| 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_types.h" |
| 21 | #include "xfs_log.h" |
| 22 | #include "xfs_trans.h" |
| 23 | #include "xfs_sb.h" |
| 24 | #include "xfs_ag.h" |
| 25 | #include "xfs_mount.h" |
| 26 | #include "xfs_bmap_btree.h" |
| 27 | #include "xfs_alloc_btree.h" |
| 28 | #include "xfs_ialloc_btree.h" |
| 29 | #include "xfs_dinode.h" |
| 30 | #include "xfs_inode.h" |
| 31 | #include "xfs_inode_item.h" |
| 32 | #include "xfs_btree.h" |
| 33 | #include "xfs_error.h" |
| 34 | #include "xfs_alloc.h" |
| 35 | #include "xfs_ialloc.h" |
| 36 | #include "xfs_fsops.h" |
| 37 | #include "xfs_itable.h" |
| 38 | #include "xfs_trans_space.h" |
| 39 | #include "xfs_rtalloc.h" |
| 40 | #include "xfs_filestream.h" |
| 41 | #include "xfs_trace.h" |
| 42 | |
| 43 | /* |
| 44 | * File system operations |
| 45 | */ |
| 46 | |
| 47 | int |
| 48 | xfs_fs_geometry( |
| 49 | xfs_mount_t *mp, |
| 50 | xfs_fsop_geom_t *geo, |
| 51 | int new_version) |
| 52 | { |
| 53 | |
| 54 | memset(geo, 0, sizeof(*geo)); |
| 55 | |
| 56 | geo->blocksize = mp->m_sb.sb_blocksize; |
| 57 | geo->rtextsize = mp->m_sb.sb_rextsize; |
| 58 | geo->agblocks = mp->m_sb.sb_agblocks; |
| 59 | geo->agcount = mp->m_sb.sb_agcount; |
| 60 | geo->logblocks = mp->m_sb.sb_logblocks; |
| 61 | geo->sectsize = mp->m_sb.sb_sectsize; |
| 62 | geo->inodesize = mp->m_sb.sb_inodesize; |
| 63 | geo->imaxpct = mp->m_sb.sb_imax_pct; |
| 64 | geo->datablocks = mp->m_sb.sb_dblocks; |
| 65 | geo->rtblocks = mp->m_sb.sb_rblocks; |
| 66 | geo->rtextents = mp->m_sb.sb_rextents; |
| 67 | geo->logstart = mp->m_sb.sb_logstart; |
| 68 | ASSERT(sizeof(geo->uuid)==sizeof(mp->m_sb.sb_uuid)); |
| 69 | memcpy(geo->uuid, &mp->m_sb.sb_uuid, sizeof(mp->m_sb.sb_uuid)); |
| 70 | if (new_version >= 2) { |
| 71 | geo->sunit = mp->m_sb.sb_unit; |
| 72 | geo->swidth = mp->m_sb.sb_width; |
| 73 | } |
| 74 | if (new_version >= 3) { |
| 75 | geo->version = XFS_FSOP_GEOM_VERSION; |
| 76 | geo->flags = |
| 77 | (xfs_sb_version_hasattr(&mp->m_sb) ? |
| 78 | XFS_FSOP_GEOM_FLAGS_ATTR : 0) | |
| 79 | (xfs_sb_version_hasnlink(&mp->m_sb) ? |
| 80 | XFS_FSOP_GEOM_FLAGS_NLINK : 0) | |
| 81 | (xfs_sb_version_hasquota(&mp->m_sb) ? |
| 82 | XFS_FSOP_GEOM_FLAGS_QUOTA : 0) | |
| 83 | (xfs_sb_version_hasalign(&mp->m_sb) ? |
| 84 | XFS_FSOP_GEOM_FLAGS_IALIGN : 0) | |
| 85 | (xfs_sb_version_hasdalign(&mp->m_sb) ? |
| 86 | XFS_FSOP_GEOM_FLAGS_DALIGN : 0) | |
| 87 | (xfs_sb_version_hasshared(&mp->m_sb) ? |
| 88 | XFS_FSOP_GEOM_FLAGS_SHARED : 0) | |
| 89 | (xfs_sb_version_hasextflgbit(&mp->m_sb) ? |
| 90 | XFS_FSOP_GEOM_FLAGS_EXTFLG : 0) | |
| 91 | (xfs_sb_version_hasdirv2(&mp->m_sb) ? |
| 92 | XFS_FSOP_GEOM_FLAGS_DIRV2 : 0) | |
| 93 | (xfs_sb_version_hassector(&mp->m_sb) ? |
| 94 | XFS_FSOP_GEOM_FLAGS_SECTOR : 0) | |
| 95 | (xfs_sb_version_hasasciici(&mp->m_sb) ? |
| 96 | XFS_FSOP_GEOM_FLAGS_DIRV2CI : 0) | |
| 97 | (xfs_sb_version_haslazysbcount(&mp->m_sb) ? |
| 98 | XFS_FSOP_GEOM_FLAGS_LAZYSB : 0) | |
| 99 | (xfs_sb_version_hasattr2(&mp->m_sb) ? |
| 100 | XFS_FSOP_GEOM_FLAGS_ATTR2 : 0) | |
| 101 | (xfs_sb_version_hasprojid32bit(&mp->m_sb) ? |
| 102 | XFS_FSOP_GEOM_FLAGS_PROJID32 : 0); |
| 103 | geo->logsectsize = xfs_sb_version_hassector(&mp->m_sb) ? |
| 104 | mp->m_sb.sb_logsectsize : BBSIZE; |
| 105 | geo->rtsectsize = mp->m_sb.sb_blocksize; |
| 106 | geo->dirblocksize = mp->m_dirblksize; |
| 107 | } |
| 108 | if (new_version >= 4) { |
| 109 | geo->flags |= |
| 110 | (xfs_sb_version_haslogv2(&mp->m_sb) ? |
| 111 | XFS_FSOP_GEOM_FLAGS_LOGV2 : 0); |
| 112 | geo->logsunit = mp->m_sb.sb_logsunit; |
| 113 | } |
| 114 | return 0; |
| 115 | } |
| 116 | |
| 117 | static struct xfs_buf * |
| 118 | xfs_growfs_get_hdr_buf( |
| 119 | struct xfs_mount *mp, |
| 120 | xfs_daddr_t blkno, |
| 121 | size_t numblks, |
| 122 | int flags) |
| 123 | { |
| 124 | struct xfs_buf *bp; |
| 125 | |
| 126 | bp = xfs_buf_get_uncached(mp->m_ddev_targp, numblks, flags); |
| 127 | if (!bp) |
| 128 | return NULL; |
| 129 | |
| 130 | xfs_buf_zero(bp, 0, BBTOB(bp->b_length)); |
| 131 | bp->b_bn = blkno; |
| 132 | bp->b_maps[0].bm_bn = blkno; |
| 133 | |
| 134 | return bp; |
| 135 | } |
| 136 | |
| 137 | static int |
| 138 | xfs_growfs_data_private( |
| 139 | xfs_mount_t *mp, /* mount point for filesystem */ |
| 140 | xfs_growfs_data_t *in) /* growfs data input struct */ |
| 141 | { |
| 142 | xfs_agf_t *agf; |
| 143 | struct xfs_agfl *agfl; |
| 144 | xfs_agi_t *agi; |
| 145 | xfs_agnumber_t agno; |
| 146 | xfs_extlen_t agsize; |
| 147 | xfs_extlen_t tmpsize; |
| 148 | xfs_alloc_rec_t *arec; |
| 149 | xfs_buf_t *bp; |
| 150 | int bucket; |
| 151 | int dpct; |
| 152 | int error; |
| 153 | xfs_agnumber_t nagcount; |
| 154 | xfs_agnumber_t nagimax = 0; |
| 155 | xfs_rfsblock_t nb, nb_mod; |
| 156 | xfs_rfsblock_t new; |
| 157 | xfs_rfsblock_t nfree; |
| 158 | xfs_agnumber_t oagcount; |
| 159 | int pct; |
| 160 | xfs_trans_t *tp; |
| 161 | |
| 162 | nb = in->newblocks; |
| 163 | pct = in->imaxpct; |
| 164 | if (nb < mp->m_sb.sb_dblocks || pct < 0 || pct > 100) |
| 165 | return XFS_ERROR(EINVAL); |
| 166 | if ((error = xfs_sb_validate_fsb_count(&mp->m_sb, nb))) |
| 167 | return error; |
| 168 | dpct = pct - mp->m_sb.sb_imax_pct; |
| 169 | bp = xfs_buf_read_uncached(mp->m_ddev_targp, |
| 170 | XFS_FSB_TO_BB(mp, nb) - XFS_FSS_TO_BB(mp, 1), |
| 171 | XFS_FSS_TO_BB(mp, 1), 0, NULL); |
| 172 | if (!bp) |
| 173 | return EIO; |
| 174 | if (bp->b_error) { |
| 175 | int error = bp->b_error; |
| 176 | xfs_buf_relse(bp); |
| 177 | return error; |
| 178 | } |
| 179 | xfs_buf_relse(bp); |
| 180 | |
| 181 | new = nb; /* use new as a temporary here */ |
| 182 | nb_mod = do_div(new, mp->m_sb.sb_agblocks); |
| 183 | nagcount = new + (nb_mod != 0); |
| 184 | if (nb_mod && nb_mod < XFS_MIN_AG_BLOCKS) { |
| 185 | nagcount--; |
| 186 | nb = (xfs_rfsblock_t)nagcount * mp->m_sb.sb_agblocks; |
| 187 | if (nb < mp->m_sb.sb_dblocks) |
| 188 | return XFS_ERROR(EINVAL); |
| 189 | } |
| 190 | new = nb - mp->m_sb.sb_dblocks; |
| 191 | oagcount = mp->m_sb.sb_agcount; |
| 192 | |
| 193 | /* allocate the new per-ag structures */ |
| 194 | if (nagcount > oagcount) { |
| 195 | error = xfs_initialize_perag(mp, nagcount, &nagimax); |
| 196 | if (error) |
| 197 | return error; |
| 198 | } |
| 199 | |
| 200 | tp = xfs_trans_alloc(mp, XFS_TRANS_GROWFS); |
| 201 | tp->t_flags |= XFS_TRANS_RESERVE; |
| 202 | if ((error = xfs_trans_reserve(tp, XFS_GROWFS_SPACE_RES(mp), |
| 203 | XFS_GROWDATA_LOG_RES(mp), 0, 0, 0))) { |
| 204 | xfs_trans_cancel(tp, 0); |
| 205 | return error; |
| 206 | } |
| 207 | |
| 208 | /* |
| 209 | * Write new AG headers to disk. Non-transactional, but written |
| 210 | * synchronously so they are completed prior to the growfs transaction |
| 211 | * being logged. |
| 212 | */ |
| 213 | nfree = 0; |
| 214 | for (agno = nagcount - 1; agno >= oagcount; agno--, new -= agsize) { |
| 215 | /* |
| 216 | * AG freespace header block |
| 217 | */ |
| 218 | bp = xfs_growfs_get_hdr_buf(mp, |
| 219 | XFS_AG_DADDR(mp, agno, XFS_AGF_DADDR(mp)), |
| 220 | XFS_FSS_TO_BB(mp, 1), 0); |
| 221 | if (!bp) { |
| 222 | error = ENOMEM; |
| 223 | goto error0; |
| 224 | } |
| 225 | |
| 226 | agf = XFS_BUF_TO_AGF(bp); |
| 227 | agf->agf_magicnum = cpu_to_be32(XFS_AGF_MAGIC); |
| 228 | agf->agf_versionnum = cpu_to_be32(XFS_AGF_VERSION); |
| 229 | agf->agf_seqno = cpu_to_be32(agno); |
| 230 | if (agno == nagcount - 1) |
| 231 | agsize = |
| 232 | nb - |
| 233 | (agno * (xfs_rfsblock_t)mp->m_sb.sb_agblocks); |
| 234 | else |
| 235 | agsize = mp->m_sb.sb_agblocks; |
| 236 | agf->agf_length = cpu_to_be32(agsize); |
| 237 | agf->agf_roots[XFS_BTNUM_BNOi] = cpu_to_be32(XFS_BNO_BLOCK(mp)); |
| 238 | agf->agf_roots[XFS_BTNUM_CNTi] = cpu_to_be32(XFS_CNT_BLOCK(mp)); |
| 239 | agf->agf_levels[XFS_BTNUM_BNOi] = cpu_to_be32(1); |
| 240 | agf->agf_levels[XFS_BTNUM_CNTi] = cpu_to_be32(1); |
| 241 | agf->agf_flfirst = 0; |
| 242 | agf->agf_fllast = cpu_to_be32(XFS_AGFL_SIZE(mp) - 1); |
| 243 | agf->agf_flcount = 0; |
| 244 | tmpsize = agsize - XFS_PREALLOC_BLOCKS(mp); |
| 245 | agf->agf_freeblks = cpu_to_be32(tmpsize); |
| 246 | agf->agf_longest = cpu_to_be32(tmpsize); |
| 247 | error = xfs_bwrite(bp); |
| 248 | xfs_buf_relse(bp); |
| 249 | if (error) |
| 250 | goto error0; |
| 251 | |
| 252 | /* |
| 253 | * AG freelist header block |
| 254 | */ |
| 255 | bp = xfs_growfs_get_hdr_buf(mp, |
| 256 | XFS_AG_DADDR(mp, agno, XFS_AGFL_DADDR(mp)), |
| 257 | XFS_FSS_TO_BB(mp, 1), 0); |
| 258 | if (!bp) { |
| 259 | error = ENOMEM; |
| 260 | goto error0; |
| 261 | } |
| 262 | |
| 263 | agfl = XFS_BUF_TO_AGFL(bp); |
| 264 | for (bucket = 0; bucket < XFS_AGFL_SIZE(mp); bucket++) |
| 265 | agfl->agfl_bno[bucket] = cpu_to_be32(NULLAGBLOCK); |
| 266 | |
| 267 | error = xfs_bwrite(bp); |
| 268 | xfs_buf_relse(bp); |
| 269 | if (error) |
| 270 | goto error0; |
| 271 | |
| 272 | /* |
| 273 | * AG inode header block |
| 274 | */ |
| 275 | bp = xfs_growfs_get_hdr_buf(mp, |
| 276 | XFS_AG_DADDR(mp, agno, XFS_AGI_DADDR(mp)), |
| 277 | XFS_FSS_TO_BB(mp, 1), 0); |
| 278 | if (!bp) { |
| 279 | error = ENOMEM; |
| 280 | goto error0; |
| 281 | } |
| 282 | |
| 283 | agi = XFS_BUF_TO_AGI(bp); |
| 284 | agi->agi_magicnum = cpu_to_be32(XFS_AGI_MAGIC); |
| 285 | agi->agi_versionnum = cpu_to_be32(XFS_AGI_VERSION); |
| 286 | agi->agi_seqno = cpu_to_be32(agno); |
| 287 | agi->agi_length = cpu_to_be32(agsize); |
| 288 | agi->agi_count = 0; |
| 289 | agi->agi_root = cpu_to_be32(XFS_IBT_BLOCK(mp)); |
| 290 | agi->agi_level = cpu_to_be32(1); |
| 291 | agi->agi_freecount = 0; |
| 292 | agi->agi_newino = cpu_to_be32(NULLAGINO); |
| 293 | agi->agi_dirino = cpu_to_be32(NULLAGINO); |
| 294 | for (bucket = 0; bucket < XFS_AGI_UNLINKED_BUCKETS; bucket++) |
| 295 | agi->agi_unlinked[bucket] = cpu_to_be32(NULLAGINO); |
| 296 | error = xfs_bwrite(bp); |
| 297 | xfs_buf_relse(bp); |
| 298 | if (error) |
| 299 | goto error0; |
| 300 | |
| 301 | /* |
| 302 | * BNO btree root block |
| 303 | */ |
| 304 | bp = xfs_growfs_get_hdr_buf(mp, |
| 305 | XFS_AGB_TO_DADDR(mp, agno, XFS_BNO_BLOCK(mp)), |
| 306 | BTOBB(mp->m_sb.sb_blocksize), 0); |
| 307 | |
| 308 | if (!bp) { |
| 309 | error = ENOMEM; |
| 310 | goto error0; |
| 311 | } |
| 312 | |
| 313 | xfs_btree_init_block(mp, bp, XFS_ABTB_MAGIC, 0, 1, 0); |
| 314 | arec = XFS_ALLOC_REC_ADDR(mp, XFS_BUF_TO_BLOCK(bp), 1); |
| 315 | arec->ar_startblock = cpu_to_be32(XFS_PREALLOC_BLOCKS(mp)); |
| 316 | arec->ar_blockcount = cpu_to_be32( |
| 317 | agsize - be32_to_cpu(arec->ar_startblock)); |
| 318 | |
| 319 | error = xfs_bwrite(bp); |
| 320 | xfs_buf_relse(bp); |
| 321 | if (error) |
| 322 | goto error0; |
| 323 | |
| 324 | /* |
| 325 | * CNT btree root block |
| 326 | */ |
| 327 | bp = xfs_growfs_get_hdr_buf(mp, |
| 328 | XFS_AGB_TO_DADDR(mp, agno, XFS_CNT_BLOCK(mp)), |
| 329 | BTOBB(mp->m_sb.sb_blocksize), 0); |
| 330 | if (!bp) { |
| 331 | error = ENOMEM; |
| 332 | goto error0; |
| 333 | } |
| 334 | |
| 335 | xfs_btree_init_block(mp, bp, XFS_ABTC_MAGIC, 0, 1, 0); |
| 336 | arec = XFS_ALLOC_REC_ADDR(mp, XFS_BUF_TO_BLOCK(bp), 1); |
| 337 | arec->ar_startblock = cpu_to_be32(XFS_PREALLOC_BLOCKS(mp)); |
| 338 | arec->ar_blockcount = cpu_to_be32( |
| 339 | agsize - be32_to_cpu(arec->ar_startblock)); |
| 340 | nfree += be32_to_cpu(arec->ar_blockcount); |
| 341 | |
| 342 | error = xfs_bwrite(bp); |
| 343 | xfs_buf_relse(bp); |
| 344 | if (error) |
| 345 | goto error0; |
| 346 | |
| 347 | /* |
| 348 | * INO btree root block |
| 349 | */ |
| 350 | bp = xfs_growfs_get_hdr_buf(mp, |
| 351 | XFS_AGB_TO_DADDR(mp, agno, XFS_IBT_BLOCK(mp)), |
| 352 | BTOBB(mp->m_sb.sb_blocksize), 0); |
| 353 | if (!bp) { |
| 354 | error = ENOMEM; |
| 355 | goto error0; |
| 356 | } |
| 357 | |
| 358 | xfs_btree_init_block(mp, bp, XFS_IBT_MAGIC, 0, 0, 0); |
| 359 | |
| 360 | error = xfs_bwrite(bp); |
| 361 | xfs_buf_relse(bp); |
| 362 | if (error) |
| 363 | goto error0; |
| 364 | } |
| 365 | xfs_trans_agblocks_delta(tp, nfree); |
| 366 | /* |
| 367 | * There are new blocks in the old last a.g. |
| 368 | */ |
| 369 | if (new) { |
| 370 | /* |
| 371 | * Change the agi length. |
| 372 | */ |
| 373 | error = xfs_ialloc_read_agi(mp, tp, agno, &bp); |
| 374 | if (error) { |
| 375 | goto error0; |
| 376 | } |
| 377 | ASSERT(bp); |
| 378 | agi = XFS_BUF_TO_AGI(bp); |
| 379 | be32_add_cpu(&agi->agi_length, new); |
| 380 | ASSERT(nagcount == oagcount || |
| 381 | be32_to_cpu(agi->agi_length) == mp->m_sb.sb_agblocks); |
| 382 | xfs_ialloc_log_agi(tp, bp, XFS_AGI_LENGTH); |
| 383 | /* |
| 384 | * Change agf length. |
| 385 | */ |
| 386 | error = xfs_alloc_read_agf(mp, tp, agno, 0, &bp); |
| 387 | if (error) { |
| 388 | goto error0; |
| 389 | } |
| 390 | ASSERT(bp); |
| 391 | agf = XFS_BUF_TO_AGF(bp); |
| 392 | be32_add_cpu(&agf->agf_length, new); |
| 393 | ASSERT(be32_to_cpu(agf->agf_length) == |
| 394 | be32_to_cpu(agi->agi_length)); |
| 395 | |
| 396 | xfs_alloc_log_agf(tp, bp, XFS_AGF_LENGTH); |
| 397 | /* |
| 398 | * Free the new space. |
| 399 | */ |
| 400 | error = xfs_free_extent(tp, XFS_AGB_TO_FSB(mp, agno, |
| 401 | be32_to_cpu(agf->agf_length) - new), new); |
| 402 | if (error) { |
| 403 | goto error0; |
| 404 | } |
| 405 | } |
| 406 | |
| 407 | /* |
| 408 | * Update changed superblock fields transactionally. These are not |
| 409 | * seen by the rest of the world until the transaction commit applies |
| 410 | * them atomically to the superblock. |
| 411 | */ |
| 412 | if (nagcount > oagcount) |
| 413 | xfs_trans_mod_sb(tp, XFS_TRANS_SB_AGCOUNT, nagcount - oagcount); |
| 414 | if (nb > mp->m_sb.sb_dblocks) |
| 415 | xfs_trans_mod_sb(tp, XFS_TRANS_SB_DBLOCKS, |
| 416 | nb - mp->m_sb.sb_dblocks); |
| 417 | if (nfree) |
| 418 | xfs_trans_mod_sb(tp, XFS_TRANS_SB_FDBLOCKS, nfree); |
| 419 | if (dpct) |
| 420 | xfs_trans_mod_sb(tp, XFS_TRANS_SB_IMAXPCT, dpct); |
| 421 | error = xfs_trans_commit(tp, 0); |
| 422 | if (error) |
| 423 | return error; |
| 424 | |
| 425 | /* New allocation groups fully initialized, so update mount struct */ |
| 426 | if (nagimax) |
| 427 | mp->m_maxagi = nagimax; |
| 428 | if (mp->m_sb.sb_imax_pct) { |
| 429 | __uint64_t icount = mp->m_sb.sb_dblocks * mp->m_sb.sb_imax_pct; |
| 430 | do_div(icount, 100); |
| 431 | mp->m_maxicount = icount << mp->m_sb.sb_inopblog; |
| 432 | } else |
| 433 | mp->m_maxicount = 0; |
| 434 | xfs_set_low_space_thresholds(mp); |
| 435 | |
| 436 | /* update secondary superblocks. */ |
| 437 | for (agno = 1; agno < nagcount; agno++) { |
| 438 | error = 0; |
| 439 | /* |
| 440 | * new secondary superblocks need to be zeroed, not read from |
| 441 | * disk as the contents of the new area we are growing into is |
| 442 | * completely unknown. |
| 443 | */ |
| 444 | if (agno < oagcount) { |
| 445 | error = xfs_trans_read_buf(mp, NULL, mp->m_ddev_targp, |
| 446 | XFS_AGB_TO_DADDR(mp, agno, XFS_SB_BLOCK(mp)), |
| 447 | XFS_FSS_TO_BB(mp, 1), 0, &bp, |
| 448 | xfs_sb_read_verify); |
| 449 | } else { |
| 450 | bp = xfs_trans_get_buf(NULL, mp->m_ddev_targp, |
| 451 | XFS_AGB_TO_DADDR(mp, agno, XFS_SB_BLOCK(mp)), |
| 452 | XFS_FSS_TO_BB(mp, 1), 0); |
| 453 | if (bp) |
| 454 | xfs_buf_zero(bp, 0, BBTOB(bp->b_length)); |
| 455 | else |
| 456 | error = ENOMEM; |
| 457 | } |
| 458 | |
| 459 | if (error) { |
| 460 | xfs_warn(mp, |
| 461 | "error %d reading secondary superblock for ag %d", |
| 462 | error, agno); |
| 463 | break; |
| 464 | } |
| 465 | xfs_sb_to_disk(XFS_BUF_TO_SBP(bp), &mp->m_sb, XFS_SB_ALL_BITS); |
| 466 | |
| 467 | /* |
| 468 | * If we get an error writing out the alternate superblocks, |
| 469 | * just issue a warning and continue. The real work is |
| 470 | * already done and committed. |
| 471 | */ |
| 472 | error = xfs_bwrite(bp); |
| 473 | xfs_buf_relse(bp); |
| 474 | if (error) { |
| 475 | xfs_warn(mp, |
| 476 | "write error %d updating secondary superblock for ag %d", |
| 477 | error, agno); |
| 478 | break; /* no point in continuing */ |
| 479 | } |
| 480 | } |
| 481 | return error; |
| 482 | |
| 483 | error0: |
| 484 | xfs_trans_cancel(tp, XFS_TRANS_ABORT); |
| 485 | return error; |
| 486 | } |
| 487 | |
| 488 | static int |
| 489 | xfs_growfs_log_private( |
| 490 | xfs_mount_t *mp, /* mount point for filesystem */ |
| 491 | xfs_growfs_log_t *in) /* growfs log input struct */ |
| 492 | { |
| 493 | xfs_extlen_t nb; |
| 494 | |
| 495 | nb = in->newblocks; |
| 496 | if (nb < XFS_MIN_LOG_BLOCKS || nb < XFS_B_TO_FSB(mp, XFS_MIN_LOG_BYTES)) |
| 497 | return XFS_ERROR(EINVAL); |
| 498 | if (nb == mp->m_sb.sb_logblocks && |
| 499 | in->isint == (mp->m_sb.sb_logstart != 0)) |
| 500 | return XFS_ERROR(EINVAL); |
| 501 | /* |
| 502 | * Moving the log is hard, need new interfaces to sync |
| 503 | * the log first, hold off all activity while moving it. |
| 504 | * Can have shorter or longer log in the same space, |
| 505 | * or transform internal to external log or vice versa. |
| 506 | */ |
| 507 | return XFS_ERROR(ENOSYS); |
| 508 | } |
| 509 | |
| 510 | /* |
| 511 | * protected versions of growfs function acquire and release locks on the mount |
| 512 | * point - exported through ioctls: XFS_IOC_FSGROWFSDATA, XFS_IOC_FSGROWFSLOG, |
| 513 | * XFS_IOC_FSGROWFSRT |
| 514 | */ |
| 515 | |
| 516 | |
| 517 | int |
| 518 | xfs_growfs_data( |
| 519 | xfs_mount_t *mp, |
| 520 | xfs_growfs_data_t *in) |
| 521 | { |
| 522 | int error; |
| 523 | |
| 524 | if (!capable(CAP_SYS_ADMIN)) |
| 525 | return XFS_ERROR(EPERM); |
| 526 | if (!mutex_trylock(&mp->m_growlock)) |
| 527 | return XFS_ERROR(EWOULDBLOCK); |
| 528 | error = xfs_growfs_data_private(mp, in); |
| 529 | mutex_unlock(&mp->m_growlock); |
| 530 | return error; |
| 531 | } |
| 532 | |
| 533 | int |
| 534 | xfs_growfs_log( |
| 535 | xfs_mount_t *mp, |
| 536 | xfs_growfs_log_t *in) |
| 537 | { |
| 538 | int error; |
| 539 | |
| 540 | if (!capable(CAP_SYS_ADMIN)) |
| 541 | return XFS_ERROR(EPERM); |
| 542 | if (!mutex_trylock(&mp->m_growlock)) |
| 543 | return XFS_ERROR(EWOULDBLOCK); |
| 544 | error = xfs_growfs_log_private(mp, in); |
| 545 | mutex_unlock(&mp->m_growlock); |
| 546 | return error; |
| 547 | } |
| 548 | |
| 549 | /* |
| 550 | * exported through ioctl XFS_IOC_FSCOUNTS |
| 551 | */ |
| 552 | |
| 553 | int |
| 554 | xfs_fs_counts( |
| 555 | xfs_mount_t *mp, |
| 556 | xfs_fsop_counts_t *cnt) |
| 557 | { |
| 558 | xfs_icsb_sync_counters(mp, XFS_ICSB_LAZY_COUNT); |
| 559 | spin_lock(&mp->m_sb_lock); |
| 560 | cnt->freedata = mp->m_sb.sb_fdblocks - XFS_ALLOC_SET_ASIDE(mp); |
| 561 | cnt->freertx = mp->m_sb.sb_frextents; |
| 562 | cnt->freeino = mp->m_sb.sb_ifree; |
| 563 | cnt->allocino = mp->m_sb.sb_icount; |
| 564 | spin_unlock(&mp->m_sb_lock); |
| 565 | return 0; |
| 566 | } |
| 567 | |
| 568 | /* |
| 569 | * exported through ioctl XFS_IOC_SET_RESBLKS & XFS_IOC_GET_RESBLKS |
| 570 | * |
| 571 | * xfs_reserve_blocks is called to set m_resblks |
| 572 | * in the in-core mount table. The number of unused reserved blocks |
| 573 | * is kept in m_resblks_avail. |
| 574 | * |
| 575 | * Reserve the requested number of blocks if available. Otherwise return |
| 576 | * as many as possible to satisfy the request. The actual number |
| 577 | * reserved are returned in outval |
| 578 | * |
| 579 | * A null inval pointer indicates that only the current reserved blocks |
| 580 | * available should be returned no settings are changed. |
| 581 | */ |
| 582 | |
| 583 | int |
| 584 | xfs_reserve_blocks( |
| 585 | xfs_mount_t *mp, |
| 586 | __uint64_t *inval, |
| 587 | xfs_fsop_resblks_t *outval) |
| 588 | { |
| 589 | __int64_t lcounter, delta, fdblks_delta; |
| 590 | __uint64_t request; |
| 591 | |
| 592 | /* If inval is null, report current values and return */ |
| 593 | if (inval == (__uint64_t *)NULL) { |
| 594 | if (!outval) |
| 595 | return EINVAL; |
| 596 | outval->resblks = mp->m_resblks; |
| 597 | outval->resblks_avail = mp->m_resblks_avail; |
| 598 | return 0; |
| 599 | } |
| 600 | |
| 601 | request = *inval; |
| 602 | |
| 603 | /* |
| 604 | * With per-cpu counters, this becomes an interesting |
| 605 | * problem. we needto work out if we are freeing or allocation |
| 606 | * blocks first, then we can do the modification as necessary. |
| 607 | * |
| 608 | * We do this under the m_sb_lock so that if we are near |
| 609 | * ENOSPC, we will hold out any changes while we work out |
| 610 | * what to do. This means that the amount of free space can |
| 611 | * change while we do this, so we need to retry if we end up |
| 612 | * trying to reserve more space than is available. |
| 613 | * |
| 614 | * We also use the xfs_mod_incore_sb() interface so that we |
| 615 | * don't have to care about whether per cpu counter are |
| 616 | * enabled, disabled or even compiled in.... |
| 617 | */ |
| 618 | retry: |
| 619 | spin_lock(&mp->m_sb_lock); |
| 620 | xfs_icsb_sync_counters_locked(mp, 0); |
| 621 | |
| 622 | /* |
| 623 | * If our previous reservation was larger than the current value, |
| 624 | * then move any unused blocks back to the free pool. |
| 625 | */ |
| 626 | fdblks_delta = 0; |
| 627 | if (mp->m_resblks > request) { |
| 628 | lcounter = mp->m_resblks_avail - request; |
| 629 | if (lcounter > 0) { /* release unused blocks */ |
| 630 | fdblks_delta = lcounter; |
| 631 | mp->m_resblks_avail -= lcounter; |
| 632 | } |
| 633 | mp->m_resblks = request; |
| 634 | } else { |
| 635 | __int64_t free; |
| 636 | |
| 637 | free = mp->m_sb.sb_fdblocks - XFS_ALLOC_SET_ASIDE(mp); |
| 638 | if (!free) |
| 639 | goto out; /* ENOSPC and fdblks_delta = 0 */ |
| 640 | |
| 641 | delta = request - mp->m_resblks; |
| 642 | lcounter = free - delta; |
| 643 | if (lcounter < 0) { |
| 644 | /* We can't satisfy the request, just get what we can */ |
| 645 | mp->m_resblks += free; |
| 646 | mp->m_resblks_avail += free; |
| 647 | fdblks_delta = -free; |
| 648 | } else { |
| 649 | fdblks_delta = -delta; |
| 650 | mp->m_resblks = request; |
| 651 | mp->m_resblks_avail += delta; |
| 652 | } |
| 653 | } |
| 654 | out: |
| 655 | if (outval) { |
| 656 | outval->resblks = mp->m_resblks; |
| 657 | outval->resblks_avail = mp->m_resblks_avail; |
| 658 | } |
| 659 | spin_unlock(&mp->m_sb_lock); |
| 660 | |
| 661 | if (fdblks_delta) { |
| 662 | /* |
| 663 | * If we are putting blocks back here, m_resblks_avail is |
| 664 | * already at its max so this will put it in the free pool. |
| 665 | * |
| 666 | * If we need space, we'll either succeed in getting it |
| 667 | * from the free block count or we'll get an enospc. If |
| 668 | * we get a ENOSPC, it means things changed while we were |
| 669 | * calculating fdblks_delta and so we should try again to |
| 670 | * see if there is anything left to reserve. |
| 671 | * |
| 672 | * Don't set the reserved flag here - we don't want to reserve |
| 673 | * the extra reserve blocks from the reserve..... |
| 674 | */ |
| 675 | int error; |
| 676 | error = xfs_icsb_modify_counters(mp, XFS_SBS_FDBLOCKS, |
| 677 | fdblks_delta, 0); |
| 678 | if (error == ENOSPC) |
| 679 | goto retry; |
| 680 | } |
| 681 | return 0; |
| 682 | } |
| 683 | |
| 684 | /* |
| 685 | * Dump a transaction into the log that contains no real change. This is needed |
| 686 | * to be able to make the log dirty or stamp the current tail LSN into the log |
| 687 | * during the covering operation. |
| 688 | * |
| 689 | * We cannot use an inode here for this - that will push dirty state back up |
| 690 | * into the VFS and then periodic inode flushing will prevent log covering from |
| 691 | * making progress. Hence we log a field in the superblock instead and use a |
| 692 | * synchronous transaction to ensure the superblock is immediately unpinned |
| 693 | * and can be written back. |
| 694 | */ |
| 695 | int |
| 696 | xfs_fs_log_dummy( |
| 697 | xfs_mount_t *mp) |
| 698 | { |
| 699 | xfs_trans_t *tp; |
| 700 | int error; |
| 701 | |
| 702 | tp = _xfs_trans_alloc(mp, XFS_TRANS_DUMMY1, KM_SLEEP); |
| 703 | error = xfs_trans_reserve(tp, 0, mp->m_sb.sb_sectsize + 128, 0, 0, |
| 704 | XFS_DEFAULT_LOG_COUNT); |
| 705 | if (error) { |
| 706 | xfs_trans_cancel(tp, 0); |
| 707 | return error; |
| 708 | } |
| 709 | |
| 710 | /* log the UUID because it is an unchanging field */ |
| 711 | xfs_mod_sb(tp, XFS_SB_UUID); |
| 712 | xfs_trans_set_sync(tp); |
| 713 | return xfs_trans_commit(tp, 0); |
| 714 | } |
| 715 | |
| 716 | int |
| 717 | xfs_fs_goingdown( |
| 718 | xfs_mount_t *mp, |
| 719 | __uint32_t inflags) |
| 720 | { |
| 721 | switch (inflags) { |
| 722 | case XFS_FSOP_GOING_FLAGS_DEFAULT: { |
| 723 | struct super_block *sb = freeze_bdev(mp->m_super->s_bdev); |
| 724 | |
| 725 | if (sb && !IS_ERR(sb)) { |
| 726 | xfs_force_shutdown(mp, SHUTDOWN_FORCE_UMOUNT); |
| 727 | thaw_bdev(sb->s_bdev, sb); |
| 728 | } |
| 729 | |
| 730 | break; |
| 731 | } |
| 732 | case XFS_FSOP_GOING_FLAGS_LOGFLUSH: |
| 733 | xfs_force_shutdown(mp, SHUTDOWN_FORCE_UMOUNT); |
| 734 | break; |
| 735 | case XFS_FSOP_GOING_FLAGS_NOLOGFLUSH: |
| 736 | xfs_force_shutdown(mp, |
| 737 | SHUTDOWN_FORCE_UMOUNT | SHUTDOWN_LOG_IO_ERROR); |
| 738 | break; |
| 739 | default: |
| 740 | return XFS_ERROR(EINVAL); |
| 741 | } |
| 742 | |
| 743 | return 0; |
| 744 | } |
| 745 | |
| 746 | /* |
| 747 | * Force a shutdown of the filesystem instantly while keeping the filesystem |
| 748 | * consistent. We don't do an unmount here; just shutdown the shop, make sure |
| 749 | * that absolutely nothing persistent happens to this filesystem after this |
| 750 | * point. |
| 751 | */ |
| 752 | void |
| 753 | xfs_do_force_shutdown( |
| 754 | xfs_mount_t *mp, |
| 755 | int flags, |
| 756 | char *fname, |
| 757 | int lnnum) |
| 758 | { |
| 759 | int logerror; |
| 760 | |
| 761 | logerror = flags & SHUTDOWN_LOG_IO_ERROR; |
| 762 | |
| 763 | if (!(flags & SHUTDOWN_FORCE_UMOUNT)) { |
| 764 | xfs_notice(mp, |
| 765 | "%s(0x%x) called from line %d of file %s. Return address = 0x%p", |
| 766 | __func__, flags, lnnum, fname, __return_address); |
| 767 | } |
| 768 | /* |
| 769 | * No need to duplicate efforts. |
| 770 | */ |
| 771 | if (XFS_FORCED_SHUTDOWN(mp) && !logerror) |
| 772 | return; |
| 773 | |
| 774 | /* |
| 775 | * This flags XFS_MOUNT_FS_SHUTDOWN, makes sure that we don't |
| 776 | * queue up anybody new on the log reservations, and wakes up |
| 777 | * everybody who's sleeping on log reservations to tell them |
| 778 | * the bad news. |
| 779 | */ |
| 780 | if (xfs_log_force_umount(mp, logerror)) |
| 781 | return; |
| 782 | |
| 783 | if (flags & SHUTDOWN_CORRUPT_INCORE) { |
| 784 | xfs_alert_tag(mp, XFS_PTAG_SHUTDOWN_CORRUPT, |
| 785 | "Corruption of in-memory data detected. Shutting down filesystem"); |
| 786 | if (XFS_ERRLEVEL_HIGH <= xfs_error_level) |
| 787 | xfs_stack_trace(); |
| 788 | } else if (!(flags & SHUTDOWN_FORCE_UMOUNT)) { |
| 789 | if (logerror) { |
| 790 | xfs_alert_tag(mp, XFS_PTAG_SHUTDOWN_LOGERROR, |
| 791 | "Log I/O Error Detected. Shutting down filesystem"); |
| 792 | } else if (flags & SHUTDOWN_DEVICE_REQ) { |
| 793 | xfs_alert_tag(mp, XFS_PTAG_SHUTDOWN_IOERROR, |
| 794 | "All device paths lost. Shutting down filesystem"); |
| 795 | } else if (!(flags & SHUTDOWN_REMOTE_REQ)) { |
| 796 | xfs_alert_tag(mp, XFS_PTAG_SHUTDOWN_IOERROR, |
| 797 | "I/O Error Detected. Shutting down filesystem"); |
| 798 | } |
| 799 | } |
| 800 | if (!(flags & SHUTDOWN_FORCE_UMOUNT)) { |
| 801 | xfs_alert(mp, |
| 802 | "Please umount the filesystem and rectify the problem(s)"); |
| 803 | } |
| 804 | } |