| 1 |
/*------------------------------------------------------------------------- |
| 2 |
* |
| 3 |
* combocid.c |
| 4 |
* Combo command ID support routines |
| 5 |
* |
| 6 |
* Before version 8.3, HeapTupleHeaderData had separate fields for cmin |
| 7 |
* and cmax. To reduce the header size, cmin and cmax are now overlaid |
| 8 |
* in the same field in the header. That usually works because you rarely |
| 9 |
* insert and delete a tuple in the same transaction, and we don't need |
| 10 |
* either field to remain valid after the originating transaction exits. |
| 11 |
* To make it work when the inserting transaction does delete the tuple, |
| 12 |
* we create a "combo" command ID and store that in the tuple header |
| 13 |
* instead of cmin and cmax. The combo command ID can be mapped to the |
| 14 |
* real cmin and cmax using a backend-private array, which is managed by |
| 15 |
* this module. |
| 16 |
* |
| 17 |
* To allow reusing existing combo CIDs, we also keep a hash table that |
| 18 |
* maps cmin,cmax pairs to combo CIDs. This keeps the data structure size |
| 19 |
* reasonable in most cases, since the number of unique pairs used by any |
| 20 |
* one transaction is likely to be small. |
| 21 |
* |
| 22 |
* With a 32-bit combo command id we can represent 2^32 distinct cmin,cmax |
| 23 |
* combinations. In the most perverse case where each command deletes a tuple |
| 24 |
* generated by every previous command, the number of combo command ids |
| 25 |
* required for N commands is N*(N+1)/2. That means that in the worst case, |
| 26 |
* that's enough for 92682 commands. In practice, you'll run out of memory |
| 27 |
* and/or disk space way before you reach that limit. |
| 28 |
* |
| 29 |
* The array and hash table are kept in TopTransactionContext, and are |
| 30 |
* destroyed at the end of each transaction. |
| 31 |
* |
| 32 |
* |
| 33 |
* Portions Copyright (c) 1996-2025, PostgreSQL Global Development Group |
| 34 |
* Portions Copyright (c) 1994, Regents of the University of California |
| 35 |
* |
| 36 |
* IDENTIFICATION |
| 37 |
* src/backend/utils/time/combocid.c |
| 38 |
* |
| 39 |
*------------------------------------------------------------------------- |
| 40 |
*/ |
| 41 |
|
| 42 |
#include "postgres.h" |
| 43 |
|
| 44 |
#include "access/htup_details.h" |
| 45 |
#include "access/xact.h" |
| 46 |
#include "miscadmin.h" |
| 47 |
#include "storage/shmem.h" |
| 48 |
#include "utils/combocid.h" |
| 49 |
#include "utils/hsearch.h" |
| 50 |
#include "utils/memutils.h" |
| 51 |
|
| 52 |
/* Hash table to lookup combo CIDs by cmin and cmax */ |
| 53 |
static HTAB *comboHash = NULL; |
| 54 |
|
| 55 |
/* Key and entry structures for the hash table */ |
| 56 |
typedef struct |
| 57 |
{ |
| 58 |
CommandId cmin; |
| 59 |
CommandId cmax; |
| 60 |
} ComboCidKeyData; |
| 61 |
|
| 62 |
typedef ComboCidKeyData *ComboCidKey; |
| 63 |
|
| 64 |
typedef struct |
| 65 |
{ |
| 66 |
ComboCidKeyData key; |
| 67 |
CommandId combocid; |
| 68 |
} ComboCidEntryData; |
| 69 |
|
| 70 |
typedef ComboCidEntryData *ComboCidEntry; |
| 71 |
|
| 72 |
/* Initial size of the hash table */ |
| 73 |
#define CCID_HASH_SIZE 100 |
| 74 |
|
| 75 |
|
| 76 |
/* |
| 77 |
* An array of cmin,cmax pairs, indexed by combo command id. |
| 78 |
* To convert a combo CID to cmin and cmax, you do a simple array lookup. |
| 79 |
*/ |
| 80 |
static ComboCidKey comboCids = NULL; |
| 81 |
static int usedComboCids = 0; /* number of elements in comboCids */ |
| 82 |
static int sizeComboCids = 0; /* allocated size of array */ |
| 83 |
|
| 84 |
/* Initial size of the array */ |
| 85 |
#define CCID_ARRAY_SIZE 100 |
| 86 |
|
| 87 |
|
| 88 |
/* prototypes for internal functions */ |
| 89 |
static CommandId GetComboCommandId(CommandId cmin, CommandId cmax); |
| 90 |
static CommandId GetRealCmin(CommandId combocid); |
| 91 |
static CommandId GetRealCmax(CommandId combocid); |
| 92 |
|
| 93 |
|
| 94 |
/**** External API ****/ |
| 95 |
|
| 96 |
/* |
| 97 |
* GetCmin and GetCmax assert that they are only called in situations where |
| 98 |
* they make sense, that is, can deliver a useful answer. If you have |
| 99 |
* reason to examine a tuple's t_cid field from a transaction other than |
| 100 |
* the originating one, use HeapTupleHeaderGetRawCommandId() directly. |
| 101 |
*/ |
| 102 |
|
| 103 |
CommandId |
| 104 |
HeapTupleHeaderGetCmin(const HeapTupleHeaderData *tup) |
| 105 |
{ |
| 106 |
CommandId cid = HeapTupleHeaderGetRawCommandId(tup); |
| 107 |
|
| 108 |
Assert(!(tup->t_infomask & HEAP_MOVED)); |
| 109 |
Assert(TransactionIdIsCurrentTransactionId(HeapTupleHeaderGetXmin(tup))); |
| 110 |
|
| 111 |
if (tup->t_infomask & HEAP_COMBOCID) |
| 112 |
return GetRealCmin(cid); |
| 113 |
else |
| 114 |
return cid; |
| 115 |
} |
| 116 |
|
| 117 |
CommandId |
| 118 |
HeapTupleHeaderGetCmax(const HeapTupleHeaderData *tup) |
| 119 |
{ |
| 120 |
CommandId cid = HeapTupleHeaderGetRawCommandId(tup); |
| 121 |
|
| 122 |
Assert(!(tup->t_infomask & HEAP_MOVED)); |
| 123 |
|
| 124 |
/* |
| 125 |
* Because GetUpdateXid() performs memory allocations if xmax is a |
| 126 |
* multixact we can't Assert() if we're inside a critical section. This |
| 127 |
* weakens the check, but not using GetCmax() inside one would complicate |
| 128 |
* things too much. |
| 129 |
*/ |
| 130 |
Assert(CritSectionCount > 0 || |
| 131 |
TransactionIdIsCurrentTransactionId(HeapTupleHeaderGetUpdateXid(tup))); |
| 132 |
|
| 133 |
if (tup->t_infomask & HEAP_COMBOCID) |
| 134 |
return GetRealCmax(cid); |
| 135 |
else |
| 136 |
return cid; |
| 137 |
} |
| 138 |
|
| 139 |
/* |
| 140 |
* Given a tuple we are about to delete, determine the correct value to store |
| 141 |
* into its t_cid field. |
| 142 |
* |
| 143 |
* If we don't need a combo CID, *cmax is unchanged and *iscombo is set to |
| 144 |
* false. If we do need one, *cmax is replaced by a combo CID and *iscombo |
| 145 |
* is set to true. |
| 146 |
* |
| 147 |
* The reason this is separate from the actual HeapTupleHeaderSetCmax() |
| 148 |
* operation is that this could fail due to out-of-memory conditions. Hence |
| 149 |
* we need to do this before entering the critical section that actually |
| 150 |
* changes the tuple in shared buffers. |
| 151 |
*/ |
| 152 |
void |
| 153 |
HeapTupleHeaderAdjustCmax(const HeapTupleHeaderData *tup, |
| 154 |
CommandId *cmax, |
| 155 |
bool *iscombo) |
| 156 |
{ |
| 157 |
/* |
| 158 |
* If we're marking a tuple deleted that was inserted by (any |
| 159 |
* subtransaction of) our transaction, we need to use a combo command id. |
| 160 |
* Test for HeapTupleHeaderXminCommitted() first, because it's cheaper |
| 161 |
* than a TransactionIdIsCurrentTransactionId call. |
| 162 |
*/ |
| 163 |
if (!HeapTupleHeaderXminCommitted(tup) && |
| 164 |
TransactionIdIsCurrentTransactionId(HeapTupleHeaderGetRawXmin(tup))) |
| 165 |
{ |
| 166 |
CommandId cmin = HeapTupleHeaderGetCmin(tup); |
| 167 |
|
| 168 |
*cmax = GetComboCommandId(cmin, *cmax); |
| 169 |
*iscombo = true; |
| 170 |
} |
| 171 |
else |
| 172 |
{ |
| 173 |
*iscombo = false; |
| 174 |
} |
| 175 |
} |
| 176 |
|
| 177 |
/* |
| 178 |
* Combo command ids are only interesting to the inserting and deleting |
| 179 |
* transaction, so we can forget about them at the end of transaction. |
| 180 |
*/ |
| 181 |
void |
| 182 |
AtEOXact_ComboCid(void) |
| 183 |
{ |
| 184 |
/* |
| 185 |
* Don't bother to pfree. These are allocated in TopTransactionContext, so |
| 186 |
* they're going to go away at the end of transaction anyway. |
| 187 |
*/ |
| 188 |
comboHash = NULL; |
| 189 |
|
| 190 |
comboCids = NULL; |
| 191 |
usedComboCids = 0; |
| 192 |
sizeComboCids = 0; |
| 193 |
} |
| 194 |
|
| 195 |
|
| 196 |
/**** Internal routines ****/ |
| 197 |
|
| 198 |
/* |
| 199 |
* Get a combo command id that maps to cmin and cmax. |
| 200 |
* |
| 201 |
* We try to reuse old combo command ids when possible. |
| 202 |
*/ |
| 203 |
static CommandId |
| 204 |
GetComboCommandId(CommandId cmin, CommandId cmax) |
| 205 |
{ |
| 206 |
CommandId combocid; |
| 207 |
ComboCidKeyData key; |
| 208 |
ComboCidEntry entry; |
| 209 |
bool found; |
| 210 |
|
| 211 |
/* |
| 212 |
* Create the hash table and array the first time we need to use combo |
| 213 |
* cids in the transaction. |
| 214 |
*/ |
| 215 |
if (comboHash == NULL) |
| 216 |
{ |
| 217 |
HASHCTL hash_ctl; |
| 218 |
|
| 219 |
/* Make array first; existence of hash table asserts array exists */ |
| 220 |
comboCids = (ComboCidKeyData *) |
| 221 |
MemoryContextAlloc(TopTransactionContext, |
| 222 |
sizeof(ComboCidKeyData) * CCID_ARRAY_SIZE); |
| 223 |
sizeComboCids = CCID_ARRAY_SIZE; |
| 224 |
usedComboCids = 0; |
| 225 |
|
| 226 |
hash_ctl.keysize = sizeof(ComboCidKeyData); |
| 227 |
hash_ctl.entrysize = sizeof(ComboCidEntryData); |
| 228 |
hash_ctl.hcxt = TopTransactionContext; |
| 229 |
|
| 230 |
comboHash = hash_create("Combo CIDs", |
| 231 |
CCID_HASH_SIZE, |
| 232 |
&hash_ctl, |
| 233 |
HASH_ELEM | HASH_BLOBS | HASH_CONTEXT); |
| 234 |
} |
| 235 |
|
| 236 |
/* |
| 237 |
* Grow the array if there's not at least one free slot. We must do this |
| 238 |
* before possibly entering a new hashtable entry, else failure to |
| 239 |
* repalloc would leave a corrupt hashtable entry behind. |
| 240 |
*/ |
| 241 |
if (usedComboCids >= sizeComboCids) |
| 242 |
{ |
| 243 |
int newsize = sizeComboCids * 2; |
| 244 |
|
| 245 |
comboCids = (ComboCidKeyData *) |
| 246 |
repalloc(comboCids, sizeof(ComboCidKeyData) * newsize); |
| 247 |
sizeComboCids = newsize; |
| 248 |
} |
| 249 |
|
| 250 |
/* Lookup or create a hash entry with the desired cmin/cmax */ |
| 251 |
|
| 252 |
/* We assume there is no struct padding in ComboCidKeyData! */ |
| 253 |
key.cmin = cmin; |
| 254 |
key.cmax = cmax; |
| 255 |
entry = (ComboCidEntry) hash_search(comboHash, |
| 256 |
&key, |
| 257 |
HASH_ENTER, |
| 258 |
&found); |
| 259 |
|
| 260 |
if (found) |
| 261 |
{ |
| 262 |
/* Reuse an existing combo CID */ |
| 263 |
return entry->combocid; |
| 264 |
} |
| 265 |
|
| 266 |
/* We have to create a new combo CID; we already made room in the array */ |
| 267 |
combocid = usedComboCids; |
| 268 |
|
| 269 |
comboCids[combocid].cmin = cmin; |
| 270 |
comboCids[combocid].cmax = cmax; |
| 271 |
usedComboCids++; |
| 272 |
|
| 273 |
entry->combocid = combocid; |
| 274 |
|
| 275 |
return combocid; |
| 276 |
} |
| 277 |
|
| 278 |
static CommandId |
| 279 |
GetRealCmin(CommandId combocid) |
| 280 |
{ |
| 281 |
Assert(combocid < usedComboCids); |
| 282 |
return comboCids[combocid].cmin; |
| 283 |
} |
| 284 |
|
| 285 |
static CommandId |
| 286 |
GetRealCmax(CommandId combocid) |
| 287 |
{ |
| 288 |
Assert(combocid < usedComboCids); |
| 289 |
return comboCids[combocid].cmax; |
| 290 |
} |
| 291 |
|
| 292 |
/* |
| 293 |
* Estimate the amount of space required to serialize the current combo CID |
| 294 |
* state. |
| 295 |
*/ |
| 296 |
Size |
| 297 |
EstimateComboCIDStateSpace(void) |
| 298 |
{ |
| 299 |
Size size; |
| 300 |
|
| 301 |
/* Add space required for saving usedComboCids */ |
| 302 |
size = sizeof(int); |
| 303 |
|
| 304 |
/* Add space required for saving ComboCidKeyData */ |
| 305 |
size = add_size(size, mul_size(sizeof(ComboCidKeyData), usedComboCids)); |
| 306 |
|
| 307 |
return size; |
| 308 |
} |
| 309 |
|
| 310 |
/* |
| 311 |
* Serialize the combo CID state into the memory, beginning at start_address. |
| 312 |
* maxsize should be at least as large as the value returned by |
| 313 |
* EstimateComboCIDStateSpace. |
| 314 |
*/ |
| 315 |
void |
| 316 |
SerializeComboCIDState(Size maxsize, char *start_address) |
| 317 |
{ |
| 318 |
char *endptr; |
| 319 |
|
| 320 |
/* First, we store the number of currently-existing combo CIDs. */ |
| 321 |
*(int *) start_address = usedComboCids; |
| 322 |
|
| 323 |
/* If maxsize is too small, throw an error. */ |
| 324 |
endptr = start_address + sizeof(int) + |
| 325 |
(sizeof(ComboCidKeyData) * usedComboCids); |
| 326 |
if (endptr < start_address || endptr > start_address + maxsize) |
| 327 |
elog(ERROR, "not enough space to serialize ComboCID state"); |
| 328 |
|
| 329 |
/* Now, copy the actual cmin/cmax pairs. */ |
| 330 |
if (usedComboCids > 0) |
| 331 |
memcpy(start_address + sizeof(int), comboCids, |
| 332 |
(sizeof(ComboCidKeyData) * usedComboCids)); |
| 333 |
} |
| 334 |
|
| 335 |
/* |
| 336 |
* Read the combo CID state at the specified address and initialize this |
| 337 |
* backend with the same combo CIDs. This is only valid in a backend that |
| 338 |
* currently has no combo CIDs (and only makes sense if the transaction state |
| 339 |
* is serialized and restored as well). |
| 340 |
*/ |
| 341 |
void |
| 342 |
RestoreComboCIDState(char *comboCIDstate) |
| 343 |
{ |
| 344 |
int num_elements; |
| 345 |
ComboCidKeyData *keydata; |
| 346 |
int i; |
| 347 |
CommandId cid; |
| 348 |
|
| 349 |
Assert(!comboCids && !comboHash); |
| 350 |
|
| 351 |
/* First, we retrieve the number of combo CIDs that were serialized. */ |
| 352 |
num_elements = *(int *) comboCIDstate; |
| 353 |
keydata = (ComboCidKeyData *) (comboCIDstate + sizeof(int)); |
| 354 |
|
| 355 |
/* Use GetComboCommandId to restore each combo CID. */ |
| 356 |
for (i = 0; i < num_elements; i++) |
| 357 |
{ |
| 358 |
cid = GetComboCommandId(keydata[i].cmin, keydata[i].cmax); |
| 359 |
|
| 360 |
/* Verify that we got the expected answer. */ |
| 361 |
if (cid != i) |
| 362 |
elog(ERROR, "unexpected command ID while restoring combo CIDs"); |
| 363 |
} |
| 364 |
} |