1/*-
2 * SPDX-License-Identifier: BSD-2-Clause
3 *
4 * Copyright 1998, 2000 Marshall Kirk McKusick. All Rights Reserved.
5 *
6 * The soft updates code is derived from the appendix of a University
7 * of Michigan technical report (Gregory R. Ganger and Yale N. Patt,
8 * "Soft Updates: A Solution to the Metadata Update Problem in File
9 * Systems", CSE-TR-254-95, August 1995).
10 *
11 * Further information about soft updates can be obtained from:
12 *
13 * Marshall Kirk McKusick http://www.mckusick.com/softdep/
14 * 1614 Oxford Street mckusick@mckusick.com
15 * Berkeley, CA 94709-1608 +1-510-843-9542
16 * USA
17 *
18 * Redistribution and use in source and binary forms, with or without
19 * modification, are permitted provided that the following conditions
20 * are met:
21 *
22 * 1. Redistributions of source code must retain the above copyright
23 * notice, this list of conditions and the following disclaimer.
24 * 2. Redistributions in binary form must reproduce the above copyright
25 * notice, this list of conditions and the following disclaimer in the
26 * documentation and/or other materials provided with the distribution.
27 *
28 * THIS SOFTWARE IS PROVIDED BY MARSHALL KIRK MCKUSICK ``AS IS'' AND ANY
29 * EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
30 * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
31 * DISCLAIMED. IN NO EVENT SHALL MARSHALL KIRK MCKUSICK BE LIABLE FOR
32 * ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
33 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
34 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
35 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
36 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
37 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
38 * SUCH DAMAGE.
39 */
40
41#include <sys/queue.h>
42
43/*
44 * Allocation dependencies are handled with undo/redo on the in-memory
45 * copy of the data. A particular data dependency is eliminated when
46 * it is ALLCOMPLETE: that is ATTACHED, DEPCOMPLETE, and COMPLETE.
47 *
48 * The ATTACHED flag means that the data is not currently being written
49 * to disk.
50 *
51 * The UNDONE flag means that the data has been rolled back to a safe
52 * state for writing to the disk. When the I/O completes, the data is
53 * restored to its current form and the state reverts to ATTACHED.
54 * The data must be locked throughout the rollback, I/O, and roll
55 * forward so that the rolled back information is never visible to
56 * user processes.
57 *
58 * The COMPLETE flag indicates that the item has been written. For example,
59 * a dependency that requires that an inode be written will be marked
60 * COMPLETE after the inode has been written to disk.
61 *
62 * The DEPCOMPLETE flag indicates the completion of any other
63 * dependencies such as the writing of a cylinder group map has been
64 * completed. A dependency structure may be freed only when both it
65 * and its dependencies have completed and any rollbacks that are in
66 * progress have finished as indicated by the set of ALLCOMPLETE flags
67 * all being set.
68 *
69 * The two MKDIR flags indicate additional dependencies that must be done
70 * when creating a new directory. MKDIR_BODY is cleared when the directory
71 * data block containing the "." and ".." entries has been written.
72 * MKDIR_PARENT is cleared when the parent inode with the increased link
73 * count for ".." has been written. When both MKDIR flags have been
74 * cleared, the DEPCOMPLETE flag is set to indicate that the directory
75 * dependencies have been completed. The writing of the directory inode
76 * itself sets the COMPLETE flag which then allows the directory entry for
77 * the new directory to be written to disk. The RMDIR flag marks a dirrem
78 * structure as representing the removal of a directory rather than a
79 * file. When the removal dependencies are completed, additional work needs
80 * to be done* (an additional decrement of the associated inode, and a
81 * decrement of the parent inode).
82 *
83 * The DIRCHG flag marks a diradd structure as representing the changing
84 * of an existing entry rather than the addition of a new one. When
85 * the update is complete the dirrem associated with the inode for
86 * the old name must be added to the worklist to do the necessary
87 * reference count decrement.
88 *
89 * The GOINGAWAY flag indicates that the data structure is frozen from
90 * further change until its dependencies have been completed and its
91 * resources freed after which it will be discarded.
92 *
93 * The IOSTARTED flag prevents multiple calls to the I/O start routine from
94 * doing multiple rollbacks.
95 *
96 * The NEWBLOCK flag marks pagedep structures that have just been allocated,
97 * so must be claimed by the inode before all dependencies are complete.
98 *
99 * The INPROGRESS flag marks worklist structures that are still on the
100 * worklist, but are being considered for action by some process.
101 *
102 * The UFS1FMT flag indicates that the inode being processed is a ufs1 format.
103 *
104 * The EXTDATA flag indicates that the allocdirect describes an
105 * extended-attributes dependency.
106 *
107 * The ONWORKLIST flag shows whether the structure is currently linked
108 * onto a worklist.
109 *
110 * The UNLINK* flags track the progress of updating the on-disk linked
111 * list of active but unlinked inodes. When an inode is first unlinked
112 * it is marked as UNLINKED. When its on-disk di_freelink has been
113 * written its UNLINKNEXT flags is set. When its predecessor in the
114 * list has its di_freelink pointing at us its UNLINKPREV is set.
115 * When the on-disk list can reach it from the superblock, its
116 * UNLINKONLIST flag is set. Once all of these flags are set, it
117 * is safe to let its last name be removed.
118 */
119#define ATTACHED 0x000001
120#define UNDONE 0x000002
121#define COMPLETE 0x000004
122#define DEPCOMPLETE 0x000008
123#define MKDIR_PARENT 0x000010 /* diradd, mkdir, jaddref, jsegdep only */
124#define MKDIR_BODY 0x000020 /* diradd, mkdir, jaddref only */
125#define RMDIR 0x000040 /* dirrem only */
126#define DIRCHG 0x000080 /* diradd, dirrem only */
127#define GOINGAWAY 0x000100 /* indirdep, jremref only */
128#define IOSTARTED 0x000200 /* inodedep, pagedep, bmsafemap only */
129#define DELAYEDFREE 0x000400 /* allocindirect free delayed. */
130#define NEWBLOCK 0x000800 /* pagedep, jaddref only */
131#define INPROGRESS 0x001000 /* dirrem, freeblks, freefrag, freefile only */
132#define UFS1FMT 0x002000 /* indirdep only */
133#define EXTDATA 0x004000 /* allocdirect only */
134#define ONWORKLIST 0x008000
135#define IOWAITING 0x010000 /* Thread is waiting for IO to complete. */
136#define ONDEPLIST 0x020000 /* Structure is on a dependency list. */
137#define UNLINKED 0x040000 /* inodedep has been unlinked. */
138#define UNLINKNEXT 0x080000 /* inodedep has valid di_freelink */
139#define UNLINKPREV 0x100000 /* inodedep is pointed at in the unlink list */
140#define UNLINKONLIST 0x200000 /* inodedep is in the unlinked list on disk */
141#define UNLINKLINKS (UNLINKNEXT | UNLINKPREV)
142#define WRITESUCCEEDED 0x400000 /* the disk write completed successfully */
143
144#define ALLCOMPLETE (ATTACHED | COMPLETE | DEPCOMPLETE)
145
146#define PRINT_SOFTDEP_FLAGS "\20\27writesucceeded\26unlinkonlist" \
147 "\25unlinkprev\24unlinknext\23unlinked\22ondeplist\21iowaiting" \
148 "\20onworklist\17extdata\16ufs1fmt\15inprogress\14newblock" \
149 "\13delayedfree\12iostarted\11goingaway\10dirchg\7rmdir\6mkdir_body" \
150 "\5mkdir_parent\4depcomplete\3complete\2undone\1attached"
151
152/*
153 * Values for each of the soft dependency types.
154 */
155#define D_UNUSED 0
156#define D_FIRST D_PAGEDEP
157#define D_PAGEDEP 1
158#define D_INODEDEP 2
159#define D_BMSAFEMAP 3
160#define D_NEWBLK 4
161#define D_ALLOCDIRECT 5
162#define D_INDIRDEP 6
163#define D_ALLOCINDIR 7
164#define D_FREEFRAG 8
165#define D_FREEBLKS 9
166#define D_FREEFILE 10
167#define D_DIRADD 11
168#define D_MKDIR 12
169#define D_DIRREM 13
170#define D_NEWDIRBLK 14
171#define D_FREEWORK 15
172#define D_FREEDEP 16
173#define D_JADDREF 17
174#define D_JREMREF 18
175#define D_JMVREF 19
176#define D_JNEWBLK 20
177#define D_JFREEBLK 21
178#define D_JFREEFRAG 22
179#define D_JSEG 23
180#define D_JSEGDEP 24
181#define D_SBDEP 25
182#define D_JTRUNC 26
183#define D_JFSYNC 27
184#define D_SENTINEL 28
185#define D_LAST D_SENTINEL
186
187/*
188 * The workitem queue.
189 *
190 * It is sometimes useful and/or necessary to clean up certain dependencies
191 * in the background rather than during execution of an application process
192 * or interrupt service routine. To realize this, we append dependency
193 * structures corresponding to such tasks to a "workitem" queue. In a soft
194 * updates implementation, most pending workitems should not wait for more
195 * than a couple of seconds, so the filesystem syncer process awakens once
196 * per second to process the items on the queue.
197 */
198
199/* LIST_HEAD(workhead, worklist); -- declared in buf.h */
200
201/*
202 * Each request can be linked onto a work queue through its worklist structure.
203 * To avoid the need for a pointer to the structure itself, this structure
204 * MUST be declared FIRST in each type in which it appears! If more than one
205 * worklist is needed in the structure, then a wk_data field must be added
206 * and the macros below changed to use it.
207 */
208struct worklist {
209 LIST_ENTRY(worklist) wk_list; /* list of work requests */
210 struct mount *wk_mp; /* Mount we live in */
211 unsigned int wk_type:8, /* type of request */
212 wk_state:24; /* state flags */
213 LIST_ENTRY(worklist) wk_all; /* list of deps of this type */
214#ifdef INVARIANTS
215 const char *wk_func; /* func where added / removed */
216 int wk_line; /* line where added / removed */
217#endif
218};
219#define WK_DATA(wk) ((void *)(wk))
220#define WK_PAGEDEP(wk) ((struct pagedep *)(wk))
221#define WK_INODEDEP(wk) ((struct inodedep *)(wk))
222#define WK_BMSAFEMAP(wk) ((struct bmsafemap *)(wk))
223#define WK_NEWBLK(wk) ((struct newblk *)(wk))
224#define WK_ALLOCDIRECT(wk) ((struct allocdirect *)(wk))
225#define WK_INDIRDEP(wk) ((struct indirdep *)(wk))
226#define WK_ALLOCINDIR(wk) ((struct allocindir *)(wk))
227#define WK_FREEFRAG(wk) ((struct freefrag *)(wk))
228#define WK_FREEBLKS(wk) ((struct freeblks *)(wk))
229#define WK_FREEWORK(wk) ((struct freework *)(wk))
230#define WK_FREEFILE(wk) ((struct freefile *)(wk))
231#define WK_DIRADD(wk) ((struct diradd *)(wk))
232#define WK_MKDIR(wk) ((struct mkdir *)(wk))
233#define WK_DIRREM(wk) ((struct dirrem *)(wk))
234#define WK_NEWDIRBLK(wk) ((struct newdirblk *)(wk))
235#define WK_JADDREF(wk) ((struct jaddref *)(wk))
236#define WK_JREMREF(wk) ((struct jremref *)(wk))
237#define WK_JMVREF(wk) ((struct jmvref *)(wk))
238#define WK_JSEGDEP(wk) ((struct jsegdep *)(wk))
239#define WK_JSEG(wk) ((struct jseg *)(wk))
240#define WK_JNEWBLK(wk) ((struct jnewblk *)(wk))
241#define WK_JFREEBLK(wk) ((struct jfreeblk *)(wk))
242#define WK_FREEDEP(wk) ((struct freedep *)(wk))
243#define WK_JFREEFRAG(wk) ((struct jfreefrag *)(wk))
244#define WK_SBDEP(wk) ((struct sbdep *)(wk))
245#define WK_JTRUNC(wk) ((struct jtrunc *)(wk))
246#define WK_JFSYNC(wk) ((struct jfsync *)(wk))
247
248/*
249 * Various types of lists
250 */
251LIST_HEAD(dirremhd, dirrem);
252LIST_HEAD(diraddhd, diradd);
253LIST_HEAD(newblkhd, newblk);
254LIST_HEAD(inodedephd, inodedep);
255LIST_HEAD(allocindirhd, allocindir);
256LIST_HEAD(allocdirecthd, allocdirect);
257TAILQ_HEAD(allocdirectlst, allocdirect);
258LIST_HEAD(indirdephd, indirdep);
259LIST_HEAD(jaddrefhd, jaddref);
260LIST_HEAD(jremrefhd, jremref);
261LIST_HEAD(jmvrefhd, jmvref);
262LIST_HEAD(jnewblkhd, jnewblk);
263LIST_HEAD(jblkdephd, jblkdep);
264LIST_HEAD(freeworkhd, freework);
265TAILQ_HEAD(freeworklst, freework);
266TAILQ_HEAD(jseglst, jseg);
267TAILQ_HEAD(inoreflst, inoref);
268TAILQ_HEAD(freeblklst, freeblks);
269
270/*
271 * The "pagedep" structure tracks the various dependencies related to
272 * a particular directory page. If a directory page has any dependencies,
273 * it will have a pagedep linked to its associated buffer. The
274 * pd_dirremhd list holds the list of dirrem requests which decrement
275 * inode reference counts. These requests are processed after the
276 * directory page with the corresponding zero'ed entries has been
277 * written. The pd_diraddhd list maintains the list of diradd requests
278 * which cannot be committed until their corresponding inode has been
279 * written to disk. Because a directory may have many new entries
280 * being created, several lists are maintained hashed on bits of the
281 * offset of the entry into the directory page to keep the lists from
282 * getting too long. Once a new directory entry has been cleared to
283 * be written, it is moved to the pd_pendinghd list. After the new
284 * entry has been written to disk it is removed from the pd_pendinghd
285 * list, any removed operations are done, and the dependency structure
286 * is freed.
287 */
288#define DAHASHSZ 5
289#define DIRADDHASH(offset) (((offset) >> 2) % DAHASHSZ)
290struct pagedep {
291 struct worklist pd_list; /* page buffer */
292# define pd_state pd_list.wk_state /* check for multiple I/O starts */
293 LIST_ENTRY(pagedep) pd_hash; /* hashed lookup */
294 ino_t pd_ino; /* associated file */
295 ufs_lbn_t pd_lbn; /* block within file */
296 struct newdirblk *pd_newdirblk; /* associated newdirblk if NEWBLOCK */
297 struct dirremhd pd_dirremhd; /* dirrem's waiting for page */
298 struct diraddhd pd_diraddhd[DAHASHSZ]; /* diradd dir entry updates */
299 struct diraddhd pd_pendinghd; /* directory entries awaiting write */
300 struct jmvrefhd pd_jmvrefhd; /* Dependent journal writes. */
301};
302
303/*
304 * The "inodedep" structure tracks the set of dependencies associated
305 * with an inode. One task that it must manage is delayed operations
306 * (i.e., work requests that must be held until the inodedep's associated
307 * inode has been written to disk). Getting an inode from its incore
308 * state to the disk requires two steps to be taken by the filesystem
309 * in this order: first the inode must be copied to its disk buffer by
310 * the VOP_UPDATE operation; second the inode's buffer must be written
311 * to disk. To ensure that both operations have happened in the required
312 * order, the inodedep maintains two lists. Delayed operations are
313 * placed on the id_inowait list. When the VOP_UPDATE is done, all
314 * operations on the id_inowait list are moved to the id_bufwait list.
315 * When the buffer is written, the items on the id_bufwait list can be
316 * safely moved to the work queue to be processed. A second task of the
317 * inodedep structure is to track the status of block allocation within
318 * the inode. Each block that is allocated is represented by an
319 * "allocdirect" structure (see below). It is linked onto the id_newinoupdt
320 * list until both its contents and its allocation in the cylinder
321 * group map have been written to disk. Once these dependencies have been
322 * satisfied, it is removed from the id_newinoupdt list and any followup
323 * actions such as releasing the previous block or fragment are placed
324 * on the id_inowait list. When an inode is updated (a VOP_UPDATE is
325 * done), the "inodedep" structure is linked onto the buffer through
326 * its worklist. Thus, it will be notified when the buffer is about
327 * to be written and when it is done. At the update time, all the
328 * elements on the id_newinoupdt list are moved to the id_inoupdt list
329 * since those changes are now relevant to the copy of the inode in the
330 * buffer. Also at update time, the tasks on the id_inowait list are
331 * moved to the id_bufwait list so that they will be executed when
332 * the updated inode has been written to disk. When the buffer containing
333 * the inode is written to disk, any updates listed on the id_inoupdt
334 * list are rolled back as they are not yet safe. Following the write,
335 * the changes are once again rolled forward and any actions on the
336 * id_bufwait list are processed (since those actions are now safe).
337 * The entries on the id_inoupdt and id_newinoupdt lists must be kept
338 * sorted by logical block number to speed the calculation of the size
339 * of the rolled back inode (see explanation in initiate_write_inodeblock).
340 * When a directory entry is created, it is represented by a diradd.
341 * The diradd is added to the id_inowait list as it cannot be safely
342 * written to disk until the inode that it represents is on disk. After
343 * the inode is written, the id_bufwait list is processed and the diradd
344 * entries are moved to the id_pendinghd list where they remain until
345 * the directory block containing the name has been written to disk.
346 * The purpose of keeping the entries on the id_pendinghd list is so that
347 * the softdep_fsync function can find and push the inode's directory
348 * name(s) as part of the fsync operation for that file.
349 */
350struct inodedep {
351 struct worklist id_list; /* buffer holding inode block */
352# define id_state id_list.wk_state /* inode dependency state */
353 LIST_ENTRY(inodedep) id_hash; /* hashed lookup */
354 TAILQ_ENTRY(inodedep) id_unlinked; /* Unlinked but ref'd inodes */
355 struct fs *id_fs; /* associated filesystem */
356 ino_t id_ino; /* dependent inode */
357 nlink_t id_nlinkdelta; /* saved effective link count */
358 nlink_t id_nlinkwrote; /* i_nlink that we wrote to disk */
359 nlink_t id_savednlink; /* Link saved during rollback */
360 LIST_ENTRY(inodedep) id_deps; /* bmsafemap's list of inodedep's */
361 struct bmsafemap *id_bmsafemap; /* related bmsafemap (if pending) */
362 struct diradd *id_mkdiradd; /* diradd for a mkdir. */
363 struct inoreflst id_inoreflst; /* Inode reference adjustments. */
364 long id_savedextsize; /* ext size saved during rollback */
365 off_t id_savedsize; /* file size saved during rollback */
366 struct dirremhd id_dirremhd; /* Removals pending. */
367 struct workhead id_pendinghd; /* entries awaiting directory write */
368 struct workhead id_bufwait; /* operations after inode written */
369 struct workhead id_inowait; /* operations waiting inode update */
370 struct allocdirectlst id_inoupdt; /* updates before inode written */
371 struct allocdirectlst id_newinoupdt; /* updates when inode written */
372 struct allocdirectlst id_extupdt; /* extdata updates pre-inode write */
373 struct allocdirectlst id_newextupdt; /* extdata updates at ino write */
374 struct freeblklst id_freeblklst; /* List of partial truncates. */
375 union {
376 struct ufs1_dinode *idu_savedino1; /* saved ufs1_dinode contents */
377 struct ufs2_dinode *idu_savedino2; /* saved ufs2_dinode contents */
378 } id_un;
379};
380#define id_savedino1 id_un.idu_savedino1
381#define id_savedino2 id_un.idu_savedino2
382
383/*
384 * A "bmsafemap" structure maintains a list of dependency structures
385 * that depend on the update of a particular cylinder group map.
386 * It has lists for newblks, allocdirects, allocindirs, and inodedeps.
387 * It is attached to the buffer of a cylinder group block when any of
388 * these things are allocated from the cylinder group. It is freed
389 * after the cylinder group map is written and the state of its
390 * dependencies are updated with DEPCOMPLETE to indicate that it has
391 * been processed.
392 */
393struct bmsafemap {
394 struct worklist sm_list; /* cylgrp buffer */
395# define sm_state sm_list.wk_state
396 LIST_ENTRY(bmsafemap) sm_hash; /* Hash links. */
397 LIST_ENTRY(bmsafemap) sm_next; /* Mount list. */
398 int sm_cg;
399 struct buf *sm_buf; /* associated buffer */
400 struct allocdirecthd sm_allocdirecthd; /* allocdirect deps */
401 struct allocdirecthd sm_allocdirectwr; /* writing allocdirect deps */
402 struct allocindirhd sm_allocindirhd; /* allocindir deps */
403 struct allocindirhd sm_allocindirwr; /* writing allocindir deps */
404 struct inodedephd sm_inodedephd; /* inodedep deps */
405 struct inodedephd sm_inodedepwr; /* writing inodedep deps */
406 struct newblkhd sm_newblkhd; /* newblk deps */
407 struct newblkhd sm_newblkwr; /* writing newblk deps */
408 struct jaddrefhd sm_jaddrefhd; /* Pending inode allocations. */
409 struct jnewblkhd sm_jnewblkhd; /* Pending block allocations. */
410 struct workhead sm_freehd; /* Freedep deps. */
411 struct workhead sm_freewr; /* Written freedeps. */
412};
413
414/*
415 * A "newblk" structure is attached to a bmsafemap structure when a block
416 * or fragment is allocated from a cylinder group. Its state is set to
417 * DEPCOMPLETE when its cylinder group map is written. It is converted to
418 * an allocdirect or allocindir allocation once the allocator calls the
419 * appropriate setup function. It will initially be linked onto a bmsafemap
420 * list. Once converted it can be linked onto the lists described for
421 * allocdirect or allocindir as described below.
422 */
423struct newblk {
424 struct worklist nb_list; /* See comment above. */
425# define nb_state nb_list.wk_state
426 LIST_ENTRY(newblk) nb_hash; /* Hashed lookup. */
427 LIST_ENTRY(newblk) nb_deps; /* Bmsafemap's list of newblks. */
428 struct jnewblk *nb_jnewblk; /* New block journal entry. */
429 struct bmsafemap *nb_bmsafemap;/* Cylgrp dep (if pending). */
430 struct freefrag *nb_freefrag; /* Fragment to be freed (if any). */
431 struct indirdephd nb_indirdeps; /* Children indirect blocks. */
432 struct workhead nb_newdirblk; /* Dir block to notify when written. */
433 struct workhead nb_jwork; /* Journal work pending. */
434 ufs2_daddr_t nb_newblkno; /* New value of block pointer. */
435};
436
437/*
438 * An "allocdirect" structure is attached to an "inodedep" when a new block
439 * or fragment is allocated and pointed to by the inode described by
440 * "inodedep". The worklist is linked to the buffer that holds the block.
441 * When the block is first allocated, it is linked to the bmsafemap
442 * structure associated with the buffer holding the cylinder group map
443 * from which it was allocated. When the cylinder group map is written
444 * to disk, ad_state has the DEPCOMPLETE flag set. When the block itself
445 * is written, the COMPLETE flag is set. Once both the cylinder group map
446 * and the data itself have been written, it is safe to write the inode
447 * that claims the block. If there was a previous fragment that had been
448 * allocated before the file was increased in size, the old fragment may
449 * be freed once the inode claiming the new block is written to disk.
450 * This ad_fragfree request is attached to the id_inowait list of the
451 * associated inodedep (pointed to by ad_inodedep) for processing after
452 * the inode is written. When a block is allocated to a directory, an
453 * fsync of a file whose name is within that block must ensure not only
454 * that the block containing the file name has been written, but also
455 * that the on-disk inode references that block. When a new directory
456 * block is created, we allocate a newdirblk structure which is linked
457 * to the associated allocdirect (on its ad_newdirblk list). When the
458 * allocdirect has been satisfied, the newdirblk structure is moved to
459 * the inodedep id_bufwait list of its directory to await the inode
460 * being written. When the inode is written, the directory entries are
461 * fully committed and can be deleted from their pagedep->id_pendinghd
462 * and inodedep->id_pendinghd lists.
463 */
464struct allocdirect {
465 struct newblk ad_block; /* Common block logic */
466# define ad_list ad_block.nb_list /* block pointer worklist */
467# define ad_state ad_list.wk_state /* block pointer state */
468 TAILQ_ENTRY(allocdirect) ad_next; /* inodedep's list of allocdirect's */
469 struct inodedep *ad_inodedep; /* associated inodedep */
470 ufs2_daddr_t ad_oldblkno; /* old value of block pointer */
471 int ad_offset; /* Pointer offset in parent. */
472 long ad_newsize; /* size of new block */
473 long ad_oldsize; /* size of old block */
474};
475#define ad_newblkno ad_block.nb_newblkno
476#define ad_freefrag ad_block.nb_freefrag
477#define ad_newdirblk ad_block.nb_newdirblk
478
479/*
480 * A single "indirdep" structure manages all allocation dependencies for
481 * pointers in an indirect block. The up-to-date state of the indirect
482 * block is stored in ir_savedata. The set of pointers that may be safely
483 * written to the disk is stored in ir_savebp. The state field is used
484 * only to track whether the buffer is currently being written (in which
485 * case it is not safe to update ir_savebp). Ir_deplisthd contains the
486 * list of allocindir structures, one for each block that needs to be
487 * written to disk. Once the block and its bitmap allocation have been
488 * written the safecopy can be updated to reflect the allocation and the
489 * allocindir structure freed. If ir_state indicates that an I/O on the
490 * indirect block is in progress when ir_savebp is to be updated, the
491 * update is deferred by placing the allocindir on the ir_donehd list.
492 * When the I/O on the indirect block completes, the entries on the
493 * ir_donehd list are processed by updating their corresponding ir_savebp
494 * pointers and then freeing the allocindir structure.
495 */
496struct indirdep {
497 struct worklist ir_list; /* buffer holding indirect block */
498# define ir_state ir_list.wk_state /* indirect block pointer state */
499 LIST_ENTRY(indirdep) ir_next; /* alloc{direct,indir} list */
500 TAILQ_HEAD(, freework) ir_trunc; /* List of truncations. */
501 caddr_t ir_saveddata; /* buffer cache contents */
502 struct buf *ir_savebp; /* buffer holding safe copy */
503 struct buf *ir_bp; /* buffer holding live copy */
504 struct allocindirhd ir_completehd; /* waiting for indirdep complete */
505 struct allocindirhd ir_writehd; /* Waiting for the pointer write. */
506 struct allocindirhd ir_donehd; /* done waiting to update safecopy */
507 struct allocindirhd ir_deplisthd; /* allocindir deps for this block */
508 struct freeblks *ir_freeblks; /* Freeblks that frees this indir. */
509};
510
511/*
512 * An "allocindir" structure is attached to an "indirdep" when a new block
513 * is allocated and pointed to by the indirect block described by the
514 * "indirdep". The worklist is linked to the buffer that holds the new block.
515 * When the block is first allocated, it is linked to the bmsafemap
516 * structure associated with the buffer holding the cylinder group map
517 * from which it was allocated. When the cylinder group map is written
518 * to disk, ai_state has the DEPCOMPLETE flag set. When the block itself
519 * is written, the COMPLETE flag is set. Once both the cylinder group map
520 * and the data itself have been written, it is safe to write the entry in
521 * the indirect block that claims the block; the "allocindir" dependency
522 * can then be freed as it is no longer applicable.
523 */
524struct allocindir {
525 struct newblk ai_block; /* Common block area */
526# define ai_state ai_block.nb_list.wk_state /* indirect pointer state */
527 LIST_ENTRY(allocindir) ai_next; /* indirdep's list of allocindir's */
528 struct indirdep *ai_indirdep; /* address of associated indirdep */
529 ufs2_daddr_t ai_oldblkno; /* old value of block pointer */
530 ufs_lbn_t ai_lbn; /* Logical block number. */
531 int ai_offset; /* Pointer offset in parent. */
532};
533#define ai_newblkno ai_block.nb_newblkno
534#define ai_freefrag ai_block.nb_freefrag
535#define ai_newdirblk ai_block.nb_newdirblk
536
537/*
538 * The allblk union is used to size the newblk structure on allocation so
539 * that it may be any one of three types.
540 */
541union allblk {
542 struct allocindir ab_allocindir;
543 struct allocdirect ab_allocdirect;
544 struct newblk ab_newblk;
545};
546
547/*
548 * A "freefrag" structure is attached to an "inodedep" when a previously
549 * allocated fragment is replaced with a larger fragment, rather than extended.
550 * The "freefrag" structure is constructed and attached when the replacement
551 * block is first allocated. It is processed after the inode claiming the
552 * bigger block that replaces it has been written to disk.
553 */
554struct freefrag {
555 struct worklist ff_list; /* id_inowait or delayed worklist */
556# define ff_state ff_list.wk_state
557 struct worklist *ff_jdep; /* Associated journal entry. */
558 struct workhead ff_jwork; /* Journal work pending. */
559 ufs2_daddr_t ff_blkno; /* fragment physical block number */
560 long ff_fragsize; /* size of fragment being deleted */
561 ino_t ff_inum; /* owning inode number */
562 __enum_uint8(vtype) ff_vtype; /* owning inode's file type */
563 int ff_key; /* trim key when deleted */
564};
565
566/*
567 * A "freeblks" structure is attached to an "inodedep" when the
568 * corresponding file's length is reduced to zero. It records all
569 * the information needed to free the blocks of a file after its
570 * zero'ed inode has been written to disk. The actual work is done
571 * by child freework structures which are responsible for individual
572 * inode pointers while freeblks is responsible for retiring the
573 * entire operation when it is complete and holding common members.
574 */
575struct freeblks {
576 struct worklist fb_list; /* id_inowait or delayed worklist */
577# define fb_state fb_list.wk_state /* inode and dirty block state */
578 TAILQ_ENTRY(freeblks) fb_next; /* List of inode truncates. */
579 struct jblkdephd fb_jblkdephd; /* Journal entries pending */
580 struct workhead fb_freeworkhd; /* Work items pending */
581 struct workhead fb_jwork; /* Journal work pending */
582 struct vnode *fb_devvp; /* filesystem device vnode */
583#ifdef QUOTA
584 struct dquot *fb_quota[MAXQUOTAS]; /* quotas to be adjusted */
585#endif
586 uint64_t fb_modrev; /* Inode revision at start of trunc. */
587 off_t fb_len; /* Length we're truncating to. */
588 ufs2_daddr_t fb_chkcnt; /* Blocks released. */
589 ino_t fb_inum; /* inode owner of blocks */
590 __enum_uint8(vtype) fb_vtype; /* inode owner's file type */
591 uid_t fb_uid; /* uid of previous owner of blocks */
592 int fb_ref; /* Children outstanding. */
593 int fb_cgwait; /* cg writes outstanding. */
594};
595
596/*
597 * A "freework" structure handles the release of a tree of blocks or a single
598 * block. Each indirect block in a tree is allocated its own freework
599 * structure so that the indirect block may be freed only when all of its
600 * children are freed. In this way we enforce the rule that an allocated
601 * block must have a valid path to a root that is journaled. Each child
602 * block acquires a reference and when the ref hits zero the parent ref
603 * is decremented. If there is no parent the freeblks ref is decremented.
604 */
605struct freework {
606 struct worklist fw_list; /* Delayed worklist. */
607# define fw_state fw_list.wk_state
608 LIST_ENTRY(freework) fw_segs; /* Seg list. */
609 TAILQ_ENTRY(freework) fw_next; /* Hash/Trunc list. */
610 struct jnewblk *fw_jnewblk; /* Journal entry to cancel. */
611 struct freeblks *fw_freeblks; /* Root of operation. */
612 struct freework *fw_parent; /* Parent indirect. */
613 struct indirdep *fw_indir; /* indirect block. */
614 ufs2_daddr_t fw_blkno; /* Our block #. */
615 ufs_lbn_t fw_lbn; /* Original lbn before free. */
616 uint16_t fw_frags; /* Number of frags. */
617 uint16_t fw_ref; /* Number of children out. */
618 uint16_t fw_off; /* Current working position. */
619 uint16_t fw_start; /* Start of partial truncate. */
620};
621
622/*
623 * A "freedep" structure is allocated to track the completion of a bitmap
624 * write for a freework. One freedep may cover many freed blocks so long
625 * as they reside in the same cylinder group. When the cg is written
626 * the freedep decrements the ref on the freework which may permit it
627 * to be freed as well.
628 */
629struct freedep {
630 struct worklist fd_list; /* Delayed worklist. */
631 struct freework *fd_freework; /* Parent freework. */
632};
633
634/*
635 * A "freefile" structure is attached to an inode when its
636 * link count is reduced to zero. It marks the inode as free in
637 * the cylinder group map after the zero'ed inode has been written
638 * to disk and any associated blocks and fragments have been freed.
639 */
640struct freefile {
641 struct worklist fx_list; /* id_inowait or delayed worklist */
642 mode_t fx_mode; /* mode of inode */
643 ino_t fx_oldinum; /* inum of the unlinked file */
644 struct vnode *fx_devvp; /* filesystem device vnode */
645 struct workhead fx_jwork; /* journal work pending. */
646};
647
648/*
649 * A "diradd" structure is linked to an "inodedep" id_inowait list when a
650 * new directory entry is allocated that references the inode described
651 * by "inodedep". When the inode itself is written (either the initial
652 * allocation for new inodes or with the increased link count for
653 * existing inodes), the COMPLETE flag is set in da_state. If the entry
654 * is for a newly allocated inode, the "inodedep" structure is associated
655 * with a bmsafemap which prevents the inode from being written to disk
656 * until the cylinder group has been updated. Thus the da_state COMPLETE
657 * flag cannot be set until the inode bitmap dependency has been removed.
658 * When creating a new file, it is safe to write the directory entry that
659 * claims the inode once the referenced inode has been written. Since
660 * writing the inode clears the bitmap dependencies, the DEPCOMPLETE flag
661 * in the diradd can be set unconditionally when creating a file. When
662 * creating a directory, there are two additional dependencies described by
663 * mkdir structures (see their description below). When these dependencies
664 * are resolved the DEPCOMPLETE flag is set in the diradd structure.
665 * If there are multiple links created to the same inode, there will be
666 * a separate diradd structure created for each link. The diradd is
667 * linked onto the pg_diraddhd list of the pagedep for the directory
668 * page that contains the entry. When a directory page is written,
669 * the pg_diraddhd list is traversed to rollback any entries that are
670 * not yet ready to be written to disk. If a directory entry is being
671 * changed (by rename) rather than added, the DIRCHG flag is set and
672 * the da_previous entry points to the entry that will be "removed"
673 * once the new entry has been committed. During rollback, entries
674 * with da_previous are replaced with the previous inode number rather
675 * than zero.
676 *
677 * The overlaying of da_pagedep and da_previous is done to keep the
678 * structure down. If a da_previous entry is present, the pointer to its
679 * pagedep is available in the associated dirrem entry. If the DIRCHG flag
680 * is set, the da_previous entry is valid; if not set the da_pagedep entry
681 * is valid. The DIRCHG flag never changes; it is set when the structure
682 * is created if appropriate and is never cleared.
683 */
684struct diradd {
685 struct worklist da_list; /* id_inowait or id_pendinghd list */
686# define da_state da_list.wk_state /* state of the new directory entry */
687 LIST_ENTRY(diradd) da_pdlist; /* pagedep holding directory block */
688 doff_t da_offset; /* offset of new dir entry in dir blk */
689 ino_t da_newinum; /* inode number for the new dir entry */
690 union {
691 struct dirrem *dau_previous; /* entry being replaced in dir change */
692 struct pagedep *dau_pagedep; /* pagedep dependency for addition */
693 } da_un;
694 struct workhead da_jwork; /* Journal work awaiting completion. */
695};
696#define da_previous da_un.dau_previous
697#define da_pagedep da_un.dau_pagedep
698
699/*
700 * Two "mkdir" structures are needed to track the additional dependencies
701 * associated with creating a new directory entry. Normally a directory
702 * addition can be committed as soon as the newly referenced inode has been
703 * written to disk with its increased link count. When a directory is
704 * created there are two additional dependencies: writing the directory
705 * data block containing the "." and ".." entries (MKDIR_BODY) and writing
706 * the parent inode with the increased link count for ".." (MKDIR_PARENT).
707 * These additional dependencies are tracked by two mkdir structures that
708 * reference the associated "diradd" structure. When they have completed,
709 * they set the DEPCOMPLETE flag on the diradd so that it knows that its
710 * extra dependencies have been completed. The md_state field is used only
711 * to identify which type of dependency the mkdir structure is tracking.
712 * It is not used in the mainline code for any purpose other than consistency
713 * checking. All the mkdir structures in the system are linked together on
714 * a list. This list is needed so that a diradd can find its associated
715 * mkdir structures and deallocate them if it is prematurely freed (as for
716 * example if a mkdir is immediately followed by a rmdir of the same directory).
717 * Here, the free of the diradd must traverse the list to find the associated
718 * mkdir structures that reference it. The deletion would be faster if the
719 * diradd structure were simply augmented to have two pointers that referenced
720 * the associated mkdir's. However, this would increase the size of the diradd
721 * structure to speed a very infrequent operation.
722 */
723struct mkdir {
724 struct worklist md_list; /* id_inowait or buffer holding dir */
725# define md_state md_list.wk_state /* type: MKDIR_PARENT or MKDIR_BODY */
726 struct diradd *md_diradd; /* associated diradd */
727 struct jaddref *md_jaddref; /* dependent jaddref. */
728 struct buf *md_buf; /* MKDIR_BODY: buffer holding dir */
729 LIST_ENTRY(mkdir) md_mkdirs; /* list of all mkdirs */
730};
731
732/*
733 * A "dirrem" structure describes an operation to decrement the link
734 * count on an inode. The dirrem structure is attached to the pg_dirremhd
735 * list of the pagedep for the directory page that contains the entry.
736 * It is processed after the directory page with the deleted entry has
737 * been written to disk.
738 */
739struct dirrem {
740 struct worklist dm_list; /* delayed worklist */
741# define dm_state dm_list.wk_state /* state of the old directory entry */
742 LIST_ENTRY(dirrem) dm_next; /* pagedep's list of dirrem's */
743 LIST_ENTRY(dirrem) dm_inonext; /* inodedep's list of dirrem's */
744 struct jremrefhd dm_jremrefhd; /* Pending remove reference deps. */
745 ino_t dm_oldinum; /* inum of the removed dir entry */
746 doff_t dm_offset; /* offset of removed dir entry in blk */
747 union {
748 struct pagedep *dmu_pagedep; /* pagedep dependency for remove */
749 ino_t dmu_dirinum; /* parent inode number (for rmdir) */
750 } dm_un;
751 struct workhead dm_jwork; /* Journal work awaiting completion. */
752};
753#define dm_pagedep dm_un.dmu_pagedep
754#define dm_dirinum dm_un.dmu_dirinum
755
756/*
757 * A "newdirblk" structure tracks the progress of a newly allocated
758 * directory block from its creation until it is claimed by its on-disk
759 * inode. When a block is allocated to a directory, an fsync of a file
760 * whose name is within that block must ensure not only that the block
761 * containing the file name has been written, but also that the on-disk
762 * inode references that block. When a new directory block is created,
763 * we allocate a newdirblk structure which is linked to the associated
764 * allocdirect (on its ad_newdirblk list). When the allocdirect has been
765 * satisfied, the newdirblk structure is moved to the inodedep id_bufwait
766 * list of its directory to await the inode being written. When the inode
767 * is written, the directory entries are fully committed and can be
768 * deleted from their pagedep->id_pendinghd and inodedep->id_pendinghd
769 * lists. Note that we could track directory blocks allocated to indirect
770 * blocks using a similar scheme with the allocindir structures. Rather
771 * than adding this level of complexity, we simply write those newly
772 * allocated indirect blocks synchronously as such allocations are rare.
773 * In the case of a new directory the . and .. links are tracked with
774 * a mkdir rather than a pagedep. In this case we track the mkdir
775 * so it can be released when it is written. A workhead is used
776 * to simplify canceling a mkdir that is removed by a subsequent dirrem.
777 */
778struct newdirblk {
779 struct worklist db_list; /* id_inowait or pg_newdirblk */
780# define db_state db_list.wk_state
781 struct pagedep *db_pagedep; /* associated pagedep */
782 struct workhead db_mkdir;
783};
784
785/*
786 * The inoref structure holds the elements common to jaddref and jremref
787 * so they may easily be queued in-order on the inodedep.
788 */
789struct inoref {
790 struct worklist if_list; /* Journal pending or jseg entries. */
791# define if_state if_list.wk_state
792 TAILQ_ENTRY(inoref) if_deps; /* Links for inodedep. */
793 struct jsegdep *if_jsegdep; /* Will track our journal record. */
794 off_t if_diroff; /* Directory offset. */
795 ino_t if_ino; /* Inode number. */
796 ino_t if_parent; /* Parent inode number. */
797 nlink_t if_nlink; /* nlink before addition. */
798 uint16_t if_mode; /* File mode, needed for IFMT. */
799};
800
801/*
802 * A "jaddref" structure tracks a new reference (link count) on an inode
803 * and prevents the link count increase and bitmap allocation until a
804 * journal entry can be written. Once the journal entry is written,
805 * the inode is put on the pendinghd of the bmsafemap and a diradd or
806 * mkdir entry is placed on the bufwait list of the inode. The DEPCOMPLETE
807 * flag is used to indicate that all of the required information for writing
808 * the journal entry is present. MKDIR_BODY and MKDIR_PARENT are used to
809 * differentiate . and .. links from regular file names. NEWBLOCK indicates
810 * a bitmap is still pending. If a new reference is canceled by a delete
811 * prior to writing the journal the jaddref write is canceled and the
812 * structure persists to prevent any disk-visible changes until it is
813 * ultimately released when the file is freed or the link is dropped again.
814 */
815struct jaddref {
816 struct inoref ja_ref; /* see inoref above. */
817# define ja_list ja_ref.if_list /* Jrnl pending, id_inowait, dm_jwork.*/
818# define ja_state ja_ref.if_list.wk_state
819 LIST_ENTRY(jaddref) ja_bmdeps; /* Links for bmsafemap. */
820 union {
821 struct diradd *jau_diradd; /* Pending diradd. */
822 struct mkdir *jau_mkdir; /* MKDIR_{PARENT,BODY} */
823 } ja_un;
824};
825#define ja_diradd ja_un.jau_diradd
826#define ja_mkdir ja_un.jau_mkdir
827#define ja_diroff ja_ref.if_diroff
828#define ja_ino ja_ref.if_ino
829#define ja_parent ja_ref.if_parent
830#define ja_mode ja_ref.if_mode
831
832/*
833 * A "jremref" structure tracks a removed reference (unlink) on an
834 * inode and prevents the directory remove from proceeding until the
835 * journal entry is written. Once the journal has been written the remove
836 * may proceed as normal.
837 */
838struct jremref {
839 struct inoref jr_ref; /* see inoref above. */
840# define jr_list jr_ref.if_list /* Linked to softdep_journal_pending. */
841# define jr_state jr_ref.if_list.wk_state
842 LIST_ENTRY(jremref) jr_deps; /* Links for dirrem. */
843 struct dirrem *jr_dirrem; /* Back pointer to dirrem. */
844};
845
846/*
847 * A "jmvref" structure tracks a name relocations within the same
848 * directory block that occur as a result of directory compaction.
849 * It prevents the updated directory entry from being written to disk
850 * until the journal entry is written. Once the journal has been
851 * written the compacted directory may be written to disk.
852 */
853struct jmvref {
854 struct worklist jm_list; /* Linked to softdep_journal_pending. */
855 LIST_ENTRY(jmvref) jm_deps; /* Jmvref on pagedep. */
856 struct pagedep *jm_pagedep; /* Back pointer to pagedep. */
857 ino_t jm_parent; /* Containing directory inode number. */
858 ino_t jm_ino; /* Inode number of our entry. */
859 off_t jm_oldoff; /* Our old offset in directory. */
860 off_t jm_newoff; /* Our new offset in directory. */
861};
862
863/*
864 * A "jnewblk" structure tracks a newly allocated block or fragment and
865 * prevents the direct or indirect block pointer as well as the cg bitmap
866 * from being written until it is logged. After it is logged the jsegdep
867 * is attached to the allocdirect or allocindir until the operation is
868 * completed or reverted. If the operation is reverted prior to the journal
869 * write the jnewblk structure is maintained to prevent the bitmaps from
870 * reaching the disk. Ultimately the jnewblk structure will be passed
871 * to the free routine as the in memory cg is modified back to the free
872 * state at which time it can be released. It may be held on any of the
873 * fx_jwork, fw_jwork, fb_jwork, ff_jwork, nb_jwork, or ir_jwork lists.
874 */
875struct jnewblk {
876 struct worklist jn_list; /* See lists above. */
877# define jn_state jn_list.wk_state
878 struct jsegdep *jn_jsegdep; /* Will track our journal record. */
879 LIST_ENTRY(jnewblk) jn_deps; /* Jnewblks on sm_jnewblkhd. */
880 struct worklist *jn_dep; /* Dependency to ref completed seg. */
881 ufs_lbn_t jn_lbn; /* Lbn to which allocated. */
882 ufs2_daddr_t jn_blkno; /* Blkno allocated */
883 ino_t jn_ino; /* Ino to which allocated. */
884 int jn_oldfrags; /* Previous fragments when extended. */
885 int jn_frags; /* Number of fragments. */
886};
887
888/*
889 * A "jblkdep" structure tracks jfreeblk and jtrunc records attached to a
890 * freeblks structure.
891 */
892struct jblkdep {
893 struct worklist jb_list; /* For softdep journal pending. */
894 struct jsegdep *jb_jsegdep; /* Reference to the jseg. */
895 struct freeblks *jb_freeblks; /* Back pointer to freeblks. */
896 LIST_ENTRY(jblkdep) jb_deps; /* Dep list on freeblks. */
897
898};
899
900/*
901 * A "jfreeblk" structure tracks the journal write for freeing a block
902 * or tree of blocks. The block pointer must not be cleared in the inode
903 * or indirect prior to the jfreeblk being written to the journal.
904 */
905struct jfreeblk {
906 struct jblkdep jf_dep; /* freeblks linkage. */
907 ufs_lbn_t jf_lbn; /* Lbn from which blocks freed. */
908 ufs2_daddr_t jf_blkno; /* Blkno being freed. */
909 ino_t jf_ino; /* Ino from which blocks freed. */
910 int jf_frags; /* Number of frags being freed. */
911};
912
913/*
914 * A "jfreefrag" tracks the freeing of a single block when a fragment is
915 * extended or an indirect page is replaced. It is not part of a larger
916 * freeblks operation.
917 */
918struct jfreefrag {
919 struct worklist fr_list; /* Linked to softdep_journal_pending. */
920# define fr_state fr_list.wk_state
921 struct jsegdep *fr_jsegdep; /* Will track our journal record. */
922 struct freefrag *fr_freefrag; /* Back pointer to freefrag. */
923 ufs_lbn_t fr_lbn; /* Lbn from which frag freed. */
924 ufs2_daddr_t fr_blkno; /* Blkno being freed. */
925 ino_t fr_ino; /* Ino from which frag freed. */
926 int fr_frags; /* Size of frag being freed. */
927};
928
929/*
930 * A "jtrunc" journals the intent to truncate an inode's data or extent area.
931 */
932struct jtrunc {
933 struct jblkdep jt_dep; /* freeblks linkage. */
934 off_t jt_size; /* Final file size. */
935 int jt_extsize; /* Final extent size. */
936 ino_t jt_ino; /* Ino being truncated. */
937};
938
939/*
940 * A "jfsync" journals the completion of an fsync which invalidates earlier
941 * jtrunc records in the journal.
942 */
943struct jfsync {
944 struct worklist jfs_list; /* For softdep journal pending. */
945 off_t jfs_size; /* Sync file size. */
946 int jfs_extsize; /* Sync extent size. */
947 ino_t jfs_ino; /* ino being synced. */
948};
949
950/*
951 * A "jsegdep" structure tracks a single reference to a written journal
952 * segment so the journal space can be reclaimed when all dependencies
953 * have been written. It can hang off of id_inowait, dm_jwork, da_jwork,
954 * nb_jwork, ff_jwork, or fb_jwork lists.
955 */
956struct jsegdep {
957 struct worklist jd_list; /* See above for lists. */
958# define jd_state jd_list.wk_state
959 struct jseg *jd_seg; /* Our journal record. */
960};
961
962/*
963 * A "jseg" structure contains all of the journal records written in a
964 * single disk write. The jaddref and jremref structures are linked into
965 * js_entries so thay may be completed when the write completes. The
966 * js_entries also include the write dependency structures: jmvref,
967 * jnewblk, jfreeblk, jfreefrag, and jtrunc. The js_refs field counts
968 * the number of entries on the js_entries list. Thus there is a single
969 * jseg entry to describe each journal write.
970 */
971struct jseg {
972 struct worklist js_list; /* b_deps link for journal */
973# define js_state js_list.wk_state
974 struct workhead js_entries; /* Entries awaiting write */
975 LIST_HEAD(, freework) js_indirs;/* List of indirects in this seg. */
976 TAILQ_ENTRY(jseg) js_next; /* List of all unfinished segments. */
977 struct jblocks *js_jblocks; /* Back pointer to block/seg list */
978 struct buf *js_buf; /* Buffer while unwritten */
979 uint64_t js_seq; /* Journal record sequence number. */
980 uint64_t js_oldseq; /* Oldest valid sequence number. */
981 int js_size; /* Size of journal record in bytes. */
982 int js_cnt; /* Total items allocated. */
983 int js_refs; /* Count of js_entries items. */
984};
985
986/*
987 * A 'sbdep' structure tracks the head of the free inode list and
988 * superblock writes. This makes sure the superblock is always pointing at
989 * the first possible unlinked inode for the suj recovery process. If a
990 * block write completes and we discover a new head is available the buf
991 * is dirtied and the dep is kept. See the description of the UNLINK*
992 * flags above for more details.
993 */
994struct sbdep {
995 struct worklist sb_list; /* b_dep linkage */
996 struct fs *sb_fs; /* Filesystem pointer within buf. */
997 struct ufsmount *sb_ump; /* Our mount structure */
998};
999
1000/*
1001 * Private journaling structures.
1002 */
1003struct jblocks {
1004 struct jseglst jb_segs; /* TAILQ of current segments. */
1005 struct jseg *jb_writeseg; /* Next write to complete. */
1006 struct jseg *jb_oldestseg; /* Oldest segment with valid entries. */
1007 struct jextent *jb_extent; /* Extent array. */
1008 uint64_t jb_nextseq; /* Next sequence number. */
1009 uint64_t jb_oldestwrseq; /* Oldest written sequence number. */
1010 uint8_t jb_needseg; /* Need a forced segment. */
1011 uint8_t jb_suspended; /* Did journal suspend writes? */
1012 int jb_avail; /* Available extents. */
1013 int jb_used; /* Last used extent. */
1014 int jb_head; /* Allocator head. */
1015 int jb_off; /* Allocator extent offset. */
1016 int jb_blocks; /* Total disk blocks covered. */
1017 int jb_free; /* Total disk blocks free. */
1018 int jb_min; /* Minimum free space. */
1019 int jb_low; /* Low on space. */
1020 int jb_age; /* Insertion time of oldest rec. */
1021};
1022
1023struct jextent {
1024 ufs2_daddr_t je_daddr; /* Disk block address. */
1025 int je_blocks; /* Disk block count. */
1026};
1027
1028/*
1029 * Hash table declarations.
1030 */
1031LIST_HEAD(mkdirlist, mkdir);
1032LIST_HEAD(pagedep_hashhead, pagedep);
1033LIST_HEAD(inodedep_hashhead, inodedep);
1034LIST_HEAD(newblk_hashhead, newblk);
1035LIST_HEAD(bmsafemap_hashhead, bmsafemap);
1036TAILQ_HEAD(indir_hashhead, freework);
1037
1038/*
1039 * Per-filesystem soft dependency data.
1040 * Allocated at mount and freed at unmount.
1041 */
1042struct mount_softdeps {
1043 struct rwlock sd_fslock; /* softdep lock */
1044 struct workhead sd_workitem_pending; /* softdep work queue */
1045 struct worklist *sd_worklist_tail; /* Tail pointer for above */
1046 struct workhead sd_journal_pending; /* journal work queue */
1047 struct worklist *sd_journal_tail; /* Tail pointer for above */
1048 struct jblocks *sd_jblocks; /* Journal block information */
1049 struct inodedeplst sd_unlinked; /* Unlinked inodes */
1050 struct bmsafemaphd sd_dirtycg; /* Dirty CGs */
1051 struct mkdirlist sd_mkdirlisthd; /* Track mkdirs */
1052 struct pagedep_hashhead *sd_pdhash; /* pagedep hash table */
1053 u_long sd_pdhashsize; /* pagedep hash table size-1 */
1054 long sd_pdnextclean; /* next hash bucket to clean */
1055 struct inodedep_hashhead *sd_idhash; /* inodedep hash table */
1056 u_long sd_idhashsize; /* inodedep hash table size-1 */
1057 long sd_idnextclean; /* next hash bucket to clean */
1058 struct newblk_hashhead *sd_newblkhash; /* newblk hash table */
1059 u_long sd_newblkhashsize; /* newblk hash table size-1 */
1060 struct bmsafemap_hashhead *sd_bmhash; /* bmsafemap hash table */
1061 u_long sd_bmhashsize; /* bmsafemap hash table size-1*/
1062 struct indir_hashhead *sd_indirhash; /* indir hash table */
1063 uint64_t sd_indirhashsize; /* indir hash table size-1 */
1064 int sd_on_journal; /* Items on the journal list */
1065 int sd_on_worklist; /* Items on the worklist */
1066 int sd_deps; /* Total dependency count */
1067 int sd_accdeps; /* accumulated dep count */
1068 int sd_req; /* Wakeup when deps hits 0. */
1069 int sd_flags; /* comm with flushing thread */
1070 int sd_cleanups; /* Calls to cleanup */
1071 struct thread *sd_flushtd; /* thread handling flushing */
1072 TAILQ_ENTRY(mount_softdeps) sd_next; /* List of softdep filesystem */
1073 struct ufsmount *sd_ump; /* our ufsmount structure */
1074 uint64_t sd_curdeps[D_LAST + 1]; /* count of current deps */
1075 struct workhead sd_alldeps[D_LAST + 1];/* Lists of all deps */
1076};
1077/*
1078 * Flags for communicating with the syncer thread.
1079 */
1080#define FLUSH_EXIT 0x0001 /* time to exit */
1081#define FLUSH_CLEANUP 0x0002 /* need to clear out softdep structures */
1082#define FLUSH_STARTING 0x0004 /* flush thread not yet started */
1083#define FLUSH_RC_ACTIVE 0x0008 /* a thread is flushing the mount point */
1084#define FLUSH_DI_ACTIVE 0x0010 /* a thread is processing delayed
1085 inactivations */
1086
1087/*
1088 * Keep the old names from when these were in the ufsmount structure.
1089 */
1090#define softdep_workitem_pending um_softdep->sd_workitem_pending
1091#define softdep_worklist_tail um_softdep->sd_worklist_tail
1092#define softdep_journal_pending um_softdep->sd_journal_pending
1093#define softdep_journal_tail um_softdep->sd_journal_tail
1094#define softdep_jblocks um_softdep->sd_jblocks
1095#define softdep_unlinked um_softdep->sd_unlinked
1096#define softdep_dirtycg um_softdep->sd_dirtycg
1097#define softdep_mkdirlisthd um_softdep->sd_mkdirlisthd
1098#define pagedep_hashtbl um_softdep->sd_pdhash
1099#define pagedep_hash_size um_softdep->sd_pdhashsize
1100#define pagedep_nextclean um_softdep->sd_pdnextclean
1101#define inodedep_hashtbl um_softdep->sd_idhash
1102#define inodedep_hash_size um_softdep->sd_idhashsize
1103#define inodedep_nextclean um_softdep->sd_idnextclean
1104#define newblk_hashtbl um_softdep->sd_newblkhash
1105#define newblk_hash_size um_softdep->sd_newblkhashsize
1106#define bmsafemap_hashtbl um_softdep->sd_bmhash
1107#define bmsafemap_hash_size um_softdep->sd_bmhashsize
1108#define indir_hashtbl um_softdep->sd_indirhash
1109#define indir_hash_size um_softdep->sd_indirhashsize
1110#define softdep_on_journal um_softdep->sd_on_journal
1111#define softdep_on_worklist um_softdep->sd_on_worklist
1112#define softdep_deps um_softdep->sd_deps
1113#define softdep_accdeps um_softdep->sd_accdeps
1114#define softdep_req um_softdep->sd_req
1115#define softdep_flags um_softdep->sd_flags
1116#define softdep_flushtd um_softdep->sd_flushtd
1117#define softdep_curdeps um_softdep->sd_curdeps
1118#define softdep_alldeps um_softdep->sd_alldeps