 | * We need to take into account alignment here to ensure that |
285 | * we don't modify the free list if we fail to have an exact | |
286 | * block. If we don't have an exact match, and every oher | |
287 | * attempt allocation attempt fails, we'll end up cancelling | |
288 | * a dirty transaction and shutting down. | |
289 | * | |
290 | * For an exact allocation, alignment must be 1, | |
291 | * however we need to take cluster alignment into account when | |
292 | * fixing up the freelist. Use the minalignslop field to | |
293 | * indicate that extra blocks might be required for alignment, | |
294 | * but not to use them in the actual exact allocation. | |
3ccb8b5f | 295 | */ |
75de2a91 DC |
296 | args.alignment = 1; |
297 | args.minalignslop = xfs_ialloc_cluster_alignment(&args) - 1; | |
298 | ||
299 | /* Allow space for the inode btree to split. */ | |
0d87e656 | 300 | args.minleft = args.mp->m_in_maxlevels - 1; |
3ccb8b5f GO |
301 | if ((error = xfs_alloc_vextent(&args))) |
302 | return error; | |
303 | } else | |
304 | args.fsbno = NULLFSBLOCK; | |
1da177e4 | 305 | |
3ccb8b5f GO |
306 | if (unlikely(args.fsbno == NULLFSBLOCK)) { |
307 | /* | |
308 | * Set the alignment for the allocation. | |
309 | * If stripe alignment is turned on then align at stripe unit | |
310 | * boundary. | |
019ff2d5 NS |
311 | * If the cluster size is smaller than a filesystem block |
312 | * then we're doing I/O for inodes in filesystem block size | |
3ccb8b5f GO |
313 | * pieces, so don't need alignment anyway. |
314 | */ | |
315 | isaligned = 0; | |
316 | if (args.mp->m_sinoalign) { | |
317 | ASSERT(!(args.mp->m_flags & XFS_MOUNT_NOALIGN)); | |
318 | args.alignment = args.mp->m_dalign; | |
319 | isaligned = 1; | |
75de2a91 DC |
320 | } else |
321 | args.alignment = xfs_ialloc_cluster_alignment(&args); | |
3ccb8b5f GO |
322 | /* |
323 | * Need to figure out where to allocate the inode blocks. | |
324 | * Ideally they should be spaced out through the a.g. | |
325 | * For now, just allocate blocks up front. | |
326 | */ | |
327 | args.agbno = be32_to_cpu(agi->agi_root); | |
85c0b2ab | 328 | args.fsbno = XFS_AGB_TO_FSB(args.mp, agno, args.agbno); |
3ccb8b5f GO |
329 | /* |
330 | * Allocate a fixed-size extent of inodes. | |
331 | */ | |
332 | args.type = XFS_ALLOCTYPE_NEAR_BNO; | |
333 | args.mod = args.total = args.wasdel = args.isfl = | |
334 | args.userdata = args.minalignslop = 0; | |
335 | args.prod = 1; | |
336 | /* | |
337 | * Allow space for the inode btree to split. | |
338 | */ | |
0d87e656 | 339 | args.minleft = args.mp->m_in_maxlevels - 1; |
3ccb8b5f GO |
340 | if ((error = xfs_alloc_vextent(&args))) |
341 | return error; | |
342 | } | |
019ff2d5 | 343 | |
1da177e4 LT |
344 | /* |
345 | * If stripe alignment is turned on, then try again with cluster | |
346 | * alignment. | |
347 | */ | |
348 | if (isaligned && args.fsbno == NULLFSBLOCK) { | |
349 | args.type = XFS_ALLOCTYPE_NEAR_BNO; | |
16259e7d | 350 | args.agbno = be32_to_cpu(agi->agi_root); |
85c0b2ab | 351 | args.fsbno = XFS_AGB_TO_FSB(args.mp, agno, args.agbno); |
75de2a91 | 352 | args.alignment = xfs_ialloc_cluster_alignment(&args); |
1da177e4 LT |
353 | if ((error = xfs_alloc_vextent(&args))) |
354 | return error; | |
355 | } | |
356 | ||
357 | if (args.fsbno == NULLFSBLOCK) { | |
358 | *alloc = 0; | |
359 | return 0; | |
360 | } | |
361 | ASSERT(args.len == args.minlen); | |
1da177e4 | 362 | |
359346a9 | 363 | /* |
85c0b2ab DC |
364 | * Stamp and write the inode buffers. |
365 | * | |
359346a9 DC |
366 | * Seed the new inode cluster with a random generation number. This |
367 | * prevents short-term reuse of generation numbers if a chunk is | |
368 | * freed and then immediately reallocated. We use random numbers | |
369 | * rather than a linear progression to prevent the next generation | |
370 | * number from being easily guessable. | |
371 | */ | |
2a30f36d CS |
372 | error = xfs_ialloc_inode_init(args.mp, tp, agno, args.agbno, |
373 | args.len, random32()); | |
d42f08f6 | 374 | |
2a30f36d CS |
375 | if (error) |
376 | return error; | |
85c0b2ab DC |
377 | /* |
378 | * Convert the results. | |
379 | */ | |
380 | newino = XFS_OFFBNO_TO_AGINO(args.mp, args.agbno, 0); | |
413d57c9 MS |
381 | be32_add_cpu(&agi->agi_count, newlen); |
382 | be32_add_cpu(&agi->agi_freecount, newlen); | |
44b56e0a DC |
383 | pag = xfs_perag_get(args.mp, agno); |
384 | pag->pagi_freecount += newlen; | |
385 | xfs_perag_put(pag); | |
16259e7d | 386 | agi->agi_newino = cpu_to_be32(newino); |
85c0b2ab | 387 | |
1da177e4 LT |
388 | /* |
389 | * Insert records describing the new inode chunk into the btree. | |
390 | */ | |
561f7d17 | 391 | cur = xfs_inobt_init_cursor(args.mp, tp, agbp, agno); |
1da177e4 LT |
392 | for (thisino = newino; |
393 | thisino < newino + newlen; | |
394 | thisino += XFS_INODES_PER_CHUNK) { | |
21875505 CH |
395 | cur->bc_rec.i.ir_startino = thisino; |
396 | cur->bc_rec.i.ir_freecount = XFS_INODES_PER_CHUNK; | |
397 | cur->bc_rec.i.ir_free = XFS_INOBT_ALL_FREE; | |
398 | error = xfs_btree_lookup(cur, XFS_LOOKUP_EQ, &i); | |
399 | if (error) { | |
1da177e4 LT |
400 | xfs_btree_del_cursor(cur, XFS_BTREE_ERROR); |
401 | return error; | |
402 | } | |
403 | ASSERT(i == 0); | |
21875505 CH |
404 | error = xfs_btree_insert(cur, &i); |
405 | if (error) { | |
1da177e4 LT |
406 | xfs_btree_del_cursor(cur, XFS_BTREE_ERROR); |
407 | return error; | |
408 | } | |
409 | ASSERT(i == 1); | |
410 | } | |
411 | xfs_btree_del_cursor(cur, XFS_BTREE_NOERROR); | |
412 | /* | |
413 | * Log allocation group header fields | |
414 | */ | |
415 | xfs_ialloc_log_agi(tp, agbp, | |
416 | XFS_AGI_COUNT | XFS_AGI_FREECOUNT | XFS_AGI_NEWINO); | |
417 | /* | |
418 | * Modify/log superblock values for inode count and inode free count. | |
419 | */ | |
420 | xfs_trans_mod_sb(tp, XFS_TRANS_SB_ICOUNT, (long)newlen); | |
421 | xfs_trans_mod_sb(tp, XFS_TRANS_SB_IFREE, (long)newlen); | |
422 | *alloc = 1; | |
423 | return 0; | |
424 | } | |
425 | ||
b8f82a4a | 426 | STATIC xfs_agnumber_t |
1da177e4 LT |
427 | xfs_ialloc_next_ag( |
428 | xfs_mount_t *mp) | |
429 | { | |
430 | xfs_agnumber_t agno; | |
431 | ||
432 | spin_lock(&mp->m_agirotor_lock); | |
433 | agno = mp->m_agirotor; | |
434 | if (++mp->m_agirotor == mp->m_maxagi) | |
435 | mp->m_agirotor = 0; | |
436 | spin_unlock(&mp->m_agirotor_lock); | |
437 | ||
438 | return agno; | |
439 | } | |
440 | ||
441 | /* | |
442 | * Select an allocation group to look for a free inode in, based on the parent | |
443 | * inode and then mode. Return the allocation group buffer. | |
444 | */ | |
445 | STATIC xfs_buf_t * /* allocation group buffer */ | |
446 | xfs_ialloc_ag_select( | |
447 | xfs_trans_t *tp, /* transaction pointer */ | |
448 | xfs_ino_t parent, /* parent directory inode number */ | |
576b1d67 | 449 | umode_t mode, /* bits set to indicate file type */ |
1da177e4 LT |
450 | int okalloc) /* ok to allocate more space */ |
451 | { | |
452 | xfs_buf_t *agbp; /* allocation group header buffer */ | |
453 | xfs_agnumber_t agcount; /* number of ag's in the filesystem */ | |
454 | xfs_agnumber_t agno; /* current ag number */ | |
455 | int flags; /* alloc buffer locking flags */ | |
456 | xfs_extlen_t ineed; /* blocks needed for inode allocation */ | |
457 | xfs_extlen_t longest = 0; /* longest extent available */ | |
458 | xfs_mount_t *mp; /* mount point structure */ | |
459 | int needspace; /* file mode implies space allocated */ | |
460 | xfs_perag_t *pag; /* per allocation group data */ | |
461 | xfs_agnumber_t pagno; /* parent (starting) ag number */ | |
462 | ||
463 | /* | |
464 | * Files of these types need at least one block if length > 0 | |
465 | * (and they won't fit in the inode, but that's hard to figure out). | |
466 | */ | |
467 | needspace = S_ISDIR(mode) || S_ISREG(mode) || S_ISLNK(mode); | |
468 | mp = tp->t_mountp; | |
469 | agcount = mp->m_maxagi; | |
470 | if (S_ISDIR(mode)) | |
471 | pagno = xfs_ialloc_next_ag(mp); | |
472 | else { | |
473 | pagno = XFS_INO_TO_AGNO(mp, parent); | |
474 | if (pagno >= agcount) | |
475 | pagno = 0; | |
476 | } | |
477 | ASSERT(pagno < agcount); | |
478 | /* | |
479 | * Loop through allocation groups, looking for one with a little | |
480 | * free space in it. Note we don't look for free inodes, exactly. | |
481 | * Instead, we include whether there is a need to allocate inodes | |
482 | * to mean that blocks must be allocated for them, | |
483 | * if none are currently free. | |
484 | */ | |
485 | agno = pagno; | |
486 | flags = XFS_ALLOC_FLAG_TRYLOCK; | |
1da177e4 | 487 | for (;;) { |
44b56e0a | 488 | pag = xfs_perag_get(mp, agno); |
1da177e4 LT |
489 | if (!pag->pagi_init) { |
490 | if (xfs_ialloc_read_agi(mp, tp, agno, &agbp)) { | |
491 | agbp = NULL; | |
492 | goto nextag; | |
493 | } | |
494 | } else | |
495 | agbp = NULL; | |
496 | ||
497 | if (!pag->pagi_inodeok) { | |
498 | xfs_ialloc_next_ag(mp); | |
499 | goto unlock_nextag; | |
500 | } | |
501 | ||
502 | /* | |
503 | * Is there enough free space for the file plus a block | |
504 | * of inodes (if we need to allocate some)? | |
505 | */ | |
506 | ineed = pag->pagi_freecount ? 0 : XFS_IALLOC_BLOCKS(mp); | |
507 | if (ineed && !pag->pagf_init) { | |
508 | if (agbp == NULL && | |
509 | xfs_ialloc_read_agi(mp, tp, agno, &agbp)) { | |
510 | agbp = NULL; | |
511 | goto nextag; | |
512 | } | |
513 | (void)xfs_alloc_pagf_init(mp, tp, agno, flags); | |
514 | } | |
515 | if (!ineed || pag->pagf_init) { | |
516 | if (ineed && !(longest = pag->pagf_longest)) | |
517 | longest = pag->pagf_flcount > 0; | |
518 | if (!ineed || | |
519 | (pag->pagf_freeblks >= needspace + ineed && | |
520 | longest >= ineed && | |
521 | okalloc)) { | |
522 | if (agbp == NULL && | |
523 | xfs_ialloc_read_agi(mp, tp, agno, &agbp)) { | |
524 | agbp = NULL; | |
525 | goto nextag; | |
526 | } | |
44b56e0a | 527 | xfs_perag_put(pag); |
1da177e4 LT |
528 | return agbp; |
529 | } | |
530 | } | |
531 | unlock_nextag: | |
532 | if (agbp) | |
533 | xfs_trans_brelse(tp, agbp); | |
534 | nextag: | |
44b56e0a | 535 | xfs_perag_put(pag); |
1da177e4 LT |
536 | /* |
537 | * No point in iterating over the rest, if we're shutting | |
538 | * down. | |
539 | */ | |
1c1c6ebc | 540 | if (XFS_FORCED_SHUTDOWN(mp)) |
1121b219 | 541 | return NULL; |
1da177e4 LT |
542 | agno++; |
543 | if (agno >= agcount) | |
544 | agno = 0; | |
545 | if (agno == pagno) { | |
1c1c6ebc | 546 | if (flags == 0) |
1121b219 | 547 | return NULL; |
1da177e4 LT |
548 | flags = 0; |
549 | } | |
550 | } | |
551 | } | |
552 | ||
4254b0bb CH |
553 | /* |
554 | * Try to retrieve the next record to the left/right from the current one. | |
555 | */ | |
556 | STATIC int | |
557 | xfs_ialloc_next_rec( | |
558 | struct xfs_btree_cur *cur, | |
559 | xfs_inobt_rec_incore_t *rec, | |
560 | int *done, | |
561 | int left) | |
562 | { | |
563 | int error; | |
564 | int i; | |
565 | ||
566 | if (left) | |
567 | error = xfs_btree_decrement(cur, 0, &i); | |
568 | else | |
569 | error = xfs_btree_increment(cur, 0, &i); | |
570 | ||
571 | if (error) | |
572 | return error; | |
573 | *done = !i; | |
574 | if (i) { | |
575 | error = xfs_inobt_get_rec(cur, rec, &i); | |
576 | if (error) | |
577 | return error; | |
578 | XFS_WANT_CORRUPTED_RETURN(i == 1); | |
579 | } | |
580 | ||
581 | return 0; | |
582 | } | |
583 | ||
bd169565 DC |
584 | STATIC int |
585 | xfs_ialloc_get_rec( | |
586 | struct xfs_btree_cur *cur, | |
587 | xfs_agino_t agino, | |
588 | xfs_inobt_rec_incore_t *rec, | |
589 | int *done, | |
590 | int left) | |
591 | { | |
592 | int error; | |
593 | int i; | |
594 | ||
595 | error = xfs_inobt_lookup(cur, agino, XFS_LOOKUP_EQ, &i); | |
596 | if (error) | |
597 | return error; | |
598 | *done = !i; | |
599 | if (i) { | |
600 | error = xfs_inobt_get_rec(cur, rec, &i); | |
601 | if (error) | |
602 | return error; | |
603 | XFS_WANT_CORRUPTED_RETURN(i == 1); | |
604 | } | |
605 | ||
606 | return 0; | |
607 | } | |
0b48db80 | 608 | |
1da177e4 LT |
609 | /* |
610 | * Visible inode allocation functions. | |
611 | */ | |
612 | ||
613 | /* | |
614 | * Allocate an inode on disk. | |
615 | * Mode is used to tell whether the new inode will need space, and whether | |
616 | * it is a directory. | |
617 | * | |
618 | * The arguments IO_agbp and alloc_done are defined to work within | |
619 | * the constraint of one allocation per transaction. | |
620 | * xfs_dialloc() is designed to be called twice if it has to do an | |
621 | * allocation to make more free inodes. On the first call, | |
622 | * IO_agbp should be set to NULL. If an inode is available, | |
623 | * i.e., xfs_dialloc() did not need to do an allocation, an inode | |
624 | * number is returned. In this case, IO_agbp would be set to the | |
625 | * current ag_buf and alloc_done set to false. | |
626 | * If an allocation needed to be done, xfs_dialloc would return | |
627 | * the current ag_buf in IO_agbp and set alloc_done to true. | |
628 | * The caller should then commit the current transaction, allocate a new | |
629 | * transaction, and call xfs_dialloc() again, passing in the previous | |
630 | * value of IO_agbp. IO_agbp should be held across the transactions. | |
631 | * Since the agbp is locked across the two calls, the second call is | |
632 | * guaranteed to have a free inode available. | |
633 | * | |
634 | * Once we successfully pick an inode its number is returned and the | |
635 | * on-disk data structures are updated. The inode itself is not read | |
636 | * in, since doing so would break ordering constraints with xfs_reclaim. | |
637 | */ | |
638 | int | |
639 | xfs_dialloc( | |
640 | xfs_trans_t *tp, /* transaction pointer */ | |
641 | xfs_ino_t parent, /* parent inode (directory) */ | |
576b1d67 | 642 | umode_t mode, /* mode bits for new inode */ |
1da177e4 LT |
643 | int okalloc, /* ok to allocate more space */ |
644 | xfs_buf_t **IO_agbp, /* in/out ag header's buffer */ | |
645 | boolean_t *alloc_done, /* true if we needed to replenish | |
646 | inode freelist */ | |
647 | xfs_ino_t *inop) /* inode number allocated */ | |
648 | { | |
649 | xfs_agnumber_t agcount; /* number of allocation groups */ | |
650 | xfs_buf_t *agbp; /* allocation group header's buffer */ | |
651 | xfs_agnumber_t agno; /* allocation group number */ | |
652 | xfs_agi_t *agi; /* allocation group header structure */ | |
653 | xfs_btree_cur_t *cur; /* inode allocation btree cursor */ | |
654 | int error; /* error return value */ | |
655 | int i; /* result code */ | |
656 | int ialloced; /* inode allocation status */ | |
657 | int noroom = 0; /* no space for inode blk allocation */ | |
658 | xfs_ino_t ino; /* fs-relative inode to be returned */ | |
659 | /* REFERENCED */ | |
660 | int j; /* result code */ | |
661 | xfs_mount_t *mp; /* file system mount structure */ | |
662 | int offset; /* index of inode in chunk */ | |
4254b0bb CH |
663 | xfs_agino_t pagino; /* parent's AG relative inode # */ |
664 | xfs_agnumber_t pagno; /* parent's AG number */ | |
61a25848 | 665 | xfs_inobt_rec_incore_t rec; /* inode allocation record */ |
1da177e4 LT |
666 | xfs_agnumber_t tagno; /* testing allocation group number */ |
667 | xfs_btree_cur_t *tcur; /* temp cursor */ | |
61a25848 | 668 | xfs_inobt_rec_incore_t trec; /* temp inode allocation record */ |
44b56e0a | 669 | struct xfs_perag *pag; |
1da177e4 LT |
670 | |
671 | ||
672 | if (*IO_agbp == NULL) { | |
673 | /* | |
674 | * We do not have an agbp, so select an initial allocation | |
675 | * group for inode allocation. | |
676 | */ | |
677 | agbp = xfs_ialloc_ag_select(tp, parent, mode, okalloc); | |
678 | /* | |
679 | * Couldn't find an allocation group satisfying the | |
680 | * criteria, give up. | |
681 | */ | |
682 | if (!agbp) { | |
683 | *inop = NULLFSINO; | |
684 | return 0; | |
685 | } | |
686 | agi = XFS_BUF_TO_AGI(agbp); | |
69ef921b | 687 | ASSERT(agi->agi_magicnum == cpu_to_be32(XFS_AGI_MAGIC)); |
1da177e4 LT |
688 | } else { |
689 | /* | |
690 | * Continue where we left off before. In this case, we | |
691 | * know that the allocation group has free inodes. | |
692 | */ | |
693 | agbp = *IO_agbp; | |
694 | agi = XFS_BUF_TO_AGI(agbp); | |
69ef921b | 695 | ASSERT(agi->agi_magicnum == cpu_to_be32(XFS_AGI_MAGIC)); |
16259e7d | 696 | ASSERT(be32_to_cpu(agi->agi_freecount) > 0); |
1da177e4 LT |
697 | } |
698 | mp = tp->t_mountp; | |
699 | agcount = mp->m_sb.sb_agcount; | |
16259e7d | 700 | agno = be32_to_cpu(agi->agi_seqno); |
1da177e4 LT |
701 | tagno = agno; |
702 | pagno = XFS_INO_TO_AGNO(mp, parent); | |
703 | pagino = XFS_INO_TO_AGINO(mp, parent); | |
704 | ||
705 | /* | |
706 | * If we have already hit the ceiling of inode blocks then clear | |
707 | * okalloc so we scan all available agi structures for a free | |
708 | * inode. | |
709 | */ | |
710 | ||
711 | if (mp->m_maxicount && | |
712 | mp->m_sb.sb_icount + XFS_IALLOC_INODES(mp) > mp->m_maxicount) { | |
713 | noroom = 1; | |
714 | okalloc = 0; | |
715 | } | |
716 | ||
717 | /* | |
718 | * Loop until we find an allocation group that either has free inodes | |
719 | * or in which we can allocate some inodes. Iterate through the | |
720 | * allocation groups upward, wrapping at the end. | |
721 | */ | |
722 | *alloc_done = B_FALSE; | |
723 | while (!agi->agi_freecount) { | |
724 | /* | |
725 | * Don't do anything if we're not supposed to allocate | |
726 | * any blocks, just go on to the next ag. | |
727 | */ | |
728 | if (okalloc) { | |
729 | /* | |
730 | * Try to allocate some new inodes in the allocation | |
731 | * group. | |
732 | */ | |
733 | if ((error = xfs_ialloc_ag_alloc(tp, agbp, &ialloced))) { | |
734 | xfs_trans_brelse(tp, agbp); | |
735 | if (error == ENOSPC) { | |
736 | *inop = NULLFSINO; | |
737 | return 0; | |
738 | } else | |
739 | return error; | |
740 | } | |
741 | if (ialloced) { | |
742 | /* | |
743 | * We successfully allocated some inodes, return | |
744 | * the current context to the caller so that it | |
745 | * can commit the current transaction and call | |
746 | * us again where we left off. | |
747 | */ | |
16259e7d | 748 | ASSERT(be32_to_cpu(agi->agi_freecount) > 0); |
1da177e4 LT |
749 | *alloc_done = B_TRUE; |
750 | *IO_agbp = agbp; | |
751 | *inop = NULLFSINO; | |
752 | return 0; | |
753 | } | |
754 | } | |
755 | /* | |
756 | * If it failed, give up on this ag. | |
757 | */ | |
758 | xfs_trans_brelse(tp, agbp); | |
759 | /* | |
760 | * Go on to the next ag: get its ag header. | |
761 | */ | |
762 | nextag: | |
763 | if (++tagno == agcount) | |
764 | tagno = 0; | |
765 | if (tagno == agno) { | |
766 | *inop = NULLFSINO; | |
767 | return noroom ? ENOSPC : 0; | |
768 | } | |
44b56e0a DC |
769 | pag = xfs_perag_get(mp, tagno); |
770 | if (pag->pagi_inodeok == 0) { | |
771 | xfs_perag_put(pag); | |
1da177e4 LT |
772 | goto nextag; |
773 | } | |
774 | error = xfs_ialloc_read_agi(mp, tp, tagno, &agbp); | |
44b56e0a | 775 | xfs_perag_put(pag); |
1da177e4 LT |
776 | if (error) |
777 | goto nextag; | |
778 | agi = XFS_BUF_TO_AGI(agbp); | |
69ef921b | 779 | ASSERT(agi->agi_magicnum == cpu_to_be32(XFS_AGI_MAGIC)); |
1da177e4 LT |
780 | } |
781 | /* | |
782 | * Here with an allocation group that has a free inode. | |
783 | * Reset agno since we may have chosen a new ag in the | |
784 | * loop above. | |
785 | */ | |
786 | agno = tagno; | |
787 | *IO_agbp = NULL; | |
44b56e0a | 788 | pag = xfs_perag_get(mp, agno); |
bd169565 DC |
789 | |
790 | restart_pagno: | |
561f7d17 | 791 | cur = xfs_inobt_init_cursor(mp, tp, agbp, be32_to_cpu(agi->agi_seqno)); |
1da177e4 LT |
792 | /* |
793 | * If pagino is 0 (this is the root inode allocation) use newino. | |
794 | * This must work because we've just allocated some. | |
795 | */ | |
796 | if (!pagino) | |
16259e7d | 797 | pagino = be32_to_cpu(agi->agi_newino); |
1da177e4 | 798 | |
0b48db80 DC |
799 | error = xfs_check_agi_freecount(cur, agi); |
800 | if (error) | |
801 | goto error0; | |
1da177e4 | 802 | |
1da177e4 | 803 | /* |
4254b0bb | 804 | * If in the same AG as the parent, try to get near the parent. |
1da177e4 LT |
805 | */ |
806 | if (pagno == agno) { | |
4254b0bb CH |
807 | int doneleft; /* done, to the left */ |
808 | int doneright; /* done, to the right */ | |
bd169565 | 809 | int searchdistance = 10; |
4254b0bb | 810 | |
21875505 | 811 | error = xfs_inobt_lookup(cur, pagino, XFS_LOOKUP_LE, &i); |
4254b0bb | 812 | if (error) |
1da177e4 | 813 | goto error0; |
4254b0bb CH |
814 | XFS_WANT_CORRUPTED_GOTO(i == 1, error0); |
815 | ||
816 | error = xfs_inobt_get_rec(cur, &rec, &j); | |
817 | if (error) | |
818 | goto error0; | |
819 | XFS_WANT_CORRUPTED_GOTO(i == 1, error0); | |
820 | ||
821 | if (rec.ir_freecount > 0) { | |
1da177e4 LT |
822 | /* |
823 | * Found a free inode in the same chunk | |
4254b0bb | 824 | * as the parent, done. |
1da177e4 | 825 | */ |
4254b0bb | 826 | goto alloc_inode; |
1da177e4 | 827 | } |
4254b0bb CH |
828 | |
829 | ||
1da177e4 | 830 | /* |
4254b0bb | 831 | * In the same AG as parent, but parent's chunk is full. |
1da177e4 | 832 | */ |
1da177e4 | 833 | |
4254b0bb CH |
834 | /* duplicate the cursor, search left & right simultaneously */ |
835 | error = xfs_btree_dup_cursor(cur, &tcur); | |
836 | if (error) | |
837 | goto error0; | |
838 | ||
bd169565 DC |
839 | /* |
840 | * Skip to last blocks looked up if same parent inode. | |
841 | */ | |
842 | if (pagino != NULLAGINO && | |
843 | pag->pagl_pagino == pagino && | |
844 | pag->pagl_leftrec != NULLAGINO && | |
845 | pag->pagl_rightrec != NULLAGINO) { | |
846 | error = xfs_ialloc_get_rec(tcur, pag->pagl_leftrec, | |
847 | &trec, &doneleft, 1); | |
848 | if (error) | |
849 | goto error1; | |
4254b0bb | 850 | |
bd169565 DC |
851 | error = xfs_ialloc_get_rec(cur, pag->pagl_rightrec, |
852 | &rec, &doneright, 0); | |
853 | if (error) | |
854 | goto error1; | |
855 | } else { | |
856 | /* search left with tcur, back up 1 record */ | |
857 | error = xfs_ialloc_next_rec(tcur, &trec, &doneleft, 1); | |
858 | if (error) | |
859 | goto error1; | |
860 | ||
861 | /* search right with cur, go forward 1 record. */ | |
862 | error = xfs_ialloc_next_rec(cur, &rec, &doneright, 0); | |
863 | if (error) | |
864 | goto error1; | |
865 | } | |
4254b0bb CH |
866 | |
867 | /* | |
868 | * Loop until we find an inode chunk with a free inode. | |
869 | */ | |
870 | while (!doneleft || !doneright) { | |
871 | int useleft; /* using left inode chunk this time */ | |
872 | ||
bd169565 DC |
873 | if (!--searchdistance) { |
874 | /* | |
875 | * Not in range - save last search | |
876 | * location and allocate a new inode | |
877 | */ | |
3b826386 | 878 | xfs_btree_del_cursor(tcur, XFS_BTREE_NOERROR); |
bd169565 DC |
879 | pag->pagl_leftrec = trec.ir_startino; |
880 | pag->pagl_rightrec = rec.ir_startino; | |
881 | pag->pagl_pagino = pagino; | |
882 | goto newino; | |
883 | } | |
884 | ||
4254b0bb CH |
885 | /* figure out the closer block if both are valid. */ |
886 | if (!doneleft && !doneright) { | |
887 | useleft = pagino - | |
888 | (trec.ir_startino + XFS_INODES_PER_CHUNK - 1) < | |
889 | rec.ir_startino - pagino; | |
890 | } else { | |
891 | useleft = !doneleft; | |
1da177e4 | 892 | } |
4254b0bb CH |
893 | |
894 | /* free inodes to the left? */ | |
895 | if (useleft && trec.ir_freecount) { | |
896 | rec = trec; | |
897 | xfs_btree_del_cursor(cur, XFS_BTREE_NOERROR); | |
898 | cur = tcur; | |
bd169565 DC |
899 | |
900 | pag->pagl_leftrec = trec.ir_startino; | |
901 | pag->pagl_rightrec = rec.ir_startino; | |
902 | pag->pagl_pagino = pagino; | |
4254b0bb | 903 | goto alloc_inode; |
1da177e4 | 904 | } |
1da177e4 | 905 | |
4254b0bb CH |
906 | /* free inodes to the right? */ |
907 | if (!useleft && rec.ir_freecount) { | |
908 | xfs_btree_del_cursor(tcur, XFS_BTREE_NOERROR); | |
bd169565 DC |
909 | |
910 | pag->pagl_leftrec = trec.ir_startino; | |
911 | pag->pagl_rightrec = rec.ir_startino; | |
912 | pag->pagl_pagino = pagino; | |
4254b0bb | 913 | goto alloc_inode; |
1da177e4 | 914 | } |
4254b0bb CH |
915 | |
916 | /* get next record to check */ | |
917 | if (useleft) { | |
918 | error = xfs_ialloc_next_rec(tcur, &trec, | |
919 | &doneleft, 1); | |
920 | } else { | |
921 | error = xfs_ialloc_next_rec(cur, &rec, | |
922 | &doneright, 0); | |
923 | } | |
924 | if (error) | |
925 | goto error1; | |
1da177e4 | 926 | } |
bd169565 DC |
927 | |
928 | /* | |
929 | * We've reached the end of the btree. because | |
930 | * we are only searching a small chunk of the | |
931 | * btree each search, there is obviously free | |
932 | * inodes closer to the parent inode than we | |
933 | * are now. restart the search again. | |
934 | */ | |
935 | pag->pagl_pagino = NULLAGINO; | |
936 | pag->pagl_leftrec = NULLAGINO; | |
937 | pag->pagl_rightrec = NULLAGINO; | |
938 | xfs_btree_del_cursor(tcur, XFS_BTREE_NOERROR); | |
939 | xfs_btree_del_cursor(cur, XFS_BTREE_NOERROR); | |
940 | goto restart_pagno; | |
1da177e4 | 941 | } |
4254b0bb | 942 | |
1da177e4 | 943 | /* |
4254b0bb | 944 | * In a different AG from the parent. |
1da177e4 LT |
945 | * See if the most recently allocated block has any free. |
946 | */ | |
bd169565 | 947 | newino: |
69ef921b | 948 | if (agi->agi_newino != cpu_to_be32(NULLAGINO)) { |
21875505 CH |
949 | error = xfs_inobt_lookup(cur, be32_to_cpu(agi->agi_newino), |
950 | XFS_LOOKUP_EQ, &i); | |
4254b0bb | 951 | if (error) |
1da177e4 | 952 | goto error0; |
4254b0bb CH |
953 | |
954 | if (i == 1) { | |
955 | error = xfs_inobt_get_rec(cur, &rec, &j); | |
956 | if (error) | |
957 | goto error0; | |
958 | ||
959 | if (j == 1 && rec.ir_freecount > 0) { | |
960 | /* | |
961 | * The last chunk allocated in the group | |
962 | * still has a free inode. | |
963 | */ | |
964 | goto alloc_inode; | |
965 | } | |
1da177e4 | 966 | } |
bd169565 | 967 | } |
4254b0bb | 968 | |
bd169565 DC |
969 | /* |
970 | * None left in the last group, search the whole AG | |
971 | */ | |
972 | error = xfs_inobt_lookup(cur, 0, XFS_LOOKUP_GE, &i); | |
973 | if (error) | |
974 | goto error0; | |
975 | XFS_WANT_CORRUPTED_GOTO(i == 1, error0); | |
976 | ||
977 | for (;;) { | |
978 | error = xfs_inobt_get_rec(cur, &rec, &i); | |
979 | if (error) | |
980 | goto error0; | |
981 | XFS_WANT_CORRUPTED_GOTO(i == 1, error0); | |
982 | if (rec.ir_freecount > 0) | |
983 | break; | |
984 | error = xfs_btree_increment(cur, 0, &i); | |
4254b0bb CH |
985 | if (error) |
986 | goto error0; | |
987 | XFS_WANT_CORRUPTED_GOTO(i == 1, error0); | |
1da177e4 | 988 | } |
4254b0bb CH |
989 | |
990 | alloc_inode: | |
9d87c319 | 991 | offset = xfs_ialloc_find_free(&rec.ir_free); |
1da177e4 LT |
992 | ASSERT(offset >= 0); |
993 | ASSERT(offset < XFS_INODES_PER_CHUNK); | |
994 | ASSERT((XFS_AGINO_TO_OFFSET(mp, rec.ir_startino) % | |
995 | XFS_INODES_PER_CHUNK) == 0); | |
996 | ino = XFS_AGINO_TO_INO(mp, agno, rec.ir_startino + offset); | |
0d87e656 | 997 | rec.ir_free &= ~XFS_INOBT_MASK(offset); |
1da177e4 | 998 | rec.ir_freecount--; |
afabc24a CH |
999 | error = xfs_inobt_update(cur, &rec); |
1000 | if (error) | |
1da177e4 | 1001 | goto error0; |
413d57c9 | 1002 | be32_add_cpu(&agi->agi_freecount, -1); |
1da177e4 | 1003 | xfs_ialloc_log_agi(tp, agbp, XFS_AGI_FREECOUNT); |
44b56e0a | 1004 | pag->pagi_freecount--; |
1da177e4 | 1005 | |
0b48db80 DC |
1006 | error = xfs_check_agi_freecount(cur, agi); |
1007 | if (error) | |
1008 | goto error0; | |
1009 | ||
1da177e4 LT |
1010 | xfs_btree_del_cursor(cur, XFS_BTREE_NOERROR); |
1011 | xfs_trans_mod_sb(tp, XFS_TRANS_SB_IFREE, -1); | |
44b56e0a | 1012 | xfs_perag_put(pag); |
1da177e4 LT |
1013 | *inop = ino; |
1014 | return 0; | |
1015 | error1: | |
1016 | xfs_btree_del_cursor(tcur, XFS_BTREE_ERROR); | |
1017 | error0: | |
1018 | xfs_btree_del_cursor(cur, XFS_BTREE_ERROR); | |
44b56e0a | 1019 | xfs_perag_put(pag); |
1da177e4 LT |
1020 | return error; |
1021 | } | |
1022 | ||
1023 | /* | |
1024 | * Free disk inode. Carefully avoids touching the incore inode, all | |
1025 | * manipulations incore are the caller's responsibility. | |
1026 | * The on-disk inode is not changed by this operation, only the | |
1027 | * btree (free inode mask) is changed. | |
1028 | */ | |
1029 | int | |
1030 | xfs_difree( | |
1031 | xfs_trans_t *tp, /* transaction pointer */ | |
1032 | xfs_ino_t inode, /* inode to be freed */ | |
1033 | xfs_bmap_free_t *flist, /* extents to free */ | |
1034 | int *delete, /* set if inode cluster was deleted */ | |
1035 | xfs_ino_t *first_ino) /* first inode in deleted cluster */ | |
1036 | { | |
1037 | /* REFERENCED */ | |
1038 | xfs_agblock_t agbno; /* block number containing inode */ | |
1039 | xfs_buf_t *agbp; /* buffer containing allocation group header */ | |
1040 | xfs_agino_t agino; /* inode number relative to allocation group */ | |
1041 | xfs_agnumber_t agno; /* allocation group number */ | |
1042 | xfs_agi_t *agi; /* allocation group header */ | |
1043 | xfs_btree_cur_t *cur; /* inode btree cursor */ | |
1044 | int error; /* error return value */ | |
1045 | int i; /* result code */ | |
1046 | int ilen; /* inodes in an inode cluster */ | |
1047 | xfs_mount_t *mp; /* mount structure for filesystem */ | |
1048 | int off; /* offset of inode in inode chunk */ | |
61a25848 | 1049 | xfs_inobt_rec_incore_t rec; /* btree record */ |
44b56e0a | 1050 | struct xfs_perag *pag; |
1da177e4 LT |
1051 | |
1052 | mp = tp->t_mountp; | |
1053 | ||
1054 | /* | |
1055 | * Break up inode number into its components. | |
1056 | */ | |
1057 | agno = XFS_INO_TO_AGNO(mp, inode); | |
1058 | if (agno >= mp->m_sb.sb_agcount) { | |
0b932ccc DC |
1059 | xfs_warn(mp, "%s: agno >= mp->m_sb.sb_agcount (%d >= %d).", |
1060 | __func__, agno, mp->m_sb.sb_agcount); | |
1da177e4 LT |
1061 | ASSERT(0); |
1062 | return XFS_ERROR(EINVAL); | |
1063 | } | |
1064 | agino = XFS_INO_TO_AGINO(mp, inode); | |
1065 | if (inode != XFS_AGINO_TO_INO(mp, agno, agino)) { | |
0b932ccc DC |
1066 | xfs_warn(mp, "%s: inode != XFS_AGINO_TO_INO() (%llu != %llu).", |
1067 | __func__, (unsigned long long)inode, | |
1068 | (unsigned long long)XFS_AGINO_TO_INO(mp, agno, agino)); | |
1da177e4 LT |
1069 | ASSERT(0); |
1070 | return XFS_ERROR(EINVAL); | |
1071 | } | |
1072 | agbno = XFS_AGINO_TO_AGBNO(mp, agino); | |
1073 | if (agbno >= mp->m_sb.sb_agblocks) { | |
0b932ccc DC |
1074 | xfs_warn(mp, "%s: agbno >= mp->m_sb.sb_agblocks (%d >= %d).", |
1075 | __func__, agbno, mp->m_sb.sb_agblocks); | |
1da177e4 LT |
1076 | ASSERT(0); |
1077 | return XFS_ERROR(EINVAL); | |
1078 | } | |
1079 | /* | |
1080 | * Get the allocation group header. | |
1081 | */ | |
1da177e4 | 1082 | error = xfs_ialloc_read_agi(mp, tp, agno, &agbp); |
1da177e4 | 1083 | if (error) { |
0b932ccc DC |
1084 | xfs_warn(mp, "%s: xfs_ialloc_read_agi() returned error %d.", |
1085 | __func__, error); | |
1da177e4 LT |
1086 | return error; |
1087 | } | |
1088 | agi = XFS_BUF_TO_AGI(agbp); | |
69ef921b | 1089 | ASSERT(agi->agi_magicnum == cpu_to_be32(XFS_AGI_MAGIC)); |
16259e7d | 1090 | ASSERT(agbno < be32_to_cpu(agi->agi_length)); |
1da177e4 LT |
1091 | /* |
1092 | * Initialize the cursor. | |
1093 | */ | |
561f7d17 | 1094 | cur = xfs_inobt_init_cursor(mp, tp, agbp, agno); |
1da177e4 | 1095 | |
0b48db80 DC |
1096 | error = xfs_check_agi_freecount(cur, agi); |
1097 | if (error) | |
1098 | goto error0; | |
1099 | ||
1da177e4 LT |
1100 | /* |
1101 | * Look for the entry describing this inode. | |
1102 | */ | |
21875505 | 1103 | if ((error = xfs_inobt_lookup(cur, agino, XFS_LOOKUP_LE, &i))) { |
0b932ccc DC |
1104 | xfs_warn(mp, "%s: xfs_inobt_lookup() returned error %d.", |
1105 | __func__, error); | |
1da177e4 LT |
1106 | goto error0; |
1107 | } | |
1108 | XFS_WANT_CORRUPTED_GOTO(i == 1, error0); | |
2e287a73 CH |
1109 | error = xfs_inobt_get_rec(cur, &rec, &i); |
1110 | if (error) { | |
0b932ccc DC |
1111 | xfs_warn(mp, "%s: xfs_inobt_get_rec() returned error %d.", |
1112 | __func__, error); | |
1da177e4 LT |
1113 | goto error0; |
1114 | } | |
1115 | XFS_WANT_CORRUPTED_GOTO(i == 1, error0); | |
1116 | /* | |
1117 | * Get the offset in the inode chunk. | |
1118 | */ | |
1119 | off = agino - rec.ir_startino; | |
1120 | ASSERT(off >= 0 && off < XFS_INODES_PER_CHUNK); | |
0d87e656 | 1121 | ASSERT(!(rec.ir_free & XFS_INOBT_MASK(off))); |
1da177e4 LT |
1122 | /* |
1123 | * Mark the inode free & increment the count. | |
1124 | */ | |
0d87e656 | 1125 | rec.ir_free |= XFS_INOBT_MASK(off); |
1da177e4 LT |
1126 | rec.ir_freecount++; |
1127 | ||
1128 | /* | |
c41564b5 | 1129 | * When an inode cluster is free, it becomes eligible for removal |
1da177e4 | 1130 | */ |
1bd960ee | 1131 | if (!(mp->m_flags & XFS_MOUNT_IKEEP) && |
1da177e4 LT |
1132 | (rec.ir_freecount == XFS_IALLOC_INODES(mp))) { |
1133 | ||
1134 | *delete = 1; | |
1135 | *first_ino = XFS_AGINO_TO_INO(mp, agno, rec.ir_startino); | |
1136 | ||
1137 | /* | |
1138 | * Remove the inode cluster from the AGI B+Tree, adjust the | |
1139 | * AGI and Superblock inode counts, and mark the disk space | |
1140 | * to be freed when the transaction is committed. | |
1141 | */ | |
1142 | ilen = XFS_IALLOC_INODES(mp); | |
413d57c9 MS |
1143 | be32_add_cpu(&agi->agi_count, -ilen); |
1144 | be32_add_cpu(&agi->agi_freecount, -(ilen - 1)); | |
1da177e4 | 1145 | xfs_ialloc_log_agi(tp, agbp, XFS_AGI_COUNT | XFS_AGI_FREECOUNT); |
44b56e0a DC |
1146 | pag = xfs_perag_get(mp, agno); |
1147 | pag->pagi_freecount -= ilen - 1; | |
1148 | xfs_perag_put(pag); | |
1da177e4 LT |
1149 | xfs_trans_mod_sb(tp, XFS_TRANS_SB_ICOUNT, -ilen); |
1150 | xfs_trans_mod_sb(tp, XFS_TRANS_SB_IFREE, -(ilen - 1)); | |
1151 | ||
91cca5df | 1152 | if ((error = xfs_btree_delete(cur, &i))) { |
0b932ccc DC |
1153 | xfs_warn(mp, "%s: xfs_btree_delete returned error %d.", |
1154 | __func__, error); | |
1da177e4 LT |
1155 | goto error0; |
1156 | } | |
1157 | ||
1158 | xfs_bmap_add_free(XFS_AGB_TO_FSB(mp, | |
1159 | agno, XFS_INO_TO_AGBNO(mp,rec.ir_startino)), | |
1160 | XFS_IALLOC_BLOCKS(mp), flist, mp); | |
1161 | } else { | |
1162 | *delete = 0; | |
1163 | ||
afabc24a CH |
1164 | error = xfs_inobt_update(cur, &rec); |
1165 | if (error) { | |
0b932ccc DC |
1166 | xfs_warn(mp, "%s: xfs_inobt_update returned error %d.", |
1167 | __func__, error); | |
1da177e4 LT |
1168 | goto error0; |
1169 | } | |
afabc24a | 1170 | |
1da177e4 LT |
1171 | /* |
1172 | * Change the inode free counts and log the ag/sb changes. | |
1173 | */ | |
413d57c9 | 1174 | be32_add_cpu(&agi->agi_freecount, 1); |
1da177e4 | 1175 | xfs_ialloc_log_agi(tp, agbp, XFS_AGI_FREECOUNT); |
44b56e0a DC |
1176 | pag = xfs_perag_get(mp, agno); |
1177 | pag->pagi_freecount++; | |
1178 | xfs_perag_put(pag); | |
1da177e4 LT |
1179 | xfs_trans_mod_sb(tp, XFS_TRANS_SB_IFREE, 1); |
1180 | } | |
1181 | ||
0b48db80 DC |
1182 | error = xfs_check_agi_freecount(cur, agi); |
1183 | if (error) | |
1184 | goto error0; | |
1da177e4 | 1185 | |
1da177e4 LT |
1186 | xfs_btree_del_cursor(cur, XFS_BTREE_NOERROR); |
1187 | return 0; | |
1188 | ||
1189 | error0: | |
1190 | xfs_btree_del_cursor(cur, XFS_BTREE_ERROR); | |
1191 | return error; | |
1192 | } | |
1193 | ||
7124fe0a DC |
1194 | STATIC int |
1195 | xfs_imap_lookup( | |
1196 | struct xfs_mount *mp, | |
1197 | struct xfs_trans *tp, | |
1198 | xfs_agnumber_t agno, | |
1199 | xfs_agino_t agino, | |
1200 | xfs_agblock_t agbno, | |
1201 | xfs_agblock_t *chunk_agbno, | |
1202 | xfs_agblock_t *offset_agbno, | |
1203 | int flags) | |
1204 | { | |
1205 | struct xfs_inobt_rec_incore rec; | |
1206 | struct xfs_btree_cur *cur; | |
1207 | struct xfs_buf *agbp; | |
7124fe0a DC |
1208 | int error; |
1209 | int i; | |
1210 | ||
1211 | error = xfs_ialloc_read_agi(mp, tp, agno, &agbp); | |
1212 | if (error) { | |
53487786 DC |
1213 | xfs_alert(mp, |
1214 | "%s: xfs_ialloc_read_agi() returned error %d, agno %d", | |
1215 | __func__, error, agno); | |
7124fe0a DC |
1216 | return error; |
1217 | } | |
1218 | ||
1219 | /* | |
4536f2ad DC |
1220 | * Lookup the inode record for the given agino. If the record cannot be |
1221 | * found, then it's an invalid inode number and we should abort. Once | |
1222 | * we have a record, we need to ensure it contains the inode number | |
1223 | * we are looking up. | |
7124fe0a DC |
1224 | */ |
1225 | cur = xfs_inobt_init_cursor(mp, tp, agbp, agno); | |
4536f2ad | 1226 | error = xfs_inobt_lookup(cur, agino, XFS_LOOKUP_LE, &i); |
7124fe0a DC |
1227 | if (!error) { |
1228 | if (i) | |
1229 | error = xfs_inobt_get_rec(cur, &rec, &i); | |
1230 | if (!error && i == 0) | |
1231 | error = EINVAL; | |
1232 | } | |
1233 | ||
1234 | xfs_trans_brelse(tp, agbp); | |
1235 | xfs_btree_del_cursor(cur, XFS_BTREE_NOERROR); | |
1236 | if (error) | |
1237 | return error; | |
1238 | ||
4536f2ad DC |
1239 | /* check that the returned record contains the required inode */ |
1240 | if (rec.ir_startino > agino || | |
1241 | rec.ir_startino + XFS_IALLOC_INODES(mp) <= agino) | |
1242 | return EINVAL; | |
1243 | ||
7124fe0a | 1244 | /* for untrusted inodes check it is allocated first */ |
1920779e | 1245 | if ((flags & XFS_IGET_UNTRUSTED) && |
7124fe0a DC |
1246 | (rec.ir_free & XFS_INOBT_MASK(agino - rec.ir_startino))) |
1247 | return EINVAL; | |
1248 | ||
1249 | *chunk_agbno = XFS_AGINO_TO_AGBNO(mp, rec.ir_startino); | |
1250 | *offset_agbno = agbno - *chunk_agbno; | |
1251 | return 0; | |
1252 | } | |
1253 | ||
1da177e4 | 1254 | /* |
94e1b69d | 1255 | * Return the location of the inode in imap, for mapping it into a buffer. |
1da177e4 | 1256 | */ |
1da177e4 | 1257 | int |
94e1b69d CH |
1258 | xfs_imap( |
1259 | xfs_mount_t *mp, /* file system mount structure */ | |
1260 | xfs_trans_t *tp, /* transaction pointer */ | |
1da177e4 | 1261 | xfs_ino_t ino, /* inode to locate */ |
94e1b69d CH |
1262 | struct xfs_imap *imap, /* location map structure */ |
1263 | uint flags) /* flags for inode btree lookup */ | |
1da177e4 LT |
1264 | { |
1265 | xfs_agblock_t agbno; /* block number of inode in the alloc group */ | |
1da177e4 LT |
1266 | xfs_agino_t agino; /* inode number within alloc group */ |
1267 | xfs_agnumber_t agno; /* allocation group number */ | |
1268 | int blks_per_cluster; /* num blocks per inode cluster */ | |
1269 | xfs_agblock_t chunk_agbno; /* first block in inode chunk */ | |
1da177e4 | 1270 | xfs_agblock_t cluster_agbno; /* first block in inode cluster */ |
1da177e4 | 1271 | int error; /* error code */ |
1da177e4 LT |
1272 | int offset; /* index of inode in its buffer */ |
1273 | int offset_agbno; /* blks from chunk start to inode */ | |
1274 | ||
1275 | ASSERT(ino != NULLFSINO); | |
94e1b69d | 1276 | |
1da177e4 LT |
1277 | /* |
1278 | * Split up the inode number into its parts. | |
1279 | */ | |
1280 | agno = XFS_INO_TO_AGNO(mp, ino); | |
1281 | agino = XFS_INO_TO_AGINO(mp, ino); | |
1282 | agbno = XFS_AGINO_TO_AGBNO(mp, agino); | |
1283 | if (agno >= mp->m_sb.sb_agcount || agbno >= mp->m_sb.sb_agblocks || | |
1284 | ino != XFS_AGINO_TO_INO(mp, agno, agino)) { | |
1285 | #ifdef DEBUG | |
1920779e DC |
1286 | /* |
1287 | * Don't output diagnostic information for untrusted inodes | |
1288 | * as they can be invalid without implying corruption. | |
1289 | */ | |
1290 | if (flags & XFS_IGET_UNTRUSTED) | |
4d1a2ed3 | 1291 | return XFS_ERROR(EINVAL); |
1da177e4 | 1292 | if (agno >= mp->m_sb.sb_agcount) { |
53487786 DC |
1293 | xfs_alert(mp, |
1294 | "%s: agno (%d) >= mp->m_sb.sb_agcount (%d)", | |
1295 | __func__, agno, mp->m_sb.sb_agcount); | |
1da177e4 LT |
1296 | } |
1297 | if (agbno >= mp->m_sb.sb_agblocks) { | |
53487786 DC |
1298 | xfs_alert(mp, |
1299 | "%s: agbno (0x%llx) >= mp->m_sb.sb_agblocks (0x%lx)", | |
1300 | __func__, (unsigned long long)agbno, | |
1301 | (unsigned long)mp->m_sb.sb_agblocks); | |
1da177e4 LT |
1302 | } |
1303 | if (ino != XFS_AGINO_TO_INO(mp, agno, agino)) { | |
53487786 DC |
1304 | xfs_alert(mp, |
1305 | "%s: ino (0x%llx) != XFS_AGINO_TO_INO() (0x%llx)", | |
1306 | __func__, ino, | |
1307 | XFS_AGINO_TO_INO(mp, agno, agino)); | |
1da177e4 | 1308 | } |
745b1f47 | 1309 | xfs_stack_trace(); |
1da177e4 LT |
1310 | #endif /* DEBUG */ |
1311 | return XFS_ERROR(EINVAL); | |
1312 | } | |
94e1b69d | 1313 | |
7124fe0a DC |
1314 | blks_per_cluster = XFS_INODE_CLUSTER_SIZE(mp) >> mp->m_sb.sb_blocklog; |
1315 | ||
1316 | /* | |
1317 | * For bulkstat and handle lookups, we have an untrusted inode number | |
1318 | * that we have to verify is valid. We cannot do this just by reading | |
1319 | * the inode buffer as it may have been unlinked and removed leaving | |
1320 | * inodes in stale state on disk. Hence we have to do a btree lookup | |
1321 | * in all cases where an untrusted inode number is passed. | |
1322 | */ | |
1920779e | 1323 | if (flags & XFS_IGET_UNTRUSTED) { |
7124fe0a DC |
1324 | error = xfs_imap_lookup(mp, tp, agno, agino, agbno, |
1325 | &chunk_agbno, &offset_agbno, flags); | |
1326 | if (error) | |
1327 | return error; | |
1328 | goto out_map; | |
1329 | } | |
1330 | ||
94e1b69d CH |
1331 | /* |
1332 | * If the inode cluster size is the same as the blocksize or | |
1333 | * smaller we get to the buffer by simple arithmetics. | |
1334 | */ | |
1335 | if (XFS_INODE_CLUSTER_SIZE(mp) <= mp->m_sb.sb_blocksize) { | |
1da177e4 LT |
1336 | offset = XFS_INO_TO_OFFSET(mp, ino); |
1337 | ASSERT(offset < mp->m_sb.sb_inopblock); | |
94e1b69d CH |
1338 | |
1339 | imap->im_blkno = XFS_AGB_TO_DADDR(mp, agno, agbno); | |
1340 | imap->im_len = XFS_FSB_TO_BB(mp, 1); | |
1341 | imap->im_boffset = (ushort)(offset << mp->m_sb.sb_inodelog); | |
1da177e4 LT |
1342 | return 0; |
1343 | } | |
94e1b69d | 1344 | |
94e1b69d CH |
1345 | /* |
1346 | * If the inode chunks are aligned then use simple maths to | |
1347 | * find the location. Otherwise we have to do a btree | |
1348 | * lookup to find the location. | |
1349 | */ | |
1da177e4 LT |
1350 | if (mp->m_inoalign_mask) { |
1351 | offset_agbno = agbno & mp->m_inoalign_mask; | |
1352 | chunk_agbno = agbno - offset_agbno; | |
1353 | } else { | |
7124fe0a DC |
1354 | error = xfs_imap_lookup(mp, tp, agno, agino, agbno, |
1355 | &chunk_agbno, &offset_agbno, flags); | |
1da177e4 LT |
1356 | if (error) |
1357 | return error; | |
1da177e4 | 1358 | } |
94e1b69d | 1359 | |
7124fe0a | 1360 | out_map: |
1da177e4 LT |
1361 | ASSERT(agbno >= chunk_agbno); |
1362 | cluster_agbno = chunk_agbno + | |
1363 | ((offset_agbno / blks_per_cluster) * blks_per_cluster); | |
1364 | offset = ((agbno - cluster_agbno) * mp->m_sb.sb_inopblock) + | |
1365 | XFS_INO_TO_OFFSET(mp, ino); | |
94e1b69d CH |
1366 | |
1367 | imap->im_blkno = XFS_AGB_TO_DADDR(mp, agno, cluster_agbno); | |
1368 | imap->im_len = XFS_FSB_TO_BB(mp, blks_per_cluster); | |
1369 | imap->im_boffset = (ushort)(offset << mp->m_sb.sb_inodelog); | |
1370 | ||
1371 | /* | |
1372 | * If the inode number maps to a block outside the bounds | |
1373 | * of the file system then return NULL rather than calling | |
1374 | * read_buf and panicing when we get an error from the | |
1375 | * driver. | |
1376 | */ | |
1377 | if ((imap->im_blkno + imap->im_len) > | |
1378 | XFS_FSB_TO_BB(mp, mp->m_sb.sb_dblocks)) { | |
53487786 DC |
1379 | xfs_alert(mp, |
1380 | "%s: (im_blkno (0x%llx) + im_len (0x%llx)) > sb_dblocks (0x%llx)", | |
1381 | __func__, (unsigned long long) imap->im_blkno, | |
94e1b69d CH |
1382 | (unsigned long long) imap->im_len, |
1383 | XFS_FSB_TO_BB(mp, mp->m_sb.sb_dblocks)); | |
1384 | return XFS_ERROR(EINVAL); | |
1385 | } | |
1da177e4 | 1386 | return 0; |
1da177e4 LT |
1387 | } |
1388 | ||
1389 | /* | |
1390 | * Compute and fill in value of m_in_maxlevels. | |
1391 | */ | |
1392 | void | |
1393 | xfs_ialloc_compute_maxlevels( | |
1394 | xfs_mount_t *mp) /* file system mount structure */ | |
1395 | { | |
1396 | int level; | |
1397 | uint maxblocks; | |
1398 | uint maxleafents; | |
1399 | int minleafrecs; | |
1400 | int minnoderecs; | |
1401 | ||
1402 | maxleafents = (1LL << XFS_INO_AGINO_BITS(mp)) >> | |
1403 | XFS_INODES_PER_CHUNK_LOG; | |
1404 | minleafrecs = mp->m_alloc_mnr[0]; | |
1405 | minnoderecs = mp->m_alloc_mnr[1]; | |
1406 | maxblocks = (maxleafents + minleafrecs - 1) / minleafrecs; | |
1407 | for (level = 1; maxblocks > 1; level++) | |
1408 | maxblocks = (maxblocks + minnoderecs - 1) / minnoderecs; | |
1409 | mp->m_in_maxlevels = level; | |
1410 | } | |
1411 | ||
1412 | /* | |
1413 | * Log specified fields for the ag hdr (inode section) | |
1414 | */ | |
1415 | void | |
1416 | xfs_ialloc_log_agi( | |
1417 | xfs_trans_t *tp, /* transaction pointer */ | |
1418 | xfs_buf_t *bp, /* allocation group header buffer */ | |
1419 | int fields) /* bitmask of fields to log */ | |
1420 | { | |
1421 | int first; /* first byte number */ | |
1422 | int last; /* last byte number */ | |
1423 | static const short offsets[] = { /* field starting offsets */ | |
1424 | /* keep in sync with bit definitions */ | |
1425 | offsetof(xfs_agi_t, agi_magicnum), | |
1426 | offsetof(xfs_agi_t, agi_versionnum), | |
1427 | offsetof(xfs_agi_t, agi_seqno), | |
1428 | offsetof(xfs_agi_t, agi_length), | |
1429 | offsetof(xfs_agi_t, agi_count), | |
1430 | offsetof(xfs_agi_t, agi_root), | |
1431 | offsetof(xfs_agi_t, agi_level), | |
1432 | offsetof(xfs_agi_t, agi_freecount), | |
1433 | offsetof(xfs_agi_t, agi_newino), | |
1434 | offsetof(xfs_agi_t, agi_dirino), | |
1435 | offsetof(xfs_agi_t, agi_unlinked), | |
1436 | sizeof(xfs_agi_t) | |
1437 | }; | |
1438 | #ifdef DEBUG | |
1439 | xfs_agi_t *agi; /* allocation group header */ | |
1440 | ||
1441 | agi = XFS_BUF_TO_AGI(bp); | |
69ef921b | 1442 | ASSERT(agi->agi_magicnum == cpu_to_be32(XFS_AGI_MAGIC)); |
1da177e4 LT |
1443 | #endif |
1444 | /* | |
1445 | * Compute byte offsets for the first and last fields. | |
1446 | */ | |
1447 | xfs_btree_offsets(fields, offsets, XFS_AGI_NUM_BITS, &first, &last); | |
1448 | /* | |
1449 | * Log the allocation group inode header buffer. | |
1450 | */ | |
1451 | xfs_trans_log_buf(tp, bp, first, last); | |
1452 | } | |
1453 | ||
5e1be0fb CH |
1454 | #ifdef DEBUG |
1455 | STATIC void | |
1456 | xfs_check_agi_unlinked( | |
1457 | struct xfs_agi *agi) | |
1458 | { | |
1459 | int i; | |
1460 | ||
1461 | for (i = 0; i < XFS_AGI_UNLINKED_BUCKETS; i++) | |
1462 | ASSERT(agi->agi_unlinked[i]); | |
1463 | } | |
1464 | #else | |
1465 | #define xfs_check_agi_unlinked(agi) | |
1466 | #endif | |
1467 | ||
1da177e4 LT |
1468 | /* |
1469 | * Read in the allocation group header (inode allocation section) | |
1470 | */ | |
1471 | int | |
5e1be0fb CH |
1472 | xfs_read_agi( |
1473 | struct xfs_mount *mp, /* file system mount structure */ | |
1474 | struct xfs_trans *tp, /* transaction pointer */ | |
1475 | xfs_agnumber_t agno, /* allocation group number */ | |
1476 | struct xfs_buf **bpp) /* allocation group hdr buf */ | |
1da177e4 | 1477 | { |
5e1be0fb CH |
1478 | struct xfs_agi *agi; /* allocation group header */ |
1479 | int agi_ok; /* agi is consistent */ | |
1480 | int error; | |
1da177e4 LT |
1481 | |
1482 | ASSERT(agno != NULLAGNUMBER); | |
5e1be0fb CH |
1483 | |
1484 | error = xfs_trans_read_buf(mp, tp, mp->m_ddev_targp, | |
1da177e4 | 1485 | XFS_AG_DADDR(mp, agno, XFS_AGI_DADDR(mp)), |
5e1be0fb | 1486 | XFS_FSS_TO_BB(mp, 1), 0, bpp); |
1da177e4 LT |
1487 | if (error) |
1488 | return error; | |
5e1be0fb | 1489 | |
5a52c2a5 | 1490 | ASSERT(!xfs_buf_geterror(*bpp)); |
5e1be0fb | 1491 | agi = XFS_BUF_TO_AGI(*bpp); |
1da177e4 LT |
1492 | |
1493 | /* | |
1494 | * Validate the magic number of the agi block. | |
1495 | */ | |
69ef921b | 1496 | agi_ok = agi->agi_magicnum == cpu_to_be32(XFS_AGI_MAGIC) && |
5e1be0fb CH |
1497 | XFS_AGI_GOOD_VERSION(be32_to_cpu(agi->agi_versionnum)) && |
1498 | be32_to_cpu(agi->agi_seqno) == agno; | |
1da177e4 LT |
1499 | if (unlikely(XFS_TEST_ERROR(!agi_ok, mp, XFS_ERRTAG_IALLOC_READ_AGI, |
1500 | XFS_RANDOM_IALLOC_READ_AGI))) { | |
5e1be0fb | 1501 | XFS_CORRUPTION_ERROR("xfs_read_agi", XFS_ERRLEVEL_LOW, |
1da177e4 | 1502 | mp, agi); |
5e1be0fb | 1503 | xfs_trans_brelse(tp, *bpp); |
1da177e4 LT |
1504 | return XFS_ERROR(EFSCORRUPTED); |
1505 | } | |
5e1be0fb | 1506 | |
38f23232 | 1507 | xfs_buf_set_ref(*bpp, XFS_AGI_REF); |
5e1be0fb CH |
1508 | |
1509 | xfs_check_agi_unlinked(agi); | |
1510 | return 0; | |
1511 | } | |
1512 | ||
1513 | int | |
1514 | xfs_ialloc_read_agi( | |
1515 | struct xfs_mount *mp, /* file system mount structure */ | |
1516 | struct xfs_trans *tp, /* transaction pointer */ | |
1517 | xfs_agnumber_t agno, /* allocation group number */ | |
1518 | struct xfs_buf **bpp) /* allocation group hdr buf */ | |
1519 | { | |
1520 | struct xfs_agi *agi; /* allocation group header */ | |
1521 | struct xfs_perag *pag; /* per allocation group data */ | |
1522 | int error; | |
1523 | ||
1524 | error = xfs_read_agi(mp, tp, agno, bpp); | |
1525 | if (error) | |
1526 | return error; | |
1527 | ||
1528 | agi = XFS_BUF_TO_AGI(*bpp); | |
44b56e0a | 1529 | pag = xfs_perag_get(mp, agno); |
1da177e4 | 1530 | if (!pag->pagi_init) { |
16259e7d | 1531 | pag->pagi_freecount = be32_to_cpu(agi->agi_freecount); |
92821e2b | 1532 | pag->pagi_count = be32_to_cpu(agi->agi_count); |
1da177e4 | 1533 | pag->pagi_init = 1; |
1da177e4 | 1534 | } |
1da177e4 | 1535 | |
5e1be0fb CH |
1536 | /* |
1537 | * It's possible for these to be out of sync if | |
1538 | * we are in the middle of a forced shutdown. | |
1539 | */ | |
1540 | ASSERT(pag->pagi_freecount == be32_to_cpu(agi->agi_freecount) || | |
1541 | XFS_FORCED_SHUTDOWN(mp)); | |
44b56e0a | 1542 | xfs_perag_put(pag); |
1da177e4 LT |
1543 | return 0; |
1544 | } | |
92821e2b DC |
1545 | |
1546 | /* | |
1547 | * Read in the agi to initialise the per-ag data in the mount structure | |
1548 | */ | |
1549 | int | |
1550 | xfs_ialloc_pagi_init( | |
1551 | xfs_mount_t *mp, /* file system mount structure */ | |
1552 | xfs_trans_t *tp, /* transaction pointer */ | |
1553 | xfs_agnumber_t agno) /* allocation group number */ | |
1554 | { | |
1555 | xfs_buf_t *bp = NULL; | |
1556 | int error; | |
1557 | ||
1558 | error = xfs_ialloc_read_agi(mp, tp, agno, &bp); | |
1559 | if (error) | |
1560 | return error; | |
1561 | if (bp) | |
1562 | xfs_trans_brelse(tp, bp); | |
1563 | return 0; | |
1564 | } |