1// SPDX-License-Identifier: GPL-2.0-or-later 2/* -*- mode: c; c-basic-offset: 8; -*- 3 * vim: noexpandtab sw=8 ts=8 sts=0: 4 * 5 * resize.c 6 * 7 * volume resize. 8 * Inspired by ext3/resize.c. 9 * 10 * Copyright (C) 2007 Oracle. All rights reserved. 11 */ 12 13#include <linux/fs.h> 14#include <linux/types.h> 15 16#include <cluster/masklog.h> 17 18#include "ocfs2.h" 19 20#include "alloc.h" 21#include "dlmglue.h" 22#include "inode.h" 23#include "journal.h" 24#include "super.h" 25#include "sysfile.h" 26#include "uptodate.h" 27#include "ocfs2_trace.h" 28 29#include "buffer_head_io.h" 30#include "suballoc.h" 31#include "resize.h" 32 33/* 34 * Check whether there are new backup superblocks exist 35 * in the last group. If there are some, mark them or clear 36 * them in the bitmap. 37 * 38 * Return how many backups we find in the last group. 39 */ 40static u16 ocfs2_calc_new_backup_super(struct inode *inode, 41 struct ocfs2_group_desc *gd, 42 u16 cl_cpg, 43 u16 old_bg_clusters, 44 int set) 45{ 46 int i; 47 u16 backups = 0; 48 u32 cluster, lgd_cluster; 49 u64 blkno, gd_blkno, lgd_blkno = le64_to_cpu(gd->bg_blkno); 50 51 for (i = 0; i < OCFS2_MAX_BACKUP_SUPERBLOCKS; i++) { 52 blkno = ocfs2_backup_super_blkno(inode->i_sb, i); 53 cluster = ocfs2_blocks_to_clusters(inode->i_sb, blkno); 54 55 gd_blkno = ocfs2_which_cluster_group(inode, cluster); 56 if (gd_blkno < lgd_blkno) 57 continue; 58 else if (gd_blkno > lgd_blkno) 59 break; 60 61 /* check if already done backup super */ 62 lgd_cluster = ocfs2_blocks_to_clusters(inode->i_sb, lgd_blkno); 63 lgd_cluster += old_bg_clusters; 64 if (lgd_cluster >= cluster) 65 continue; 66 67 if (set) 68 ocfs2_set_bit(cluster % cl_cpg, 69 (unsigned long *)gd->bg_bitmap); 70 else 71 ocfs2_clear_bit(cluster % cl_cpg, 72 (unsigned long *)gd->bg_bitmap); 73 backups++; 74 } 75 76 return backups; 77} 78 79static int ocfs2_update_last_group_and_inode(handle_t *handle, 80 struct inode *bm_inode, 81 struct buffer_head *bm_bh, 82 struct buffer_head *group_bh, 83 u32 first_new_cluster, 84 int new_clusters) 85{ 86 int ret = 0; 87 struct ocfs2_super *osb = OCFS2_SB(bm_inode->i_sb); 88 struct ocfs2_dinode *fe = (struct ocfs2_dinode *) bm_bh->b_data; 89 struct ocfs2_chain_list *cl = &fe->id2.i_chain; 90 struct ocfs2_chain_rec *cr; 91 struct ocfs2_group_desc *group; 92 u16 chain, num_bits, backups = 0; 93 u16 cl_bpc = le16_to_cpu(cl->cl_bpc); 94 u16 cl_cpg = le16_to_cpu(cl->cl_cpg); 95 u16 old_bg_clusters; 96 97 trace_ocfs2_update_last_group_and_inode(new_clusters, 98 first_new_cluster); 99 100 ret = ocfs2_journal_access_gd(handle, INODE_CACHE(bm_inode), 101 group_bh, OCFS2_JOURNAL_ACCESS_WRITE); 102 if (ret < 0) { 103 mlog_errno(ret); 104 goto out; 105 } 106 107 group = (struct ocfs2_group_desc *)group_bh->b_data; 108 109 old_bg_clusters = le16_to_cpu(group->bg_bits) / cl_bpc; 110 /* update the group first. */ 111 num_bits = new_clusters * cl_bpc; 112 le16_add_cpu(&group->bg_bits, num_bits); 113 le16_add_cpu(&group->bg_free_bits_count, num_bits); 114 115 /* 116 * check whether there are some new backup superblocks exist in 117 * this group and update the group bitmap accordingly. 118 */ 119 if (OCFS2_HAS_COMPAT_FEATURE(osb->sb, 120 OCFS2_FEATURE_COMPAT_BACKUP_SB)) { 121 backups = ocfs2_calc_new_backup_super(bm_inode, 122 group, 123 cl_cpg, old_bg_clusters, 1); 124 le16_add_cpu(&group->bg_free_bits_count, -1 * backups); 125 } 126 127 ocfs2_journal_dirty(handle, group_bh); 128 129 /* update the inode accordingly. */ 130 ret = ocfs2_journal_access_di(handle, INODE_CACHE(bm_inode), bm_bh, 131 OCFS2_JOURNAL_ACCESS_WRITE); 132 if (ret < 0) { 133 mlog_errno(ret); 134 goto out_rollback; 135 } 136 137 chain = le16_to_cpu(group->bg_chain); 138 cr = (&cl->cl_recs[chain]); 139 le32_add_cpu(&cr->c_total, num_bits); 140 le32_add_cpu(&cr->c_free, num_bits); 141 le32_add_cpu(&fe->id1.bitmap1.i_total, num_bits); 142 le32_add_cpu(&fe->i_clusters, new_clusters); 143 144 if (backups) { 145 le32_add_cpu(&cr->c_free, -1 * backups); 146 le32_add_cpu(&fe->id1.bitmap1.i_used, backups); 147 } 148 149 spin_lock(&OCFS2_I(bm_inode)->ip_lock); 150 OCFS2_I(bm_inode)->ip_clusters = le32_to_cpu(fe->i_clusters); 151 le64_add_cpu(&fe->i_size, (u64)new_clusters << osb->s_clustersize_bits); 152 spin_unlock(&OCFS2_I(bm_inode)->ip_lock); 153 i_size_write(bm_inode, le64_to_cpu(fe->i_size)); 154 155 ocfs2_journal_dirty(handle, bm_bh); 156 157out_rollback: 158 if (ret < 0) { 159 ocfs2_calc_new_backup_super(bm_inode, 160 group, 161 cl_cpg, old_bg_clusters, 0); 162 le16_add_cpu(&group->bg_free_bits_count, backups); 163 le16_add_cpu(&group->bg_bits, -1 * num_bits); 164 le16_add_cpu(&group->bg_free_bits_count, -1 * num_bits); 165 } 166out: 167 if (ret) 168 mlog_errno(ret); 169 return ret; 170} 171 172static int update_backups(struct inode * inode, u32 clusters, char *data) 173{ 174 int i, ret = 0; 175 u32 cluster; 176 u64 blkno; 177 struct buffer_head *backup = NULL; 178 struct ocfs2_dinode *backup_di = NULL; 179 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); 180 181 /* calculate the real backups we need to update. */ 182 for (i = 0; i < OCFS2_MAX_BACKUP_SUPERBLOCKS; i++) { 183 blkno = ocfs2_backup_super_blkno(inode->i_sb, i); 184 cluster = ocfs2_blocks_to_clusters(inode->i_sb, blkno); 185 if (cluster >= clusters) 186 break; 187 188 ret = ocfs2_read_blocks_sync(osb, blkno, 1, &backup); 189 if (ret < 0) { 190 mlog_errno(ret); 191 break; 192 } 193 194 memcpy(backup->b_data, data, inode->i_sb->s_blocksize); 195 196 backup_di = (struct ocfs2_dinode *)backup->b_data; 197 backup_di->i_blkno = cpu_to_le64(blkno); 198 199 ret = ocfs2_write_super_or_backup(osb, backup); 200 brelse(backup); 201 backup = NULL; 202 if (ret < 0) { 203 mlog_errno(ret); 204 break; 205 } 206 } 207 208 return ret; 209} 210 211static void ocfs2_update_super_and_backups(struct inode *inode, 212 int new_clusters) 213{ 214 int ret; 215 u32 clusters = 0; 216 struct buffer_head *super_bh = NULL; 217 struct ocfs2_dinode *super_di = NULL; 218 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); 219 220 /* 221 * update the superblock last. 222 * It doesn't matter if the write failed. 223 */ 224 ret = ocfs2_read_blocks_sync(osb, OCFS2_SUPER_BLOCK_BLKNO, 1, 225 &super_bh); 226 if (ret < 0) { 227 mlog_errno(ret); 228 goto out; 229 } 230 231 super_di = (struct ocfs2_dinode *)super_bh->b_data; 232 le32_add_cpu(&super_di->i_clusters, new_clusters); 233 clusters = le32_to_cpu(super_di->i_clusters); 234 235 ret = ocfs2_write_super_or_backup(osb, super_bh); 236 if (ret < 0) { 237 mlog_errno(ret); 238 goto out; 239 } 240 241 if (OCFS2_HAS_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_COMPAT_BACKUP_SB)) 242 ret = update_backups(inode, clusters, super_bh->b_data); 243 244out: 245 brelse(super_bh); 246 if (ret) 247 printk(KERN_WARNING "ocfs2: Failed to update super blocks on %s" 248 " during fs resize. This condition is not fatal," 249 " but fsck.ocfs2 should be run to fix it\n", 250 osb->dev_str); 251 return; 252} 253 254/* 255 * Extend the filesystem to the new number of clusters specified. This entry 256 * point is only used to extend the current filesystem to the end of the last 257 * existing group. 258 */ 259int ocfs2_group_extend(struct inode * inode, int new_clusters) 260{ 261 int ret; 262 handle_t *handle; 263 struct buffer_head *main_bm_bh = NULL; 264 struct buffer_head *group_bh = NULL; 265 struct inode *main_bm_inode = NULL; 266 struct ocfs2_dinode *fe = NULL; 267 struct ocfs2_group_desc *group = NULL; 268 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); 269 u16 cl_bpc; 270 u32 first_new_cluster; 271 u64 lgd_blkno; 272 273 if (ocfs2_is_hard_readonly(osb) || ocfs2_is_soft_readonly(osb)) 274 return -EROFS; 275 276 if (new_clusters < 0) 277 return -EINVAL; 278 else if (new_clusters == 0) 279 return 0; 280 281 main_bm_inode = ocfs2_get_system_file_inode(osb, 282 GLOBAL_BITMAP_SYSTEM_INODE, 283 OCFS2_INVALID_SLOT); 284 if (!main_bm_inode) { 285 ret = -EINVAL; 286 mlog_errno(ret); 287 goto out; 288 } 289 290 inode_lock(main_bm_inode); 291 292 ret = ocfs2_inode_lock(main_bm_inode, &main_bm_bh, 1); 293 if (ret < 0) { 294 mlog_errno(ret); 295 goto out_mutex; 296 } 297 298 fe = (struct ocfs2_dinode *)main_bm_bh->b_data; 299 300 /* main_bm_bh is validated by inode read inside ocfs2_inode_lock(), 301 * so any corruption is a code bug. */ 302 BUG_ON(!OCFS2_IS_VALID_DINODE(fe)); 303 304 if (le16_to_cpu(fe->id2.i_chain.cl_cpg) != 305 ocfs2_group_bitmap_size(osb->sb, 0, 306 osb->s_feature_incompat) * 8) { 307 mlog(ML_ERROR, "The disk is too old and small. " 308 "Force to do offline resize."); 309 ret = -EINVAL; 310 goto out_unlock; 311 } 312 313 first_new_cluster = le32_to_cpu(fe->i_clusters); 314 lgd_blkno = ocfs2_which_cluster_group(main_bm_inode, 315 first_new_cluster - 1); 316 317 ret = ocfs2_read_group_descriptor(main_bm_inode, fe, lgd_blkno, 318 &group_bh); 319 if (ret < 0) { 320 mlog_errno(ret); 321 goto out_unlock; 322 } 323 group = (struct ocfs2_group_desc *)group_bh->b_data; 324 325 cl_bpc = le16_to_cpu(fe->id2.i_chain.cl_bpc); 326 if (le16_to_cpu(group->bg_bits) / cl_bpc + new_clusters > 327 le16_to_cpu(fe->id2.i_chain.cl_cpg)) { 328 ret = -EINVAL; 329 goto out_unlock; 330 } 331 332 333 trace_ocfs2_group_extend( 334 (unsigned long long)le64_to_cpu(group->bg_blkno), new_clusters); 335 336 handle = ocfs2_start_trans(osb, OCFS2_GROUP_EXTEND_CREDITS); 337 if (IS_ERR(handle)) { 338 mlog_errno(PTR_ERR(handle)); 339 ret = -EINVAL; 340 goto out_unlock; 341 } 342 343 /* update the last group descriptor and inode. */ 344 ret = ocfs2_update_last_group_and_inode(handle, main_bm_inode, 345 main_bm_bh, group_bh, 346 first_new_cluster, 347 new_clusters); 348 if (ret) { 349 mlog_errno(ret); 350 goto out_commit; 351 } 352 353 ocfs2_update_super_and_backups(main_bm_inode, new_clusters); 354 355out_commit: 356 ocfs2_commit_trans(osb, handle); 357out_unlock: 358 brelse(group_bh); 359 brelse(main_bm_bh); 360 361 ocfs2_inode_unlock(main_bm_inode, 1); 362 363out_mutex: 364 inode_unlock(main_bm_inode); 365 iput(main_bm_inode); 366 367out: 368 return ret; 369} 370 371static int ocfs2_check_new_group(struct inode *inode, 372 struct ocfs2_dinode *di, 373 struct ocfs2_new_group_input *input, 374 struct buffer_head *group_bh) 375{ 376 int ret; 377 struct ocfs2_group_desc *gd = 378 (struct ocfs2_group_desc *)group_bh->b_data; 379 u16 cl_bpc = le16_to_cpu(di->id2.i_chain.cl_bpc); 380 381 ret = ocfs2_check_group_descriptor(inode->i_sb, di, group_bh); 382 if (ret) 383 goto out; 384 385 ret = -EINVAL; 386 if (le16_to_cpu(gd->bg_chain) != input->chain) 387 mlog(ML_ERROR, "Group descriptor # %llu has bad chain %u " 388 "while input has %u set.\n", 389 (unsigned long long)le64_to_cpu(gd->bg_blkno), 390 le16_to_cpu(gd->bg_chain), input->chain); 391 else if (le16_to_cpu(gd->bg_bits) != input->clusters * cl_bpc) 392 mlog(ML_ERROR, "Group descriptor # %llu has bit count %u but " 393 "input has %u clusters set\n", 394 (unsigned long long)le64_to_cpu(gd->bg_blkno), 395 le16_to_cpu(gd->bg_bits), input->clusters); 396 else if (le16_to_cpu(gd->bg_free_bits_count) != input->frees * cl_bpc) 397 mlog(ML_ERROR, "Group descriptor # %llu has free bit count %u " 398 "but it should have %u set\n", 399 (unsigned long long)le64_to_cpu(gd->bg_blkno), 400 le16_to_cpu(gd->bg_bits), 401 input->frees * cl_bpc); 402 else 403 ret = 0; 404 405out: 406 return ret; 407} 408 409static int ocfs2_verify_group_and_input(struct inode *inode, 410 struct ocfs2_dinode *di, 411 struct ocfs2_new_group_input *input, 412 struct buffer_head *group_bh) 413{ 414 u16 cl_count = le16_to_cpu(di->id2.i_chain.cl_count); 415 u16 cl_cpg = le16_to_cpu(di->id2.i_chain.cl_cpg); 416 u16 next_free = le16_to_cpu(di->id2.i_chain.cl_next_free_rec); 417 u32 cluster = ocfs2_blocks_to_clusters(inode->i_sb, input->group); 418 u32 total_clusters = le32_to_cpu(di->i_clusters); 419 int ret = -EINVAL; 420 421 if (cluster < total_clusters) 422 mlog(ML_ERROR, "add a group which is in the current volume.\n"); 423 else if (input->chain >= cl_count) 424 mlog(ML_ERROR, "input chain exceeds the limit.\n"); 425 else if (next_free != cl_count && next_free != input->chain) 426 mlog(ML_ERROR, 427 "the add group should be in chain %u\n", next_free); 428 else if (total_clusters + input->clusters < total_clusters) 429 mlog(ML_ERROR, "add group's clusters overflow.\n"); 430 else if (input->clusters > cl_cpg) 431 mlog(ML_ERROR, "the cluster exceeds the maximum of a group\n"); 432 else if (input->frees > input->clusters) 433 mlog(ML_ERROR, "the free cluster exceeds the total clusters\n"); 434 else if (total_clusters % cl_cpg != 0) 435 mlog(ML_ERROR, 436 "the last group isn't full. Use group extend first.\n"); 437 else if (input->group != ocfs2_which_cluster_group(inode, cluster)) 438 mlog(ML_ERROR, "group blkno is invalid\n"); 439 else if ((ret = ocfs2_check_new_group(inode, di, input, group_bh))) 440 mlog(ML_ERROR, "group descriptor check failed.\n"); 441 else 442 ret = 0; 443 444 return ret; 445} 446 447/* Add a new group descriptor to global_bitmap. */ 448int ocfs2_group_add(struct inode *inode, struct ocfs2_new_group_input *input) 449{ 450 int ret; 451 handle_t *handle; 452 struct buffer_head *main_bm_bh = NULL; 453 struct inode *main_bm_inode = NULL; 454 struct ocfs2_dinode *fe = NULL; 455 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); 456 struct buffer_head *group_bh = NULL; 457 struct ocfs2_group_desc *group = NULL; 458 struct ocfs2_chain_list *cl; 459 struct ocfs2_chain_rec *cr; 460 u16 cl_bpc; 461 u64 bg_ptr; 462 463 if (ocfs2_is_hard_readonly(osb) || ocfs2_is_soft_readonly(osb)) 464 return -EROFS; 465 466 main_bm_inode = ocfs2_get_system_file_inode(osb, 467 GLOBAL_BITMAP_SYSTEM_INODE, 468 OCFS2_INVALID_SLOT); 469 if (!main_bm_inode) { 470 ret = -EINVAL; 471 mlog_errno(ret); 472 goto out; 473 } 474 475 inode_lock(main_bm_inode); 476 477 ret = ocfs2_inode_lock(main_bm_inode, &main_bm_bh, 1); 478 if (ret < 0) { 479 mlog_errno(ret); 480 goto out_mutex; 481 } 482 483 fe = (struct ocfs2_dinode *)main_bm_bh->b_data; 484 485 if (le16_to_cpu(fe->id2.i_chain.cl_cpg) != 486 ocfs2_group_bitmap_size(osb->sb, 0, 487 osb->s_feature_incompat) * 8) { 488 mlog(ML_ERROR, "The disk is too old and small." 489 " Force to do offline resize."); 490 ret = -EINVAL; 491 goto out_unlock; 492 } 493 494 ret = ocfs2_read_blocks_sync(osb, input->group, 1, &group_bh); 495 if (ret < 0) { 496 mlog(ML_ERROR, "Can't read the group descriptor # %llu " 497 "from the device.", (unsigned long long)input->group); 498 goto out_unlock; 499 } 500 501 ocfs2_set_new_buffer_uptodate(INODE_CACHE(inode), group_bh); 502 503 ret = ocfs2_verify_group_and_input(main_bm_inode, fe, input, group_bh); 504 if (ret) { 505 mlog_errno(ret); 506 goto out_free_group_bh; 507 } 508 509 trace_ocfs2_group_add((unsigned long long)input->group, 510 input->chain, input->clusters, input->frees); 511 512 handle = ocfs2_start_trans(osb, OCFS2_GROUP_ADD_CREDITS); 513 if (IS_ERR(handle)) { 514 mlog_errno(PTR_ERR(handle)); 515 ret = -EINVAL; 516 goto out_free_group_bh; 517 } 518 519 cl_bpc = le16_to_cpu(fe->id2.i_chain.cl_bpc); 520 cl = &fe->id2.i_chain; 521 cr = &cl->cl_recs[input->chain]; 522 523 ret = ocfs2_journal_access_gd(handle, INODE_CACHE(main_bm_inode), 524 group_bh, OCFS2_JOURNAL_ACCESS_WRITE); 525 if (ret < 0) { 526 mlog_errno(ret); 527 goto out_commit; 528 } 529 530 group = (struct ocfs2_group_desc *)group_bh->b_data; 531 bg_ptr = le64_to_cpu(group->bg_next_group); 532 group->bg_next_group = cr->c_blkno; 533 ocfs2_journal_dirty(handle, group_bh); 534 535 ret = ocfs2_journal_access_di(handle, INODE_CACHE(main_bm_inode), 536 main_bm_bh, OCFS2_JOURNAL_ACCESS_WRITE); 537 if (ret < 0) { 538 group->bg_next_group = cpu_to_le64(bg_ptr); 539 mlog_errno(ret); 540 goto out_commit; 541 } 542 543 if (input->chain == le16_to_cpu(cl->cl_next_free_rec)) { 544 le16_add_cpu(&cl->cl_next_free_rec, 1); 545 memset(cr, 0, sizeof(struct ocfs2_chain_rec)); 546 } 547 548 cr->c_blkno = cpu_to_le64(input->group); 549 le32_add_cpu(&cr->c_total, input->clusters * cl_bpc); 550 le32_add_cpu(&cr->c_free, input->frees * cl_bpc); 551 552 le32_add_cpu(&fe->id1.bitmap1.i_total, input->clusters *cl_bpc); 553 le32_add_cpu(&fe->id1.bitmap1.i_used, 554 (input->clusters - input->frees) * cl_bpc); 555 le32_add_cpu(&fe->i_clusters, input->clusters); 556 557 ocfs2_journal_dirty(handle, main_bm_bh); 558 559 spin_lock(&OCFS2_I(main_bm_inode)->ip_lock); 560 OCFS2_I(main_bm_inode)->ip_clusters = le32_to_cpu(fe->i_clusters); 561 le64_add_cpu(&fe->i_size, (u64)input->clusters << osb->s_clustersize_bits); 562 spin_unlock(&OCFS2_I(main_bm_inode)->ip_lock); 563 i_size_write(main_bm_inode, le64_to_cpu(fe->i_size)); 564 565 ocfs2_update_super_and_backups(main_bm_inode, input->clusters); 566 567out_commit: 568 ocfs2_commit_trans(osb, handle); 569 570out_free_group_bh: 571 brelse(group_bh); 572 573out_unlock: 574 brelse(main_bm_bh); 575 576 ocfs2_inode_unlock(main_bm_inode, 1); 577 578out_mutex: 579 inode_unlock(main_bm_inode); 580 iput(main_bm_inode); 581 582out: 583 return ret; 584} 585