1/*- 2 * Copyright 1998, 2000 Marshall Kirk McKusick. All Rights Reserved. 3 * 4 * The soft updates code is derived from the appendix of a University 5 * of Michigan technical report (Gregory R. Ganger and Yale N. Patt, 6 * "Soft Updates: A Solution to the Metadata Update Problem in File 7 * Systems", CSE-TR-254-95, August 1995). 8 * 9 * Further information about soft updates can be obtained from: 10 * 11 * Marshall Kirk McKusick http://www.mckusick.com/softdep/ 12 * 1614 Oxford Street mckusick@mckusick.com 13 * Berkeley, CA 94709-1608 +1-510-843-9542 14 * USA 15 * 16 * Redistribution and use in source and binary forms, with or without 17 * modification, are permitted provided that the following conditions 18 * are met: 19 * 20 * 1. Redistributions of source code must retain the above copyright 21 * notice, this list of conditions and the following disclaimer. 22 * 2. Redistributions in binary form must reproduce the above copyright 23 * notice, this list of conditions and the following disclaimer in the 24 * documentation and/or other materials provided with the distribution. 25 * 26 * THIS SOFTWARE IS PROVIDED BY MARSHALL KIRK MCKUSICK ``AS IS'' AND ANY 27 * EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED 28 * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE 29 * DISCLAIMED. IN NO EVENT SHALL MARSHALL KIRK MCKUSICK BE LIABLE FOR 30 * ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 31 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 32 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 33 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 34 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 35 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 36 * SUCH DAMAGE. 37 * 38 * @(#)softdep.h 9.7 (McKusick) 6/21/00 39 * $FreeBSD$ 40 */ 41 42#include <sys/queue.h> 43 44/* 45 * Allocation dependencies are handled with undo/redo on the in-memory 46 * copy of the data. A particular data dependency is eliminated when 47 * it is ALLCOMPLETE: that is ATTACHED, DEPCOMPLETE, and COMPLETE. 48 * 49 * The ATTACHED flag means that the data is not currently being written 50 * to disk. 51 * 52 * The UNDONE flag means that the data has been rolled back to a safe 53 * state for writing to the disk. When the I/O completes, the data is 54 * restored to its current form and the state reverts to ATTACHED. 55 * The data must be locked throughout the rollback, I/O, and roll 56 * forward so that the rolled back information is never visible to 57 * user processes. 58 * 59 * The COMPLETE flag indicates that the item has been written. For example, 60 * a dependency that requires that an inode be written will be marked 61 * COMPLETE after the inode has been written to disk. 62 * 63 * The DEPCOMPLETE flag indicates the completion of any other 64 * dependencies such as the writing of a cylinder group map has been 65 * completed. A dependency structure may be freed only when both it 66 * and its dependencies have completed and any rollbacks that are in 67 * progress have finished as indicated by the set of ALLCOMPLETE flags 68 * all being set. 69 * 70 * The two MKDIR flags indicate additional dependencies that must be done 71 * when creating a new directory. MKDIR_BODY is cleared when the directory 72 * data block containing the "." and ".." entries has been written. 73 * MKDIR_PARENT is cleared when the parent inode with the increased link 74 * count for ".." has been written. When both MKDIR flags have been 75 * cleared, the DEPCOMPLETE flag is set to indicate that the directory 76 * dependencies have been completed. The writing of the directory inode 77 * itself sets the COMPLETE flag which then allows the directory entry for 78 * the new directory to be written to disk. The RMDIR flag marks a dirrem 79 * structure as representing the removal of a directory rather than a 80 * file. When the removal dependencies are completed, additional work needs 81 * to be done* (an additional decrement of the associated inode, and a 82 * decrement of the parent inode). 83 * 84 * The DIRCHG flag marks a diradd structure as representing the changing 85 * of an existing entry rather than the addition of a new one. When 86 * the update is complete the dirrem associated with the inode for 87 * the old name must be added to the worklist to do the necessary 88 * reference count decrement. 89 * 90 * The GOINGAWAY flag indicates that the data structure is frozen from 91 * further change until its dependencies have been completed and its 92 * resources freed after which it will be discarded. 93 * 94 * The IOSTARTED flag prevents multiple calls to the I/O start routine from 95 * doing multiple rollbacks. 96 * 97 * The NEWBLOCK flag marks pagedep structures that have just been allocated, 98 * so must be claimed by the inode before all dependencies are complete. 99 * 100 * The INPROGRESS flag marks worklist structures that are still on the 101 * worklist, but are being considered for action by some process. 102 * 103 * The UFS1FMT flag indicates that the inode being processed is a ufs1 format. 104 * 105 * The EXTDATA flag indicates that the allocdirect describes an 106 * extended-attributes dependency. 107 * 108 * The ONWORKLIST flag shows whether the structure is currently linked 109 * onto a worklist. 110 * 111 * The UNLINK* flags track the progress of updating the on-disk linked 112 * list of active but unlinked inodes. When an inode is first unlinked 113 * it is marked as UNLINKED. When its on-disk di_freelink has been 114 * written its UNLINKNEXT flags is set. When its predecessor in the 115 * list has its di_freelink pointing at us its UNLINKPREV is set. 116 * When the on-disk list can reach it from the superblock, its 117 * UNLINKONLIST flag is set. Once all of these flags are set, it 118 * is safe to let its last name be removed. 119 */ 120#define ATTACHED 0x000001 121#define UNDONE 0x000002 122#define COMPLETE 0x000004 123#define DEPCOMPLETE 0x000008 124#define MKDIR_PARENT 0x000010 /* diradd, mkdir, jaddref, jsegdep only */ 125#define MKDIR_BODY 0x000020 /* diradd, mkdir, jaddref only */ 126#define RMDIR 0x000040 /* dirrem only */ 127#define DIRCHG 0x000080 /* diradd, dirrem only */ 128#define GOINGAWAY 0x000100 /* indirdep, jremref only */ 129#define IOSTARTED 0x000200 /* inodedep, pagedep, bmsafemap only */ 130#define DELAYEDFREE 0x000400 /* allocindirect free delayed. */ 131#define NEWBLOCK 0x000800 /* pagedep, jaddref only */ 132#define INPROGRESS 0x001000 /* dirrem, freeblks, freefrag, freefile only */ 133#define UFS1FMT 0x002000 /* indirdep only */ 134#define EXTDATA 0x004000 /* allocdirect only */ 135#define ONWORKLIST 0x008000 136#define IOWAITING 0x010000 /* Thread is waiting for IO to complete. */ 137#define ONDEPLIST 0x020000 /* Structure is on a dependency list. */ 138#define UNLINKED 0x040000 /* inodedep has been unlinked. */ 139#define UNLINKNEXT 0x080000 /* inodedep has valid di_freelink */ 140#define UNLINKPREV 0x100000 /* inodedep is pointed at in the unlink list */ 141#define UNLINKONLIST 0x200000 /* inodedep is in the unlinked list on disk */ 142#define UNLINKLINKS (UNLINKNEXT | UNLINKPREV) 143 144#define ALLCOMPLETE (ATTACHED | COMPLETE | DEPCOMPLETE) 145 146/* 147 * Values for each of the soft dependency types. 148 */ 149#define D_PAGEDEP 0 150#define D_INODEDEP 1 151#define D_BMSAFEMAP 2 152#define D_NEWBLK 3 153#define D_ALLOCDIRECT 4 154#define D_INDIRDEP 5 155#define D_ALLOCINDIR 6 156#define D_FREEFRAG 7 157#define D_FREEBLKS 8 158#define D_FREEFILE 9 159#define D_DIRADD 10 160#define D_MKDIR 11 161#define D_DIRREM 12 162#define D_NEWDIRBLK 13 163#define D_FREEWORK 14 164#define D_FREEDEP 15 165#define D_JADDREF 16 166#define D_JREMREF 17 167#define D_JMVREF 18 168#define D_JNEWBLK 19 169#define D_JFREEBLK 20 170#define D_JFREEFRAG 21 171#define D_JSEG 22 172#define D_JSEGDEP 23 173#define D_SBDEP 24 174#define D_JTRUNC 25 175#define D_JFSYNC 26 176#define D_SENTINEL 27 177#define D_LAST D_SENTINEL 178 179/* 180 * The workitem queue. 181 * 182 * It is sometimes useful and/or necessary to clean up certain dependencies 183 * in the background rather than during execution of an application process 184 * or interrupt service routine. To realize this, we append dependency 185 * structures corresponding to such tasks to a "workitem" queue. In a soft 186 * updates implementation, most pending workitems should not wait for more 187 * than a couple of seconds, so the filesystem syncer process awakens once 188 * per second to process the items on the queue. 189 */ 190 191/* LIST_HEAD(workhead, worklist); -- declared in buf.h */ 192 193/* 194 * Each request can be linked onto a work queue through its worklist structure. 195 * To avoid the need for a pointer to the structure itself, this structure 196 * MUST be declared FIRST in each type in which it appears! If more than one 197 * worklist is needed in the structure, then a wk_data field must be added 198 * and the macros below changed to use it. 199 */ 200struct worklist { 201 LIST_ENTRY(worklist) wk_list; /* list of work requests */ 202 struct mount *wk_mp; /* Mount we live in */ 203 unsigned int wk_type:8, /* type of request */ 204 wk_state:24; /* state flags */ 205}; 206#define WK_DATA(wk) ((void *)(wk)) 207#define WK_PAGEDEP(wk) ((struct pagedep *)(wk)) 208#define WK_INODEDEP(wk) ((struct inodedep *)(wk)) 209#define WK_BMSAFEMAP(wk) ((struct bmsafemap *)(wk)) 210#define WK_NEWBLK(wk) ((struct newblk *)(wk)) 211#define WK_ALLOCDIRECT(wk) ((struct allocdirect *)(wk)) 212#define WK_INDIRDEP(wk) ((struct indirdep *)(wk)) 213#define WK_ALLOCINDIR(wk) ((struct allocindir *)(wk)) 214#define WK_FREEFRAG(wk) ((struct freefrag *)(wk)) 215#define WK_FREEBLKS(wk) ((struct freeblks *)(wk)) 216#define WK_FREEWORK(wk) ((struct freework *)(wk)) 217#define WK_FREEFILE(wk) ((struct freefile *)(wk)) 218#define WK_DIRADD(wk) ((struct diradd *)(wk)) 219#define WK_MKDIR(wk) ((struct mkdir *)(wk)) 220#define WK_DIRREM(wk) ((struct dirrem *)(wk)) 221#define WK_NEWDIRBLK(wk) ((struct newdirblk *)(wk)) 222#define WK_JADDREF(wk) ((struct jaddref *)(wk)) 223#define WK_JREMREF(wk) ((struct jremref *)(wk)) 224#define WK_JMVREF(wk) ((struct jmvref *)(wk)) 225#define WK_JSEGDEP(wk) ((struct jsegdep *)(wk)) 226#define WK_JSEG(wk) ((struct jseg *)(wk)) 227#define WK_JNEWBLK(wk) ((struct jnewblk *)(wk)) 228#define WK_JFREEBLK(wk) ((struct jfreeblk *)(wk)) 229#define WK_FREEDEP(wk) ((struct freedep *)(wk)) 230#define WK_JFREEFRAG(wk) ((struct jfreefrag *)(wk)) 231#define WK_SBDEP(wk) ((struct sbdep *)(wk)) 232#define WK_JTRUNC(wk) ((struct jtrunc *)(wk)) 233#define WK_JFSYNC(wk) ((struct jfsync *)(wk)) 234 235/* 236 * Various types of lists 237 */ 238LIST_HEAD(dirremhd, dirrem); 239LIST_HEAD(diraddhd, diradd); 240LIST_HEAD(newblkhd, newblk); 241LIST_HEAD(inodedephd, inodedep); 242LIST_HEAD(allocindirhd, allocindir); 243LIST_HEAD(allocdirecthd, allocdirect); 244TAILQ_HEAD(allocdirectlst, allocdirect); 245LIST_HEAD(indirdephd, indirdep); 246LIST_HEAD(jaddrefhd, jaddref); 247LIST_HEAD(jremrefhd, jremref); 248LIST_HEAD(jmvrefhd, jmvref); 249LIST_HEAD(jnewblkhd, jnewblk); 250LIST_HEAD(jblkdephd, jblkdep); 251LIST_HEAD(freeworkhd, freework); 252TAILQ_HEAD(freeworklst, freework); 253TAILQ_HEAD(jseglst, jseg); 254TAILQ_HEAD(inoreflst, inoref); 255TAILQ_HEAD(freeblklst, freeblks); 256 257/* 258 * The "pagedep" structure tracks the various dependencies related to 259 * a particular directory page. If a directory page has any dependencies, 260 * it will have a pagedep linked to its associated buffer. The 261 * pd_dirremhd list holds the list of dirrem requests which decrement 262 * inode reference counts. These requests are processed after the 263 * directory page with the corresponding zero'ed entries has been 264 * written. The pd_diraddhd list maintains the list of diradd requests 265 * which cannot be committed until their corresponding inode has been 266 * written to disk. Because a directory may have many new entries 267 * being created, several lists are maintained hashed on bits of the 268 * offset of the entry into the directory page to keep the lists from 269 * getting too long. Once a new directory entry has been cleared to 270 * be written, it is moved to the pd_pendinghd list. After the new 271 * entry has been written to disk it is removed from the pd_pendinghd 272 * list, any removed operations are done, and the dependency structure 273 * is freed. 274 */ 275#define DAHASHSZ 5 276#define DIRADDHASH(offset) (((offset) >> 2) % DAHASHSZ) 277struct pagedep { 278 struct worklist pd_list; /* page buffer */ 279# define pd_state pd_list.wk_state /* check for multiple I/O starts */ 280 LIST_ENTRY(pagedep) pd_hash; /* hashed lookup */ 281 ino_t pd_ino; /* associated file */ 282 ufs_lbn_t pd_lbn; /* block within file */ 283 struct newdirblk *pd_newdirblk; /* associated newdirblk if NEWBLOCK */ 284 struct dirremhd pd_dirremhd; /* dirrem's waiting for page */ 285 struct diraddhd pd_diraddhd[DAHASHSZ]; /* diradd dir entry updates */ 286 struct diraddhd pd_pendinghd; /* directory entries awaiting write */ 287 struct jmvrefhd pd_jmvrefhd; /* Dependent journal writes. */ 288}; 289 290/* 291 * The "inodedep" structure tracks the set of dependencies associated 292 * with an inode. One task that it must manage is delayed operations 293 * (i.e., work requests that must be held until the inodedep's associated 294 * inode has been written to disk). Getting an inode from its incore 295 * state to the disk requires two steps to be taken by the filesystem 296 * in this order: first the inode must be copied to its disk buffer by 297 * the VOP_UPDATE operation; second the inode's buffer must be written 298 * to disk. To ensure that both operations have happened in the required 299 * order, the inodedep maintains two lists. Delayed operations are 300 * placed on the id_inowait list. When the VOP_UPDATE is done, all 301 * operations on the id_inowait list are moved to the id_bufwait list. 302 * When the buffer is written, the items on the id_bufwait list can be 303 * safely moved to the work queue to be processed. A second task of the 304 * inodedep structure is to track the status of block allocation within 305 * the inode. Each block that is allocated is represented by an 306 * "allocdirect" structure (see below). It is linked onto the id_newinoupdt 307 * list until both its contents and its allocation in the cylinder 308 * group map have been written to disk. Once these dependencies have been 309 * satisfied, it is removed from the id_newinoupdt list and any followup 310 * actions such as releasing the previous block or fragment are placed 311 * on the id_inowait list. When an inode is updated (a VOP_UPDATE is 312 * done), the "inodedep" structure is linked onto the buffer through 313 * its worklist. Thus, it will be notified when the buffer is about 314 * to be written and when it is done. At the update time, all the 315 * elements on the id_newinoupdt list are moved to the id_inoupdt list 316 * since those changes are now relevant to the copy of the inode in the 317 * buffer. Also at update time, the tasks on the id_inowait list are 318 * moved to the id_bufwait list so that they will be executed when 319 * the updated inode has been written to disk. When the buffer containing 320 * the inode is written to disk, any updates listed on the id_inoupdt 321 * list are rolled back as they are not yet safe. Following the write, 322 * the changes are once again rolled forward and any actions on the 323 * id_bufwait list are processed (since those actions are now safe). 324 * The entries on the id_inoupdt and id_newinoupdt lists must be kept 325 * sorted by logical block number to speed the calculation of the size 326 * of the rolled back inode (see explanation in initiate_write_inodeblock). 327 * When a directory entry is created, it is represented by a diradd. 328 * The diradd is added to the id_inowait list as it cannot be safely 329 * written to disk until the inode that it represents is on disk. After 330 * the inode is written, the id_bufwait list is processed and the diradd 331 * entries are moved to the id_pendinghd list where they remain until 332 * the directory block containing the name has been written to disk. 333 * The purpose of keeping the entries on the id_pendinghd list is so that 334 * the softdep_fsync function can find and push the inode's directory 335 * name(s) as part of the fsync operation for that file. 336 */ 337struct inodedep { 338 struct worklist id_list; /* buffer holding inode block */ 339# define id_state id_list.wk_state /* inode dependency state */ 340 LIST_ENTRY(inodedep) id_hash; /* hashed lookup */ 341 TAILQ_ENTRY(inodedep) id_unlinked; /* Unlinked but ref'd inodes */ 342 struct fs *id_fs; /* associated filesystem */ 343 ino_t id_ino; /* dependent inode */ 344 nlink_t id_nlinkdelta; /* saved effective link count */ 345 nlink_t id_savednlink; /* Link saved during rollback */ 346 LIST_ENTRY(inodedep) id_deps; /* bmsafemap's list of inodedep's */ 347 struct bmsafemap *id_bmsafemap; /* related bmsafemap (if pending) */ 348 struct diradd *id_mkdiradd; /* diradd for a mkdir. */ 349 struct inoreflst id_inoreflst; /* Inode reference adjustments. */ 350 long id_savedextsize; /* ext size saved during rollback */ 351 off_t id_savedsize; /* file size saved during rollback */ 352 struct dirremhd id_dirremhd; /* Removals pending. */ 353 struct workhead id_pendinghd; /* entries awaiting directory write */ 354 struct workhead id_bufwait; /* operations after inode written */ 355 struct workhead id_inowait; /* operations waiting inode update */ 356 struct allocdirectlst id_inoupdt; /* updates before inode written */ 357 struct allocdirectlst id_newinoupdt; /* updates when inode written */ 358 struct allocdirectlst id_extupdt; /* extdata updates pre-inode write */ 359 struct allocdirectlst id_newextupdt; /* extdata updates at ino write */ 360 struct freeblklst id_freeblklst; /* List of partial truncates. */ 361 union { 362 struct ufs1_dinode *idu_savedino1; /* saved ufs1_dinode contents */ 363 struct ufs2_dinode *idu_savedino2; /* saved ufs2_dinode contents */ 364 } id_un; 365}; 366#define id_savedino1 id_un.idu_savedino1 367#define id_savedino2 id_un.idu_savedino2 368 369/* 370 * A "bmsafemap" structure maintains a list of dependency structures 371 * that depend on the update of a particular cylinder group map. 372 * It has lists for newblks, allocdirects, allocindirs, and inodedeps. 373 * It is attached to the buffer of a cylinder group block when any of 374 * these things are allocated from the cylinder group. It is freed 375 * after the cylinder group map is written and the state of its 376 * dependencies are updated with DEPCOMPLETE to indicate that it has 377 * been processed. 378 */ 379struct bmsafemap { 380 struct worklist sm_list; /* cylgrp buffer */ 381# define sm_state sm_list.wk_state 382 LIST_ENTRY(bmsafemap) sm_hash; /* Hash links. */ 383 LIST_ENTRY(bmsafemap) sm_next; /* Mount list. */ 384 int sm_cg; 385 struct buf *sm_buf; /* associated buffer */ 386 struct allocdirecthd sm_allocdirecthd; /* allocdirect deps */ 387 struct allocdirecthd sm_allocdirectwr; /* writing allocdirect deps */ 388 struct allocindirhd sm_allocindirhd; /* allocindir deps */ 389 struct allocindirhd sm_allocindirwr; /* writing allocindir deps */ 390 struct inodedephd sm_inodedephd; /* inodedep deps */ 391 struct inodedephd sm_inodedepwr; /* writing inodedep deps */ 392 struct newblkhd sm_newblkhd; /* newblk deps */ 393 struct newblkhd sm_newblkwr; /* writing newblk deps */ 394 struct jaddrefhd sm_jaddrefhd; /* Pending inode allocations. */ 395 struct jnewblkhd sm_jnewblkhd; /* Pending block allocations. */ 396 struct workhead sm_freehd; /* Freedep deps. */ 397 struct workhead sm_freewr; /* Written freedeps. */ 398}; 399 400/* 401 * A "newblk" structure is attached to a bmsafemap structure when a block 402 * or fragment is allocated from a cylinder group. Its state is set to 403 * DEPCOMPLETE when its cylinder group map is written. It is converted to 404 * an allocdirect or allocindir allocation once the allocator calls the 405 * appropriate setup function. It will initially be linked onto a bmsafemap 406 * list. Once converted it can be linked onto the lists described for 407 * allocdirect or allocindir as described below. 408 */ 409struct newblk { 410 struct worklist nb_list; /* See comment above. */ 411# define nb_state nb_list.wk_state 412 LIST_ENTRY(newblk) nb_hash; /* Hashed lookup. */ 413 LIST_ENTRY(newblk) nb_deps; /* Bmsafemap's list of newblks. */ 414 struct jnewblk *nb_jnewblk; /* New block journal entry. */ 415 struct bmsafemap *nb_bmsafemap;/* Cylgrp dep (if pending). */ 416 struct freefrag *nb_freefrag; /* Fragment to be freed (if any). */ 417 struct indirdephd nb_indirdeps; /* Children indirect blocks. */ 418 struct workhead nb_newdirblk; /* Dir block to notify when written. */ 419 struct workhead nb_jwork; /* Journal work pending. */ 420 ufs2_daddr_t nb_newblkno; /* New value of block pointer. */ 421}; 422 423/* 424 * An "allocdirect" structure is attached to an "inodedep" when a new block 425 * or fragment is allocated and pointed to by the inode described by 426 * "inodedep". The worklist is linked to the buffer that holds the block. 427 * When the block is first allocated, it is linked to the bmsafemap 428 * structure associated with the buffer holding the cylinder group map 429 * from which it was allocated. When the cylinder group map is written 430 * to disk, ad_state has the DEPCOMPLETE flag set. When the block itself 431 * is written, the COMPLETE flag is set. Once both the cylinder group map 432 * and the data itself have been written, it is safe to write the inode 433 * that claims the block. If there was a previous fragment that had been 434 * allocated before the file was increased in size, the old fragment may 435 * be freed once the inode claiming the new block is written to disk. 436 * This ad_fragfree request is attached to the id_inowait list of the 437 * associated inodedep (pointed to by ad_inodedep) for processing after 438 * the inode is written. When a block is allocated to a directory, an 439 * fsync of a file whose name is within that block must ensure not only 440 * that the block containing the file name has been written, but also 441 * that the on-disk inode references that block. When a new directory 442 * block is created, we allocate a newdirblk structure which is linked 443 * to the associated allocdirect (on its ad_newdirblk list). When the 444 * allocdirect has been satisfied, the newdirblk structure is moved to 445 * the inodedep id_bufwait list of its directory to await the inode 446 * being written. When the inode is written, the directory entries are 447 * fully committed and can be deleted from their pagedep->id_pendinghd 448 * and inodedep->id_pendinghd lists. 449 */ 450struct allocdirect { 451 struct newblk ad_block; /* Common block logic */ 452# define ad_list ad_block.nb_list /* block pointer worklist */ 453# define ad_state ad_list.wk_state /* block pointer state */ 454 TAILQ_ENTRY(allocdirect) ad_next; /* inodedep's list of allocdirect's */ 455 struct inodedep *ad_inodedep; /* associated inodedep */ 456 ufs2_daddr_t ad_oldblkno; /* old value of block pointer */ 457 int ad_offset; /* Pointer offset in parent. */ 458 long ad_newsize; /* size of new block */ 459 long ad_oldsize; /* size of old block */ 460}; 461#define ad_newblkno ad_block.nb_newblkno 462#define ad_freefrag ad_block.nb_freefrag 463#define ad_newdirblk ad_block.nb_newdirblk 464 465/* 466 * A single "indirdep" structure manages all allocation dependencies for 467 * pointers in an indirect block. The up-to-date state of the indirect 468 * block is stored in ir_savedata. The set of pointers that may be safely 469 * written to the disk is stored in ir_safecopy. The state field is used 470 * only to track whether the buffer is currently being written (in which 471 * case it is not safe to update ir_safecopy). Ir_deplisthd contains the 472 * list of allocindir structures, one for each block that needs to be 473 * written to disk. Once the block and its bitmap allocation have been 474 * written the safecopy can be updated to reflect the allocation and the 475 * allocindir structure freed. If ir_state indicates that an I/O on the 476 * indirect block is in progress when ir_safecopy is to be updated, the 477 * update is deferred by placing the allocindir on the ir_donehd list. 478 * When the I/O on the indirect block completes, the entries on the 479 * ir_donehd list are processed by updating their corresponding ir_safecopy 480 * pointers and then freeing the allocindir structure. 481 */ 482struct indirdep { 483 struct worklist ir_list; /* buffer holding indirect block */ 484# define ir_state ir_list.wk_state /* indirect block pointer state */ 485 LIST_ENTRY(indirdep) ir_next; /* alloc{direct,indir} list */ 486 TAILQ_HEAD(, freework) ir_trunc; /* List of truncations. */ 487 caddr_t ir_saveddata; /* buffer cache contents */ 488 struct buf *ir_savebp; /* buffer holding safe copy */ 489 struct buf *ir_bp; /* buffer holding live copy */ 490 struct allocindirhd ir_completehd; /* waiting for indirdep complete */ 491 struct allocindirhd ir_writehd; /* Waiting for the pointer write. */ 492 struct allocindirhd ir_donehd; /* done waiting to update safecopy */ 493 struct allocindirhd ir_deplisthd; /* allocindir deps for this block */ 494 struct freeblks *ir_freeblks; /* Freeblks that frees this indir. */ 495}; 496 497/* 498 * An "allocindir" structure is attached to an "indirdep" when a new block 499 * is allocated and pointed to by the indirect block described by the 500 * "indirdep". The worklist is linked to the buffer that holds the new block. 501 * When the block is first allocated, it is linked to the bmsafemap 502 * structure associated with the buffer holding the cylinder group map 503 * from which it was allocated. When the cylinder group map is written 504 * to disk, ai_state has the DEPCOMPLETE flag set. When the block itself 505 * is written, the COMPLETE flag is set. Once both the cylinder group map 506 * and the data itself have been written, it is safe to write the entry in 507 * the indirect block that claims the block; the "allocindir" dependency 508 * can then be freed as it is no longer applicable. 509 */ 510struct allocindir { 511 struct newblk ai_block; /* Common block area */ 512# define ai_state ai_block.nb_list.wk_state /* indirect pointer state */ 513 LIST_ENTRY(allocindir) ai_next; /* indirdep's list of allocindir's */ 514 struct indirdep *ai_indirdep; /* address of associated indirdep */ 515 ufs2_daddr_t ai_oldblkno; /* old value of block pointer */ 516 ufs_lbn_t ai_lbn; /* Logical block number. */ 517 int ai_offset; /* Pointer offset in parent. */ 518}; 519#define ai_newblkno ai_block.nb_newblkno 520#define ai_freefrag ai_block.nb_freefrag 521#define ai_newdirblk ai_block.nb_newdirblk 522 523/* 524 * The allblk union is used to size the newblk structure on allocation so 525 * that it may be any one of three types. 526 */ 527union allblk { 528 struct allocindir ab_allocindir; 529 struct allocdirect ab_allocdirect; 530 struct newblk ab_newblk; 531}; 532 533/* 534 * A "freefrag" structure is attached to an "inodedep" when a previously 535 * allocated fragment is replaced with a larger fragment, rather than extended. 536 * The "freefrag" structure is constructed and attached when the replacement 537 * block is first allocated. It is processed after the inode claiming the 538 * bigger block that replaces it has been written to disk. 539 */ 540struct freefrag { 541 struct worklist ff_list; /* id_inowait or delayed worklist */ 542# define ff_state ff_list.wk_state 543 struct worklist *ff_jdep; /* Associated journal entry. */ 544 struct workhead ff_jwork; /* Journal work pending. */ 545 ufs2_daddr_t ff_blkno; /* fragment physical block number */ 546 long ff_fragsize; /* size of fragment being deleted */ 547 ino_t ff_inum; /* owning inode number */ 548 enum vtype ff_vtype; /* owning inode's file type */ 549}; 550 551/* 552 * A "freeblks" structure is attached to an "inodedep" when the 553 * corresponding file's length is reduced to zero. It records all 554 * the information needed to free the blocks of a file after its 555 * zero'ed inode has been written to disk. The actual work is done 556 * by child freework structures which are responsible for individual 557 * inode pointers while freeblks is responsible for retiring the 558 * entire operation when it is complete and holding common members. 559 */ 560struct freeblks { 561 struct worklist fb_list; /* id_inowait or delayed worklist */ 562# define fb_state fb_list.wk_state /* inode and dirty block state */ 563 TAILQ_ENTRY(freeblks) fb_next; /* List of inode truncates. */ 564 struct jblkdephd fb_jblkdephd; /* Journal entries pending */ 565 struct workhead fb_freeworkhd; /* Work items pending */ 566 struct workhead fb_jwork; /* Journal work pending */ 567 struct vnode *fb_devvp; /* filesystem device vnode */ 568#ifdef QUOTA 569 struct dquot *fb_quota[MAXQUOTAS]; /* quotas to be adjusted */ 570#endif 571 uint64_t fb_modrev; /* Inode revision at start of trunc. */ 572 off_t fb_len; /* Length we're truncating to. */ 573 ufs2_daddr_t fb_chkcnt; /* Blocks released. */ 574 ino_t fb_inum; /* inode owner of blocks */ 575 enum vtype fb_vtype; /* inode owner's file type */ 576 uid_t fb_uid; /* uid of previous owner of blocks */ 577 int fb_ref; /* Children outstanding. */ 578 int fb_cgwait; /* cg writes outstanding. */ 579}; 580 581/* 582 * A "freework" structure handles the release of a tree of blocks or a single 583 * block. Each indirect block in a tree is allocated its own freework 584 * structure so that the indirect block may be freed only when all of its 585 * children are freed. In this way we enforce the rule that an allocated 586 * block must have a valid path to a root that is journaled. Each child 587 * block acquires a reference and when the ref hits zero the parent ref 588 * is decremented. If there is no parent the freeblks ref is decremented. 589 */ 590struct freework { 591 struct worklist fw_list; /* Delayed worklist. */ 592# define fw_state fw_list.wk_state 593 LIST_ENTRY(freework) fw_segs; /* Seg list. */ 594 TAILQ_ENTRY(freework) fw_next; /* Hash/Trunc list. */ 595 struct jnewblk *fw_jnewblk; /* Journal entry to cancel. */ 596 struct freeblks *fw_freeblks; /* Root of operation. */ 597 struct freework *fw_parent; /* Parent indirect. */ 598 struct indirdep *fw_indir; /* indirect block. */ 599 ufs2_daddr_t fw_blkno; /* Our block #. */ 600 ufs_lbn_t fw_lbn; /* Original lbn before free. */ 601 uint16_t fw_frags; /* Number of frags. */ 602 uint16_t fw_ref; /* Number of children out. */ 603 uint16_t fw_off; /* Current working position. */ 604 uint16_t fw_start; /* Start of partial truncate. */ 605}; 606 607/* 608 * A "freedep" structure is allocated to track the completion of a bitmap 609 * write for a freework. One freedep may cover many freed blocks so long 610 * as they reside in the same cylinder group. When the cg is written 611 * the freedep decrements the ref on the freework which may permit it 612 * to be freed as well. 613 */ 614struct freedep { 615 struct worklist fd_list; /* Delayed worklist. */ 616 struct freework *fd_freework; /* Parent freework. */ 617}; 618 619/* 620 * A "freefile" structure is attached to an inode when its 621 * link count is reduced to zero. It marks the inode as free in 622 * the cylinder group map after the zero'ed inode has been written 623 * to disk and any associated blocks and fragments have been freed. 624 */ 625struct freefile { 626 struct worklist fx_list; /* id_inowait or delayed worklist */ 627 mode_t fx_mode; /* mode of inode */ 628 ino_t fx_oldinum; /* inum of the unlinked file */ 629 struct vnode *fx_devvp; /* filesystem device vnode */ 630 struct workhead fx_jwork; /* journal work pending. */ 631}; 632 633/* 634 * A "diradd" structure is linked to an "inodedep" id_inowait list when a 635 * new directory entry is allocated that references the inode described 636 * by "inodedep". When the inode itself is written (either the initial 637 * allocation for new inodes or with the increased link count for 638 * existing inodes), the COMPLETE flag is set in da_state. If the entry 639 * is for a newly allocated inode, the "inodedep" structure is associated 640 * with a bmsafemap which prevents the inode from being written to disk 641 * until the cylinder group has been updated. Thus the da_state COMPLETE 642 * flag cannot be set until the inode bitmap dependency has been removed. 643 * When creating a new file, it is safe to write the directory entry that 644 * claims the inode once the referenced inode has been written. Since 645 * writing the inode clears the bitmap dependencies, the DEPCOMPLETE flag 646 * in the diradd can be set unconditionally when creating a file. When 647 * creating a directory, there are two additional dependencies described by 648 * mkdir structures (see their description below). When these dependencies 649 * are resolved the DEPCOMPLETE flag is set in the diradd structure. 650 * If there are multiple links created to the same inode, there will be 651 * a separate diradd structure created for each link. The diradd is 652 * linked onto the pg_diraddhd list of the pagedep for the directory 653 * page that contains the entry. When a directory page is written, 654 * the pg_diraddhd list is traversed to rollback any entries that are 655 * not yet ready to be written to disk. If a directory entry is being 656 * changed (by rename) rather than added, the DIRCHG flag is set and 657 * the da_previous entry points to the entry that will be "removed" 658 * once the new entry has been committed. During rollback, entries 659 * with da_previous are replaced with the previous inode number rather 660 * than zero. 661 * 662 * The overlaying of da_pagedep and da_previous is done to keep the 663 * structure down. If a da_previous entry is present, the pointer to its 664 * pagedep is available in the associated dirrem entry. If the DIRCHG flag 665 * is set, the da_previous entry is valid; if not set the da_pagedep entry 666 * is valid. The DIRCHG flag never changes; it is set when the structure 667 * is created if appropriate and is never cleared. 668 */ 669struct diradd { 670 struct worklist da_list; /* id_inowait or id_pendinghd list */ 671# define da_state da_list.wk_state /* state of the new directory entry */ 672 LIST_ENTRY(diradd) da_pdlist; /* pagedep holding directory block */ 673 doff_t da_offset; /* offset of new dir entry in dir blk */ 674 ino_t da_newinum; /* inode number for the new dir entry */ 675 union { 676 struct dirrem *dau_previous; /* entry being replaced in dir change */ 677 struct pagedep *dau_pagedep; /* pagedep dependency for addition */ 678 } da_un; 679 struct workhead da_jwork; /* Journal work awaiting completion. */ 680}; 681#define da_previous da_un.dau_previous 682#define da_pagedep da_un.dau_pagedep 683 684/* 685 * Two "mkdir" structures are needed to track the additional dependencies 686 * associated with creating a new directory entry. Normally a directory 687 * addition can be committed as soon as the newly referenced inode has been 688 * written to disk with its increased link count. When a directory is 689 * created there are two additional dependencies: writing the directory 690 * data block containing the "." and ".." entries (MKDIR_BODY) and writing 691 * the parent inode with the increased link count for ".." (MKDIR_PARENT). 692 * These additional dependencies are tracked by two mkdir structures that 693 * reference the associated "diradd" structure. When they have completed, 694 * they set the DEPCOMPLETE flag on the diradd so that it knows that its 695 * extra dependencies have been completed. The md_state field is used only 696 * to identify which type of dependency the mkdir structure is tracking. 697 * It is not used in the mainline code for any purpose other than consistency 698 * checking. All the mkdir structures in the system are linked together on 699 * a list. This list is needed so that a diradd can find its associated 700 * mkdir structures and deallocate them if it is prematurely freed (as for 701 * example if a mkdir is immediately followed by a rmdir of the same directory). 702 * Here, the free of the diradd must traverse the list to find the associated 703 * mkdir structures that reference it. The deletion would be faster if the 704 * diradd structure were simply augmented to have two pointers that referenced 705 * the associated mkdir's. However, this would increase the size of the diradd 706 * structure to speed a very infrequent operation. 707 */ 708struct mkdir { 709 struct worklist md_list; /* id_inowait or buffer holding dir */ 710# define md_state md_list.wk_state /* type: MKDIR_PARENT or MKDIR_BODY */ 711 struct diradd *md_diradd; /* associated diradd */ 712 struct jaddref *md_jaddref; /* dependent jaddref. */ 713 struct buf *md_buf; /* MKDIR_BODY: buffer holding dir */ 714 LIST_ENTRY(mkdir) md_mkdirs; /* list of all mkdirs */ 715}; 716 717/* 718 * A "dirrem" structure describes an operation to decrement the link 719 * count on an inode. The dirrem structure is attached to the pg_dirremhd 720 * list of the pagedep for the directory page that contains the entry. 721 * It is processed after the directory page with the deleted entry has 722 * been written to disk. 723 */ 724struct dirrem { 725 struct worklist dm_list; /* delayed worklist */ 726# define dm_state dm_list.wk_state /* state of the old directory entry */ 727 LIST_ENTRY(dirrem) dm_next; /* pagedep's list of dirrem's */ 728 LIST_ENTRY(dirrem) dm_inonext; /* inodedep's list of dirrem's */ 729 struct jremrefhd dm_jremrefhd; /* Pending remove reference deps. */ 730 ino_t dm_oldinum; /* inum of the removed dir entry */ 731 doff_t dm_offset; /* offset of removed dir entry in blk */ 732 union { 733 struct pagedep *dmu_pagedep; /* pagedep dependency for remove */ 734 ino_t dmu_dirinum; /* parent inode number (for rmdir) */ 735 } dm_un; 736 struct workhead dm_jwork; /* Journal work awaiting completion. */ 737}; 738#define dm_pagedep dm_un.dmu_pagedep 739#define dm_dirinum dm_un.dmu_dirinum 740 741/* 742 * A "newdirblk" structure tracks the progress of a newly allocated 743 * directory block from its creation until it is claimed by its on-disk 744 * inode. When a block is allocated to a directory, an fsync of a file 745 * whose name is within that block must ensure not only that the block 746 * containing the file name has been written, but also that the on-disk 747 * inode references that block. When a new directory block is created, 748 * we allocate a newdirblk structure which is linked to the associated 749 * allocdirect (on its ad_newdirblk list). When the allocdirect has been 750 * satisfied, the newdirblk structure is moved to the inodedep id_bufwait 751 * list of its directory to await the inode being written. When the inode 752 * is written, the directory entries are fully committed and can be 753 * deleted from their pagedep->id_pendinghd and inodedep->id_pendinghd 754 * lists. Note that we could track directory blocks allocated to indirect 755 * blocks using a similar scheme with the allocindir structures. Rather 756 * than adding this level of complexity, we simply write those newly 757 * allocated indirect blocks synchronously as such allocations are rare. 758 * In the case of a new directory the . and .. links are tracked with 759 * a mkdir rather than a pagedep. In this case we track the mkdir 760 * so it can be released when it is written. A workhead is used 761 * to simplify canceling a mkdir that is removed by a subsequent dirrem. 762 */ 763struct newdirblk { 764 struct worklist db_list; /* id_inowait or pg_newdirblk */ 765# define db_state db_list.wk_state 766 struct pagedep *db_pagedep; /* associated pagedep */ 767 struct workhead db_mkdir; 768}; 769 770/* 771 * The inoref structure holds the elements common to jaddref and jremref 772 * so they may easily be queued in-order on the inodedep. 773 */ 774struct inoref { 775 struct worklist if_list; /* Journal pending or jseg entries. */ 776# define if_state if_list.wk_state 777 TAILQ_ENTRY(inoref) if_deps; /* Links for inodedep. */ 778 struct jsegdep *if_jsegdep; /* Will track our journal record. */ 779 off_t if_diroff; /* Directory offset. */ 780 ino_t if_ino; /* Inode number. */ 781 ino_t if_parent; /* Parent inode number. */ 782 nlink_t if_nlink; /* nlink before addition. */ 783 uint16_t if_mode; /* File mode, needed for IFMT. */ 784}; 785 786/* 787 * A "jaddref" structure tracks a new reference (link count) on an inode 788 * and prevents the link count increase and bitmap allocation until a 789 * journal entry can be written. Once the journal entry is written, 790 * the inode is put on the pendinghd of the bmsafemap and a diradd or 791 * mkdir entry is placed on the bufwait list of the inode. The DEPCOMPLETE 792 * flag is used to indicate that all of the required information for writing 793 * the journal entry is present. MKDIR_BODY and MKDIR_PARENT are used to 794 * differentiate . and .. links from regular file names. NEWBLOCK indicates 795 * a bitmap is still pending. If a new reference is canceled by a delete 796 * prior to writing the journal the jaddref write is canceled and the 797 * structure persists to prevent any disk-visible changes until it is 798 * ultimately released when the file is freed or the link is dropped again. 799 */ 800struct jaddref { 801 struct inoref ja_ref; /* see inoref above. */ 802# define ja_list ja_ref.if_list /* Jrnl pending, id_inowait, dm_jwork.*/ 803# define ja_state ja_ref.if_list.wk_state 804 LIST_ENTRY(jaddref) ja_bmdeps; /* Links for bmsafemap. */ 805 union { 806 struct diradd *jau_diradd; /* Pending diradd. */ 807 struct mkdir *jau_mkdir; /* MKDIR_{PARENT,BODY} */ 808 } ja_un; 809}; 810#define ja_diradd ja_un.jau_diradd 811#define ja_mkdir ja_un.jau_mkdir 812#define ja_diroff ja_ref.if_diroff 813#define ja_ino ja_ref.if_ino 814#define ja_parent ja_ref.if_parent 815#define ja_mode ja_ref.if_mode 816 817/* 818 * A "jremref" structure tracks a removed reference (unlink) on an 819 * inode and prevents the directory remove from proceeding until the 820 * journal entry is written. Once the journal has been written the remove 821 * may proceed as normal. 822 */ 823struct jremref { 824 struct inoref jr_ref; /* see inoref above. */ 825# define jr_list jr_ref.if_list /* Linked to softdep_journal_pending. */ 826# define jr_state jr_ref.if_list.wk_state 827 LIST_ENTRY(jremref) jr_deps; /* Links for dirrem. */ 828 struct dirrem *jr_dirrem; /* Back pointer to dirrem. */ 829}; 830 831/* 832 * A "jmvref" structure tracks a name relocations within the same 833 * directory block that occur as a result of directory compaction. 834 * It prevents the updated directory entry from being written to disk 835 * until the journal entry is written. Once the journal has been 836 * written the compacted directory may be written to disk. 837 */ 838struct jmvref { 839 struct worklist jm_list; /* Linked to softdep_journal_pending. */ 840 LIST_ENTRY(jmvref) jm_deps; /* Jmvref on pagedep. */ 841 struct pagedep *jm_pagedep; /* Back pointer to pagedep. */ 842 ino_t jm_parent; /* Containing directory inode number. */ 843 ino_t jm_ino; /* Inode number of our entry. */ 844 off_t jm_oldoff; /* Our old offset in directory. */ 845 off_t jm_newoff; /* Our new offset in directory. */ 846}; 847 848/* 849 * A "jnewblk" structure tracks a newly allocated block or fragment and 850 * prevents the direct or indirect block pointer as well as the cg bitmap 851 * from being written until it is logged. After it is logged the jsegdep 852 * is attached to the allocdirect or allocindir until the operation is 853 * completed or reverted. If the operation is reverted prior to the journal 854 * write the jnewblk structure is maintained to prevent the bitmaps from 855 * reaching the disk. Ultimately the jnewblk structure will be passed 856 * to the free routine as the in memory cg is modified back to the free 857 * state at which time it can be released. It may be held on any of the 858 * fx_jwork, fw_jwork, fb_jwork, ff_jwork, nb_jwork, or ir_jwork lists. 859 */ 860struct jnewblk { 861 struct worklist jn_list; /* See lists above. */ 862# define jn_state jn_list.wk_state 863 struct jsegdep *jn_jsegdep; /* Will track our journal record. */ 864 LIST_ENTRY(jnewblk) jn_deps; /* Jnewblks on sm_jnewblkhd. */ 865 struct worklist *jn_dep; /* Dependency to ref completed seg. */ 866 ufs_lbn_t jn_lbn; /* Lbn to which allocated. */ 867 ufs2_daddr_t jn_blkno; /* Blkno allocated */ 868 ino_t jn_ino; /* Ino to which allocated. */ 869 int jn_oldfrags; /* Previous fragments when extended. */ 870 int jn_frags; /* Number of fragments. */ 871}; 872 873/* 874 * A "jblkdep" structure tracks jfreeblk and jtrunc records attached to a 875 * freeblks structure. 876 */ 877struct jblkdep { 878 struct worklist jb_list; /* For softdep journal pending. */ 879 struct jsegdep *jb_jsegdep; /* Reference to the jseg. */ 880 struct freeblks *jb_freeblks; /* Back pointer to freeblks. */ 881 LIST_ENTRY(jblkdep) jb_deps; /* Dep list on freeblks. */ 882 883}; 884 885/* 886 * A "jfreeblk" structure tracks the journal write for freeing a block 887 * or tree of blocks. The block pointer must not be cleared in the inode 888 * or indirect prior to the jfreeblk being written to the journal. 889 */ 890struct jfreeblk { 891 struct jblkdep jf_dep; /* freeblks linkage. */ 892 ufs_lbn_t jf_lbn; /* Lbn from which blocks freed. */ 893 ufs2_daddr_t jf_blkno; /* Blkno being freed. */ 894 ino_t jf_ino; /* Ino from which blocks freed. */ 895 int jf_frags; /* Number of frags being freed. */ 896}; 897 898/* 899 * A "jfreefrag" tracks the freeing of a single block when a fragment is 900 * extended or an indirect page is replaced. It is not part of a larger 901 * freeblks operation. 902 */ 903struct jfreefrag { 904 struct worklist fr_list; /* Linked to softdep_journal_pending. */ 905# define fr_state fr_list.wk_state 906 struct jsegdep *fr_jsegdep; /* Will track our journal record. */ 907 struct freefrag *fr_freefrag; /* Back pointer to freefrag. */ 908 ufs_lbn_t fr_lbn; /* Lbn from which frag freed. */ 909 ufs2_daddr_t fr_blkno; /* Blkno being freed. */ 910 ino_t fr_ino; /* Ino from which frag freed. */ 911 int fr_frags; /* Size of frag being freed. */ 912}; 913 914/* 915 * A "jtrunc" journals the intent to truncate an inode's data or extent area. 916 */ 917struct jtrunc { 918 struct jblkdep jt_dep; /* freeblks linkage. */ 919 off_t jt_size; /* Final file size. */ 920 int jt_extsize; /* Final extent size. */ 921 ino_t jt_ino; /* Ino being truncated. */ 922}; 923 924/* 925 * A "jfsync" journals the completion of an fsync which invalidates earlier 926 * jtrunc records in the journal. 927 */ 928struct jfsync { 929 struct worklist jfs_list; /* For softdep journal pending. */ 930 off_t jfs_size; /* Sync file size. */ 931 int jfs_extsize; /* Sync extent size. */ 932 ino_t jfs_ino; /* ino being synced. */ 933}; 934 935/* 936 * A "jsegdep" structure tracks a single reference to a written journal 937 * segment so the journal space can be reclaimed when all dependencies 938 * have been written. It can hang off of id_inowait, dm_jwork, da_jwork, 939 * nb_jwork, ff_jwork, or fb_jwork lists. 940 */ 941struct jsegdep { 942 struct worklist jd_list; /* See above for lists. */ 943# define jd_state jd_list.wk_state 944 struct jseg *jd_seg; /* Our journal record. */ 945}; 946 947/* 948 * A "jseg" structure contains all of the journal records written in a 949 * single disk write. The jaddref and jremref structures are linked into 950 * js_entries so thay may be completed when the write completes. The 951 * js_entries also include the write dependency structures: jmvref, 952 * jnewblk, jfreeblk, jfreefrag, and jtrunc. The js_refs field counts 953 * the number of entries on the js_entries list. Thus there is a single 954 * jseg entry to describe each journal write. 955 */ 956struct jseg { 957 struct worklist js_list; /* b_deps link for journal */ 958# define js_state js_list.wk_state 959 struct workhead js_entries; /* Entries awaiting write */ 960 LIST_HEAD(, freework) js_indirs;/* List of indirects in this seg. */ 961 TAILQ_ENTRY(jseg) js_next; /* List of all unfinished segments. */ 962 struct jblocks *js_jblocks; /* Back pointer to block/seg list */ 963 struct buf *js_buf; /* Buffer while unwritten */ 964 uint64_t js_seq; /* Journal record sequence number. */ 965 uint64_t js_oldseq; /* Oldest valid sequence number. */ 966 int js_size; /* Size of journal record in bytes. */ 967 int js_cnt; /* Total items allocated. */ 968 int js_refs; /* Count of js_entries items. */ 969}; 970 971/* 972 * A 'sbdep' structure tracks the head of the free inode list and 973 * superblock writes. This makes sure the superblock is always pointing at 974 * the first possible unlinked inode for the suj recovery process. If a 975 * block write completes and we discover a new head is available the buf 976 * is dirtied and the dep is kept. See the description of the UNLINK* 977 * flags above for more details. 978 */ 979struct sbdep { 980 struct worklist sb_list; /* b_dep linkage */ 981 struct fs *sb_fs; /* Filesystem pointer within buf. */ 982 struct ufsmount *sb_ump; /* Our mount structure */ 983}; 984 985/* 986 * Private journaling structures. 987 */ 988struct jblocks { 989 struct jseglst jb_segs; /* TAILQ of current segments. */ 990 struct jseg *jb_writeseg; /* Next write to complete. */ 991 struct jseg *jb_oldestseg; /* Oldest segment with valid entries. */ 992 struct jextent *jb_extent; /* Extent array. */ 993 uint64_t jb_nextseq; /* Next sequence number. */ 994 uint64_t jb_oldestwrseq; /* Oldest written sequence number. */ 995 uint8_t jb_needseg; /* Need a forced segment. */ 996 uint8_t jb_suspended; /* Did journal suspend writes? */ 997 int jb_avail; /* Available extents. */ 998 int jb_used; /* Last used extent. */ 999 int jb_head; /* Allocator head. */ 1000 int jb_off; /* Allocator extent offset. */ 1001 int jb_blocks; /* Total disk blocks covered. */ 1002 int jb_free; /* Total disk blocks free. */ 1003 int jb_min; /* Minimum free space. */ 1004 int jb_low; /* Low on space. */ 1005 int jb_age; /* Insertion time of oldest rec. */ 1006}; 1007 1008struct jextent { 1009 ufs2_daddr_t je_daddr; /* Disk block address. */ 1010 int je_blocks; /* Disk block count. */ 1011}; 1012 1013/* 1014 * Hash table declarations. 1015 */ 1016LIST_HEAD(mkdirlist, mkdir); 1017LIST_HEAD(pagedep_hashhead, pagedep); 1018LIST_HEAD(inodedep_hashhead, inodedep); 1019LIST_HEAD(newblk_hashhead, newblk); 1020LIST_HEAD(bmsafemap_hashhead, bmsafemap); 1021TAILQ_HEAD(indir_hashhead, freework); 1022 1023/* 1024 * Per-filesystem soft dependency data. 1025 * Allocated at mount and freed at unmount. 1026 */ 1027struct mount_softdeps { 1028 struct rwlock sd_fslock; /* softdep lock */ 1029 struct workhead sd_workitem_pending; /* softdep work queue */ 1030 struct worklist *sd_worklist_tail; /* Tail pointer for above */ 1031 struct workhead sd_journal_pending; /* journal work queue */ 1032 struct worklist *sd_journal_tail; /* Tail pointer for above */ 1033 struct jblocks *sd_jblocks; /* Journal block information */ 1034 struct inodedeplst sd_unlinked; /* Unlinked inodes */ 1035 struct bmsafemaphd sd_dirtycg; /* Dirty CGs */ 1036 struct mkdirlist sd_mkdirlisthd; /* Track mkdirs */ 1037 struct pagedep_hashhead *sd_pdhash; /* pagedep hash table */ 1038 u_long sd_pdhashsize; /* pagedep hash table size-1 */ 1039 long sd_pdnextclean; /* next hash bucket to clean */ 1040 struct inodedep_hashhead *sd_idhash; /* inodedep hash table */ 1041 u_long sd_idhashsize; /* inodedep hash table size-1 */ 1042 long sd_idnextclean; /* next hash bucket to clean */ 1043 struct newblk_hashhead *sd_newblkhash; /* newblk hash table */ 1044 u_long sd_newblkhashsize; /* newblk hash table size-1 */ 1045 struct bmsafemap_hashhead *sd_bmhash; /* bmsafemap hash table */ 1046 u_long sd_bmhashsize; /* bmsafemap hash table size-1*/ 1047 struct indir_hashhead *sd_indirhash; /* indir hash table */ 1048 u_long sd_indirhashsize; /* indir hash table size-1 */ 1049 int sd_on_journal; /* Items on the journal list */ 1050 int sd_on_worklist; /* Items on the worklist */ 1051 int sd_deps; /* Total dependency count */ 1052 int sd_accdeps; /* accumulated dep count */ 1053 int sd_req; /* Wakeup when deps hits 0. */ 1054 int sd_flags; /* comm with flushing thread */ 1055 int sd_cleanups; /* Calls to cleanup */ 1056 struct thread *sd_flushtd; /* thread handling flushing */ 1057 TAILQ_ENTRY(mount_softdeps) sd_next; /* List of softdep filesystem */ 1058 struct ufsmount *sd_ump; /* our ufsmount structure */ 1059 u_long sd_curdeps[D_LAST + 1]; /* count of current deps */ 1060}; 1061/* 1062 * Flags for communicating with the syncer thread. 1063 */ 1064#define FLUSH_EXIT 0x0001 /* time to exit */ 1065#define FLUSH_CLEANUP 0x0002 /* need to clear out softdep structures */ 1066#define FLUSH_STARTING 0x0004 /* flush thread not yet started */ 1067 1068/* 1069 * Keep the old names from when these were in the ufsmount structure. 1070 */ 1071#define softdep_workitem_pending um_softdep->sd_workitem_pending 1072#define softdep_worklist_tail um_softdep->sd_worklist_tail 1073#define softdep_journal_pending um_softdep->sd_journal_pending 1074#define softdep_journal_tail um_softdep->sd_journal_tail 1075#define softdep_jblocks um_softdep->sd_jblocks 1076#define softdep_unlinked um_softdep->sd_unlinked 1077#define softdep_dirtycg um_softdep->sd_dirtycg 1078#define softdep_mkdirlisthd um_softdep->sd_mkdirlisthd 1079#define pagedep_hashtbl um_softdep->sd_pdhash 1080#define pagedep_hash_size um_softdep->sd_pdhashsize 1081#define pagedep_nextclean um_softdep->sd_pdnextclean 1082#define inodedep_hashtbl um_softdep->sd_idhash 1083#define inodedep_hash_size um_softdep->sd_idhashsize 1084#define inodedep_nextclean um_softdep->sd_idnextclean 1085#define newblk_hashtbl um_softdep->sd_newblkhash 1086#define newblk_hash_size um_softdep->sd_newblkhashsize 1087#define bmsafemap_hashtbl um_softdep->sd_bmhash 1088#define bmsafemap_hash_size um_softdep->sd_bmhashsize 1089#define indir_hashtbl um_softdep->sd_indirhash 1090#define indir_hash_size um_softdep->sd_indirhashsize 1091#define softdep_on_journal um_softdep->sd_on_journal 1092#define softdep_on_worklist um_softdep->sd_on_worklist 1093#define softdep_deps um_softdep->sd_deps 1094#define softdep_accdeps um_softdep->sd_accdeps 1095#define softdep_req um_softdep->sd_req 1096#define softdep_flags um_softdep->sd_flags 1097#define softdep_flushtd um_softdep->sd_flushtd 1098#define softdep_curdeps um_softdep->sd_curdeps 1099