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

863 lines
24 KiB
C
Raw Normal View History

/*-------------------------------------------------------------------------
*
* plancache.c
* Plan cache management.
*
* We can store a cached plan in either fully-planned format, or just
* parsed-and-rewritten if the caller wishes to postpone planning until
* actual parameter values are available. CachedPlanSource has the same
* contents either way, but CachedPlan contains a list of PlannedStmts
* and bare utility statements in the first case, or a list of Query nodes
* in the second case.
*
* The plan cache manager itself is principally responsible for tracking
* whether cached plans should be invalidated because of schema changes in
* the tables they depend on. When (and if) the next demand for a cached
* plan occurs, the query will be replanned. Note that this could result
* in an error, for example if a column referenced by the query is no
* longer present. The creator of a cached plan can specify whether it
* is allowable for the query to change output tupdesc on replan (this
* could happen with "SELECT *" for example) --- if so, it's up to the
* caller to notice changes and cope with them.
*
* Currently, we use only relcache invalidation events to invalidate plans.
* This means that changes such as modification of a function definition do
* not invalidate plans using the function. This is not 100% OK --- for
* example, changing a SQL function that's been inlined really ought to
* cause invalidation of the plan that it's been inlined into --- but the
* cost of tracking additional types of object seems much higher than the
* gain, so we're just ignoring them for now.
*
*
* Portions Copyright (c) 1996-2007, PostgreSQL Global Development Group
* Portions Copyright (c) 1994, Regents of the University of California
*
* IDENTIFICATION
* $PostgreSQL: pgsql/src/backend/utils/cache/plancache.c,v 1.1 2007/03/13 00:33:42 tgl Exp $
*
*-------------------------------------------------------------------------
*/
#include "postgres.h"
#include "utils/plancache.h"
#include "executor/executor.h"
#include "optimizer/clauses.h"
#include "storage/lmgr.h"
#include "tcop/pquery.h"
#include "tcop/tcopprot.h"
#include "tcop/utility.h"
#include "utils/inval.h"
#include "utils/memutils.h"
#include "utils/resowner.h"
typedef struct
{
void (*callback) ();
void *arg;
} ScanQueryWalkerContext;
typedef struct
{
Oid inval_relid;
CachedPlan *plan;
} InvalRelidContext;
static List *cached_plans_list = NIL;
static void StoreCachedPlan(CachedPlanSource *plansource, List *stmt_list,
MemoryContext plan_context);
static void AcquireExecutorLocks(List *stmt_list, bool acquire);
static void AcquirePlannerLocks(List *stmt_list, bool acquire);
static void LockRelid(Oid relid, LOCKMODE lockmode, void *arg);
static void UnlockRelid(Oid relid, LOCKMODE lockmode, void *arg);
static void ScanQueryForRelids(Query *parsetree,
void (*callback) (),
void *arg);
static bool ScanQueryWalker(Node *node, ScanQueryWalkerContext *context);
static bool rowmark_member(List *rowMarks, int rt_index);
static TupleDesc ComputeResultDesc(List *stmt_list);
static void PlanCacheCallback(Datum arg, Oid relid);
static void InvalRelid(Oid relid, LOCKMODE lockmode,
InvalRelidContext *context);
/*
* InitPlanCache: initialize module during InitPostgres.
*
* All we need to do is hook into inval.c's callback list.
*/
void
InitPlanCache(void)
{
CacheRegisterRelcacheCallback(PlanCacheCallback, (Datum) 0);
}
/*
* CreateCachedPlan: initially create a plan cache entry.
*
* The caller must already have successfully parsed/planned the query;
* about all that we do here is copy it into permanent storage.
*
* raw_parse_tree: output of raw_parser()
* query_string: original query text (can be NULL if not available, but
* that is discouraged because it degrades error message quality)
* commandTag: compile-time-constant tag for query, or NULL if empty query
* param_types: array of parameter type OIDs, or NULL if none
* num_params: number of parameters
* stmt_list: list of PlannedStmts/utility stmts, or list of Query trees
* fully_planned: are we caching planner or rewriter output?
* fixed_result: TRUE to disallow changes in result tupdesc
*/
CachedPlanSource *
CreateCachedPlan(Node *raw_parse_tree,
const char *query_string,
const char *commandTag,
Oid *param_types,
int num_params,
List *stmt_list,
bool fully_planned,
bool fixed_result)
{
CachedPlanSource *plansource;
MemoryContext source_context;
MemoryContext oldcxt;
/*
* Make a dedicated memory context for the CachedPlanSource and its
* subsidiary data. We expect it can be pretty small.
*/
source_context = AllocSetContextCreate(CacheMemoryContext,
"CachedPlanSource",
ALLOCSET_SMALL_MINSIZE,
ALLOCSET_SMALL_INITSIZE,
ALLOCSET_SMALL_MAXSIZE);
/*
* Create and fill the CachedPlanSource struct within the new context.
*/
oldcxt = MemoryContextSwitchTo(source_context);
plansource = (CachedPlanSource *) palloc(sizeof(CachedPlanSource));
plansource->raw_parse_tree = copyObject(raw_parse_tree);
plansource->query_string = query_string ? pstrdup(query_string) : NULL;
plansource->commandTag = commandTag; /* no copying needed */
if (num_params > 0)
{
plansource->param_types = (Oid *) palloc(num_params * sizeof(Oid));
memcpy(plansource->param_types, param_types, num_params * sizeof(Oid));
}
else
plansource->param_types = NULL;
plansource->num_params = num_params;
plansource->fully_planned = fully_planned;
plansource->fixed_result = fixed_result;
plansource->generation = 0; /* StoreCachedPlan will increment */
plansource->resultDesc = ComputeResultDesc(stmt_list);
plansource->plan = NULL;
plansource->context = source_context;
plansource->orig_plan = NULL;
/*
* Copy the current output plans into the plancache entry.
*/
StoreCachedPlan(plansource, stmt_list, NULL);
/*
* Now we can add the entry to the list of cached plans. The List nodes
* live in CacheMemoryContext.
*/
MemoryContextSwitchTo(CacheMemoryContext);
cached_plans_list = lappend(cached_plans_list, plansource);
MemoryContextSwitchTo(oldcxt);
return plansource;
}
/*
* FastCreateCachedPlan: create a plan cache entry with minimal data copying.
*
* For plans that aren't expected to live very long, the copying overhead of
* CreateCachedPlan is annoying. We provide this variant entry point in which
* the caller has already placed all the data in a suitable memory context.
* The source data and completed plan are in the same context, since this
* avoids extra copy steps during plan construction. If the query ever does
* need replanning, we'll generate a separate new CachedPlan at that time, but
* the CachedPlanSource and the initial CachedPlan share the caller-provided
* context and go away together when neither is needed any longer. (Because
* the parser and planner generate extra cruft in addition to their real
* output, this approach means that the context probably contains a bunch of
* useless junk as well as the useful trees. Hence, this method is a
* space-for-time tradeoff, which is worth making for plans expected to be
* short-lived.)
*
* raw_parse_tree, query_string, param_types, and stmt_list must reside in the
* given context, which must have adequate lifespan (recommendation: make it a
* child of CacheMemoryContext). Otherwise the API is the same as
* CreateCachedPlan.
*/
CachedPlanSource *
FastCreateCachedPlan(Node *raw_parse_tree,
char *query_string,
const char *commandTag,
Oid *param_types,
int num_params,
List *stmt_list,
bool fully_planned,
bool fixed_result,
MemoryContext context)
{
CachedPlanSource *plansource;
MemoryContext oldcxt;
/*
* Create and fill the CachedPlanSource struct within the given context.
*/
oldcxt = MemoryContextSwitchTo(context);
plansource = (CachedPlanSource *) palloc(sizeof(CachedPlanSource));
plansource->raw_parse_tree = raw_parse_tree;
plansource->query_string = query_string;
plansource->commandTag = commandTag; /* no copying needed */
plansource->param_types = param_types;
plansource->num_params = num_params;
plansource->fully_planned = fully_planned;
plansource->fixed_result = fixed_result;
plansource->generation = 0; /* StoreCachedPlan will increment */
plansource->resultDesc = ComputeResultDesc(stmt_list);
plansource->plan = NULL;
plansource->context = context;
plansource->orig_plan = NULL;
/*
* Store the current output plans into the plancache entry.
*/
StoreCachedPlan(plansource, stmt_list, context);
/*
* Since the context is owned by the CachedPlan, advance its refcount.
*/
plansource->orig_plan = plansource->plan;
plansource->orig_plan->refcount++;
/*
* Now we can add the entry to the list of cached plans. The List nodes
* live in CacheMemoryContext.
*/
MemoryContextSwitchTo(CacheMemoryContext);
cached_plans_list = lappend(cached_plans_list, plansource);
MemoryContextSwitchTo(oldcxt);
return plansource;
}
/*
* StoreCachedPlan: store a built or rebuilt plan into a plancache entry.
*
* Common subroutine for CreateCachedPlan and RevalidateCachedPlan.
*/
static void
StoreCachedPlan(CachedPlanSource *plansource,
List *stmt_list,
MemoryContext plan_context)
{
CachedPlan *plan;
MemoryContext oldcxt;
if (plan_context == NULL)
{
/*
* Make a dedicated memory context for the CachedPlan and its
* subsidiary data.
*/
plan_context = AllocSetContextCreate(CacheMemoryContext,
"CachedPlan",
ALLOCSET_DEFAULT_MINSIZE,
ALLOCSET_DEFAULT_INITSIZE,
ALLOCSET_DEFAULT_MAXSIZE);
/*
* Copy supplied data into the new context.
*/
oldcxt = MemoryContextSwitchTo(plan_context);
stmt_list = (List *) copyObject(stmt_list);
}
else
{
/* Assume subsidiary data is in the given context */
oldcxt = MemoryContextSwitchTo(plan_context);
}
/*
* Create and fill the CachedPlan struct within the new context.
*/
plan = (CachedPlan *) palloc(sizeof(CachedPlan));
plan->stmt_list = stmt_list;
plan->fully_planned = plansource->fully_planned;
plan->dead = false;
plan->refcount = 1; /* for the parent's link */
plan->generation = ++(plansource->generation);
plan->context = plan_context;
Assert(plansource->plan == NULL);
plansource->plan = plan;
MemoryContextSwitchTo(oldcxt);
}
/*
* DropCachedPlan: destroy a cached plan.
*
* Actually this only destroys the CachedPlanSource: the referenced CachedPlan
* is released, but not destroyed until its refcount goes to zero. That
* handles the situation where DropCachedPlan is called while the plan is
* still in use.
*/
void
DropCachedPlan(CachedPlanSource *plansource)
{
/* Validity check that we were given a CachedPlanSource */
Assert(list_member_ptr(cached_plans_list, plansource));
/* Remove it from the list */
cached_plans_list = list_delete_ptr(cached_plans_list, plansource);
/* Decrement child CachePlan's refcount and drop if no longer needed */
if (plansource->plan)
ReleaseCachedPlan(plansource->plan, false);
/*
* If CachedPlanSource has independent storage, just drop it. Otherwise
* decrement the refcount on the CachePlan that owns the storage.
*/
if (plansource->orig_plan == NULL)
{
/* Remove the CachedPlanSource and all subsidiary data */
MemoryContextDelete(plansource->context);
}
else
{
Assert(plansource->context == plansource->orig_plan->context);
ReleaseCachedPlan(plansource->orig_plan, false);
}
}
/*
* RevalidateCachedPlan: prepare for re-use of a previously cached plan.
*
* What we do here is re-acquire locks and rebuild the plan if necessary.
* On return, the plan is valid and we have sufficient locks to begin
* execution (or planning, if not fully_planned).
*
* On return, the refcount of the plan has been incremented; a later
* ReleaseCachedPlan() call is expected. The refcount has been reported
* to the CurrentResourceOwner if useResOwner is true.
*
* Note: if any replanning activity is required, the caller's memory context
* is used for that work.
*/
CachedPlan *
RevalidateCachedPlan(CachedPlanSource *plansource, bool useResOwner)
{
CachedPlan *plan;
/* Validity check that we were given a CachedPlanSource */
Assert(list_member_ptr(cached_plans_list, plansource));
/*
* If the plan currently appears valid, acquire locks on the referenced
* objects; then check again. We need to do it this way to cover the
* race condition that an invalidation message arrives before we get
* the lock.
*/
plan = plansource->plan;
if (plan && !plan->dead)
{
/*
* Plan must have positive refcount because it is referenced by
* plansource; so no need to fear it disappears under us here.
*/
Assert(plan->refcount > 0);
if (plan->fully_planned)
AcquireExecutorLocks(plan->stmt_list, true);
else
AcquirePlannerLocks(plan->stmt_list, true);
/*
* By now, if any invalidation has happened, PlanCacheCallback
* will have marked the plan dead.
*/
if (plan->dead)
{
/* Ooops, the race case happened. Release useless locks. */
if (plan->fully_planned)
AcquireExecutorLocks(plan->stmt_list, false);
else
AcquirePlannerLocks(plan->stmt_list, false);
}
}
/*
* If plan has been invalidated, unlink it from the parent and release it.
*/
if (plan && plan->dead)
{
plansource->plan = NULL;
ReleaseCachedPlan(plan, false);
plan = NULL;
}
/*
* Build a new plan if needed.
*/
if (!plan)
{
List *slist;
TupleDesc resultDesc;
/*
* Run parse analysis and rule rewriting. The parser tends to
* scribble on its input, so we must copy the raw parse tree to
* prevent corruption of the cache. Note that we do not use
* parse_analyze_varparams(), assuming that the caller never wants the
* parameter types to change from the original values.
*/
slist = pg_analyze_and_rewrite(copyObject(plansource->raw_parse_tree),
plansource->query_string,
plansource->param_types,
plansource->num_params);
if (plansource->fully_planned)
{
/*
* Generate plans for queries. Assume snapshot is not set yet
* (XXX this may be wasteful, won't all callers have done that?)
*/
slist = pg_plan_queries(slist, NULL, true);
}
/*
* Check or update the result tupdesc. XXX should we use a weaker
* condition than equalTupleDescs() here?
*/
resultDesc = ComputeResultDesc(slist);
if (resultDesc == NULL && plansource->resultDesc == NULL)
{
/* OK, doesn't return tuples */
}
else if (resultDesc == NULL || plansource->resultDesc == NULL ||
!equalTupleDescs(resultDesc, plansource->resultDesc))
{
MemoryContext oldcxt;
/* can we give a better error message? */
if (plansource->fixed_result)
ereport(ERROR,
(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
errmsg("cached plan must not change result type")));
oldcxt = MemoryContextSwitchTo(plansource->context);
if (resultDesc)
resultDesc = CreateTupleDescCopy(resultDesc);
if (plansource->resultDesc)
FreeTupleDesc(plansource->resultDesc);
plansource->resultDesc = resultDesc;
MemoryContextSwitchTo(oldcxt);
}
/*
* Store the plans into the plancache entry, advancing the generation
* count.
*/
StoreCachedPlan(plansource, slist, NULL);
plan = plansource->plan;
}
/*
* Last step: flag the plan as in use by caller.
*/
if (useResOwner)
ResourceOwnerEnlargePlanCacheRefs(CurrentResourceOwner);
plan->refcount++;
if (useResOwner)
ResourceOwnerRememberPlanCacheRef(CurrentResourceOwner, plan);
return plan;
}
/*
* ReleaseCachedPlan: release active use of a cached plan.
*
* This decrements the reference count, and frees the plan if the count
* has thereby gone to zero. If useResOwner is true, it is assumed that
* the reference count is managed by the CurrentResourceOwner.
*
* Note: useResOwner = false is used for releasing references that are in
* persistent data structures, such as the parent CachedPlanSource or a
* Portal. Transient references should be protected by a resource owner.
*/
void
ReleaseCachedPlan(CachedPlan *plan, bool useResOwner)
{
if (useResOwner)
ResourceOwnerForgetPlanCacheRef(CurrentResourceOwner, plan);
Assert(plan->refcount > 0);
plan->refcount--;
if (plan->refcount == 0)
MemoryContextDelete(plan->context);
}
/*
* AcquireExecutorLocks: acquire locks needed for execution of a fully-planned
* cached plan; or release them if acquire is false.
*/
static void
AcquireExecutorLocks(List *stmt_list, bool acquire)
{
ListCell *lc1;
foreach(lc1, stmt_list)
{
PlannedStmt *plannedstmt = (PlannedStmt *) lfirst(lc1);
int rt_index;
ListCell *lc2;
Assert(!IsA(plannedstmt, Query));
if (!IsA(plannedstmt, PlannedStmt))
continue; /* Ignore utility statements */
rt_index = 0;
foreach(lc2, plannedstmt->rtable)
{
RangeTblEntry *rte = (RangeTblEntry *) lfirst(lc2);
LOCKMODE lockmode;
rt_index++;
if (rte->rtekind != RTE_RELATION)
continue;
/*
* Acquire the appropriate type of lock on each relation OID.
* Note that we don't actually try to open the rel, and hence
* will not fail if it's been dropped entirely --- we'll just
* transiently acquire a non-conflicting lock.
*/
if (list_member_int(plannedstmt->resultRelations, rt_index))
lockmode = RowExclusiveLock;
else if (rowmark_member(plannedstmt->rowMarks, rt_index))
lockmode = RowShareLock;
else
lockmode = AccessShareLock;
if (acquire)
LockRelationOid(rte->relid, lockmode);
else
UnlockRelationOid(rte->relid, lockmode);
}
}
}
/*
* AcquirePlannerLocks: acquire locks needed for planning and execution of a
* not-fully-planned cached plan; or release them if acquire is false.
*
* Note that we don't actually try to open the relations, and hence will not
* fail if one has been dropped entirely --- we'll just transiently acquire
* a non-conflicting lock.
*/
static void
AcquirePlannerLocks(List *stmt_list, bool acquire)
{
ListCell *lc;
foreach(lc, stmt_list)
{
Query *query = (Query *) lfirst(lc);
Assert(IsA(query, Query));
if (acquire)
ScanQueryForRelids(query, LockRelid, NULL);
else
ScanQueryForRelids(query, UnlockRelid, NULL);
}
}
/*
* ScanQueryForRelids callback functions for AcquirePlannerLocks
*/
static void
LockRelid(Oid relid, LOCKMODE lockmode, void *arg)
{
LockRelationOid(relid, lockmode);
}
static void
UnlockRelid(Oid relid, LOCKMODE lockmode, void *arg)
{
UnlockRelationOid(relid, lockmode);
}
/*
* ScanQueryForRelids: recursively scan one Query and apply the callback
* function to each relation OID found therein. The callback function
* takes the arguments relation OID, lockmode, pointer arg.
*/
static void
ScanQueryForRelids(Query *parsetree,
void (*callback) (),
void *arg)
{
ListCell *lc;
int rt_index;
/*
* First, process RTEs of the current query level.
*/
rt_index = 0;
foreach(lc, parsetree->rtable)
{
RangeTblEntry *rte = (RangeTblEntry *) lfirst(lc);
LOCKMODE lockmode;
rt_index++;
switch (rte->rtekind)
{
case RTE_RELATION:
/*
* Determine the lock type required for this RTE.
*/
if (rt_index == parsetree->resultRelation)
lockmode = RowExclusiveLock;
else if (rowmark_member(parsetree->rowMarks, rt_index))
lockmode = RowShareLock;
else
lockmode = AccessShareLock;
(*callback) (rte->relid, lockmode, arg);
break;
case RTE_SUBQUERY:
/*
* The subquery RTE itself is all right, but we have to
* recurse to process the represented subquery.
*/
ScanQueryForRelids(rte->subquery, callback, arg);
break;
default:
/* ignore other types of RTEs */
break;
}
}
/*
* Recurse into sublink subqueries, too. But we already did the ones in
* the rtable.
*/
if (parsetree->hasSubLinks)
{
ScanQueryWalkerContext context;
context.callback = callback;
context.arg = arg;
query_tree_walker(parsetree, ScanQueryWalker,
(void *) &context,
QTW_IGNORE_RT_SUBQUERIES);
}
}
/*
* Walker to find sublink subqueries for ScanQueryForRelids
*/
static bool
ScanQueryWalker(Node *node, ScanQueryWalkerContext *context)
{
if (node == NULL)
return false;
if (IsA(node, SubLink))
{
SubLink *sub = (SubLink *) node;
/* Do what we came for */
ScanQueryForRelids((Query *) sub->subselect,
context->callback, context->arg);
/* Fall through to process lefthand args of SubLink */
}
/*
* Do NOT recurse into Query nodes, because ScanQueryForRelids
* already processed subselects of subselects for us.
*/
return expression_tree_walker(node, ScanQueryWalker,
(void *) context);
}
/*
* rowmark_member: check whether an RT index appears in a RowMarkClause list.
*/
static bool
rowmark_member(List *rowMarks, int rt_index)
{
ListCell *l;
foreach(l, rowMarks)
{
RowMarkClause *rc = (RowMarkClause *) lfirst(l);
if (rc->rti == rt_index)
return true;
}
return false;
}
/*
* ComputeResultDesc: given a list of either fully-planned statements or
* Queries, determine the result tupledesc it will produce. Returns NULL
* if the execution will not return tuples.
*
* Note: the result is created or copied into current memory context.
*/
static TupleDesc
ComputeResultDesc(List *stmt_list)
{
Node *node;
Query *query;
PlannedStmt *pstmt;
switch (ChoosePortalStrategy(stmt_list))
{
case PORTAL_ONE_SELECT:
node = (Node *) linitial(stmt_list);
if (IsA(node, Query))
{
query = (Query *) node;
return ExecCleanTypeFromTL(query->targetList, false);
}
if (IsA(node, PlannedStmt))
{
pstmt = (PlannedStmt *) node;
return ExecCleanTypeFromTL(pstmt->planTree->targetlist, false);
}
/* other cases shouldn't happen, but return NULL */
break;
case PORTAL_ONE_RETURNING:
node = PortalListGetPrimaryStmt(stmt_list);
if (IsA(node, Query))
{
query = (Query *) node;
Assert(query->returningList);
return ExecCleanTypeFromTL(query->returningList, false);
}
if (IsA(node, PlannedStmt))
{
pstmt = (PlannedStmt *) node;
Assert(pstmt->returningLists);
return ExecCleanTypeFromTL((List *) linitial(pstmt->returningLists), false);
}
/* other cases shouldn't happen, but return NULL */
break;
case PORTAL_UTIL_SELECT:
node = (Node *) linitial(stmt_list);
if (IsA(node, Query))
{
query = (Query *) node;
Assert(query->utilityStmt);
return UtilityTupleDescriptor(query->utilityStmt);
}
/* else it's a bare utility statement */
return UtilityTupleDescriptor(node);
case PORTAL_MULTI_QUERY:
/* will not return tuples */
break;
}
return NULL;
}
/*
* PlanCacheCallback
* Relcache inval callback function
*/
static void
PlanCacheCallback(Datum arg, Oid relid)
{
ListCell *lc1;
ListCell *lc2;
foreach(lc1, cached_plans_list)
{
CachedPlanSource *plansource = (CachedPlanSource *) lfirst(lc1);
CachedPlan *plan = plansource->plan;
/* No work if it's already invalidated */
if (!plan || plan->dead)
continue;
if (plan->fully_planned)
{
foreach(lc2, plan->stmt_list)
{
PlannedStmt *plannedstmt = (PlannedStmt *) lfirst(lc2);
ListCell *lc3;
Assert(!IsA(plannedstmt, Query));
if (!IsA(plannedstmt, PlannedStmt))
continue; /* Ignore utility statements */
foreach(lc3, plannedstmt->rtable)
{
RangeTblEntry *rte = (RangeTblEntry *) lfirst(lc3);
if (rte->rtekind != RTE_RELATION)
continue;
if (relid == rte->relid)
{
/* Invalidate the plan! */
plan->dead = true;
break; /* out of rangetable scan */
}
}
if (plan->dead)
break; /* out of stmt_list scan */
}
}
else
{
/*
* For not-fully-planned entries we use ScanQueryForRelids,
* since a recursive traversal is needed. The callback API
* is a bit tedious but avoids duplication of coding.
*/
InvalRelidContext context;
context.inval_relid = relid;
context.plan = plan;
foreach(lc2, plan->stmt_list)
{
Query *query = (Query *) lfirst(lc2);
Assert(IsA(query, Query));
ScanQueryForRelids(query, InvalRelid, (void *) &context);
}
}
}
}
/*
* ScanQueryForRelids callback function for PlanCacheCallback
*/
static void
InvalRelid(Oid relid, LOCKMODE lockmode, InvalRelidContext *context)
{
if (relid == context->inval_relid)
context->plan->dead = true;
}