postgresql/src/backend/utils/cache/syscache.c

Ignoring revisions in .git-blame-ignore-revs. Click here to bypass and see the normal blame view.

682 lines
18 KiB
C
Raw Normal View History

/*-------------------------------------------------------------------------
*
* syscache.c
* System cache management routines
*
* Portions Copyright (c) 1996-2024, PostgreSQL Global Development Group
* Portions Copyright (c) 1994, Regents of the University of California
*
*
* IDENTIFICATION
2010-09-20 22:08:53 +02:00
* src/backend/utils/cache/syscache.c
*
* NOTES
* These routines allow the parser/planner/executor to perform
* rapid lookups on the contents of the system catalogs.
*
2011-06-18 23:37:30 +02:00
* see utils/syscache.h for a list of the cache IDs
*
*-------------------------------------------------------------------------
*/
1996-11-03 07:54:38 +01:00
#include "postgres.h"
#include "access/htup_details.h"
#include "access/sysattr.h"
#include "catalog/pg_db_role_setting_d.h"
#include "catalog/pg_depend_d.h"
#include "catalog/pg_description_d.h"
#include "catalog/pg_seclabel_d.h"
#include "catalog/pg_shdepend_d.h"
#include "catalog/pg_shdescription_d.h"
#include "catalog/pg_shseclabel_d.h"
#include "common/int.h"
#include "lib/qunique.h"
#include "utils/catcache.h"
#include "utils/lsyscache.h"
#include "utils/rel.h"
#include "utils/syscache.h"
/*---------------------------------------------------------------------------
Adding system caches:
There must be a unique index underlying each syscache (ie, an index
whose key is the same as that of the cache). If there is not one
already, add the definition for it to include/catalog/pg_*.h using
DECLARE_UNIQUE_INDEX.
(Adding an index requires a catversion.h update, while simply
adding/deleting caches only requires a recompile.)
Add a MAKE_SYSCACHE call to the same pg_*.h file specifying the name of
your cache, the underlying index, and the initial number of hash buckets.
The number of hash buckets must be a power of 2. It's reasonable to
set this to the number of entries that might be in the particular cache
in a medium-size database.
Finally, any place your relation gets heap_insert() or
heap_update() calls, use CatalogTupleInsert() or CatalogTupleUpdate()
instead, which also update indexes. The heap_* calls do not do that.
*---------------------------------------------------------------------------
*/
/*
* struct cachedesc: information defining a single syscache
*/
struct cachedesc
{
Oid reloid; /* OID of the relation being cached */
Oid indoid; /* OID of index relation for this cache */
int nkeys; /* # of keys needed for cache lookup */
int key[4]; /* attribute numbers of key attrs */
int nbuckets; /* number of hash buckets for this cache */
};
/* Macro to provide nkeys and key array with convenient syntax. */
#define KEY(...) VA_ARGS_NARGS(__VA_ARGS__), { __VA_ARGS__ }
#include "catalog/syscache_info.h"
StaticAssertDecl(lengthof(cacheinfo) == SysCacheSize,
"SysCacheSize does not match syscache.c's array");
Use a safer method for determining whether relcache init file is stale. When we invalidate the relcache entry for a system catalog or index, we must also delete the relcache "init file" if the init file contains a copy of that rel's entry. The old way of doing this relied on a specially maintained list of the OIDs of relations present in the init file: we made the list either when reading the file in, or when writing the file out. The problem is that when writing the file out, we included only rels present in our local relcache, which might have already suffered some deletions due to relcache inval events. In such cases we correctly decided not to overwrite the real init file with incomplete data --- but we still used the incomplete initFileRelationIds list for the rest of the current session. This could result in wrong decisions about whether the session's own actions require deletion of the init file, potentially allowing an init file created by some other concurrent session to be left around even though it's been made stale. Since we don't support changing the schema of a system catalog at runtime, the only likely scenario in which this would cause a problem in the field involves a "vacuum full" on a catalog concurrently with other activity, and even then it's far from easy to provoke. Remarkably, this has been broken since 2002 (in commit 786340441706ac1957a031f11ad1c2e5b6e18314), but we had never seen a reproducible test case until recently. If it did happen in the field, the symptoms would probably involve unexpected "cache lookup failed" errors to begin with, then "could not open file" failures after the next checkpoint, as all accesses to the affected catalog stopped working. Recovery would require manually removing the stale "pg_internal.init" file. To fix, get rid of the initFileRelationIds list, and instead consult syscache.c's list of relations used in catalog caches to decide whether a relation is included in the init file. This should be a tad more efficient anyway, since we're replacing linear search of a list with ~100 entries with a binary search. It's a bit ugly that the init file contents are now so directly tied to the catalog caches, but in practice that won't make much difference. Back-patch to all supported branches.
2015-06-07 21:32:09 +02:00
static CatCache *SysCache[SysCacheSize];
2000-02-18 10:30:20 +01:00
static bool CacheInitialized = false;
Use a safer method for determining whether relcache init file is stale. When we invalidate the relcache entry for a system catalog or index, we must also delete the relcache "init file" if the init file contains a copy of that rel's entry. The old way of doing this relied on a specially maintained list of the OIDs of relations present in the init file: we made the list either when reading the file in, or when writing the file out. The problem is that when writing the file out, we included only rels present in our local relcache, which might have already suffered some deletions due to relcache inval events. In such cases we correctly decided not to overwrite the real init file with incomplete data --- but we still used the incomplete initFileRelationIds list for the rest of the current session. This could result in wrong decisions about whether the session's own actions require deletion of the init file, potentially allowing an init file created by some other concurrent session to be left around even though it's been made stale. Since we don't support changing the schema of a system catalog at runtime, the only likely scenario in which this would cause a problem in the field involves a "vacuum full" on a catalog concurrently with other activity, and even then it's far from easy to provoke. Remarkably, this has been broken since 2002 (in commit 786340441706ac1957a031f11ad1c2e5b6e18314), but we had never seen a reproducible test case until recently. If it did happen in the field, the symptoms would probably involve unexpected "cache lookup failed" errors to begin with, then "could not open file" failures after the next checkpoint, as all accesses to the affected catalog stopped working. Recovery would require manually removing the stale "pg_internal.init" file. To fix, get rid of the initFileRelationIds list, and instead consult syscache.c's list of relations used in catalog caches to decide whether a relation is included in the init file. This should be a tad more efficient anyway, since we're replacing linear search of a list with ~100 entries with a binary search. It's a bit ugly that the init file contents are now so directly tied to the catalog caches, but in practice that won't make much difference. Back-patch to all supported branches.
2015-06-07 21:32:09 +02:00
/* Sorted array of OIDs of tables that have caches on them */
static Oid SysCacheRelationOid[SysCacheSize];
static int SysCacheRelationOidSize;
Use a safer method for determining whether relcache init file is stale. When we invalidate the relcache entry for a system catalog or index, we must also delete the relcache "init file" if the init file contains a copy of that rel's entry. The old way of doing this relied on a specially maintained list of the OIDs of relations present in the init file: we made the list either when reading the file in, or when writing the file out. The problem is that when writing the file out, we included only rels present in our local relcache, which might have already suffered some deletions due to relcache inval events. In such cases we correctly decided not to overwrite the real init file with incomplete data --- but we still used the incomplete initFileRelationIds list for the rest of the current session. This could result in wrong decisions about whether the session's own actions require deletion of the init file, potentially allowing an init file created by some other concurrent session to be left around even though it's been made stale. Since we don't support changing the schema of a system catalog at runtime, the only likely scenario in which this would cause a problem in the field involves a "vacuum full" on a catalog concurrently with other activity, and even then it's far from easy to provoke. Remarkably, this has been broken since 2002 (in commit 786340441706ac1957a031f11ad1c2e5b6e18314), but we had never seen a reproducible test case until recently. If it did happen in the field, the symptoms would probably involve unexpected "cache lookup failed" errors to begin with, then "could not open file" failures after the next checkpoint, as all accesses to the affected catalog stopped working. Recovery would require manually removing the stale "pg_internal.init" file. To fix, get rid of the initFileRelationIds list, and instead consult syscache.c's list of relations used in catalog caches to decide whether a relation is included in the init file. This should be a tad more efficient anyway, since we're replacing linear search of a list with ~100 entries with a binary search. It's a bit ugly that the init file contents are now so directly tied to the catalog caches, but in practice that won't make much difference. Back-patch to all supported branches.
2015-06-07 21:32:09 +02:00
/* Sorted array of OIDs of tables and indexes used by caches */
static Oid SysCacheSupportingRelOid[SysCacheSize * 2];
static int SysCacheSupportingRelOidSize;
static int oid_compare(const void *a, const void *b);
Use a safer method for determining whether relcache init file is stale. When we invalidate the relcache entry for a system catalog or index, we must also delete the relcache "init file" if the init file contains a copy of that rel's entry. The old way of doing this relied on a specially maintained list of the OIDs of relations present in the init file: we made the list either when reading the file in, or when writing the file out. The problem is that when writing the file out, we included only rels present in our local relcache, which might have already suffered some deletions due to relcache inval events. In such cases we correctly decided not to overwrite the real init file with incomplete data --- but we still used the incomplete initFileRelationIds list for the rest of the current session. This could result in wrong decisions about whether the session's own actions require deletion of the init file, potentially allowing an init file created by some other concurrent session to be left around even though it's been made stale. Since we don't support changing the schema of a system catalog at runtime, the only likely scenario in which this would cause a problem in the field involves a "vacuum full" on a catalog concurrently with other activity, and even then it's far from easy to provoke. Remarkably, this has been broken since 2002 (in commit 786340441706ac1957a031f11ad1c2e5b6e18314), but we had never seen a reproducible test case until recently. If it did happen in the field, the symptoms would probably involve unexpected "cache lookup failed" errors to begin with, then "could not open file" failures after the next checkpoint, as all accesses to the affected catalog stopped working. Recovery would require manually removing the stale "pg_internal.init" file. To fix, get rid of the initFileRelationIds list, and instead consult syscache.c's list of relations used in catalog caches to decide whether a relation is included in the init file. This should be a tad more efficient anyway, since we're replacing linear search of a list with ~100 entries with a binary search. It's a bit ugly that the init file contents are now so directly tied to the catalog caches, but in practice that won't make much difference. Back-patch to all supported branches.
2015-06-07 21:32:09 +02:00
/*
* InitCatalogCache - initialize the caches
*
* Note that no database access is done here; we only allocate memory
* and initialize the cache structure. Interrogation of the database
* to complete initialization of a cache happens upon first use
* of that cache.
*/
void
InitCatalogCache(void)
{
int cacheId;
Assert(!CacheInitialized);
Use a safer method for determining whether relcache init file is stale. When we invalidate the relcache entry for a system catalog or index, we must also delete the relcache "init file" if the init file contains a copy of that rel's entry. The old way of doing this relied on a specially maintained list of the OIDs of relations present in the init file: we made the list either when reading the file in, or when writing the file out. The problem is that when writing the file out, we included only rels present in our local relcache, which might have already suffered some deletions due to relcache inval events. In such cases we correctly decided not to overwrite the real init file with incomplete data --- but we still used the incomplete initFileRelationIds list for the rest of the current session. This could result in wrong decisions about whether the session's own actions require deletion of the init file, potentially allowing an init file created by some other concurrent session to be left around even though it's been made stale. Since we don't support changing the schema of a system catalog at runtime, the only likely scenario in which this would cause a problem in the field involves a "vacuum full" on a catalog concurrently with other activity, and even then it's far from easy to provoke. Remarkably, this has been broken since 2002 (in commit 786340441706ac1957a031f11ad1c2e5b6e18314), but we had never seen a reproducible test case until recently. If it did happen in the field, the symptoms would probably involve unexpected "cache lookup failed" errors to begin with, then "could not open file" failures after the next checkpoint, as all accesses to the affected catalog stopped working. Recovery would require manually removing the stale "pg_internal.init" file. To fix, get rid of the initFileRelationIds list, and instead consult syscache.c's list of relations used in catalog caches to decide whether a relation is included in the init file. This should be a tad more efficient anyway, since we're replacing linear search of a list with ~100 entries with a binary search. It's a bit ugly that the init file contents are now so directly tied to the catalog caches, but in practice that won't make much difference. Back-patch to all supported branches.
2015-06-07 21:32:09 +02:00
SysCacheRelationOidSize = SysCacheSupportingRelOidSize = 0;
for (cacheId = 0; cacheId < SysCacheSize; cacheId++)
{
/*
* Assert that every enumeration value defined in syscache.h has been
* populated in the cacheinfo array.
*/
Assert(OidIsValid(cacheinfo[cacheId].reloid));
Assert(OidIsValid(cacheinfo[cacheId].indoid));
/* .nbuckets and .key[] are checked by InitCatCache() */
SysCache[cacheId] = InitCatCache(cacheId,
cacheinfo[cacheId].reloid,
cacheinfo[cacheId].indoid,
cacheinfo[cacheId].nkeys,
cacheinfo[cacheId].key,
cacheinfo[cacheId].nbuckets);
if (!PointerIsValid(SysCache[cacheId]))
elog(ERROR, "could not initialize cache %u (%d)",
cacheinfo[cacheId].reloid, cacheId);
Use a safer method for determining whether relcache init file is stale. When we invalidate the relcache entry for a system catalog or index, we must also delete the relcache "init file" if the init file contains a copy of that rel's entry. The old way of doing this relied on a specially maintained list of the OIDs of relations present in the init file: we made the list either when reading the file in, or when writing the file out. The problem is that when writing the file out, we included only rels present in our local relcache, which might have already suffered some deletions due to relcache inval events. In such cases we correctly decided not to overwrite the real init file with incomplete data --- but we still used the incomplete initFileRelationIds list for the rest of the current session. This could result in wrong decisions about whether the session's own actions require deletion of the init file, potentially allowing an init file created by some other concurrent session to be left around even though it's been made stale. Since we don't support changing the schema of a system catalog at runtime, the only likely scenario in which this would cause a problem in the field involves a "vacuum full" on a catalog concurrently with other activity, and even then it's far from easy to provoke. Remarkably, this has been broken since 2002 (in commit 786340441706ac1957a031f11ad1c2e5b6e18314), but we had never seen a reproducible test case until recently. If it did happen in the field, the symptoms would probably involve unexpected "cache lookup failed" errors to begin with, then "could not open file" failures after the next checkpoint, as all accesses to the affected catalog stopped working. Recovery would require manually removing the stale "pg_internal.init" file. To fix, get rid of the initFileRelationIds list, and instead consult syscache.c's list of relations used in catalog caches to decide whether a relation is included in the init file. This should be a tad more efficient anyway, since we're replacing linear search of a list with ~100 entries with a binary search. It's a bit ugly that the init file contents are now so directly tied to the catalog caches, but in practice that won't make much difference. Back-patch to all supported branches.
2015-06-07 21:32:09 +02:00
/* Accumulate data for OID lists, too */
SysCacheRelationOid[SysCacheRelationOidSize++] =
cacheinfo[cacheId].reloid;
Use a safer method for determining whether relcache init file is stale. When we invalidate the relcache entry for a system catalog or index, we must also delete the relcache "init file" if the init file contains a copy of that rel's entry. The old way of doing this relied on a specially maintained list of the OIDs of relations present in the init file: we made the list either when reading the file in, or when writing the file out. The problem is that when writing the file out, we included only rels present in our local relcache, which might have already suffered some deletions due to relcache inval events. In such cases we correctly decided not to overwrite the real init file with incomplete data --- but we still used the incomplete initFileRelationIds list for the rest of the current session. This could result in wrong decisions about whether the session's own actions require deletion of the init file, potentially allowing an init file created by some other concurrent session to be left around even though it's been made stale. Since we don't support changing the schema of a system catalog at runtime, the only likely scenario in which this would cause a problem in the field involves a "vacuum full" on a catalog concurrently with other activity, and even then it's far from easy to provoke. Remarkably, this has been broken since 2002 (in commit 786340441706ac1957a031f11ad1c2e5b6e18314), but we had never seen a reproducible test case until recently. If it did happen in the field, the symptoms would probably involve unexpected "cache lookup failed" errors to begin with, then "could not open file" failures after the next checkpoint, as all accesses to the affected catalog stopped working. Recovery would require manually removing the stale "pg_internal.init" file. To fix, get rid of the initFileRelationIds list, and instead consult syscache.c's list of relations used in catalog caches to decide whether a relation is included in the init file. This should be a tad more efficient anyway, since we're replacing linear search of a list with ~100 entries with a binary search. It's a bit ugly that the init file contents are now so directly tied to the catalog caches, but in practice that won't make much difference. Back-patch to all supported branches.
2015-06-07 21:32:09 +02:00
SysCacheSupportingRelOid[SysCacheSupportingRelOidSize++] =
cacheinfo[cacheId].reloid;
SysCacheSupportingRelOid[SysCacheSupportingRelOidSize++] =
cacheinfo[cacheId].indoid;
/* see comments for RelationInvalidatesSnapshotsOnly */
Assert(!RelationInvalidatesSnapshotsOnly(cacheinfo[cacheId].reloid));
}
Use a safer method for determining whether relcache init file is stale. When we invalidate the relcache entry for a system catalog or index, we must also delete the relcache "init file" if the init file contains a copy of that rel's entry. The old way of doing this relied on a specially maintained list of the OIDs of relations present in the init file: we made the list either when reading the file in, or when writing the file out. The problem is that when writing the file out, we included only rels present in our local relcache, which might have already suffered some deletions due to relcache inval events. In such cases we correctly decided not to overwrite the real init file with incomplete data --- but we still used the incomplete initFileRelationIds list for the rest of the current session. This could result in wrong decisions about whether the session's own actions require deletion of the init file, potentially allowing an init file created by some other concurrent session to be left around even though it's been made stale. Since we don't support changing the schema of a system catalog at runtime, the only likely scenario in which this would cause a problem in the field involves a "vacuum full" on a catalog concurrently with other activity, and even then it's far from easy to provoke. Remarkably, this has been broken since 2002 (in commit 786340441706ac1957a031f11ad1c2e5b6e18314), but we had never seen a reproducible test case until recently. If it did happen in the field, the symptoms would probably involve unexpected "cache lookup failed" errors to begin with, then "could not open file" failures after the next checkpoint, as all accesses to the affected catalog stopped working. Recovery would require manually removing the stale "pg_internal.init" file. To fix, get rid of the initFileRelationIds list, and instead consult syscache.c's list of relations used in catalog caches to decide whether a relation is included in the init file. This should be a tad more efficient anyway, since we're replacing linear search of a list with ~100 entries with a binary search. It's a bit ugly that the init file contents are now so directly tied to the catalog caches, but in practice that won't make much difference. Back-patch to all supported branches.
2015-06-07 21:32:09 +02:00
Assert(SysCacheRelationOidSize <= lengthof(SysCacheRelationOid));
Assert(SysCacheSupportingRelOidSize <= lengthof(SysCacheSupportingRelOid));
/* Sort and de-dup OID arrays, so we can use binary search. */
qsort(SysCacheRelationOid, SysCacheRelationOidSize,
sizeof(Oid), oid_compare);
SysCacheRelationOidSize =
qunique(SysCacheRelationOid, SysCacheRelationOidSize, sizeof(Oid),
oid_compare);
qsort(SysCacheSupportingRelOid, SysCacheSupportingRelOidSize,
sizeof(Oid), oid_compare);
SysCacheSupportingRelOidSize =
qunique(SysCacheSupportingRelOid, SysCacheSupportingRelOidSize,
sizeof(Oid), oid_compare);
Use a safer method for determining whether relcache init file is stale. When we invalidate the relcache entry for a system catalog or index, we must also delete the relcache "init file" if the init file contains a copy of that rel's entry. The old way of doing this relied on a specially maintained list of the OIDs of relations present in the init file: we made the list either when reading the file in, or when writing the file out. The problem is that when writing the file out, we included only rels present in our local relcache, which might have already suffered some deletions due to relcache inval events. In such cases we correctly decided not to overwrite the real init file with incomplete data --- but we still used the incomplete initFileRelationIds list for the rest of the current session. This could result in wrong decisions about whether the session's own actions require deletion of the init file, potentially allowing an init file created by some other concurrent session to be left around even though it's been made stale. Since we don't support changing the schema of a system catalog at runtime, the only likely scenario in which this would cause a problem in the field involves a "vacuum full" on a catalog concurrently with other activity, and even then it's far from easy to provoke. Remarkably, this has been broken since 2002 (in commit 786340441706ac1957a031f11ad1c2e5b6e18314), but we had never seen a reproducible test case until recently. If it did happen in the field, the symptoms would probably involve unexpected "cache lookup failed" errors to begin with, then "could not open file" failures after the next checkpoint, as all accesses to the affected catalog stopped working. Recovery would require manually removing the stale "pg_internal.init" file. To fix, get rid of the initFileRelationIds list, and instead consult syscache.c's list of relations used in catalog caches to decide whether a relation is included in the init file. This should be a tad more efficient anyway, since we're replacing linear search of a list with ~100 entries with a binary search. It's a bit ugly that the init file contents are now so directly tied to the catalog caches, but in practice that won't make much difference. Back-patch to all supported branches.
2015-06-07 21:32:09 +02:00
2000-02-18 10:30:20 +01:00
CacheInitialized = true;
}
/*
* InitCatalogCachePhase2 - finish initializing the caches
*
* Finish initializing all the caches, including necessary database
* access.
*
* This is *not* essential; normally we allow syscaches to be initialized
* on first use. However, it is useful as a mechanism to preload the
* relcache with entries for the most-commonly-used system catalogs.
* Therefore, we invoke this routine when we need to write a new relcache
* init file.
*/
void
InitCatalogCachePhase2(void)
{
int cacheId;
Assert(CacheInitialized);
for (cacheId = 0; cacheId < SysCacheSize; cacheId++)
InitCatCachePhase2(SysCache[cacheId], true);
}
/*
* SearchSysCache
*
* A layer on top of SearchCatCache that does the initialization and
* key-setting for you.
*
* Returns the cache copy of the tuple if one is found, NULL if not.
* The tuple is the 'cache' copy and must NOT be modified!
*
* When the caller is done using the tuple, call ReleaseSysCache()
* to release the reference count grabbed by SearchSysCache(). If this
* is not done, the tuple will remain locked in cache until end of
* transaction, which is tolerable but not desirable.
*
* CAUTION: The tuple that is returned must NOT be freed by the caller!
*/
HeapTuple
SearchSysCache(int cacheId,
Datum key1,
Datum key2,
Datum key3,
Datum key4)
{
Improve sys/catcache performance. The following are the individual improvements: 1) Avoidance of FunctionCallInfo based function calls, replaced by more efficient functions with a native C argument interface. 2) Don't extract columns from a cache entry's tuple whenever matching entries - instead store them as a Datum array. This also allows to get rid of having to build dummy tuples for negative & list entries, and of a hack for dealing with cstring vs. text weirdness. 3) Reorder members of catcache.h struct, so imortant entries are more likely to be on one cacheline. 4) Allowing the compiler to specialize critical SearchCatCache for a specific number of attributes allows to unroll loops and avoid other nkeys dependant initialization. 5) Only initializing the ScanKey when necessary, i.e. catcache misses, greatly reduces cache unnecessary cpu cache misses. 6) Split of the cache-miss case from the hash lookup, reducing stack allocations etc in the common case. 7) CatCTup and their corresponding heaptuple are allocated in one piece. This results in making cache lookups themselves roughly three times as fast - full-system benchmarks obviously improve less than that. I've also evaluated further techniques: - replace open coded hash with simplehash - the list walk right now shows up in profiles. Unfortunately it's not easy to do so safely as an entry's memory location can change at various times, which doesn't work well with the refcounting and cache invalidation. - Cacheline-aligning CatCTup entries - helps some with performance, but the win isn't big and the code for it is ugly, because the tuples have to be freed as well. - add more proper functions, rather than macros for SearchSysCacheCopyN etc., but right now they don't show up in profiles. The reason the macro wrapper for syscache.c/h have to be changed, rather than just catcache, is that doing otherwise would require exposing the SysCache array to the outside. That might be a good idea anyway, but it's for another day. Author: Andres Freund Reviewed-By: Robert Haas Discussion: https://postgr.es/m/20170914061207.zxotvyopetm7lrrp@alap3.anarazel.de
2017-10-13 22:16:50 +02:00
Assert(cacheId >= 0 && cacheId < SysCacheSize &&
PointerIsValid(SysCache[cacheId]));
return SearchCatCache(SysCache[cacheId], key1, key2, key3, key4);
}
Improve sys/catcache performance. The following are the individual improvements: 1) Avoidance of FunctionCallInfo based function calls, replaced by more efficient functions with a native C argument interface. 2) Don't extract columns from a cache entry's tuple whenever matching entries - instead store them as a Datum array. This also allows to get rid of having to build dummy tuples for negative & list entries, and of a hack for dealing with cstring vs. text weirdness. 3) Reorder members of catcache.h struct, so imortant entries are more likely to be on one cacheline. 4) Allowing the compiler to specialize critical SearchCatCache for a specific number of attributes allows to unroll loops and avoid other nkeys dependant initialization. 5) Only initializing the ScanKey when necessary, i.e. catcache misses, greatly reduces cache unnecessary cpu cache misses. 6) Split of the cache-miss case from the hash lookup, reducing stack allocations etc in the common case. 7) CatCTup and their corresponding heaptuple are allocated in one piece. This results in making cache lookups themselves roughly three times as fast - full-system benchmarks obviously improve less than that. I've also evaluated further techniques: - replace open coded hash with simplehash - the list walk right now shows up in profiles. Unfortunately it's not easy to do so safely as an entry's memory location can change at various times, which doesn't work well with the refcounting and cache invalidation. - Cacheline-aligning CatCTup entries - helps some with performance, but the win isn't big and the code for it is ugly, because the tuples have to be freed as well. - add more proper functions, rather than macros for SearchSysCacheCopyN etc., but right now they don't show up in profiles. The reason the macro wrapper for syscache.c/h have to be changed, rather than just catcache, is that doing otherwise would require exposing the SysCache array to the outside. That might be a good idea anyway, but it's for another day. Author: Andres Freund Reviewed-By: Robert Haas Discussion: https://postgr.es/m/20170914061207.zxotvyopetm7lrrp@alap3.anarazel.de
2017-10-13 22:16:50 +02:00
HeapTuple
SearchSysCache1(int cacheId,
Datum key1)
{
Assert(cacheId >= 0 && cacheId < SysCacheSize &&
PointerIsValid(SysCache[cacheId]));
Assert(SysCache[cacheId]->cc_nkeys == 1);
return SearchCatCache1(SysCache[cacheId], key1);
}
HeapTuple
SearchSysCache2(int cacheId,
Datum key1, Datum key2)
{
Assert(cacheId >= 0 && cacheId < SysCacheSize &&
PointerIsValid(SysCache[cacheId]));
Assert(SysCache[cacheId]->cc_nkeys == 2);
return SearchCatCache2(SysCache[cacheId], key1, key2);
}
HeapTuple
SearchSysCache3(int cacheId,
Datum key1, Datum key2, Datum key3)
{
Assert(cacheId >= 0 && cacheId < SysCacheSize &&
PointerIsValid(SysCache[cacheId]));
Assert(SysCache[cacheId]->cc_nkeys == 3);
return SearchCatCache3(SysCache[cacheId], key1, key2, key3);
}
HeapTuple
SearchSysCache4(int cacheId,
Datum key1, Datum key2, Datum key3, Datum key4)
{
Assert(cacheId >= 0 && cacheId < SysCacheSize &&
PointerIsValid(SysCache[cacheId]));
Assert(SysCache[cacheId]->cc_nkeys == 4);
return SearchCatCache4(SysCache[cacheId], key1, key2, key3, key4);
}
/*
* ReleaseSysCache
* Release previously grabbed reference count on a tuple
*/
void
ReleaseSysCache(HeapTuple tuple)
{
ReleaseCatCache(tuple);
}
/*
* SearchSysCacheCopy
*
* A convenience routine that does SearchSysCache and (if successful)
* returns a modifiable copy of the syscache entry. The original
* syscache entry is released before returning. The caller should
* heap_freetuple() the result when done with it.
*/
HeapTuple
SearchSysCacheCopy(int cacheId,
Datum key1,
Datum key2,
Datum key3,
Datum key4)
{
HeapTuple tuple,
newtuple;
tuple = SearchSysCache(cacheId, key1, key2, key3, key4);
if (!HeapTupleIsValid(tuple))
return tuple;
newtuple = heap_copytuple(tuple);
ReleaseSysCache(tuple);
return newtuple;
}
/*
* SearchSysCacheExists
*
* A convenience routine that just probes to see if a tuple can be found.
* No lock is retained on the syscache entry.
*/
bool
SearchSysCacheExists(int cacheId,
Datum key1,
Datum key2,
Datum key3,
Datum key4)
{
HeapTuple tuple;
tuple = SearchSysCache(cacheId, key1, key2, key3, key4);
if (!HeapTupleIsValid(tuple))
return false;
ReleaseSysCache(tuple);
return true;
}
/*
* GetSysCacheOid
*
Remove WITH OIDS support, change oid catalog column visibility. Previously tables declared WITH OIDS, including a significant fraction of the catalog tables, stored the oid column not as a normal column, but as part of the tuple header. This special column was not shown by default, which was somewhat odd, as it's often (consider e.g. pg_class.oid) one of the more important parts of a row. Neither pg_dump nor COPY included the contents of the oid column by default. The fact that the oid column was not an ordinary column necessitated a significant amount of special case code to support oid columns. That already was painful for the existing, but upcoming work aiming to make table storage pluggable, would have required expanding and duplicating that "specialness" significantly. WITH OIDS has been deprecated since 2005 (commit ff02d0a05280e0). Remove it. Removing includes: - CREATE TABLE and ALTER TABLE syntax for declaring the table to be WITH OIDS has been removed (WITH (oids[ = true]) will error out) - pg_dump does not support dumping tables declared WITH OIDS and will issue a warning when dumping one (and ignore the oid column). - restoring an pg_dump archive with pg_restore will warn when restoring a table with oid contents (and ignore the oid column) - COPY will refuse to load binary dump that includes oids. - pg_upgrade will error out when encountering tables declared WITH OIDS, they have to be altered to remove the oid column first. - Functionality to access the oid of the last inserted row (like plpgsql's RESULT_OID, spi's SPI_lastoid, ...) has been removed. The syntax for declaring a table WITHOUT OIDS (or WITH (oids = false) for CREATE TABLE) is still supported. While that requires a bit of support code, it seems unnecessary to break applications / dumps that do not use oids, and are explicit about not using them. The biggest user of WITH OID columns was postgres' catalog. This commit changes all 'magic' oid columns to be columns that are normally declared and stored. To reduce unnecessary query breakage all the newly added columns are still named 'oid', even if a table's column naming scheme would indicate 'reloid' or such. This obviously requires adapting a lot code, mostly replacing oid access via HeapTupleGetOid() with access to the underlying Form_pg_*->oid column. The bootstrap process now assigns oids for all oid columns in genbki.pl that do not have an explicit value (starting at the largest oid previously used), only oids assigned later by oids will be above FirstBootstrapObjectId. As the oid column now is a normal column the special bootstrap syntax for oids has been removed. Oids are not automatically assigned during insertion anymore, all backend code explicitly assigns oids with GetNewOidWithIndex(). For the rare case that insertions into the catalog via SQL are called for the new pg_nextoid() function can be used (which only works on catalog tables). The fact that oid columns on system tables are now normal columns means that they will be included in the set of columns expanded by * (i.e. SELECT * FROM pg_class will now include the table's oid, previously it did not). It'd not technically be hard to hide oid column by default, but that'd mean confusing behavior would either have to be carried forward forever, or it'd cause breakage down the line. While it's not unlikely that further adjustments are needed, the scope/invasiveness of the patch makes it worthwhile to get merge this now. It's painful to maintain externally, too complicated to commit after the code code freeze, and a dependency of a number of other patches. Catversion bump, for obvious reasons. Author: Andres Freund, with contributions by John Naylor Discussion: https://postgr.es/m/20180930034810.ywp2c7awz7opzcfr@alap3.anarazel.de
2018-11-21 00:36:57 +01:00
* A convenience routine that does SearchSysCache and returns the OID in the
* oidcol column of the found tuple, or InvalidOid if no tuple could be found.
* No lock is retained on the syscache entry.
*/
Oid
GetSysCacheOid(int cacheId,
Remove WITH OIDS support, change oid catalog column visibility. Previously tables declared WITH OIDS, including a significant fraction of the catalog tables, stored the oid column not as a normal column, but as part of the tuple header. This special column was not shown by default, which was somewhat odd, as it's often (consider e.g. pg_class.oid) one of the more important parts of a row. Neither pg_dump nor COPY included the contents of the oid column by default. The fact that the oid column was not an ordinary column necessitated a significant amount of special case code to support oid columns. That already was painful for the existing, but upcoming work aiming to make table storage pluggable, would have required expanding and duplicating that "specialness" significantly. WITH OIDS has been deprecated since 2005 (commit ff02d0a05280e0). Remove it. Removing includes: - CREATE TABLE and ALTER TABLE syntax for declaring the table to be WITH OIDS has been removed (WITH (oids[ = true]) will error out) - pg_dump does not support dumping tables declared WITH OIDS and will issue a warning when dumping one (and ignore the oid column). - restoring an pg_dump archive with pg_restore will warn when restoring a table with oid contents (and ignore the oid column) - COPY will refuse to load binary dump that includes oids. - pg_upgrade will error out when encountering tables declared WITH OIDS, they have to be altered to remove the oid column first. - Functionality to access the oid of the last inserted row (like plpgsql's RESULT_OID, spi's SPI_lastoid, ...) has been removed. The syntax for declaring a table WITHOUT OIDS (or WITH (oids = false) for CREATE TABLE) is still supported. While that requires a bit of support code, it seems unnecessary to break applications / dumps that do not use oids, and are explicit about not using them. The biggest user of WITH OID columns was postgres' catalog. This commit changes all 'magic' oid columns to be columns that are normally declared and stored. To reduce unnecessary query breakage all the newly added columns are still named 'oid', even if a table's column naming scheme would indicate 'reloid' or such. This obviously requires adapting a lot code, mostly replacing oid access via HeapTupleGetOid() with access to the underlying Form_pg_*->oid column. The bootstrap process now assigns oids for all oid columns in genbki.pl that do not have an explicit value (starting at the largest oid previously used), only oids assigned later by oids will be above FirstBootstrapObjectId. As the oid column now is a normal column the special bootstrap syntax for oids has been removed. Oids are not automatically assigned during insertion anymore, all backend code explicitly assigns oids with GetNewOidWithIndex(). For the rare case that insertions into the catalog via SQL are called for the new pg_nextoid() function can be used (which only works on catalog tables). The fact that oid columns on system tables are now normal columns means that they will be included in the set of columns expanded by * (i.e. SELECT * FROM pg_class will now include the table's oid, previously it did not). It'd not technically be hard to hide oid column by default, but that'd mean confusing behavior would either have to be carried forward forever, or it'd cause breakage down the line. While it's not unlikely that further adjustments are needed, the scope/invasiveness of the patch makes it worthwhile to get merge this now. It's painful to maintain externally, too complicated to commit after the code code freeze, and a dependency of a number of other patches. Catversion bump, for obvious reasons. Author: Andres Freund, with contributions by John Naylor Discussion: https://postgr.es/m/20180930034810.ywp2c7awz7opzcfr@alap3.anarazel.de
2018-11-21 00:36:57 +01:00
AttrNumber oidcol,
Datum key1,
Datum key2,
Datum key3,
Datum key4)
{
HeapTuple tuple;
Remove WITH OIDS support, change oid catalog column visibility. Previously tables declared WITH OIDS, including a significant fraction of the catalog tables, stored the oid column not as a normal column, but as part of the tuple header. This special column was not shown by default, which was somewhat odd, as it's often (consider e.g. pg_class.oid) one of the more important parts of a row. Neither pg_dump nor COPY included the contents of the oid column by default. The fact that the oid column was not an ordinary column necessitated a significant amount of special case code to support oid columns. That already was painful for the existing, but upcoming work aiming to make table storage pluggable, would have required expanding and duplicating that "specialness" significantly. WITH OIDS has been deprecated since 2005 (commit ff02d0a05280e0). Remove it. Removing includes: - CREATE TABLE and ALTER TABLE syntax for declaring the table to be WITH OIDS has been removed (WITH (oids[ = true]) will error out) - pg_dump does not support dumping tables declared WITH OIDS and will issue a warning when dumping one (and ignore the oid column). - restoring an pg_dump archive with pg_restore will warn when restoring a table with oid contents (and ignore the oid column) - COPY will refuse to load binary dump that includes oids. - pg_upgrade will error out when encountering tables declared WITH OIDS, they have to be altered to remove the oid column first. - Functionality to access the oid of the last inserted row (like plpgsql's RESULT_OID, spi's SPI_lastoid, ...) has been removed. The syntax for declaring a table WITHOUT OIDS (or WITH (oids = false) for CREATE TABLE) is still supported. While that requires a bit of support code, it seems unnecessary to break applications / dumps that do not use oids, and are explicit about not using them. The biggest user of WITH OID columns was postgres' catalog. This commit changes all 'magic' oid columns to be columns that are normally declared and stored. To reduce unnecessary query breakage all the newly added columns are still named 'oid', even if a table's column naming scheme would indicate 'reloid' or such. This obviously requires adapting a lot code, mostly replacing oid access via HeapTupleGetOid() with access to the underlying Form_pg_*->oid column. The bootstrap process now assigns oids for all oid columns in genbki.pl that do not have an explicit value (starting at the largest oid previously used), only oids assigned later by oids will be above FirstBootstrapObjectId. As the oid column now is a normal column the special bootstrap syntax for oids has been removed. Oids are not automatically assigned during insertion anymore, all backend code explicitly assigns oids with GetNewOidWithIndex(). For the rare case that insertions into the catalog via SQL are called for the new pg_nextoid() function can be used (which only works on catalog tables). The fact that oid columns on system tables are now normal columns means that they will be included in the set of columns expanded by * (i.e. SELECT * FROM pg_class will now include the table's oid, previously it did not). It'd not technically be hard to hide oid column by default, but that'd mean confusing behavior would either have to be carried forward forever, or it'd cause breakage down the line. While it's not unlikely that further adjustments are needed, the scope/invasiveness of the patch makes it worthwhile to get merge this now. It's painful to maintain externally, too complicated to commit after the code code freeze, and a dependency of a number of other patches. Catversion bump, for obvious reasons. Author: Andres Freund, with contributions by John Naylor Discussion: https://postgr.es/m/20180930034810.ywp2c7awz7opzcfr@alap3.anarazel.de
2018-11-21 00:36:57 +01:00
bool isNull;
Oid result;
tuple = SearchSysCache(cacheId, key1, key2, key3, key4);
if (!HeapTupleIsValid(tuple))
return InvalidOid;
Remove WITH OIDS support, change oid catalog column visibility. Previously tables declared WITH OIDS, including a significant fraction of the catalog tables, stored the oid column not as a normal column, but as part of the tuple header. This special column was not shown by default, which was somewhat odd, as it's often (consider e.g. pg_class.oid) one of the more important parts of a row. Neither pg_dump nor COPY included the contents of the oid column by default. The fact that the oid column was not an ordinary column necessitated a significant amount of special case code to support oid columns. That already was painful for the existing, but upcoming work aiming to make table storage pluggable, would have required expanding and duplicating that "specialness" significantly. WITH OIDS has been deprecated since 2005 (commit ff02d0a05280e0). Remove it. Removing includes: - CREATE TABLE and ALTER TABLE syntax for declaring the table to be WITH OIDS has been removed (WITH (oids[ = true]) will error out) - pg_dump does not support dumping tables declared WITH OIDS and will issue a warning when dumping one (and ignore the oid column). - restoring an pg_dump archive with pg_restore will warn when restoring a table with oid contents (and ignore the oid column) - COPY will refuse to load binary dump that includes oids. - pg_upgrade will error out when encountering tables declared WITH OIDS, they have to be altered to remove the oid column first. - Functionality to access the oid of the last inserted row (like plpgsql's RESULT_OID, spi's SPI_lastoid, ...) has been removed. The syntax for declaring a table WITHOUT OIDS (or WITH (oids = false) for CREATE TABLE) is still supported. While that requires a bit of support code, it seems unnecessary to break applications / dumps that do not use oids, and are explicit about not using them. The biggest user of WITH OID columns was postgres' catalog. This commit changes all 'magic' oid columns to be columns that are normally declared and stored. To reduce unnecessary query breakage all the newly added columns are still named 'oid', even if a table's column naming scheme would indicate 'reloid' or such. This obviously requires adapting a lot code, mostly replacing oid access via HeapTupleGetOid() with access to the underlying Form_pg_*->oid column. The bootstrap process now assigns oids for all oid columns in genbki.pl that do not have an explicit value (starting at the largest oid previously used), only oids assigned later by oids will be above FirstBootstrapObjectId. As the oid column now is a normal column the special bootstrap syntax for oids has been removed. Oids are not automatically assigned during insertion anymore, all backend code explicitly assigns oids with GetNewOidWithIndex(). For the rare case that insertions into the catalog via SQL are called for the new pg_nextoid() function can be used (which only works on catalog tables). The fact that oid columns on system tables are now normal columns means that they will be included in the set of columns expanded by * (i.e. SELECT * FROM pg_class will now include the table's oid, previously it did not). It'd not technically be hard to hide oid column by default, but that'd mean confusing behavior would either have to be carried forward forever, or it'd cause breakage down the line. While it's not unlikely that further adjustments are needed, the scope/invasiveness of the patch makes it worthwhile to get merge this now. It's painful to maintain externally, too complicated to commit after the code code freeze, and a dependency of a number of other patches. Catversion bump, for obvious reasons. Author: Andres Freund, with contributions by John Naylor Discussion: https://postgr.es/m/20180930034810.ywp2c7awz7opzcfr@alap3.anarazel.de
2018-11-21 00:36:57 +01:00
result = heap_getattr(tuple, oidcol,
SysCache[cacheId]->cc_tupdesc,
&isNull);
Assert(!isNull); /* columns used as oids should never be NULL */
ReleaseSysCache(tuple);
return result;
}
/*
* SearchSysCacheAttName
*
* This routine is equivalent to SearchSysCache on the ATTNAME cache,
* except that it will return NULL if the found attribute is marked
* attisdropped. This is convenient for callers that want to act as
* though dropped attributes don't exist.
*/
HeapTuple
SearchSysCacheAttName(Oid relid, const char *attname)
{
HeapTuple tuple;
tuple = SearchSysCache2(ATTNAME,
ObjectIdGetDatum(relid),
CStringGetDatum(attname));
if (!HeapTupleIsValid(tuple))
return NULL;
if (((Form_pg_attribute) GETSTRUCT(tuple))->attisdropped)
{
ReleaseSysCache(tuple);
return NULL;
}
return tuple;
}
/*
* SearchSysCacheCopyAttName
*
* As above, an attisdropped-aware version of SearchSysCacheCopy.
*/
HeapTuple
SearchSysCacheCopyAttName(Oid relid, const char *attname)
{
HeapTuple tuple,
newtuple;
tuple = SearchSysCacheAttName(relid, attname);
if (!HeapTupleIsValid(tuple))
return tuple;
newtuple = heap_copytuple(tuple);
ReleaseSysCache(tuple);
return newtuple;
}
/*
* SearchSysCacheExistsAttName
*
* As above, an attisdropped-aware version of SearchSysCacheExists.
*/
bool
SearchSysCacheExistsAttName(Oid relid, const char *attname)
{
HeapTuple tuple;
tuple = SearchSysCacheAttName(relid, attname);
if (!HeapTupleIsValid(tuple))
return false;
ReleaseSysCache(tuple);
return true;
}
/*
* SearchSysCacheAttNum
*
* This routine is equivalent to SearchSysCache on the ATTNUM cache,
* except that it will return NULL if the found attribute is marked
* attisdropped. This is convenient for callers that want to act as
* though dropped attributes don't exist.
*/
HeapTuple
SearchSysCacheAttNum(Oid relid, int16 attnum)
{
HeapTuple tuple;
tuple = SearchSysCache2(ATTNUM,
ObjectIdGetDatum(relid),
Int16GetDatum(attnum));
if (!HeapTupleIsValid(tuple))
return NULL;
if (((Form_pg_attribute) GETSTRUCT(tuple))->attisdropped)
{
ReleaseSysCache(tuple);
return NULL;
}
return tuple;
}
/*
* SearchSysCacheCopyAttNum
*
* As above, an attisdropped-aware version of SearchSysCacheCopy.
*/
HeapTuple
SearchSysCacheCopyAttNum(Oid relid, int16 attnum)
{
HeapTuple tuple,
newtuple;
tuple = SearchSysCacheAttNum(relid, attnum);
if (!HeapTupleIsValid(tuple))
return NULL;
newtuple = heap_copytuple(tuple);
ReleaseSysCache(tuple);
return newtuple;
}
/*
* SysCacheGetAttr
*
* Given a tuple previously fetched by SearchSysCache(),
* extract a specific attribute.
*
* This is equivalent to using heap_getattr() on a tuple fetched
* from a non-cached relation. Usually, this is only used for attributes
* that could be NULL or variable length; the fixed-size attributes in
* a system table are accessed just by mapping the tuple onto the C struct
* declarations from include/catalog/.
*
* As with heap_getattr(), if the attribute is of a pass-by-reference type
* then a pointer into the tuple data area is returned --- the caller must
* not modify or pfree the datum!
*
* Note: it is legal to use SysCacheGetAttr() with a cacheId referencing
* a different cache for the same catalog the tuple was fetched from.
*/
Datum
SysCacheGetAttr(int cacheId, HeapTuple tup,
AttrNumber attributeNumber,
bool *isNull)
{
/*
* We just need to get the TupleDesc out of the cache entry, and then we
* can apply heap_getattr(). Normally the cache control data is already
* valid (because the caller recently fetched the tuple via this same
* cache), but there are cases where we have to initialize the cache here.
*/
if (cacheId < 0 || cacheId >= SysCacheSize ||
!PointerIsValid(SysCache[cacheId]))
2011-06-18 23:37:30 +02:00
elog(ERROR, "invalid cache ID: %d", cacheId);
if (!PointerIsValid(SysCache[cacheId]->cc_tupdesc))
{
InitCatCachePhase2(SysCache[cacheId], false);
Assert(PointerIsValid(SysCache[cacheId]->cc_tupdesc));
}
return heap_getattr(tup, attributeNumber,
SysCache[cacheId]->cc_tupdesc,
isNull);
}
/*
* SysCacheGetAttrNotNull
*
* As above, a version of SysCacheGetAttr which knows that the attr cannot
* be NULL.
*/
Datum
SysCacheGetAttrNotNull(int cacheId, HeapTuple tup,
AttrNumber attributeNumber)
{
bool isnull;
Datum attr;
attr = SysCacheGetAttr(cacheId, tup, attributeNumber, &isnull);
if (isnull)
{
elog(ERROR,
"unexpected null value in cached tuple for catalog %s column %s",
get_rel_name(cacheinfo[cacheId].reloid),
NameStr(TupleDescAttr(SysCache[cacheId]->cc_tupdesc, attributeNumber - 1)->attname));
}
return attr;
}
/*
* GetSysCacheHashValue
*
* Get the hash value that would be used for a tuple in the specified cache
* with the given search keys.
*
* The reason for exposing this as part of the API is that the hash value is
* exposed in cache invalidation operations, so there are places outside the
* catcache code that need to be able to compute the hash values.
*/
uint32
GetSysCacheHashValue(int cacheId,
Datum key1,
Datum key2,
Datum key3,
Datum key4)
{
if (cacheId < 0 || cacheId >= SysCacheSize ||
!PointerIsValid(SysCache[cacheId]))
elog(ERROR, "invalid cache ID: %d", cacheId);
return GetCatCacheHashValue(SysCache[cacheId], key1, key2, key3, key4);
}
/*
* List-search interface
*/
struct catclist *
SearchSysCacheList(int cacheId, int nkeys,
Datum key1, Datum key2, Datum key3)
{
if (cacheId < 0 || cacheId >= SysCacheSize ||
!PointerIsValid(SysCache[cacheId]))
2011-06-18 23:37:30 +02:00
elog(ERROR, "invalid cache ID: %d", cacheId);
return SearchCatCacheList(SysCache[cacheId], nkeys,
key1, key2, key3);
}
2017-05-13 00:17:29 +02:00
/*
* SysCacheInvalidate
*
* Invalidate entries in the specified cache, given a hash value.
* See CatCacheInvalidate() for more info.
*
* This routine is only quasi-public: it should only be used by inval.c.
*/
void
SysCacheInvalidate(int cacheId, uint32 hashValue)
{
if (cacheId < 0 || cacheId >= SysCacheSize)
elog(ERROR, "invalid cache ID: %d", cacheId);
/* if this cache isn't initialized yet, no need to do anything */
if (!PointerIsValid(SysCache[cacheId]))
return;
CatCacheInvalidate(SysCache[cacheId], hashValue);
}
/*
* Certain relations that do not have system caches send snapshot invalidation
* messages in lieu of catcache messages. This is for the benefit of
* GetCatalogSnapshot(), which can then reuse its existing MVCC snapshot
* for scanning one of those catalogs, rather than taking a new one, if no
* invalidation has been received.
*
* Relations that have syscaches need not (and must not) be listed here. The
* catcache invalidation messages will also flush the snapshot. If you add a
* syscache for one of these relations, remove it from this list.
*/
bool
RelationInvalidatesSnapshotsOnly(Oid relid)
{
switch (relid)
{
case DbRoleSettingRelationId:
case DependRelationId:
case SharedDependRelationId:
case DescriptionRelationId:
case SharedDescriptionRelationId:
case SecLabelRelationId:
case SharedSecLabelRelationId:
return true;
default:
break;
}
return false;
}
/*
* Test whether a relation has a system cache.
*/
bool
RelationHasSysCache(Oid relid)
{
int low = 0,
high = SysCacheRelationOidSize - 1;
while (low <= high)
{
int middle = low + (high - low) / 2;
if (SysCacheRelationOid[middle] == relid)
return true;
if (SysCacheRelationOid[middle] < relid)
low = middle + 1;
else
high = middle - 1;
}
return false;
}
Use a safer method for determining whether relcache init file is stale. When we invalidate the relcache entry for a system catalog or index, we must also delete the relcache "init file" if the init file contains a copy of that rel's entry. The old way of doing this relied on a specially maintained list of the OIDs of relations present in the init file: we made the list either when reading the file in, or when writing the file out. The problem is that when writing the file out, we included only rels present in our local relcache, which might have already suffered some deletions due to relcache inval events. In such cases we correctly decided not to overwrite the real init file with incomplete data --- but we still used the incomplete initFileRelationIds list for the rest of the current session. This could result in wrong decisions about whether the session's own actions require deletion of the init file, potentially allowing an init file created by some other concurrent session to be left around even though it's been made stale. Since we don't support changing the schema of a system catalog at runtime, the only likely scenario in which this would cause a problem in the field involves a "vacuum full" on a catalog concurrently with other activity, and even then it's far from easy to provoke. Remarkably, this has been broken since 2002 (in commit 786340441706ac1957a031f11ad1c2e5b6e18314), but we had never seen a reproducible test case until recently. If it did happen in the field, the symptoms would probably involve unexpected "cache lookup failed" errors to begin with, then "could not open file" failures after the next checkpoint, as all accesses to the affected catalog stopped working. Recovery would require manually removing the stale "pg_internal.init" file. To fix, get rid of the initFileRelationIds list, and instead consult syscache.c's list of relations used in catalog caches to decide whether a relation is included in the init file. This should be a tad more efficient anyway, since we're replacing linear search of a list with ~100 entries with a binary search. It's a bit ugly that the init file contents are now so directly tied to the catalog caches, but in practice that won't make much difference. Back-patch to all supported branches.
2015-06-07 21:32:09 +02:00
/*
* Test whether a relation supports a system cache, ie it is either a
* cached table or the index used for a cache.
*/
bool
RelationSupportsSysCache(Oid relid)
{
int low = 0,
high = SysCacheSupportingRelOidSize - 1;
while (low <= high)
{
int middle = low + (high - low) / 2;
if (SysCacheSupportingRelOid[middle] == relid)
return true;
if (SysCacheSupportingRelOid[middle] < relid)
low = middle + 1;
else
high = middle - 1;
}
return false;
}
/*
* OID comparator for qsort
*/
static int
oid_compare(const void *a, const void *b)
{
Use a safer method for determining whether relcache init file is stale. When we invalidate the relcache entry for a system catalog or index, we must also delete the relcache "init file" if the init file contains a copy of that rel's entry. The old way of doing this relied on a specially maintained list of the OIDs of relations present in the init file: we made the list either when reading the file in, or when writing the file out. The problem is that when writing the file out, we included only rels present in our local relcache, which might have already suffered some deletions due to relcache inval events. In such cases we correctly decided not to overwrite the real init file with incomplete data --- but we still used the incomplete initFileRelationIds list for the rest of the current session. This could result in wrong decisions about whether the session's own actions require deletion of the init file, potentially allowing an init file created by some other concurrent session to be left around even though it's been made stale. Since we don't support changing the schema of a system catalog at runtime, the only likely scenario in which this would cause a problem in the field involves a "vacuum full" on a catalog concurrently with other activity, and even then it's far from easy to provoke. Remarkably, this has been broken since 2002 (in commit 786340441706ac1957a031f11ad1c2e5b6e18314), but we had never seen a reproducible test case until recently. If it did happen in the field, the symptoms would probably involve unexpected "cache lookup failed" errors to begin with, then "could not open file" failures after the next checkpoint, as all accesses to the affected catalog stopped working. Recovery would require manually removing the stale "pg_internal.init" file. To fix, get rid of the initFileRelationIds list, and instead consult syscache.c's list of relations used in catalog caches to decide whether a relation is included in the init file. This should be a tad more efficient anyway, since we're replacing linear search of a list with ~100 entries with a binary search. It's a bit ugly that the init file contents are now so directly tied to the catalog caches, but in practice that won't make much difference. Back-patch to all supported branches.
2015-06-07 21:32:09 +02:00
Oid oa = *((const Oid *) a);
Oid ob = *((const Oid *) b);
return pg_cmp_u32(oa, ob);
}