]>
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 CH |
59 | /* |
60 | * Lookup the record equal to ino in the btree given by cur. | |
61 | */ | |
62 | STATIC int /* error */ | |
63 | xfs_inobt_lookup_eq( | |
64 | struct xfs_btree_cur *cur, /* btree cursor */ | |
65 | xfs_agino_t ino, /* starting inode of chunk */ | |
66 | __int32_t fcnt, /* free inode count */ | |
67 | xfs_inofree_t free, /* free inode mask */ | |
68 | int *stat) /* success/failure */ | |
69 | { | |
70 | cur->bc_rec.i.ir_startino = ino; | |
71 | cur->bc_rec.i.ir_freecount = fcnt; | |
72 | cur->bc_rec.i.ir_free = free; | |
73 | return xfs_btree_lookup(cur, XFS_LOOKUP_EQ, stat); | |
74 | } | |
75 | ||
76 | /* | |
77 | * Lookup the first record greater than or equal to ino | |
78 | * in the btree given by cur. | |
79 | */ | |
80 | int /* error */ | |
81 | xfs_inobt_lookup_ge( | |
82 | struct xfs_btree_cur *cur, /* btree cursor */ | |
83 | xfs_agino_t ino, /* starting inode of chunk */ | |
84 | __int32_t fcnt, /* free inode count */ | |
85 | xfs_inofree_t free, /* free inode mask */ | |
86 | int *stat) /* success/failure */ | |
87 | { | |
88 | cur->bc_rec.i.ir_startino = ino; | |
89 | cur->bc_rec.i.ir_freecount = fcnt; | |
90 | cur->bc_rec.i.ir_free = free; | |
91 | return xfs_btree_lookup(cur, XFS_LOOKUP_GE, stat); | |
92 | } | |
93 | ||
94 | /* | |
95 | * Lookup the first record less than or equal to ino | |
96 | * in the btree given by cur. | |
97 | */ | |
98 | int /* error */ | |
99 | xfs_inobt_lookup_le( | |
100 | struct xfs_btree_cur *cur, /* btree cursor */ | |
101 | xfs_agino_t ino, /* starting inode of chunk */ | |
102 | __int32_t fcnt, /* free inode count */ | |
103 | xfs_inofree_t free, /* free inode mask */ | |
104 | int *stat) /* success/failure */ | |
105 | { | |
106 | cur->bc_rec.i.ir_startino = ino; | |
107 | cur->bc_rec.i.ir_freecount = fcnt; | |
108 | cur->bc_rec.i.ir_free = free; | |
109 | return xfs_btree_lookup(cur, XFS_LOOKUP_LE, stat); | |
110 | } | |
111 | ||
278d0ca1 | 112 | /* |
afabc24a | 113 | * Update the record referred to by cur to the value given. |
278d0ca1 CH |
114 | * This either works (return 0) or gets an EFSCORRUPTED error. |
115 | */ | |
116 | STATIC int /* error */ | |
117 | xfs_inobt_update( | |
118 | struct xfs_btree_cur *cur, /* btree cursor */ | |
afabc24a | 119 | xfs_inobt_rec_incore_t *irec) /* btree record */ |
278d0ca1 CH |
120 | { |
121 | union xfs_btree_rec rec; | |
122 | ||
afabc24a CH |
123 | rec.inobt.ir_startino = cpu_to_be32(irec->ir_startino); |
124 | rec.inobt.ir_freecount = cpu_to_be32(irec->ir_freecount); | |
125 | rec.inobt.ir_free = cpu_to_be64(irec->ir_free); | |
278d0ca1 CH |
126 | return xfs_btree_update(cur, &rec); |
127 | } | |
128 | ||
8cc938fe CH |
129 | /* |
130 | * Get the data from the pointed-to record. | |
131 | */ | |
132 | int /* error */ | |
133 | xfs_inobt_get_rec( | |
134 | struct xfs_btree_cur *cur, /* btree cursor */ | |
2e287a73 | 135 | xfs_inobt_rec_incore_t *irec, /* btree record */ |
8cc938fe CH |
136 | int *stat) /* output: success/failure */ |
137 | { | |
138 | union xfs_btree_rec *rec; | |
139 | int error; | |
140 | ||
141 | error = xfs_btree_get_rec(cur, &rec, stat); | |
142 | if (!error && *stat == 1) { | |
2e287a73 CH |
143 | irec->ir_startino = be32_to_cpu(rec->inobt.ir_startino); |
144 | irec->ir_freecount = be32_to_cpu(rec->inobt.ir_freecount); | |
145 | irec->ir_free = be64_to_cpu(rec->inobt.ir_free); | |
8cc938fe CH |
146 | } |
147 | return error; | |
148 | } | |
149 | ||
85c0b2ab DC |
150 | /* |
151 | * Initialise a new set of inodes. | |
152 | */ | |
153 | STATIC void | |
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, | |
203 | mp->m_bsize * blks_per_cluster, | |
204 | XFS_BUF_LOCK); | |
205 | ASSERT(fbuf); | |
206 | ASSERT(!XFS_BUF_GETERROR(fbuf)); | |
207 | ||
208 | /* | |
209 | * Initialize all inodes in this buffer and then log them. | |
210 | * | |
211 | * XXX: It would be much better if we had just one transaction | |
212 | * to log a whole cluster of inodes instead of all the | |
213 | * individual transactions causing a lot of log traffic. | |
214 | */ | |
215 | xfs_biozero(fbuf, 0, ninodes << mp->m_sb.sb_inodelog); | |
216 | for (i = 0; i < ninodes; i++) { | |
217 | int ioffset = i << mp->m_sb.sb_inodelog; | |
218 | uint isize = sizeof(struct xfs_dinode); | |
219 | ||
220 | free = xfs_make_iptr(mp, fbuf, i); | |
221 | free->di_magic = cpu_to_be16(XFS_DINODE_MAGIC); | |
222 | free->di_version = version; | |
223 | free->di_gen = cpu_to_be32(gen); | |
224 | free->di_next_unlinked = cpu_to_be32(NULLAGINO); | |
225 | xfs_trans_log_buf(tp, fbuf, ioffset, ioffset + isize - 1); | |
226 | } | |
227 | xfs_trans_inode_alloc_buf(tp, fbuf); | |
228 | } | |
229 | } | |
230 | ||
1da177e4 LT |
231 | /* |
232 | * Allocate new inodes in the allocation group specified by agbp. | |
233 | * Return 0 for success, else error code. | |
234 | */ | |
235 | STATIC int /* error code or 0 */ | |
236 | xfs_ialloc_ag_alloc( | |
237 | xfs_trans_t *tp, /* transaction pointer */ | |
238 | xfs_buf_t *agbp, /* alloc group buffer */ | |
239 | int *alloc) | |
240 | { | |
241 | xfs_agi_t *agi; /* allocation group header */ | |
242 | xfs_alloc_arg_t args; /* allocation argument structure */ | |
1da177e4 | 243 | xfs_btree_cur_t *cur; /* inode btree cursor */ |
92821e2b | 244 | xfs_agnumber_t agno; |
1da177e4 | 245 | int error; |
85c0b2ab | 246 | int i; |
1da177e4 LT |
247 | xfs_agino_t newino; /* new first inode's number */ |
248 | xfs_agino_t newlen; /* new number of inodes */ | |
1da177e4 | 249 | xfs_agino_t thisino; /* current inode number, for loop */ |
3ccb8b5f | 250 | int isaligned = 0; /* inode allocation at stripe unit */ |
1da177e4 | 251 | /* boundary */ |
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 | */ | |
85c0b2ab DC |
372 | xfs_ialloc_inode_init(args.mp, tp, agno, args.agbno, args.len, |
373 | random32()); | |
d42f08f6 | 374 | |
85c0b2ab DC |
375 | /* |
376 | * Convert the results. | |
377 | */ | |
378 | newino = XFS_OFFBNO_TO_AGINO(args.mp, args.agbno, 0); | |
413d57c9 MS |
379 | be32_add_cpu(&agi->agi_count, newlen); |
380 | be32_add_cpu(&agi->agi_freecount, newlen); | |
1da177e4 | 381 | down_read(&args.mp->m_peraglock); |
92821e2b | 382 | args.mp->m_perag[agno].pagi_freecount += newlen; |
1da177e4 | 383 | up_read(&args.mp->m_peraglock); |
16259e7d | 384 | agi->agi_newino = cpu_to_be32(newino); |
85c0b2ab | 385 | |
1da177e4 LT |
386 | /* |
387 | * Insert records describing the new inode chunk into the btree. | |
388 | */ | |
561f7d17 | 389 | cur = xfs_inobt_init_cursor(args.mp, tp, agbp, agno); |
1da177e4 LT |
390 | for (thisino = newino; |
391 | thisino < newino + newlen; | |
392 | thisino += XFS_INODES_PER_CHUNK) { | |
393 | if ((error = xfs_inobt_lookup_eq(cur, thisino, | |
394 | XFS_INODES_PER_CHUNK, XFS_INOBT_ALL_FREE, &i))) { | |
395 | xfs_btree_del_cursor(cur, XFS_BTREE_ERROR); | |
396 | return error; | |
397 | } | |
398 | ASSERT(i == 0); | |
4b22a571 | 399 | if ((error = xfs_btree_insert(cur, &i))) { |
1da177e4 LT |
400 | xfs_btree_del_cursor(cur, XFS_BTREE_ERROR); |
401 | return error; | |
402 | } | |
403 | ASSERT(i == 1); | |
404 | } | |
405 | xfs_btree_del_cursor(cur, XFS_BTREE_NOERROR); | |
406 | /* | |
407 | * Log allocation group header fields | |
408 | */ | |
409 | xfs_ialloc_log_agi(tp, agbp, | |
410 | XFS_AGI_COUNT | XFS_AGI_FREECOUNT | XFS_AGI_NEWINO); | |
411 | /* | |
412 | * Modify/log superblock values for inode count and inode free count. | |
413 | */ | |
414 | xfs_trans_mod_sb(tp, XFS_TRANS_SB_ICOUNT, (long)newlen); | |
415 | xfs_trans_mod_sb(tp, XFS_TRANS_SB_IFREE, (long)newlen); | |
416 | *alloc = 1; | |
417 | return 0; | |
418 | } | |
419 | ||
7989cb8e | 420 | STATIC_INLINE xfs_agnumber_t |
1da177e4 LT |
421 | xfs_ialloc_next_ag( |
422 | xfs_mount_t *mp) | |
423 | { | |
424 | xfs_agnumber_t agno; | |
425 | ||
426 | spin_lock(&mp->m_agirotor_lock); | |
427 | agno = mp->m_agirotor; | |
428 | if (++mp->m_agirotor == mp->m_maxagi) | |
429 | mp->m_agirotor = 0; | |
430 | spin_unlock(&mp->m_agirotor_lock); | |
431 | ||
432 | return agno; | |
433 | } | |
434 | ||
435 | /* | |
436 | * Select an allocation group to look for a free inode in, based on the parent | |
437 | * inode and then mode. Return the allocation group buffer. | |
438 | */ | |
439 | STATIC xfs_buf_t * /* allocation group buffer */ | |
440 | xfs_ialloc_ag_select( | |
441 | xfs_trans_t *tp, /* transaction pointer */ | |
442 | xfs_ino_t parent, /* parent directory inode number */ | |
443 | mode_t mode, /* bits set to indicate file type */ | |
444 | int okalloc) /* ok to allocate more space */ | |
445 | { | |
446 | xfs_buf_t *agbp; /* allocation group header buffer */ | |
447 | xfs_agnumber_t agcount; /* number of ag's in the filesystem */ | |
448 | xfs_agnumber_t agno; /* current ag number */ | |
449 | int flags; /* alloc buffer locking flags */ | |
450 | xfs_extlen_t ineed; /* blocks needed for inode allocation */ | |
451 | xfs_extlen_t longest = 0; /* longest extent available */ | |
452 | xfs_mount_t *mp; /* mount point structure */ | |
453 | int needspace; /* file mode implies space allocated */ | |
454 | xfs_perag_t *pag; /* per allocation group data */ | |
455 | xfs_agnumber_t pagno; /* parent (starting) ag number */ | |
456 | ||
457 | /* | |
458 | * Files of these types need at least one block if length > 0 | |
459 | * (and they won't fit in the inode, but that's hard to figure out). | |
460 | */ | |
461 | needspace = S_ISDIR(mode) || S_ISREG(mode) || S_ISLNK(mode); | |
462 | mp = tp->t_mountp; | |
463 | agcount = mp->m_maxagi; | |
464 | if (S_ISDIR(mode)) | |
465 | pagno = xfs_ialloc_next_ag(mp); | |
466 | else { | |
467 | pagno = XFS_INO_TO_AGNO(mp, parent); | |
468 | if (pagno >= agcount) | |
469 | pagno = 0; | |
470 | } | |
471 | ASSERT(pagno < agcount); | |
472 | /* | |
473 | * Loop through allocation groups, looking for one with a little | |
474 | * free space in it. Note we don't look for free inodes, exactly. | |
475 | * Instead, we include whether there is a need to allocate inodes | |
476 | * to mean that blocks must be allocated for them, | |
477 | * if none are currently free. | |
478 | */ | |
479 | agno = pagno; | |
480 | flags = XFS_ALLOC_FLAG_TRYLOCK; | |
481 | down_read(&mp->m_peraglock); | |
482 | for (;;) { | |
483 | pag = &mp->m_perag[agno]; | |
484 | if (!pag->pagi_init) { | |
485 | if (xfs_ialloc_read_agi(mp, tp, agno, &agbp)) { | |
486 | agbp = NULL; | |
487 | goto nextag; | |
488 | } | |
489 | } else | |
490 | agbp = NULL; | |
491 | ||
492 | if (!pag->pagi_inodeok) { | |
493 | xfs_ialloc_next_ag(mp); | |
494 | goto unlock_nextag; | |
495 | } | |
496 | ||
497 | /* | |
498 | * Is there enough free space for the file plus a block | |
499 | * of inodes (if we need to allocate some)? | |
500 | */ | |
501 | ineed = pag->pagi_freecount ? 0 : XFS_IALLOC_BLOCKS(mp); | |
502 | if (ineed && !pag->pagf_init) { | |
503 | if (agbp == NULL && | |
504 | xfs_ialloc_read_agi(mp, tp, agno, &agbp)) { | |
505 | agbp = NULL; | |
506 | goto nextag; | |
507 | } | |
508 | (void)xfs_alloc_pagf_init(mp, tp, agno, flags); | |
509 | } | |
510 | if (!ineed || pag->pagf_init) { | |
511 | if (ineed && !(longest = pag->pagf_longest)) | |
512 | longest = pag->pagf_flcount > 0; | |
513 | if (!ineed || | |
514 | (pag->pagf_freeblks >= needspace + ineed && | |
515 | longest >= ineed && | |
516 | okalloc)) { | |
517 | if (agbp == NULL && | |
518 | xfs_ialloc_read_agi(mp, tp, agno, &agbp)) { | |
519 | agbp = NULL; | |
520 | goto nextag; | |
521 | } | |
522 | up_read(&mp->m_peraglock); | |
523 | return agbp; | |
524 | } | |
525 | } | |
526 | unlock_nextag: | |
527 | if (agbp) | |
528 | xfs_trans_brelse(tp, agbp); | |
529 | nextag: | |
530 | /* | |
531 | * No point in iterating over the rest, if we're shutting | |
532 | * down. | |
533 | */ | |
534 | if (XFS_FORCED_SHUTDOWN(mp)) { | |
535 | up_read(&mp->m_peraglock); | |
1121b219 | 536 | return NULL; |
1da177e4 LT |
537 | } |
538 | agno++; | |
539 | if (agno >= agcount) | |
540 | agno = 0; | |
541 | if (agno == pagno) { | |
542 | if (flags == 0) { | |
543 | up_read(&mp->m_peraglock); | |
1121b219 | 544 | return NULL; |
1da177e4 LT |
545 | } |
546 | flags = 0; | |
547 | } | |
548 | } | |
549 | } | |
550 | ||
551 | /* | |
552 | * Visible inode allocation functions. | |
553 | */ | |
554 | ||
555 | /* | |
556 | * Allocate an inode on disk. | |
557 | * Mode is used to tell whether the new inode will need space, and whether | |
558 | * it is a directory. | |
559 | * | |
560 | * The arguments IO_agbp and alloc_done are defined to work within | |
561 | * the constraint of one allocation per transaction. | |
562 | * xfs_dialloc() is designed to be called twice if it has to do an | |
563 | * allocation to make more free inodes. On the first call, | |
564 | * IO_agbp should be set to NULL. If an inode is available, | |
565 | * i.e., xfs_dialloc() did not need to do an allocation, an inode | |
566 | * number is returned. In this case, IO_agbp would be set to the | |
567 | * current ag_buf and alloc_done set to false. | |
568 | * If an allocation needed to be done, xfs_dialloc would return | |
569 | * the current ag_buf in IO_agbp and set alloc_done to true. | |
570 | * The caller should then commit the current transaction, allocate a new | |
571 | * transaction, and call xfs_dialloc() again, passing in the previous | |
572 | * value of IO_agbp. IO_agbp should be held across the transactions. | |
573 | * Since the agbp is locked across the two calls, the second call is | |
574 | * guaranteed to have a free inode available. | |
575 | * | |
576 | * Once we successfully pick an inode its number is returned and the | |
577 | * on-disk data structures are updated. The inode itself is not read | |
578 | * in, since doing so would break ordering constraints with xfs_reclaim. | |
579 | */ | |
580 | int | |
581 | xfs_dialloc( | |
582 | xfs_trans_t *tp, /* transaction pointer */ | |
583 | xfs_ino_t parent, /* parent inode (directory) */ | |
584 | mode_t mode, /* mode bits for new inode */ | |
585 | int okalloc, /* ok to allocate more space */ | |
586 | xfs_buf_t **IO_agbp, /* in/out ag header's buffer */ | |
587 | boolean_t *alloc_done, /* true if we needed to replenish | |
588 | inode freelist */ | |
589 | xfs_ino_t *inop) /* inode number allocated */ | |
590 | { | |
591 | xfs_agnumber_t agcount; /* number of allocation groups */ | |
592 | xfs_buf_t *agbp; /* allocation group header's buffer */ | |
593 | xfs_agnumber_t agno; /* allocation group number */ | |
594 | xfs_agi_t *agi; /* allocation group header structure */ | |
595 | xfs_btree_cur_t *cur; /* inode allocation btree cursor */ | |
596 | int error; /* error return value */ | |
597 | int i; /* result code */ | |
598 | int ialloced; /* inode allocation status */ | |
599 | int noroom = 0; /* no space for inode blk allocation */ | |
600 | xfs_ino_t ino; /* fs-relative inode to be returned */ | |
601 | /* REFERENCED */ | |
602 | int j; /* result code */ | |
603 | xfs_mount_t *mp; /* file system mount structure */ | |
604 | int offset; /* index of inode in chunk */ | |
605 | xfs_agino_t pagino; /* parent's a.g. relative inode # */ | |
606 | xfs_agnumber_t pagno; /* parent's allocation group number */ | |
61a25848 | 607 | xfs_inobt_rec_incore_t rec; /* inode allocation record */ |
1da177e4 LT |
608 | xfs_agnumber_t tagno; /* testing allocation group number */ |
609 | xfs_btree_cur_t *tcur; /* temp cursor */ | |
61a25848 | 610 | xfs_inobt_rec_incore_t trec; /* temp inode allocation record */ |
1da177e4 LT |
611 | |
612 | ||
613 | if (*IO_agbp == NULL) { | |
614 | /* | |
615 | * We do not have an agbp, so select an initial allocation | |
616 | * group for inode allocation. | |
617 | */ | |
618 | agbp = xfs_ialloc_ag_select(tp, parent, mode, okalloc); | |
619 | /* | |
620 | * Couldn't find an allocation group satisfying the | |
621 | * criteria, give up. | |
622 | */ | |
623 | if (!agbp) { | |
624 | *inop = NULLFSINO; | |
625 | return 0; | |
626 | } | |
627 | agi = XFS_BUF_TO_AGI(agbp); | |
16259e7d | 628 | ASSERT(be32_to_cpu(agi->agi_magicnum) == XFS_AGI_MAGIC); |
1da177e4 LT |
629 | } else { |
630 | /* | |
631 | * Continue where we left off before. In this case, we | |
632 | * know that the allocation group has free inodes. | |
633 | */ | |
634 | agbp = *IO_agbp; | |
635 | agi = XFS_BUF_TO_AGI(agbp); | |
16259e7d CH |
636 | ASSERT(be32_to_cpu(agi->agi_magicnum) == XFS_AGI_MAGIC); |
637 | ASSERT(be32_to_cpu(agi->agi_freecount) > 0); | |
1da177e4 LT |
638 | } |
639 | mp = tp->t_mountp; | |
640 | agcount = mp->m_sb.sb_agcount; | |
16259e7d | 641 | agno = be32_to_cpu(agi->agi_seqno); |
1da177e4 LT |
642 | tagno = agno; |
643 | pagno = XFS_INO_TO_AGNO(mp, parent); | |
644 | pagino = XFS_INO_TO_AGINO(mp, parent); | |
645 | ||
646 | /* | |
647 | * If we have already hit the ceiling of inode blocks then clear | |
648 | * okalloc so we scan all available agi structures for a free | |
649 | * inode. | |
650 | */ | |
651 | ||
652 | if (mp->m_maxicount && | |
653 | mp->m_sb.sb_icount + XFS_IALLOC_INODES(mp) > mp->m_maxicount) { | |
654 | noroom = 1; | |
655 | okalloc = 0; | |
656 | } | |
657 | ||
658 | /* | |
659 | * Loop until we find an allocation group that either has free inodes | |
660 | * or in which we can allocate some inodes. Iterate through the | |
661 | * allocation groups upward, wrapping at the end. | |
662 | */ | |
663 | *alloc_done = B_FALSE; | |
664 | while (!agi->agi_freecount) { | |
665 | /* | |
666 | * Don't do anything if we're not supposed to allocate | |
667 | * any blocks, just go on to the next ag. | |
668 | */ | |
669 | if (okalloc) { | |
670 | /* | |
671 | * Try to allocate some new inodes in the allocation | |
672 | * group. | |
673 | */ | |
674 | if ((error = xfs_ialloc_ag_alloc(tp, agbp, &ialloced))) { | |
675 | xfs_trans_brelse(tp, agbp); | |
676 | if (error == ENOSPC) { | |
677 | *inop = NULLFSINO; | |
678 | return 0; | |
679 | } else | |
680 | return error; | |
681 | } | |
682 | if (ialloced) { | |
683 | /* | |
684 | * We successfully allocated some inodes, return | |
685 | * the current context to the caller so that it | |
686 | * can commit the current transaction and call | |
687 | * us again where we left off. | |
688 | */ | |
16259e7d | 689 | ASSERT(be32_to_cpu(agi->agi_freecount) > 0); |
1da177e4 LT |
690 | *alloc_done = B_TRUE; |
691 | *IO_agbp = agbp; | |
692 | *inop = NULLFSINO; | |
693 | return 0; | |
694 | } | |
695 | } | |
696 | /* | |
697 | * If it failed, give up on this ag. | |
698 | */ | |
699 | xfs_trans_brelse(tp, agbp); | |
700 | /* | |
701 | * Go on to the next ag: get its ag header. | |
702 | */ | |
703 | nextag: | |
704 | if (++tagno == agcount) | |
705 | tagno = 0; | |
706 | if (tagno == agno) { | |
707 | *inop = NULLFSINO; | |
708 | return noroom ? ENOSPC : 0; | |
709 | } | |
710 | down_read(&mp->m_peraglock); | |
711 | if (mp->m_perag[tagno].pagi_inodeok == 0) { | |
712 | up_read(&mp->m_peraglock); | |
713 | goto nextag; | |
714 | } | |
715 | error = xfs_ialloc_read_agi(mp, tp, tagno, &agbp); | |
716 | up_read(&mp->m_peraglock); | |
717 | if (error) | |
718 | goto nextag; | |
719 | agi = XFS_BUF_TO_AGI(agbp); | |
16259e7d | 720 | ASSERT(be32_to_cpu(agi->agi_magicnum) == XFS_AGI_MAGIC); |
1da177e4 LT |
721 | } |
722 | /* | |
723 | * Here with an allocation group that has a free inode. | |
724 | * Reset agno since we may have chosen a new ag in the | |
725 | * loop above. | |
726 | */ | |
727 | agno = tagno; | |
728 | *IO_agbp = NULL; | |
561f7d17 | 729 | cur = xfs_inobt_init_cursor(mp, tp, agbp, be32_to_cpu(agi->agi_seqno)); |
1da177e4 LT |
730 | /* |
731 | * If pagino is 0 (this is the root inode allocation) use newino. | |
732 | * This must work because we've just allocated some. | |
733 | */ | |
734 | if (!pagino) | |
16259e7d | 735 | pagino = be32_to_cpu(agi->agi_newino); |
1da177e4 LT |
736 | #ifdef DEBUG |
737 | if (cur->bc_nlevels == 1) { | |
738 | int freecount = 0; | |
739 | ||
740 | if ((error = xfs_inobt_lookup_ge(cur, 0, 0, 0, &i))) | |
741 | goto error0; | |
742 | XFS_WANT_CORRUPTED_GOTO(i == 1, error0); | |
743 | do { | |
2e287a73 CH |
744 | error = xfs_inobt_get_rec(cur, &rec, &i); |
745 | if (error) | |
1da177e4 LT |
746 | goto error0; |
747 | XFS_WANT_CORRUPTED_GOTO(i == 1, error0); | |
748 | freecount += rec.ir_freecount; | |
637aa50f | 749 | if ((error = xfs_btree_increment(cur, 0, &i))) |
1da177e4 LT |
750 | goto error0; |
751 | } while (i == 1); | |
752 | ||
16259e7d | 753 | ASSERT(freecount == be32_to_cpu(agi->agi_freecount) || |
1da177e4 LT |
754 | XFS_FORCED_SHUTDOWN(mp)); |
755 | } | |
756 | #endif | |
757 | /* | |
758 | * If in the same a.g. as the parent, try to get near the parent. | |
759 | */ | |
760 | if (pagno == agno) { | |
761 | if ((error = xfs_inobt_lookup_le(cur, pagino, 0, 0, &i))) | |
762 | goto error0; | |
763 | if (i != 0 && | |
2e287a73 | 764 | (error = xfs_inobt_get_rec(cur, &rec, &j)) == 0 && |
1da177e4 LT |
765 | j == 1 && |
766 | rec.ir_freecount > 0) { | |
767 | /* | |
768 | * Found a free inode in the same chunk | |
769 | * as parent, done. | |
770 | */ | |
771 | } | |
772 | /* | |
773 | * In the same a.g. as parent, but parent's chunk is full. | |
774 | */ | |
775 | else { | |
776 | int doneleft; /* done, to the left */ | |
777 | int doneright; /* done, to the right */ | |
778 | ||
779 | if (error) | |
780 | goto error0; | |
781 | ASSERT(i == 1); | |
782 | ASSERT(j == 1); | |
783 | /* | |
784 | * Duplicate the cursor, search left & right | |
785 | * simultaneously. | |
786 | */ | |
787 | if ((error = xfs_btree_dup_cursor(cur, &tcur))) | |
788 | goto error0; | |
789 | /* | |
790 | * Search left with tcur, back up 1 record. | |
791 | */ | |
8df4da4a | 792 | if ((error = xfs_btree_decrement(tcur, 0, &i))) |
1da177e4 LT |
793 | goto error1; |
794 | doneleft = !i; | |
795 | if (!doneleft) { | |
2e287a73 CH |
796 | error = xfs_inobt_get_rec(tcur, &trec, &i); |
797 | if (error) | |
1da177e4 LT |
798 | goto error1; |
799 | XFS_WANT_CORRUPTED_GOTO(i == 1, error1); | |
800 | } | |
801 | /* | |
802 | * Search right with cur, go forward 1 record. | |
803 | */ | |
637aa50f | 804 | if ((error = xfs_btree_increment(cur, 0, &i))) |
1da177e4 LT |
805 | goto error1; |
806 | doneright = !i; | |
807 | if (!doneright) { | |
2e287a73 CH |
808 | error = xfs_inobt_get_rec(cur, &rec, &i); |
809 | if (error) | |
1da177e4 LT |
810 | goto error1; |
811 | XFS_WANT_CORRUPTED_GOTO(i == 1, error1); | |
812 | } | |
813 | /* | |
814 | * Loop until we find the closest inode chunk | |
815 | * with a free one. | |
816 | */ | |
817 | while (!doneleft || !doneright) { | |
818 | int useleft; /* using left inode | |
819 | chunk this time */ | |
820 | ||
821 | /* | |
822 | * Figure out which block is closer, | |
823 | * if both are valid. | |
824 | */ | |
825 | if (!doneleft && !doneright) | |
826 | useleft = | |
827 | pagino - | |
828 | (trec.ir_startino + | |
829 | XFS_INODES_PER_CHUNK - 1) < | |
830 | rec.ir_startino - pagino; | |
831 | else | |
832 | useleft = !doneleft; | |
833 | /* | |
834 | * If checking the left, does it have | |
835 | * free inodes? | |
836 | */ | |
837 | if (useleft && trec.ir_freecount) { | |
838 | /* | |
839 | * Yes, set it up as the chunk to use. | |
840 | */ | |
841 | rec = trec; | |
842 | xfs_btree_del_cursor(cur, | |
843 | XFS_BTREE_NOERROR); | |
844 | cur = tcur; | |
845 | break; | |
846 | } | |
847 | /* | |
848 | * If checking the right, does it have | |
849 | * free inodes? | |
850 | */ | |
851 | if (!useleft && rec.ir_freecount) { | |
852 | /* | |
853 | * Yes, it's already set up. | |
854 | */ | |
855 | xfs_btree_del_cursor(tcur, | |
856 | XFS_BTREE_NOERROR); | |
857 | break; | |
858 | } | |
859 | /* | |
860 | * If used the left, get another one | |
861 | * further left. | |
862 | */ | |
863 | if (useleft) { | |
8df4da4a | 864 | if ((error = xfs_btree_decrement(tcur, 0, |
1da177e4 LT |
865 | &i))) |
866 | goto error1; | |
867 | doneleft = !i; | |
868 | if (!doneleft) { | |
2e287a73 CH |
869 | error = xfs_inobt_get_rec( |
870 | tcur, &trec, &i); | |
871 | if (error) | |
1da177e4 LT |
872 | goto error1; |
873 | XFS_WANT_CORRUPTED_GOTO(i == 1, | |
874 | error1); | |
875 | } | |
876 | } | |
877 | /* | |
878 | * If used the right, get another one | |
879 | * further right. | |
880 | */ | |
881 | else { | |
637aa50f | 882 | if ((error = xfs_btree_increment(cur, 0, |
1da177e4 LT |
883 | &i))) |
884 | goto error1; | |
885 | doneright = !i; | |
886 | if (!doneright) { | |
2e287a73 CH |
887 | error = xfs_inobt_get_rec( |
888 | cur, &rec, &i); | |
889 | if (error) | |
1da177e4 LT |
890 | goto error1; |
891 | XFS_WANT_CORRUPTED_GOTO(i == 1, | |
892 | error1); | |
893 | } | |
894 | } | |
895 | } | |
896 | ASSERT(!doneleft || !doneright); | |
897 | } | |
898 | } | |
899 | /* | |
900 | * In a different a.g. from the parent. | |
901 | * See if the most recently allocated block has any free. | |
902 | */ | |
16259e7d | 903 | else if (be32_to_cpu(agi->agi_newino) != NULLAGINO) { |
1da177e4 | 904 | if ((error = xfs_inobt_lookup_eq(cur, |
16259e7d | 905 | be32_to_cpu(agi->agi_newino), 0, 0, &i))) |
1da177e4 LT |
906 | goto error0; |
907 | if (i == 1 && | |
2e287a73 | 908 | (error = xfs_inobt_get_rec(cur, &rec, &j)) == 0 && |
1da177e4 LT |
909 | j == 1 && |
910 | rec.ir_freecount > 0) { | |
911 | /* | |
912 | * The last chunk allocated in the group still has | |
913 | * a free inode. | |
914 | */ | |
915 | } | |
916 | /* | |
917 | * None left in the last group, search the whole a.g. | |
918 | */ | |
919 | else { | |
920 | if (error) | |
921 | goto error0; | |
922 | if ((error = xfs_inobt_lookup_ge(cur, 0, 0, 0, &i))) | |
923 | goto error0; | |
924 | ASSERT(i == 1); | |
925 | for (;;) { | |
2e287a73 CH |
926 | error = xfs_inobt_get_rec(cur, &rec, &i); |
927 | if (error) | |
1da177e4 LT |
928 | goto error0; |
929 | XFS_WANT_CORRUPTED_GOTO(i == 1, error0); | |
930 | if (rec.ir_freecount > 0) | |
931 | break; | |
637aa50f | 932 | if ((error = xfs_btree_increment(cur, 0, &i))) |
1da177e4 LT |
933 | goto error0; |
934 | XFS_WANT_CORRUPTED_GOTO(i == 1, error0); | |
935 | } | |
936 | } | |
937 | } | |
9d87c319 | 938 | offset = xfs_ialloc_find_free(&rec.ir_free); |
1da177e4 LT |
939 | ASSERT(offset >= 0); |
940 | ASSERT(offset < XFS_INODES_PER_CHUNK); | |
941 | ASSERT((XFS_AGINO_TO_OFFSET(mp, rec.ir_startino) % | |
942 | XFS_INODES_PER_CHUNK) == 0); | |
943 | ino = XFS_AGINO_TO_INO(mp, agno, rec.ir_startino + offset); | |
0d87e656 | 944 | rec.ir_free &= ~XFS_INOBT_MASK(offset); |
1da177e4 | 945 | rec.ir_freecount--; |
afabc24a CH |
946 | error = xfs_inobt_update(cur, &rec); |
947 | if (error) | |
1da177e4 | 948 | goto error0; |
413d57c9 | 949 | be32_add_cpu(&agi->agi_freecount, -1); |
1da177e4 LT |
950 | xfs_ialloc_log_agi(tp, agbp, XFS_AGI_FREECOUNT); |
951 | down_read(&mp->m_peraglock); | |
952 | mp->m_perag[tagno].pagi_freecount--; | |
953 | up_read(&mp->m_peraglock); | |
954 | #ifdef DEBUG | |
955 | if (cur->bc_nlevels == 1) { | |
956 | int freecount = 0; | |
957 | ||
958 | if ((error = xfs_inobt_lookup_ge(cur, 0, 0, 0, &i))) | |
959 | goto error0; | |
960 | do { | |
2e287a73 CH |
961 | error = xfs_inobt_get_rec(cur, &rec, &i); |
962 | if (error) | |
1da177e4 LT |
963 | goto error0; |
964 | XFS_WANT_CORRUPTED_GOTO(i == 1, error0); | |
965 | freecount += rec.ir_freecount; | |
637aa50f | 966 | if ((error = xfs_btree_increment(cur, 0, &i))) |
1da177e4 LT |
967 | goto error0; |
968 | } while (i == 1); | |
16259e7d | 969 | ASSERT(freecount == be32_to_cpu(agi->agi_freecount) || |
1da177e4 LT |
970 | XFS_FORCED_SHUTDOWN(mp)); |
971 | } | |
972 | #endif | |
973 | xfs_btree_del_cursor(cur, XFS_BTREE_NOERROR); | |
974 | xfs_trans_mod_sb(tp, XFS_TRANS_SB_IFREE, -1); | |
975 | *inop = ino; | |
976 | return 0; | |
977 | error1: | |
978 | xfs_btree_del_cursor(tcur, XFS_BTREE_ERROR); | |
979 | error0: | |
980 | xfs_btree_del_cursor(cur, XFS_BTREE_ERROR); | |
981 | return error; | |
982 | } | |
983 | ||
984 | /* | |
985 | * Free disk inode. Carefully avoids touching the incore inode, all | |
986 | * manipulations incore are the caller's responsibility. | |
987 | * The on-disk inode is not changed by this operation, only the | |
988 | * btree (free inode mask) is changed. | |
989 | */ | |
990 | int | |
991 | xfs_difree( | |
992 | xfs_trans_t *tp, /* transaction pointer */ | |
993 | xfs_ino_t inode, /* inode to be freed */ | |
994 | xfs_bmap_free_t *flist, /* extents to free */ | |
995 | int *delete, /* set if inode cluster was deleted */ | |
996 | xfs_ino_t *first_ino) /* first inode in deleted cluster */ | |
997 | { | |
998 | /* REFERENCED */ | |
999 | xfs_agblock_t agbno; /* block number containing inode */ | |
1000 | xfs_buf_t *agbp; /* buffer containing allocation group header */ | |
1001 | xfs_agino_t agino; /* inode number relative to allocation group */ | |
1002 | xfs_agnumber_t agno; /* allocation group number */ | |
1003 | xfs_agi_t *agi; /* allocation group header */ | |
1004 | xfs_btree_cur_t *cur; /* inode btree cursor */ | |
1005 | int error; /* error return value */ | |
1006 | int i; /* result code */ | |
1007 | int ilen; /* inodes in an inode cluster */ | |
1008 | xfs_mount_t *mp; /* mount structure for filesystem */ | |
1009 | int off; /* offset of inode in inode chunk */ | |
61a25848 | 1010 | xfs_inobt_rec_incore_t rec; /* btree record */ |
1da177e4 LT |
1011 | |
1012 | mp = tp->t_mountp; | |
1013 | ||
1014 | /* | |
1015 | * Break up inode number into its components. | |
1016 | */ | |
1017 | agno = XFS_INO_TO_AGNO(mp, inode); | |
1018 | if (agno >= mp->m_sb.sb_agcount) { | |
1019 | cmn_err(CE_WARN, | |
1020 | "xfs_difree: agno >= mp->m_sb.sb_agcount (%d >= %d) on %s. Returning EINVAL.", | |
1021 | agno, mp->m_sb.sb_agcount, mp->m_fsname); | |
1022 | ASSERT(0); | |
1023 | return XFS_ERROR(EINVAL); | |
1024 | } | |
1025 | agino = XFS_INO_TO_AGINO(mp, inode); | |
1026 | if (inode != XFS_AGINO_TO_INO(mp, agno, agino)) { | |
1027 | cmn_err(CE_WARN, | |
da1650a5 CH |
1028 | "xfs_difree: inode != XFS_AGINO_TO_INO() " |
1029 | "(%llu != %llu) on %s. Returning EINVAL.", | |
1030 | (unsigned long long)inode, | |
1031 | (unsigned long long)XFS_AGINO_TO_INO(mp, agno, agino), | |
1032 | mp->m_fsname); | |
1da177e4 LT |
1033 | ASSERT(0); |
1034 | return XFS_ERROR(EINVAL); | |
1035 | } | |
1036 | agbno = XFS_AGINO_TO_AGBNO(mp, agino); | |
1037 | if (agbno >= mp->m_sb.sb_agblocks) { | |
1038 | cmn_err(CE_WARN, | |
1039 | "xfs_difree: agbno >= mp->m_sb.sb_agblocks (%d >= %d) on %s. Returning EINVAL.", | |
1040 | agbno, mp->m_sb.sb_agblocks, mp->m_fsname); | |
1041 | ASSERT(0); | |
1042 | return XFS_ERROR(EINVAL); | |
1043 | } | |
1044 | /* | |
1045 | * Get the allocation group header. | |
1046 | */ | |
1047 | down_read(&mp->m_peraglock); | |
1048 | error = xfs_ialloc_read_agi(mp, tp, agno, &agbp); | |
1049 | up_read(&mp->m_peraglock); | |
1050 | if (error) { | |
1051 | cmn_err(CE_WARN, | |
1052 | "xfs_difree: xfs_ialloc_read_agi() returned an error %d on %s. Returning error.", | |
1053 | error, mp->m_fsname); | |
1054 | return error; | |
1055 | } | |
1056 | agi = XFS_BUF_TO_AGI(agbp); | |
16259e7d CH |
1057 | ASSERT(be32_to_cpu(agi->agi_magicnum) == XFS_AGI_MAGIC); |
1058 | ASSERT(agbno < be32_to_cpu(agi->agi_length)); | |
1da177e4 LT |
1059 | /* |
1060 | * Initialize the cursor. | |
1061 | */ | |
561f7d17 | 1062 | cur = xfs_inobt_init_cursor(mp, tp, agbp, agno); |
1da177e4 LT |
1063 | #ifdef DEBUG |
1064 | if (cur->bc_nlevels == 1) { | |
1065 | int freecount = 0; | |
1066 | ||
1067 | if ((error = xfs_inobt_lookup_ge(cur, 0, 0, 0, &i))) | |
1068 | goto error0; | |
1069 | do { | |
2e287a73 CH |
1070 | error = xfs_inobt_get_rec(cur, &rec, &i); |
1071 | if (error) | |
1da177e4 LT |
1072 | goto error0; |
1073 | if (i) { | |
1074 | freecount += rec.ir_freecount; | |
637aa50f | 1075 | if ((error = xfs_btree_increment(cur, 0, &i))) |
1da177e4 LT |
1076 | goto error0; |
1077 | } | |
1078 | } while (i == 1); | |
16259e7d | 1079 | ASSERT(freecount == be32_to_cpu(agi->agi_freecount) || |
1da177e4 LT |
1080 | XFS_FORCED_SHUTDOWN(mp)); |
1081 | } | |
1082 | #endif | |
1083 | /* | |
1084 | * Look for the entry describing this inode. | |
1085 | */ | |
1086 | if ((error = xfs_inobt_lookup_le(cur, agino, 0, 0, &i))) { | |
1087 | cmn_err(CE_WARN, | |
1088 | "xfs_difree: xfs_inobt_lookup_le returned() an error %d on %s. Returning error.", | |
1089 | error, mp->m_fsname); | |
1090 | goto error0; | |
1091 | } | |
1092 | XFS_WANT_CORRUPTED_GOTO(i == 1, error0); | |
2e287a73 CH |
1093 | error = xfs_inobt_get_rec(cur, &rec, &i); |
1094 | if (error) { | |
1da177e4 LT |
1095 | cmn_err(CE_WARN, |
1096 | "xfs_difree: xfs_inobt_get_rec() returned an error %d on %s. Returning error.", | |
1097 | error, mp->m_fsname); | |
1098 | goto error0; | |
1099 | } | |
1100 | XFS_WANT_CORRUPTED_GOTO(i == 1, error0); | |
1101 | /* | |
1102 | * Get the offset in the inode chunk. | |
1103 | */ | |
1104 | off = agino - rec.ir_startino; | |
1105 | ASSERT(off >= 0 && off < XFS_INODES_PER_CHUNK); | |
0d87e656 | 1106 | ASSERT(!(rec.ir_free & XFS_INOBT_MASK(off))); |
1da177e4 LT |
1107 | /* |
1108 | * Mark the inode free & increment the count. | |
1109 | */ | |
0d87e656 | 1110 | rec.ir_free |= XFS_INOBT_MASK(off); |
1da177e4 LT |
1111 | rec.ir_freecount++; |
1112 | ||
1113 | /* | |
c41564b5 | 1114 | * When an inode cluster is free, it becomes eligible for removal |
1da177e4 | 1115 | */ |
1bd960ee | 1116 | if (!(mp->m_flags & XFS_MOUNT_IKEEP) && |
1da177e4 LT |
1117 | (rec.ir_freecount == XFS_IALLOC_INODES(mp))) { |
1118 | ||
1119 | *delete = 1; | |
1120 | *first_ino = XFS_AGINO_TO_INO(mp, agno, rec.ir_startino); | |
1121 | ||
1122 | /* | |
1123 | * Remove the inode cluster from the AGI B+Tree, adjust the | |
1124 | * AGI and Superblock inode counts, and mark the disk space | |
1125 | * to be freed when the transaction is committed. | |
1126 | */ | |
1127 | ilen = XFS_IALLOC_INODES(mp); | |
413d57c9 MS |
1128 | be32_add_cpu(&agi->agi_count, -ilen); |
1129 | be32_add_cpu(&agi->agi_freecount, -(ilen - 1)); | |
1da177e4 LT |
1130 | xfs_ialloc_log_agi(tp, agbp, XFS_AGI_COUNT | XFS_AGI_FREECOUNT); |
1131 | down_read(&mp->m_peraglock); | |
1132 | mp->m_perag[agno].pagi_freecount -= ilen - 1; | |
1133 | up_read(&mp->m_peraglock); | |
1134 | xfs_trans_mod_sb(tp, XFS_TRANS_SB_ICOUNT, -ilen); | |
1135 | xfs_trans_mod_sb(tp, XFS_TRANS_SB_IFREE, -(ilen - 1)); | |
1136 | ||
91cca5df CH |
1137 | if ((error = xfs_btree_delete(cur, &i))) { |
1138 | cmn_err(CE_WARN, "xfs_difree: xfs_btree_delete returned an error %d on %s.\n", | |
1da177e4 LT |
1139 | error, mp->m_fsname); |
1140 | goto error0; | |
1141 | } | |
1142 | ||
1143 | xfs_bmap_add_free(XFS_AGB_TO_FSB(mp, | |
1144 | agno, XFS_INO_TO_AGBNO(mp,rec.ir_startino)), | |
1145 | XFS_IALLOC_BLOCKS(mp), flist, mp); | |
1146 | } else { | |
1147 | *delete = 0; | |
1148 | ||
afabc24a CH |
1149 | error = xfs_inobt_update(cur, &rec); |
1150 | if (error) { | |
1da177e4 | 1151 | cmn_err(CE_WARN, |
afabc24a | 1152 | "xfs_difree: xfs_inobt_update returned an error %d on %s.", |
1da177e4 LT |
1153 | error, mp->m_fsname); |
1154 | goto error0; | |
1155 | } | |
afabc24a | 1156 | |
1da177e4 LT |
1157 | /* |
1158 | * Change the inode free counts and log the ag/sb changes. | |
1159 | */ | |
413d57c9 | 1160 | be32_add_cpu(&agi->agi_freecount, 1); |
1da177e4 LT |
1161 | xfs_ialloc_log_agi(tp, agbp, XFS_AGI_FREECOUNT); |
1162 | down_read(&mp->m_peraglock); | |
1163 | mp->m_perag[agno].pagi_freecount++; | |
1164 | up_read(&mp->m_peraglock); | |
1165 | xfs_trans_mod_sb(tp, XFS_TRANS_SB_IFREE, 1); | |
1166 | } | |
1167 | ||
1168 | #ifdef DEBUG | |
1169 | if (cur->bc_nlevels == 1) { | |
1170 | int freecount = 0; | |
1171 | ||
1172 | if ((error = xfs_inobt_lookup_ge(cur, 0, 0, 0, &i))) | |
1173 | goto error0; | |
1174 | do { | |
2e287a73 CH |
1175 | error = xfs_inobt_get_rec(cur, &rec, &i); |
1176 | if (error) | |
1da177e4 LT |
1177 | goto error0; |
1178 | if (i) { | |
1179 | freecount += rec.ir_freecount; | |
637aa50f | 1180 | if ((error = xfs_btree_increment(cur, 0, &i))) |
1da177e4 LT |
1181 | goto error0; |
1182 | } | |
1183 | } while (i == 1); | |
16259e7d | 1184 | ASSERT(freecount == be32_to_cpu(agi->agi_freecount) || |
1da177e4 LT |
1185 | XFS_FORCED_SHUTDOWN(mp)); |
1186 | } | |
1187 | #endif | |
1188 | xfs_btree_del_cursor(cur, XFS_BTREE_NOERROR); | |
1189 | return 0; | |
1190 | ||
1191 | error0: | |
1192 | xfs_btree_del_cursor(cur, XFS_BTREE_ERROR); | |
1193 | return error; | |
1194 | } | |
1195 | ||
1196 | /* | |
94e1b69d | 1197 | * Return the location of the inode in imap, for mapping it into a buffer. |
1da177e4 | 1198 | */ |
1da177e4 | 1199 | int |
94e1b69d CH |
1200 | xfs_imap( |
1201 | xfs_mount_t *mp, /* file system mount structure */ | |
1202 | xfs_trans_t *tp, /* transaction pointer */ | |
1da177e4 | 1203 | xfs_ino_t ino, /* inode to locate */ |
94e1b69d CH |
1204 | struct xfs_imap *imap, /* location map structure */ |
1205 | uint flags) /* flags for inode btree lookup */ | |
1da177e4 LT |
1206 | { |
1207 | xfs_agblock_t agbno; /* block number of inode in the alloc group */ | |
1da177e4 LT |
1208 | xfs_agino_t agino; /* inode number within alloc group */ |
1209 | xfs_agnumber_t agno; /* allocation group number */ | |
1210 | int blks_per_cluster; /* num blocks per inode cluster */ | |
1211 | xfs_agblock_t chunk_agbno; /* first block in inode chunk */ | |
1da177e4 | 1212 | xfs_agblock_t cluster_agbno; /* first block in inode cluster */ |
1da177e4 | 1213 | int error; /* error code */ |
1da177e4 LT |
1214 | int offset; /* index of inode in its buffer */ |
1215 | int offset_agbno; /* blks from chunk start to inode */ | |
1216 | ||
1217 | ASSERT(ino != NULLFSINO); | |
94e1b69d | 1218 | |
1da177e4 LT |
1219 | /* |
1220 | * Split up the inode number into its parts. | |
1221 | */ | |
1222 | agno = XFS_INO_TO_AGNO(mp, ino); | |
1223 | agino = XFS_INO_TO_AGINO(mp, ino); | |
1224 | agbno = XFS_AGINO_TO_AGBNO(mp, agino); | |
1225 | if (agno >= mp->m_sb.sb_agcount || agbno >= mp->m_sb.sb_agblocks || | |
1226 | ino != XFS_AGINO_TO_INO(mp, agno, agino)) { | |
1227 | #ifdef DEBUG | |
4d1a2ed3 | 1228 | /* no diagnostics for bulkstat, ino comes from userspace */ |
b48d8d64 | 1229 | if (flags & XFS_IGET_BULKSTAT) |
4d1a2ed3 | 1230 | return XFS_ERROR(EINVAL); |
1da177e4 LT |
1231 | if (agno >= mp->m_sb.sb_agcount) { |
1232 | xfs_fs_cmn_err(CE_ALERT, mp, | |
94e1b69d | 1233 | "xfs_imap: agno (%d) >= " |
1da177e4 LT |
1234 | "mp->m_sb.sb_agcount (%d)", |
1235 | agno, mp->m_sb.sb_agcount); | |
1236 | } | |
1237 | if (agbno >= mp->m_sb.sb_agblocks) { | |
1238 | xfs_fs_cmn_err(CE_ALERT, mp, | |
94e1b69d | 1239 | "xfs_imap: agbno (0x%llx) >= " |
1da177e4 LT |
1240 | "mp->m_sb.sb_agblocks (0x%lx)", |
1241 | (unsigned long long) agbno, | |
1242 | (unsigned long) mp->m_sb.sb_agblocks); | |
1243 | } | |
1244 | if (ino != XFS_AGINO_TO_INO(mp, agno, agino)) { | |
1245 | xfs_fs_cmn_err(CE_ALERT, mp, | |
94e1b69d | 1246 | "xfs_imap: ino (0x%llx) != " |
1da177e4 LT |
1247 | "XFS_AGINO_TO_INO(mp, agno, agino) " |
1248 | "(0x%llx)", | |
1249 | ino, XFS_AGINO_TO_INO(mp, agno, agino)); | |
1250 | } | |
745b1f47 | 1251 | xfs_stack_trace(); |
1da177e4 LT |
1252 | #endif /* DEBUG */ |
1253 | return XFS_ERROR(EINVAL); | |
1254 | } | |
94e1b69d CH |
1255 | |
1256 | /* | |
1257 | * If the inode cluster size is the same as the blocksize or | |
1258 | * smaller we get to the buffer by simple arithmetics. | |
1259 | */ | |
1260 | if (XFS_INODE_CLUSTER_SIZE(mp) <= mp->m_sb.sb_blocksize) { | |
1da177e4 LT |
1261 | offset = XFS_INO_TO_OFFSET(mp, ino); |
1262 | ASSERT(offset < mp->m_sb.sb_inopblock); | |
94e1b69d CH |
1263 | |
1264 | imap->im_blkno = XFS_AGB_TO_DADDR(mp, agno, agbno); | |
1265 | imap->im_len = XFS_FSB_TO_BB(mp, 1); | |
1266 | imap->im_boffset = (ushort)(offset << mp->m_sb.sb_inodelog); | |
1da177e4 LT |
1267 | return 0; |
1268 | } | |
94e1b69d | 1269 | |
1da177e4 | 1270 | blks_per_cluster = XFS_INODE_CLUSTER_SIZE(mp) >> mp->m_sb.sb_blocklog; |
94e1b69d CH |
1271 | |
1272 | /* | |
1273 | * If we get a block number passed from bulkstat we can use it to | |
1274 | * find the buffer easily. | |
1275 | */ | |
1276 | if (imap->im_blkno) { | |
1da177e4 LT |
1277 | offset = XFS_INO_TO_OFFSET(mp, ino); |
1278 | ASSERT(offset < mp->m_sb.sb_inopblock); | |
94e1b69d | 1279 | |
9d87c319 | 1280 | cluster_agbno = xfs_daddr_to_agbno(mp, imap->im_blkno); |
94e1b69d CH |
1281 | offset += (agbno - cluster_agbno) * mp->m_sb.sb_inopblock; |
1282 | ||
1283 | imap->im_len = XFS_FSB_TO_BB(mp, blks_per_cluster); | |
1284 | imap->im_boffset = (ushort)(offset << mp->m_sb.sb_inodelog); | |
1da177e4 LT |
1285 | return 0; |
1286 | } | |
94e1b69d CH |
1287 | |
1288 | /* | |
1289 | * If the inode chunks are aligned then use simple maths to | |
1290 | * find the location. Otherwise we have to do a btree | |
1291 | * lookup to find the location. | |
1292 | */ | |
1da177e4 LT |
1293 | if (mp->m_inoalign_mask) { |
1294 | offset_agbno = agbno & mp->m_inoalign_mask; | |
1295 | chunk_agbno = agbno - offset_agbno; | |
1296 | } else { | |
94e1b69d | 1297 | xfs_btree_cur_t *cur; /* inode btree cursor */ |
2e287a73 | 1298 | xfs_inobt_rec_incore_t chunk_rec; |
94e1b69d CH |
1299 | xfs_buf_t *agbp; /* agi buffer */ |
1300 | int i; /* temp state */ | |
1301 | ||
1da177e4 LT |
1302 | down_read(&mp->m_peraglock); |
1303 | error = xfs_ialloc_read_agi(mp, tp, agno, &agbp); | |
1304 | up_read(&mp->m_peraglock); | |
1305 | if (error) { | |
94e1b69d | 1306 | xfs_fs_cmn_err(CE_ALERT, mp, "xfs_imap: " |
1da177e4 LT |
1307 | "xfs_ialloc_read_agi() returned " |
1308 | "error %d, agno %d", | |
1309 | error, agno); | |
1da177e4 LT |
1310 | return error; |
1311 | } | |
94e1b69d | 1312 | |
561f7d17 | 1313 | cur = xfs_inobt_init_cursor(mp, tp, agbp, agno); |
94e1b69d CH |
1314 | error = xfs_inobt_lookup_le(cur, agino, 0, 0, &i); |
1315 | if (error) { | |
1316 | xfs_fs_cmn_err(CE_ALERT, mp, "xfs_imap: " | |
1da177e4 | 1317 | "xfs_inobt_lookup_le() failed"); |
1da177e4 LT |
1318 | goto error0; |
1319 | } | |
94e1b69d | 1320 | |
2e287a73 | 1321 | error = xfs_inobt_get_rec(cur, &chunk_rec, &i); |
94e1b69d CH |
1322 | if (error) { |
1323 | xfs_fs_cmn_err(CE_ALERT, mp, "xfs_imap: " | |
1da177e4 | 1324 | "xfs_inobt_get_rec() failed"); |
1da177e4 LT |
1325 | goto error0; |
1326 | } | |
1327 | if (i == 0) { | |
1328 | #ifdef DEBUG | |
94e1b69d | 1329 | xfs_fs_cmn_err(CE_ALERT, mp, "xfs_imap: " |
1da177e4 LT |
1330 | "xfs_inobt_get_rec() failed"); |
1331 | #endif /* DEBUG */ | |
1332 | error = XFS_ERROR(EINVAL); | |
1333 | } | |
94e1b69d | 1334 | error0: |
1da177e4 LT |
1335 | xfs_trans_brelse(tp, agbp); |
1336 | xfs_btree_del_cursor(cur, XFS_BTREE_NOERROR); | |
1337 | if (error) | |
1338 | return error; | |
2e287a73 | 1339 | chunk_agbno = XFS_AGINO_TO_AGBNO(mp, chunk_rec.ir_startino); |
1da177e4 LT |
1340 | offset_agbno = agbno - chunk_agbno; |
1341 | } | |
94e1b69d | 1342 | |
1da177e4 LT |
1343 | ASSERT(agbno >= chunk_agbno); |
1344 | cluster_agbno = chunk_agbno + | |
1345 | ((offset_agbno / blks_per_cluster) * blks_per_cluster); | |
1346 | offset = ((agbno - cluster_agbno) * mp->m_sb.sb_inopblock) + | |
1347 | XFS_INO_TO_OFFSET(mp, ino); | |
94e1b69d CH |
1348 | |
1349 | imap->im_blkno = XFS_AGB_TO_DADDR(mp, agno, cluster_agbno); | |
1350 | imap->im_len = XFS_FSB_TO_BB(mp, blks_per_cluster); | |
1351 | imap->im_boffset = (ushort)(offset << mp->m_sb.sb_inodelog); | |
1352 | ||
1353 | /* | |
1354 | * If the inode number maps to a block outside the bounds | |
1355 | * of the file system then return NULL rather than calling | |
1356 | * read_buf and panicing when we get an error from the | |
1357 | * driver. | |
1358 | */ | |
1359 | if ((imap->im_blkno + imap->im_len) > | |
1360 | XFS_FSB_TO_BB(mp, mp->m_sb.sb_dblocks)) { | |
1361 | xfs_fs_cmn_err(CE_ALERT, mp, "xfs_imap: " | |
1362 | "(imap->im_blkno (0x%llx) + imap->im_len (0x%llx)) > " | |
1363 | " XFS_FSB_TO_BB(mp, mp->m_sb.sb_dblocks) (0x%llx)", | |
1364 | (unsigned long long) imap->im_blkno, | |
1365 | (unsigned long long) imap->im_len, | |
1366 | XFS_FSB_TO_BB(mp, mp->m_sb.sb_dblocks)); | |
1367 | return XFS_ERROR(EINVAL); | |
1368 | } | |
1369 | ||
1da177e4 | 1370 | return 0; |
1da177e4 LT |
1371 | } |
1372 | ||
1373 | /* | |
1374 | * Compute and fill in value of m_in_maxlevels. | |
1375 | */ | |
1376 | void | |
1377 | xfs_ialloc_compute_maxlevels( | |
1378 | xfs_mount_t *mp) /* file system mount structure */ | |
1379 | { | |
1380 | int level; | |
1381 | uint maxblocks; | |
1382 | uint maxleafents; | |
1383 | int minleafrecs; | |
1384 | int minnoderecs; | |
1385 | ||
1386 | maxleafents = (1LL << XFS_INO_AGINO_BITS(mp)) >> | |
1387 | XFS_INODES_PER_CHUNK_LOG; | |
1388 | minleafrecs = mp->m_alloc_mnr[0]; | |
1389 | minnoderecs = mp->m_alloc_mnr[1]; | |
1390 | maxblocks = (maxleafents + minleafrecs - 1) / minleafrecs; | |
1391 | for (level = 1; maxblocks > 1; level++) | |
1392 | maxblocks = (maxblocks + minnoderecs - 1) / minnoderecs; | |
1393 | mp->m_in_maxlevels = level; | |
1394 | } | |
1395 | ||
1396 | /* | |
1397 | * Log specified fields for the ag hdr (inode section) | |
1398 | */ | |
1399 | void | |
1400 | xfs_ialloc_log_agi( | |
1401 | xfs_trans_t *tp, /* transaction pointer */ | |
1402 | xfs_buf_t *bp, /* allocation group header buffer */ | |
1403 | int fields) /* bitmask of fields to log */ | |
1404 | { | |
1405 | int first; /* first byte number */ | |
1406 | int last; /* last byte number */ | |
1407 | static const short offsets[] = { /* field starting offsets */ | |
1408 | /* keep in sync with bit definitions */ | |
1409 | offsetof(xfs_agi_t, agi_magicnum), | |
1410 | offsetof(xfs_agi_t, agi_versionnum), | |
1411 | offsetof(xfs_agi_t, agi_seqno), | |
1412 | offsetof(xfs_agi_t, agi_length), | |
1413 | offsetof(xfs_agi_t, agi_count), | |
1414 | offsetof(xfs_agi_t, agi_root), | |
1415 | offsetof(xfs_agi_t, agi_level), | |
1416 | offsetof(xfs_agi_t, agi_freecount), | |
1417 | offsetof(xfs_agi_t, agi_newino), | |
1418 | offsetof(xfs_agi_t, agi_dirino), | |
1419 | offsetof(xfs_agi_t, agi_unlinked), | |
1420 | sizeof(xfs_agi_t) | |
1421 | }; | |
1422 | #ifdef DEBUG | |
1423 | xfs_agi_t *agi; /* allocation group header */ | |
1424 | ||
1425 | agi = XFS_BUF_TO_AGI(bp); | |
16259e7d | 1426 | ASSERT(be32_to_cpu(agi->agi_magicnum) == XFS_AGI_MAGIC); |
1da177e4 LT |
1427 | #endif |
1428 | /* | |
1429 | * Compute byte offsets for the first and last fields. | |
1430 | */ | |
1431 | xfs_btree_offsets(fields, offsets, XFS_AGI_NUM_BITS, &first, &last); | |
1432 | /* | |
1433 | * Log the allocation group inode header buffer. | |
1434 | */ | |
1435 | xfs_trans_log_buf(tp, bp, first, last); | |
1436 | } | |
1437 | ||
5e1be0fb CH |
1438 | #ifdef DEBUG |
1439 | STATIC void | |
1440 | xfs_check_agi_unlinked( | |
1441 | struct xfs_agi *agi) | |
1442 | { | |
1443 | int i; | |
1444 | ||
1445 | for (i = 0; i < XFS_AGI_UNLINKED_BUCKETS; i++) | |
1446 | ASSERT(agi->agi_unlinked[i]); | |
1447 | } | |
1448 | #else | |
1449 | #define xfs_check_agi_unlinked(agi) | |
1450 | #endif | |
1451 | ||
1da177e4 LT |
1452 | /* |
1453 | * Read in the allocation group header (inode allocation section) | |
1454 | */ | |
1455 | int | |
5e1be0fb CH |
1456 | xfs_read_agi( |
1457 | struct xfs_mount *mp, /* file system mount structure */ | |
1458 | struct xfs_trans *tp, /* transaction pointer */ | |
1459 | xfs_agnumber_t agno, /* allocation group number */ | |
1460 | struct xfs_buf **bpp) /* allocation group hdr buf */ | |
1da177e4 | 1461 | { |
5e1be0fb CH |
1462 | struct xfs_agi *agi; /* allocation group header */ |
1463 | int agi_ok; /* agi is consistent */ | |
1464 | int error; | |
1da177e4 LT |
1465 | |
1466 | ASSERT(agno != NULLAGNUMBER); | |
5e1be0fb CH |
1467 | |
1468 | error = xfs_trans_read_buf(mp, tp, mp->m_ddev_targp, | |
1da177e4 | 1469 | XFS_AG_DADDR(mp, agno, XFS_AGI_DADDR(mp)), |
5e1be0fb | 1470 | XFS_FSS_TO_BB(mp, 1), 0, bpp); |
1da177e4 LT |
1471 | if (error) |
1472 | return error; | |
5e1be0fb CH |
1473 | |
1474 | ASSERT(*bpp && !XFS_BUF_GETERROR(*bpp)); | |
1475 | agi = XFS_BUF_TO_AGI(*bpp); | |
1da177e4 LT |
1476 | |
1477 | /* | |
1478 | * Validate the magic number of the agi block. | |
1479 | */ | |
5e1be0fb CH |
1480 | agi_ok = be32_to_cpu(agi->agi_magicnum) == XFS_AGI_MAGIC && |
1481 | XFS_AGI_GOOD_VERSION(be32_to_cpu(agi->agi_versionnum)) && | |
1482 | be32_to_cpu(agi->agi_seqno) == agno; | |
1da177e4 LT |
1483 | if (unlikely(XFS_TEST_ERROR(!agi_ok, mp, XFS_ERRTAG_IALLOC_READ_AGI, |
1484 | XFS_RANDOM_IALLOC_READ_AGI))) { | |
5e1be0fb | 1485 | XFS_CORRUPTION_ERROR("xfs_read_agi", XFS_ERRLEVEL_LOW, |
1da177e4 | 1486 | mp, agi); |
5e1be0fb | 1487 | xfs_trans_brelse(tp, *bpp); |
1da177e4 LT |
1488 | return XFS_ERROR(EFSCORRUPTED); |
1489 | } | |
5e1be0fb CH |
1490 | |
1491 | XFS_BUF_SET_VTYPE_REF(*bpp, B_FS_AGI, XFS_AGI_REF); | |
1492 | ||
1493 | xfs_check_agi_unlinked(agi); | |
1494 | return 0; | |
1495 | } | |
1496 | ||
1497 | int | |
1498 | xfs_ialloc_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 */ | |
1503 | { | |
1504 | struct xfs_agi *agi; /* allocation group header */ | |
1505 | struct xfs_perag *pag; /* per allocation group data */ | |
1506 | int error; | |
1507 | ||
1508 | error = xfs_read_agi(mp, tp, agno, bpp); | |
1509 | if (error) | |
1510 | return error; | |
1511 | ||
1512 | agi = XFS_BUF_TO_AGI(*bpp); | |
1da177e4 | 1513 | pag = &mp->m_perag[agno]; |
5e1be0fb | 1514 | |
1da177e4 | 1515 | if (!pag->pagi_init) { |
16259e7d | 1516 | pag->pagi_freecount = be32_to_cpu(agi->agi_freecount); |
92821e2b | 1517 | pag->pagi_count = be32_to_cpu(agi->agi_count); |
1da177e4 | 1518 | pag->pagi_init = 1; |
1da177e4 | 1519 | } |
1da177e4 | 1520 | |
5e1be0fb CH |
1521 | /* |
1522 | * It's possible for these to be out of sync if | |
1523 | * we are in the middle of a forced shutdown. | |
1524 | */ | |
1525 | ASSERT(pag->pagi_freecount == be32_to_cpu(agi->agi_freecount) || | |
1526 | XFS_FORCED_SHUTDOWN(mp)); | |
1da177e4 LT |
1527 | return 0; |
1528 | } | |
92821e2b DC |
1529 | |
1530 | /* | |
1531 | * Read in the agi to initialise the per-ag data in the mount structure | |
1532 | */ | |
1533 | int | |
1534 | xfs_ialloc_pagi_init( | |
1535 | xfs_mount_t *mp, /* file system mount structure */ | |
1536 | xfs_trans_t *tp, /* transaction pointer */ | |
1537 | xfs_agnumber_t agno) /* allocation group number */ | |
1538 | { | |
1539 | xfs_buf_t *bp = NULL; | |
1540 | int error; | |
1541 | ||
1542 | error = xfs_ialloc_read_agi(mp, tp, agno, &bp); | |
1543 | if (error) | |
1544 | return error; | |
1545 | if (bp) | |
1546 | xfs_trans_brelse(tp, bp); | |
1547 | return 0; | |
1548 | } |