1 /* 2 * Copyright (c) 2007-2008 The DragonFly Project. All rights reserved. 3 * 4 * This code is derived from software contributed to The DragonFly Project 5 * by Matthew Dillon <dillon@backplane.com> 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 11 * 1. Redistributions of source code must retain the above copyright 12 * notice, this list of conditions and the following disclaimer. 13 * 2. Redistributions in binary form must reproduce the above copyright 14 * notice, this list of conditions and the following disclaimer in 15 * the documentation and/or other materials provided with the 16 * distribution. 17 * 3. Neither the name of The DragonFly Project nor the names of its 18 * contributors may be used to endorse or promote products derived 19 * from this software without specific, prior written permission. 20 * 21 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS 22 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT 23 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS 24 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE 25 * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, 26 * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING, 27 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; 28 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED 29 * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, 30 * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT 31 * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 32 * SUCH DAMAGE. 33 * 34 * $DragonFly: src/sys/vfs/hammer/hammer.h,v 1.130 2008/11/13 02:18:43 dillon Exp $ 35 */ 36 /* 37 * This header file contains structures used internally by the HAMMERFS 38 * implementation. See hammer_disk.h for on-disk structures. 39 */ 40 41 #include <sys/param.h> 42 #include <sys/types.h> 43 #include <sys/kernel.h> 44 #include <sys/conf.h> 45 #include <sys/systm.h> 46 #include <sys/tree.h> 47 #include <sys/malloc.h> 48 #include <sys/mount.h> 49 #include <sys/mountctl.h> 50 #include <sys/vnode.h> 51 #include <sys/proc.h> 52 #include <sys/priv.h> 53 #include <sys/stat.h> 54 #include <sys/globaldata.h> 55 #include <sys/lockf.h> 56 #include <sys/buf.h> 57 #include <sys/queue.h> 58 #include <sys/ktr.h> 59 #include <sys/globaldata.h> 60 61 #include <sys/buf2.h> 62 #include <sys/signal2.h> 63 #include "hammer_disk.h" 64 #include "hammer_mount.h" 65 #include "hammer_ioctl.h" 66 67 #if defined(_KERNEL) || defined(_KERNEL_STRUCTURES) 68 69 MALLOC_DECLARE(M_HAMMER); 70 71 /* 72 * Kernel trace 73 */ 74 #if !defined(KTR_HAMMER) 75 #define KTR_HAMMER KTR_ALL 76 #endif 77 KTR_INFO_MASTER_EXTERN(hammer); 78 79 /* 80 * Misc structures 81 */ 82 struct hammer_mount; 83 84 /* 85 * Key structure used for custom RB tree inode lookups. This prototypes 86 * the function hammer_ino_rb_tree_RB_LOOKUP_INFO(root, info). 87 */ 88 typedef struct hammer_inode_info { 89 int64_t obj_id; /* (key) object identifier */ 90 hammer_tid_t obj_asof; /* (key) snapshot transid or 0 */ 91 u_int32_t obj_localization; /* (key) pseudo-fs */ 92 union { 93 struct hammer_btree_leaf_elm *leaf; 94 } u; 95 } *hammer_inode_info_t; 96 97 typedef enum hammer_transaction_type { 98 HAMMER_TRANS_RO, 99 HAMMER_TRANS_STD, 100 HAMMER_TRANS_FLS 101 } hammer_transaction_type_t; 102 103 /* 104 * HAMMER Transaction tracking 105 */ 106 struct hammer_transaction { 107 hammer_transaction_type_t type; 108 struct hammer_mount *hmp; 109 hammer_tid_t tid; 110 u_int64_t time; 111 u_int32_t time32; 112 int sync_lock_refs; 113 int flags; 114 struct hammer_volume *rootvol; 115 }; 116 117 typedef struct hammer_transaction *hammer_transaction_t; 118 119 #define HAMMER_TRANSF_NEWINODE 0x0001 120 #define HAMMER_TRANSF_DIDIO 0x0002 121 #define HAMMER_TRANSF_CRCDOM 0x0004 /* EDOM on CRC error, less critical */ 122 123 /* 124 * HAMMER locks 125 */ 126 struct hammer_lock { 127 int refs; /* active references delay writes */ 128 int lockcount; /* lock count for exclusive/shared access */ 129 int wanted; 130 int exwanted; /* number of threads waiting for ex lock */ 131 struct thread *locktd; 132 }; 133 134 static __inline int 135 hammer_islocked(struct hammer_lock *lock) 136 { 137 return(lock->lockcount != 0); 138 } 139 140 static __inline int 141 hammer_isactive(struct hammer_lock *lock) 142 { 143 return(lock->refs != 0); 144 } 145 146 static __inline int 147 hammer_islastref(struct hammer_lock *lock) 148 { 149 return(lock->refs == 1); 150 } 151 152 /* 153 * Return if we specifically own the lock exclusively. 154 */ 155 static __inline int 156 hammer_lock_excl_owned(struct hammer_lock *lock, thread_t td) 157 { 158 if (lock->lockcount > 0 && lock->locktd == td) 159 return(1); 160 return(0); 161 } 162 163 /* 164 * Flush state, used by various structures 165 */ 166 typedef enum hammer_inode_state { 167 HAMMER_FST_IDLE, 168 HAMMER_FST_SETUP, 169 HAMMER_FST_FLUSH 170 } hammer_inode_state_t; 171 172 TAILQ_HEAD(hammer_record_list, hammer_record); 173 174 /* 175 * Pseudo-filesystem extended data tracking 176 */ 177 struct hammer_pfs_rb_tree; 178 struct hammer_pseudofs_inmem; 179 RB_HEAD(hammer_pfs_rb_tree, hammer_pseudofs_inmem); 180 RB_PROTOTYPE2(hammer_pfs_rb_tree, hammer_pseudofs_inmem, rb_node, 181 hammer_pfs_rb_compare, u_int32_t); 182 183 struct hammer_pseudofs_inmem { 184 RB_ENTRY(hammer_pseudofs_inmem) rb_node; 185 struct hammer_lock lock; 186 u_int32_t localization; 187 hammer_tid_t create_tid; 188 int flags; 189 udev_t fsid_udev; 190 struct hammer_pseudofs_data pfsd; 191 }; 192 193 typedef struct hammer_pseudofs_inmem *hammer_pseudofs_inmem_t; 194 195 #define HAMMER_PFSM_DELETED 0x0001 196 197 /* 198 * Cache object ids. A fixed number of objid cache structures are 199 * created to reserve object id's for newly created files in multiples 200 * of 100,000, localized to a particular directory, and recycled as 201 * needed. This allows parallel create operations in different 202 * directories to retain fairly localized object ids which in turn 203 * improves reblocking performance and layout. 204 */ 205 #define OBJID_CACHE_SIZE 1024 206 #define OBJID_CACHE_BULK 100000 207 208 typedef struct hammer_objid_cache { 209 TAILQ_ENTRY(hammer_objid_cache) entry; 210 struct hammer_inode *dip; 211 hammer_tid_t next_tid; 212 int count; 213 } *hammer_objid_cache_t; 214 215 /* 216 * Associate an inode with a B-Tree node to cache search start positions 217 */ 218 typedef struct hammer_node_cache { 219 TAILQ_ENTRY(hammer_node_cache) entry; 220 struct hammer_node *node; 221 struct hammer_inode *ip; 222 } *hammer_node_cache_t; 223 224 TAILQ_HEAD(hammer_node_cache_list, hammer_node_cache); 225 226 /* 227 * Structure used to organize flush groups. Flush groups must be 228 * organized into chunks in order to avoid blowing out the UNDO FIFO. 229 * Without this a 'sync' could end up flushing 50,000 inodes in a single 230 * transaction. 231 */ 232 struct hammer_flush_group { 233 TAILQ_ENTRY(hammer_flush_group) flush_entry; 234 TAILQ_HEAD(, hammer_inode) flush_list; 235 int unused01; /* inode load */ 236 int total_count; /* record load */ 237 int running; /* group is running */ 238 int closed; 239 int refs; 240 }; 241 242 typedef struct hammer_flush_group *hammer_flush_group_t; 243 244 TAILQ_HEAD(hammer_flush_group_list, hammer_flush_group); 245 246 /* 247 * Structure used to represent an inode in-memory. 248 * 249 * The record and data associated with an inode may be out of sync with 250 * the disk (xDIRTY flags), or not even on the disk at all (ONDISK flag 251 * clear). 252 * 253 * An inode may also hold a cache of unsynchronized records, used for 254 * database and directories only. Unsynchronized regular file data is 255 * stored in the buffer cache. 256 * 257 * NOTE: A file which is created and destroyed within the initial 258 * synchronization period can wind up not doing any disk I/O at all. 259 * 260 * Finally, an inode may cache numerous disk-referencing B-Tree cursors. 261 */ 262 struct hammer_ino_rb_tree; 263 struct hammer_inode; 264 RB_HEAD(hammer_ino_rb_tree, hammer_inode); 265 RB_PROTOTYPEX(hammer_ino_rb_tree, INFO, hammer_inode, rb_node, 266 hammer_ino_rb_compare, hammer_inode_info_t); 267 268 struct hammer_rec_rb_tree; 269 struct hammer_record; 270 RB_HEAD(hammer_rec_rb_tree, hammer_record); 271 RB_PROTOTYPEX(hammer_rec_rb_tree, INFO, hammer_record, rb_node, 272 hammer_rec_rb_compare, hammer_btree_leaf_elm_t); 273 274 TAILQ_HEAD(hammer_node_list, hammer_node); 275 276 struct hammer_inode { 277 RB_ENTRY(hammer_inode) rb_node; 278 hammer_inode_state_t flush_state; 279 hammer_flush_group_t flush_group; 280 TAILQ_ENTRY(hammer_inode) flush_entry; 281 struct hammer_record_list target_list; /* target of dependant recs */ 282 int64_t obj_id; /* (key) object identifier */ 283 hammer_tid_t obj_asof; /* (key) snapshot or 0 */ 284 u_int32_t obj_localization; /* (key) pseudo-fs */ 285 struct hammer_mount *hmp; 286 hammer_objid_cache_t objid_cache; 287 int flags; 288 int error; /* flush error */ 289 int cursor_ip_refs; /* sanity */ 290 int rsv_recs; 291 struct vnode *vp; 292 hammer_pseudofs_inmem_t pfsm; 293 struct lockf advlock; 294 struct hammer_lock lock; /* sync copy interlock */ 295 off_t trunc_off; 296 struct hammer_btree_leaf_elm ino_leaf; /* in-memory cache */ 297 struct hammer_inode_data ino_data; /* in-memory cache */ 298 struct hammer_rec_rb_tree rec_tree; /* in-memory cache */ 299 int rec_generation; 300 struct hammer_node_cache cache[2]; /* search initiate cache */ 301 302 /* 303 * When a demark is created to synchronize an inode to 304 * disk, certain fields are copied so the front-end VOPs 305 * can continue to run in parallel with the synchronization 306 * occuring in the background. 307 */ 308 int sync_flags; /* to-sync flags cache */ 309 off_t sync_trunc_off; /* to-sync truncation */ 310 off_t save_trunc_off; /* write optimization */ 311 struct hammer_btree_leaf_elm sync_ino_leaf; /* to-sync cache */ 312 struct hammer_inode_data sync_ino_data; /* to-sync cache */ 313 }; 314 315 typedef struct hammer_inode *hammer_inode_t; 316 317 #define VTOI(vp) ((struct hammer_inode *)(vp)->v_data) 318 319 #define HAMMER_INODE_DDIRTY 0x0001 /* in-memory ino_data is dirty */ 320 /* (not including atime/mtime) */ 321 #define HAMMER_INODE_RSV_INODES 0x0002 /* hmp->rsv_inodes bumped */ 322 #define HAMMER_INODE_CONN_DOWN 0x0004 /* include in downward recursion */ 323 #define HAMMER_INODE_XDIRTY 0x0008 /* in-memory records */ 324 #define HAMMER_INODE_ONDISK 0x0010 /* inode is on-disk (else not yet) */ 325 #define HAMMER_INODE_FLUSH 0x0020 /* flush on last ref */ 326 #define HAMMER_INODE_DELETED 0x0080 /* inode delete (backend) */ 327 #define HAMMER_INODE_DELONDISK 0x0100 /* delete synchronized to disk */ 328 #define HAMMER_INODE_RO 0x0200 /* read-only (because of as-of) */ 329 #define HAMMER_INODE_VHELD 0x0400 /* vnode held on sync */ 330 #define HAMMER_INODE_DONDISK 0x0800 /* data records may be on disk */ 331 #define HAMMER_INODE_BUFS 0x1000 /* dirty high level bps present */ 332 #define HAMMER_INODE_REFLUSH 0x2000 /* flush on dependancy / reflush */ 333 #define HAMMER_INODE_RECLAIM 0x4000 /* trying to reclaim */ 334 #define HAMMER_INODE_FLUSHW 0x8000 /* Someone waiting for flush */ 335 336 #define HAMMER_INODE_TRUNCATED 0x00010000 337 #define HAMMER_INODE_DELETING 0x00020000 /* inode delete request (frontend)*/ 338 #define HAMMER_INODE_RESIGNAL 0x00040000 /* re-signal on re-flush */ 339 #define HAMMER_INODE_ATIME 0x00100000 /* in-memory atime modified */ 340 #define HAMMER_INODE_MTIME 0x00200000 /* in-memory mtime modified */ 341 #define HAMMER_INODE_WOULDBLOCK 0x00400000 /* re-issue to new flush group */ 342 #define HAMMER_INODE_DUMMY 0x00800000 /* dummy inode covering bad file */ 343 344 #define HAMMER_INODE_MODMASK (HAMMER_INODE_DDIRTY| \ 345 HAMMER_INODE_XDIRTY|HAMMER_INODE_BUFS| \ 346 HAMMER_INODE_ATIME|HAMMER_INODE_MTIME| \ 347 HAMMER_INODE_TRUNCATED|HAMMER_INODE_DELETING) 348 349 #define HAMMER_INODE_MODMASK_NOXDIRTY \ 350 (HAMMER_INODE_MODMASK & ~HAMMER_INODE_XDIRTY) 351 352 #define HAMMER_FLUSH_GROUP_SIZE 64 353 354 #define HAMMER_FLUSH_SIGNAL 0x0001 355 #define HAMMER_FLUSH_RECURSION 0x0002 356 357 /* 358 * Used by the inode reclaim code to pipeline reclaims and avoid 359 * blowing out kernel memory or letting the flusher get too far 360 * behind. The reclaim wakes up when count reaches 0 or the 361 * timer expires. 362 */ 363 struct hammer_reclaim { 364 TAILQ_ENTRY(hammer_reclaim) entry; 365 int count; 366 }; 367 368 #define HAMMER_RECLAIM_FLUSH 2000 369 #define HAMMER_RECLAIM_WAIT 4000 370 371 /* 372 * Structure used to represent an unsynchronized record in-memory. These 373 * records typically represent directory entries. Only non-historical 374 * records are kept in-memory. 375 * 376 * Records are organized as a per-inode RB-Tree. If the inode is not 377 * on disk then neither are any records and the in-memory record tree 378 * represents the entire contents of the inode. If the inode is on disk 379 * then the on-disk B-Tree is scanned in parallel with the in-memory 380 * RB-Tree to synthesize the current state of the file. 381 * 382 * Records are also used to enforce the ordering of directory create/delete 383 * operations. A new inode will not be flushed to disk unless its related 384 * directory entry is also being flushed at the same time. A directory entry 385 * will not be removed unless its related inode is also being removed at the 386 * same time. 387 */ 388 typedef enum hammer_record_type { 389 HAMMER_MEM_RECORD_GENERAL, /* misc record */ 390 HAMMER_MEM_RECORD_INODE, /* inode record */ 391 HAMMER_MEM_RECORD_ADD, /* positive memory cache record */ 392 HAMMER_MEM_RECORD_DEL, /* negative delete-on-disk record */ 393 HAMMER_MEM_RECORD_DATA /* bulk-data record w/on-disk ref */ 394 } hammer_record_type_t; 395 396 struct hammer_record { 397 RB_ENTRY(hammer_record) rb_node; 398 TAILQ_ENTRY(hammer_record) target_entry; 399 hammer_inode_state_t flush_state; 400 hammer_flush_group_t flush_group; 401 hammer_record_type_t type; 402 struct hammer_lock lock; 403 struct hammer_reserve *resv; 404 struct hammer_inode *ip; 405 struct hammer_inode *target_ip; 406 struct hammer_btree_leaf_elm leaf; 407 union hammer_data_ondisk *data; 408 int flags; 409 hammer_off_t zone2_offset; /* direct-write only */ 410 }; 411 412 typedef struct hammer_record *hammer_record_t; 413 414 /* 415 * Record flags. Note that FE can only be set by the frontend if the 416 * record has not been interlocked by the backend w/ BE. 417 */ 418 #define HAMMER_RECF_ALLOCDATA 0x0001 419 #define HAMMER_RECF_ONRBTREE 0x0002 420 #define HAMMER_RECF_DELETED_FE 0x0004 /* deleted (frontend) */ 421 #define HAMMER_RECF_DELETED_BE 0x0008 /* deleted (backend) */ 422 #define HAMMER_RECF_COMMITTED 0x0010 /* committed to the B-Tree */ 423 #define HAMMER_RECF_INTERLOCK_BE 0x0020 /* backend interlock */ 424 #define HAMMER_RECF_WANTED 0x0040 /* wanted by the frontend */ 425 #define HAMMER_RECF_CONVERT_DELETE 0x0100 /* special case */ 426 #define HAMMER_RECF_DIRECT_IO 0x0200 /* related direct I/O running*/ 427 #define HAMMER_RECF_DIRECT_WAIT 0x0400 /* related direct I/O running*/ 428 #define HAMMER_RECF_DIRECT_INVAL 0x0800 /* buffer alias invalidation */ 429 430 /* 431 * hammer_delete_at_cursor() flags 432 */ 433 #define HAMMER_DELETE_ADJUST 0x0001 434 #define HAMMER_DELETE_DESTROY 0x0002 435 436 /* 437 * In-memory structures representing on-disk structures. 438 */ 439 struct hammer_volume; 440 struct hammer_buffer; 441 struct hammer_node; 442 struct hammer_undo; 443 struct hammer_reserve; 444 445 RB_HEAD(hammer_vol_rb_tree, hammer_volume); 446 RB_HEAD(hammer_buf_rb_tree, hammer_buffer); 447 RB_HEAD(hammer_nod_rb_tree, hammer_node); 448 RB_HEAD(hammer_und_rb_tree, hammer_undo); 449 RB_HEAD(hammer_res_rb_tree, hammer_reserve); 450 451 RB_PROTOTYPE2(hammer_vol_rb_tree, hammer_volume, rb_node, 452 hammer_vol_rb_compare, int32_t); 453 RB_PROTOTYPE2(hammer_buf_rb_tree, hammer_buffer, rb_node, 454 hammer_buf_rb_compare, hammer_off_t); 455 RB_PROTOTYPE2(hammer_nod_rb_tree, hammer_node, rb_node, 456 hammer_nod_rb_compare, hammer_off_t); 457 RB_PROTOTYPE2(hammer_und_rb_tree, hammer_undo, rb_node, 458 hammer_und_rb_compare, hammer_off_t); 459 RB_PROTOTYPE2(hammer_res_rb_tree, hammer_reserve, rb_node, 460 hammer_res_rb_compare, hammer_off_t); 461 462 /* 463 * IO management - embedded at the head of various in-memory structures 464 * 465 * VOLUME - hammer_volume containing meta-data 466 * META_BUFFER - hammer_buffer containing meta-data 467 * DATA_BUFFER - hammer_buffer containing pure-data 468 * 469 * Dirty volume headers and dirty meta-data buffers are locked until the 470 * flusher can sequence them out. Dirty pure-data buffers can be written. 471 * Clean buffers can be passively released. 472 */ 473 typedef enum hammer_io_type { 474 HAMMER_STRUCTURE_VOLUME, 475 HAMMER_STRUCTURE_META_BUFFER, 476 HAMMER_STRUCTURE_UNDO_BUFFER, 477 HAMMER_STRUCTURE_DATA_BUFFER 478 } hammer_io_type_t; 479 480 union hammer_io_structure; 481 struct hammer_io; 482 483 struct worklist { 484 LIST_ENTRY(worklist) node; 485 }; 486 487 TAILQ_HEAD(hammer_io_list, hammer_io); 488 typedef struct hammer_io_list *hammer_io_list_t; 489 490 struct hammer_io { 491 struct worklist worklist; 492 struct hammer_lock lock; 493 enum hammer_io_type type; 494 struct hammer_mount *hmp; 495 struct hammer_volume *volume; 496 TAILQ_ENTRY(hammer_io) mod_entry; /* list entry if modified */ 497 hammer_io_list_t mod_list; 498 struct buf *bp; 499 int64_t offset; /* zone-2 offset */ 500 int bytes; /* buffer cache buffer size */ 501 int loading; /* loading/unloading interlock */ 502 int modify_refs; 503 504 u_int modified : 1; /* bp's data was modified */ 505 u_int released : 1; /* bp released (w/ B_LOCKED set) */ 506 u_int running : 1; /* bp write IO in progress */ 507 u_int waiting : 1; /* someone is waiting on us */ 508 u_int validated : 1; /* ondisk has been validated */ 509 u_int waitdep : 1; /* flush waits for dependancies */ 510 u_int recovered : 1; /* has recovery ref */ 511 u_int waitmod : 1; /* waiting for modify_refs */ 512 u_int reclaim : 1; /* reclaim requested */ 513 u_int gencrc : 1; /* crc needs to be generated */ 514 u_int ioerror : 1; /* abort on io-error */ 515 }; 516 517 typedef struct hammer_io *hammer_io_t; 518 519 #define HAMMER_CLUSTER_SIZE (64 * 1024) 520 #if HAMMER_CLUSTER_SIZE > MAXBSIZE 521 #undef HAMMER_CLUSTER_SIZE 522 #define HAMMER_CLUSTER_SIZE MAXBSIZE 523 #endif 524 #define HAMMER_CLUSTER_BUFS (HAMMER_CLUSTER_SIZE / HAMMER_BUFSIZE) 525 526 /* 527 * In-memory volume representing on-disk buffer 528 */ 529 struct hammer_volume { 530 struct hammer_io io; 531 RB_ENTRY(hammer_volume) rb_node; 532 struct hammer_volume_ondisk *ondisk; 533 int32_t vol_no; 534 int64_t nblocks; /* note: special calculation for statfs */ 535 int64_t buffer_base; /* base offset of buffer 0 */ 536 hammer_off_t maxbuf_off; /* Maximum buffer offset (zone-2) */ 537 hammer_off_t maxraw_off; /* Maximum raw offset for device */ 538 char *vol_name; 539 struct vnode *devvp; 540 int vol_flags; 541 }; 542 543 typedef struct hammer_volume *hammer_volume_t; 544 545 /* 546 * In-memory buffer (other then volume, super-cluster, or cluster), 547 * representing an on-disk buffer. 548 */ 549 struct hammer_buffer { 550 struct hammer_io io; 551 RB_ENTRY(hammer_buffer) rb_node; 552 void *ondisk; 553 hammer_off_t zoneX_offset; 554 hammer_off_t zone2_offset; 555 struct hammer_reserve *resv; 556 struct hammer_node_list clist; 557 }; 558 559 typedef struct hammer_buffer *hammer_buffer_t; 560 561 /* 562 * In-memory B-Tree node, representing an on-disk B-Tree node. 563 * 564 * This is a hang-on structure which is backed by a hammer_buffer, 565 * indexed by a hammer_cluster, and used for fine-grained locking of 566 * B-Tree nodes in order to properly control lock ordering. A hammer_buffer 567 * can contain multiple nodes representing wildly disassociated portions 568 * of the B-Tree so locking cannot be done on a buffer-by-buffer basis. 569 * 570 * This structure uses a cluster-relative index to reduce the number 571 * of layers required to access it, and also because all on-disk B-Tree 572 * references are cluster-relative offsets. 573 */ 574 struct hammer_node { 575 struct hammer_lock lock; /* node-by-node lock */ 576 TAILQ_ENTRY(hammer_node) entry; /* per-buffer linkage */ 577 RB_ENTRY(hammer_node) rb_node; /* per-cluster linkage */ 578 hammer_off_t node_offset; /* full offset spec */ 579 struct hammer_mount *hmp; 580 struct hammer_buffer *buffer; /* backing buffer */ 581 hammer_node_ondisk_t ondisk; /* ptr to on-disk structure */ 582 TAILQ_HEAD(, hammer_cursor) cursor_list; /* deadlock recovery */ 583 struct hammer_node_cache_list cache_list; /* passive caches */ 584 int flags; 585 int loading; /* load interlock */ 586 }; 587 588 #define HAMMER_NODE_DELETED 0x0001 589 #define HAMMER_NODE_FLUSH 0x0002 590 #define HAMMER_NODE_CRCGOOD 0x0004 591 #define HAMMER_NODE_NEEDSCRC 0x0008 592 #define HAMMER_NODE_NEEDSMIRROR 0x0010 593 #define HAMMER_NODE_CRCBAD 0x0020 594 595 #define HAMMER_NODE_CRCANY (HAMMER_NODE_CRCGOOD | HAMMER_NODE_CRCBAD) 596 597 typedef struct hammer_node *hammer_node_t; 598 599 /* 600 * List of locked nodes. This structure is used to lock potentially large 601 * numbers of nodes as an aid for complex B-Tree operations. 602 */ 603 struct hammer_node_lock; 604 TAILQ_HEAD(hammer_node_lock_list, hammer_node_lock); 605 606 struct hammer_node_lock { 607 TAILQ_ENTRY(hammer_node_lock) entry; 608 struct hammer_node_lock_list list; 609 struct hammer_node_lock *parent; 610 hammer_node_t node; 611 hammer_node_ondisk_t copy; /* copy of on-disk data */ 612 int index; /* index of this node in parent */ 613 int count; /* count children */ 614 int flags; 615 }; 616 617 typedef struct hammer_node_lock *hammer_node_lock_t; 618 619 #define HAMMER_NODE_LOCK_UPDATED 0x0001 620 621 /* 622 * Common I/O management structure - embedded in in-memory structures 623 * which are backed by filesystem buffers. 624 */ 625 union hammer_io_structure { 626 struct hammer_io io; 627 struct hammer_volume volume; 628 struct hammer_buffer buffer; 629 }; 630 631 typedef union hammer_io_structure *hammer_io_structure_t; 632 633 /* 634 * The reserve structure prevents the blockmap from allocating 635 * out of a reserved bigblock. Such reservations are used by 636 * the direct-write mechanism. 637 * 638 * The structure is also used to hold off on reallocations of 639 * big blocks from the freemap until flush dependancies have 640 * been dealt with. 641 */ 642 struct hammer_reserve { 643 RB_ENTRY(hammer_reserve) rb_node; 644 TAILQ_ENTRY(hammer_reserve) delay_entry; 645 int flush_group; 646 int flags; 647 int refs; 648 int zone; 649 int append_off; 650 hammer_off_t zone_offset; 651 }; 652 653 typedef struct hammer_reserve *hammer_reserve_t; 654 655 #define HAMMER_RESF_ONDELAY 0x0001 656 #define HAMMER_RESF_LAYER2FREE 0x0002 657 658 #include "hammer_cursor.h" 659 660 /* 661 * The undo structure tracks recent undos to avoid laying down duplicate 662 * undos within a flush group, saving us a significant amount of overhead. 663 * 664 * This is strictly a heuristic. 665 */ 666 #define HAMMER_MAX_UNDOS 1024 667 #define HAMMER_MAX_FLUSHERS 4 668 669 struct hammer_undo { 670 RB_ENTRY(hammer_undo) rb_node; 671 TAILQ_ENTRY(hammer_undo) lru_entry; 672 hammer_off_t offset; 673 int bytes; 674 }; 675 676 typedef struct hammer_undo *hammer_undo_t; 677 678 struct hammer_flusher_info; 679 TAILQ_HEAD(hammer_flusher_info_list, hammer_flusher_info); 680 681 struct hammer_flusher { 682 int signal; /* flusher thread sequencer */ 683 int act; /* currently active flush group */ 684 int done; /* set to act when complete */ 685 int next; /* next flush group */ 686 int group_lock; /* lock sequencing of the next flush */ 687 int exiting; /* request master exit */ 688 thread_t td; /* master flusher thread */ 689 hammer_tid_t tid; /* last flushed transaction id */ 690 int finalize_want; /* serialize finalization */ 691 struct hammer_lock finalize_lock; /* serialize finalization */ 692 struct hammer_transaction trans; /* shared transaction */ 693 struct hammer_flusher_info_list run_list; 694 struct hammer_flusher_info_list ready_list; 695 }; 696 697 /* 698 * Internal hammer mount data structure 699 */ 700 struct hammer_mount { 701 struct mount *mp; 702 /*struct vnode *rootvp;*/ 703 struct hammer_ino_rb_tree rb_inos_root; 704 struct hammer_vol_rb_tree rb_vols_root; 705 struct hammer_nod_rb_tree rb_nods_root; 706 struct hammer_und_rb_tree rb_undo_root; 707 struct hammer_res_rb_tree rb_resv_root; 708 struct hammer_buf_rb_tree rb_bufs_root; 709 struct hammer_pfs_rb_tree rb_pfsm_root; 710 struct hammer_volume *rootvol; 711 struct hammer_base_elm root_btree_beg; 712 struct hammer_base_elm root_btree_end; 713 714 struct malloc_type *m_misc; 715 struct malloc_type *m_inodes; 716 717 int flags; /* HAMMER_MOUNT_xxx flags */ 718 int hflags; 719 int ronly; 720 int nvolumes; 721 int volume_iterator; 722 int master_id; /* -1 or 0-15 - clustering and mirroring */ 723 int version; /* hammer filesystem version to use */ 724 int rsv_inodes; /* reserved space due to dirty inodes */ 725 int64_t rsv_databytes; /* reserved space due to record data */ 726 int rsv_recs; /* reserved space due to dirty records */ 727 int rsv_fromdelay; /* bigblocks reserved due to flush delay */ 728 int undo_rec_limit; /* based on size of undo area */ 729 int last_newrecords; 730 int count_newrecords; 731 732 int inode_reclaims; /* inodes pending reclaim by flusher */ 733 int count_inodes; /* total number of inodes */ 734 int count_iqueued; /* inodes queued to flusher */ 735 736 struct hammer_flusher flusher; 737 738 u_int check_interrupt; 739 uuid_t fsid; 740 struct hammer_io_list volu_list; /* dirty undo buffers */ 741 struct hammer_io_list undo_list; /* dirty undo buffers */ 742 struct hammer_io_list data_list; /* dirty data buffers */ 743 struct hammer_io_list alt_data_list; /* dirty data buffers */ 744 struct hammer_io_list meta_list; /* dirty meta bufs */ 745 struct hammer_io_list lose_list; /* loose buffers */ 746 int locked_dirty_space; /* meta/volu count */ 747 int io_running_space; 748 int objid_cache_count; 749 int error; /* critical I/O error */ 750 struct krate krate; /* rate limited kprintf */ 751 hammer_tid_t asof; /* snapshot mount */ 752 hammer_tid_t next_tid; 753 hammer_tid_t flush_tid1; /* flusher tid sequencing */ 754 hammer_tid_t flush_tid2; /* flusher tid sequencing */ 755 int64_t copy_stat_freebigblocks; /* number of free bigblocks */ 756 757 struct netexport export; 758 struct hammer_lock sync_lock; 759 struct hammer_lock free_lock; 760 struct hammer_lock undo_lock; 761 struct hammer_lock blkmap_lock; 762 struct hammer_blockmap blockmap[HAMMER_MAX_ZONES]; 763 struct hammer_undo undos[HAMMER_MAX_UNDOS]; 764 int undo_alloc; 765 TAILQ_HEAD(, hammer_undo) undo_lru_list; 766 TAILQ_HEAD(, hammer_reserve) delay_list; 767 struct hammer_flush_group_list flush_group_list; 768 hammer_flush_group_t next_flush_group; 769 TAILQ_HEAD(, hammer_objid_cache) objid_cache_list; 770 TAILQ_HEAD(, hammer_reclaim) reclaim_list; 771 }; 772 773 typedef struct hammer_mount *hammer_mount_t; 774 775 #define HAMMER_MOUNT_CRITICAL_ERROR 0x0001 776 #define HAMMER_MOUNT_FLUSH_RECOVERY 0x0002 777 778 struct hammer_sync_info { 779 int error; 780 int waitfor; 781 }; 782 783 #endif 784 785 /* 786 * checkspace slop (8MB chunks), higher numbers are more conservative. 787 */ 788 #define HAMMER_CHKSPC_REBLOCK 25 789 #define HAMMER_CHKSPC_MIRROR 20 790 #define HAMMER_CHKSPC_WRITE 20 791 #define HAMMER_CHKSPC_CREATE 20 792 #define HAMMER_CHKSPC_REMOVE 10 793 #define HAMMER_CHKSPC_EMERGENCY 0 794 795 #if defined(_KERNEL) 796 797 extern struct vop_ops hammer_vnode_vops; 798 extern struct vop_ops hammer_spec_vops; 799 extern struct vop_ops hammer_fifo_vops; 800 extern struct bio_ops hammer_bioops; 801 802 extern int hammer_debug_io; 803 extern int hammer_debug_general; 804 extern int hammer_debug_debug; 805 extern int hammer_debug_inode; 806 extern int hammer_debug_locks; 807 extern int hammer_debug_btree; 808 extern int hammer_debug_tid; 809 extern int hammer_debug_recover; 810 extern int hammer_debug_recover_faults; 811 extern int hammer_cluster_enable; 812 extern int hammer_count_fsyncs; 813 extern int hammer_count_inodes; 814 extern int hammer_count_iqueued; 815 extern int hammer_count_reclaiming; 816 extern int hammer_count_records; 817 extern int hammer_count_record_datas; 818 extern int hammer_count_volumes; 819 extern int hammer_count_buffers; 820 extern int hammer_count_nodes; 821 extern int64_t hammer_count_extra_space_used; 822 extern int64_t hammer_stats_btree_lookups; 823 extern int64_t hammer_stats_btree_searches; 824 extern int64_t hammer_stats_btree_inserts; 825 extern int64_t hammer_stats_btree_deletes; 826 extern int64_t hammer_stats_btree_elements; 827 extern int64_t hammer_stats_btree_splits; 828 extern int64_t hammer_stats_btree_iterations; 829 extern int64_t hammer_stats_record_iterations; 830 extern int64_t hammer_stats_file_read; 831 extern int64_t hammer_stats_file_write; 832 extern int64_t hammer_stats_file_iopsr; 833 extern int64_t hammer_stats_file_iopsw; 834 extern int64_t hammer_stats_disk_read; 835 extern int64_t hammer_stats_disk_write; 836 extern int64_t hammer_stats_inode_flushes; 837 extern int64_t hammer_stats_commits; 838 extern int hammer_count_dirtybufspace; 839 extern int hammer_count_refedbufs; 840 extern int hammer_count_reservations; 841 extern int hammer_count_io_running_read; 842 extern int hammer_count_io_running_write; 843 extern int hammer_count_io_locked; 844 extern int hammer_limit_dirtybufspace; 845 extern int hammer_limit_recs; 846 extern int hammer_bio_count; 847 extern int hammer_verify_zone; 848 extern int hammer_verify_data; 849 extern int hammer_write_mode; 850 extern int hammer_autoflush; 851 extern int64_t hammer_contention_count; 852 853 void hammer_critical_error(hammer_mount_t hmp, hammer_inode_t ip, 854 int error, const char *msg); 855 int hammer_vop_inactive(struct vop_inactive_args *); 856 int hammer_vop_reclaim(struct vop_reclaim_args *); 857 int hammer_get_vnode(struct hammer_inode *ip, struct vnode **vpp); 858 struct hammer_inode *hammer_get_inode(hammer_transaction_t trans, 859 hammer_inode_t dip, int64_t obj_id, 860 hammer_tid_t asof, u_int32_t localization, 861 int flags, int *errorp); 862 struct hammer_inode *hammer_get_dummy_inode(hammer_transaction_t trans, 863 hammer_inode_t dip, int64_t obj_id, 864 hammer_tid_t asof, u_int32_t localization, 865 int flags, int *errorp); 866 void hammer_scan_inode_snapshots(hammer_mount_t hmp, 867 hammer_inode_info_t iinfo, 868 int (*callback)(hammer_inode_t ip, void *data), 869 void *data); 870 void hammer_put_inode(struct hammer_inode *ip); 871 void hammer_put_inode_ref(struct hammer_inode *ip); 872 void hammer_inode_waitreclaims(hammer_mount_t hmp); 873 void hammer_inode_waithard(hammer_mount_t hmp); 874 875 int hammer_unload_volume(hammer_volume_t volume, void *data __unused); 876 int hammer_adjust_volume_mode(hammer_volume_t volume, void *data __unused); 877 878 int hammer_unload_buffer(hammer_buffer_t buffer, void *data __unused); 879 int hammer_install_volume(hammer_mount_t hmp, const char *volname, 880 struct vnode *devvp); 881 int hammer_mountcheck_volumes(hammer_mount_t hmp); 882 883 int hammer_mem_add(hammer_record_t record); 884 int hammer_ip_lookup(hammer_cursor_t cursor); 885 int hammer_ip_first(hammer_cursor_t cursor); 886 int hammer_ip_next(hammer_cursor_t cursor); 887 int hammer_ip_resolve_data(hammer_cursor_t cursor); 888 int hammer_ip_delete_record(hammer_cursor_t cursor, hammer_inode_t ip, 889 hammer_tid_t tid); 890 int hammer_delete_at_cursor(hammer_cursor_t cursor, int delete_flags, 891 hammer_tid_t delete_tid, u_int32_t delete_ts, 892 int track, int64_t *stat_bytes); 893 int hammer_ip_check_directory_empty(hammer_transaction_t trans, 894 hammer_inode_t ip); 895 int hammer_sync_hmp(hammer_mount_t hmp, int waitfor); 896 int hammer_queue_inodes_flusher(hammer_mount_t hmp, int waitfor); 897 898 hammer_record_t 899 hammer_alloc_mem_record(hammer_inode_t ip, int data_len); 900 void hammer_flush_record_done(hammer_record_t record, int error); 901 void hammer_wait_mem_record_ident(hammer_record_t record, const char *ident); 902 void hammer_rel_mem_record(hammer_record_t record); 903 904 int hammer_cursor_up(hammer_cursor_t cursor); 905 int hammer_cursor_up_locked(hammer_cursor_t cursor); 906 int hammer_cursor_down(hammer_cursor_t cursor); 907 int hammer_cursor_upgrade(hammer_cursor_t cursor); 908 int hammer_cursor_upgrade_node(hammer_cursor_t cursor); 909 void hammer_cursor_downgrade(hammer_cursor_t cursor); 910 int hammer_cursor_seek(hammer_cursor_t cursor, hammer_node_t node, 911 int index); 912 void hammer_lock_ex_ident(struct hammer_lock *lock, const char *ident); 913 int hammer_lock_ex_try(struct hammer_lock *lock); 914 void hammer_lock_sh(struct hammer_lock *lock); 915 int hammer_lock_sh_try(struct hammer_lock *lock); 916 int hammer_lock_upgrade(struct hammer_lock *lock); 917 void hammer_lock_downgrade(struct hammer_lock *lock); 918 int hammer_lock_status(struct hammer_lock *lock); 919 void hammer_unlock(struct hammer_lock *lock); 920 void hammer_ref(struct hammer_lock *lock); 921 void hammer_unref(struct hammer_lock *lock); 922 923 void hammer_sync_lock_ex(hammer_transaction_t trans); 924 void hammer_sync_lock_sh(hammer_transaction_t trans); 925 int hammer_sync_lock_sh_try(hammer_transaction_t trans); 926 void hammer_sync_unlock(hammer_transaction_t trans); 927 928 u_int32_t hammer_to_unix_xid(uuid_t *uuid); 929 void hammer_guid_to_uuid(uuid_t *uuid, u_int32_t guid); 930 void hammer_time_to_timespec(u_int64_t xtime, struct timespec *ts); 931 u_int64_t hammer_timespec_to_time(struct timespec *ts); 932 int hammer_str_to_tid(const char *str, int *ispfsp, 933 hammer_tid_t *tidp, u_int32_t *localizationp); 934 int hammer_is_atatext(const char *name, int len); 935 hammer_tid_t hammer_alloc_objid(hammer_mount_t hmp, hammer_inode_t dip); 936 void hammer_clear_objid(hammer_inode_t dip); 937 void hammer_destroy_objid_cache(hammer_mount_t hmp); 938 939 int hammer_enter_undo_history(hammer_mount_t hmp, hammer_off_t offset, 940 int bytes); 941 void hammer_clear_undo_history(hammer_mount_t hmp); 942 enum vtype hammer_get_vnode_type(u_int8_t obj_type); 943 int hammer_get_dtype(u_int8_t obj_type); 944 u_int8_t hammer_get_obj_type(enum vtype vtype); 945 int64_t hammer_directory_namekey(hammer_inode_t dip, const void *name, int len, 946 u_int32_t *max_iterationsp); 947 int hammer_nohistory(hammer_inode_t ip); 948 949 int hammer_init_cursor(hammer_transaction_t trans, hammer_cursor_t cursor, 950 hammer_node_cache_t cache, hammer_inode_t ip); 951 void hammer_normalize_cursor(hammer_cursor_t cursor); 952 void hammer_done_cursor(hammer_cursor_t cursor); 953 int hammer_recover_cursor(hammer_cursor_t cursor); 954 void hammer_unlock_cursor(hammer_cursor_t cursor); 955 int hammer_lock_cursor(hammer_cursor_t cursor); 956 hammer_cursor_t hammer_push_cursor(hammer_cursor_t ocursor); 957 void hammer_pop_cursor(hammer_cursor_t ocursor, hammer_cursor_t ncursor); 958 959 void hammer_cursor_replaced_node(hammer_node_t onode, hammer_node_t nnode); 960 void hammer_cursor_removed_node(hammer_node_t onode, hammer_node_t parent, 961 int index); 962 void hammer_cursor_split_node(hammer_node_t onode, hammer_node_t nnode, 963 int index); 964 void hammer_cursor_moved_element(hammer_node_t onode, hammer_node_t nnode, 965 int oindex, int nindex); 966 void hammer_cursor_parent_changed(hammer_node_t node, hammer_node_t oparent, 967 hammer_node_t nparent, int nindex); 968 void hammer_cursor_inserted_element(hammer_node_t node, int index); 969 void hammer_cursor_deleted_element(hammer_node_t node, int index); 970 971 int hammer_btree_lookup(hammer_cursor_t cursor); 972 int hammer_btree_first(hammer_cursor_t cursor); 973 int hammer_btree_last(hammer_cursor_t cursor); 974 int hammer_btree_extract(hammer_cursor_t cursor, int flags); 975 int hammer_btree_iterate(hammer_cursor_t cursor); 976 int hammer_btree_iterate_reverse(hammer_cursor_t cursor); 977 int hammer_btree_insert(hammer_cursor_t cursor, 978 hammer_btree_leaf_elm_t elm, int *doprop); 979 int hammer_btree_delete(hammer_cursor_t cursor); 980 void hammer_btree_do_propagation(hammer_cursor_t cursor, 981 hammer_pseudofs_inmem_t pfsm, 982 hammer_btree_leaf_elm_t leaf); 983 int hammer_btree_cmp(hammer_base_elm_t key1, hammer_base_elm_t key2); 984 int hammer_btree_chkts(hammer_tid_t ts, hammer_base_elm_t key); 985 int hammer_btree_correct_rhb(hammer_cursor_t cursor, hammer_tid_t tid); 986 int hammer_btree_correct_lhb(hammer_cursor_t cursor, hammer_tid_t tid); 987 988 int btree_set_parent(hammer_transaction_t trans, hammer_node_t node, 989 hammer_btree_elm_t elm); 990 void hammer_node_lock_init(hammer_node_lock_t parent, hammer_node_t node); 991 int hammer_btree_lock_children(hammer_cursor_t cursor, int depth, 992 hammer_node_lock_t parent); 993 void hammer_btree_lock_copy(hammer_cursor_t cursor, 994 hammer_node_lock_t parent); 995 int hammer_btree_sync_copy(hammer_cursor_t cursor, 996 hammer_node_lock_t parent); 997 void hammer_btree_unlock_children(hammer_cursor_t cursor, 998 hammer_node_lock_t parent); 999 int hammer_btree_search_node(hammer_base_elm_t elm, hammer_node_ondisk_t node); 1000 hammer_node_t hammer_btree_get_parent(hammer_transaction_t trans, 1001 hammer_node_t node, int *parent_indexp, 1002 int *errorp, int try_exclusive); 1003 1004 void hammer_print_btree_node(hammer_node_ondisk_t ondisk); 1005 void hammer_print_btree_elm(hammer_btree_elm_t elm, u_int8_t type, int i); 1006 1007 void *hammer_bread(struct hammer_mount *hmp, hammer_off_t off, 1008 int *errorp, struct hammer_buffer **bufferp); 1009 void *hammer_bnew(struct hammer_mount *hmp, hammer_off_t off, 1010 int *errorp, struct hammer_buffer **bufferp); 1011 void *hammer_bread_ext(struct hammer_mount *hmp, hammer_off_t off, int bytes, 1012 int *errorp, struct hammer_buffer **bufferp); 1013 void *hammer_bnew_ext(struct hammer_mount *hmp, hammer_off_t off, int bytes, 1014 int *errorp, struct hammer_buffer **bufferp); 1015 1016 hammer_volume_t hammer_get_root_volume(hammer_mount_t hmp, int *errorp); 1017 1018 hammer_volume_t hammer_get_volume(hammer_mount_t hmp, 1019 int32_t vol_no, int *errorp); 1020 hammer_buffer_t hammer_get_buffer(hammer_mount_t hmp, hammer_off_t buf_offset, 1021 int bytes, int isnew, int *errorp); 1022 void hammer_sync_buffers(hammer_mount_t hmp, 1023 hammer_off_t base_offset, int bytes); 1024 int hammer_del_buffers(hammer_mount_t hmp, 1025 hammer_off_t base_offset, 1026 hammer_off_t zone2_offset, int bytes, 1027 int report_conflicts); 1028 1029 int hammer_ref_volume(hammer_volume_t volume); 1030 int hammer_ref_buffer(hammer_buffer_t buffer); 1031 void hammer_flush_buffer_nodes(hammer_buffer_t buffer); 1032 1033 void hammer_rel_volume(hammer_volume_t volume, int flush); 1034 void hammer_rel_buffer(hammer_buffer_t buffer, int flush); 1035 1036 int hammer_vfs_export(struct mount *mp, int op, 1037 const struct export_args *export); 1038 hammer_node_t hammer_get_node(hammer_transaction_t trans, 1039 hammer_off_t node_offset, int isnew, int *errorp); 1040 void hammer_ref_node(hammer_node_t node); 1041 hammer_node_t hammer_ref_node_safe(hammer_transaction_t trans, 1042 hammer_node_cache_t cache, int *errorp); 1043 void hammer_rel_node(hammer_node_t node); 1044 void hammer_delete_node(hammer_transaction_t trans, 1045 hammer_node_t node); 1046 void hammer_cache_node(hammer_node_cache_t cache, 1047 hammer_node_t node); 1048 void hammer_uncache_node(hammer_node_cache_t cache); 1049 void hammer_flush_node(hammer_node_t node); 1050 1051 void hammer_dup_buffer(struct hammer_buffer **bufferp, 1052 struct hammer_buffer *buffer); 1053 hammer_node_t hammer_alloc_btree(hammer_transaction_t trans, int *errorp); 1054 void *hammer_alloc_data(hammer_transaction_t trans, int32_t data_len, 1055 u_int16_t rec_type, hammer_off_t *data_offsetp, 1056 struct hammer_buffer **data_bufferp, int *errorp); 1057 1058 int hammer_generate_undo(hammer_transaction_t trans, hammer_io_t io, 1059 hammer_off_t zone1_offset, void *base, int len); 1060 1061 void hammer_put_volume(struct hammer_volume *volume, int flush); 1062 void hammer_put_buffer(struct hammer_buffer *buffer, int flush); 1063 1064 hammer_off_t hammer_freemap_alloc(hammer_transaction_t trans, 1065 hammer_off_t owner, int *errorp); 1066 void hammer_freemap_free(hammer_transaction_t trans, hammer_off_t phys_offset, 1067 hammer_off_t owner, int *errorp); 1068 int hammer_checkspace(hammer_mount_t hmp, int slop); 1069 hammer_off_t hammer_blockmap_alloc(hammer_transaction_t trans, int zone, 1070 int bytes, int *errorp); 1071 hammer_reserve_t hammer_blockmap_reserve(hammer_mount_t hmp, int zone, 1072 int bytes, hammer_off_t *zone_offp, int *errorp); 1073 void hammer_blockmap_reserve_complete(hammer_mount_t hmp, 1074 hammer_reserve_t resv); 1075 void hammer_reserve_clrdelay(hammer_mount_t hmp, hammer_reserve_t resv); 1076 void hammer_blockmap_free(hammer_transaction_t trans, 1077 hammer_off_t bmap_off, int bytes); 1078 int hammer_blockmap_finalize(hammer_transaction_t trans, 1079 hammer_reserve_t resv, 1080 hammer_off_t bmap_off, int bytes); 1081 int hammer_blockmap_getfree(hammer_mount_t hmp, hammer_off_t bmap_off, 1082 int *curp, int *errorp); 1083 hammer_off_t hammer_blockmap_lookup(hammer_mount_t hmp, hammer_off_t bmap_off, 1084 int *errorp); 1085 hammer_off_t hammer_undo_lookup(hammer_mount_t hmp, hammer_off_t bmap_off, 1086 int *errorp); 1087 int64_t hammer_undo_used(hammer_transaction_t trans); 1088 int64_t hammer_undo_space(hammer_transaction_t trans); 1089 int64_t hammer_undo_max(hammer_mount_t hmp); 1090 1091 void hammer_start_transaction(struct hammer_transaction *trans, 1092 struct hammer_mount *hmp); 1093 void hammer_simple_transaction(struct hammer_transaction *trans, 1094 struct hammer_mount *hmp); 1095 void hammer_start_transaction_fls(struct hammer_transaction *trans, 1096 struct hammer_mount *hmp); 1097 void hammer_done_transaction(struct hammer_transaction *trans); 1098 1099 void hammer_modify_inode(hammer_inode_t ip, int flags); 1100 void hammer_flush_inode(hammer_inode_t ip, int flags); 1101 void hammer_flush_inode_done(hammer_inode_t ip, int error); 1102 void hammer_wait_inode(hammer_inode_t ip); 1103 1104 int hammer_create_inode(struct hammer_transaction *trans, struct vattr *vap, 1105 struct ucred *cred, struct hammer_inode *dip, 1106 hammer_pseudofs_inmem_t pfsm, 1107 struct hammer_inode **ipp); 1108 void hammer_rel_inode(hammer_inode_t ip, int flush); 1109 int hammer_reload_inode(hammer_inode_t ip, void *arg __unused); 1110 int hammer_ino_rb_compare(hammer_inode_t ip1, hammer_inode_t ip2); 1111 int hammer_destroy_inode_callback(hammer_inode_t ip, void *data __unused); 1112 1113 int hammer_sync_inode(hammer_transaction_t trans, hammer_inode_t ip); 1114 void hammer_test_inode(hammer_inode_t dip); 1115 void hammer_inode_unloadable_check(hammer_inode_t ip, int getvp); 1116 1117 int hammer_ip_add_directory(struct hammer_transaction *trans, 1118 hammer_inode_t dip, const char *name, int bytes, 1119 hammer_inode_t nip); 1120 int hammer_ip_del_directory(struct hammer_transaction *trans, 1121 hammer_cursor_t cursor, hammer_inode_t dip, 1122 hammer_inode_t ip); 1123 hammer_record_t hammer_ip_add_bulk(hammer_inode_t ip, off_t file_offset, 1124 void *data, int bytes, int *errorp); 1125 int hammer_ip_frontend_trunc(struct hammer_inode *ip, off_t file_size); 1126 int hammer_ip_add_record(struct hammer_transaction *trans, 1127 hammer_record_t record); 1128 int hammer_ip_delete_range(hammer_cursor_t cursor, hammer_inode_t ip, 1129 int64_t ran_beg, int64_t ran_end, int truncating); 1130 int hammer_ip_delete_clean(hammer_cursor_t cursor, hammer_inode_t ip, 1131 int *countp); 1132 int hammer_ip_sync_data(hammer_cursor_t cursor, hammer_inode_t ip, 1133 int64_t offset, void *data, int bytes); 1134 int hammer_ip_sync_record(hammer_transaction_t trans, hammer_record_t rec); 1135 int hammer_ip_sync_record_cursor(hammer_cursor_t cursor, hammer_record_t rec); 1136 hammer_pseudofs_inmem_t hammer_load_pseudofs(hammer_transaction_t trans, 1137 u_int32_t localization, int *errorp); 1138 int hammer_mkroot_pseudofs(hammer_transaction_t trans, struct ucred *cred, 1139 hammer_pseudofs_inmem_t pfsm); 1140 int hammer_save_pseudofs(hammer_transaction_t trans, 1141 hammer_pseudofs_inmem_t pfsm); 1142 int hammer_unload_pseudofs(hammer_transaction_t trans, u_int32_t localization); 1143 void hammer_rel_pseudofs(hammer_mount_t hmp, hammer_pseudofs_inmem_t pfsm); 1144 int hammer_ioctl(hammer_inode_t ip, u_long com, caddr_t data, int fflag, 1145 struct ucred *cred); 1146 1147 void hammer_io_init(hammer_io_t io, hammer_volume_t volume, 1148 enum hammer_io_type type); 1149 int hammer_io_read(struct vnode *devvp, struct hammer_io *io, 1150 hammer_off_t limit); 1151 int hammer_io_new(struct vnode *devvp, struct hammer_io *io); 1152 int hammer_io_inval(hammer_volume_t volume, hammer_off_t zone2_offset); 1153 struct buf *hammer_io_release(struct hammer_io *io, int flush); 1154 void hammer_io_flush(struct hammer_io *io); 1155 void hammer_io_wait(struct hammer_io *io); 1156 void hammer_io_waitdep(struct hammer_io *io); 1157 void hammer_io_wait_all(hammer_mount_t hmp, const char *ident); 1158 int hammer_io_direct_read(hammer_mount_t hmp, struct bio *bio, 1159 hammer_btree_leaf_elm_t leaf); 1160 int hammer_io_direct_write(hammer_mount_t hmp, hammer_record_t record, 1161 struct bio *bio); 1162 void hammer_io_direct_wait(hammer_record_t record); 1163 void hammer_io_direct_uncache(hammer_mount_t hmp, hammer_btree_leaf_elm_t leaf); 1164 void hammer_io_write_interlock(hammer_io_t io); 1165 void hammer_io_done_interlock(hammer_io_t io); 1166 void hammer_io_clear_modify(struct hammer_io *io, int inval); 1167 void hammer_io_clear_modlist(struct hammer_io *io); 1168 void hammer_io_flush_sync(hammer_mount_t hmp); 1169 1170 void hammer_modify_volume(hammer_transaction_t trans, hammer_volume_t volume, 1171 void *base, int len); 1172 void hammer_modify_buffer(hammer_transaction_t trans, hammer_buffer_t buffer, 1173 void *base, int len); 1174 void hammer_modify_volume_done(hammer_volume_t volume); 1175 void hammer_modify_buffer_done(hammer_buffer_t buffer); 1176 1177 int hammer_ioc_reblock(hammer_transaction_t trans, hammer_inode_t ip, 1178 struct hammer_ioc_reblock *reblock); 1179 int hammer_ioc_rebalance(hammer_transaction_t trans, hammer_inode_t ip, 1180 struct hammer_ioc_rebalance *rebal); 1181 int hammer_ioc_prune(hammer_transaction_t trans, hammer_inode_t ip, 1182 struct hammer_ioc_prune *prune); 1183 int hammer_ioc_mirror_read(hammer_transaction_t trans, hammer_inode_t ip, 1184 struct hammer_ioc_mirror_rw *mirror); 1185 int hammer_ioc_mirror_write(hammer_transaction_t trans, hammer_inode_t ip, 1186 struct hammer_ioc_mirror_rw *mirror); 1187 int hammer_ioc_set_pseudofs(hammer_transaction_t trans, hammer_inode_t ip, 1188 struct ucred *cred, struct hammer_ioc_pseudofs_rw *pfs); 1189 int hammer_ioc_get_pseudofs(hammer_transaction_t trans, hammer_inode_t ip, 1190 struct hammer_ioc_pseudofs_rw *pfs); 1191 int hammer_ioc_destroy_pseudofs(hammer_transaction_t trans, hammer_inode_t ip, 1192 struct hammer_ioc_pseudofs_rw *pfs); 1193 int hammer_ioc_downgrade_pseudofs(hammer_transaction_t trans, hammer_inode_t ip, 1194 struct hammer_ioc_pseudofs_rw *pfs); 1195 int hammer_ioc_upgrade_pseudofs(hammer_transaction_t trans, hammer_inode_t ip, 1196 struct hammer_ioc_pseudofs_rw *pfs); 1197 int hammer_ioc_wait_pseudofs(hammer_transaction_t trans, hammer_inode_t ip, 1198 struct hammer_ioc_pseudofs_rw *pfs); 1199 1200 int hammer_signal_check(hammer_mount_t hmp); 1201 1202 void hammer_flusher_create(hammer_mount_t hmp); 1203 void hammer_flusher_destroy(hammer_mount_t hmp); 1204 void hammer_flusher_sync(hammer_mount_t hmp); 1205 int hammer_flusher_async(hammer_mount_t hmp, hammer_flush_group_t flg); 1206 int hammer_flusher_async_one(hammer_mount_t hmp); 1207 void hammer_flusher_wait(hammer_mount_t hmp, int seq); 1208 void hammer_flusher_wait_next(hammer_mount_t hmp); 1209 int hammer_flusher_meta_limit(hammer_mount_t hmp); 1210 int hammer_flusher_meta_halflimit(hammer_mount_t hmp); 1211 int hammer_flusher_undo_exhausted(hammer_transaction_t trans, int quarter); 1212 void hammer_flusher_clean_loose_ios(hammer_mount_t hmp); 1213 void hammer_flusher_finalize(hammer_transaction_t trans, int final); 1214 int hammer_flusher_haswork(hammer_mount_t hmp); 1215 1216 1217 int hammer_recover(hammer_mount_t hmp, hammer_volume_t rootvol); 1218 void hammer_recover_flush_buffers(hammer_mount_t hmp, 1219 hammer_volume_t root_volume, int final); 1220 1221 void hammer_crc_set_blockmap(hammer_blockmap_t blockmap); 1222 void hammer_crc_set_volume(hammer_volume_ondisk_t ondisk); 1223 void hammer_crc_set_leaf(void *data, hammer_btree_leaf_elm_t leaf); 1224 1225 int hammer_crc_test_blockmap(hammer_blockmap_t blockmap); 1226 int hammer_crc_test_volume(hammer_volume_ondisk_t ondisk); 1227 int hammer_crc_test_btree(hammer_node_ondisk_t ondisk); 1228 int hammer_crc_test_leaf(void *data, hammer_btree_leaf_elm_t leaf); 1229 void hkprintf(const char *ctl, ...); 1230 udev_t hammer_fsid_to_udev(uuid_t *uuid); 1231 1232 1233 int hammer_blocksize(int64_t file_offset); 1234 int64_t hammer_blockdemarc(int64_t file_offset1, int64_t file_offset2); 1235 1236 #endif 1237 1238 static __inline void 1239 hammer_wait_mem_record(hammer_record_t record) 1240 { 1241 hammer_wait_mem_record_ident(record, "hmmwai"); 1242 } 1243 1244 static __inline void 1245 hammer_lock_ex(struct hammer_lock *lock) 1246 { 1247 hammer_lock_ex_ident(lock, "hmrlck"); 1248 } 1249 1250 /* 1251 * Indicate that a B-Tree node is being modified. 1252 */ 1253 static __inline void 1254 hammer_modify_node_noundo(hammer_transaction_t trans, hammer_node_t node) 1255 { 1256 KKASSERT((node->flags & HAMMER_NODE_CRCBAD) == 0); 1257 hammer_modify_buffer(trans, node->buffer, NULL, 0); 1258 } 1259 1260 static __inline void 1261 hammer_modify_node_all(hammer_transaction_t trans, struct hammer_node *node) 1262 { 1263 KKASSERT((node->flags & HAMMER_NODE_CRCBAD) == 0); 1264 hammer_modify_buffer(trans, node->buffer, 1265 node->ondisk, sizeof(*node->ondisk)); 1266 } 1267 1268 static __inline void 1269 hammer_modify_node(hammer_transaction_t trans, hammer_node_t node, 1270 void *base, int len) 1271 { 1272 hammer_crc_t *crcptr; 1273 1274 KKASSERT((char *)base >= (char *)node->ondisk && 1275 (char *)base + len <= 1276 (char *)node->ondisk + sizeof(*node->ondisk)); 1277 KKASSERT((node->flags & HAMMER_NODE_CRCBAD) == 0); 1278 hammer_modify_buffer(trans, node->buffer, base, len); 1279 crcptr = &node->ondisk->crc; 1280 hammer_modify_buffer(trans, node->buffer, crcptr, sizeof(hammer_crc_t)); 1281 --node->buffer->io.modify_refs; /* only want one ref */ 1282 } 1283 1284 /* 1285 * Indicate that the specified modifications have been completed. 1286 * 1287 * Do not try to generate the crc here, it's very expensive to do and a 1288 * sequence of insertions or deletions can result in many calls to this 1289 * function on the same node. 1290 */ 1291 static __inline void 1292 hammer_modify_node_done(hammer_node_t node) 1293 { 1294 node->flags |= HAMMER_NODE_CRCGOOD; 1295 if ((node->flags & HAMMER_NODE_NEEDSCRC) == 0) { 1296 node->flags |= HAMMER_NODE_NEEDSCRC; 1297 node->buffer->io.gencrc = 1; 1298 hammer_ref_node(node); 1299 } 1300 hammer_modify_buffer_done(node->buffer); 1301 } 1302 1303 #define hammer_modify_volume_field(trans, vol, field) \ 1304 hammer_modify_volume(trans, vol, &(vol)->ondisk->field, \ 1305 sizeof((vol)->ondisk->field)) 1306 1307 #define hammer_modify_node_field(trans, node, field) \ 1308 hammer_modify_node(trans, node, &(node)->ondisk->field, \ 1309 sizeof((node)->ondisk->field)) 1310 1311