| 1 |
/*------------------------------------------------------------------------- |
| 2 |
* |
| 3 |
* pruneheap.c |
| 4 |
* heap page pruning and HOT-chain management code |
| 5 |
* |
| 6 |
* Portions Copyright (c) 1996-2025, PostgreSQL Global Development Group |
| 7 |
* Portions Copyright (c) 1994, Regents of the University of California |
| 8 |
* |
| 9 |
* |
| 10 |
* IDENTIFICATION |
| 11 |
* src/backend/access/heap/pruneheap.c |
| 12 |
* |
| 13 |
*------------------------------------------------------------------------- |
| 14 |
*/ |
| 15 |
#include "postgres.h" |
| 16 |
|
| 17 |
#include "access/heapam.h" |
| 18 |
#include "access/heapam_xlog.h" |
| 19 |
#include "access/htup_details.h" |
| 20 |
#include "access/multixact.h" |
| 21 |
#include "access/transam.h" |
| 22 |
#include "access/xlog.h" |
| 23 |
#include "access/xloginsert.h" |
| 24 |
#include "commands/vacuum.h" |
| 25 |
#include "executor/instrument.h" |
| 26 |
#include "miscadmin.h" |
| 27 |
#include "pgstat.h" |
| 28 |
#include "storage/bufmgr.h" |
| 29 |
#include "utils/rel.h" |
| 30 |
#include "utils/snapmgr.h" |
| 31 |
|
| 32 |
/* Working data for heap_page_prune_and_freeze() and subroutines */ |
| 33 |
typedef struct |
| 34 |
{ |
| 35 |
/*------------------------------------------------------- |
| 36 |
* Arguments passed to heap_page_prune_and_freeze() |
| 37 |
*------------------------------------------------------- |
| 38 |
*/ |
| 39 |
|
| 40 |
/* tuple visibility test, initialized for the relation */ |
| 41 |
GlobalVisState *vistest; |
| 42 |
/* whether or not dead items can be set LP_UNUSED during pruning */ |
| 43 |
bool mark_unused_now; |
| 44 |
/* whether to attempt freezing tuples */ |
| 45 |
bool freeze; |
| 46 |
struct VacuumCutoffs *cutoffs; |
| 47 |
|
| 48 |
/*------------------------------------------------------- |
| 49 |
* Fields describing what to do to the page |
| 50 |
*------------------------------------------------------- |
| 51 |
*/ |
| 52 |
TransactionId new_prune_xid; /* new prune hint value */ |
| 53 |
TransactionId latest_xid_removed; |
| 54 |
int nredirected; /* numbers of entries in arrays below */ |
| 55 |
int ndead; |
| 56 |
int nunused; |
| 57 |
int nfrozen; |
| 58 |
/* arrays that accumulate indexes of items to be changed */ |
| 59 |
OffsetNumber redirected[MaxHeapTuplesPerPage * 2]; |
| 60 |
OffsetNumber nowdead[MaxHeapTuplesPerPage]; |
| 61 |
OffsetNumber nowunused[MaxHeapTuplesPerPage]; |
| 62 |
HeapTupleFreeze frozen[MaxHeapTuplesPerPage]; |
| 63 |
|
| 64 |
/*------------------------------------------------------- |
| 65 |
* Working state for HOT chain processing |
| 66 |
*------------------------------------------------------- |
| 67 |
*/ |
| 68 |
|
| 69 |
/* |
| 70 |
* 'root_items' contains offsets of all LP_REDIRECT line pointers and |
| 71 |
* normal non-HOT tuples. They can be stand-alone items or the first item |
| 72 |
* in a HOT chain. 'heaponly_items' contains heap-only tuples which can |
| 73 |
* only be removed as part of a HOT chain. |
| 74 |
*/ |
| 75 |
int nroot_items; |
| 76 |
OffsetNumber root_items[MaxHeapTuplesPerPage]; |
| 77 |
int nheaponly_items; |
| 78 |
OffsetNumber heaponly_items[MaxHeapTuplesPerPage]; |
| 79 |
|
| 80 |
/* |
| 81 |
* processed[offnum] is true if item at offnum has been processed. |
| 82 |
* |
| 83 |
* This needs to be MaxHeapTuplesPerPage + 1 long as FirstOffsetNumber is |
| 84 |
* 1. Otherwise every access would need to subtract 1. |
| 85 |
*/ |
| 86 |
bool processed[MaxHeapTuplesPerPage + 1]; |
| 87 |
|
| 88 |
/* |
| 89 |
* Tuple visibility is only computed once for each tuple, for correctness |
| 90 |
* and efficiency reasons; see comment in heap_page_prune_and_freeze() for |
| 91 |
* details. This is of type int8[], instead of HTSV_Result[], so we can |
| 92 |
* use -1 to indicate no visibility has been computed, e.g. for LP_DEAD |
| 93 |
* items. |
| 94 |
* |
| 95 |
* This needs to be MaxHeapTuplesPerPage + 1 long as FirstOffsetNumber is |
| 96 |
* 1. Otherwise every access would need to subtract 1. |
| 97 |
*/ |
| 98 |
int8 htsv[MaxHeapTuplesPerPage + 1]; |
| 99 |
|
| 100 |
/* |
| 101 |
* Freezing-related state. |
| 102 |
*/ |
| 103 |
HeapPageFreeze pagefrz; |
| 104 |
|
| 105 |
/*------------------------------------------------------- |
| 106 |
* Information about what was done |
| 107 |
* |
| 108 |
* These fields are not used by pruning itself for the most part, but are |
| 109 |
* used to collect information about what was pruned and what state the |
| 110 |
* page is in after pruning, for the benefit of the caller. They are |
| 111 |
* copied to the caller's PruneFreezeResult at the end. |
| 112 |
* ------------------------------------------------------- |
| 113 |
*/ |
| 114 |
|
| 115 |
int ndeleted; /* Number of tuples deleted from the page */ |
| 116 |
|
| 117 |
/* Number of live and recently dead tuples, after pruning */ |
| 118 |
int live_tuples; |
| 119 |
int recently_dead_tuples; |
| 120 |
|
| 121 |
/* Whether or not the page makes rel truncation unsafe */ |
| 122 |
bool hastup; |
| 123 |
|
| 124 |
/* |
| 125 |
* LP_DEAD items on the page after pruning. Includes existing LP_DEAD |
| 126 |
* items |
| 127 |
*/ |
| 128 |
int lpdead_items; /* number of items in the array */ |
| 129 |
OffsetNumber *deadoffsets; /* points directly to presult->deadoffsets */ |
| 130 |
|
| 131 |
/* |
| 132 |
* all_visible and all_frozen indicate if the all-visible and all-frozen |
| 133 |
* bits in the visibility map can be set for this page after pruning. |
| 134 |
* |
| 135 |
* visibility_cutoff_xid is the newest xmin of live tuples on the page. |
| 136 |
* The caller can use it as the conflict horizon, when setting the VM |
| 137 |
* bits. It is only valid if we froze some tuples, and all_frozen is |
| 138 |
* true. |
| 139 |
* |
| 140 |
* NOTE: all_visible and all_frozen don't include LP_DEAD items. That's |
| 141 |
* convenient for heap_page_prune_and_freeze(), to use them to decide |
| 142 |
* whether to freeze the page or not. The all_visible and all_frozen |
| 143 |
* values returned to the caller are adjusted to include LP_DEAD items at |
| 144 |
* the end. |
| 145 |
* |
| 146 |
* all_frozen should only be considered valid if all_visible is also set; |
| 147 |
* we don't bother to clear the all_frozen flag every time we clear the |
| 148 |
* all_visible flag. |
| 149 |
*/ |
| 150 |
bool all_visible; |
| 151 |
bool all_frozen; |
| 152 |
TransactionId visibility_cutoff_xid; |
| 153 |
} PruneState; |
| 154 |
|
| 155 |
/* Local functions */ |
| 156 |
static HTSV_Result heap_prune_satisfies_vacuum(PruneState *prstate, |
| 157 |
HeapTuple tup, |
| 158 |
Buffer buffer); |
| 159 |
static inline HTSV_Result htsv_get_valid_status(int status); |
| 160 |
static void heap_prune_chain(Page page, BlockNumber blockno, OffsetNumber maxoff, |
| 161 |
OffsetNumber rootoffnum, PruneState *prstate); |
| 162 |
static void heap_prune_record_prunable(PruneState *prstate, TransactionId xid); |
| 163 |
static void heap_prune_record_redirect(PruneState *prstate, |
| 164 |
OffsetNumber offnum, OffsetNumber rdoffnum, |
| 165 |
bool was_normal); |
| 166 |
static void heap_prune_record_dead(PruneState *prstate, OffsetNumber offnum, |
| 167 |
bool was_normal); |
| 168 |
static void heap_prune_record_dead_or_unused(PruneState *prstate, OffsetNumber offnum, |
| 169 |
bool was_normal); |
| 170 |
static void heap_prune_record_unused(PruneState *prstate, OffsetNumber offnum, bool was_normal); |
| 171 |
|
| 172 |
static void heap_prune_record_unchanged_lp_unused(Page page, PruneState *prstate, OffsetNumber offnum); |
| 173 |
static void heap_prune_record_unchanged_lp_normal(Page page, PruneState *prstate, OffsetNumber offnum); |
| 174 |
static void heap_prune_record_unchanged_lp_dead(Page page, PruneState *prstate, OffsetNumber offnum); |
| 175 |
static void heap_prune_record_unchanged_lp_redirect(PruneState *prstate, OffsetNumber offnum); |
| 176 |
|
| 177 |
static void page_verify_redirects(Page page); |
| 178 |
|
| 179 |
|
| 180 |
/* |
| 181 |
* Optionally prune and repair fragmentation in the specified page. |
| 182 |
* |
| 183 |
* This is an opportunistic function. It will perform housekeeping |
| 184 |
* only if the page heuristically looks like a candidate for pruning and we |
| 185 |
* can acquire buffer cleanup lock without blocking. |
| 186 |
* |
| 187 |
* Note: this is called quite often. It's important that it fall out quickly |
| 188 |
* if there's not any use in pruning. |
| 189 |
* |
| 190 |
* Caller must have pin on the buffer, and must *not* have a lock on it. |
| 191 |
*/ |
| 192 |
void |
| 193 |
heap_page_prune_opt(Relation relation, Buffer buffer) |
| 194 |
{ |
| 195 |
Page page = BufferGetPage(buffer); |
| 196 |
TransactionId prune_xid; |
| 197 |
GlobalVisState *vistest; |
| 198 |
Size minfree; |
| 199 |
|
| 200 |
/* |
| 201 |
* We can't write WAL in recovery mode, so there's no point trying to |
| 202 |
* clean the page. The primary will likely issue a cleaning WAL record |
| 203 |
* soon anyway, so this is no particular loss. |
| 204 |
*/ |
| 205 |
if (RecoveryInProgress()) |
| 206 |
return; |
| 207 |
|
| 208 |
/* |
| 209 |
* First check whether there's any chance there's something to prune, |
| 210 |
* determining the appropriate horizon is a waste if there's no prune_xid |
| 211 |
* (i.e. no updates/deletes left potentially dead tuples around). |
| 212 |
*/ |
| 213 |
prune_xid = ((PageHeader) page)->pd_prune_xid; |
| 214 |
if (!TransactionIdIsValid(prune_xid)) |
| 215 |
return; |
| 216 |
|
| 217 |
/* |
| 218 |
* Check whether prune_xid indicates that there may be dead rows that can |
| 219 |
* be cleaned up. |
| 220 |
*/ |
| 221 |
vistest = GlobalVisTestFor(relation); |
| 222 |
|
| 223 |
if (!GlobalVisTestIsRemovableXid(vistest, prune_xid)) |
| 224 |
return; |
| 225 |
|
| 226 |
/* |
| 227 |
* We prune when a previous UPDATE failed to find enough space on the page |
| 228 |
* for a new tuple version, or when free space falls below the relation's |
| 229 |
* fill-factor target (but not less than 10%). |
| 230 |
* |
| 231 |
* Checking free space here is questionable since we aren't holding any |
| 232 |
* lock on the buffer; in the worst case we could get a bogus answer. It's |
| 233 |
* unlikely to be *seriously* wrong, though, since reading either pd_lower |
| 234 |
* or pd_upper is probably atomic. Avoiding taking a lock seems more |
| 235 |
* important than sometimes getting a wrong answer in what is after all |
| 236 |
* just a heuristic estimate. |
| 237 |
*/ |
| 238 |
minfree = RelationGetTargetPageFreeSpace(relation, |
| 239 |
HEAP_DEFAULT_FILLFACTOR); |
| 240 |
minfree = Max(minfree, BLCKSZ / 10); |
| 241 |
|
| 242 |
if (PageIsFull(page) || PageGetHeapFreeSpace(page) < minfree) |
| 243 |
{ |
| 244 |
/* OK, try to get exclusive buffer lock */ |
| 245 |
if (!ConditionalLockBufferForCleanup(buffer)) |
| 246 |
return; |
| 247 |
|
| 248 |
/* |
| 249 |
* Now that we have buffer lock, get accurate information about the |
| 250 |
* page's free space, and recheck the heuristic about whether to |
| 251 |
* prune. |
| 252 |
*/ |
| 253 |
if (PageIsFull(page) || PageGetHeapFreeSpace(page) < minfree) |
| 254 |
{ |
| 255 |
OffsetNumber dummy_off_loc; |
| 256 |
PruneFreezeResult presult; |
| 257 |
|
| 258 |
/* |
| 259 |
* For now, pass mark_unused_now as false regardless of whether or |
| 260 |
* not the relation has indexes, since we cannot safely determine |
| 261 |
* that during on-access pruning with the current implementation. |
| 262 |
*/ |
| 263 |
heap_page_prune_and_freeze(relation, buffer, vistest, 0, |
| 264 |
NULL, &presult, PRUNE_ON_ACCESS, &dummy_off_loc, NULL, NULL); |
| 265 |
|
| 266 |
/* |
| 267 |
* Report the number of tuples reclaimed to pgstats. This is |
| 268 |
* presult.ndeleted minus the number of newly-LP_DEAD-set items. |
| 269 |
* |
| 270 |
* We derive the number of dead tuples like this to avoid totally |
| 271 |
* forgetting about items that were set to LP_DEAD, since they |
| 272 |
* still need to be cleaned up by VACUUM. We only want to count |
| 273 |
* heap-only tuples that just became LP_UNUSED in our report, |
| 274 |
* which don't. |
| 275 |
* |
| 276 |
* VACUUM doesn't have to compensate in the same way when it |
| 277 |
* tracks ndeleted, since it will set the same LP_DEAD items to |
| 278 |
* LP_UNUSED separately. |
| 279 |
*/ |
| 280 |
if (presult.ndeleted > presult.nnewlpdead) |
| 281 |
pgstat_update_heap_dead_tuples(relation, |
| 282 |
presult.ndeleted - presult.nnewlpdead); |
| 283 |
} |
| 284 |
|
| 285 |
/* And release buffer lock */ |
| 286 |
LockBuffer(buffer, BUFFER_LOCK_UNLOCK); |
| 287 |
|
| 288 |
/* |
| 289 |
* We avoid reuse of any free space created on the page by unrelated |
| 290 |
* UPDATEs/INSERTs by opting to not update the FSM at this point. The |
| 291 |
* free space should be reused by UPDATEs to *this* page. |
| 292 |
*/ |
| 293 |
} |
| 294 |
} |
| 295 |
|
| 296 |
|
| 297 |
/* |
| 298 |
* Prune and repair fragmentation and potentially freeze tuples on the |
| 299 |
* specified page. |
| 300 |
* |
| 301 |
* Caller must have pin and buffer cleanup lock on the page. Note that we |
| 302 |
* don't update the FSM information for page on caller's behalf. Caller might |
| 303 |
* also need to account for a reduction in the length of the line pointer |
| 304 |
* array following array truncation by us. |
| 305 |
* |
| 306 |
* If the HEAP_PRUNE_FREEZE option is set, we will also freeze tuples if it's |
| 307 |
* required in order to advance relfrozenxid / relminmxid, or if it's |
| 308 |
* considered advantageous for overall system performance to do so now. The |
| 309 |
* 'cutoffs', 'presult', 'new_relfrozen_xid' and 'new_relmin_mxid' arguments |
| 310 |
* are required when freezing. When HEAP_PRUNE_FREEZE option is set, we also |
| 311 |
* set presult->all_visible and presult->all_frozen on exit, to indicate if |
| 312 |
* the VM bits can be set. They are always set to false when the |
| 313 |
* HEAP_PRUNE_FREEZE option is not set, because at the moment only callers |
| 314 |
* that also freeze need that information. |
| 315 |
* |
| 316 |
* vistest is used to distinguish whether tuples are DEAD or RECENTLY_DEAD |
| 317 |
* (see heap_prune_satisfies_vacuum). |
| 318 |
* |
| 319 |
* options: |
| 320 |
* MARK_UNUSED_NOW indicates that dead items can be set LP_UNUSED during |
| 321 |
* pruning. |
| 322 |
* |
| 323 |
* FREEZE indicates that we will also freeze tuples, and will return |
| 324 |
* 'all_visible', 'all_frozen' flags to the caller. |
| 325 |
* |
| 326 |
* cutoffs contains the freeze cutoffs, established by VACUUM at the beginning |
| 327 |
* of vacuuming the relation. Required if HEAP_PRUNE_FREEZE option is set. |
| 328 |
* cutoffs->OldestXmin is also used to determine if dead tuples are |
| 329 |
* HEAPTUPLE_RECENTLY_DEAD or HEAPTUPLE_DEAD. |
| 330 |
* |
| 331 |
* presult contains output parameters needed by callers, such as the number of |
| 332 |
* tuples removed and the offsets of dead items on the page after pruning. |
| 333 |
* heap_page_prune_and_freeze() is responsible for initializing it. Required |
| 334 |
* by all callers. |
| 335 |
* |
| 336 |
* reason indicates why the pruning is performed. It is included in the WAL |
| 337 |
* record for debugging and analysis purposes, but otherwise has no effect. |
| 338 |
* |
| 339 |
* off_loc is the offset location required by the caller to use in error |
| 340 |
* callback. |
| 341 |
* |
| 342 |
* new_relfrozen_xid and new_relmin_mxid must provided by the caller if the |
| 343 |
* HEAP_PRUNE_FREEZE option is set. On entry, they contain the oldest XID and |
| 344 |
* multi-XID seen on the relation so far. They will be updated with oldest |
| 345 |
* values present on the page after pruning. After processing the whole |
| 346 |
* relation, VACUUM can use these values as the new relfrozenxid/relminmxid |
| 347 |
* for the relation. |
| 348 |
*/ |
| 349 |
void |
| 350 |
heap_page_prune_and_freeze(Relation relation, Buffer buffer, |
| 351 |
GlobalVisState *vistest, |
| 352 |
int options, |
| 353 |
struct VacuumCutoffs *cutoffs, |
| 354 |
PruneFreezeResult *presult, |
| 355 |
PruneReason reason, |
| 356 |
OffsetNumber *off_loc, |
| 357 |
TransactionId *new_relfrozen_xid, |
| 358 |
MultiXactId *new_relmin_mxid) |
| 359 |
{ |
| 360 |
Page page = BufferGetPage(buffer); |
| 361 |
BlockNumber blockno = BufferGetBlockNumber(buffer); |
| 362 |
OffsetNumber offnum, |
| 363 |
maxoff; |
| 364 |
PruneState prstate; |
| 365 |
HeapTupleData tup; |
| 366 |
bool do_freeze; |
| 367 |
bool do_prune; |
| 368 |
bool do_hint; |
| 369 |
bool hint_bit_fpi; |
| 370 |
int64 fpi_before = pgWalUsage.wal_fpi; |
| 371 |
|
| 372 |
/* Copy parameters to prstate */ |
| 373 |
prstate.vistest = vistest; |
| 374 |
prstate.mark_unused_now = (options & HEAP_PAGE_PRUNE_MARK_UNUSED_NOW) != 0; |
| 375 |
prstate.freeze = (options & HEAP_PAGE_PRUNE_FREEZE) != 0; |
| 376 |
prstate.cutoffs = cutoffs; |
| 377 |
|
| 378 |
/* |
| 379 |
* Our strategy is to scan the page and make lists of items to change, |
| 380 |
* then apply the changes within a critical section. This keeps as much |
| 381 |
* logic as possible out of the critical section, and also ensures that |
| 382 |
* WAL replay will work the same as the normal case. |
| 383 |
* |
| 384 |
* First, initialize the new pd_prune_xid value to zero (indicating no |
| 385 |
* prunable tuples). If we find any tuples which may soon become |
| 386 |
* prunable, we will save the lowest relevant XID in new_prune_xid. Also |
| 387 |
* initialize the rest of our working state. |
| 388 |
*/ |
| 389 |
prstate.new_prune_xid = InvalidTransactionId; |
| 390 |
prstate.latest_xid_removed = InvalidTransactionId; |
| 391 |
prstate.nredirected = prstate.ndead = prstate.nunused = prstate.nfrozen = 0; |
| 392 |
prstate.nroot_items = 0; |
| 393 |
prstate.nheaponly_items = 0; |
| 394 |
|
| 395 |
/* initialize page freezing working state */ |
| 396 |
prstate.pagefrz.freeze_required = false; |
| 397 |
if (prstate.freeze) |
| 398 |
{ |
| 399 |
Assert(new_relfrozen_xid && new_relmin_mxid); |
| 400 |
prstate.pagefrz.FreezePageRelfrozenXid = *new_relfrozen_xid; |
| 401 |
prstate.pagefrz.NoFreezePageRelfrozenXid = *new_relfrozen_xid; |
| 402 |
prstate.pagefrz.FreezePageRelminMxid = *new_relmin_mxid; |
| 403 |
prstate.pagefrz.NoFreezePageRelminMxid = *new_relmin_mxid; |
| 404 |
} |
| 405 |
else |
| 406 |
{ |
| 407 |
Assert(new_relfrozen_xid == NULL && new_relmin_mxid == NULL); |
| 408 |
prstate.pagefrz.FreezePageRelminMxid = InvalidMultiXactId; |
| 409 |
prstate.pagefrz.NoFreezePageRelminMxid = InvalidMultiXactId; |
| 410 |
prstate.pagefrz.FreezePageRelfrozenXid = InvalidTransactionId; |
| 411 |
prstate.pagefrz.NoFreezePageRelfrozenXid = InvalidTransactionId; |
| 412 |
} |
| 413 |
|
| 414 |
prstate.ndeleted = 0; |
| 415 |
prstate.live_tuples = 0; |
| 416 |
prstate.recently_dead_tuples = 0; |
| 417 |
prstate.hastup = false; |
| 418 |
prstate.lpdead_items = 0; |
| 419 |
prstate.deadoffsets = presult->deadoffsets; |
| 420 |
|
| 421 |
/* |
| 422 |
* Caller may update the VM after we're done. We can keep track of |
| 423 |
* whether the page will be all-visible and all-frozen after pruning and |
| 424 |
* freezing to help the caller to do that. |
| 425 |
* |
| 426 |
* Currently, only VACUUM sets the VM bits. To save the effort, only do |
| 427 |
* the bookkeeping if the caller needs it. Currently, that's tied to |
| 428 |
* HEAP_PAGE_PRUNE_FREEZE, but it could be a separate flag if you wanted |
| 429 |
* to update the VM bits without also freezing or freeze without also |
| 430 |
* setting the VM bits. |
| 431 |
* |
| 432 |
* In addition to telling the caller whether it can set the VM bit, we |
| 433 |
* also use 'all_visible' and 'all_frozen' for our own decision-making. If |
| 434 |
* the whole page would become frozen, we consider opportunistically |
| 435 |
* freezing tuples. We will not be able to freeze the whole page if there |
| 436 |
* are tuples present that are not visible to everyone or if there are |
| 437 |
* dead tuples which are not yet removable. However, dead tuples which |
| 438 |
* will be removed by the end of vacuuming should not preclude us from |
| 439 |
* opportunistically freezing. Because of that, we do not clear |
| 440 |
* all_visible when we see LP_DEAD items. We fix that at the end of the |
| 441 |
* function, when we return the value to the caller, so that the caller |
| 442 |
* doesn't set the VM bit incorrectly. |
| 443 |
*/ |
| 444 |
if (prstate.freeze) |
| 445 |
{ |
| 446 |
prstate.all_visible = true; |
| 447 |
prstate.all_frozen = true; |
| 448 |
} |
| 449 |
else |
| 450 |
{ |
| 451 |
/* |
| 452 |
* Initializing to false allows skipping the work to update them in |
| 453 |
* heap_prune_record_unchanged_lp_normal(). |
| 454 |
*/ |
| 455 |
prstate.all_visible = false; |
| 456 |
prstate.all_frozen = false; |
| 457 |
} |
| 458 |
|
| 459 |
/* |
| 460 |
* The visibility cutoff xid is the newest xmin of live tuples on the |
| 461 |
* page. In the common case, this will be set as the conflict horizon the |
| 462 |
* caller can use for updating the VM. If, at the end of freezing and |
| 463 |
* pruning, the page is all-frozen, there is no possibility that any |
| 464 |
* running transaction on the standby does not see tuples on the page as |
| 465 |
* all-visible, so the conflict horizon remains InvalidTransactionId. |
| 466 |
*/ |
| 467 |
prstate.visibility_cutoff_xid = InvalidTransactionId; |
| 468 |
|
| 469 |
maxoff = PageGetMaxOffsetNumber(page); |
| 470 |
tup.t_tableOid = RelationGetRelid(relation); |
| 471 |
|
| 472 |
/* |
| 473 |
* Determine HTSV for all tuples, and queue them up for processing as HOT |
| 474 |
* chain roots or as heap-only items. |
| 475 |
* |
| 476 |
* Determining HTSV only once for each tuple is required for correctness, |
| 477 |
* to deal with cases where running HTSV twice could result in different |
| 478 |
* results. For example, RECENTLY_DEAD can turn to DEAD if another |
| 479 |
* checked item causes GlobalVisTestIsRemovableFullXid() to update the |
| 480 |
* horizon, or INSERT_IN_PROGRESS can change to DEAD if the inserting |
| 481 |
* transaction aborts. |
| 482 |
* |
| 483 |
* It's also good for performance. Most commonly tuples within a page are |
| 484 |
* stored at decreasing offsets (while the items are stored at increasing |
| 485 |
* offsets). When processing all tuples on a page this leads to reading |
| 486 |
* memory at decreasing offsets within a page, with a variable stride. |
| 487 |
* That's hard for CPU prefetchers to deal with. Processing the items in |
| 488 |
* reverse order (and thus the tuples in increasing order) increases |
| 489 |
* prefetching efficiency significantly / decreases the number of cache |
| 490 |
* misses. |
| 491 |
*/ |
| 492 |
for (offnum = maxoff; |
| 493 |
offnum >= FirstOffsetNumber; |
| 494 |
offnum = OffsetNumberPrev(offnum)) |
| 495 |
{ |
| 496 |
ItemId itemid = PageGetItemId(page, offnum); |
| 497 |
HeapTupleHeader htup; |
| 498 |
|
| 499 |
/* |
| 500 |
* Set the offset number so that we can display it along with any |
| 501 |
* error that occurred while processing this tuple. |
| 502 |
*/ |
| 503 |
*off_loc = offnum; |
| 504 |
|
| 505 |
prstate.processed[offnum] = false; |
| 506 |
prstate.htsv[offnum] = -1; |
| 507 |
|
| 508 |
/* Nothing to do if slot doesn't contain a tuple */ |
| 509 |
if (!ItemIdIsUsed(itemid)) |
| 510 |
{ |
| 511 |
heap_prune_record_unchanged_lp_unused(page, &prstate, offnum); |
| 512 |
continue; |
| 513 |
} |
| 514 |
|
| 515 |
if (ItemIdIsDead(itemid)) |
| 516 |
{ |
| 517 |
/* |
| 518 |
* If the caller set mark_unused_now true, we can set dead line |
| 519 |
* pointers LP_UNUSED now. |
| 520 |
*/ |
| 521 |
if (unlikely(prstate.mark_unused_now)) |
| 522 |
heap_prune_record_unused(&prstate, offnum, false); |
| 523 |
else |
| 524 |
heap_prune_record_unchanged_lp_dead(page, &prstate, offnum); |
| 525 |
continue; |
| 526 |
} |
| 527 |
|
| 528 |
if (ItemIdIsRedirected(itemid)) |
| 529 |
{ |
| 530 |
/* This is the start of a HOT chain */ |
| 531 |
prstate.root_items[prstate.nroot_items++] = offnum; |
| 532 |
continue; |
| 533 |
} |
| 534 |
|
| 535 |
Assert(ItemIdIsNormal(itemid)); |
| 536 |
|
| 537 |
/* |
| 538 |
* Get the tuple's visibility status and queue it up for processing. |
| 539 |
*/ |
| 540 |
htup = (HeapTupleHeader) PageGetItem(page, itemid); |
| 541 |
tup.t_data = htup; |
| 542 |
tup.t_len = ItemIdGetLength(itemid); |
| 543 |
ItemPointerSet(&tup.t_self, blockno, offnum); |
| 544 |
|
| 545 |
prstate.htsv[offnum] = heap_prune_satisfies_vacuum(&prstate, &tup, |
| 546 |
buffer); |
| 547 |
|
| 548 |
if (!HeapTupleHeaderIsHeapOnly(htup)) |
| 549 |
prstate.root_items[prstate.nroot_items++] = offnum; |
| 550 |
else |
| 551 |
prstate.heaponly_items[prstate.nheaponly_items++] = offnum; |
| 552 |
} |
| 553 |
|
| 554 |
/* |
| 555 |
* If checksums are enabled, heap_prune_satisfies_vacuum() may have caused |
| 556 |
* an FPI to be emitted. |
| 557 |
*/ |
| 558 |
hint_bit_fpi = fpi_before != pgWalUsage.wal_fpi; |
| 559 |
|
| 560 |
/* |
| 561 |
* Process HOT chains. |
| 562 |
* |
| 563 |
* We added the items to the array starting from 'maxoff', so by |
| 564 |
* processing the array in reverse order, we process the items in |
| 565 |
* ascending offset number order. The order doesn't matter for |
| 566 |
* correctness, but some quick micro-benchmarking suggests that this is |
| 567 |
* faster. (Earlier PostgreSQL versions, which scanned all the items on |
| 568 |
* the page instead of using the root_items array, also did it in |
| 569 |
* ascending offset number order.) |
| 570 |
*/ |
| 571 |
for (int i = prstate.nroot_items - 1; i >= 0; i--) |
| 572 |
{ |
| 573 |
offnum = prstate.root_items[i]; |
| 574 |
|
| 575 |
/* Ignore items already processed as part of an earlier chain */ |
| 576 |
if (prstate.processed[offnum]) |
| 577 |
continue; |
| 578 |
|
| 579 |
/* see preceding loop */ |
| 580 |
*off_loc = offnum; |
| 581 |
|
| 582 |
/* Process this item or chain of items */ |
| 583 |
heap_prune_chain(page, blockno, maxoff, offnum, &prstate); |
| 584 |
} |
| 585 |
|
| 586 |
/* |
| 587 |
* Process any heap-only tuples that were not already processed as part of |
| 588 |
* a HOT chain. |
| 589 |
*/ |
| 590 |
for (int i = prstate.nheaponly_items - 1; i >= 0; i--) |
| 591 |
{ |
| 592 |
offnum = prstate.heaponly_items[i]; |
| 593 |
|
| 594 |
if (prstate.processed[offnum]) |
| 595 |
continue; |
| 596 |
|
| 597 |
/* see preceding loop */ |
| 598 |
*off_loc = offnum; |
| 599 |
|
| 600 |
/* |
| 601 |
* If the tuple is DEAD and doesn't chain to anything else, mark it |
| 602 |
* unused. (If it does chain, we can only remove it as part of |
| 603 |
* pruning its chain.) |
| 604 |
* |
| 605 |
* We need this primarily to handle aborted HOT updates, that is, |
| 606 |
* XMIN_INVALID heap-only tuples. Those might not be linked to by any |
| 607 |
* chain, since the parent tuple might be re-updated before any |
| 608 |
* pruning occurs. So we have to be able to reap them separately from |
| 609 |
* chain-pruning. (Note that HeapTupleHeaderIsHotUpdated will never |
| 610 |
* return true for an XMIN_INVALID tuple, so this code will work even |
| 611 |
* when there were sequential updates within the aborted transaction.) |
| 612 |
*/ |
| 613 |
if (prstate.htsv[offnum] == HEAPTUPLE_DEAD) |
| 614 |
{ |
| 615 |
ItemId itemid = PageGetItemId(page, offnum); |
| 616 |
HeapTupleHeader htup = (HeapTupleHeader) PageGetItem(page, itemid); |
| 617 |
|
| 618 |
if (likely(!HeapTupleHeaderIsHotUpdated(htup))) |
| 619 |
{ |
| 620 |
HeapTupleHeaderAdvanceConflictHorizon(htup, |
| 621 |
&prstate.latest_xid_removed); |
| 622 |
heap_prune_record_unused(&prstate, offnum, true); |
| 623 |
} |
| 624 |
else |
| 625 |
{ |
| 626 |
/* |
| 627 |
* This tuple should've been processed and removed as part of |
| 628 |
* a HOT chain, so something's wrong. To preserve evidence, |
| 629 |
* we don't dare to remove it. We cannot leave behind a DEAD |
| 630 |
* tuple either, because that will cause VACUUM to error out. |
| 631 |
* Throwing an error with a distinct error message seems like |
| 632 |
* the least bad option. |
| 633 |
*/ |
| 634 |
elog(ERROR, "dead heap-only tuple (%u, %d) is not linked to from any HOT chain", |
| 635 |
blockno, offnum); |
| 636 |
} |
| 637 |
} |
| 638 |
else |
| 639 |
heap_prune_record_unchanged_lp_normal(page, &prstate, offnum); |
| 640 |
} |
| 641 |
|
| 642 |
/* We should now have processed every tuple exactly once */ |
| 643 |
#ifdef USE_ASSERT_CHECKING |
| 644 |
for (offnum = FirstOffsetNumber; |
| 645 |
offnum <= maxoff; |
| 646 |
offnum = OffsetNumberNext(offnum)) |
| 647 |
{ |
| 648 |
*off_loc = offnum; |
| 649 |
|
| 650 |
Assert(prstate.processed[offnum]); |
| 651 |
} |
| 652 |
#endif |
| 653 |
|
| 654 |
/* Clear the offset information once we have processed the given page. */ |
| 655 |
*off_loc = InvalidOffsetNumber; |
| 656 |
|
| 657 |
do_prune = prstate.nredirected > 0 || |
| 658 |
prstate.ndead > 0 || |
| 659 |
prstate.nunused > 0; |
| 660 |
|
| 661 |
/* |
| 662 |
* Even if we don't prune anything, if we found a new value for the |
| 663 |
* pd_prune_xid field or the page was marked full, we will update the hint |
| 664 |
* bit. |
| 665 |
*/ |
| 666 |
do_hint = ((PageHeader) page)->pd_prune_xid != prstate.new_prune_xid || |
| 667 |
PageIsFull(page); |
| 668 |
|
| 669 |
/* |
| 670 |
* Decide if we want to go ahead with freezing according to the freeze |
| 671 |
* plans we prepared, or not. |
| 672 |
*/ |
| 673 |
do_freeze = false; |
| 674 |
if (prstate.freeze) |
| 675 |
{ |
| 676 |
if (prstate.pagefrz.freeze_required) |
| 677 |
{ |
| 678 |
/* |
| 679 |
* heap_prepare_freeze_tuple indicated that at least one XID/MXID |
| 680 |
* from before FreezeLimit/MultiXactCutoff is present. Must |
| 681 |
* freeze to advance relfrozenxid/relminmxid. |
| 682 |
*/ |
| 683 |
do_freeze = true; |
| 684 |
} |
| 685 |
else |
| 686 |
{ |
| 687 |
/* |
| 688 |
* Opportunistically freeze the page if we are generating an FPI |
| 689 |
* anyway and if doing so means that we can set the page |
| 690 |
* all-frozen afterwards (might not happen until VACUUM's final |
| 691 |
* heap pass). |
| 692 |
* |
| 693 |
* XXX: Previously, we knew if pruning emitted an FPI by checking |
| 694 |
* pgWalUsage.wal_fpi before and after pruning. Once the freeze |
| 695 |
* and prune records were combined, this heuristic couldn't be |
| 696 |
* used anymore. The opportunistic freeze heuristic must be |
| 697 |
* improved; however, for now, try to approximate the old logic. |
| 698 |
*/ |
| 699 |
if (prstate.all_visible && prstate.all_frozen && prstate.nfrozen > 0) |
| 700 |
{ |
| 701 |
/* |
| 702 |
* Freezing would make the page all-frozen. Have already |
| 703 |
* emitted an FPI or will do so anyway? |
| 704 |
*/ |
| 705 |
if (RelationNeedsWAL(relation)) |
| 706 |
{ |
| 707 |
if (hint_bit_fpi) |
| 708 |
do_freeze = true; |
| 709 |
else if (do_prune) |
| 710 |
{ |
| 711 |
if (XLogCheckBufferNeedsBackup(buffer)) |
| 712 |
do_freeze = true; |
| 713 |
} |
| 714 |
else if (do_hint) |
| 715 |
{ |
| 716 |
if (XLogHintBitIsNeeded() && XLogCheckBufferNeedsBackup(buffer)) |
| 717 |
do_freeze = true; |
| 718 |
} |
| 719 |
} |
| 720 |
} |
| 721 |
} |
| 722 |
} |
| 723 |
|
| 724 |
if (do_freeze) |
| 725 |
{ |
| 726 |
/* |
| 727 |
* Validate the tuples we will be freezing before entering the |
| 728 |
* critical section. |
| 729 |
*/ |
| 730 |
heap_pre_freeze_checks(buffer, prstate.frozen, prstate.nfrozen); |
| 731 |
} |
| 732 |
else if (prstate.nfrozen > 0) |
| 733 |
{ |
| 734 |
/* |
| 735 |
* The page contained some tuples that were not already frozen, and we |
| 736 |
* chose not to freeze them now. The page won't be all-frozen then. |
| 737 |
*/ |
| 738 |
Assert(!prstate.pagefrz.freeze_required); |
| 739 |
|
| 740 |
prstate.all_frozen = false; |
| 741 |
prstate.nfrozen = 0; /* avoid miscounts in instrumentation */ |
| 742 |
} |
| 743 |
else |
| 744 |
{ |
| 745 |
/* |
| 746 |
* We have no freeze plans to execute. The page might already be |
| 747 |
* all-frozen (perhaps only following pruning), though. Such pages |
| 748 |
* can be marked all-frozen in the VM by our caller, even though none |
| 749 |
* of its tuples were newly frozen here. |
| 750 |
*/ |
| 751 |
} |
| 752 |
|
| 753 |
/* Any error while applying the changes is critical */ |
| 754 |
START_CRIT_SECTION(); |
| 755 |
|
| 756 |
if (do_hint) |
| 757 |
{ |
| 758 |
/* |
| 759 |
* Update the page's pd_prune_xid field to either zero, or the lowest |
| 760 |
* XID of any soon-prunable tuple. |
| 761 |
*/ |
| 762 |
((PageHeader) page)->pd_prune_xid = prstate.new_prune_xid; |
| 763 |
|
| 764 |
/* |
| 765 |
* Also clear the "page is full" flag, since there's no point in |
| 766 |
* repeating the prune/defrag process until something else happens to |
| 767 |
* the page. |
| 768 |
*/ |
| 769 |
PageClearFull(page); |
| 770 |
|
| 771 |
/* |
| 772 |
* If that's all we had to do to the page, this is a non-WAL-logged |
| 773 |
* hint. If we are going to freeze or prune the page, we will mark |
| 774 |
* the buffer dirty below. |
| 775 |
*/ |
| 776 |
if (!do_freeze && !do_prune) |
| 777 |
MarkBufferDirtyHint(buffer, true); |
| 778 |
} |
| 779 |
|
| 780 |
if (do_prune || do_freeze) |
| 781 |
{ |
| 782 |
/* Apply the planned item changes and repair page fragmentation. */ |
| 783 |
if (do_prune) |
| 784 |
{ |
| 785 |
heap_page_prune_execute(buffer, false, |
| 786 |
prstate.redirected, prstate.nredirected, |
| 787 |
prstate.nowdead, prstate.ndead, |
| 788 |
prstate.nowunused, prstate.nunused); |
| 789 |
} |
| 790 |
|
| 791 |
if (do_freeze) |
| 792 |
heap_freeze_prepared_tuples(buffer, prstate.frozen, prstate.nfrozen); |
| 793 |
|
| 794 |
MarkBufferDirty(buffer); |
| 795 |
|
| 796 |
/* |
| 797 |
* Emit a WAL XLOG_HEAP2_PRUNE_FREEZE record showing what we did |
| 798 |
*/ |
| 799 |
if (RelationNeedsWAL(relation)) |
| 800 |
{ |
| 801 |
/* |
| 802 |
* The snapshotConflictHorizon for the whole record should be the |
| 803 |
* most conservative of all the horizons calculated for any of the |
| 804 |
* possible modifications. If this record will prune tuples, any |
| 805 |
* transactions on the standby older than the youngest xmax of the |
| 806 |
* most recently removed tuple this record will prune will |
| 807 |
* conflict. If this record will freeze tuples, any transactions |
| 808 |
* on the standby with xids older than the youngest tuple this |
| 809 |
* record will freeze will conflict. |
| 810 |
*/ |
| 811 |
TransactionId frz_conflict_horizon = InvalidTransactionId; |
| 812 |
TransactionId conflict_xid; |
| 813 |
|
| 814 |
/* |
| 815 |
* We can use the visibility_cutoff_xid as our cutoff for |
| 816 |
* conflicts when the whole page is eligible to become all-frozen |
| 817 |
* in the VM once we're done with it. Otherwise we generate a |
| 818 |
* conservative cutoff by stepping back from OldestXmin. |
| 819 |
*/ |
| 820 |
if (do_freeze) |
| 821 |
{ |
| 822 |
if (prstate.all_visible && prstate.all_frozen) |
| 823 |
frz_conflict_horizon = prstate.visibility_cutoff_xid; |
| 824 |
else |
| 825 |
{ |
| 826 |
/* Avoids false conflicts when hot_standby_feedback in use */ |
| 827 |
frz_conflict_horizon = prstate.cutoffs->OldestXmin; |
| 828 |
TransactionIdRetreat(frz_conflict_horizon); |
| 829 |
} |
| 830 |
} |
| 831 |
|
| 832 |
if (TransactionIdFollows(frz_conflict_horizon, prstate.latest_xid_removed)) |
| 833 |
conflict_xid = frz_conflict_horizon; |
| 834 |
else |
| 835 |
conflict_xid = prstate.latest_xid_removed; |
| 836 |
|
| 837 |
log_heap_prune_and_freeze(relation, buffer, |
| 838 |
conflict_xid, |
| 839 |
true, reason, |
| 840 |
prstate.frozen, prstate.nfrozen, |
| 841 |
prstate.redirected, prstate.nredirected, |
| 842 |
prstate.nowdead, prstate.ndead, |
| 843 |
prstate.nowunused, prstate.nunused); |
| 844 |
} |
| 845 |
} |
| 846 |
|
| 847 |
END_CRIT_SECTION(); |
| 848 |
|
| 849 |
/* Copy information back for caller */ |
| 850 |
presult->ndeleted = prstate.ndeleted; |
| 851 |
presult->nnewlpdead = prstate.ndead; |
| 852 |
presult->nfrozen = prstate.nfrozen; |
| 853 |
presult->live_tuples = prstate.live_tuples; |
| 854 |
presult->recently_dead_tuples = prstate.recently_dead_tuples; |
| 855 |
|
| 856 |
/* |
| 857 |
* It was convenient to ignore LP_DEAD items in all_visible earlier on to |
| 858 |
* make the choice of whether or not to freeze the page unaffected by the |
| 859 |
* short-term presence of LP_DEAD items. These LP_DEAD items were |
| 860 |
* effectively assumed to be LP_UNUSED items in the making. It doesn't |
| 861 |
* matter which vacuum heap pass (initial pass or final pass) ends up |
| 862 |
* setting the page all-frozen, as long as the ongoing VACUUM does it. |
| 863 |
* |
| 864 |
* Now that freezing has been finalized, unset all_visible if there are |
| 865 |
* any LP_DEAD items on the page. It needs to reflect the present state |
| 866 |
* of the page, as expected by our caller. |
| 867 |
*/ |
| 868 |
if (prstate.all_visible && prstate.lpdead_items == 0) |
| 869 |
{ |
| 870 |
presult->all_visible = prstate.all_visible; |
| 871 |
presult->all_frozen = prstate.all_frozen; |
| 872 |
} |
| 873 |
else |
| 874 |
{ |
| 875 |
presult->all_visible = false; |
| 876 |
presult->all_frozen = false; |
| 877 |
} |
| 878 |
|
| 879 |
presult->hastup = prstate.hastup; |
| 880 |
|
| 881 |
/* |
| 882 |
* For callers planning to update the visibility map, the conflict horizon |
| 883 |
* for that record must be the newest xmin on the page. However, if the |
| 884 |
* page is completely frozen, there can be no conflict and the |
| 885 |
* vm_conflict_horizon should remain InvalidTransactionId. This includes |
| 886 |
* the case that we just froze all the tuples; the prune-freeze record |
| 887 |
* included the conflict XID already so the caller doesn't need it. |
| 888 |
*/ |
| 889 |
if (presult->all_frozen) |
| 890 |
presult->vm_conflict_horizon = InvalidTransactionId; |
| 891 |
else |
| 892 |
presult->vm_conflict_horizon = prstate.visibility_cutoff_xid; |
| 893 |
|
| 894 |
presult->lpdead_items = prstate.lpdead_items; |
| 895 |
/* the presult->deadoffsets array was already filled in */ |
| 896 |
|
| 897 |
if (prstate.freeze) |
| 898 |
{ |
| 899 |
if (presult->nfrozen > 0) |
| 900 |
{ |
| 901 |
*new_relfrozen_xid = prstate.pagefrz.FreezePageRelfrozenXid; |
| 902 |
*new_relmin_mxid = prstate.pagefrz.FreezePageRelminMxid; |
| 903 |
} |
| 904 |
else |
| 905 |
{ |
| 906 |
*new_relfrozen_xid = prstate.pagefrz.NoFreezePageRelfrozenXid; |
| 907 |
*new_relmin_mxid = prstate.pagefrz.NoFreezePageRelminMxid; |
| 908 |
} |
| 909 |
} |
| 910 |
} |
| 911 |
|
| 912 |
|
| 913 |
/* |
| 914 |
* Perform visibility checks for heap pruning. |
| 915 |
*/ |
| 916 |
static HTSV_Result |
| 917 |
heap_prune_satisfies_vacuum(PruneState *prstate, HeapTuple tup, Buffer buffer) |
| 918 |
{ |
| 919 |
HTSV_Result res; |
| 920 |
TransactionId dead_after; |
| 921 |
|
| 922 |
res = HeapTupleSatisfiesVacuumHorizon(tup, buffer, &dead_after); |
| 923 |
|
| 924 |
if (res != HEAPTUPLE_RECENTLY_DEAD) |
| 925 |
return res; |
| 926 |
|
| 927 |
/* |
| 928 |
* For VACUUM, we must be sure to prune tuples with xmax older than |
| 929 |
* OldestXmin -- a visibility cutoff determined at the beginning of |
| 930 |
* vacuuming the relation. OldestXmin is used for freezing determination |
| 931 |
* and we cannot freeze dead tuples' xmaxes. |
| 932 |
*/ |
| 933 |
if (prstate->cutoffs && |
| 934 |
TransactionIdIsValid(prstate->cutoffs->OldestXmin) && |
| 935 |
NormalTransactionIdPrecedes(dead_after, prstate->cutoffs->OldestXmin)) |
| 936 |
return HEAPTUPLE_DEAD; |
| 937 |
|
| 938 |
/* |
| 939 |
* Determine whether or not the tuple is considered dead when compared |
| 940 |
* with the provided GlobalVisState. On-access pruning does not provide |
| 941 |
* VacuumCutoffs. And for vacuum, even if the tuple's xmax is not older |
| 942 |
* than OldestXmin, GlobalVisTestIsRemovableXid() could find the row dead |
| 943 |
* if the GlobalVisState has been updated since the beginning of vacuuming |
| 944 |
* the relation. |
| 945 |
*/ |
| 946 |
if (GlobalVisTestIsRemovableXid(prstate->vistest, dead_after)) |
| 947 |
return HEAPTUPLE_DEAD; |
| 948 |
|
| 949 |
return res; |
| 950 |
} |
| 951 |
|
| 952 |
|
| 953 |
/* |
| 954 |
* Pruning calculates tuple visibility once and saves the results in an array |
| 955 |
* of int8. See PruneState.htsv for details. This helper function is meant |
| 956 |
* to guard against examining visibility status array members which have not |
| 957 |
* yet been computed. |
| 958 |
*/ |
| 959 |
static inline HTSV_Result |
| 960 |
htsv_get_valid_status(int status) |
| 961 |
{ |
| 962 |
Assert(status >= HEAPTUPLE_DEAD && |
| 963 |
status <= HEAPTUPLE_DELETE_IN_PROGRESS); |
| 964 |
return (HTSV_Result) status; |
| 965 |
} |
| 966 |
|
| 967 |
/* |
| 968 |
* Prune specified line pointer or a HOT chain originating at line pointer. |
| 969 |
* |
| 970 |
* Tuple visibility information is provided in prstate->htsv. |
| 971 |
* |
| 972 |
* If the item is an index-referenced tuple (i.e. not a heap-only tuple), |
| 973 |
* the HOT chain is pruned by removing all DEAD tuples at the start of the HOT |
| 974 |
* chain. We also prune any RECENTLY_DEAD tuples preceding a DEAD tuple. |
| 975 |
* This is OK because a RECENTLY_DEAD tuple preceding a DEAD tuple is really |
| 976 |
* DEAD, our visibility test is just too coarse to detect it. |
| 977 |
* |
| 978 |
* Pruning must never leave behind a DEAD tuple that still has tuple storage. |
| 979 |
* VACUUM isn't prepared to deal with that case. |
| 980 |
* |
| 981 |
* The root line pointer is redirected to the tuple immediately after the |
| 982 |
* latest DEAD tuple. If all tuples in the chain are DEAD, the root line |
| 983 |
* pointer is marked LP_DEAD. (This includes the case of a DEAD simple |
| 984 |
* tuple, which we treat as a chain of length 1.) |
| 985 |
* |
| 986 |
* We don't actually change the page here. We just add entries to the arrays in |
| 987 |
* prstate showing the changes to be made. Items to be redirected are added |
| 988 |
* to the redirected[] array (two entries per redirection); items to be set to |
| 989 |
* LP_DEAD state are added to nowdead[]; and items to be set to LP_UNUSED |
| 990 |
* state are added to nowunused[]. We perform bookkeeping of live tuples, |
| 991 |
* visibility etc. based on what the page will look like after the changes |
| 992 |
* applied. All that bookkeeping is performed in the heap_prune_record_*() |
| 993 |
* subroutines. The division of labor is that heap_prune_chain() decides the |
| 994 |
* fate of each tuple, ie. whether it's going to be removed, redirected or |
| 995 |
* left unchanged, and the heap_prune_record_*() subroutines update PruneState |
| 996 |
* based on that outcome. |
| 997 |
*/ |
| 998 |
static void |
| 999 |
heap_prune_chain(Page page, BlockNumber blockno, OffsetNumber maxoff, |
| 1000 |
OffsetNumber rootoffnum, PruneState *prstate) |
| 1001 |
{ |
| 1002 |
TransactionId priorXmax = InvalidTransactionId; |
| 1003 |
ItemId rootlp; |
| 1004 |
OffsetNumber offnum; |
| 1005 |
OffsetNumber chainitems[MaxHeapTuplesPerPage]; |
| 1006 |
|
| 1007 |
/* |
| 1008 |
* After traversing the HOT chain, ndeadchain is the index in chainitems |
| 1009 |
* of the first live successor after the last dead item. |
| 1010 |
*/ |
| 1011 |
int ndeadchain = 0, |
| 1012 |
nchain = 0; |
| 1013 |
|
| 1014 |
rootlp = PageGetItemId(page, rootoffnum); |
| 1015 |
|
| 1016 |
/* Start from the root tuple */ |
| 1017 |
offnum = rootoffnum; |
| 1018 |
|
| 1019 |
/* while not end of the chain */ |
| 1020 |
for (;;) |
| 1021 |
{ |
| 1022 |
HeapTupleHeader htup; |
| 1023 |
ItemId lp; |
| 1024 |
|
| 1025 |
/* Sanity check (pure paranoia) */ |
| 1026 |
if (offnum < FirstOffsetNumber) |
| 1027 |
break; |
| 1028 |
|
| 1029 |
/* |
| 1030 |
* An offset past the end of page's line pointer array is possible |
| 1031 |
* when the array was truncated (original item must have been unused) |
| 1032 |
*/ |
| 1033 |
if (offnum > maxoff) |
| 1034 |
break; |
| 1035 |
|
| 1036 |
/* If item is already processed, stop --- it must not be same chain */ |
| 1037 |
if (prstate->processed[offnum]) |
| 1038 |
break; |
| 1039 |
|
| 1040 |
lp = PageGetItemId(page, offnum); |
| 1041 |
|
| 1042 |
/* |
| 1043 |
* Unused item obviously isn't part of the chain. Likewise, a dead |
| 1044 |
* line pointer can't be part of the chain. Both of those cases were |
| 1045 |
* already marked as processed. |
| 1046 |
*/ |
| 1047 |
Assert(ItemIdIsUsed(lp)); |
| 1048 |
Assert(!ItemIdIsDead(lp)); |
| 1049 |
|
| 1050 |
/* |
| 1051 |
* If we are looking at the redirected root line pointer, jump to the |
| 1052 |
* first normal tuple in the chain. If we find a redirect somewhere |
| 1053 |
* else, stop --- it must not be same chain. |
| 1054 |
*/ |
| 1055 |
if (ItemIdIsRedirected(lp)) |
| 1056 |
{ |
| 1057 |
if (nchain > 0) |
| 1058 |
break; /* not at start of chain */ |
| 1059 |
chainitems[nchain++] = offnum; |
| 1060 |
offnum = ItemIdGetRedirect(rootlp); |
| 1061 |
continue; |
| 1062 |
} |
| 1063 |
|
| 1064 |
Assert(ItemIdIsNormal(lp)); |
| 1065 |
|
| 1066 |
htup = (HeapTupleHeader) PageGetItem(page, lp); |
| 1067 |
|
| 1068 |
/* |
| 1069 |
* Check the tuple XMIN against prior XMAX, if any |
| 1070 |
*/ |
| 1071 |
if (TransactionIdIsValid(priorXmax) && |
| 1072 |
!TransactionIdEquals(HeapTupleHeaderGetXmin(htup), priorXmax)) |
| 1073 |
break; |
| 1074 |
|
| 1075 |
/* |
| 1076 |
* OK, this tuple is indeed a member of the chain. |
| 1077 |
*/ |
| 1078 |
chainitems[nchain++] = offnum; |
| 1079 |
|
| 1080 |
switch (htsv_get_valid_status(prstate->htsv[offnum])) |
| 1081 |
{ |
| 1082 |
case HEAPTUPLE_DEAD: |
| 1083 |
|
| 1084 |
/* Remember the last DEAD tuple seen */ |
| 1085 |
ndeadchain = nchain; |
| 1086 |
HeapTupleHeaderAdvanceConflictHorizon(htup, |
| 1087 |
&prstate->latest_xid_removed); |
| 1088 |
/* Advance to next chain member */ |
| 1089 |
break; |
| 1090 |
|
| 1091 |
case HEAPTUPLE_RECENTLY_DEAD: |
| 1092 |
|
| 1093 |
/* |
| 1094 |
* We don't need to advance the conflict horizon for |
| 1095 |
* RECENTLY_DEAD tuples, even if we are removing them. This |
| 1096 |
* is because we only remove RECENTLY_DEAD tuples if they |
| 1097 |
* precede a DEAD tuple, and the DEAD tuple must have been |
| 1098 |
* inserted by a newer transaction than the RECENTLY_DEAD |
| 1099 |
* tuple by virtue of being later in the chain. We will have |
| 1100 |
* advanced the conflict horizon for the DEAD tuple. |
| 1101 |
*/ |
| 1102 |
|
| 1103 |
/* |
| 1104 |
* Advance past RECENTLY_DEAD tuples just in case there's a |
| 1105 |
* DEAD one after them. We have to make sure that we don't |
| 1106 |
* miss any DEAD tuples, since DEAD tuples that still have |
| 1107 |
* tuple storage after pruning will confuse VACUUM. |
| 1108 |
*/ |
| 1109 |
break; |
| 1110 |
|
| 1111 |
case HEAPTUPLE_DELETE_IN_PROGRESS: |
| 1112 |
case HEAPTUPLE_LIVE: |
| 1113 |
case HEAPTUPLE_INSERT_IN_PROGRESS: |
| 1114 |
goto process_chain; |
| 1115 |
|
| 1116 |
default: |
| 1117 |
elog(ERROR, "unexpected HeapTupleSatisfiesVacuum result"); |
| 1118 |
goto process_chain; |
| 1119 |
} |
| 1120 |
|
| 1121 |
/* |
| 1122 |
* If the tuple is not HOT-updated, then we are at the end of this |
| 1123 |
* HOT-update chain. |
| 1124 |
*/ |
| 1125 |
if (!HeapTupleHeaderIsHotUpdated(htup)) |
| 1126 |
goto process_chain; |
| 1127 |
|
| 1128 |
/* HOT implies it can't have moved to different partition */ |
| 1129 |
Assert(!HeapTupleHeaderIndicatesMovedPartitions(htup)); |
| 1130 |
|
| 1131 |
/* |
| 1132 |
* Advance to next chain member. |
| 1133 |
*/ |
| 1134 |
Assert(ItemPointerGetBlockNumber(&htup->t_ctid) == blockno); |
| 1135 |
offnum = ItemPointerGetOffsetNumber(&htup->t_ctid); |
| 1136 |
priorXmax = HeapTupleHeaderGetUpdateXid(htup); |
| 1137 |
} |
| 1138 |
|
| 1139 |
if (ItemIdIsRedirected(rootlp) && nchain < 2) |
| 1140 |
{ |
| 1141 |
/* |
| 1142 |
* We found a redirect item that doesn't point to a valid follow-on |
| 1143 |
* item. This can happen if the loop in heap_page_prune_and_freeze() |
| 1144 |
* caused us to visit the dead successor of a redirect item before |
| 1145 |
* visiting the redirect item. We can clean up by setting the |
| 1146 |
* redirect item to LP_DEAD state or LP_UNUSED if the caller |
| 1147 |
* indicated. |
| 1148 |
*/ |
| 1149 |
heap_prune_record_dead_or_unused(prstate, rootoffnum, false); |
| 1150 |
return; |
| 1151 |
} |
| 1152 |
|
| 1153 |
process_chain: |
| 1154 |
|
| 1155 |
if (ndeadchain == 0) |
| 1156 |
{ |
| 1157 |
/* |
| 1158 |
* No DEAD tuple was found, so the chain is entirely composed of |
| 1159 |
* normal, unchanged tuples. Leave it alone. |
| 1160 |
*/ |
| 1161 |
int i = 0; |
| 1162 |
|
| 1163 |
if (ItemIdIsRedirected(rootlp)) |
| 1164 |
{ |
| 1165 |
heap_prune_record_unchanged_lp_redirect(prstate, rootoffnum); |
| 1166 |
i++; |
| 1167 |
} |
| 1168 |
for (; i < nchain; i++) |
| 1169 |
heap_prune_record_unchanged_lp_normal(page, prstate, chainitems[i]); |
| 1170 |
} |
| 1171 |
else if (ndeadchain == nchain) |
| 1172 |
{ |
| 1173 |
/* |
| 1174 |
* The entire chain is dead. Mark the root line pointer LP_DEAD, and |
| 1175 |
* fully remove the other tuples in the chain. |
| 1176 |
*/ |
| 1177 |
heap_prune_record_dead_or_unused(prstate, rootoffnum, ItemIdIsNormal(rootlp)); |
| 1178 |
for (int i = 1; i < nchain; i++) |
| 1179 |
heap_prune_record_unused(prstate, chainitems[i], true); |
| 1180 |
} |
| 1181 |
else |
| 1182 |
{ |
| 1183 |
/* |
| 1184 |
* We found a DEAD tuple in the chain. Redirect the root line pointer |
| 1185 |
* to the first non-DEAD tuple, and mark as unused each intermediate |
| 1186 |
* item that we are able to remove from the chain. |
| 1187 |
*/ |
| 1188 |
heap_prune_record_redirect(prstate, rootoffnum, chainitems[ndeadchain], |
| 1189 |
ItemIdIsNormal(rootlp)); |
| 1190 |
for (int i = 1; i < ndeadchain; i++) |
| 1191 |
heap_prune_record_unused(prstate, chainitems[i], true); |
| 1192 |
|
| 1193 |
/* the rest of tuples in the chain are normal, unchanged tuples */ |
| 1194 |
for (int i = ndeadchain; i < nchain; i++) |
| 1195 |
heap_prune_record_unchanged_lp_normal(page, prstate, chainitems[i]); |
| 1196 |
} |
| 1197 |
} |
| 1198 |
|
| 1199 |
/* Record lowest soon-prunable XID */ |
| 1200 |
static void |
| 1201 |
heap_prune_record_prunable(PruneState *prstate, TransactionId xid) |
| 1202 |
{ |
| 1203 |
/* |
| 1204 |
* This should exactly match the PageSetPrunable macro. We can't store |
| 1205 |
* directly into the page header yet, so we update working state. |
| 1206 |
*/ |
| 1207 |
Assert(TransactionIdIsNormal(xid)); |
| 1208 |
if (!TransactionIdIsValid(prstate->new_prune_xid) || |
| 1209 |
TransactionIdPrecedes(xid, prstate->new_prune_xid)) |
| 1210 |
prstate->new_prune_xid = xid; |
| 1211 |
} |
| 1212 |
|
| 1213 |
/* Record line pointer to be redirected */ |
| 1214 |
static void |
| 1215 |
heap_prune_record_redirect(PruneState *prstate, |
| 1216 |
OffsetNumber offnum, OffsetNumber rdoffnum, |
| 1217 |
bool was_normal) |
| 1218 |
{ |
| 1219 |
Assert(!prstate->processed[offnum]); |
| 1220 |
prstate->processed[offnum] = true; |
| 1221 |
|
| 1222 |
/* |
| 1223 |
* Do not mark the redirect target here. It needs to be counted |
| 1224 |
* separately as an unchanged tuple. |
| 1225 |
*/ |
| 1226 |
|
| 1227 |
Assert(prstate->nredirected < MaxHeapTuplesPerPage); |
| 1228 |
prstate->redirected[prstate->nredirected * 2] = offnum; |
| 1229 |
prstate->redirected[prstate->nredirected * 2 + 1] = rdoffnum; |
| 1230 |
|
| 1231 |
prstate->nredirected++; |
| 1232 |
|
| 1233 |
/* |
| 1234 |
* If the root entry had been a normal tuple, we are deleting it, so count |
| 1235 |
* it in the result. But changing a redirect (even to DEAD state) doesn't |
| 1236 |
* count. |
| 1237 |
*/ |
| 1238 |
if (was_normal) |
| 1239 |
prstate->ndeleted++; |
| 1240 |
|
| 1241 |
prstate->hastup = true; |
| 1242 |
} |
| 1243 |
|
| 1244 |
/* Record line pointer to be marked dead */ |
| 1245 |
static void |
| 1246 |
heap_prune_record_dead(PruneState *prstate, OffsetNumber offnum, |
| 1247 |
bool was_normal) |
| 1248 |
{ |
| 1249 |
Assert(!prstate->processed[offnum]); |
| 1250 |
prstate->processed[offnum] = true; |
| 1251 |
|
| 1252 |
Assert(prstate->ndead < MaxHeapTuplesPerPage); |
| 1253 |
prstate->nowdead[prstate->ndead] = offnum; |
| 1254 |
prstate->ndead++; |
| 1255 |
|
| 1256 |
/* |
| 1257 |
* Deliberately delay unsetting all_visible until later during pruning. |
| 1258 |
* Removable dead tuples shouldn't preclude freezing the page. |
| 1259 |
*/ |
| 1260 |
|
| 1261 |
/* Record the dead offset for vacuum */ |
| 1262 |
prstate->deadoffsets[prstate->lpdead_items++] = offnum; |
| 1263 |
|
| 1264 |
/* |
| 1265 |
* If the root entry had been a normal tuple, we are deleting it, so count |
| 1266 |
* it in the result. But changing a redirect (even to DEAD state) doesn't |
| 1267 |
* count. |
| 1268 |
*/ |
| 1269 |
if (was_normal) |
| 1270 |
prstate->ndeleted++; |
| 1271 |
} |
| 1272 |
|
| 1273 |
/* |
| 1274 |
* Depending on whether or not the caller set mark_unused_now to true, record that a |
| 1275 |
* line pointer should be marked LP_DEAD or LP_UNUSED. There are other cases in |
| 1276 |
* which we will mark line pointers LP_UNUSED, but we will not mark line |
| 1277 |
* pointers LP_DEAD if mark_unused_now is true. |
| 1278 |
*/ |
| 1279 |
static void |
| 1280 |
heap_prune_record_dead_or_unused(PruneState *prstate, OffsetNumber offnum, |
| 1281 |
bool was_normal) |
| 1282 |
{ |
| 1283 |
/* |
| 1284 |
* If the caller set mark_unused_now to true, we can remove dead tuples |
| 1285 |
* during pruning instead of marking their line pointers dead. Set this |
| 1286 |
* tuple's line pointer LP_UNUSED. We hint that this option is less |
| 1287 |
* likely. |
| 1288 |
*/ |
| 1289 |
if (unlikely(prstate->mark_unused_now)) |
| 1290 |
heap_prune_record_unused(prstate, offnum, was_normal); |
| 1291 |
else |
| 1292 |
heap_prune_record_dead(prstate, offnum, was_normal); |
| 1293 |
} |
| 1294 |
|
| 1295 |
/* Record line pointer to be marked unused */ |
| 1296 |
static void |
| 1297 |
heap_prune_record_unused(PruneState *prstate, OffsetNumber offnum, bool was_normal) |
| 1298 |
{ |
| 1299 |
Assert(!prstate->processed[offnum]); |
| 1300 |
prstate->processed[offnum] = true; |
| 1301 |
|
| 1302 |
Assert(prstate->nunused < MaxHeapTuplesPerPage); |
| 1303 |
prstate->nowunused[prstate->nunused] = offnum; |
| 1304 |
prstate->nunused++; |
| 1305 |
|
| 1306 |
/* |
| 1307 |
* If the root entry had been a normal tuple, we are deleting it, so count |
| 1308 |
* it in the result. But changing a redirect (even to DEAD state) doesn't |
| 1309 |
* count. |
| 1310 |
*/ |
| 1311 |
if (was_normal) |
| 1312 |
prstate->ndeleted++; |
| 1313 |
} |
| 1314 |
|
| 1315 |
/* |
| 1316 |
* Record an unused line pointer that is left unchanged. |
| 1317 |
*/ |
| 1318 |
static void |
| 1319 |
heap_prune_record_unchanged_lp_unused(Page page, PruneState *prstate, OffsetNumber offnum) |
| 1320 |
{ |
| 1321 |
Assert(!prstate->processed[offnum]); |
| 1322 |
prstate->processed[offnum] = true; |
| 1323 |
} |
| 1324 |
|
| 1325 |
/* |
| 1326 |
* Record line pointer that is left unchanged. We consider freezing it, and |
| 1327 |
* update bookkeeping of tuple counts and page visibility. |
| 1328 |
*/ |
| 1329 |
static void |
| 1330 |
heap_prune_record_unchanged_lp_normal(Page page, PruneState *prstate, OffsetNumber offnum) |
| 1331 |
{ |
| 1332 |
HeapTupleHeader htup; |
| 1333 |
|
| 1334 |
Assert(!prstate->processed[offnum]); |
| 1335 |
prstate->processed[offnum] = true; |
| 1336 |
|
| 1337 |
prstate->hastup = true; /* the page is not empty */ |
| 1338 |
|
| 1339 |
/* |
| 1340 |
* The criteria for counting a tuple as live in this block need to match |
| 1341 |
* what analyze.c's acquire_sample_rows() does, otherwise VACUUM and |
| 1342 |
* ANALYZE may produce wildly different reltuples values, e.g. when there |
| 1343 |
* are many recently-dead tuples. |
| 1344 |
* |
| 1345 |
* The logic here is a bit simpler than acquire_sample_rows(), as VACUUM |
| 1346 |
* can't run inside a transaction block, which makes some cases impossible |
| 1347 |
* (e.g. in-progress insert from the same transaction). |
| 1348 |
* |
| 1349 |
* HEAPTUPLE_DEAD are handled by the other heap_prune_record_*() |
| 1350 |
* subroutines. They don't count dead items like acquire_sample_rows() |
| 1351 |
* does, because we assume that all dead items will become LP_UNUSED |
| 1352 |
* before VACUUM finishes. This difference is only superficial. VACUUM |
| 1353 |
* effectively agrees with ANALYZE about DEAD items, in the end. VACUUM |
| 1354 |
* won't remember LP_DEAD items, but only because they're not supposed to |
| 1355 |
* be left behind when it is done. (Cases where we bypass index vacuuming |
| 1356 |
* will violate this optimistic assumption, but the overall impact of that |
| 1357 |
* should be negligible.) |
| 1358 |
*/ |
| 1359 |
htup = (HeapTupleHeader) PageGetItem(page, PageGetItemId(page, offnum)); |
| 1360 |
|
| 1361 |
switch (prstate->htsv[offnum]) |
| 1362 |
{ |
| 1363 |
case HEAPTUPLE_LIVE: |
| 1364 |
|
| 1365 |
/* |
| 1366 |
* Count it as live. Not only is this natural, but it's also what |
| 1367 |
* acquire_sample_rows() does. |
| 1368 |
*/ |
| 1369 |
prstate->live_tuples++; |
| 1370 |
|
| 1371 |
/* |
| 1372 |
* Is the tuple definitely visible to all transactions? |
| 1373 |
* |
| 1374 |
* NB: Like with per-tuple hint bits, we can't set the |
| 1375 |
* PD_ALL_VISIBLE flag if the inserter committed asynchronously. |
| 1376 |
* See SetHintBits for more info. Check that the tuple is hinted |
| 1377 |
* xmin-committed because of that. |
| 1378 |
*/ |
| 1379 |
if (prstate->all_visible) |
| 1380 |
{ |
| 1381 |
TransactionId xmin; |
| 1382 |
|
| 1383 |
if (!HeapTupleHeaderXminCommitted(htup)) |
| 1384 |
{ |
| 1385 |
prstate->all_visible = false; |
| 1386 |
break; |
| 1387 |
} |
| 1388 |
|
| 1389 |
/* |
| 1390 |
* The inserter definitely committed. But is it old enough |
| 1391 |
* that everyone sees it as committed? A FrozenTransactionId |
| 1392 |
* is seen as committed to everyone. Otherwise, we check if |
| 1393 |
* there is a snapshot that considers this xid to still be |
| 1394 |
* running, and if so, we don't consider the page all-visible. |
| 1395 |
*/ |
| 1396 |
xmin = HeapTupleHeaderGetXmin(htup); |
| 1397 |
|
| 1398 |
/* |
| 1399 |
* For now always use prstate->cutoffs for this test, because |
| 1400 |
* we only update 'all_visible' when freezing is requested. We |
| 1401 |
* could use GlobalVisTestIsRemovableXid instead, if a |
| 1402 |
* non-freezing caller wanted to set the VM bit. |
| 1403 |
*/ |
| 1404 |
Assert(prstate->cutoffs); |
| 1405 |
if (!TransactionIdPrecedes(xmin, prstate->cutoffs->OldestXmin)) |
| 1406 |
{ |
| 1407 |
prstate->all_visible = false; |
| 1408 |
break; |
| 1409 |
} |
| 1410 |
|
| 1411 |
/* Track newest xmin on page. */ |
| 1412 |
if (TransactionIdFollows(xmin, prstate->visibility_cutoff_xid) && |
| 1413 |
TransactionIdIsNormal(xmin)) |
| 1414 |
prstate->visibility_cutoff_xid = xmin; |
| 1415 |
} |
| 1416 |
break; |
| 1417 |
|
| 1418 |
case HEAPTUPLE_RECENTLY_DEAD: |
| 1419 |
prstate->recently_dead_tuples++; |
| 1420 |
prstate->all_visible = false; |
| 1421 |
|
| 1422 |
/* |
| 1423 |
* This tuple will soon become DEAD. Update the hint field so |
| 1424 |
* that the page is reconsidered for pruning in future. |
| 1425 |
*/ |
| 1426 |
heap_prune_record_prunable(prstate, |
| 1427 |
HeapTupleHeaderGetUpdateXid(htup)); |
| 1428 |
break; |
| 1429 |
|
| 1430 |
case HEAPTUPLE_INSERT_IN_PROGRESS: |
| 1431 |
|
| 1432 |
/* |
| 1433 |
* We do not count these rows as live, because we expect the |
| 1434 |
* inserting transaction to update the counters at commit, and we |
| 1435 |
* assume that will happen only after we report our results. This |
| 1436 |
* assumption is a bit shaky, but it is what acquire_sample_rows() |
| 1437 |
* does, so be consistent. |
| 1438 |
*/ |
| 1439 |
prstate->all_visible = false; |
| 1440 |
|
| 1441 |
/* |
| 1442 |
* If we wanted to optimize for aborts, we might consider marking |
| 1443 |
* the page prunable when we see INSERT_IN_PROGRESS. But we |
| 1444 |
* don't. See related decisions about when to mark the page |
| 1445 |
* prunable in heapam.c. |
| 1446 |
*/ |
| 1447 |
break; |
| 1448 |
|
| 1449 |
case HEAPTUPLE_DELETE_IN_PROGRESS: |
| 1450 |
|
| 1451 |
/* |
| 1452 |
* This an expected case during concurrent vacuum. Count such |
| 1453 |
* rows as live. As above, we assume the deleting transaction |
| 1454 |
* will commit and update the counters after we report. |
| 1455 |
*/ |
| 1456 |
prstate->live_tuples++; |
| 1457 |
prstate->all_visible = false; |
| 1458 |
|
| 1459 |
/* |
| 1460 |
* This tuple may soon become DEAD. Update the hint field so that |
| 1461 |
* the page is reconsidered for pruning in future. |
| 1462 |
*/ |
| 1463 |
heap_prune_record_prunable(prstate, |
| 1464 |
HeapTupleHeaderGetUpdateXid(htup)); |
| 1465 |
break; |
| 1466 |
|
| 1467 |
default: |
| 1468 |
|
| 1469 |
/* |
| 1470 |
* DEAD tuples should've been passed to heap_prune_record_dead() |
| 1471 |
* or heap_prune_record_unused() instead. |
| 1472 |
*/ |
| 1473 |
elog(ERROR, "unexpected HeapTupleSatisfiesVacuum result %d", |
| 1474 |
prstate->htsv[offnum]); |
| 1475 |
break; |
| 1476 |
} |
| 1477 |
|
| 1478 |
/* Consider freezing any normal tuples which will not be removed */ |
| 1479 |
if (prstate->freeze) |
| 1480 |
{ |
| 1481 |
bool totally_frozen; |
| 1482 |
|
| 1483 |
if ((heap_prepare_freeze_tuple(htup, |
| 1484 |
prstate->cutoffs, |
| 1485 |
&prstate->pagefrz, |
| 1486 |
&prstate->frozen[prstate->nfrozen], |
| 1487 |
&totally_frozen))) |
| 1488 |
{ |
| 1489 |
/* Save prepared freeze plan for later */ |
| 1490 |
prstate->frozen[prstate->nfrozen++].offset = offnum; |
| 1491 |
} |
| 1492 |
|
| 1493 |
/* |
| 1494 |
* If any tuple isn't either totally frozen already or eligible to |
| 1495 |
* become totally frozen (according to its freeze plan), then the page |
| 1496 |
* definitely cannot be set all-frozen in the visibility map later on. |
| 1497 |
*/ |
| 1498 |
if (!totally_frozen) |
| 1499 |
prstate->all_frozen = false; |
| 1500 |
} |
| 1501 |
} |
| 1502 |
|
| 1503 |
|
| 1504 |
/* |
| 1505 |
* Record line pointer that was already LP_DEAD and is left unchanged. |
| 1506 |
*/ |
| 1507 |
static void |
| 1508 |
heap_prune_record_unchanged_lp_dead(Page page, PruneState *prstate, OffsetNumber offnum) |
| 1509 |
{ |
| 1510 |
Assert(!prstate->processed[offnum]); |
| 1511 |
prstate->processed[offnum] = true; |
| 1512 |
|
| 1513 |
/* |
| 1514 |
* Deliberately don't set hastup for LP_DEAD items. We make the soft |
| 1515 |
* assumption that any LP_DEAD items encountered here will become |
| 1516 |
* LP_UNUSED later on, before count_nondeletable_pages is reached. If we |
| 1517 |
* don't make this assumption then rel truncation will only happen every |
| 1518 |
* other VACUUM, at most. Besides, VACUUM must treat |
| 1519 |
* hastup/nonempty_pages as provisional no matter how LP_DEAD items are |
| 1520 |
* handled (handled here, or handled later on). |
| 1521 |
* |
| 1522 |
* Similarly, don't unset all_visible until later, at the end of |
| 1523 |
* heap_page_prune_and_freeze(). This will allow us to attempt to freeze |
| 1524 |
* the page after pruning. As long as we unset it before updating the |
| 1525 |
* visibility map, this will be correct. |
| 1526 |
*/ |
| 1527 |
|
| 1528 |
/* Record the dead offset for vacuum */ |
| 1529 |
prstate->deadoffsets[prstate->lpdead_items++] = offnum; |
| 1530 |
} |
| 1531 |
|
| 1532 |
/* |
| 1533 |
* Record LP_REDIRECT that is left unchanged. |
| 1534 |
*/ |
| 1535 |
static void |
| 1536 |
heap_prune_record_unchanged_lp_redirect(PruneState *prstate, OffsetNumber offnum) |
| 1537 |
{ |
| 1538 |
/* |
| 1539 |
* A redirect line pointer doesn't count as a live tuple. |
| 1540 |
* |
| 1541 |
* If we leave a redirect line pointer in place, there will be another |
| 1542 |
* tuple on the page that it points to. We will do the bookkeeping for |
| 1543 |
* that separately. So we have nothing to do here, except remember that |
| 1544 |
* we processed this item. |
| 1545 |
*/ |
| 1546 |
Assert(!prstate->processed[offnum]); |
| 1547 |
prstate->processed[offnum] = true; |
| 1548 |
} |
| 1549 |
|
| 1550 |
/* |
| 1551 |
* Perform the actual page changes needed by heap_page_prune_and_freeze(). |
| 1552 |
* |
| 1553 |
* If 'lp_truncate_only' is set, we are merely marking LP_DEAD line pointers |
| 1554 |
* as unused, not redirecting or removing anything else. The |
| 1555 |
* PageRepairFragmentation() call is skipped in that case. |
| 1556 |
* |
| 1557 |
* If 'lp_truncate_only' is not set, the caller must hold a cleanup lock on |
| 1558 |
* the buffer. If it is set, an ordinary exclusive lock suffices. |
| 1559 |
*/ |
| 1560 |
void |
| 1561 |
heap_page_prune_execute(Buffer buffer, bool lp_truncate_only, |
| 1562 |
OffsetNumber *redirected, int nredirected, |
| 1563 |
OffsetNumber *nowdead, int ndead, |
| 1564 |
OffsetNumber *nowunused, int nunused) |
| 1565 |
{ |
| 1566 |
Page page = (Page) BufferGetPage(buffer); |
| 1567 |
OffsetNumber *offnum; |
| 1568 |
HeapTupleHeader htup PG_USED_FOR_ASSERTS_ONLY; |
| 1569 |
|
| 1570 |
/* Shouldn't be called unless there's something to do */ |
| 1571 |
Assert(nredirected > 0 || ndead > 0 || nunused > 0); |
| 1572 |
|
| 1573 |
/* If 'lp_truncate_only', we can only remove already-dead line pointers */ |
| 1574 |
Assert(!lp_truncate_only || (nredirected == 0 && ndead == 0)); |
| 1575 |
|
| 1576 |
/* Update all redirected line pointers */ |
| 1577 |
offnum = redirected; |
| 1578 |
for (int i = 0; i < nredirected; i++) |
| 1579 |
{ |
| 1580 |
OffsetNumber fromoff = *offnum++; |
| 1581 |
OffsetNumber tooff = *offnum++; |
| 1582 |
ItemId fromlp = PageGetItemId(page, fromoff); |
| 1583 |
ItemId tolp PG_USED_FOR_ASSERTS_ONLY; |
| 1584 |
|
| 1585 |
#ifdef USE_ASSERT_CHECKING |
| 1586 |
|
| 1587 |
/* |
| 1588 |
* Any existing item that we set as an LP_REDIRECT (any 'from' item) |
| 1589 |
* must be the first item from a HOT chain. If the item has tuple |
| 1590 |
* storage then it can't be a heap-only tuple. Otherwise we are just |
| 1591 |
* maintaining an existing LP_REDIRECT from an existing HOT chain that |
| 1592 |
* has been pruned at least once before now. |
| 1593 |
*/ |
| 1594 |
if (!ItemIdIsRedirected(fromlp)) |
| 1595 |
{ |
| 1596 |
Assert(ItemIdHasStorage(fromlp) && ItemIdIsNormal(fromlp)); |
| 1597 |
|
| 1598 |
htup = (HeapTupleHeader) PageGetItem(page, fromlp); |
| 1599 |
Assert(!HeapTupleHeaderIsHeapOnly(htup)); |
| 1600 |
} |
| 1601 |
else |
| 1602 |
{ |
| 1603 |
/* We shouldn't need to redundantly set the redirect */ |
| 1604 |
Assert(ItemIdGetRedirect(fromlp) != tooff); |
| 1605 |
} |
| 1606 |
|
| 1607 |
/* |
| 1608 |
* The item that we're about to set as an LP_REDIRECT (the 'from' |
| 1609 |
* item) will point to an existing item (the 'to' item) that is |
| 1610 |
* already a heap-only tuple. There can be at most one LP_REDIRECT |
| 1611 |
* item per HOT chain. |
| 1612 |
* |
| 1613 |
* We need to keep around an LP_REDIRECT item (after original |
| 1614 |
* non-heap-only root tuple gets pruned away) so that it's always |
| 1615 |
* possible for VACUUM to easily figure out what TID to delete from |
| 1616 |
* indexes when an entire HOT chain becomes dead. A heap-only tuple |
| 1617 |
* can never become LP_DEAD; an LP_REDIRECT item or a regular heap |
| 1618 |
* tuple can. |
| 1619 |
* |
| 1620 |
* This check may miss problems, e.g. the target of a redirect could |
| 1621 |
* be marked as unused subsequently. The page_verify_redirects() check |
| 1622 |
* below will catch such problems. |
| 1623 |
*/ |
| 1624 |
tolp = PageGetItemId(page, tooff); |
| 1625 |
Assert(ItemIdHasStorage(tolp) && ItemIdIsNormal(tolp)); |
| 1626 |
htup = (HeapTupleHeader) PageGetItem(page, tolp); |
| 1627 |
Assert(HeapTupleHeaderIsHeapOnly(htup)); |
| 1628 |
#endif |
| 1629 |
|
| 1630 |
ItemIdSetRedirect(fromlp, tooff); |
| 1631 |
} |
| 1632 |
|
| 1633 |
/* Update all now-dead line pointers */ |
| 1634 |
offnum = nowdead; |
| 1635 |
for (int i = 0; i < ndead; i++) |
| 1636 |
{ |
| 1637 |
OffsetNumber off = *offnum++; |
| 1638 |
ItemId lp = PageGetItemId(page, off); |
| 1639 |
|
| 1640 |
#ifdef USE_ASSERT_CHECKING |
| 1641 |
|
| 1642 |
/* |
| 1643 |
* An LP_DEAD line pointer must be left behind when the original item |
| 1644 |
* (which is dead to everybody) could still be referenced by a TID in |
| 1645 |
* an index. This should never be necessary with any individual |
| 1646 |
* heap-only tuple item, though. (It's not clear how much of a problem |
| 1647 |
* that would be, but there is no reason to allow it.) |
| 1648 |
*/ |
| 1649 |
if (ItemIdHasStorage(lp)) |
| 1650 |
{ |
| 1651 |
Assert(ItemIdIsNormal(lp)); |
| 1652 |
htup = (HeapTupleHeader) PageGetItem(page, lp); |
| 1653 |
Assert(!HeapTupleHeaderIsHeapOnly(htup)); |
| 1654 |
} |
| 1655 |
else |
| 1656 |
{ |
| 1657 |
/* Whole HOT chain becomes dead */ |
| 1658 |
Assert(ItemIdIsRedirected(lp)); |
| 1659 |
} |
| 1660 |
#endif |
| 1661 |
|
| 1662 |
ItemIdSetDead(lp); |
| 1663 |
} |
| 1664 |
|
| 1665 |
/* Update all now-unused line pointers */ |
| 1666 |
offnum = nowunused; |
| 1667 |
for (int i = 0; i < nunused; i++) |
| 1668 |
{ |
| 1669 |
OffsetNumber off = *offnum++; |
| 1670 |
ItemId lp = PageGetItemId(page, off); |
| 1671 |
|
| 1672 |
#ifdef USE_ASSERT_CHECKING |
| 1673 |
|
| 1674 |
if (lp_truncate_only) |
| 1675 |
{ |
| 1676 |
/* Setting LP_DEAD to LP_UNUSED in vacuum's second pass */ |
| 1677 |
Assert(ItemIdIsDead(lp) && !ItemIdHasStorage(lp)); |
| 1678 |
} |
| 1679 |
else |
| 1680 |
{ |
| 1681 |
/* |
| 1682 |
* When heap_page_prune_and_freeze() was called, mark_unused_now |
| 1683 |
* may have been passed as true, which allows would-be LP_DEAD |
| 1684 |
* items to be made LP_UNUSED instead. This is only possible if |
| 1685 |
* the relation has no indexes. If there are any dead items, then |
| 1686 |
* mark_unused_now was not true and every item being marked |
| 1687 |
* LP_UNUSED must refer to a heap-only tuple. |
| 1688 |
*/ |
| 1689 |
if (ndead > 0) |
| 1690 |
{ |
| 1691 |
Assert(ItemIdHasStorage(lp) && ItemIdIsNormal(lp)); |
| 1692 |
htup = (HeapTupleHeader) PageGetItem(page, lp); |
| 1693 |
Assert(HeapTupleHeaderIsHeapOnly(htup)); |
| 1694 |
} |
| 1695 |
else |
| 1696 |
Assert(ItemIdIsUsed(lp)); |
| 1697 |
} |
| 1698 |
|
| 1699 |
#endif |
| 1700 |
|
| 1701 |
ItemIdSetUnused(lp); |
| 1702 |
} |
| 1703 |
|
| 1704 |
if (lp_truncate_only) |
| 1705 |
PageTruncateLinePointerArray(page); |
| 1706 |
else |
| 1707 |
{ |
| 1708 |
/* |
| 1709 |
* Finally, repair any fragmentation, and update the page's hint bit |
| 1710 |
* about whether it has free pointers. |
| 1711 |
*/ |
| 1712 |
PageRepairFragmentation(page); |
| 1713 |
|
| 1714 |
/* |
| 1715 |
* Now that the page has been modified, assert that redirect items |
| 1716 |
* still point to valid targets. |
| 1717 |
*/ |
| 1718 |
page_verify_redirects(page); |
| 1719 |
} |
| 1720 |
} |
| 1721 |
|
| 1722 |
|
| 1723 |
/* |
| 1724 |
* If built with assertions, verify that all LP_REDIRECT items point to a |
| 1725 |
* valid item. |
| 1726 |
* |
| 1727 |
* One way that bugs related to HOT pruning show is redirect items pointing to |
| 1728 |
* removed tuples. It's not trivial to reliably check that marking an item |
| 1729 |
* unused will not orphan a redirect item during heap_prune_chain() / |
| 1730 |
* heap_page_prune_execute(), so we additionally check the whole page after |
| 1731 |
* pruning. Without this check such bugs would typically only cause asserts |
| 1732 |
* later, potentially well after the corruption has been introduced. |
| 1733 |
* |
| 1734 |
* Also check comments in heap_page_prune_execute()'s redirection loop. |
| 1735 |
*/ |
| 1736 |
static void |
| 1737 |
page_verify_redirects(Page page) |
| 1738 |
{ |
| 1739 |
#ifdef USE_ASSERT_CHECKING |
| 1740 |
OffsetNumber offnum; |
| 1741 |
OffsetNumber maxoff; |
| 1742 |
|
| 1743 |
maxoff = PageGetMaxOffsetNumber(page); |
| 1744 |
for (offnum = FirstOffsetNumber; |
| 1745 |
offnum <= maxoff; |
| 1746 |
offnum = OffsetNumberNext(offnum)) |
| 1747 |
{ |
| 1748 |
ItemId itemid = PageGetItemId(page, offnum); |
| 1749 |
OffsetNumber targoff; |
| 1750 |
ItemId targitem; |
| 1751 |
HeapTupleHeader htup; |
| 1752 |
|
| 1753 |
if (!ItemIdIsRedirected(itemid)) |
| 1754 |
continue; |
| 1755 |
|
| 1756 |
targoff = ItemIdGetRedirect(itemid); |
| 1757 |
targitem = PageGetItemId(page, targoff); |
| 1758 |
|
| 1759 |
Assert(ItemIdIsUsed(targitem)); |
| 1760 |
Assert(ItemIdIsNormal(targitem)); |
| 1761 |
Assert(ItemIdHasStorage(targitem)); |
| 1762 |
htup = (HeapTupleHeader) PageGetItem(page, targitem); |
| 1763 |
Assert(HeapTupleHeaderIsHeapOnly(htup)); |
| 1764 |
} |
| 1765 |
#endif |
| 1766 |
} |
| 1767 |
|
| 1768 |
|
| 1769 |
/* |
| 1770 |
* For all items in this page, find their respective root line pointers. |
| 1771 |
* If item k is part of a HOT-chain with root at item j, then we set |
| 1772 |
* root_offsets[k - 1] = j. |
| 1773 |
* |
| 1774 |
* The passed-in root_offsets array must have MaxHeapTuplesPerPage entries. |
| 1775 |
* Unused entries are filled with InvalidOffsetNumber (zero). |
| 1776 |
* |
| 1777 |
* The function must be called with at least share lock on the buffer, to |
| 1778 |
* prevent concurrent prune operations. |
| 1779 |
* |
| 1780 |
* Note: The information collected here is valid only as long as the caller |
| 1781 |
* holds a pin on the buffer. Once pin is released, a tuple might be pruned |
| 1782 |
* and reused by a completely unrelated tuple. |
| 1783 |
*/ |
| 1784 |
void |
| 1785 |
heap_get_root_tuples(Page page, OffsetNumber *root_offsets) |
| 1786 |
{ |
| 1787 |
OffsetNumber offnum, |
| 1788 |
maxoff; |
| 1789 |
|
| 1790 |
MemSet(root_offsets, InvalidOffsetNumber, |
| 1791 |
MaxHeapTuplesPerPage * sizeof(OffsetNumber)); |
| 1792 |
|
| 1793 |
maxoff = PageGetMaxOffsetNumber(page); |
| 1794 |
for (offnum = FirstOffsetNumber; offnum <= maxoff; offnum = OffsetNumberNext(offnum)) |
| 1795 |
{ |
| 1796 |
ItemId lp = PageGetItemId(page, offnum); |
| 1797 |
HeapTupleHeader htup; |
| 1798 |
OffsetNumber nextoffnum; |
| 1799 |
TransactionId priorXmax; |
| 1800 |
|
| 1801 |
/* skip unused and dead items */ |
| 1802 |
if (!ItemIdIsUsed(lp) || ItemIdIsDead(lp)) |
| 1803 |
continue; |
| 1804 |
|
| 1805 |
if (ItemIdIsNormal(lp)) |
| 1806 |
{ |
| 1807 |
htup = (HeapTupleHeader) PageGetItem(page, lp); |
| 1808 |
|
| 1809 |
/* |
| 1810 |
* Check if this tuple is part of a HOT-chain rooted at some other |
| 1811 |
* tuple. If so, skip it for now; we'll process it when we find |
| 1812 |
* its root. |
| 1813 |
*/ |
| 1814 |
if (HeapTupleHeaderIsHeapOnly(htup)) |
| 1815 |
continue; |
| 1816 |
|
| 1817 |
/* |
| 1818 |
* This is either a plain tuple or the root of a HOT-chain. |
| 1819 |
* Remember it in the mapping. |
| 1820 |
*/ |
| 1821 |
root_offsets[offnum - 1] = offnum; |
| 1822 |
|
| 1823 |
/* If it's not the start of a HOT-chain, we're done with it */ |
| 1824 |
if (!HeapTupleHeaderIsHotUpdated(htup)) |
| 1825 |
continue; |
| 1826 |
|
| 1827 |
/* Set up to scan the HOT-chain */ |
| 1828 |
nextoffnum = ItemPointerGetOffsetNumber(&htup->t_ctid); |
| 1829 |
priorXmax = HeapTupleHeaderGetUpdateXid(htup); |
| 1830 |
} |
| 1831 |
else |
| 1832 |
{ |
| 1833 |
/* Must be a redirect item. We do not set its root_offsets entry */ |
| 1834 |
Assert(ItemIdIsRedirected(lp)); |
| 1835 |
/* Set up to scan the HOT-chain */ |
| 1836 |
nextoffnum = ItemIdGetRedirect(lp); |
| 1837 |
priorXmax = InvalidTransactionId; |
| 1838 |
} |
| 1839 |
|
| 1840 |
/* |
| 1841 |
* Now follow the HOT-chain and collect other tuples in the chain. |
| 1842 |
* |
| 1843 |
* Note: Even though this is a nested loop, the complexity of the |
| 1844 |
* function is O(N) because a tuple in the page should be visited not |
| 1845 |
* more than twice, once in the outer loop and once in HOT-chain |
| 1846 |
* chases. |
| 1847 |
*/ |
| 1848 |
for (;;) |
| 1849 |
{ |
| 1850 |
/* Sanity check (pure paranoia) */ |
| 1851 |
if (nextoffnum < FirstOffsetNumber) |
| 1852 |
break; |
| 1853 |
|
| 1854 |
/* |
| 1855 |
* An offset past the end of page's line pointer array is possible |
| 1856 |
* when the array was truncated |
| 1857 |
*/ |
| 1858 |
if (nextoffnum > maxoff) |
| 1859 |
break; |
| 1860 |
|
| 1861 |
lp = PageGetItemId(page, nextoffnum); |
| 1862 |
|
| 1863 |
/* Check for broken chains */ |
| 1864 |
if (!ItemIdIsNormal(lp)) |
| 1865 |
break; |
| 1866 |
|
| 1867 |
htup = (HeapTupleHeader) PageGetItem(page, lp); |
| 1868 |
|
| 1869 |
if (TransactionIdIsValid(priorXmax) && |
| 1870 |
!TransactionIdEquals(priorXmax, HeapTupleHeaderGetXmin(htup))) |
| 1871 |
break; |
| 1872 |
|
| 1873 |
/* Remember the root line pointer for this item */ |
| 1874 |
root_offsets[nextoffnum - 1] = offnum; |
| 1875 |
|
| 1876 |
/* Advance to next chain member, if any */ |
| 1877 |
if (!HeapTupleHeaderIsHotUpdated(htup)) |
| 1878 |
break; |
| 1879 |
|
| 1880 |
/* HOT implies it can't have moved to different partition */ |
| 1881 |
Assert(!HeapTupleHeaderIndicatesMovedPartitions(htup)); |
| 1882 |
|
| 1883 |
nextoffnum = ItemPointerGetOffsetNumber(&htup->t_ctid); |
| 1884 |
priorXmax = HeapTupleHeaderGetUpdateXid(htup); |
| 1885 |
} |
| 1886 |
} |
| 1887 |
} |
| 1888 |
|
| 1889 |
|
| 1890 |
/* |
| 1891 |
* Compare fields that describe actions required to freeze tuple with caller's |
| 1892 |
* open plan. If everything matches then the frz tuple plan is equivalent to |
| 1893 |
* caller's plan. |
| 1894 |
*/ |
| 1895 |
static inline bool |
| 1896 |
heap_log_freeze_eq(xlhp_freeze_plan *plan, HeapTupleFreeze *frz) |
| 1897 |
{ |
| 1898 |
if (plan->xmax == frz->xmax && |
| 1899 |
plan->t_infomask2 == frz->t_infomask2 && |
| 1900 |
plan->t_infomask == frz->t_infomask && |
| 1901 |
plan->frzflags == frz->frzflags) |
| 1902 |
return true; |
| 1903 |
|
| 1904 |
/* Caller must call heap_log_freeze_new_plan again for frz */ |
| 1905 |
return false; |
| 1906 |
} |
| 1907 |
|
| 1908 |
/* |
| 1909 |
* Comparator used to deduplicate the freeze plans used in WAL records. |
| 1910 |
*/ |
| 1911 |
static int |
| 1912 |
heap_log_freeze_cmp(const void *arg1, const void *arg2) |
| 1913 |
{ |
| 1914 |
HeapTupleFreeze *frz1 = (HeapTupleFreeze *) arg1; |
| 1915 |
HeapTupleFreeze *frz2 = (HeapTupleFreeze *) arg2; |
| 1916 |
|
| 1917 |
if (frz1->xmax < frz2->xmax) |
| 1918 |
return -1; |
| 1919 |
else if (frz1->xmax > frz2->xmax) |
| 1920 |
return 1; |
| 1921 |
|
| 1922 |
if (frz1->t_infomask2 < frz2->t_infomask2) |
| 1923 |
return -1; |
| 1924 |
else if (frz1->t_infomask2 > frz2->t_infomask2) |
| 1925 |
return 1; |
| 1926 |
|
| 1927 |
if (frz1->t_infomask < frz2->t_infomask) |
| 1928 |
return -1; |
| 1929 |
else if (frz1->t_infomask > frz2->t_infomask) |
| 1930 |
return 1; |
| 1931 |
|
| 1932 |
if (frz1->frzflags < frz2->frzflags) |
| 1933 |
return -1; |
| 1934 |
else if (frz1->frzflags > frz2->frzflags) |
| 1935 |
return 1; |
| 1936 |
|
| 1937 |
/* |
| 1938 |
* heap_log_freeze_eq would consider these tuple-wise plans to be equal. |
| 1939 |
* (So the tuples will share a single canonical freeze plan.) |
| 1940 |
* |
| 1941 |
* We tiebreak on page offset number to keep each freeze plan's page |
| 1942 |
* offset number array individually sorted. (Unnecessary, but be tidy.) |
| 1943 |
*/ |
| 1944 |
if (frz1->offset < frz2->offset) |
| 1945 |
return -1; |
| 1946 |
else if (frz1->offset > frz2->offset) |
| 1947 |
return 1; |
| 1948 |
|
| 1949 |
Assert(false); |
| 1950 |
return 0; |
| 1951 |
} |
| 1952 |
|
| 1953 |
/* |
| 1954 |
* Start new plan initialized using tuple-level actions. At least one tuple |
| 1955 |
* will have steps required to freeze described by caller's plan during REDO. |
| 1956 |
*/ |
| 1957 |
static inline void |
| 1958 |
heap_log_freeze_new_plan(xlhp_freeze_plan *plan, HeapTupleFreeze *frz) |
| 1959 |
{ |
| 1960 |
plan->xmax = frz->xmax; |
| 1961 |
plan->t_infomask2 = frz->t_infomask2; |
| 1962 |
plan->t_infomask = frz->t_infomask; |
| 1963 |
plan->frzflags = frz->frzflags; |
| 1964 |
plan->ntuples = 1; /* for now */ |
| 1965 |
} |
| 1966 |
|
| 1967 |
/* |
| 1968 |
* Deduplicate tuple-based freeze plans so that each distinct set of |
| 1969 |
* processing steps is only stored once in the WAL record. |
| 1970 |
* Called during original execution of freezing (for logged relations). |
| 1971 |
* |
| 1972 |
* Return value is number of plans set in *plans_out for caller. Also writes |
| 1973 |
* an array of offset numbers into *offsets_out output argument for caller |
| 1974 |
* (actually there is one array per freeze plan, but that's not of immediate |
| 1975 |
* concern to our caller). |
| 1976 |
*/ |
| 1977 |
static int |
| 1978 |
heap_log_freeze_plan(HeapTupleFreeze *tuples, int ntuples, |
| 1979 |
xlhp_freeze_plan *plans_out, |
| 1980 |
OffsetNumber *offsets_out) |
| 1981 |
{ |
| 1982 |
int nplans = 0; |
| 1983 |
|
| 1984 |
/* Sort tuple-based freeze plans in the order required to deduplicate */ |
| 1985 |
qsort(tuples, ntuples, sizeof(HeapTupleFreeze), heap_log_freeze_cmp); |
| 1986 |
|
| 1987 |
for (int i = 0; i < ntuples; i++) |
| 1988 |
{ |
| 1989 |
HeapTupleFreeze *frz = tuples + i; |
| 1990 |
|
| 1991 |
if (i == 0) |
| 1992 |
{ |
| 1993 |
/* New canonical freeze plan starting with first tup */ |
| 1994 |
heap_log_freeze_new_plan(plans_out, frz); |
| 1995 |
nplans++; |
| 1996 |
} |
| 1997 |
else if (heap_log_freeze_eq(plans_out, frz)) |
| 1998 |
{ |
| 1999 |
/* tup matches open canonical plan -- include tup in it */ |
| 2000 |
Assert(offsets_out[i - 1] < frz->offset); |
| 2001 |
plans_out->ntuples++; |
| 2002 |
} |
| 2003 |
else |
| 2004 |
{ |
| 2005 |
/* Tup doesn't match current plan -- done with it now */ |
| 2006 |
plans_out++; |
| 2007 |
|
| 2008 |
/* New canonical freeze plan starting with this tup */ |
| 2009 |
heap_log_freeze_new_plan(plans_out, frz); |
| 2010 |
nplans++; |
| 2011 |
} |
| 2012 |
|
| 2013 |
/* |
| 2014 |
* Save page offset number in dedicated buffer in passing. |
| 2015 |
* |
| 2016 |
* REDO routine relies on the record's offset numbers array grouping |
| 2017 |
* offset numbers by freeze plan. The sort order within each grouping |
| 2018 |
* is ascending offset number order, just to keep things tidy. |
| 2019 |
*/ |
| 2020 |
offsets_out[i] = frz->offset; |
| 2021 |
} |
| 2022 |
|
| 2023 |
Assert(nplans > 0 && nplans <= ntuples); |
| 2024 |
|
| 2025 |
return nplans; |
| 2026 |
} |
| 2027 |
|
| 2028 |
/* |
| 2029 |
* Write an XLOG_HEAP2_PRUNE_FREEZE WAL record |
| 2030 |
* |
| 2031 |
* This is used for several different page maintenance operations: |
| 2032 |
* |
| 2033 |
* - Page pruning, in VACUUM's 1st pass or on access: Some items are |
| 2034 |
* redirected, some marked dead, and some removed altogether. |
| 2035 |
* |
| 2036 |
* - Freezing: Items are marked as 'frozen'. |
| 2037 |
* |
| 2038 |
* - Vacuum, 2nd pass: Items that are already LP_DEAD are marked as unused. |
| 2039 |
* |
| 2040 |
* They have enough commonalities that we use a single WAL record for them |
| 2041 |
* all. |
| 2042 |
* |
| 2043 |
* If replaying the record requires a cleanup lock, pass cleanup_lock = true. |
| 2044 |
* Replaying 'redirected' or 'dead' items always requires a cleanup lock, but |
| 2045 |
* replaying 'unused' items depends on whether they were all previously marked |
| 2046 |
* as dead. |
| 2047 |
* |
| 2048 |
* Note: This function scribbles on the 'frozen' array. |
| 2049 |
* |
| 2050 |
* Note: This is called in a critical section, so careful what you do here. |
| 2051 |
*/ |
| 2052 |
void |
| 2053 |
log_heap_prune_and_freeze(Relation relation, Buffer buffer, |
| 2054 |
TransactionId conflict_xid, |
| 2055 |
bool cleanup_lock, |
| 2056 |
PruneReason reason, |
| 2057 |
HeapTupleFreeze *frozen, int nfrozen, |
| 2058 |
OffsetNumber *redirected, int nredirected, |
| 2059 |
OffsetNumber *dead, int ndead, |
| 2060 |
OffsetNumber *unused, int nunused) |
| 2061 |
{ |
| 2062 |
xl_heap_prune xlrec; |
| 2063 |
XLogRecPtr recptr; |
| 2064 |
uint8 info; |
| 2065 |
|
| 2066 |
/* The following local variables hold data registered in the WAL record: */ |
| 2067 |
xlhp_freeze_plan plans[MaxHeapTuplesPerPage]; |
| 2068 |
xlhp_freeze_plans freeze_plans; |
| 2069 |
xlhp_prune_items redirect_items; |
| 2070 |
xlhp_prune_items dead_items; |
| 2071 |
xlhp_prune_items unused_items; |
| 2072 |
OffsetNumber frz_offsets[MaxHeapTuplesPerPage]; |
| 2073 |
|
| 2074 |
xlrec.flags = 0; |
| 2075 |
|
| 2076 |
/* |
| 2077 |
* Prepare data for the buffer. The arrays are not actually in the |
| 2078 |
* buffer, but we pretend that they are. When XLogInsert stores a full |
| 2079 |
* page image, the arrays can be omitted. |
| 2080 |
*/ |
| 2081 |
XLogBeginInsert(); |
| 2082 |
XLogRegisterBuffer(0, buffer, REGBUF_STANDARD); |
| 2083 |
if (nfrozen > 0) |
| 2084 |
{ |
| 2085 |
int nplans; |
| 2086 |
|
| 2087 |
xlrec.flags |= XLHP_HAS_FREEZE_PLANS; |
| 2088 |
|
| 2089 |
/* |
| 2090 |
* Prepare deduplicated representation for use in the WAL record. This |
| 2091 |
* destructively sorts frozen tuples array in-place. |
| 2092 |
*/ |
| 2093 |
nplans = heap_log_freeze_plan(frozen, nfrozen, plans, frz_offsets); |
| 2094 |
|
| 2095 |
freeze_plans.nplans = nplans; |
| 2096 |
XLogRegisterBufData(0, &freeze_plans, |
| 2097 |
offsetof(xlhp_freeze_plans, plans)); |
| 2098 |
XLogRegisterBufData(0, plans, |
| 2099 |
sizeof(xlhp_freeze_plan) * nplans); |
| 2100 |
} |
| 2101 |
if (nredirected > 0) |
| 2102 |
{ |
| 2103 |
xlrec.flags |= XLHP_HAS_REDIRECTIONS; |
| 2104 |
|
| 2105 |
redirect_items.ntargets = nredirected; |
| 2106 |
XLogRegisterBufData(0, &redirect_items, |
| 2107 |
offsetof(xlhp_prune_items, data)); |
| 2108 |
XLogRegisterBufData(0, redirected, |
| 2109 |
sizeof(OffsetNumber[2]) * nredirected); |
| 2110 |
} |
| 2111 |
if (ndead > 0) |
| 2112 |
{ |
| 2113 |
xlrec.flags |= XLHP_HAS_DEAD_ITEMS; |
| 2114 |
|
| 2115 |
dead_items.ntargets = ndead; |
| 2116 |
XLogRegisterBufData(0, &dead_items, |
| 2117 |
offsetof(xlhp_prune_items, data)); |
| 2118 |
XLogRegisterBufData(0, dead, |
| 2119 |
sizeof(OffsetNumber) * ndead); |
| 2120 |
} |
| 2121 |
if (nunused > 0) |
| 2122 |
{ |
| 2123 |
xlrec.flags |= XLHP_HAS_NOW_UNUSED_ITEMS; |
| 2124 |
|
| 2125 |
unused_items.ntargets = nunused; |
| 2126 |
XLogRegisterBufData(0, &unused_items, |
| 2127 |
offsetof(xlhp_prune_items, data)); |
| 2128 |
XLogRegisterBufData(0, unused, |
| 2129 |
sizeof(OffsetNumber) * nunused); |
| 2130 |
} |
| 2131 |
if (nfrozen > 0) |
| 2132 |
XLogRegisterBufData(0, frz_offsets, |
| 2133 |
sizeof(OffsetNumber) * nfrozen); |
| 2134 |
|
| 2135 |
/* |
| 2136 |
* Prepare the main xl_heap_prune record. We already set the XLHP_HAS_* |
| 2137 |
* flag above. |
| 2138 |
*/ |
| 2139 |
if (RelationIsAccessibleInLogicalDecoding(relation)) |
| 2140 |
xlrec.flags |= XLHP_IS_CATALOG_REL; |
| 2141 |
if (TransactionIdIsValid(conflict_xid)) |
| 2142 |
xlrec.flags |= XLHP_HAS_CONFLICT_HORIZON; |
| 2143 |
if (cleanup_lock) |
| 2144 |
xlrec.flags |= XLHP_CLEANUP_LOCK; |
| 2145 |
else |
| 2146 |
{ |
| 2147 |
Assert(nredirected == 0 && ndead == 0); |
| 2148 |
/* also, any items in 'unused' must've been LP_DEAD previously */ |
| 2149 |
} |
| 2150 |
XLogRegisterData(&xlrec, SizeOfHeapPrune); |
| 2151 |
if (TransactionIdIsValid(conflict_xid)) |
| 2152 |
XLogRegisterData(&conflict_xid, sizeof(TransactionId)); |
| 2153 |
|
| 2154 |
switch (reason) |
| 2155 |
{ |
| 2156 |
case PRUNE_ON_ACCESS: |
| 2157 |
info = XLOG_HEAP2_PRUNE_ON_ACCESS; |
| 2158 |
break; |
| 2159 |
case PRUNE_VACUUM_SCAN: |
| 2160 |
info = XLOG_HEAP2_PRUNE_VACUUM_SCAN; |
| 2161 |
break; |
| 2162 |
case PRUNE_VACUUM_CLEANUP: |
| 2163 |
info = XLOG_HEAP2_PRUNE_VACUUM_CLEANUP; |
| 2164 |
break; |
| 2165 |
default: |
| 2166 |
elog(ERROR, "unrecognized prune reason: %d", (int) reason); |
| 2167 |
break; |
| 2168 |
} |
| 2169 |
recptr = XLogInsert(RM_HEAP2_ID, info); |
| 2170 |
|
| 2171 |
PageSetLSN(BufferGetPage(buffer), recptr); |
| 2172 |
} |