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 LT |
27 | #include "xfs_dir2.h" |
28 | #include "xfs_dmapi.h" | |
29 | #include "xfs_mount.h" | |
1da177e4 | 30 | #include "xfs_bmap_btree.h" |
a844f451 | 31 | #include "xfs_alloc_btree.h" |
1da177e4 | 32 | #include "xfs_ialloc_btree.h" |
1da177e4 | 33 | #include "xfs_dir2_sf.h" |
a844f451 | 34 | #include "xfs_attr_sf.h" |
1da177e4 LT |
35 | #include "xfs_dinode.h" |
36 | #include "xfs_inode.h" | |
a844f451 NS |
37 | #include "xfs_btree.h" |
38 | #include "xfs_ialloc.h" | |
1da177e4 | 39 | #include "xfs_alloc.h" |
1da177e4 LT |
40 | #include "xfs_rtalloc.h" |
41 | #include "xfs_error.h" | |
42 | #include "xfs_bmap.h" | |
43 | ||
1da177e4 LT |
44 | |
45 | /* | |
46 | * Allocation group level functions. | |
47 | */ | |
75de2a91 DC |
48 | static inline int |
49 | xfs_ialloc_cluster_alignment( | |
50 | xfs_alloc_arg_t *args) | |
51 | { | |
52 | if (xfs_sb_version_hasalign(&args->mp->m_sb) && | |
53 | args->mp->m_sb.sb_inoalignmt >= | |
54 | XFS_B_TO_FSBT(args->mp, XFS_INODE_CLUSTER_SIZE(args->mp))) | |
55 | return args->mp->m_sb.sb_inoalignmt; | |
56 | return 1; | |
57 | } | |
1da177e4 | 58 | |
fe033cc8 | 59 | /* |
21875505 | 60 | * Lookup a record by ino in the btree given by cur. |
fe033cc8 | 61 | */ |
81e25176 | 62 | int /* error */ |
21875505 | 63 | xfs_inobt_lookup( |
fe033cc8 CH |
64 | struct xfs_btree_cur *cur, /* btree cursor */ |
65 | xfs_agino_t ino, /* starting inode of chunk */ | |
21875505 | 66 | xfs_lookup_t dir, /* <=, >=, == */ |
fe033cc8 CH |
67 | int *stat) /* success/failure */ |
68 | { | |
69 | cur->bc_rec.i.ir_startino = ino; | |
21875505 CH |
70 | cur->bc_rec.i.ir_freecount = 0; |
71 | cur->bc_rec.i.ir_free = 0; | |
72 | return xfs_btree_lookup(cur, dir, stat); | |
fe033cc8 CH |
73 | } |
74 | ||
278d0ca1 | 75 | /* |
afabc24a | 76 | * Update the record referred to by cur to the value given. |
278d0ca1 CH |
77 | * This either works (return 0) or gets an EFSCORRUPTED error. |
78 | */ | |
79 | STATIC int /* error */ | |
80 | xfs_inobt_update( | |
81 | struct xfs_btree_cur *cur, /* btree cursor */ | |
afabc24a | 82 | xfs_inobt_rec_incore_t *irec) /* btree record */ |
278d0ca1 CH |
83 | { |
84 | union xfs_btree_rec rec; | |
85 | ||
afabc24a CH |
86 | rec.inobt.ir_startino = cpu_to_be32(irec->ir_startino); |
87 | rec.inobt.ir_freecount = cpu_to_be32(irec->ir_freecount); | |
88 | rec.inobt.ir_free = cpu_to_be64(irec->ir_free); | |
278d0ca1 CH |
89 | return xfs_btree_update(cur, &rec); |
90 | } | |
91 | ||
8cc938fe CH |
92 | /* |
93 | * Get the data from the pointed-to record. | |
94 | */ | |
95 | int /* error */ | |
96 | xfs_inobt_get_rec( | |
97 | struct xfs_btree_cur *cur, /* btree cursor */ | |
2e287a73 | 98 | xfs_inobt_rec_incore_t *irec, /* btree record */ |
8cc938fe CH |
99 | int *stat) /* output: success/failure */ |
100 | { | |
101 | union xfs_btree_rec *rec; | |
102 | int error; | |
103 | ||
104 | error = xfs_btree_get_rec(cur, &rec, stat); | |
105 | if (!error && *stat == 1) { | |
2e287a73 CH |
106 | irec->ir_startino = be32_to_cpu(rec->inobt.ir_startino); |
107 | irec->ir_freecount = be32_to_cpu(rec->inobt.ir_freecount); | |
108 | irec->ir_free = be64_to_cpu(rec->inobt.ir_free); | |
8cc938fe CH |
109 | } |
110 | return error; | |
111 | } | |
112 | ||
0b48db80 DC |
113 | /* |
114 | * Verify that the number of free inodes in the AGI is correct. | |
115 | */ | |
116 | #ifdef DEBUG | |
117 | STATIC int | |
118 | xfs_check_agi_freecount( | |
119 | struct xfs_btree_cur *cur, | |
120 | struct xfs_agi *agi) | |
121 | { | |
122 | if (cur->bc_nlevels == 1) { | |
123 | xfs_inobt_rec_incore_t rec; | |
124 | int freecount = 0; | |
125 | int error; | |
126 | int i; | |
127 | ||
21875505 | 128 | error = xfs_inobt_lookup(cur, 0, XFS_LOOKUP_GE, &i); |
0b48db80 DC |
129 | if (error) |
130 | return error; | |
131 | ||
132 | do { | |
133 | error = xfs_inobt_get_rec(cur, &rec, &i); | |
134 | if (error) | |
135 | return error; | |
136 | ||
137 | if (i) { | |
138 | freecount += rec.ir_freecount; | |
139 | error = xfs_btree_increment(cur, 0, &i); | |
140 | if (error) | |
141 | return error; | |
142 | } | |
143 | } while (i == 1); | |
144 | ||
145 | if (!XFS_FORCED_SHUTDOWN(cur->bc_mp)) | |
146 | ASSERT(freecount == be32_to_cpu(agi->agi_freecount)); | |
147 | } | |
148 | return 0; | |
149 | } | |
150 | #else | |
151 | #define xfs_check_agi_freecount(cur, agi) 0 | |
152 | #endif | |
153 | ||
85c0b2ab DC |
154 | /* |
155 | * Initialise a new set of inodes. | |
156 | */ | |
157 | STATIC void | |
158 | xfs_ialloc_inode_init( | |
159 | struct xfs_mount *mp, | |
160 | struct xfs_trans *tp, | |
161 | xfs_agnumber_t agno, | |
162 | xfs_agblock_t agbno, | |
163 | xfs_agblock_t length, | |
164 | unsigned int gen) | |
165 | { | |
166 | struct xfs_buf *fbuf; | |
167 | struct xfs_dinode *free; | |
168 | int blks_per_cluster, nbufs, ninodes; | |
169 | int version; | |
170 | int i, j; | |
171 | xfs_daddr_t d; | |
172 | ||
173 | /* | |
174 | * Loop over the new block(s), filling in the inodes. | |
175 | * For small block sizes, manipulate the inodes in buffers | |
176 | * which are multiples of the blocks size. | |
177 | */ | |
178 | if (mp->m_sb.sb_blocksize >= XFS_INODE_CLUSTER_SIZE(mp)) { | |
179 | blks_per_cluster = 1; | |
180 | nbufs = length; | |
181 | ninodes = mp->m_sb.sb_inopblock; | |
182 | } else { | |
183 | blks_per_cluster = XFS_INODE_CLUSTER_SIZE(mp) / | |
184 | mp->m_sb.sb_blocksize; | |
185 | nbufs = length / blks_per_cluster; | |
186 | ninodes = blks_per_cluster * mp->m_sb.sb_inopblock; | |
187 | } | |
188 | ||
189 | /* | |
190 | * Figure out what version number to use in the inodes we create. | |
191 | * If the superblock version has caught up to the one that supports | |
192 | * the new inode format, then use the new inode version. Otherwise | |
193 | * use the old version so that old kernels will continue to be | |
194 | * able to use the file system. | |
195 | */ | |
196 | if (xfs_sb_version_hasnlink(&mp->m_sb)) | |
197 | version = 2; | |
198 | else | |
199 | version = 1; | |
200 | ||
201 | for (j = 0; j < nbufs; j++) { | |
202 | /* | |
203 | * Get the block. | |
204 | */ | |
205 | d = XFS_AGB_TO_DADDR(mp, agno, agbno + (j * blks_per_cluster)); | |
206 | fbuf = xfs_trans_get_buf(tp, mp->m_ddev_targp, d, | |
207 | mp->m_bsize * blks_per_cluster, | |
0cadda1c | 208 | XBF_LOCK); |
85c0b2ab DC |
209 | ASSERT(fbuf); |
210 | ASSERT(!XFS_BUF_GETERROR(fbuf)); | |
211 | ||
212 | /* | |
213 | * Initialize all inodes in this buffer and then log them. | |
214 | * | |
215 | * XXX: It would be much better if we had just one transaction | |
216 | * to log a whole cluster of inodes instead of all the | |
217 | * individual transactions causing a lot of log traffic. | |
218 | */ | |
219 | xfs_biozero(fbuf, 0, ninodes << mp->m_sb.sb_inodelog); | |
220 | for (i = 0; i < ninodes; i++) { | |
221 | int ioffset = i << mp->m_sb.sb_inodelog; | |
222 | uint isize = sizeof(struct xfs_dinode); | |
223 | ||
224 | free = xfs_make_iptr(mp, fbuf, i); | |
225 | free->di_magic = cpu_to_be16(XFS_DINODE_MAGIC); | |
226 | free->di_version = version; | |
227 | free->di_gen = cpu_to_be32(gen); | |
228 | free->di_next_unlinked = cpu_to_be32(NULLAGINO); | |
229 | xfs_trans_log_buf(tp, fbuf, ioffset, ioffset + isize - 1); | |
230 | } | |
231 | xfs_trans_inode_alloc_buf(tp, fbuf); | |
232 | } | |
233 | } | |
234 | ||
1da177e4 LT |
235 | /* |
236 | * Allocate new inodes in the allocation group specified by agbp. | |
237 | * Return 0 for success, else error code. | |
238 | */ | |
239 | STATIC int /* error code or 0 */ | |
240 | xfs_ialloc_ag_alloc( | |
241 | xfs_trans_t *tp, /* transaction pointer */ | |
242 | xfs_buf_t *agbp, /* alloc group buffer */ | |
243 | int *alloc) | |
244 | { | |
245 | xfs_agi_t *agi; /* allocation group header */ | |
246 | xfs_alloc_arg_t args; /* allocation argument structure */ | |
1da177e4 | 247 | xfs_btree_cur_t *cur; /* inode btree cursor */ |
92821e2b | 248 | xfs_agnumber_t agno; |
1da177e4 | 249 | int error; |
85c0b2ab | 250 | int i; |
1da177e4 LT |
251 | xfs_agino_t newino; /* new first inode's number */ |
252 | xfs_agino_t newlen; /* new number of inodes */ | |
1da177e4 | 253 | xfs_agino_t thisino; /* current inode number, for loop */ |
3ccb8b5f | 254 | int isaligned = 0; /* inode allocation at stripe unit */ |
1da177e4 | 255 | /* boundary */ |
44b56e0a | 256 | struct xfs_perag *pag; |
1da177e4 LT |
257 | |
258 | args.tp = tp; | |
259 | args.mp = tp->t_mountp; | |
260 | ||
261 | /* | |
262 | * Locking will ensure that we don't have two callers in here | |
263 | * at one time. | |
264 | */ | |
265 | newlen = XFS_IALLOC_INODES(args.mp); | |
266 | if (args.mp->m_maxicount && | |
267 | args.mp->m_sb.sb_icount + newlen > args.mp->m_maxicount) | |
268 | return XFS_ERROR(ENOSPC); | |
269 | args.minlen = args.maxlen = XFS_IALLOC_BLOCKS(args.mp); | |
270 | /* | |
3ccb8b5f GO |
271 | * First try to allocate inodes contiguous with the last-allocated |
272 | * chunk of inodes. If the filesystem is striped, this will fill | |
273 | * an entire stripe unit with inodes. | |
274 | */ | |
1da177e4 | 275 | agi = XFS_BUF_TO_AGI(agbp); |
3ccb8b5f | 276 | newino = be32_to_cpu(agi->agi_newino); |
85c0b2ab | 277 | agno = be32_to_cpu(agi->agi_seqno); |
019ff2d5 NS |
278 | args.agbno = XFS_AGINO_TO_AGBNO(args.mp, newino) + |
279 | XFS_IALLOC_BLOCKS(args.mp); | |
280 | if (likely(newino != NULLAGINO && | |
281 | (args.agbno < be32_to_cpu(agi->agi_length)))) { | |
85c0b2ab | 282 | args.fsbno = XFS_AGB_TO_FSB(args.mp, agno, args.agbno); |
3ccb8b5f GO |
283 | args.type = XFS_ALLOCTYPE_THIS_BNO; |
284 | args.mod = args.total = args.wasdel = args.isfl = | |
285 | args.userdata = args.minalignslop = 0; | |
286 | args.prod = 1; | |
75de2a91 | 287 | |
3ccb8b5f | 288 | /* |
75de2a91 DC |
289 | * We need to take into account alignment here to ensure that |
290 | * we don't modify the free list if we fail to have an exact | |
291 | * block. If we don't have an exact match, and every oher | |
292 | * attempt allocation attempt fails, we'll end up cancelling | |
293 | * a dirty transaction and shutting down. | |
294 | * | |
295 | * For an exact allocation, alignment must be 1, | |
296 | * however we need to take cluster alignment into account when | |
297 | * fixing up the freelist. Use the minalignslop field to | |
298 | * indicate that extra blocks might be required for alignment, | |
299 | * but not to use them in the actual exact allocation. | |
3ccb8b5f | 300 | */ |
75de2a91 DC |
301 | args.alignment = 1; |
302 | args.minalignslop = xfs_ialloc_cluster_alignment(&args) - 1; | |
303 | ||
304 | /* Allow space for the inode btree to split. */ | |
0d87e656 | 305 | args.minleft = args.mp->m_in_maxlevels - 1; |
3ccb8b5f GO |
306 | if ((error = xfs_alloc_vextent(&args))) |
307 | return error; | |
308 | } else | |
309 | args.fsbno = NULLFSBLOCK; | |
1da177e4 | 310 | |
3ccb8b5f GO |
311 | if (unlikely(args.fsbno == NULLFSBLOCK)) { |
312 | /* | |
313 | * Set the alignment for the allocation. | |
314 | * If stripe alignment is turned on then align at stripe unit | |
315 | * boundary. | |
019ff2d5 NS |
316 | * If the cluster size is smaller than a filesystem block |
317 | * then we're doing I/O for inodes in filesystem block size | |
3ccb8b5f GO |
318 | * pieces, so don't need alignment anyway. |
319 | */ | |
320 | isaligned = 0; | |
321 | if (args.mp->m_sinoalign) { | |
322 | ASSERT(!(args.mp->m_flags & XFS_MOUNT_NOALIGN)); | |
323 | args.alignment = args.mp->m_dalign; | |
324 | isaligned = 1; | |
75de2a91 DC |
325 | } else |
326 | args.alignment = xfs_ialloc_cluster_alignment(&args); | |
3ccb8b5f GO |
327 | /* |
328 | * Need to figure out where to allocate the inode blocks. | |
329 | * Ideally they should be spaced out through the a.g. | |
330 | * For now, just allocate blocks up front. | |
331 | */ | |
332 | args.agbno = be32_to_cpu(agi->agi_root); | |
85c0b2ab | 333 | args.fsbno = XFS_AGB_TO_FSB(args.mp, agno, args.agbno); |
3ccb8b5f GO |
334 | /* |
335 | * Allocate a fixed-size extent of inodes. | |
336 | */ | |
337 | args.type = XFS_ALLOCTYPE_NEAR_BNO; | |
338 | args.mod = args.total = args.wasdel = args.isfl = | |
339 | args.userdata = args.minalignslop = 0; | |
340 | args.prod = 1; | |
341 | /* | |
342 | * Allow space for the inode btree to split. | |
343 | */ | |
0d87e656 | 344 | args.minleft = args.mp->m_in_maxlevels - 1; |
3ccb8b5f GO |
345 | if ((error = xfs_alloc_vextent(&args))) |
346 | return error; | |
347 | } | |
019ff2d5 | 348 | |
1da177e4 LT |
349 | /* |
350 | * If stripe alignment is turned on, then try again with cluster | |
351 | * alignment. | |
352 | */ | |
353 | if (isaligned && args.fsbno == NULLFSBLOCK) { | |
354 | args.type = XFS_ALLOCTYPE_NEAR_BNO; | |
16259e7d | 355 | args.agbno = be32_to_cpu(agi->agi_root); |
85c0b2ab | 356 | args.fsbno = XFS_AGB_TO_FSB(args.mp, agno, args.agbno); |
75de2a91 | 357 | args.alignment = xfs_ialloc_cluster_alignment(&args); |
1da177e4 LT |
358 | if ((error = xfs_alloc_vextent(&args))) |
359 | return error; | |
360 | } | |
361 | ||
362 | if (args.fsbno == NULLFSBLOCK) { | |
363 | *alloc = 0; | |
364 | return 0; | |
365 | } | |
366 | ASSERT(args.len == args.minlen); | |
1da177e4 | 367 | |
359346a9 | 368 | /* |
85c0b2ab DC |
369 | * Stamp and write the inode buffers. |
370 | * | |
359346a9 DC |
371 | * Seed the new inode cluster with a random generation number. This |
372 | * prevents short-term reuse of generation numbers if a chunk is | |
373 | * freed and then immediately reallocated. We use random numbers | |
374 | * rather than a linear progression to prevent the next generation | |
375 | * number from being easily guessable. | |
376 | */ | |
85c0b2ab DC |
377 | xfs_ialloc_inode_init(args.mp, tp, agno, args.agbno, args.len, |
378 | random32()); | |
d42f08f6 | 379 | |
85c0b2ab DC |
380 | /* |
381 | * Convert the results. | |
382 | */ | |
383 | newino = XFS_OFFBNO_TO_AGINO(args.mp, args.agbno, 0); | |
413d57c9 MS |
384 | be32_add_cpu(&agi->agi_count, newlen); |
385 | be32_add_cpu(&agi->agi_freecount, newlen); | |
44b56e0a DC |
386 | pag = xfs_perag_get(args.mp, agno); |
387 | pag->pagi_freecount += newlen; | |
388 | xfs_perag_put(pag); | |
16259e7d | 389 | agi->agi_newino = cpu_to_be32(newino); |
85c0b2ab | 390 | |
1da177e4 LT |
391 | /* |
392 | * Insert records describing the new inode chunk into the btree. | |
393 | */ | |
561f7d17 | 394 | cur = xfs_inobt_init_cursor(args.mp, tp, agbp, agno); |
1da177e4 LT |
395 | for (thisino = newino; |
396 | thisino < newino + newlen; | |
397 | thisino += XFS_INODES_PER_CHUNK) { | |
21875505 CH |
398 | cur->bc_rec.i.ir_startino = thisino; |
399 | cur->bc_rec.i.ir_freecount = XFS_INODES_PER_CHUNK; | |
400 | cur->bc_rec.i.ir_free = XFS_INOBT_ALL_FREE; | |
401 | error = xfs_btree_lookup(cur, XFS_LOOKUP_EQ, &i); | |
402 | if (error) { | |
1da177e4 LT |
403 | xfs_btree_del_cursor(cur, XFS_BTREE_ERROR); |
404 | return error; | |
405 | } | |
406 | ASSERT(i == 0); | |
21875505 CH |
407 | error = xfs_btree_insert(cur, &i); |
408 | if (error) { | |
1da177e4 LT |
409 | xfs_btree_del_cursor(cur, XFS_BTREE_ERROR); |
410 | return error; | |
411 | } | |
412 | ASSERT(i == 1); | |
413 | } | |
414 | xfs_btree_del_cursor(cur, XFS_BTREE_NOERROR); | |
415 | /* | |
416 | * Log allocation group header fields | |
417 | */ | |
418 | xfs_ialloc_log_agi(tp, agbp, | |
419 | XFS_AGI_COUNT | XFS_AGI_FREECOUNT | XFS_AGI_NEWINO); | |
420 | /* | |
421 | * Modify/log superblock values for inode count and inode free count. | |
422 | */ | |
423 | xfs_trans_mod_sb(tp, XFS_TRANS_SB_ICOUNT, (long)newlen); | |
424 | xfs_trans_mod_sb(tp, XFS_TRANS_SB_IFREE, (long)newlen); | |
425 | *alloc = 1; | |
426 | return 0; | |
427 | } | |
428 | ||
b8f82a4a | 429 | STATIC xfs_agnumber_t |
1da177e4 LT |
430 | xfs_ialloc_next_ag( |
431 | xfs_mount_t *mp) | |
432 | { | |
433 | xfs_agnumber_t agno; | |
434 | ||
435 | spin_lock(&mp->m_agirotor_lock); | |
436 | agno = mp->m_agirotor; | |
437 | if (++mp->m_agirotor == mp->m_maxagi) | |
438 | mp->m_agirotor = 0; | |
439 | spin_unlock(&mp->m_agirotor_lock); | |
440 | ||
441 | return agno; | |
442 | } | |
443 | ||
444 | /* | |
445 | * Select an allocation group to look for a free inode in, based on the parent | |
446 | * inode and then mode. Return the allocation group buffer. | |
447 | */ | |
448 | STATIC xfs_buf_t * /* allocation group buffer */ | |
449 | xfs_ialloc_ag_select( | |
450 | xfs_trans_t *tp, /* transaction pointer */ | |
451 | xfs_ino_t parent, /* parent directory inode number */ | |
452 | mode_t mode, /* bits set to indicate file type */ | |
453 | int okalloc) /* ok to allocate more space */ | |
454 | { | |
455 | xfs_buf_t *agbp; /* allocation group header buffer */ | |
456 | xfs_agnumber_t agcount; /* number of ag's in the filesystem */ | |
457 | xfs_agnumber_t agno; /* current ag number */ | |
458 | int flags; /* alloc buffer locking flags */ | |
459 | xfs_extlen_t ineed; /* blocks needed for inode allocation */ | |
460 | xfs_extlen_t longest = 0; /* longest extent available */ | |
461 | xfs_mount_t *mp; /* mount point structure */ | |
462 | int needspace; /* file mode implies space allocated */ | |
463 | xfs_perag_t *pag; /* per allocation group data */ | |
464 | xfs_agnumber_t pagno; /* parent (starting) ag number */ | |
465 | ||
466 | /* | |
467 | * Files of these types need at least one block if length > 0 | |
468 | * (and they won't fit in the inode, but that's hard to figure out). | |
469 | */ | |
470 | needspace = S_ISDIR(mode) || S_ISREG(mode) || S_ISLNK(mode); | |
471 | mp = tp->t_mountp; | |
472 | agcount = mp->m_maxagi; | |
473 | if (S_ISDIR(mode)) | |
474 | pagno = xfs_ialloc_next_ag(mp); | |
475 | else { | |
476 | pagno = XFS_INO_TO_AGNO(mp, parent); | |
477 | if (pagno >= agcount) | |
478 | pagno = 0; | |
479 | } | |
480 | ASSERT(pagno < agcount); | |
481 | /* | |
482 | * Loop through allocation groups, looking for one with a little | |
483 | * free space in it. Note we don't look for free inodes, exactly. | |
484 | * Instead, we include whether there is a need to allocate inodes | |
485 | * to mean that blocks must be allocated for them, | |
486 | * if none are currently free. | |
487 | */ | |
488 | agno = pagno; | |
489 | flags = XFS_ALLOC_FLAG_TRYLOCK; | |
1da177e4 | 490 | for (;;) { |
44b56e0a | 491 | pag = xfs_perag_get(mp, agno); |
1da177e4 LT |
492 | if (!pag->pagi_init) { |
493 | if (xfs_ialloc_read_agi(mp, tp, agno, &agbp)) { | |
494 | agbp = NULL; | |
495 | goto nextag; | |
496 | } | |
497 | } else | |
498 | agbp = NULL; | |
499 | ||
500 | if (!pag->pagi_inodeok) { | |
501 | xfs_ialloc_next_ag(mp); | |
502 | goto unlock_nextag; | |
503 | } | |
504 | ||
505 | /* | |
506 | * Is there enough free space for the file plus a block | |
507 | * of inodes (if we need to allocate some)? | |
508 | */ | |
509 | ineed = pag->pagi_freecount ? 0 : XFS_IALLOC_BLOCKS(mp); | |
510 | if (ineed && !pag->pagf_init) { | |
511 | if (agbp == NULL && | |
512 | xfs_ialloc_read_agi(mp, tp, agno, &agbp)) { | |
513 | agbp = NULL; | |
514 | goto nextag; | |
515 | } | |
516 | (void)xfs_alloc_pagf_init(mp, tp, agno, flags); | |
517 | } | |
518 | if (!ineed || pag->pagf_init) { | |
519 | if (ineed && !(longest = pag->pagf_longest)) | |
520 | longest = pag->pagf_flcount > 0; | |
521 | if (!ineed || | |
522 | (pag->pagf_freeblks >= needspace + ineed && | |
523 | longest >= ineed && | |
524 | okalloc)) { | |
525 | if (agbp == NULL && | |
526 | xfs_ialloc_read_agi(mp, tp, agno, &agbp)) { | |
527 | agbp = NULL; | |
528 | goto nextag; | |
529 | } | |
44b56e0a | 530 | xfs_perag_put(pag); |
1da177e4 LT |
531 | return agbp; |
532 | } | |
533 | } | |
534 | unlock_nextag: | |
535 | if (agbp) | |
536 | xfs_trans_brelse(tp, agbp); | |
537 | nextag: | |
44b56e0a | 538 | xfs_perag_put(pag); |
1da177e4 LT |
539 | /* |
540 | * No point in iterating over the rest, if we're shutting | |
541 | * down. | |
542 | */ | |
1c1c6ebc | 543 | if (XFS_FORCED_SHUTDOWN(mp)) |
1121b219 | 544 | return NULL; |
1da177e4 LT |
545 | agno++; |
546 | if (agno >= agcount) | |
547 | agno = 0; | |
548 | if (agno == pagno) { | |
1c1c6ebc | 549 | if (flags == 0) |
1121b219 | 550 | return NULL; |
1da177e4 LT |
551 | flags = 0; |
552 | } | |
553 | } | |
554 | } | |
555 | ||
4254b0bb CH |
556 | /* |
557 | * Try to retrieve the next record to the left/right from the current one. | |
558 | */ | |
559 | STATIC int | |
560 | xfs_ialloc_next_rec( | |
561 | struct xfs_btree_cur *cur, | |
562 | xfs_inobt_rec_incore_t *rec, | |
563 | int *done, | |
564 | int left) | |
565 | { | |
566 | int error; | |
567 | int i; | |
568 | ||
569 | if (left) | |
570 | error = xfs_btree_decrement(cur, 0, &i); | |
571 | else | |
572 | error = xfs_btree_increment(cur, 0, &i); | |
573 | ||
574 | if (error) | |
575 | return error; | |
576 | *done = !i; | |
577 | if (i) { | |
578 | error = xfs_inobt_get_rec(cur, rec, &i); | |
579 | if (error) | |
580 | return error; | |
581 | XFS_WANT_CORRUPTED_RETURN(i == 1); | |
582 | } | |
583 | ||
584 | return 0; | |
585 | } | |
586 | ||
bd169565 DC |
587 | STATIC int |
588 | xfs_ialloc_get_rec( | |
589 | struct xfs_btree_cur *cur, | |
590 | xfs_agino_t agino, | |
591 | xfs_inobt_rec_incore_t *rec, | |
592 | int *done, | |
593 | int left) | |
594 | { | |
595 | int error; | |
596 | int i; | |
597 | ||
598 | error = xfs_inobt_lookup(cur, agino, XFS_LOOKUP_EQ, &i); | |
599 | if (error) | |
600 | return error; | |
601 | *done = !i; | |
602 | if (i) { | |
603 | error = xfs_inobt_get_rec(cur, rec, &i); | |
604 | if (error) | |
605 | return error; | |
606 | XFS_WANT_CORRUPTED_RETURN(i == 1); | |
607 | } | |
608 | ||
609 | return 0; | |
610 | } | |
0b48db80 | 611 | |
1da177e4 LT |
612 | /* |
613 | * Visible inode allocation functions. | |
614 | */ | |
615 | ||
616 | /* | |
617 | * Allocate an inode on disk. | |
618 | * Mode is used to tell whether the new inode will need space, and whether | |
619 | * it is a directory. | |
620 | * | |
621 | * The arguments IO_agbp and alloc_done are defined to work within | |
622 | * the constraint of one allocation per transaction. | |
623 | * xfs_dialloc() is designed to be called twice if it has to do an | |
624 | * allocation to make more free inodes. On the first call, | |
625 | * IO_agbp should be set to NULL. If an inode is available, | |
626 | * i.e., xfs_dialloc() did not need to do an allocation, an inode | |
627 | * number is returned. In this case, IO_agbp would be set to the | |
628 | * current ag_buf and alloc_done set to false. | |
629 | * If an allocation needed to be done, xfs_dialloc would return | |
630 | * the current ag_buf in IO_agbp and set alloc_done to true. | |
631 | * The caller should then commit the current transaction, allocate a new | |
632 | * transaction, and call xfs_dialloc() again, passing in the previous | |
633 | * value of IO_agbp. IO_agbp should be held across the transactions. | |
634 | * Since the agbp is locked across the two calls, the second call is | |
635 | * guaranteed to have a free inode available. | |
636 | * | |
637 | * Once we successfully pick an inode its number is returned and the | |
638 | * on-disk data structures are updated. The inode itself is not read | |
639 | * in, since doing so would break ordering constraints with xfs_reclaim. | |
640 | */ | |
641 | int | |
642 | xfs_dialloc( | |
643 | xfs_trans_t *tp, /* transaction pointer */ | |
644 | xfs_ino_t parent, /* parent inode (directory) */ | |
645 | mode_t mode, /* mode bits for new inode */ | |
646 | int okalloc, /* ok to allocate more space */ | |
647 | xfs_buf_t **IO_agbp, /* in/out ag header's buffer */ | |
648 | boolean_t *alloc_done, /* true if we needed to replenish | |
649 | inode freelist */ | |
650 | xfs_ino_t *inop) /* inode number allocated */ | |
651 | { | |
652 | xfs_agnumber_t agcount; /* number of allocation groups */ | |
653 | xfs_buf_t *agbp; /* allocation group header's buffer */ | |
654 | xfs_agnumber_t agno; /* allocation group number */ | |
655 | xfs_agi_t *agi; /* allocation group header structure */ | |
656 | xfs_btree_cur_t *cur; /* inode allocation btree cursor */ | |
657 | int error; /* error return value */ | |
658 | int i; /* result code */ | |
659 | int ialloced; /* inode allocation status */ | |
660 | int noroom = 0; /* no space for inode blk allocation */ | |
661 | xfs_ino_t ino; /* fs-relative inode to be returned */ | |
662 | /* REFERENCED */ | |
663 | int j; /* result code */ | |
664 | xfs_mount_t *mp; /* file system mount structure */ | |
665 | int offset; /* index of inode in chunk */ | |
4254b0bb CH |
666 | xfs_agino_t pagino; /* parent's AG relative inode # */ |
667 | xfs_agnumber_t pagno; /* parent's AG number */ | |
61a25848 | 668 | xfs_inobt_rec_incore_t rec; /* inode allocation record */ |
1da177e4 LT |
669 | xfs_agnumber_t tagno; /* testing allocation group number */ |
670 | xfs_btree_cur_t *tcur; /* temp cursor */ | |
61a25848 | 671 | xfs_inobt_rec_incore_t trec; /* temp inode allocation record */ |
44b56e0a | 672 | struct xfs_perag *pag; |
1da177e4 LT |
673 | |
674 | ||
675 | if (*IO_agbp == NULL) { | |
676 | /* | |
677 | * We do not have an agbp, so select an initial allocation | |
678 | * group for inode allocation. | |
679 | */ | |
680 | agbp = xfs_ialloc_ag_select(tp, parent, mode, okalloc); | |
681 | /* | |
682 | * Couldn't find an allocation group satisfying the | |
683 | * criteria, give up. | |
684 | */ | |
685 | if (!agbp) { | |
686 | *inop = NULLFSINO; | |
687 | return 0; | |
688 | } | |
689 | agi = XFS_BUF_TO_AGI(agbp); | |
16259e7d | 690 | ASSERT(be32_to_cpu(agi->agi_magicnum) == XFS_AGI_MAGIC); |
1da177e4 LT |
691 | } else { |
692 | /* | |
693 | * Continue where we left off before. In this case, we | |
694 | * know that the allocation group has free inodes. | |
695 | */ | |
696 | agbp = *IO_agbp; | |
697 | agi = XFS_BUF_TO_AGI(agbp); | |
16259e7d CH |
698 | ASSERT(be32_to_cpu(agi->agi_magicnum) == XFS_AGI_MAGIC); |
699 | ASSERT(be32_to_cpu(agi->agi_freecount) > 0); | |
1da177e4 LT |
700 | } |
701 | mp = tp->t_mountp; | |
702 | agcount = mp->m_sb.sb_agcount; | |
16259e7d | 703 | agno = be32_to_cpu(agi->agi_seqno); |
1da177e4 LT |
704 | tagno = agno; |
705 | pagno = XFS_INO_TO_AGNO(mp, parent); | |
706 | pagino = XFS_INO_TO_AGINO(mp, parent); | |
707 | ||
708 | /* | |
709 | * If we have already hit the ceiling of inode blocks then clear | |
710 | * okalloc so we scan all available agi structures for a free | |
711 | * inode. | |
712 | */ | |
713 | ||
714 | if (mp->m_maxicount && | |
715 | mp->m_sb.sb_icount + XFS_IALLOC_INODES(mp) > mp->m_maxicount) { | |
716 | noroom = 1; | |
717 | okalloc = 0; | |
718 | } | |
719 | ||
720 | /* | |
721 | * Loop until we find an allocation group that either has free inodes | |
722 | * or in which we can allocate some inodes. Iterate through the | |
723 | * allocation groups upward, wrapping at the end. | |
724 | */ | |
725 | *alloc_done = B_FALSE; | |
726 | while (!agi->agi_freecount) { | |
727 | /* | |
728 | * Don't do anything if we're not supposed to allocate | |
729 | * any blocks, just go on to the next ag. | |
730 | */ | |
731 | if (okalloc) { | |
732 | /* | |
733 | * Try to allocate some new inodes in the allocation | |
734 | * group. | |
735 | */ | |
736 | if ((error = xfs_ialloc_ag_alloc(tp, agbp, &ialloced))) { | |
737 | xfs_trans_brelse(tp, agbp); | |
738 | if (error == ENOSPC) { | |
739 | *inop = NULLFSINO; | |
740 | return 0; | |
741 | } else | |
742 | return error; | |
743 | } | |
744 | if (ialloced) { | |
745 | /* | |
746 | * We successfully allocated some inodes, return | |
747 | * the current context to the caller so that it | |
748 | * can commit the current transaction and call | |
749 | * us again where we left off. | |
750 | */ | |
16259e7d | 751 | ASSERT(be32_to_cpu(agi->agi_freecount) > 0); |
1da177e4 LT |
752 | *alloc_done = B_TRUE; |
753 | *IO_agbp = agbp; | |
754 | *inop = NULLFSINO; | |
755 | return 0; | |
756 | } | |
757 | } | |
758 | /* | |
759 | * If it failed, give up on this ag. | |
760 | */ | |
761 | xfs_trans_brelse(tp, agbp); | |
762 | /* | |
763 | * Go on to the next ag: get its ag header. | |
764 | */ | |
765 | nextag: | |
766 | if (++tagno == agcount) | |
767 | tagno = 0; | |
768 | if (tagno == agno) { | |
769 | *inop = NULLFSINO; | |
770 | return noroom ? ENOSPC : 0; | |
771 | } | |
44b56e0a DC |
772 | pag = xfs_perag_get(mp, tagno); |
773 | if (pag->pagi_inodeok == 0) { | |
774 | xfs_perag_put(pag); | |
1da177e4 LT |
775 | goto nextag; |
776 | } | |
777 | error = xfs_ialloc_read_agi(mp, tp, tagno, &agbp); | |
44b56e0a | 778 | xfs_perag_put(pag); |
1da177e4 LT |
779 | if (error) |
780 | goto nextag; | |
781 | agi = XFS_BUF_TO_AGI(agbp); | |
16259e7d | 782 | ASSERT(be32_to_cpu(agi->agi_magicnum) == XFS_AGI_MAGIC); |
1da177e4 LT |
783 | } |
784 | /* | |
785 | * Here with an allocation group that has a free inode. | |
786 | * Reset agno since we may have chosen a new ag in the | |
787 | * loop above. | |
788 | */ | |
789 | agno = tagno; | |
790 | *IO_agbp = NULL; | |
44b56e0a | 791 | pag = xfs_perag_get(mp, agno); |
bd169565 DC |
792 | |
793 | restart_pagno: | |
561f7d17 | 794 | cur = xfs_inobt_init_cursor(mp, tp, agbp, be32_to_cpu(agi->agi_seqno)); |
1da177e4 LT |
795 | /* |
796 | * If pagino is 0 (this is the root inode allocation) use newino. | |
797 | * This must work because we've just allocated some. | |
798 | */ | |
799 | if (!pagino) | |
16259e7d | 800 | pagino = be32_to_cpu(agi->agi_newino); |
1da177e4 | 801 | |
0b48db80 DC |
802 | error = xfs_check_agi_freecount(cur, agi); |
803 | if (error) | |
804 | goto error0; | |
1da177e4 | 805 | |
1da177e4 | 806 | /* |
4254b0bb | 807 | * If in the same AG as the parent, try to get near the parent. |
1da177e4 LT |
808 | */ |
809 | if (pagno == agno) { | |
4254b0bb CH |
810 | int doneleft; /* done, to the left */ |
811 | int doneright; /* done, to the right */ | |
bd169565 | 812 | int searchdistance = 10; |
4254b0bb | 813 | |
21875505 | 814 | error = xfs_inobt_lookup(cur, pagino, XFS_LOOKUP_LE, &i); |
4254b0bb | 815 | if (error) |
1da177e4 | 816 | goto error0; |
4254b0bb CH |
817 | XFS_WANT_CORRUPTED_GOTO(i == 1, error0); |
818 | ||
819 | error = xfs_inobt_get_rec(cur, &rec, &j); | |
820 | if (error) | |
821 | goto error0; | |
822 | XFS_WANT_CORRUPTED_GOTO(i == 1, error0); | |
823 | ||
824 | if (rec.ir_freecount > 0) { | |
1da177e4 LT |
825 | /* |
826 | * Found a free inode in the same chunk | |
4254b0bb | 827 | * as the parent, done. |
1da177e4 | 828 | */ |
4254b0bb | 829 | goto alloc_inode; |
1da177e4 | 830 | } |
4254b0bb CH |
831 | |
832 | ||
1da177e4 | 833 | /* |
4254b0bb | 834 | * In the same AG as parent, but parent's chunk is full. |
1da177e4 | 835 | */ |
1da177e4 | 836 | |
4254b0bb CH |
837 | /* duplicate the cursor, search left & right simultaneously */ |
838 | error = xfs_btree_dup_cursor(cur, &tcur); | |
839 | if (error) | |
840 | goto error0; | |
841 | ||
bd169565 DC |
842 | /* |
843 | * Skip to last blocks looked up if same parent inode. | |
844 | */ | |
845 | if (pagino != NULLAGINO && | |
846 | pag->pagl_pagino == pagino && | |
847 | pag->pagl_leftrec != NULLAGINO && | |
848 | pag->pagl_rightrec != NULLAGINO) { | |
849 | error = xfs_ialloc_get_rec(tcur, pag->pagl_leftrec, | |
850 | &trec, &doneleft, 1); | |
851 | if (error) | |
852 | goto error1; | |
4254b0bb | 853 | |
bd169565 DC |
854 | error = xfs_ialloc_get_rec(cur, pag->pagl_rightrec, |
855 | &rec, &doneright, 0); | |
856 | if (error) | |
857 | goto error1; | |
858 | } else { | |
859 | /* search left with tcur, back up 1 record */ | |
860 | error = xfs_ialloc_next_rec(tcur, &trec, &doneleft, 1); | |
861 | if (error) | |
862 | goto error1; | |
863 | ||
864 | /* search right with cur, go forward 1 record. */ | |
865 | error = xfs_ialloc_next_rec(cur, &rec, &doneright, 0); | |
866 | if (error) | |
867 | goto error1; | |
868 | } | |
4254b0bb CH |
869 | |
870 | /* | |
871 | * Loop until we find an inode chunk with a free inode. | |
872 | */ | |
873 | while (!doneleft || !doneright) { | |
874 | int useleft; /* using left inode chunk this time */ | |
875 | ||
bd169565 DC |
876 | if (!--searchdistance) { |
877 | /* | |
878 | * Not in range - save last search | |
879 | * location and allocate a new inode | |
880 | */ | |
3b826386 | 881 | xfs_btree_del_cursor(tcur, XFS_BTREE_NOERROR); |
bd169565 DC |
882 | pag->pagl_leftrec = trec.ir_startino; |
883 | pag->pagl_rightrec = rec.ir_startino; | |
884 | pag->pagl_pagino = pagino; | |
885 | goto newino; | |
886 | } | |
887 | ||
4254b0bb CH |
888 | /* figure out the closer block if both are valid. */ |
889 | if (!doneleft && !doneright) { | |
890 | useleft = pagino - | |
891 | (trec.ir_startino + XFS_INODES_PER_CHUNK - 1) < | |
892 | rec.ir_startino - pagino; | |
893 | } else { | |
894 | useleft = !doneleft; | |
1da177e4 | 895 | } |
4254b0bb CH |
896 | |
897 | /* free inodes to the left? */ | |
898 | if (useleft && trec.ir_freecount) { | |
899 | rec = trec; | |
900 | xfs_btree_del_cursor(cur, XFS_BTREE_NOERROR); | |
901 | cur = tcur; | |
bd169565 DC |
902 | |
903 | pag->pagl_leftrec = trec.ir_startino; | |
904 | pag->pagl_rightrec = rec.ir_startino; | |
905 | pag->pagl_pagino = pagino; | |
4254b0bb | 906 | goto alloc_inode; |
1da177e4 | 907 | } |
1da177e4 | 908 | |
4254b0bb CH |
909 | /* free inodes to the right? */ |
910 | if (!useleft && rec.ir_freecount) { | |
911 | xfs_btree_del_cursor(tcur, XFS_BTREE_NOERROR); | |
bd169565 DC |
912 | |
913 | pag->pagl_leftrec = trec.ir_startino; | |
914 | pag->pagl_rightrec = rec.ir_startino; | |
915 | pag->pagl_pagino = pagino; | |
4254b0bb | 916 | goto alloc_inode; |
1da177e4 | 917 | } |
4254b0bb CH |
918 | |
919 | /* get next record to check */ | |
920 | if (useleft) { | |
921 | error = xfs_ialloc_next_rec(tcur, &trec, | |
922 | &doneleft, 1); | |
923 | } else { | |
924 | error = xfs_ialloc_next_rec(cur, &rec, | |
925 | &doneright, 0); | |
926 | } | |
927 | if (error) | |
928 | goto error1; | |
1da177e4 | 929 | } |
bd169565 DC |
930 | |
931 | /* | |
932 | * We've reached the end of the btree. because | |
933 | * we are only searching a small chunk of the | |
934 | * btree each search, there is obviously free | |
935 | * inodes closer to the parent inode than we | |
936 | * are now. restart the search again. | |
937 | */ | |
938 | pag->pagl_pagino = NULLAGINO; | |
939 | pag->pagl_leftrec = NULLAGINO; | |
940 | pag->pagl_rightrec = NULLAGINO; | |
941 | xfs_btree_del_cursor(tcur, XFS_BTREE_NOERROR); | |
942 | xfs_btree_del_cursor(cur, XFS_BTREE_NOERROR); | |
943 | goto restart_pagno; | |
1da177e4 | 944 | } |
4254b0bb | 945 | |
1da177e4 | 946 | /* |
4254b0bb | 947 | * In a different AG from the parent. |
1da177e4 LT |
948 | * See if the most recently allocated block has any free. |
949 | */ | |
bd169565 DC |
950 | newino: |
951 | if (be32_to_cpu(agi->agi_newino) != NULLAGINO) { | |
21875505 CH |
952 | error = xfs_inobt_lookup(cur, be32_to_cpu(agi->agi_newino), |
953 | XFS_LOOKUP_EQ, &i); | |
4254b0bb | 954 | if (error) |
1da177e4 | 955 | goto error0; |
4254b0bb CH |
956 | |
957 | if (i == 1) { | |
958 | error = xfs_inobt_get_rec(cur, &rec, &j); | |
959 | if (error) | |
960 | goto error0; | |
961 | ||
962 | if (j == 1 && rec.ir_freecount > 0) { | |
963 | /* | |
964 | * The last chunk allocated in the group | |
965 | * still has a free inode. | |
966 | */ | |
967 | goto alloc_inode; | |
968 | } | |
1da177e4 | 969 | } |
bd169565 | 970 | } |
4254b0bb | 971 | |
bd169565 DC |
972 | /* |
973 | * None left in the last group, search the whole AG | |
974 | */ | |
975 | error = xfs_inobt_lookup(cur, 0, XFS_LOOKUP_GE, &i); | |
976 | if (error) | |
977 | goto error0; | |
978 | XFS_WANT_CORRUPTED_GOTO(i == 1, error0); | |
979 | ||
980 | for (;;) { | |
981 | error = xfs_inobt_get_rec(cur, &rec, &i); | |
982 | if (error) | |
983 | goto error0; | |
984 | XFS_WANT_CORRUPTED_GOTO(i == 1, error0); | |
985 | if (rec.ir_freecount > 0) | |
986 | break; | |
987 | error = xfs_btree_increment(cur, 0, &i); | |
4254b0bb CH |
988 | if (error) |
989 | goto error0; | |
990 | XFS_WANT_CORRUPTED_GOTO(i == 1, error0); | |
1da177e4 | 991 | } |
4254b0bb CH |
992 | |
993 | alloc_inode: | |
9d87c319 | 994 | offset = xfs_ialloc_find_free(&rec.ir_free); |
1da177e4 LT |
995 | ASSERT(offset >= 0); |
996 | ASSERT(offset < XFS_INODES_PER_CHUNK); | |
997 | ASSERT((XFS_AGINO_TO_OFFSET(mp, rec.ir_startino) % | |
998 | XFS_INODES_PER_CHUNK) == 0); | |
999 | ino = XFS_AGINO_TO_INO(mp, agno, rec.ir_startino + offset); | |
0d87e656 | 1000 | rec.ir_free &= ~XFS_INOBT_MASK(offset); |
1da177e4 | 1001 | rec.ir_freecount--; |
afabc24a CH |
1002 | error = xfs_inobt_update(cur, &rec); |
1003 | if (error) | |
1da177e4 | 1004 | goto error0; |
413d57c9 | 1005 | be32_add_cpu(&agi->agi_freecount, -1); |
1da177e4 | 1006 | xfs_ialloc_log_agi(tp, agbp, XFS_AGI_FREECOUNT); |
44b56e0a | 1007 | pag->pagi_freecount--; |
1da177e4 | 1008 | |
0b48db80 DC |
1009 | error = xfs_check_agi_freecount(cur, agi); |
1010 | if (error) | |
1011 | goto error0; | |
1012 | ||
1da177e4 LT |
1013 | xfs_btree_del_cursor(cur, XFS_BTREE_NOERROR); |
1014 | xfs_trans_mod_sb(tp, XFS_TRANS_SB_IFREE, -1); | |
44b56e0a | 1015 | xfs_perag_put(pag); |
1da177e4 LT |
1016 | *inop = ino; |
1017 | return 0; | |
1018 | error1: | |
1019 | xfs_btree_del_cursor(tcur, XFS_BTREE_ERROR); | |
1020 | error0: | |
1021 | xfs_btree_del_cursor(cur, XFS_BTREE_ERROR); | |
44b56e0a | 1022 | xfs_perag_put(pag); |
1da177e4 LT |
1023 | return error; |
1024 | } | |
1025 | ||
1026 | /* | |
1027 | * Free disk inode. Carefully avoids touching the incore inode, all | |
1028 | * manipulations incore are the caller's responsibility. | |
1029 | * The on-disk inode is not changed by this operation, only the | |
1030 | * btree (free inode mask) is changed. | |
1031 | */ | |
1032 | int | |
1033 | xfs_difree( | |
1034 | xfs_trans_t *tp, /* transaction pointer */ | |
1035 | xfs_ino_t inode, /* inode to be freed */ | |
1036 | xfs_bmap_free_t *flist, /* extents to free */ | |
1037 | int *delete, /* set if inode cluster was deleted */ | |
1038 | xfs_ino_t *first_ino) /* first inode in deleted cluster */ | |
1039 | { | |
1040 | /* REFERENCED */ | |
1041 | xfs_agblock_t agbno; /* block number containing inode */ | |
1042 | xfs_buf_t *agbp; /* buffer containing allocation group header */ | |
1043 | xfs_agino_t agino; /* inode number relative to allocation group */ | |
1044 | xfs_agnumber_t agno; /* allocation group number */ | |
1045 | xfs_agi_t *agi; /* allocation group header */ | |
1046 | xfs_btree_cur_t *cur; /* inode btree cursor */ | |
1047 | int error; /* error return value */ | |
1048 | int i; /* result code */ | |
1049 | int ilen; /* inodes in an inode cluster */ | |
1050 | xfs_mount_t *mp; /* mount structure for filesystem */ | |
1051 | int off; /* offset of inode in inode chunk */ | |
61a25848 | 1052 | xfs_inobt_rec_incore_t rec; /* btree record */ |
44b56e0a | 1053 | struct xfs_perag *pag; |
1da177e4 LT |
1054 | |
1055 | mp = tp->t_mountp; | |
1056 | ||
1057 | /* | |
1058 | * Break up inode number into its components. | |
1059 | */ | |
1060 | agno = XFS_INO_TO_AGNO(mp, inode); | |
1061 | if (agno >= mp->m_sb.sb_agcount) { | |
1062 | cmn_err(CE_WARN, | |
1063 | "xfs_difree: agno >= mp->m_sb.sb_agcount (%d >= %d) on %s. Returning EINVAL.", | |
1064 | agno, mp->m_sb.sb_agcount, mp->m_fsname); | |
1065 | ASSERT(0); | |
1066 | return XFS_ERROR(EINVAL); | |
1067 | } | |
1068 | agino = XFS_INO_TO_AGINO(mp, inode); | |
1069 | if (inode != XFS_AGINO_TO_INO(mp, agno, agino)) { | |
1070 | cmn_err(CE_WARN, | |
da1650a5 CH |
1071 | "xfs_difree: inode != XFS_AGINO_TO_INO() " |
1072 | "(%llu != %llu) on %s. Returning EINVAL.", | |
1073 | (unsigned long long)inode, | |
1074 | (unsigned long long)XFS_AGINO_TO_INO(mp, agno, agino), | |
1075 | mp->m_fsname); | |
1da177e4 LT |
1076 | ASSERT(0); |
1077 | return XFS_ERROR(EINVAL); | |
1078 | } | |
1079 | agbno = XFS_AGINO_TO_AGBNO(mp, agino); | |
1080 | if (agbno >= mp->m_sb.sb_agblocks) { | |
1081 | cmn_err(CE_WARN, | |
1082 | "xfs_difree: agbno >= mp->m_sb.sb_agblocks (%d >= %d) on %s. Returning EINVAL.", | |
1083 | agbno, mp->m_sb.sb_agblocks, mp->m_fsname); | |
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 LT |
1091 | if (error) { |
1092 | cmn_err(CE_WARN, | |
1093 | "xfs_difree: xfs_ialloc_read_agi() returned an error %d on %s. Returning error.", | |
1094 | error, mp->m_fsname); | |
1095 | return error; | |
1096 | } | |
1097 | agi = XFS_BUF_TO_AGI(agbp); | |
16259e7d CH |
1098 | ASSERT(be32_to_cpu(agi->agi_magicnum) == XFS_AGI_MAGIC); |
1099 | ASSERT(agbno < be32_to_cpu(agi->agi_length)); | |
1da177e4 LT |
1100 | /* |
1101 | * Initialize the cursor. | |
1102 | */ | |
561f7d17 | 1103 | cur = xfs_inobt_init_cursor(mp, tp, agbp, agno); |
1da177e4 | 1104 | |
0b48db80 DC |
1105 | error = xfs_check_agi_freecount(cur, agi); |
1106 | if (error) | |
1107 | goto error0; | |
1108 | ||
1da177e4 LT |
1109 | /* |
1110 | * Look for the entry describing this inode. | |
1111 | */ | |
21875505 | 1112 | if ((error = xfs_inobt_lookup(cur, agino, XFS_LOOKUP_LE, &i))) { |
1da177e4 | 1113 | cmn_err(CE_WARN, |
21875505 | 1114 | "xfs_difree: xfs_inobt_lookup returned() an error %d on %s. Returning error.", |
1da177e4 LT |
1115 | error, mp->m_fsname); |
1116 | goto error0; | |
1117 | } | |
1118 | XFS_WANT_CORRUPTED_GOTO(i == 1, error0); | |
2e287a73 CH |
1119 | error = xfs_inobt_get_rec(cur, &rec, &i); |
1120 | if (error) { | |
1da177e4 LT |
1121 | cmn_err(CE_WARN, |
1122 | "xfs_difree: xfs_inobt_get_rec() returned an error %d on %s. Returning error.", | |
1123 | error, mp->m_fsname); | |
1124 | goto error0; | |
1125 | } | |
1126 | XFS_WANT_CORRUPTED_GOTO(i == 1, error0); | |
1127 | /* | |
1128 | * Get the offset in the inode chunk. | |
1129 | */ | |
1130 | off = agino - rec.ir_startino; | |
1131 | ASSERT(off >= 0 && off < XFS_INODES_PER_CHUNK); | |
0d87e656 | 1132 | ASSERT(!(rec.ir_free & XFS_INOBT_MASK(off))); |
1da177e4 LT |
1133 | /* |
1134 | * Mark the inode free & increment the count. | |
1135 | */ | |
0d87e656 | 1136 | rec.ir_free |= XFS_INOBT_MASK(off); |
1da177e4 LT |
1137 | rec.ir_freecount++; |
1138 | ||
1139 | /* | |
c41564b5 | 1140 | * When an inode cluster is free, it becomes eligible for removal |
1da177e4 | 1141 | */ |
1bd960ee | 1142 | if (!(mp->m_flags & XFS_MOUNT_IKEEP) && |
1da177e4 LT |
1143 | (rec.ir_freecount == XFS_IALLOC_INODES(mp))) { |
1144 | ||
1145 | *delete = 1; | |
1146 | *first_ino = XFS_AGINO_TO_INO(mp, agno, rec.ir_startino); | |
1147 | ||
1148 | /* | |
1149 | * Remove the inode cluster from the AGI B+Tree, adjust the | |
1150 | * AGI and Superblock inode counts, and mark the disk space | |
1151 | * to be freed when the transaction is committed. | |
1152 | */ | |
1153 | ilen = XFS_IALLOC_INODES(mp); | |
413d57c9 MS |
1154 | be32_add_cpu(&agi->agi_count, -ilen); |
1155 | be32_add_cpu(&agi->agi_freecount, -(ilen - 1)); | |
1da177e4 | 1156 | xfs_ialloc_log_agi(tp, agbp, XFS_AGI_COUNT | XFS_AGI_FREECOUNT); |
44b56e0a DC |
1157 | pag = xfs_perag_get(mp, agno); |
1158 | pag->pagi_freecount -= ilen - 1; | |
1159 | xfs_perag_put(pag); | |
1da177e4 LT |
1160 | xfs_trans_mod_sb(tp, XFS_TRANS_SB_ICOUNT, -ilen); |
1161 | xfs_trans_mod_sb(tp, XFS_TRANS_SB_IFREE, -(ilen - 1)); | |
1162 | ||
91cca5df CH |
1163 | if ((error = xfs_btree_delete(cur, &i))) { |
1164 | cmn_err(CE_WARN, "xfs_difree: xfs_btree_delete returned an error %d on %s.\n", | |
1da177e4 LT |
1165 | error, mp->m_fsname); |
1166 | goto error0; | |
1167 | } | |
1168 | ||
1169 | xfs_bmap_add_free(XFS_AGB_TO_FSB(mp, | |
1170 | agno, XFS_INO_TO_AGBNO(mp,rec.ir_startino)), | |
1171 | XFS_IALLOC_BLOCKS(mp), flist, mp); | |
1172 | } else { | |
1173 | *delete = 0; | |
1174 | ||
afabc24a CH |
1175 | error = xfs_inobt_update(cur, &rec); |
1176 | if (error) { | |
1da177e4 | 1177 | cmn_err(CE_WARN, |
afabc24a | 1178 | "xfs_difree: xfs_inobt_update returned an error %d on %s.", |
1da177e4 LT |
1179 | error, mp->m_fsname); |
1180 | goto error0; | |
1181 | } | |
afabc24a | 1182 | |
1da177e4 LT |
1183 | /* |
1184 | * Change the inode free counts and log the ag/sb changes. | |
1185 | */ | |
413d57c9 | 1186 | be32_add_cpu(&agi->agi_freecount, 1); |
1da177e4 | 1187 | xfs_ialloc_log_agi(tp, agbp, XFS_AGI_FREECOUNT); |
44b56e0a DC |
1188 | pag = xfs_perag_get(mp, agno); |
1189 | pag->pagi_freecount++; | |
1190 | xfs_perag_put(pag); | |
1da177e4 LT |
1191 | xfs_trans_mod_sb(tp, XFS_TRANS_SB_IFREE, 1); |
1192 | } | |
1193 | ||
0b48db80 DC |
1194 | error = xfs_check_agi_freecount(cur, agi); |
1195 | if (error) | |
1196 | goto error0; | |
1da177e4 | 1197 | |
1da177e4 LT |
1198 | xfs_btree_del_cursor(cur, XFS_BTREE_NOERROR); |
1199 | return 0; | |
1200 | ||
1201 | error0: | |
1202 | xfs_btree_del_cursor(cur, XFS_BTREE_ERROR); | |
1203 | return error; | |
1204 | } | |
1205 | ||
7124fe0a DC |
1206 | STATIC int |
1207 | xfs_imap_lookup( | |
1208 | struct xfs_mount *mp, | |
1209 | struct xfs_trans *tp, | |
1210 | xfs_agnumber_t agno, | |
1211 | xfs_agino_t agino, | |
1212 | xfs_agblock_t agbno, | |
1213 | xfs_agblock_t *chunk_agbno, | |
1214 | xfs_agblock_t *offset_agbno, | |
1215 | int flags) | |
1216 | { | |
1217 | struct xfs_inobt_rec_incore rec; | |
1218 | struct xfs_btree_cur *cur; | |
1219 | struct xfs_buf *agbp; | |
1220 | xfs_agino_t startino; | |
1221 | int error; | |
1222 | int i; | |
1223 | ||
1224 | error = xfs_ialloc_read_agi(mp, tp, agno, &agbp); | |
1225 | if (error) { | |
1226 | xfs_fs_cmn_err(CE_ALERT, mp, "xfs_imap: " | |
1227 | "xfs_ialloc_read_agi() returned " | |
1228 | "error %d, agno %d", | |
1229 | error, agno); | |
1230 | return error; | |
1231 | } | |
1232 | ||
1233 | /* | |
1234 | * derive and lookup the exact inode record for the given agino. If the | |
1235 | * record cannot be found, then it's an invalid inode number and we | |
1236 | * should abort. | |
1237 | */ | |
1238 | cur = xfs_inobt_init_cursor(mp, tp, agbp, agno); | |
1239 | startino = agino & ~(XFS_IALLOC_INODES(mp) - 1); | |
1240 | error = xfs_inobt_lookup(cur, startino, XFS_LOOKUP_EQ, &i); | |
1241 | if (!error) { | |
1242 | if (i) | |
1243 | error = xfs_inobt_get_rec(cur, &rec, &i); | |
1244 | if (!error && i == 0) | |
1245 | error = EINVAL; | |
1246 | } | |
1247 | ||
1248 | xfs_trans_brelse(tp, agbp); | |
1249 | xfs_btree_del_cursor(cur, XFS_BTREE_NOERROR); | |
1250 | if (error) | |
1251 | return error; | |
1252 | ||
1253 | /* for untrusted inodes check it is allocated first */ | |
1254 | if ((flags & XFS_IGET_BULKSTAT) && | |
1255 | (rec.ir_free & XFS_INOBT_MASK(agino - rec.ir_startino))) | |
1256 | return EINVAL; | |
1257 | ||
1258 | *chunk_agbno = XFS_AGINO_TO_AGBNO(mp, rec.ir_startino); | |
1259 | *offset_agbno = agbno - *chunk_agbno; | |
1260 | return 0; | |
1261 | } | |
1262 | ||
1da177e4 | 1263 | /* |
94e1b69d | 1264 | * Return the location of the inode in imap, for mapping it into a buffer. |
1da177e4 | 1265 | */ |
1da177e4 | 1266 | int |
94e1b69d CH |
1267 | xfs_imap( |
1268 | xfs_mount_t *mp, /* file system mount structure */ | |
1269 | xfs_trans_t *tp, /* transaction pointer */ | |
1da177e4 | 1270 | xfs_ino_t ino, /* inode to locate */ |
94e1b69d CH |
1271 | struct xfs_imap *imap, /* location map structure */ |
1272 | uint flags) /* flags for inode btree lookup */ | |
1da177e4 LT |
1273 | { |
1274 | xfs_agblock_t agbno; /* block number of inode in the alloc group */ | |
1da177e4 LT |
1275 | xfs_agino_t agino; /* inode number within alloc group */ |
1276 | xfs_agnumber_t agno; /* allocation group number */ | |
1277 | int blks_per_cluster; /* num blocks per inode cluster */ | |
1278 | xfs_agblock_t chunk_agbno; /* first block in inode chunk */ | |
1da177e4 | 1279 | xfs_agblock_t cluster_agbno; /* first block in inode cluster */ |
1da177e4 | 1280 | int error; /* error code */ |
1da177e4 LT |
1281 | int offset; /* index of inode in its buffer */ |
1282 | int offset_agbno; /* blks from chunk start to inode */ | |
1283 | ||
1284 | ASSERT(ino != NULLFSINO); | |
94e1b69d | 1285 | |
1da177e4 LT |
1286 | /* |
1287 | * Split up the inode number into its parts. | |
1288 | */ | |
1289 | agno = XFS_INO_TO_AGNO(mp, ino); | |
1290 | agino = XFS_INO_TO_AGINO(mp, ino); | |
1291 | agbno = XFS_AGINO_TO_AGBNO(mp, agino); | |
1292 | if (agno >= mp->m_sb.sb_agcount || agbno >= mp->m_sb.sb_agblocks || | |
1293 | ino != XFS_AGINO_TO_INO(mp, agno, agino)) { | |
1294 | #ifdef DEBUG | |
4d1a2ed3 | 1295 | /* no diagnostics for bulkstat, ino comes from userspace */ |
b48d8d64 | 1296 | if (flags & XFS_IGET_BULKSTAT) |
4d1a2ed3 | 1297 | return XFS_ERROR(EINVAL); |
1da177e4 LT |
1298 | if (agno >= mp->m_sb.sb_agcount) { |
1299 | xfs_fs_cmn_err(CE_ALERT, mp, | |
94e1b69d | 1300 | "xfs_imap: agno (%d) >= " |
1da177e4 LT |
1301 | "mp->m_sb.sb_agcount (%d)", |
1302 | agno, mp->m_sb.sb_agcount); | |
1303 | } | |
1304 | if (agbno >= mp->m_sb.sb_agblocks) { | |
1305 | xfs_fs_cmn_err(CE_ALERT, mp, | |
94e1b69d | 1306 | "xfs_imap: agbno (0x%llx) >= " |
1da177e4 LT |
1307 | "mp->m_sb.sb_agblocks (0x%lx)", |
1308 | (unsigned long long) agbno, | |
1309 | (unsigned long) mp->m_sb.sb_agblocks); | |
1310 | } | |
1311 | if (ino != XFS_AGINO_TO_INO(mp, agno, agino)) { | |
1312 | xfs_fs_cmn_err(CE_ALERT, mp, | |
94e1b69d | 1313 | "xfs_imap: ino (0x%llx) != " |
1da177e4 LT |
1314 | "XFS_AGINO_TO_INO(mp, agno, agino) " |
1315 | "(0x%llx)", | |
1316 | ino, XFS_AGINO_TO_INO(mp, agno, agino)); | |
1317 | } | |
745b1f47 | 1318 | xfs_stack_trace(); |
1da177e4 LT |
1319 | #endif /* DEBUG */ |
1320 | return XFS_ERROR(EINVAL); | |
1321 | } | |
94e1b69d | 1322 | |
7124fe0a DC |
1323 | blks_per_cluster = XFS_INODE_CLUSTER_SIZE(mp) >> mp->m_sb.sb_blocklog; |
1324 | ||
1325 | /* | |
1326 | * For bulkstat and handle lookups, we have an untrusted inode number | |
1327 | * that we have to verify is valid. We cannot do this just by reading | |
1328 | * the inode buffer as it may have been unlinked and removed leaving | |
1329 | * inodes in stale state on disk. Hence we have to do a btree lookup | |
1330 | * in all cases where an untrusted inode number is passed. | |
1331 | */ | |
1332 | if (flags & XFS_IGET_BULKSTAT) { | |
1333 | error = xfs_imap_lookup(mp, tp, agno, agino, agbno, | |
1334 | &chunk_agbno, &offset_agbno, flags); | |
1335 | if (error) | |
1336 | return error; | |
1337 | goto out_map; | |
1338 | } | |
1339 | ||
94e1b69d CH |
1340 | /* |
1341 | * If the inode cluster size is the same as the blocksize or | |
1342 | * smaller we get to the buffer by simple arithmetics. | |
1343 | */ | |
1344 | if (XFS_INODE_CLUSTER_SIZE(mp) <= mp->m_sb.sb_blocksize) { | |
1da177e4 LT |
1345 | offset = XFS_INO_TO_OFFSET(mp, ino); |
1346 | ASSERT(offset < mp->m_sb.sb_inopblock); | |
94e1b69d CH |
1347 | |
1348 | imap->im_blkno = XFS_AGB_TO_DADDR(mp, agno, agbno); | |
1349 | imap->im_len = XFS_FSB_TO_BB(mp, 1); | |
1350 | imap->im_boffset = (ushort)(offset << mp->m_sb.sb_inodelog); | |
1da177e4 LT |
1351 | return 0; |
1352 | } | |
94e1b69d | 1353 | |
94e1b69d | 1354 | /* |
7124fe0a | 1355 | * If we get a block number passed we can use it to |
94e1b69d CH |
1356 | * find the buffer easily. |
1357 | */ | |
1358 | if (imap->im_blkno) { | |
1da177e4 LT |
1359 | offset = XFS_INO_TO_OFFSET(mp, ino); |
1360 | ASSERT(offset < mp->m_sb.sb_inopblock); | |
94e1b69d | 1361 | |
9d87c319 | 1362 | cluster_agbno = xfs_daddr_to_agbno(mp, imap->im_blkno); |
94e1b69d CH |
1363 | offset += (agbno - cluster_agbno) * mp->m_sb.sb_inopblock; |
1364 | ||
1365 | imap->im_len = XFS_FSB_TO_BB(mp, blks_per_cluster); | |
1366 | imap->im_boffset = (ushort)(offset << mp->m_sb.sb_inodelog); | |
1da177e4 LT |
1367 | return 0; |
1368 | } | |
94e1b69d CH |
1369 | |
1370 | /* | |
1371 | * If the inode chunks are aligned then use simple maths to | |
1372 | * find the location. Otherwise we have to do a btree | |
1373 | * lookup to find the location. | |
1374 | */ | |
1da177e4 LT |
1375 | if (mp->m_inoalign_mask) { |
1376 | offset_agbno = agbno & mp->m_inoalign_mask; | |
1377 | chunk_agbno = agbno - offset_agbno; | |
1378 | } else { | |
7124fe0a DC |
1379 | error = xfs_imap_lookup(mp, tp, agno, agino, agbno, |
1380 | &chunk_agbno, &offset_agbno, flags); | |
1da177e4 LT |
1381 | if (error) |
1382 | return error; | |
1da177e4 | 1383 | } |
94e1b69d | 1384 | |
7124fe0a | 1385 | out_map: |
1da177e4 LT |
1386 | ASSERT(agbno >= chunk_agbno); |
1387 | cluster_agbno = chunk_agbno + | |
1388 | ((offset_agbno / blks_per_cluster) * blks_per_cluster); | |
1389 | offset = ((agbno - cluster_agbno) * mp->m_sb.sb_inopblock) + | |
1390 | XFS_INO_TO_OFFSET(mp, ino); | |
94e1b69d CH |
1391 | |
1392 | imap->im_blkno = XFS_AGB_TO_DADDR(mp, agno, cluster_agbno); | |
1393 | imap->im_len = XFS_FSB_TO_BB(mp, blks_per_cluster); | |
1394 | imap->im_boffset = (ushort)(offset << mp->m_sb.sb_inodelog); | |
1395 | ||
1396 | /* | |
1397 | * If the inode number maps to a block outside the bounds | |
1398 | * of the file system then return NULL rather than calling | |
1399 | * read_buf and panicing when we get an error from the | |
1400 | * driver. | |
1401 | */ | |
1402 | if ((imap->im_blkno + imap->im_len) > | |
1403 | XFS_FSB_TO_BB(mp, mp->m_sb.sb_dblocks)) { | |
1404 | xfs_fs_cmn_err(CE_ALERT, mp, "xfs_imap: " | |
1405 | "(imap->im_blkno (0x%llx) + imap->im_len (0x%llx)) > " | |
1406 | " XFS_FSB_TO_BB(mp, mp->m_sb.sb_dblocks) (0x%llx)", | |
1407 | (unsigned long long) imap->im_blkno, | |
1408 | (unsigned long long) imap->im_len, | |
1409 | XFS_FSB_TO_BB(mp, mp->m_sb.sb_dblocks)); | |
1410 | return XFS_ERROR(EINVAL); | |
1411 | } | |
1da177e4 | 1412 | return 0; |
1da177e4 LT |
1413 | } |
1414 | ||
1415 | /* | |
1416 | * Compute and fill in value of m_in_maxlevels. | |
1417 | */ | |
1418 | void | |
1419 | xfs_ialloc_compute_maxlevels( | |
1420 | xfs_mount_t *mp) /* file system mount structure */ | |
1421 | { | |
1422 | int level; | |
1423 | uint maxblocks; | |
1424 | uint maxleafents; | |
1425 | int minleafrecs; | |
1426 | int minnoderecs; | |
1427 | ||
1428 | maxleafents = (1LL << XFS_INO_AGINO_BITS(mp)) >> | |
1429 | XFS_INODES_PER_CHUNK_LOG; | |
1430 | minleafrecs = mp->m_alloc_mnr[0]; | |
1431 | minnoderecs = mp->m_alloc_mnr[1]; | |
1432 | maxblocks = (maxleafents + minleafrecs - 1) / minleafrecs; | |
1433 | for (level = 1; maxblocks > 1; level++) | |
1434 | maxblocks = (maxblocks + minnoderecs - 1) / minnoderecs; | |
1435 | mp->m_in_maxlevels = level; | |
1436 | } | |
1437 | ||
1438 | /* | |
1439 | * Log specified fields for the ag hdr (inode section) | |
1440 | */ | |
1441 | void | |
1442 | xfs_ialloc_log_agi( | |
1443 | xfs_trans_t *tp, /* transaction pointer */ | |
1444 | xfs_buf_t *bp, /* allocation group header buffer */ | |
1445 | int fields) /* bitmask of fields to log */ | |
1446 | { | |
1447 | int first; /* first byte number */ | |
1448 | int last; /* last byte number */ | |
1449 | static const short offsets[] = { /* field starting offsets */ | |
1450 | /* keep in sync with bit definitions */ | |
1451 | offsetof(xfs_agi_t, agi_magicnum), | |
1452 | offsetof(xfs_agi_t, agi_versionnum), | |
1453 | offsetof(xfs_agi_t, agi_seqno), | |
1454 | offsetof(xfs_agi_t, agi_length), | |
1455 | offsetof(xfs_agi_t, agi_count), | |
1456 | offsetof(xfs_agi_t, agi_root), | |
1457 | offsetof(xfs_agi_t, agi_level), | |
1458 | offsetof(xfs_agi_t, agi_freecount), | |
1459 | offsetof(xfs_agi_t, agi_newino), | |
1460 | offsetof(xfs_agi_t, agi_dirino), | |
1461 | offsetof(xfs_agi_t, agi_unlinked), | |
1462 | sizeof(xfs_agi_t) | |
1463 | }; | |
1464 | #ifdef DEBUG | |
1465 | xfs_agi_t *agi; /* allocation group header */ | |
1466 | ||
1467 | agi = XFS_BUF_TO_AGI(bp); | |
16259e7d | 1468 | ASSERT(be32_to_cpu(agi->agi_magicnum) == XFS_AGI_MAGIC); |
1da177e4 LT |
1469 | #endif |
1470 | /* | |
1471 | * Compute byte offsets for the first and last fields. | |
1472 | */ | |
1473 | xfs_btree_offsets(fields, offsets, XFS_AGI_NUM_BITS, &first, &last); | |
1474 | /* | |
1475 | * Log the allocation group inode header buffer. | |
1476 | */ | |
1477 | xfs_trans_log_buf(tp, bp, first, last); | |
1478 | } | |
1479 | ||
5e1be0fb CH |
1480 | #ifdef DEBUG |
1481 | STATIC void | |
1482 | xfs_check_agi_unlinked( | |
1483 | struct xfs_agi *agi) | |
1484 | { | |
1485 | int i; | |
1486 | ||
1487 | for (i = 0; i < XFS_AGI_UNLINKED_BUCKETS; i++) | |
1488 | ASSERT(agi->agi_unlinked[i]); | |
1489 | } | |
1490 | #else | |
1491 | #define xfs_check_agi_unlinked(agi) | |
1492 | #endif | |
1493 | ||
1da177e4 LT |
1494 | /* |
1495 | * Read in the allocation group header (inode allocation section) | |
1496 | */ | |
1497 | int | |
5e1be0fb CH |
1498 | xfs_read_agi( |
1499 | struct xfs_mount *mp, /* file system mount structure */ | |
1500 | struct xfs_trans *tp, /* transaction pointer */ | |
1501 | xfs_agnumber_t agno, /* allocation group number */ | |
1502 | struct xfs_buf **bpp) /* allocation group hdr buf */ | |
1da177e4 | 1503 | { |
5e1be0fb CH |
1504 | struct xfs_agi *agi; /* allocation group header */ |
1505 | int agi_ok; /* agi is consistent */ | |
1506 | int error; | |
1da177e4 LT |
1507 | |
1508 | ASSERT(agno != NULLAGNUMBER); | |
5e1be0fb CH |
1509 | |
1510 | error = xfs_trans_read_buf(mp, tp, mp->m_ddev_targp, | |
1da177e4 | 1511 | XFS_AG_DADDR(mp, agno, XFS_AGI_DADDR(mp)), |
5e1be0fb | 1512 | XFS_FSS_TO_BB(mp, 1), 0, bpp); |
1da177e4 LT |
1513 | if (error) |
1514 | return error; | |
5e1be0fb CH |
1515 | |
1516 | ASSERT(*bpp && !XFS_BUF_GETERROR(*bpp)); | |
1517 | agi = XFS_BUF_TO_AGI(*bpp); | |
1da177e4 LT |
1518 | |
1519 | /* | |
1520 | * Validate the magic number of the agi block. | |
1521 | */ | |
5e1be0fb CH |
1522 | agi_ok = be32_to_cpu(agi->agi_magicnum) == XFS_AGI_MAGIC && |
1523 | XFS_AGI_GOOD_VERSION(be32_to_cpu(agi->agi_versionnum)) && | |
1524 | be32_to_cpu(agi->agi_seqno) == agno; | |
1da177e4 LT |
1525 | if (unlikely(XFS_TEST_ERROR(!agi_ok, mp, XFS_ERRTAG_IALLOC_READ_AGI, |
1526 | XFS_RANDOM_IALLOC_READ_AGI))) { | |
5e1be0fb | 1527 | XFS_CORRUPTION_ERROR("xfs_read_agi", XFS_ERRLEVEL_LOW, |
1da177e4 | 1528 | mp, agi); |
5e1be0fb | 1529 | xfs_trans_brelse(tp, *bpp); |
1da177e4 LT |
1530 | return XFS_ERROR(EFSCORRUPTED); |
1531 | } | |
5e1be0fb CH |
1532 | |
1533 | XFS_BUF_SET_VTYPE_REF(*bpp, B_FS_AGI, XFS_AGI_REF); | |
1534 | ||
1535 | xfs_check_agi_unlinked(agi); | |
1536 | return 0; | |
1537 | } | |
1538 | ||
1539 | int | |
1540 | xfs_ialloc_read_agi( | |
1541 | struct xfs_mount *mp, /* file system mount structure */ | |
1542 | struct xfs_trans *tp, /* transaction pointer */ | |
1543 | xfs_agnumber_t agno, /* allocation group number */ | |
1544 | struct xfs_buf **bpp) /* allocation group hdr buf */ | |
1545 | { | |
1546 | struct xfs_agi *agi; /* allocation group header */ | |
1547 | struct xfs_perag *pag; /* per allocation group data */ | |
1548 | int error; | |
1549 | ||
1550 | error = xfs_read_agi(mp, tp, agno, bpp); | |
1551 | if (error) | |
1552 | return error; | |
1553 | ||
1554 | agi = XFS_BUF_TO_AGI(*bpp); | |
44b56e0a | 1555 | pag = xfs_perag_get(mp, agno); |
1da177e4 | 1556 | if (!pag->pagi_init) { |
16259e7d | 1557 | pag->pagi_freecount = be32_to_cpu(agi->agi_freecount); |
92821e2b | 1558 | pag->pagi_count = be32_to_cpu(agi->agi_count); |
1da177e4 | 1559 | pag->pagi_init = 1; |
1da177e4 | 1560 | } |
1da177e4 | 1561 | |
5e1be0fb CH |
1562 | /* |
1563 | * It's possible for these to be out of sync if | |
1564 | * we are in the middle of a forced shutdown. | |
1565 | */ | |
1566 | ASSERT(pag->pagi_freecount == be32_to_cpu(agi->agi_freecount) || | |
1567 | XFS_FORCED_SHUTDOWN(mp)); | |
44b56e0a | 1568 | xfs_perag_put(pag); |
1da177e4 LT |
1569 | return 0; |
1570 | } | |
92821e2b DC |
1571 | |
1572 | /* | |
1573 | * Read in the agi to initialise the per-ag data in the mount structure | |
1574 | */ | |
1575 | int | |
1576 | xfs_ialloc_pagi_init( | |
1577 | xfs_mount_t *mp, /* file system mount structure */ | |
1578 | xfs_trans_t *tp, /* transaction pointer */ | |
1579 | xfs_agnumber_t agno) /* allocation group number */ | |
1580 | { | |
1581 | xfs_buf_t *bp = NULL; | |
1582 | int error; | |
1583 | ||
1584 | error = xfs_ialloc_read_agi(mp, tp, agno, &bp); | |
1585 | if (error) | |
1586 | return error; | |
1587 | if (bp) | |
1588 | xfs_trans_brelse(tp, bp); | |
1589 | return 0; | |
1590 | } |