1 /* 2 * Copyright (C) Sistina Software, Inc. 1997-2003 All rights reserved. 3 * Copyright (C) 2004-2008 Red Hat, Inc. All rights reserved. 4 * 5 * This copyrighted material is made available to anyone wishing to use, 6 * modify, copy, or redistribute it subject to the terms and conditions 7 * of the GNU General Public License version 2. 8 */ 9 10 #ifndef __INCORE_DOT_H__ 11 #define __INCORE_DOT_H__ 12 13 #include <linux/fs.h> 14 #include <linux/kobject.h> 15 #include <linux/workqueue.h> 16 #include <linux/dlm.h> 17 #include <linux/buffer_head.h> 18 #include <linux/rcupdate.h> 19 #include <linux/rculist_bl.h> 20 #include <linux/completion.h> 21 #include <linux/rbtree.h> 22 23 #define DIO_WAIT 0x00000010 24 #define DIO_METADATA 0x00000020 25 26 struct gfs2_log_operations; 27 struct gfs2_log_element; 28 struct gfs2_holder; 29 struct gfs2_glock; 30 struct gfs2_quota_data; 31 struct gfs2_trans; 32 struct gfs2_ail; 33 struct gfs2_jdesc; 34 struct gfs2_sbd; 35 struct lm_lockops; 36 37 typedef void (*gfs2_glop_bh_t) (struct gfs2_glock *gl, unsigned int ret); 38 39 struct gfs2_log_header_host { 40 u64 lh_sequence; /* Sequence number of this transaction */ 41 u32 lh_flags; /* GFS2_LOG_HEAD_... */ 42 u32 lh_tail; /* Block number of log tail */ 43 u32 lh_blkno; 44 u32 lh_hash; 45 }; 46 47 /* 48 * Structure of operations that are associated with each 49 * type of element in the log. 50 */ 51 52 struct gfs2_log_operations { 53 void (*lo_add) (struct gfs2_sbd *sdp, struct gfs2_log_element *le); 54 void (*lo_before_commit) (struct gfs2_sbd *sdp); 55 void (*lo_after_commit) (struct gfs2_sbd *sdp, struct gfs2_ail *ai); 56 void (*lo_before_scan) (struct gfs2_jdesc *jd, 57 struct gfs2_log_header_host *head, int pass); 58 int (*lo_scan_elements) (struct gfs2_jdesc *jd, unsigned int start, 59 struct gfs2_log_descriptor *ld, __be64 *ptr, 60 int pass); 61 void (*lo_after_scan) (struct gfs2_jdesc *jd, int error, int pass); 62 const char *lo_name; 63 }; 64 65 struct gfs2_log_element { 66 struct list_head le_list; 67 const struct gfs2_log_operations *le_ops; 68 }; 69 70 #define GBF_FULL 1 71 72 struct gfs2_bitmap { 73 struct buffer_head *bi_bh; 74 char *bi_clone; 75 unsigned long bi_flags; 76 u32 bi_offset; 77 u32 bi_start; 78 u32 bi_len; 79 }; 80 81 struct gfs2_rgrpd { 82 struct rb_node rd_node; /* Link with superblock */ 83 struct gfs2_glock *rd_gl; /* Glock for this rgrp */ 84 u64 rd_addr; /* grp block disk address */ 85 u64 rd_data0; /* first data location */ 86 u32 rd_length; /* length of rgrp header in fs blocks */ 87 u32 rd_data; /* num of data blocks in rgrp */ 88 u32 rd_bitbytes; /* number of bytes in data bitmaps */ 89 u32 rd_free; 90 u32 rd_free_clone; 91 u32 rd_dinodes; 92 u64 rd_igeneration; 93 struct gfs2_bitmap *rd_bits; 94 struct gfs2_sbd *rd_sbd; 95 u32 rd_last_alloc; 96 u32 rd_flags; 97 #define GFS2_RDF_CHECK 0x10000000 /* check for unlinked inodes */ 98 #define GFS2_RDF_UPTODATE 0x20000000 /* rg is up to date */ 99 #define GFS2_RDF_ERROR 0x40000000 /* error in rg */ 100 #define GFS2_RDF_MASK 0xf0000000 /* mask for internal flags */ 101 }; 102 103 enum gfs2_state_bits { 104 BH_Pinned = BH_PrivateStart, 105 BH_Escaped = BH_PrivateStart + 1, 106 BH_Zeronew = BH_PrivateStart + 2, 107 }; 108 109 BUFFER_FNS(Pinned, pinned) 110 TAS_BUFFER_FNS(Pinned, pinned) 111 BUFFER_FNS(Escaped, escaped) 112 TAS_BUFFER_FNS(Escaped, escaped) 113 BUFFER_FNS(Zeronew, zeronew) 114 TAS_BUFFER_FNS(Zeronew, zeronew) 115 116 struct gfs2_bufdata { 117 struct buffer_head *bd_bh; 118 struct gfs2_glock *bd_gl; 119 120 union { 121 struct list_head list_tr; 122 u64 blkno; 123 } u; 124 #define bd_list_tr u.list_tr 125 #define bd_blkno u.blkno 126 127 struct gfs2_log_element bd_le; 128 129 struct gfs2_ail *bd_ail; 130 struct list_head bd_ail_st_list; 131 struct list_head bd_ail_gl_list; 132 }; 133 134 /* 135 * Internally, we prefix things with gdlm_ and GDLM_ (for gfs-dlm) since a 136 * prefix of lock_dlm_ gets awkward. 137 */ 138 139 #define GDLM_STRNAME_BYTES 25 140 #define GDLM_LVB_SIZE 32 141 142 enum { 143 DFL_BLOCK_LOCKS = 0, 144 }; 145 146 struct lm_lockname { 147 u64 ln_number; 148 unsigned int ln_type; 149 }; 150 151 #define lm_name_equal(name1, name2) \ 152 (((name1)->ln_number == (name2)->ln_number) && \ 153 ((name1)->ln_type == (name2)->ln_type)) 154 155 156 struct gfs2_glock_operations { 157 void (*go_xmote_th) (struct gfs2_glock *gl); 158 int (*go_xmote_bh) (struct gfs2_glock *gl, struct gfs2_holder *gh); 159 void (*go_inval) (struct gfs2_glock *gl, int flags); 160 int (*go_demote_ok) (const struct gfs2_glock *gl); 161 int (*go_lock) (struct gfs2_holder *gh); 162 void (*go_unlock) (struct gfs2_holder *gh); 163 int (*go_dump)(struct seq_file *seq, const struct gfs2_glock *gl); 164 void (*go_callback) (struct gfs2_glock *gl); 165 const int go_type; 166 const unsigned long go_flags; 167 #define GLOF_ASPACE 1 168 }; 169 170 enum { 171 /* States */ 172 HIF_HOLDER = 6, /* Set for gh that "holds" the glock */ 173 HIF_FIRST = 7, 174 HIF_WAIT = 10, 175 }; 176 177 struct gfs2_holder { 178 struct list_head gh_list; 179 180 struct gfs2_glock *gh_gl; 181 struct pid *gh_owner_pid; 182 unsigned int gh_state; 183 unsigned gh_flags; 184 185 int gh_error; 186 unsigned long gh_iflags; /* HIF_... */ 187 unsigned long gh_ip; 188 }; 189 190 enum { 191 GLF_LOCK = 1, 192 GLF_DEMOTE = 3, 193 GLF_PENDING_DEMOTE = 4, 194 GLF_DEMOTE_IN_PROGRESS = 5, 195 GLF_DIRTY = 6, 196 GLF_LFLUSH = 7, 197 GLF_INVALIDATE_IN_PROGRESS = 8, 198 GLF_REPLY_PENDING = 9, 199 GLF_INITIAL = 10, 200 GLF_FROZEN = 11, 201 GLF_QUEUED = 12, 202 GLF_LRU = 13, 203 GLF_OBJECT = 14, /* Used only for tracing */ 204 }; 205 206 struct gfs2_glock { 207 struct hlist_bl_node gl_list; 208 unsigned long gl_flags; /* GLF_... */ 209 struct lm_lockname gl_name; 210 atomic_t gl_ref; 211 212 spinlock_t gl_spin; 213 214 /* State fields protected by gl_spin */ 215 unsigned int gl_state:2, /* Current state */ 216 gl_target:2, /* Target state */ 217 gl_demote_state:2, /* State requested by remote node */ 218 gl_req:2, /* State in last dlm request */ 219 gl_reply:8; /* Last reply from the dlm */ 220 221 unsigned int gl_hash; 222 unsigned long gl_demote_time; /* time of first demote request */ 223 long gl_hold_time; 224 struct list_head gl_holders; 225 226 const struct gfs2_glock_operations *gl_ops; 227 char gl_strname[GDLM_STRNAME_BYTES]; 228 struct dlm_lksb gl_lksb; 229 char gl_lvb[32]; 230 unsigned long gl_tchange; 231 void *gl_object; 232 233 struct list_head gl_lru; 234 235 struct gfs2_sbd *gl_sbd; 236 237 struct list_head gl_ail_list; 238 atomic_t gl_ail_count; 239 atomic_t gl_revokes; 240 struct delayed_work gl_work; 241 struct work_struct gl_delete; 242 struct rcu_head gl_rcu; 243 }; 244 245 #define GFS2_MIN_LVB_SIZE 32 /* Min size of LVB that gfs2 supports */ 246 247 struct gfs2_alloc { 248 /* Quota stuff */ 249 struct gfs2_quota_data *al_qd[2*MAXQUOTAS]; 250 struct gfs2_holder al_qd_ghs[2*MAXQUOTAS]; 251 unsigned int al_qd_num; 252 253 u32 al_requested; /* Filled in by caller of gfs2_inplace_reserve() */ 254 u32 al_alloced; /* Filled in by gfs2_alloc_*() */ 255 256 /* Filled in by gfs2_inplace_reserve() */ 257 struct gfs2_holder al_rgd_gh; 258 }; 259 260 enum { 261 GIF_INVALID = 0, 262 GIF_QD_LOCKED = 1, 263 GIF_ALLOC_FAILED = 2, 264 GIF_SW_PAGED = 3, 265 }; 266 267 268 struct gfs2_inode { 269 struct inode i_inode; 270 u64 i_no_addr; 271 u64 i_no_formal_ino; 272 u64 i_generation; 273 u64 i_eattr; 274 unsigned long i_flags; /* GIF_... */ 275 struct gfs2_glock *i_gl; /* Move into i_gh? */ 276 struct gfs2_holder i_iopen_gh; 277 struct gfs2_holder i_gh; /* for prepare/commit_write only */ 278 struct gfs2_alloc *i_alloc; 279 struct gfs2_rgrpd *i_rgd; 280 u64 i_goal; /* goal block for allocations */ 281 struct rw_semaphore i_rw_mutex; 282 struct list_head i_trunc_list; 283 __be64 *i_hash_cache; 284 u32 i_entries; 285 u32 i_diskflags; 286 u8 i_height; 287 u8 i_depth; 288 }; 289 290 /* 291 * Since i_inode is the first element of struct gfs2_inode, 292 * this is effectively a cast. 293 */ 294 static inline struct gfs2_inode *GFS2_I(struct inode *inode) 295 { 296 return container_of(inode, struct gfs2_inode, i_inode); 297 } 298 299 static inline struct gfs2_sbd *GFS2_SB(const struct inode *inode) 300 { 301 return inode->i_sb->s_fs_info; 302 } 303 304 struct gfs2_file { 305 struct mutex f_fl_mutex; 306 struct gfs2_holder f_fl_gh; 307 }; 308 309 struct gfs2_revoke_replay { 310 struct list_head rr_list; 311 u64 rr_blkno; 312 unsigned int rr_where; 313 }; 314 315 enum { 316 QDF_USER = 0, 317 QDF_CHANGE = 1, 318 QDF_LOCKED = 2, 319 QDF_REFRESH = 3, 320 }; 321 322 struct gfs2_quota_data { 323 struct list_head qd_list; 324 struct list_head qd_reclaim; 325 326 atomic_t qd_count; 327 328 u32 qd_id; 329 unsigned long qd_flags; /* QDF_... */ 330 331 s64 qd_change; 332 s64 qd_change_sync; 333 334 unsigned int qd_slot; 335 unsigned int qd_slot_count; 336 337 struct buffer_head *qd_bh; 338 struct gfs2_quota_change *qd_bh_qc; 339 unsigned int qd_bh_count; 340 341 struct gfs2_glock *qd_gl; 342 struct gfs2_quota_lvb qd_qb; 343 344 u64 qd_sync_gen; 345 unsigned long qd_last_warn; 346 }; 347 348 struct gfs2_trans { 349 unsigned long tr_ip; 350 351 unsigned int tr_blocks; 352 unsigned int tr_revokes; 353 unsigned int tr_reserved; 354 355 struct gfs2_holder tr_t_gh; 356 357 int tr_touched; 358 359 unsigned int tr_num_buf; 360 unsigned int tr_num_buf_new; 361 unsigned int tr_num_databuf_new; 362 unsigned int tr_num_buf_rm; 363 unsigned int tr_num_databuf_rm; 364 struct list_head tr_list_buf; 365 366 unsigned int tr_num_revoke; 367 unsigned int tr_num_revoke_rm; 368 }; 369 370 struct gfs2_ail { 371 struct list_head ai_list; 372 373 unsigned int ai_first; 374 struct list_head ai_ail1_list; 375 struct list_head ai_ail2_list; 376 }; 377 378 struct gfs2_journal_extent { 379 struct list_head extent_list; 380 381 unsigned int lblock; /* First logical block */ 382 u64 dblock; /* First disk block */ 383 u64 blocks; 384 }; 385 386 struct gfs2_jdesc { 387 struct list_head jd_list; 388 struct list_head extent_list; 389 struct work_struct jd_work; 390 struct inode *jd_inode; 391 unsigned long jd_flags; 392 #define JDF_RECOVERY 1 393 unsigned int jd_jid; 394 unsigned int jd_blocks; 395 }; 396 397 struct gfs2_statfs_change_host { 398 s64 sc_total; 399 s64 sc_free; 400 s64 sc_dinodes; 401 }; 402 403 #define GFS2_QUOTA_DEFAULT GFS2_QUOTA_OFF 404 #define GFS2_QUOTA_OFF 0 405 #define GFS2_QUOTA_ACCOUNT 1 406 #define GFS2_QUOTA_ON 2 407 408 #define GFS2_DATA_DEFAULT GFS2_DATA_ORDERED 409 #define GFS2_DATA_WRITEBACK 1 410 #define GFS2_DATA_ORDERED 2 411 412 #define GFS2_ERRORS_DEFAULT GFS2_ERRORS_WITHDRAW 413 #define GFS2_ERRORS_WITHDRAW 0 414 #define GFS2_ERRORS_CONTINUE 1 /* place holder for future feature */ 415 #define GFS2_ERRORS_RO 2 /* place holder for future feature */ 416 #define GFS2_ERRORS_PANIC 3 417 418 struct gfs2_args { 419 char ar_lockproto[GFS2_LOCKNAME_LEN]; /* Name of the Lock Protocol */ 420 char ar_locktable[GFS2_LOCKNAME_LEN]; /* Name of the Lock Table */ 421 char ar_hostdata[GFS2_LOCKNAME_LEN]; /* Host specific data */ 422 unsigned int ar_spectator:1; /* Don't get a journal */ 423 unsigned int ar_localflocks:1; /* Let the VFS do flock|fcntl */ 424 unsigned int ar_debug:1; /* Oops on errors */ 425 unsigned int ar_posix_acl:1; /* Enable posix acls */ 426 unsigned int ar_quota:2; /* off/account/on */ 427 unsigned int ar_suiddir:1; /* suiddir support */ 428 unsigned int ar_data:2; /* ordered/writeback */ 429 unsigned int ar_meta:1; /* mount metafs */ 430 unsigned int ar_discard:1; /* discard requests */ 431 unsigned int ar_errors:2; /* errors=withdraw | panic */ 432 unsigned int ar_nobarrier:1; /* do not send barriers */ 433 int ar_commit; /* Commit interval */ 434 int ar_statfs_quantum; /* The fast statfs interval */ 435 int ar_quota_quantum; /* The quota interval */ 436 int ar_statfs_percent; /* The % change to force sync */ 437 }; 438 439 struct gfs2_tune { 440 spinlock_t gt_spin; 441 442 unsigned int gt_logd_secs; 443 444 unsigned int gt_quota_simul_sync; /* Max quotavals to sync at once */ 445 unsigned int gt_quota_warn_period; /* Secs between quota warn msgs */ 446 unsigned int gt_quota_scale_num; /* Numerator */ 447 unsigned int gt_quota_scale_den; /* Denominator */ 448 unsigned int gt_quota_quantum; /* Secs between syncs to quota file */ 449 unsigned int gt_new_files_jdata; 450 unsigned int gt_max_readahead; /* Max bytes to read-ahead from disk */ 451 unsigned int gt_complain_secs; 452 unsigned int gt_statfs_quantum; 453 unsigned int gt_statfs_slow; 454 }; 455 456 enum { 457 SDF_JOURNAL_CHECKED = 0, 458 SDF_JOURNAL_LIVE = 1, 459 SDF_SHUTDOWN = 2, 460 SDF_NOBARRIERS = 3, 461 SDF_NORECOVERY = 4, 462 SDF_DEMOTE = 5, 463 SDF_NOJOURNALID = 6, 464 }; 465 466 #define GFS2_FSNAME_LEN 256 467 468 struct gfs2_inum_host { 469 u64 no_formal_ino; 470 u64 no_addr; 471 }; 472 473 struct gfs2_sb_host { 474 u32 sb_magic; 475 u32 sb_type; 476 u32 sb_format; 477 478 u32 sb_fs_format; 479 u32 sb_multihost_format; 480 u32 sb_bsize; 481 u32 sb_bsize_shift; 482 483 struct gfs2_inum_host sb_master_dir; 484 struct gfs2_inum_host sb_root_dir; 485 486 char sb_lockproto[GFS2_LOCKNAME_LEN]; 487 char sb_locktable[GFS2_LOCKNAME_LEN]; 488 }; 489 490 /* 491 * lm_mount() return values 492 * 493 * ls_jid - the journal ID this node should use 494 * ls_first - this node is the first to mount the file system 495 * ls_lockspace - lock module's context for this file system 496 * ls_ops - lock module's functions 497 */ 498 499 struct lm_lockstruct { 500 int ls_jid; 501 unsigned int ls_first; 502 unsigned int ls_first_done; 503 unsigned int ls_nodir; 504 const struct lm_lockops *ls_ops; 505 unsigned long ls_flags; 506 dlm_lockspace_t *ls_dlm; 507 508 int ls_recover_jid_done; 509 int ls_recover_jid_status; 510 }; 511 512 struct gfs2_sbd { 513 struct super_block *sd_vfs; 514 struct kobject sd_kobj; 515 unsigned long sd_flags; /* SDF_... */ 516 struct gfs2_sb_host sd_sb; 517 518 /* Constants computed on mount */ 519 520 u32 sd_fsb2bb; 521 u32 sd_fsb2bb_shift; 522 u32 sd_diptrs; /* Number of pointers in a dinode */ 523 u32 sd_inptrs; /* Number of pointers in a indirect block */ 524 u32 sd_jbsize; /* Size of a journaled data block */ 525 u32 sd_hash_bsize; /* sizeof(exhash block) */ 526 u32 sd_hash_bsize_shift; 527 u32 sd_hash_ptrs; /* Number of pointers in a hash block */ 528 u32 sd_qc_per_block; 529 u32 sd_max_dirres; /* Max blocks needed to add a directory entry */ 530 u32 sd_max_height; /* Max height of a file's metadata tree */ 531 u64 sd_heightsize[GFS2_MAX_META_HEIGHT + 1]; 532 u32 sd_max_jheight; /* Max height of journaled file's meta tree */ 533 u64 sd_jheightsize[GFS2_MAX_META_HEIGHT + 1]; 534 535 struct gfs2_args sd_args; /* Mount arguments */ 536 struct gfs2_tune sd_tune; /* Filesystem tuning structure */ 537 538 /* Lock Stuff */ 539 540 struct lm_lockstruct sd_lockstruct; 541 struct gfs2_holder sd_live_gh; 542 struct gfs2_glock *sd_rename_gl; 543 struct gfs2_glock *sd_trans_gl; 544 wait_queue_head_t sd_glock_wait; 545 atomic_t sd_glock_disposal; 546 struct completion sd_locking_init; 547 548 /* Inode Stuff */ 549 550 struct dentry *sd_master_dir; 551 struct dentry *sd_root_dir; 552 553 struct inode *sd_jindex; 554 struct inode *sd_statfs_inode; 555 struct inode *sd_sc_inode; 556 struct inode *sd_qc_inode; 557 struct inode *sd_rindex; 558 struct inode *sd_quota_inode; 559 560 /* StatFS stuff */ 561 562 spinlock_t sd_statfs_spin; 563 struct gfs2_statfs_change_host sd_statfs_master; 564 struct gfs2_statfs_change_host sd_statfs_local; 565 int sd_statfs_force_sync; 566 567 /* Resource group stuff */ 568 569 int sd_rindex_uptodate; 570 spinlock_t sd_rindex_spin; 571 struct mutex sd_rindex_mutex; 572 struct rb_root sd_rindex_tree; 573 unsigned int sd_rgrps; 574 unsigned int sd_max_rg_data; 575 576 /* Journal index stuff */ 577 578 struct list_head sd_jindex_list; 579 spinlock_t sd_jindex_spin; 580 struct mutex sd_jindex_mutex; 581 unsigned int sd_journals; 582 583 struct gfs2_jdesc *sd_jdesc; 584 struct gfs2_holder sd_journal_gh; 585 struct gfs2_holder sd_jinode_gh; 586 587 struct gfs2_holder sd_sc_gh; 588 struct gfs2_holder sd_qc_gh; 589 590 /* Daemon stuff */ 591 592 struct task_struct *sd_logd_process; 593 struct task_struct *sd_quotad_process; 594 595 /* Quota stuff */ 596 597 struct list_head sd_quota_list; 598 atomic_t sd_quota_count; 599 struct mutex sd_quota_mutex; 600 wait_queue_head_t sd_quota_wait; 601 struct list_head sd_trunc_list; 602 spinlock_t sd_trunc_lock; 603 604 unsigned int sd_quota_slots; 605 unsigned int sd_quota_chunks; 606 unsigned char **sd_quota_bitmap; 607 608 u64 sd_quota_sync_gen; 609 610 /* Log stuff */ 611 612 spinlock_t sd_log_lock; 613 614 unsigned int sd_log_blks_reserved; 615 unsigned int sd_log_commited_buf; 616 unsigned int sd_log_commited_databuf; 617 int sd_log_commited_revoke; 618 619 atomic_t sd_log_pinned; 620 unsigned int sd_log_num_buf; 621 unsigned int sd_log_num_revoke; 622 unsigned int sd_log_num_rg; 623 unsigned int sd_log_num_databuf; 624 625 struct list_head sd_log_le_buf; 626 struct list_head sd_log_le_revoke; 627 struct list_head sd_log_le_rg; 628 struct list_head sd_log_le_databuf; 629 struct list_head sd_log_le_ordered; 630 631 atomic_t sd_log_thresh1; 632 atomic_t sd_log_thresh2; 633 atomic_t sd_log_blks_free; 634 wait_queue_head_t sd_log_waitq; 635 wait_queue_head_t sd_logd_waitq; 636 637 u64 sd_log_sequence; 638 unsigned int sd_log_head; 639 unsigned int sd_log_tail; 640 int sd_log_idle; 641 642 struct rw_semaphore sd_log_flush_lock; 643 atomic_t sd_log_in_flight; 644 wait_queue_head_t sd_log_flush_wait; 645 646 unsigned int sd_log_flush_head; 647 u64 sd_log_flush_wrapped; 648 649 spinlock_t sd_ail_lock; 650 struct list_head sd_ail1_list; 651 struct list_head sd_ail2_list; 652 653 /* Replay stuff */ 654 655 struct list_head sd_revoke_list; 656 unsigned int sd_replay_tail; 657 658 unsigned int sd_found_blocks; 659 unsigned int sd_found_revokes; 660 unsigned int sd_replayed_blocks; 661 662 /* For quiescing the filesystem */ 663 664 struct gfs2_holder sd_freeze_gh; 665 struct mutex sd_freeze_lock; 666 unsigned int sd_freeze_count; 667 668 char sd_fsname[GFS2_FSNAME_LEN]; 669 char sd_table_name[GFS2_FSNAME_LEN]; 670 char sd_proto_name[GFS2_FSNAME_LEN]; 671 672 /* Debugging crud */ 673 674 unsigned long sd_last_warning; 675 struct dentry *debugfs_dir; /* debugfs directory */ 676 struct dentry *debugfs_dentry_glocks; /* for debugfs */ 677 }; 678 679 #endif /* __INCORE_DOT_H__ */ 680 681