| 1 | /* |
| 2 | * Copyright (c) 2000-2005 Silicon Graphics, Inc. |
| 3 | * All Rights Reserved. |
| 4 | * |
| 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 |
| 7 | * published by the Free Software Foundation. |
| 8 | * |
| 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. |
| 13 | * |
| 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 |
| 17 | */ |
| 18 | #include "xfs.h" |
| 19 | #include "xfs_fs.h" |
| 20 | #include "xfs_shared.h" |
| 21 | #include "xfs_format.h" |
| 22 | #include "xfs_log_format.h" |
| 23 | #include "xfs_trans_resv.h" |
| 24 | #include "xfs_sb.h" |
| 25 | #include "xfs_mount.h" |
| 26 | #include "xfs_defer.h" |
| 27 | #include "xfs_da_format.h" |
| 28 | #include "xfs_da_btree.h" |
| 29 | #include "xfs_inode.h" |
| 30 | #include "xfs_trans.h" |
| 31 | #include "xfs_inode_item.h" |
| 32 | #include "xfs_error.h" |
| 33 | #include "xfs_btree.h" |
| 34 | #include "xfs_alloc_btree.h" |
| 35 | #include "xfs_alloc.h" |
| 36 | #include "xfs_ialloc.h" |
| 37 | #include "xfs_fsops.h" |
| 38 | #include "xfs_itable.h" |
| 39 | #include "xfs_trans_space.h" |
| 40 | #include "xfs_rtalloc.h" |
| 41 | #include "xfs_trace.h" |
| 42 | #include "xfs_log.h" |
| 43 | #include "xfs_filestream.h" |
| 44 | |
| 45 | /* |
| 46 | * File system operations |
| 47 | */ |
| 48 | |
| 49 | int |
| 50 | xfs_fs_geometry( |
| 51 | xfs_mount_t *mp, |
| 52 | xfs_fsop_geom_t *geo, |
| 53 | int new_version) |
| 54 | { |
| 55 | |
| 56 | memset(geo, 0, sizeof(*geo)); |
| 57 | |
| 58 | geo->blocksize = mp->m_sb.sb_blocksize; |
| 59 | geo->rtextsize = mp->m_sb.sb_rextsize; |
| 60 | geo->agblocks = mp->m_sb.sb_agblocks; |
| 61 | geo->agcount = mp->m_sb.sb_agcount; |
| 62 | geo->logblocks = mp->m_sb.sb_logblocks; |
| 63 | geo->sectsize = mp->m_sb.sb_sectsize; |
| 64 | geo->inodesize = mp->m_sb.sb_inodesize; |
| 65 | geo->imaxpct = mp->m_sb.sb_imax_pct; |
| 66 | geo->datablocks = mp->m_sb.sb_dblocks; |
| 67 | geo->rtblocks = mp->m_sb.sb_rblocks; |
| 68 | geo->rtextents = mp->m_sb.sb_rextents; |
| 69 | geo->logstart = mp->m_sb.sb_logstart; |
| 70 | ASSERT(sizeof(geo->uuid)==sizeof(mp->m_sb.sb_uuid)); |
| 71 | memcpy(geo->uuid, &mp->m_sb.sb_uuid, sizeof(mp->m_sb.sb_uuid)); |
| 72 | if (new_version >= 2) { |
| 73 | geo->sunit = mp->m_sb.sb_unit; |
| 74 | geo->swidth = mp->m_sb.sb_width; |
| 75 | } |
| 76 | if (new_version >= 3) { |
| 77 | geo->version = XFS_FSOP_GEOM_VERSION; |
| 78 | geo->flags = XFS_FSOP_GEOM_FLAGS_NLINK | |
| 79 | XFS_FSOP_GEOM_FLAGS_DIRV2 | |
| 80 | (xfs_sb_version_hasattr(&mp->m_sb) ? |
| 81 | XFS_FSOP_GEOM_FLAGS_ATTR : 0) | |
| 82 | (xfs_sb_version_hasquota(&mp->m_sb) ? |
| 83 | XFS_FSOP_GEOM_FLAGS_QUOTA : 0) | |
| 84 | (xfs_sb_version_hasalign(&mp->m_sb) ? |
| 85 | XFS_FSOP_GEOM_FLAGS_IALIGN : 0) | |
| 86 | (xfs_sb_version_hasdalign(&mp->m_sb) ? |
| 87 | XFS_FSOP_GEOM_FLAGS_DALIGN : 0) | |
| 88 | (xfs_sb_version_hasextflgbit(&mp->m_sb) ? |
| 89 | XFS_FSOP_GEOM_FLAGS_EXTFLG : 0) | |
| 90 | (xfs_sb_version_hassector(&mp->m_sb) ? |
| 91 | XFS_FSOP_GEOM_FLAGS_SECTOR : 0) | |
| 92 | (xfs_sb_version_hasasciici(&mp->m_sb) ? |
| 93 | XFS_FSOP_GEOM_FLAGS_DIRV2CI : 0) | |
| 94 | (xfs_sb_version_haslazysbcount(&mp->m_sb) ? |
| 95 | XFS_FSOP_GEOM_FLAGS_LAZYSB : 0) | |
| 96 | (xfs_sb_version_hasattr2(&mp->m_sb) ? |
| 97 | XFS_FSOP_GEOM_FLAGS_ATTR2 : 0) | |
| 98 | (xfs_sb_version_hasprojid32bit(&mp->m_sb) ? |
| 99 | XFS_FSOP_GEOM_FLAGS_PROJID32 : 0) | |
| 100 | (xfs_sb_version_hascrc(&mp->m_sb) ? |
| 101 | XFS_FSOP_GEOM_FLAGS_V5SB : 0) | |
| 102 | (xfs_sb_version_hasftype(&mp->m_sb) ? |
| 103 | XFS_FSOP_GEOM_FLAGS_FTYPE : 0) | |
| 104 | (xfs_sb_version_hasfinobt(&mp->m_sb) ? |
| 105 | XFS_FSOP_GEOM_FLAGS_FINOBT : 0) | |
| 106 | (xfs_sb_version_hassparseinodes(&mp->m_sb) ? |
| 107 | XFS_FSOP_GEOM_FLAGS_SPINODES : 0); |
| 108 | geo->logsectsize = xfs_sb_version_hassector(&mp->m_sb) ? |
| 109 | mp->m_sb.sb_logsectsize : BBSIZE; |
| 110 | geo->rtsectsize = mp->m_sb.sb_blocksize; |
| 111 | geo->dirblocksize = mp->m_dir_geo->blksize; |
| 112 | } |
| 113 | if (new_version >= 4) { |
| 114 | geo->flags |= |
| 115 | (xfs_sb_version_haslogv2(&mp->m_sb) ? |
| 116 | XFS_FSOP_GEOM_FLAGS_LOGV2 : 0); |
| 117 | geo->logsunit = mp->m_sb.sb_logsunit; |
| 118 | } |
| 119 | return 0; |
| 120 | } |
| 121 | |
| 122 | static struct xfs_buf * |
| 123 | xfs_growfs_get_hdr_buf( |
| 124 | struct xfs_mount *mp, |
| 125 | xfs_daddr_t blkno, |
| 126 | size_t numblks, |
| 127 | int flags, |
| 128 | const struct xfs_buf_ops *ops) |
| 129 | { |
| 130 | struct xfs_buf *bp; |
| 131 | |
| 132 | bp = xfs_buf_get_uncached(mp->m_ddev_targp, numblks, flags); |
| 133 | if (!bp) |
| 134 | return NULL; |
| 135 | |
| 136 | xfs_buf_zero(bp, 0, BBTOB(bp->b_length)); |
| 137 | bp->b_bn = blkno; |
| 138 | bp->b_maps[0].bm_bn = blkno; |
| 139 | bp->b_ops = ops; |
| 140 | |
| 141 | return bp; |
| 142 | } |
| 143 | |
| 144 | static int |
| 145 | xfs_growfs_data_private( |
| 146 | xfs_mount_t *mp, /* mount point for filesystem */ |
| 147 | xfs_growfs_data_t *in) /* growfs data input struct */ |
| 148 | { |
| 149 | xfs_agf_t *agf; |
| 150 | struct xfs_agfl *agfl; |
| 151 | xfs_agi_t *agi; |
| 152 | xfs_agnumber_t agno; |
| 153 | xfs_extlen_t agsize; |
| 154 | xfs_extlen_t tmpsize; |
| 155 | xfs_alloc_rec_t *arec; |
| 156 | xfs_buf_t *bp; |
| 157 | int bucket; |
| 158 | int dpct; |
| 159 | int error, saved_error = 0; |
| 160 | xfs_agnumber_t nagcount; |
| 161 | xfs_agnumber_t nagimax = 0; |
| 162 | xfs_rfsblock_t nb, nb_mod; |
| 163 | xfs_rfsblock_t new; |
| 164 | xfs_rfsblock_t nfree; |
| 165 | xfs_agnumber_t oagcount; |
| 166 | int pct; |
| 167 | xfs_trans_t *tp; |
| 168 | |
| 169 | nb = in->newblocks; |
| 170 | pct = in->imaxpct; |
| 171 | if (nb < mp->m_sb.sb_dblocks || pct < 0 || pct > 100) |
| 172 | return -EINVAL; |
| 173 | if ((error = xfs_sb_validate_fsb_count(&mp->m_sb, nb))) |
| 174 | return error; |
| 175 | dpct = pct - mp->m_sb.sb_imax_pct; |
| 176 | error = xfs_buf_read_uncached(mp->m_ddev_targp, |
| 177 | XFS_FSB_TO_BB(mp, nb) - XFS_FSS_TO_BB(mp, 1), |
| 178 | XFS_FSS_TO_BB(mp, 1), 0, &bp, NULL); |
| 179 | if (error) |
| 180 | return error; |
| 181 | xfs_buf_relse(bp); |
| 182 | |
| 183 | new = nb; /* use new as a temporary here */ |
| 184 | nb_mod = do_div(new, mp->m_sb.sb_agblocks); |
| 185 | nagcount = new + (nb_mod != 0); |
| 186 | if (nb_mod && nb_mod < XFS_MIN_AG_BLOCKS) { |
| 187 | nagcount--; |
| 188 | nb = (xfs_rfsblock_t)nagcount * mp->m_sb.sb_agblocks; |
| 189 | if (nb < mp->m_sb.sb_dblocks) |
| 190 | return -EINVAL; |
| 191 | } |
| 192 | new = nb - mp->m_sb.sb_dblocks; |
| 193 | oagcount = mp->m_sb.sb_agcount; |
| 194 | |
| 195 | /* allocate the new per-ag structures */ |
| 196 | if (nagcount > oagcount) { |
| 197 | error = xfs_initialize_perag(mp, nagcount, &nagimax); |
| 198 | if (error) |
| 199 | return error; |
| 200 | } |
| 201 | |
| 202 | error = xfs_trans_alloc(mp, &M_RES(mp)->tr_growdata, |
| 203 | XFS_GROWFS_SPACE_RES(mp), 0, XFS_TRANS_RESERVE, &tp); |
| 204 | if (error) |
| 205 | return error; |
| 206 | |
| 207 | /* |
| 208 | * Write new AG headers to disk. Non-transactional, but written |
| 209 | * synchronously so they are completed prior to the growfs transaction |
| 210 | * being logged. |
| 211 | */ |
| 212 | nfree = 0; |
| 213 | for (agno = nagcount - 1; agno >= oagcount; agno--, new -= agsize) { |
| 214 | __be32 *agfl_bno; |
| 215 | |
| 216 | /* |
| 217 | * AG freespace header block |
| 218 | */ |
| 219 | bp = xfs_growfs_get_hdr_buf(mp, |
| 220 | XFS_AG_DADDR(mp, agno, XFS_AGF_DADDR(mp)), |
| 221 | XFS_FSS_TO_BB(mp, 1), 0, |
| 222 | &xfs_agf_buf_ops); |
| 223 | if (!bp) { |
| 224 | error = -ENOMEM; |
| 225 | goto error0; |
| 226 | } |
| 227 | |
| 228 | agf = XFS_BUF_TO_AGF(bp); |
| 229 | agf->agf_magicnum = cpu_to_be32(XFS_AGF_MAGIC); |
| 230 | agf->agf_versionnum = cpu_to_be32(XFS_AGF_VERSION); |
| 231 | agf->agf_seqno = cpu_to_be32(agno); |
| 232 | if (agno == nagcount - 1) |
| 233 | agsize = |
| 234 | nb - |
| 235 | (agno * (xfs_rfsblock_t)mp->m_sb.sb_agblocks); |
| 236 | else |
| 237 | agsize = mp->m_sb.sb_agblocks; |
| 238 | agf->agf_length = cpu_to_be32(agsize); |
| 239 | agf->agf_roots[XFS_BTNUM_BNOi] = cpu_to_be32(XFS_BNO_BLOCK(mp)); |
| 240 | agf->agf_roots[XFS_BTNUM_CNTi] = cpu_to_be32(XFS_CNT_BLOCK(mp)); |
| 241 | agf->agf_levels[XFS_BTNUM_BNOi] = cpu_to_be32(1); |
| 242 | agf->agf_levels[XFS_BTNUM_CNTi] = cpu_to_be32(1); |
| 243 | agf->agf_flfirst = cpu_to_be32(1); |
| 244 | agf->agf_fllast = 0; |
| 245 | agf->agf_flcount = 0; |
| 246 | tmpsize = agsize - mp->m_ag_prealloc_blocks; |
| 247 | agf->agf_freeblks = cpu_to_be32(tmpsize); |
| 248 | agf->agf_longest = cpu_to_be32(tmpsize); |
| 249 | if (xfs_sb_version_hascrc(&mp->m_sb)) |
| 250 | uuid_copy(&agf->agf_uuid, &mp->m_sb.sb_meta_uuid); |
| 251 | |
| 252 | error = xfs_bwrite(bp); |
| 253 | xfs_buf_relse(bp); |
| 254 | if (error) |
| 255 | goto error0; |
| 256 | |
| 257 | /* |
| 258 | * AG freelist header block |
| 259 | */ |
| 260 | bp = xfs_growfs_get_hdr_buf(mp, |
| 261 | XFS_AG_DADDR(mp, agno, XFS_AGFL_DADDR(mp)), |
| 262 | XFS_FSS_TO_BB(mp, 1), 0, |
| 263 | &xfs_agfl_buf_ops); |
| 264 | if (!bp) { |
| 265 | error = -ENOMEM; |
| 266 | goto error0; |
| 267 | } |
| 268 | |
| 269 | agfl = XFS_BUF_TO_AGFL(bp); |
| 270 | if (xfs_sb_version_hascrc(&mp->m_sb)) { |
| 271 | agfl->agfl_magicnum = cpu_to_be32(XFS_AGFL_MAGIC); |
| 272 | agfl->agfl_seqno = cpu_to_be32(agno); |
| 273 | uuid_copy(&agfl->agfl_uuid, &mp->m_sb.sb_meta_uuid); |
| 274 | } |
| 275 | |
| 276 | agfl_bno = XFS_BUF_TO_AGFL_BNO(mp, bp); |
| 277 | for (bucket = 0; bucket < XFS_AGFL_SIZE(mp); bucket++) |
| 278 | agfl_bno[bucket] = cpu_to_be32(NULLAGBLOCK); |
| 279 | |
| 280 | error = xfs_bwrite(bp); |
| 281 | xfs_buf_relse(bp); |
| 282 | if (error) |
| 283 | goto error0; |
| 284 | |
| 285 | /* |
| 286 | * AG inode header block |
| 287 | */ |
| 288 | bp = xfs_growfs_get_hdr_buf(mp, |
| 289 | XFS_AG_DADDR(mp, agno, XFS_AGI_DADDR(mp)), |
| 290 | XFS_FSS_TO_BB(mp, 1), 0, |
| 291 | &xfs_agi_buf_ops); |
| 292 | if (!bp) { |
| 293 | error = -ENOMEM; |
| 294 | goto error0; |
| 295 | } |
| 296 | |
| 297 | agi = XFS_BUF_TO_AGI(bp); |
| 298 | agi->agi_magicnum = cpu_to_be32(XFS_AGI_MAGIC); |
| 299 | agi->agi_versionnum = cpu_to_be32(XFS_AGI_VERSION); |
| 300 | agi->agi_seqno = cpu_to_be32(agno); |
| 301 | agi->agi_length = cpu_to_be32(agsize); |
| 302 | agi->agi_count = 0; |
| 303 | agi->agi_root = cpu_to_be32(XFS_IBT_BLOCK(mp)); |
| 304 | agi->agi_level = cpu_to_be32(1); |
| 305 | agi->agi_freecount = 0; |
| 306 | agi->agi_newino = cpu_to_be32(NULLAGINO); |
| 307 | agi->agi_dirino = cpu_to_be32(NULLAGINO); |
| 308 | if (xfs_sb_version_hascrc(&mp->m_sb)) |
| 309 | uuid_copy(&agi->agi_uuid, &mp->m_sb.sb_meta_uuid); |
| 310 | if (xfs_sb_version_hasfinobt(&mp->m_sb)) { |
| 311 | agi->agi_free_root = cpu_to_be32(XFS_FIBT_BLOCK(mp)); |
| 312 | agi->agi_free_level = cpu_to_be32(1); |
| 313 | } |
| 314 | for (bucket = 0; bucket < XFS_AGI_UNLINKED_BUCKETS; bucket++) |
| 315 | agi->agi_unlinked[bucket] = cpu_to_be32(NULLAGINO); |
| 316 | |
| 317 | error = xfs_bwrite(bp); |
| 318 | xfs_buf_relse(bp); |
| 319 | if (error) |
| 320 | goto error0; |
| 321 | |
| 322 | /* |
| 323 | * BNO btree root block |
| 324 | */ |
| 325 | bp = xfs_growfs_get_hdr_buf(mp, |
| 326 | XFS_AGB_TO_DADDR(mp, agno, XFS_BNO_BLOCK(mp)), |
| 327 | BTOBB(mp->m_sb.sb_blocksize), 0, |
| 328 | &xfs_allocbt_buf_ops); |
| 329 | |
| 330 | if (!bp) { |
| 331 | error = -ENOMEM; |
| 332 | goto error0; |
| 333 | } |
| 334 | |
| 335 | if (xfs_sb_version_hascrc(&mp->m_sb)) |
| 336 | xfs_btree_init_block(mp, bp, XFS_ABTB_CRC_MAGIC, 0, 1, |
| 337 | agno, XFS_BTREE_CRC_BLOCKS); |
| 338 | else |
| 339 | xfs_btree_init_block(mp, bp, XFS_ABTB_MAGIC, 0, 1, |
| 340 | agno, 0); |
| 341 | |
| 342 | arec = XFS_ALLOC_REC_ADDR(mp, XFS_BUF_TO_BLOCK(bp), 1); |
| 343 | arec->ar_startblock = cpu_to_be32(mp->m_ag_prealloc_blocks); |
| 344 | arec->ar_blockcount = cpu_to_be32( |
| 345 | agsize - be32_to_cpu(arec->ar_startblock)); |
| 346 | |
| 347 | error = xfs_bwrite(bp); |
| 348 | xfs_buf_relse(bp); |
| 349 | if (error) |
| 350 | goto error0; |
| 351 | |
| 352 | /* |
| 353 | * CNT btree root block |
| 354 | */ |
| 355 | bp = xfs_growfs_get_hdr_buf(mp, |
| 356 | XFS_AGB_TO_DADDR(mp, agno, XFS_CNT_BLOCK(mp)), |
| 357 | BTOBB(mp->m_sb.sb_blocksize), 0, |
| 358 | &xfs_allocbt_buf_ops); |
| 359 | if (!bp) { |
| 360 | error = -ENOMEM; |
| 361 | goto error0; |
| 362 | } |
| 363 | |
| 364 | if (xfs_sb_version_hascrc(&mp->m_sb)) |
| 365 | xfs_btree_init_block(mp, bp, XFS_ABTC_CRC_MAGIC, 0, 1, |
| 366 | agno, XFS_BTREE_CRC_BLOCKS); |
| 367 | else |
| 368 | xfs_btree_init_block(mp, bp, XFS_ABTC_MAGIC, 0, 1, |
| 369 | agno, 0); |
| 370 | |
| 371 | arec = XFS_ALLOC_REC_ADDR(mp, XFS_BUF_TO_BLOCK(bp), 1); |
| 372 | arec->ar_startblock = cpu_to_be32(mp->m_ag_prealloc_blocks); |
| 373 | arec->ar_blockcount = cpu_to_be32( |
| 374 | agsize - be32_to_cpu(arec->ar_startblock)); |
| 375 | nfree += be32_to_cpu(arec->ar_blockcount); |
| 376 | |
| 377 | error = xfs_bwrite(bp); |
| 378 | xfs_buf_relse(bp); |
| 379 | if (error) |
| 380 | goto error0; |
| 381 | |
| 382 | /* |
| 383 | * INO btree root block |
| 384 | */ |
| 385 | bp = xfs_growfs_get_hdr_buf(mp, |
| 386 | XFS_AGB_TO_DADDR(mp, agno, XFS_IBT_BLOCK(mp)), |
| 387 | BTOBB(mp->m_sb.sb_blocksize), 0, |
| 388 | &xfs_inobt_buf_ops); |
| 389 | if (!bp) { |
| 390 | error = -ENOMEM; |
| 391 | goto error0; |
| 392 | } |
| 393 | |
| 394 | if (xfs_sb_version_hascrc(&mp->m_sb)) |
| 395 | xfs_btree_init_block(mp, bp, XFS_IBT_CRC_MAGIC, 0, 0, |
| 396 | agno, XFS_BTREE_CRC_BLOCKS); |
| 397 | else |
| 398 | xfs_btree_init_block(mp, bp, XFS_IBT_MAGIC, 0, 0, |
| 399 | agno, 0); |
| 400 | |
| 401 | error = xfs_bwrite(bp); |
| 402 | xfs_buf_relse(bp); |
| 403 | if (error) |
| 404 | goto error0; |
| 405 | |
| 406 | /* |
| 407 | * FINO btree root block |
| 408 | */ |
| 409 | if (xfs_sb_version_hasfinobt(&mp->m_sb)) { |
| 410 | bp = xfs_growfs_get_hdr_buf(mp, |
| 411 | XFS_AGB_TO_DADDR(mp, agno, XFS_FIBT_BLOCK(mp)), |
| 412 | BTOBB(mp->m_sb.sb_blocksize), 0, |
| 413 | &xfs_inobt_buf_ops); |
| 414 | if (!bp) { |
| 415 | error = -ENOMEM; |
| 416 | goto error0; |
| 417 | } |
| 418 | |
| 419 | if (xfs_sb_version_hascrc(&mp->m_sb)) |
| 420 | xfs_btree_init_block(mp, bp, XFS_FIBT_CRC_MAGIC, |
| 421 | 0, 0, agno, |
| 422 | XFS_BTREE_CRC_BLOCKS); |
| 423 | else |
| 424 | xfs_btree_init_block(mp, bp, XFS_FIBT_MAGIC, 0, |
| 425 | 0, agno, 0); |
| 426 | |
| 427 | error = xfs_bwrite(bp); |
| 428 | xfs_buf_relse(bp); |
| 429 | if (error) |
| 430 | goto error0; |
| 431 | } |
| 432 | |
| 433 | } |
| 434 | xfs_trans_agblocks_delta(tp, nfree); |
| 435 | /* |
| 436 | * There are new blocks in the old last a.g. |
| 437 | */ |
| 438 | if (new) { |
| 439 | /* |
| 440 | * Change the agi length. |
| 441 | */ |
| 442 | error = xfs_ialloc_read_agi(mp, tp, agno, &bp); |
| 443 | if (error) { |
| 444 | goto error0; |
| 445 | } |
| 446 | ASSERT(bp); |
| 447 | agi = XFS_BUF_TO_AGI(bp); |
| 448 | be32_add_cpu(&agi->agi_length, new); |
| 449 | ASSERT(nagcount == oagcount || |
| 450 | be32_to_cpu(agi->agi_length) == mp->m_sb.sb_agblocks); |
| 451 | xfs_ialloc_log_agi(tp, bp, XFS_AGI_LENGTH); |
| 452 | /* |
| 453 | * Change agf length. |
| 454 | */ |
| 455 | error = xfs_alloc_read_agf(mp, tp, agno, 0, &bp); |
| 456 | if (error) { |
| 457 | goto error0; |
| 458 | } |
| 459 | ASSERT(bp); |
| 460 | agf = XFS_BUF_TO_AGF(bp); |
| 461 | be32_add_cpu(&agf->agf_length, new); |
| 462 | ASSERT(be32_to_cpu(agf->agf_length) == |
| 463 | be32_to_cpu(agi->agi_length)); |
| 464 | |
| 465 | xfs_alloc_log_agf(tp, bp, XFS_AGF_LENGTH); |
| 466 | /* |
| 467 | * Free the new space. |
| 468 | */ |
| 469 | error = xfs_free_extent(tp, XFS_AGB_TO_FSB(mp, agno, |
| 470 | be32_to_cpu(agf->agf_length) - new), new); |
| 471 | if (error) { |
| 472 | goto error0; |
| 473 | } |
| 474 | } |
| 475 | |
| 476 | /* |
| 477 | * Update changed superblock fields transactionally. These are not |
| 478 | * seen by the rest of the world until the transaction commit applies |
| 479 | * them atomically to the superblock. |
| 480 | */ |
| 481 | if (nagcount > oagcount) |
| 482 | xfs_trans_mod_sb(tp, XFS_TRANS_SB_AGCOUNT, nagcount - oagcount); |
| 483 | if (nb > mp->m_sb.sb_dblocks) |
| 484 | xfs_trans_mod_sb(tp, XFS_TRANS_SB_DBLOCKS, |
| 485 | nb - mp->m_sb.sb_dblocks); |
| 486 | if (nfree) |
| 487 | xfs_trans_mod_sb(tp, XFS_TRANS_SB_FDBLOCKS, nfree); |
| 488 | if (dpct) |
| 489 | xfs_trans_mod_sb(tp, XFS_TRANS_SB_IMAXPCT, dpct); |
| 490 | xfs_trans_set_sync(tp); |
| 491 | error = xfs_trans_commit(tp); |
| 492 | if (error) |
| 493 | return error; |
| 494 | |
| 495 | /* New allocation groups fully initialized, so update mount struct */ |
| 496 | if (nagimax) |
| 497 | mp->m_maxagi = nagimax; |
| 498 | if (mp->m_sb.sb_imax_pct) { |
| 499 | __uint64_t icount = mp->m_sb.sb_dblocks * mp->m_sb.sb_imax_pct; |
| 500 | do_div(icount, 100); |
| 501 | mp->m_maxicount = icount << mp->m_sb.sb_inopblog; |
| 502 | } else |
| 503 | mp->m_maxicount = 0; |
| 504 | xfs_set_low_space_thresholds(mp); |
| 505 | |
| 506 | /* update secondary superblocks. */ |
| 507 | for (agno = 1; agno < nagcount; agno++) { |
| 508 | error = 0; |
| 509 | /* |
| 510 | * new secondary superblocks need to be zeroed, not read from |
| 511 | * disk as the contents of the new area we are growing into is |
| 512 | * completely unknown. |
| 513 | */ |
| 514 | if (agno < oagcount) { |
| 515 | error = xfs_trans_read_buf(mp, NULL, mp->m_ddev_targp, |
| 516 | XFS_AGB_TO_DADDR(mp, agno, XFS_SB_BLOCK(mp)), |
| 517 | XFS_FSS_TO_BB(mp, 1), 0, &bp, |
| 518 | &xfs_sb_buf_ops); |
| 519 | } else { |
| 520 | bp = xfs_trans_get_buf(NULL, mp->m_ddev_targp, |
| 521 | XFS_AGB_TO_DADDR(mp, agno, XFS_SB_BLOCK(mp)), |
| 522 | XFS_FSS_TO_BB(mp, 1), 0); |
| 523 | if (bp) { |
| 524 | bp->b_ops = &xfs_sb_buf_ops; |
| 525 | xfs_buf_zero(bp, 0, BBTOB(bp->b_length)); |
| 526 | } else |
| 527 | error = -ENOMEM; |
| 528 | } |
| 529 | |
| 530 | /* |
| 531 | * If we get an error reading or writing alternate superblocks, |
| 532 | * continue. xfs_repair chooses the "best" superblock based |
| 533 | * on most matches; if we break early, we'll leave more |
| 534 | * superblocks un-updated than updated, and xfs_repair may |
| 535 | * pick them over the properly-updated primary. |
| 536 | */ |
| 537 | if (error) { |
| 538 | xfs_warn(mp, |
| 539 | "error %d reading secondary superblock for ag %d", |
| 540 | error, agno); |
| 541 | saved_error = error; |
| 542 | continue; |
| 543 | } |
| 544 | xfs_sb_to_disk(XFS_BUF_TO_SBP(bp), &mp->m_sb); |
| 545 | |
| 546 | error = xfs_bwrite(bp); |
| 547 | xfs_buf_relse(bp); |
| 548 | if (error) { |
| 549 | xfs_warn(mp, |
| 550 | "write error %d updating secondary superblock for ag %d", |
| 551 | error, agno); |
| 552 | saved_error = error; |
| 553 | continue; |
| 554 | } |
| 555 | } |
| 556 | return saved_error ? saved_error : error; |
| 557 | |
| 558 | error0: |
| 559 | xfs_trans_cancel(tp); |
| 560 | return error; |
| 561 | } |
| 562 | |
| 563 | static int |
| 564 | xfs_growfs_log_private( |
| 565 | xfs_mount_t *mp, /* mount point for filesystem */ |
| 566 | xfs_growfs_log_t *in) /* growfs log input struct */ |
| 567 | { |
| 568 | xfs_extlen_t nb; |
| 569 | |
| 570 | nb = in->newblocks; |
| 571 | if (nb < XFS_MIN_LOG_BLOCKS || nb < XFS_B_TO_FSB(mp, XFS_MIN_LOG_BYTES)) |
| 572 | return -EINVAL; |
| 573 | if (nb == mp->m_sb.sb_logblocks && |
| 574 | in->isint == (mp->m_sb.sb_logstart != 0)) |
| 575 | return -EINVAL; |
| 576 | /* |
| 577 | * Moving the log is hard, need new interfaces to sync |
| 578 | * the log first, hold off all activity while moving it. |
| 579 | * Can have shorter or longer log in the same space, |
| 580 | * or transform internal to external log or vice versa. |
| 581 | */ |
| 582 | return -ENOSYS; |
| 583 | } |
| 584 | |
| 585 | /* |
| 586 | * protected versions of growfs function acquire and release locks on the mount |
| 587 | * point - exported through ioctls: XFS_IOC_FSGROWFSDATA, XFS_IOC_FSGROWFSLOG, |
| 588 | * XFS_IOC_FSGROWFSRT |
| 589 | */ |
| 590 | |
| 591 | |
| 592 | int |
| 593 | xfs_growfs_data( |
| 594 | xfs_mount_t *mp, |
| 595 | xfs_growfs_data_t *in) |
| 596 | { |
| 597 | int error; |
| 598 | |
| 599 | if (!capable(CAP_SYS_ADMIN)) |
| 600 | return -EPERM; |
| 601 | if (!mutex_trylock(&mp->m_growlock)) |
| 602 | return -EWOULDBLOCK; |
| 603 | error = xfs_growfs_data_private(mp, in); |
| 604 | /* |
| 605 | * Increment the generation unconditionally, the error could be from |
| 606 | * updating the secondary superblocks, in which case the new size |
| 607 | * is live already. |
| 608 | */ |
| 609 | mp->m_generation++; |
| 610 | mutex_unlock(&mp->m_growlock); |
| 611 | return error; |
| 612 | } |
| 613 | |
| 614 | int |
| 615 | xfs_growfs_log( |
| 616 | xfs_mount_t *mp, |
| 617 | xfs_growfs_log_t *in) |
| 618 | { |
| 619 | int error; |
| 620 | |
| 621 | if (!capable(CAP_SYS_ADMIN)) |
| 622 | return -EPERM; |
| 623 | if (!mutex_trylock(&mp->m_growlock)) |
| 624 | return -EWOULDBLOCK; |
| 625 | error = xfs_growfs_log_private(mp, in); |
| 626 | mutex_unlock(&mp->m_growlock); |
| 627 | return error; |
| 628 | } |
| 629 | |
| 630 | /* |
| 631 | * exported through ioctl XFS_IOC_FSCOUNTS |
| 632 | */ |
| 633 | |
| 634 | int |
| 635 | xfs_fs_counts( |
| 636 | xfs_mount_t *mp, |
| 637 | xfs_fsop_counts_t *cnt) |
| 638 | { |
| 639 | cnt->allocino = percpu_counter_read_positive(&mp->m_icount); |
| 640 | cnt->freeino = percpu_counter_read_positive(&mp->m_ifree); |
| 641 | cnt->freedata = percpu_counter_read_positive(&mp->m_fdblocks) - |
| 642 | XFS_ALLOC_SET_ASIDE(mp); |
| 643 | |
| 644 | spin_lock(&mp->m_sb_lock); |
| 645 | cnt->freertx = mp->m_sb.sb_frextents; |
| 646 | spin_unlock(&mp->m_sb_lock); |
| 647 | return 0; |
| 648 | } |
| 649 | |
| 650 | /* |
| 651 | * exported through ioctl XFS_IOC_SET_RESBLKS & XFS_IOC_GET_RESBLKS |
| 652 | * |
| 653 | * xfs_reserve_blocks is called to set m_resblks |
| 654 | * in the in-core mount table. The number of unused reserved blocks |
| 655 | * is kept in m_resblks_avail. |
| 656 | * |
| 657 | * Reserve the requested number of blocks if available. Otherwise return |
| 658 | * as many as possible to satisfy the request. The actual number |
| 659 | * reserved are returned in outval |
| 660 | * |
| 661 | * A null inval pointer indicates that only the current reserved blocks |
| 662 | * available should be returned no settings are changed. |
| 663 | */ |
| 664 | |
| 665 | int |
| 666 | xfs_reserve_blocks( |
| 667 | xfs_mount_t *mp, |
| 668 | __uint64_t *inval, |
| 669 | xfs_fsop_resblks_t *outval) |
| 670 | { |
| 671 | __int64_t lcounter, delta; |
| 672 | __int64_t fdblks_delta = 0; |
| 673 | __uint64_t request; |
| 674 | __int64_t free; |
| 675 | int error = 0; |
| 676 | |
| 677 | /* If inval is null, report current values and return */ |
| 678 | if (inval == (__uint64_t *)NULL) { |
| 679 | if (!outval) |
| 680 | return -EINVAL; |
| 681 | outval->resblks = mp->m_resblks; |
| 682 | outval->resblks_avail = mp->m_resblks_avail; |
| 683 | return 0; |
| 684 | } |
| 685 | |
| 686 | request = *inval; |
| 687 | |
| 688 | /* |
| 689 | * With per-cpu counters, this becomes an interesting problem. we need |
| 690 | * to work out if we are freeing or allocation blocks first, then we can |
| 691 | * do the modification as necessary. |
| 692 | * |
| 693 | * We do this under the m_sb_lock so that if we are near ENOSPC, we will |
| 694 | * hold out any changes while we work out what to do. This means that |
| 695 | * the amount of free space can change while we do this, so we need to |
| 696 | * retry if we end up trying to reserve more space than is available. |
| 697 | */ |
| 698 | spin_lock(&mp->m_sb_lock); |
| 699 | |
| 700 | /* |
| 701 | * If our previous reservation was larger than the current value, |
| 702 | * then move any unused blocks back to the free pool. Modify the resblks |
| 703 | * counters directly since we shouldn't have any problems unreserving |
| 704 | * space. |
| 705 | */ |
| 706 | if (mp->m_resblks > request) { |
| 707 | lcounter = mp->m_resblks_avail - request; |
| 708 | if (lcounter > 0) { /* release unused blocks */ |
| 709 | fdblks_delta = lcounter; |
| 710 | mp->m_resblks_avail -= lcounter; |
| 711 | } |
| 712 | mp->m_resblks = request; |
| 713 | if (fdblks_delta) { |
| 714 | spin_unlock(&mp->m_sb_lock); |
| 715 | error = xfs_mod_fdblocks(mp, fdblks_delta, 0); |
| 716 | spin_lock(&mp->m_sb_lock); |
| 717 | } |
| 718 | |
| 719 | goto out; |
| 720 | } |
| 721 | |
| 722 | /* |
| 723 | * If the request is larger than the current reservation, reserve the |
| 724 | * blocks before we update the reserve counters. Sample m_fdblocks and |
| 725 | * perform a partial reservation if the request exceeds free space. |
| 726 | */ |
| 727 | error = -ENOSPC; |
| 728 | do { |
| 729 | free = percpu_counter_sum(&mp->m_fdblocks) - |
| 730 | XFS_ALLOC_SET_ASIDE(mp); |
| 731 | if (!free) |
| 732 | break; |
| 733 | |
| 734 | delta = request - mp->m_resblks; |
| 735 | lcounter = free - delta; |
| 736 | if (lcounter < 0) |
| 737 | /* We can't satisfy the request, just get what we can */ |
| 738 | fdblks_delta = free; |
| 739 | else |
| 740 | fdblks_delta = delta; |
| 741 | |
| 742 | /* |
| 743 | * We'll either succeed in getting space from the free block |
| 744 | * count or we'll get an ENOSPC. If we get a ENOSPC, it means |
| 745 | * things changed while we were calculating fdblks_delta and so |
| 746 | * we should try again to see if there is anything left to |
| 747 | * reserve. |
| 748 | * |
| 749 | * Don't set the reserved flag here - we don't want to reserve |
| 750 | * the extra reserve blocks from the reserve..... |
| 751 | */ |
| 752 | spin_unlock(&mp->m_sb_lock); |
| 753 | error = xfs_mod_fdblocks(mp, -fdblks_delta, 0); |
| 754 | spin_lock(&mp->m_sb_lock); |
| 755 | } while (error == -ENOSPC); |
| 756 | |
| 757 | /* |
| 758 | * Update the reserve counters if blocks have been successfully |
| 759 | * allocated. |
| 760 | */ |
| 761 | if (!error && fdblks_delta) { |
| 762 | mp->m_resblks += fdblks_delta; |
| 763 | mp->m_resblks_avail += fdblks_delta; |
| 764 | } |
| 765 | |
| 766 | out: |
| 767 | if (outval) { |
| 768 | outval->resblks = mp->m_resblks; |
| 769 | outval->resblks_avail = mp->m_resblks_avail; |
| 770 | } |
| 771 | |
| 772 | spin_unlock(&mp->m_sb_lock); |
| 773 | return error; |
| 774 | } |
| 775 | |
| 776 | int |
| 777 | xfs_fs_goingdown( |
| 778 | xfs_mount_t *mp, |
| 779 | __uint32_t inflags) |
| 780 | { |
| 781 | switch (inflags) { |
| 782 | case XFS_FSOP_GOING_FLAGS_DEFAULT: { |
| 783 | struct super_block *sb = freeze_bdev(mp->m_super->s_bdev); |
| 784 | |
| 785 | if (sb && !IS_ERR(sb)) { |
| 786 | xfs_force_shutdown(mp, SHUTDOWN_FORCE_UMOUNT); |
| 787 | thaw_bdev(sb->s_bdev, sb); |
| 788 | } |
| 789 | |
| 790 | break; |
| 791 | } |
| 792 | case XFS_FSOP_GOING_FLAGS_LOGFLUSH: |
| 793 | xfs_force_shutdown(mp, SHUTDOWN_FORCE_UMOUNT); |
| 794 | break; |
| 795 | case XFS_FSOP_GOING_FLAGS_NOLOGFLUSH: |
| 796 | xfs_force_shutdown(mp, |
| 797 | SHUTDOWN_FORCE_UMOUNT | SHUTDOWN_LOG_IO_ERROR); |
| 798 | break; |
| 799 | default: |
| 800 | return -EINVAL; |
| 801 | } |
| 802 | |
| 803 | return 0; |
| 804 | } |
| 805 | |
| 806 | /* |
| 807 | * Force a shutdown of the filesystem instantly while keeping the filesystem |
| 808 | * consistent. We don't do an unmount here; just shutdown the shop, make sure |
| 809 | * that absolutely nothing persistent happens to this filesystem after this |
| 810 | * point. |
| 811 | */ |
| 812 | void |
| 813 | xfs_do_force_shutdown( |
| 814 | xfs_mount_t *mp, |
| 815 | int flags, |
| 816 | char *fname, |
| 817 | int lnnum) |
| 818 | { |
| 819 | int logerror; |
| 820 | |
| 821 | logerror = flags & SHUTDOWN_LOG_IO_ERROR; |
| 822 | |
| 823 | if (!(flags & SHUTDOWN_FORCE_UMOUNT)) { |
| 824 | xfs_notice(mp, |
| 825 | "%s(0x%x) called from line %d of file %s. Return address = 0x%p", |
| 826 | __func__, flags, lnnum, fname, __return_address); |
| 827 | } |
| 828 | /* |
| 829 | * No need to duplicate efforts. |
| 830 | */ |
| 831 | if (XFS_FORCED_SHUTDOWN(mp) && !logerror) |
| 832 | return; |
| 833 | |
| 834 | /* |
| 835 | * This flags XFS_MOUNT_FS_SHUTDOWN, makes sure that we don't |
| 836 | * queue up anybody new on the log reservations, and wakes up |
| 837 | * everybody who's sleeping on log reservations to tell them |
| 838 | * the bad news. |
| 839 | */ |
| 840 | if (xfs_log_force_umount(mp, logerror)) |
| 841 | return; |
| 842 | |
| 843 | if (flags & SHUTDOWN_CORRUPT_INCORE) { |
| 844 | xfs_alert_tag(mp, XFS_PTAG_SHUTDOWN_CORRUPT, |
| 845 | "Corruption of in-memory data detected. Shutting down filesystem"); |
| 846 | if (XFS_ERRLEVEL_HIGH <= xfs_error_level) |
| 847 | xfs_stack_trace(); |
| 848 | } else if (!(flags & SHUTDOWN_FORCE_UMOUNT)) { |
| 849 | if (logerror) { |
| 850 | xfs_alert_tag(mp, XFS_PTAG_SHUTDOWN_LOGERROR, |
| 851 | "Log I/O Error Detected. Shutting down filesystem"); |
| 852 | } else if (flags & SHUTDOWN_DEVICE_REQ) { |
| 853 | xfs_alert_tag(mp, XFS_PTAG_SHUTDOWN_IOERROR, |
| 854 | "All device paths lost. Shutting down filesystem"); |
| 855 | } else if (!(flags & SHUTDOWN_REMOTE_REQ)) { |
| 856 | xfs_alert_tag(mp, XFS_PTAG_SHUTDOWN_IOERROR, |
| 857 | "I/O Error Detected. Shutting down filesystem"); |
| 858 | } |
| 859 | } |
| 860 | if (!(flags & SHUTDOWN_FORCE_UMOUNT)) { |
| 861 | xfs_alert(mp, |
| 862 | "Please umount the filesystem and rectify the problem(s)"); |
| 863 | } |
| 864 | } |