1996-07-09 02:22:35 -04:00
|
|
|
/*-------------------------------------------------------------------------
|
|
|
|
|
*
|
1999-02-13 18:22:53 -05:00
|
|
|
* bootstrap.c
|
1997-09-07 01:04:48 -04:00
|
|
|
* routines to support running postgres in 'bootstrap' mode
|
|
|
|
|
* bootstrap mode is used to create the initial template database
|
1996-07-09 02:22:35 -04:00
|
|
|
*
|
2014-01-07 16:05:30 -05:00
|
|
|
* Portions Copyright (c) 1996-2014, PostgreSQL Global Development Group
|
2000-01-26 00:58:53 -05:00
|
|
|
* Portions Copyright (c) 1994, Regents of the University of California
|
1996-07-09 02:22:35 -04:00
|
|
|
*
|
|
|
|
|
* IDENTIFICATION
|
2010-09-20 16:08:53 -04:00
|
|
|
* src/backend/bootstrap/bootstrap.c
|
1996-07-09 02:22:35 -04:00
|
|
|
*
|
|
|
|
|
*-------------------------------------------------------------------------
|
|
|
|
|
*/
|
2000-11-04 07:43:24 -05:00
|
|
|
#include "postgres.h"
|
|
|
|
|
|
2008-02-16 23:21:05 -05:00
|
|
|
#include <time.h>
|
1999-07-15 23:14:30 -04:00
|
|
|
#include <unistd.h>
|
1996-10-23 03:42:13 -04:00
|
|
|
#include <signal.h>
|
1998-04-26 00:12:15 -04:00
|
|
|
|
2012-08-30 16:15:44 -04:00
|
|
|
#include "access/htup_details.h"
|
1998-04-26 00:12:15 -04:00
|
|
|
#include "bootstrap/bootstrap.h"
|
|
|
|
|
#include "catalog/index.h"
|
2011-02-08 16:04:18 -05:00
|
|
|
#include "catalog/pg_collation.h"
|
1998-04-26 00:12:15 -04:00
|
|
|
#include "catalog/pg_type.h"
|
|
|
|
|
#include "libpq/pqsignal.h"
|
1999-07-16 01:00:38 -04:00
|
|
|
#include "miscadmin.h"
|
2005-04-14 16:03:27 -04:00
|
|
|
#include "nodes/makefuncs.h"
|
2014-02-15 14:31:30 -05:00
|
|
|
#include "pg_getopt.h"
|
2004-05-29 18:48:23 -04:00
|
|
|
#include "postmaster/bgwriter.h"
|
2011-11-02 10:25:01 -04:00
|
|
|
#include "postmaster/startup.h"
|
2007-07-24 00:54:09 -04:00
|
|
|
#include "postmaster/walwriter.h"
|
2010-01-15 04:19:10 -05:00
|
|
|
#include "replication/walreceiver.h"
|
2008-05-11 20:00:54 -04:00
|
|
|
#include "storage/bufmgr.h"
|
2013-04-30 07:27:12 -04:00
|
|
|
#include "storage/bufpage.h"
|
2002-05-04 20:03:29 -04:00
|
|
|
#include "storage/ipc.h"
|
2011-09-04 01:13:16 -04:00
|
|
|
#include "storage/proc.h"
|
1998-04-26 00:12:15 -04:00
|
|
|
#include "tcop/tcopprot.h"
|
|
|
|
|
#include "utils/builtins.h"
|
2000-05-28 13:56:29 -04:00
|
|
|
#include "utils/fmgroids.h"
|
2005-05-06 13:24:55 -04:00
|
|
|
#include "utils/memutils.h"
|
2004-05-28 01:13:32 -04:00
|
|
|
#include "utils/ps_status.h"
|
2011-02-23 12:18:09 -05:00
|
|
|
#include "utils/rel.h"
|
2010-02-07 15:48:13 -05:00
|
|
|
#include "utils/relmapper.h"
|
2008-03-26 17:10:39 -04:00
|
|
|
#include "utils/tqual.h"
|
2000-06-27 23:33:33 -04:00
|
|
|
|
2013-05-29 16:58:43 -04:00
|
|
|
uint32 bootstrap_data_checksum_version = 0; /* No checksum */
|
2013-03-22 09:54:07 -04:00
|
|
|
|
1996-10-21 04:31:23 -04:00
|
|
|
|
2000-06-05 03:29:25 -04:00
|
|
|
#define ALLOC(t, c) ((t *) calloc((unsigned)(c), sizeof(t)))
|
1996-07-09 02:22:35 -04:00
|
|
|
|
2007-03-07 08:35:03 -05:00
|
|
|
static void CheckerModeMain(void);
|
|
|
|
|
static void BootstrapModeMain(void);
|
2004-05-29 18:48:23 -04:00
|
|
|
static void bootstrap_signals(void);
|
2007-03-07 08:35:03 -05:00
|
|
|
static void ShutdownAuxiliaryProcess(int code, Datum arg);
|
1998-08-31 23:29:17 -04:00
|
|
|
static Form_pg_attribute AllocateAttribute(void);
|
1998-08-24 15:04:04 -04:00
|
|
|
static Oid gettype(char *type);
|
1997-09-07 22:41:22 -04:00
|
|
|
static void cleanup(void);
|
1996-10-23 03:42:13 -04:00
|
|
|
|
1996-07-09 02:22:35 -04:00
|
|
|
/* ----------------
|
1997-09-07 01:04:48 -04:00
|
|
|
* global variables
|
1996-07-09 02:22:35 -04:00
|
|
|
* ----------------
|
|
|
|
|
*/
|
2002-04-27 17:24:34 -04:00
|
|
|
|
2013-05-29 16:58:43 -04:00
|
|
|
AuxProcType MyAuxProcType = NotAnAuxProcess; /* declared in miscadmin.h */
|
Fix management of pendingOpsTable in auxiliary processes.
mdinit() was misusing IsBootstrapProcessingMode() to decide whether to
create an fsync pending-operations table in the current process. This led
to creating a table not only in the startup and checkpointer processes as
intended, but also in the bgwriter process, not to mention other auxiliary
processes such as walwriter and walreceiver. Creation of the table in the
bgwriter is fatal, because it absorbs fsync requests that should have gone
to the checkpointer; instead they just sit in bgwriter local memory and are
never acted on. So writes performed by the bgwriter were not being fsync'd
which could result in data loss after an OS crash. I think there is no
live bug with respect to walwriter and walreceiver because those never
perform any writes of shared buffers; but the potential is there for
future breakage in those processes too.
To fix, make AuxiliaryProcessMain() export the current process's
AuxProcType as a global variable, and then make mdinit() test directly for
the types of aux process that should have a pendingOpsTable. Having done
that, we might as well also get rid of the random bool flags such as
am_walreceiver that some of the aux processes had grown. (Note that we
could not have fixed the bug by examining those variables in mdinit(),
because it's called from BaseInit() which is run by AuxiliaryProcessMain()
before entering any of the process-type-specific code.)
Back-patch to 9.2, where the problem was introduced by the split-up of
bgwriter and checkpointer processes. The bogus pendingOpsTable exists
in walwriter and walreceiver processes in earlier branches, but absent
any evidence that it causes actual problems there, I'll leave the older
branches alone.
2012-07-18 15:28:10 -04:00
|
|
|
|
2002-04-27 17:24:34 -04:00
|
|
|
Relation boot_reldesc; /* current relation descriptor */
|
|
|
|
|
|
2009-09-26 21:32:11 -04:00
|
|
|
Form_pg_attribute attrtypes[MAXATTR]; /* points to attribute info */
|
|
|
|
|
int numattr; /* number of attributes for cur. rel */
|
1996-07-09 02:22:35 -04:00
|
|
|
|
|
|
|
|
|
2009-09-26 21:32:11 -04:00
|
|
|
/*
|
1996-07-09 02:22:35 -04:00
|
|
|
* Basic information associated with each type. This is used before
|
2010-09-02 21:34:55 -04:00
|
|
|
* pg_type is filled, so it has to cover the datatypes used as column types
|
|
|
|
|
* in the core "bootstrapped" catalogs.
|
1996-07-09 02:22:35 -04:00
|
|
|
*
|
1997-09-07 01:04:48 -04:00
|
|
|
* XXX several of these input/output functions do catalog scans
|
2014-05-06 12:12:18 -04:00
|
|
|
* (e.g., F_REGPROCIN scans pg_proc). this obviously creates some
|
1997-09-07 01:04:48 -04:00
|
|
|
* order dependencies in the catalog creation process.
|
1996-07-09 02:22:35 -04:00
|
|
|
*/
|
1997-09-07 01:04:48 -04:00
|
|
|
struct typinfo
|
|
|
|
|
{
|
1997-09-07 22:41:22 -04:00
|
|
|
char name[NAMEDATALEN];
|
|
|
|
|
Oid oid;
|
|
|
|
|
Oid elem;
|
|
|
|
|
int16 len;
|
2004-04-01 16:28:47 -05:00
|
|
|
bool byval;
|
|
|
|
|
char align;
|
|
|
|
|
char storage;
|
2011-02-08 16:04:18 -05:00
|
|
|
Oid collation;
|
1997-09-07 22:41:22 -04:00
|
|
|
Oid inproc;
|
|
|
|
|
Oid outproc;
|
1996-07-09 02:22:35 -04:00
|
|
|
};
|
|
|
|
|
|
2004-04-01 16:28:47 -05:00
|
|
|
static const struct typinfo TypInfo[] = {
|
2011-02-08 16:04:18 -05:00
|
|
|
{"bool", BOOLOID, 0, 1, true, 'c', 'p', InvalidOid,
|
2004-08-29 01:07:03 -04:00
|
|
|
F_BOOLIN, F_BOOLOUT},
|
2011-02-08 16:04:18 -05:00
|
|
|
{"bytea", BYTEAOID, 0, -1, false, 'i', 'x', InvalidOid,
|
2004-08-29 01:07:03 -04:00
|
|
|
F_BYTEAIN, F_BYTEAOUT},
|
2011-02-08 16:04:18 -05:00
|
|
|
{"char", CHAROID, 0, 1, true, 'c', 'p', InvalidOid,
|
2004-08-29 01:07:03 -04:00
|
|
|
F_CHARIN, F_CHAROUT},
|
2011-02-08 16:04:18 -05:00
|
|
|
{"int2", INT2OID, 0, 2, true, 's', 'p', InvalidOid,
|
2004-08-29 01:07:03 -04:00
|
|
|
F_INT2IN, F_INT2OUT},
|
2011-02-08 16:04:18 -05:00
|
|
|
{"int4", INT4OID, 0, 4, true, 'i', 'p', InvalidOid,
|
2004-08-29 01:07:03 -04:00
|
|
|
F_INT4IN, F_INT4OUT},
|
2011-02-08 16:04:18 -05:00
|
|
|
{"float4", FLOAT4OID, 0, 4, FLOAT4PASSBYVAL, 'i', 'p', InvalidOid,
|
2007-01-21 20:35:23 -05:00
|
|
|
F_FLOAT4IN, F_FLOAT4OUT},
|
2011-02-08 16:04:18 -05:00
|
|
|
{"name", NAMEOID, CHAROID, NAMEDATALEN, false, 'c', 'p', InvalidOid,
|
2007-01-21 20:35:23 -05:00
|
|
|
F_NAMEIN, F_NAMEOUT},
|
2011-02-08 16:04:18 -05:00
|
|
|
{"regclass", REGCLASSOID, 0, 4, true, 'i', 'p', InvalidOid,
|
2004-08-29 01:07:03 -04:00
|
|
|
F_REGCLASSIN, F_REGCLASSOUT},
|
2011-02-08 16:04:18 -05:00
|
|
|
{"regproc", REGPROCOID, 0, 4, true, 'i', 'p', InvalidOid,
|
2007-01-21 20:35:23 -05:00
|
|
|
F_REGPROCIN, F_REGPROCOUT},
|
2011-02-08 16:04:18 -05:00
|
|
|
{"regtype", REGTYPEOID, 0, 4, true, 'i', 'p', InvalidOid,
|
2004-08-29 01:07:03 -04:00
|
|
|
F_REGTYPEIN, F_REGTYPEOUT},
|
2011-02-08 16:04:18 -05:00
|
|
|
{"text", TEXTOID, 0, -1, false, 'i', 'x', DEFAULT_COLLATION_OID,
|
2004-08-29 01:07:03 -04:00
|
|
|
F_TEXTIN, F_TEXTOUT},
|
2011-02-08 16:04:18 -05:00
|
|
|
{"oid", OIDOID, 0, 4, true, 'i', 'p', InvalidOid,
|
2004-08-29 01:07:03 -04:00
|
|
|
F_OIDIN, F_OIDOUT},
|
2011-02-08 16:04:18 -05:00
|
|
|
{"tid", TIDOID, 0, 6, false, 's', 'p', InvalidOid,
|
2004-08-29 01:07:03 -04:00
|
|
|
F_TIDIN, F_TIDOUT},
|
2011-02-08 16:04:18 -05:00
|
|
|
{"xid", XIDOID, 0, 4, true, 'i', 'p', InvalidOid,
|
2004-08-29 01:07:03 -04:00
|
|
|
F_XIDIN, F_XIDOUT},
|
2011-02-08 16:04:18 -05:00
|
|
|
{"cid", CIDOID, 0, 4, true, 'i', 'p', InvalidOid,
|
2004-08-29 01:07:03 -04:00
|
|
|
F_CIDIN, F_CIDOUT},
|
2011-02-08 16:04:18 -05:00
|
|
|
{"pg_node_tree", PGNODETREEOID, 0, -1, false, 'i', 'x', DEFAULT_COLLATION_OID,
|
2010-09-02 21:34:55 -04:00
|
|
|
F_PG_NODE_TREE_IN, F_PG_NODE_TREE_OUT},
|
2011-02-08 16:04:18 -05:00
|
|
|
{"int2vector", INT2VECTOROID, INT2OID, -1, false, 'i', 'p', InvalidOid,
|
2004-08-29 01:07:03 -04:00
|
|
|
F_INT2VECTORIN, F_INT2VECTOROUT},
|
2011-02-08 16:04:18 -05:00
|
|
|
{"oidvector", OIDVECTOROID, OIDOID, -1, false, 'i', 'p', InvalidOid,
|
2004-08-29 01:07:03 -04:00
|
|
|
F_OIDVECTORIN, F_OIDVECTOROUT},
|
2011-02-08 16:04:18 -05:00
|
|
|
{"_int4", INT4ARRAYOID, INT4OID, -1, false, 'i', 'x', InvalidOid,
|
2004-08-29 01:07:03 -04:00
|
|
|
F_ARRAY_IN, F_ARRAY_OUT},
|
2011-02-08 16:04:18 -05:00
|
|
|
{"_text", 1009, TEXTOID, -1, false, 'i', 'x', DEFAULT_COLLATION_OID,
|
2004-08-29 01:07:03 -04:00
|
|
|
F_ARRAY_IN, F_ARRAY_OUT},
|
2011-02-08 16:04:18 -05:00
|
|
|
{"_oid", 1028, OIDOID, -1, false, 'i', 'x', InvalidOid,
|
2005-03-29 14:44:23 -05:00
|
|
|
F_ARRAY_IN, F_ARRAY_OUT},
|
2011-02-08 16:04:18 -05:00
|
|
|
{"_char", 1002, CHAROID, -1, false, 'i', 'x', InvalidOid,
|
2005-03-29 14:44:23 -05:00
|
|
|
F_ARRAY_IN, F_ARRAY_OUT},
|
2011-02-08 16:04:18 -05:00
|
|
|
{"_aclitem", 1034, ACLITEMOID, -1, false, 'i', 'x', InvalidOid,
|
2004-08-29 01:07:03 -04:00
|
|
|
F_ARRAY_IN, F_ARRAY_OUT}
|
1996-07-09 02:22:35 -04:00
|
|
|
};
|
|
|
|
|
|
2004-08-29 01:07:03 -04:00
|
|
|
static const int n_types = sizeof(TypInfo) / sizeof(struct typinfo);
|
1996-07-09 02:22:35 -04:00
|
|
|
|
1997-09-07 01:04:48 -04:00
|
|
|
struct typmap
|
|
|
|
|
{ /* a hack */
|
1997-09-07 22:41:22 -04:00
|
|
|
Oid am_oid;
|
1998-08-31 23:29:17 -04:00
|
|
|
FormData_pg_type am_typ;
|
1996-07-09 02:22:35 -04:00
|
|
|
};
|
|
|
|
|
|
2004-01-07 13:56:30 -05:00
|
|
|
static struct typmap **Typ = NULL;
|
|
|
|
|
static struct typmap *Ap = NULL;
|
1997-09-07 01:04:48 -04:00
|
|
|
|
2009-09-26 21:32:11 -04:00
|
|
|
static Datum values[MAXATTR]; /* current row's attribute values */
|
2008-11-01 21:45:28 -04:00
|
|
|
static bool Nulls[MAXATTR];
|
1997-09-07 01:04:48 -04:00
|
|
|
|
2001-03-22 01:16:21 -05:00
|
|
|
static MemoryContext nogc = NULL; /* special no-gc mem context */
|
1997-09-07 01:04:48 -04:00
|
|
|
|
1996-07-09 02:22:35 -04:00
|
|
|
/*
|
1997-09-07 01:04:48 -04:00
|
|
|
* At bootstrap time, we first declare all the indices to be built, and
|
|
|
|
|
* then build them. The IndexList structure stores enough information
|
|
|
|
|
* to allow us to build the indices after they've been declared.
|
1996-07-09 02:22:35 -04:00
|
|
|
*/
|
|
|
|
|
|
1997-09-07 01:04:48 -04:00
|
|
|
typedef struct _IndexList
|
|
|
|
|
{
|
2002-03-26 14:17:02 -05:00
|
|
|
Oid il_heap;
|
|
|
|
|
Oid il_ind;
|
2000-07-14 18:18:02 -04:00
|
|
|
IndexInfo *il_info;
|
1997-09-07 01:04:48 -04:00
|
|
|
struct _IndexList *il_next;
|
1997-09-08 17:56:23 -04:00
|
|
|
} IndexList;
|
1996-07-09 02:22:35 -04:00
|
|
|
|
2004-01-07 13:56:30 -05:00
|
|
|
static IndexList *ILHead = NULL;
|
1997-09-07 01:04:48 -04:00
|
|
|
|
1996-07-09 02:22:35 -04:00
|
|
|
|
2004-05-28 01:13:32 -04:00
|
|
|
/*
|
2007-03-07 08:35:03 -05:00
|
|
|
* AuxiliaryProcessMain
|
2004-05-28 01:13:32 -04:00
|
|
|
*
|
2007-03-07 08:35:03 -05:00
|
|
|
* The main entry point for auxiliary processes, such as the bgwriter,
|
2010-01-15 04:19:10 -05:00
|
|
|
* walwriter, walreceiver, bootstrapper and the shared memory checker code.
|
1996-07-09 02:22:35 -04:00
|
|
|
*
|
2007-03-07 08:35:03 -05:00
|
|
|
* This code is here just because of historical reasons.
|
1996-07-09 02:22:35 -04:00
|
|
|
*/
|
2007-03-07 08:35:03 -05:00
|
|
|
void
|
|
|
|
|
AuxiliaryProcessMain(int argc, char *argv[])
|
1996-07-09 02:22:35 -04:00
|
|
|
{
|
2005-01-14 16:08:44 -05:00
|
|
|
char *progname = argv[0];
|
1997-09-07 22:41:22 -04:00
|
|
|
int flag;
|
2004-10-07 21:36:36 -04:00
|
|
|
char *userDoption = NULL;
|
1997-09-07 01:04:48 -04:00
|
|
|
|
2001-03-22 01:16:21 -05:00
|
|
|
/*
|
|
|
|
|
* initialize globals
|
1997-09-07 01:04:48 -04:00
|
|
|
*/
|
1998-01-25 00:15:15 -05:00
|
|
|
MyProcPid = getpid();
|
1997-09-07 01:04:48 -04:00
|
|
|
|
2007-08-02 19:39:45 -04:00
|
|
|
MyStartTime = time(NULL);
|
|
|
|
|
|
2004-07-31 13:57:11 -04:00
|
|
|
/* Compute paths, if we didn't inherit them from postmaster */
|
|
|
|
|
if (my_exec_path[0] == '\0')
|
|
|
|
|
{
|
2005-01-14 16:08:44 -05:00
|
|
|
if (find_my_exec(progname, my_exec_path) < 0)
|
2004-07-31 13:57:11 -04:00
|
|
|
elog(FATAL, "%s: could not locate my own executable path",
|
2005-01-14 16:08:44 -05:00
|
|
|
progname);
|
2004-07-31 13:57:11 -04:00
|
|
|
}
|
|
|
|
|
|
2001-03-22 01:16:21 -05:00
|
|
|
/*
|
|
|
|
|
* process command arguments
|
1997-09-07 01:04:48 -04:00
|
|
|
*/
|
|
|
|
|
|
|
|
|
|
/* Set defaults, to be overriden by explicit options below */
|
2003-12-24 22:52:51 -05:00
|
|
|
if (!IsUnderPostmaster)
|
2002-05-16 21:19:19 -04:00
|
|
|
InitializeGUCOptions();
|
1997-09-07 01:04:48 -04:00
|
|
|
|
2006-06-18 11:38:37 -04:00
|
|
|
/* Ignore the initial --boot argument, if present */
|
|
|
|
|
if (argc > 1 && strcmp(argv[1], "--boot") == 0)
|
2004-05-28 01:13:32 -04:00
|
|
|
{
|
|
|
|
|
argv++;
|
|
|
|
|
argc--;
|
|
|
|
|
}
|
|
|
|
|
|
Fix management of pendingOpsTable in auxiliary processes.
mdinit() was misusing IsBootstrapProcessingMode() to decide whether to
create an fsync pending-operations table in the current process. This led
to creating a table not only in the startup and checkpointer processes as
intended, but also in the bgwriter process, not to mention other auxiliary
processes such as walwriter and walreceiver. Creation of the table in the
bgwriter is fatal, because it absorbs fsync requests that should have gone
to the checkpointer; instead they just sit in bgwriter local memory and are
never acted on. So writes performed by the bgwriter were not being fsync'd
which could result in data loss after an OS crash. I think there is no
live bug with respect to walwriter and walreceiver because those never
perform any writes of shared buffers; but the potential is there for
future breakage in those processes too.
To fix, make AuxiliaryProcessMain() export the current process's
AuxProcType as a global variable, and then make mdinit() test directly for
the types of aux process that should have a pendingOpsTable. Having done
that, we might as well also get rid of the random bool flags such as
am_walreceiver that some of the aux processes had grown. (Note that we
could not have fixed the bug by examining those variables in mdinit(),
because it's called from BaseInit() which is run by AuxiliaryProcessMain()
before entering any of the process-type-specific code.)
Back-patch to 9.2, where the problem was introduced by the split-up of
bgwriter and checkpointer processes. The bogus pendingOpsTable exists
in walwriter and walreceiver processes in earlier branches, but absent
any evidence that it causes actual problems there, I'll leave the older
branches alone.
2012-07-18 15:28:10 -04:00
|
|
|
/* If no -x argument, we are a CheckerProcess */
|
|
|
|
|
MyAuxProcType = CheckerProcess;
|
|
|
|
|
|
2013-03-22 09:54:07 -04:00
|
|
|
while ((flag = getopt(argc, argv, "B:c:d:D:Fkr:x:-:")) != -1)
|
1997-09-07 01:04:48 -04:00
|
|
|
{
|
|
|
|
|
switch (flag)
|
|
|
|
|
{
|
2006-01-05 05:07:46 -05:00
|
|
|
case 'B':
|
|
|
|
|
SetConfigOption("shared_buffers", optarg, PGC_POSTMASTER, PGC_S_ARGV);
|
|
|
|
|
break;
|
1997-09-07 22:41:22 -04:00
|
|
|
case 'D':
|
2012-10-12 13:35:40 -04:00
|
|
|
userDoption = strdup(optarg);
|
1997-09-07 22:41:22 -04:00
|
|
|
break;
|
|
|
|
|
case 'd':
|
2002-09-04 16:31:48 -04:00
|
|
|
{
|
|
|
|
|
/* Turn on debugging for the bootstrap process. */
|
2013-10-13 00:09:18 -04:00
|
|
|
char *debugstr;
|
2002-09-04 16:31:48 -04:00
|
|
|
|
2013-10-13 00:09:18 -04:00
|
|
|
debugstr = psprintf("debug%s", optarg);
|
2002-11-14 18:53:27 -05:00
|
|
|
SetConfigOption("log_min_messages", debugstr,
|
2002-09-04 16:31:48 -04:00
|
|
|
PGC_POSTMASTER, PGC_S_ARGV);
|
|
|
|
|
SetConfigOption("client_min_messages", debugstr,
|
|
|
|
|
PGC_POSTMASTER, PGC_S_ARGV);
|
|
|
|
|
pfree(debugstr);
|
|
|
|
|
}
|
1997-09-07 22:41:22 -04:00
|
|
|
break;
|
|
|
|
|
case 'F':
|
2002-02-22 20:31:37 -05:00
|
|
|
SetConfigOption("fsync", "false", PGC_POSTMASTER, PGC_S_ARGV);
|
1997-09-07 22:41:22 -04:00
|
|
|
break;
|
2013-03-22 09:54:07 -04:00
|
|
|
case 'k':
|
2013-04-30 07:27:12 -04:00
|
|
|
bootstrap_data_checksum_version = PG_DATA_CHECKSUM_VERSION;
|
2013-03-22 09:54:07 -04:00
|
|
|
break;
|
2006-01-05 05:07:46 -05:00
|
|
|
case 'r':
|
2007-02-10 09:58:55 -05:00
|
|
|
strlcpy(OutputFileName, optarg, MAXPGPATH);
|
2001-08-24 20:31:17 -04:00
|
|
|
break;
|
1999-10-06 17:58:18 -04:00
|
|
|
case 'x':
|
Fix management of pendingOpsTable in auxiliary processes.
mdinit() was misusing IsBootstrapProcessingMode() to decide whether to
create an fsync pending-operations table in the current process. This led
to creating a table not only in the startup and checkpointer processes as
intended, but also in the bgwriter process, not to mention other auxiliary
processes such as walwriter and walreceiver. Creation of the table in the
bgwriter is fatal, because it absorbs fsync requests that should have gone
to the checkpointer; instead they just sit in bgwriter local memory and are
never acted on. So writes performed by the bgwriter were not being fsync'd
which could result in data loss after an OS crash. I think there is no
live bug with respect to walwriter and walreceiver because those never
perform any writes of shared buffers; but the potential is there for
future breakage in those processes too.
To fix, make AuxiliaryProcessMain() export the current process's
AuxProcType as a global variable, and then make mdinit() test directly for
the types of aux process that should have a pendingOpsTable. Having done
that, we might as well also get rid of the random bool flags such as
am_walreceiver that some of the aux processes had grown. (Note that we
could not have fixed the bug by examining those variables in mdinit(),
because it's called from BaseInit() which is run by AuxiliaryProcessMain()
before entering any of the process-type-specific code.)
Back-patch to 9.2, where the problem was introduced by the split-up of
bgwriter and checkpointer processes. The bogus pendingOpsTable exists
in walwriter and walreceiver processes in earlier branches, but absent
any evidence that it causes actual problems there, I'll leave the older
branches alone.
2012-07-18 15:28:10 -04:00
|
|
|
MyAuxProcType = atoi(optarg);
|
1999-10-06 17:58:18 -04:00
|
|
|
break;
|
2003-07-14 20:11:14 -04:00
|
|
|
case 'c':
|
|
|
|
|
case '-':
|
|
|
|
|
{
|
|
|
|
|
char *name,
|
|
|
|
|
*value;
|
|
|
|
|
|
|
|
|
|
ParseLongOption(optarg, &name, &value);
|
|
|
|
|
if (!value)
|
|
|
|
|
{
|
|
|
|
|
if (flag == '-')
|
2003-07-22 19:30:39 -04:00
|
|
|
ereport(ERROR,
|
|
|
|
|
(errcode(ERRCODE_SYNTAX_ERROR),
|
|
|
|
|
errmsg("--%s requires a value",
|
|
|
|
|
optarg)));
|
2003-07-14 20:11:14 -04:00
|
|
|
else
|
2003-07-22 19:30:39 -04:00
|
|
|
ereport(ERROR,
|
|
|
|
|
(errcode(ERRCODE_SYNTAX_ERROR),
|
|
|
|
|
errmsg("-c %s requires a value",
|
|
|
|
|
optarg)));
|
2003-07-14 20:11:14 -04:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
SetConfigOption(name, value, PGC_POSTMASTER, PGC_S_ARGV);
|
|
|
|
|
free(name);
|
|
|
|
|
if (value)
|
|
|
|
|
free(value);
|
|
|
|
|
break;
|
|
|
|
|
}
|
1997-09-07 22:41:22 -04:00
|
|
|
default:
|
2006-06-18 11:38:37 -04:00
|
|
|
write_stderr("Try \"%s --help\" for more information.\n",
|
|
|
|
|
progname);
|
|
|
|
|
proc_exit(1);
|
1997-09-07 22:41:22 -04:00
|
|
|
break;
|
1997-09-07 01:04:48 -04:00
|
|
|
}
|
2003-05-06 19:34:56 -04:00
|
|
|
}
|
1997-09-07 01:04:48 -04:00
|
|
|
|
2007-02-15 21:10:07 -05:00
|
|
|
if (argc != optind)
|
2006-06-18 11:38:37 -04:00
|
|
|
{
|
|
|
|
|
write_stderr("%s: invalid command-line arguments\n", progname);
|
|
|
|
|
proc_exit(1);
|
|
|
|
|
}
|
1997-09-07 01:04:48 -04:00
|
|
|
|
2004-05-28 01:13:32 -04:00
|
|
|
/*
|
|
|
|
|
* Identify myself via ps
|
|
|
|
|
*/
|
2003-09-02 15:04:12 -04:00
|
|
|
if (IsUnderPostmaster)
|
|
|
|
|
{
|
2004-05-28 01:13:32 -04:00
|
|
|
const char *statmsg;
|
2003-12-24 22:52:51 -05:00
|
|
|
|
Fix management of pendingOpsTable in auxiliary processes.
mdinit() was misusing IsBootstrapProcessingMode() to decide whether to
create an fsync pending-operations table in the current process. This led
to creating a table not only in the startup and checkpointer processes as
intended, but also in the bgwriter process, not to mention other auxiliary
processes such as walwriter and walreceiver. Creation of the table in the
bgwriter is fatal, because it absorbs fsync requests that should have gone
to the checkpointer; instead they just sit in bgwriter local memory and are
never acted on. So writes performed by the bgwriter were not being fsync'd
which could result in data loss after an OS crash. I think there is no
live bug with respect to walwriter and walreceiver because those never
perform any writes of shared buffers; but the potential is there for
future breakage in those processes too.
To fix, make AuxiliaryProcessMain() export the current process's
AuxProcType as a global variable, and then make mdinit() test directly for
the types of aux process that should have a pendingOpsTable. Having done
that, we might as well also get rid of the random bool flags such as
am_walreceiver that some of the aux processes had grown. (Note that we
could not have fixed the bug by examining those variables in mdinit(),
because it's called from BaseInit() which is run by AuxiliaryProcessMain()
before entering any of the process-type-specific code.)
Back-patch to 9.2, where the problem was introduced by the split-up of
bgwriter and checkpointer processes. The bogus pendingOpsTable exists
in walwriter and walreceiver processes in earlier branches, but absent
any evidence that it causes actual problems there, I'll leave the older
branches alone.
2012-07-18 15:28:10 -04:00
|
|
|
switch (MyAuxProcType)
|
2004-05-28 01:13:32 -04:00
|
|
|
{
|
2007-03-07 08:35:03 -05:00
|
|
|
case StartupProcess:
|
2004-05-29 18:48:23 -04:00
|
|
|
statmsg = "startup process";
|
2004-05-28 01:13:32 -04:00
|
|
|
break;
|
2007-03-07 08:35:03 -05:00
|
|
|
case BgWriterProcess:
|
2004-05-29 18:48:23 -04:00
|
|
|
statmsg = "writer process";
|
2004-05-28 01:13:32 -04:00
|
|
|
break;
|
2011-11-01 13:14:47 -04:00
|
|
|
case CheckpointerProcess:
|
|
|
|
|
statmsg = "checkpointer process";
|
|
|
|
|
break;
|
2007-07-24 00:54:09 -04:00
|
|
|
case WalWriterProcess:
|
|
|
|
|
statmsg = "wal writer process";
|
|
|
|
|
break;
|
2010-01-15 04:19:10 -05:00
|
|
|
case WalReceiverProcess:
|
|
|
|
|
statmsg = "wal receiver process";
|
|
|
|
|
break;
|
2004-05-28 01:13:32 -04:00
|
|
|
default:
|
2004-05-29 18:48:23 -04:00
|
|
|
statmsg = "??? process";
|
2004-05-28 01:13:32 -04:00
|
|
|
break;
|
|
|
|
|
}
|
2006-06-27 18:16:44 -04:00
|
|
|
init_ps_display(statmsg, "", "", "");
|
2003-09-02 15:04:12 -04:00
|
|
|
}
|
2004-05-28 01:13:32 -04:00
|
|
|
|
|
|
|
|
/* Acquire configuration parameters, unless inherited from postmaster */
|
|
|
|
|
if (!IsUnderPostmaster)
|
|
|
|
|
{
|
2005-01-14 16:08:44 -05:00
|
|
|
if (!SelectConfigFiles(userDoption, progname))
|
2004-10-07 21:36:36 -04:00
|
|
|
proc_exit(1);
|
2004-05-28 01:13:32 -04:00
|
|
|
}
|
2004-05-21 01:08:06 -04:00
|
|
|
|
2004-10-07 21:36:36 -04:00
|
|
|
/* Validate we have been given a reasonable-looking DataDir */
|
|
|
|
|
Assert(DataDir);
|
|
|
|
|
ValidatePgVersion(DataDir);
|
|
|
|
|
|
2005-07-04 00:51:52 -04:00
|
|
|
/* Change into DataDir (if under postmaster, should be done already) */
|
|
|
|
|
if (!IsUnderPostmaster)
|
|
|
|
|
ChangeToDataDir();
|
|
|
|
|
|
2004-05-29 18:48:23 -04:00
|
|
|
/* If standalone, create lockfile for data directory */
|
|
|
|
|
if (!IsUnderPostmaster)
|
2005-07-04 00:51:52 -04:00
|
|
|
CreateDataDirLockFile(false);
|
1999-10-06 17:58:18 -04:00
|
|
|
|
XLOG (and related) changes:
* Store two past checkpoint locations, not just one, in pg_control.
On startup, we fall back to the older checkpoint if the newer one
is unreadable. Also, a physical copy of the newest checkpoint record
is kept in pg_control for possible use in disaster recovery (ie,
complete loss of pg_xlog). Also add a version number for pg_control
itself. Remove archdir from pg_control; it ought to be a GUC
parameter, not a special case (not that it's implemented yet anyway).
* Suppress successive checkpoint records when nothing has been entered
in the WAL log since the last one. This is not so much to avoid I/O
as to make it actually useful to keep track of the last two
checkpoints. If the things are right next to each other then there's
not a lot of redundancy gained...
* Change CRC scheme to a true 64-bit CRC, not a pair of 32-bit CRCs
on alternate bytes. Polynomial borrowed from ECMA DLT1 standard.
* Fix XLOG record length handling so that it will work at BLCKSZ = 32k.
* Change XID allocation to work more like OID allocation. (This is of
dubious necessity, but I think it's a good idea anyway.)
* Fix a number of minor bugs, such as off-by-one logic for XLOG file
wraparound at the 4 gig mark.
* Add documentation and clean up some coding infelicities; move file
format declarations out to include files where planned contrib
utilities can get at them.
* Checkpoint will now occur every CHECKPOINT_SEGMENTS log segments or
every CHECKPOINT_TIMEOUT seconds, whichever comes first. It is also
possible to force a checkpoint by sending SIGUSR1 to the postmaster
(undocumented feature...)
* Defend against kill -9 postmaster by storing shmem block's key and ID
in postmaster.pid lockfile, and checking at startup to ensure that no
processes are still connected to old shmem block (if it still exists).
* Switch backends to accept SIGQUIT rather than SIGUSR1 for emergency
stop, for symmetry with postmaster and xlog utilities. Clean up signal
handling in bootstrap.c so that xlog utilities launched by postmaster
will react to signals better.
* Standalone bootstrap now grabs lockfile in target directory, as added
insurance against running it in parallel with live postmaster.
2001-03-12 20:17:06 -05:00
|
|
|
SetProcessingMode(BootstrapProcessing);
|
2006-01-05 05:07:46 -05:00
|
|
|
IgnoreSystemIndexes = true;
|
XLOG (and related) changes:
* Store two past checkpoint locations, not just one, in pg_control.
On startup, we fall back to the older checkpoint if the newer one
is unreadable. Also, a physical copy of the newest checkpoint record
is kept in pg_control for possible use in disaster recovery (ie,
complete loss of pg_xlog). Also add a version number for pg_control
itself. Remove archdir from pg_control; it ought to be a GUC
parameter, not a special case (not that it's implemented yet anyway).
* Suppress successive checkpoint records when nothing has been entered
in the WAL log since the last one. This is not so much to avoid I/O
as to make it actually useful to keep track of the last two
checkpoints. If the things are right next to each other then there's
not a lot of redundancy gained...
* Change CRC scheme to a true 64-bit CRC, not a pair of 32-bit CRCs
on alternate bytes. Polynomial borrowed from ECMA DLT1 standard.
* Fix XLOG record length handling so that it will work at BLCKSZ = 32k.
* Change XID allocation to work more like OID allocation. (This is of
dubious necessity, but I think it's a good idea anyway.)
* Fix a number of minor bugs, such as off-by-one logic for XLOG file
wraparound at the 4 gig mark.
* Add documentation and clean up some coding infelicities; move file
format declarations out to include files where planned contrib
utilities can get at them.
* Checkpoint will now occur every CHECKPOINT_SEGMENTS log segments or
every CHECKPOINT_TIMEOUT seconds, whichever comes first. It is also
possible to force a checkpoint by sending SIGUSR1 to the postmaster
(undocumented feature...)
* Defend against kill -9 postmaster by storing shmem block's key and ID
in postmaster.pid lockfile, and checking at startup to ensure that no
processes are still connected to old shmem block (if it still exists).
* Switch backends to accept SIGQUIT rather than SIGUSR1 for emergency
stop, for symmetry with postmaster and xlog utilities. Clean up signal
handling in bootstrap.c so that xlog utilities launched by postmaster
will react to signals better.
* Standalone bootstrap now grabs lockfile in target directory, as added
insurance against running it in parallel with live postmaster.
2001-03-12 20:17:06 -05:00
|
|
|
|
2013-01-02 16:39:20 -05:00
|
|
|
/* Initialize MaxBackends (if under postmaster, was done already) */
|
|
|
|
|
if (!IsUnderPostmaster)
|
|
|
|
|
InitializeMaxBackends();
|
|
|
|
|
|
XLOG (and related) changes:
* Store two past checkpoint locations, not just one, in pg_control.
On startup, we fall back to the older checkpoint if the newer one
is unreadable. Also, a physical copy of the newest checkpoint record
is kept in pg_control for possible use in disaster recovery (ie,
complete loss of pg_xlog). Also add a version number for pg_control
itself. Remove archdir from pg_control; it ought to be a GUC
parameter, not a special case (not that it's implemented yet anyway).
* Suppress successive checkpoint records when nothing has been entered
in the WAL log since the last one. This is not so much to avoid I/O
as to make it actually useful to keep track of the last two
checkpoints. If the things are right next to each other then there's
not a lot of redundancy gained...
* Change CRC scheme to a true 64-bit CRC, not a pair of 32-bit CRCs
on alternate bytes. Polynomial borrowed from ECMA DLT1 standard.
* Fix XLOG record length handling so that it will work at BLCKSZ = 32k.
* Change XID allocation to work more like OID allocation. (This is of
dubious necessity, but I think it's a good idea anyway.)
* Fix a number of minor bugs, such as off-by-one logic for XLOG file
wraparound at the 4 gig mark.
* Add documentation and clean up some coding infelicities; move file
format declarations out to include files where planned contrib
utilities can get at them.
* Checkpoint will now occur every CHECKPOINT_SEGMENTS log segments or
every CHECKPOINT_TIMEOUT seconds, whichever comes first. It is also
possible to force a checkpoint by sending SIGUSR1 to the postmaster
(undocumented feature...)
* Defend against kill -9 postmaster by storing shmem block's key and ID
in postmaster.pid lockfile, and checking at startup to ensure that no
processes are still connected to old shmem block (if it still exists).
* Switch backends to accept SIGQUIT rather than SIGUSR1 for emergency
stop, for symmetry with postmaster and xlog utilities. Clean up signal
handling in bootstrap.c so that xlog utilities launched by postmaster
will react to signals better.
* Standalone bootstrap now grabs lockfile in target directory, as added
insurance against running it in parallel with live postmaster.
2001-03-12 20:17:06 -05:00
|
|
|
BaseInit();
|
|
|
|
|
|
2005-08-07 23:12:16 -04:00
|
|
|
/*
|
2007-03-07 08:35:03 -05:00
|
|
|
* When we are an auxiliary process, we aren't going to do the full
|
2006-10-03 20:30:14 -04:00
|
|
|
* InitPostgres pushups, but there are a couple of things that need to get
|
2007-03-07 08:35:03 -05:00
|
|
|
* lit up even in an auxiliary process.
|
2005-08-07 23:12:16 -04:00
|
|
|
*/
|
2002-09-25 16:31:40 -04:00
|
|
|
if (IsUnderPostmaster)
|
2003-11-19 10:55:08 -05:00
|
|
|
{
|
2006-01-04 16:06:32 -05:00
|
|
|
/*
|
|
|
|
|
* Create a PGPROC so we can use LWLocks. In the EXEC_BACKEND case,
|
|
|
|
|
* this was already done by SubPostmasterMain().
|
|
|
|
|
*/
|
|
|
|
|
#ifndef EXEC_BACKEND
|
2007-03-07 08:35:03 -05:00
|
|
|
InitAuxiliaryProcess();
|
2006-01-04 16:06:32 -05:00
|
|
|
#endif
|
2005-08-07 23:12:16 -04:00
|
|
|
|
2009-07-31 16:26:23 -04:00
|
|
|
/*
|
2014-05-06 12:12:18 -04:00
|
|
|
* Assign the ProcSignalSlot for an auxiliary process. Since it
|
2009-07-31 16:26:23 -04:00
|
|
|
* doesn't have a BackendId, the slot is statically allocated based on
|
2014-05-06 12:12:18 -04:00
|
|
|
* the auxiliary process type (MyAuxProcType). Backends use slots
|
Fix management of pendingOpsTable in auxiliary processes.
mdinit() was misusing IsBootstrapProcessingMode() to decide whether to
create an fsync pending-operations table in the current process. This led
to creating a table not only in the startup and checkpointer processes as
intended, but also in the bgwriter process, not to mention other auxiliary
processes such as walwriter and walreceiver. Creation of the table in the
bgwriter is fatal, because it absorbs fsync requests that should have gone
to the checkpointer; instead they just sit in bgwriter local memory and are
never acted on. So writes performed by the bgwriter were not being fsync'd
which could result in data loss after an OS crash. I think there is no
live bug with respect to walwriter and walreceiver because those never
perform any writes of shared buffers; but the potential is there for
future breakage in those processes too.
To fix, make AuxiliaryProcessMain() export the current process's
AuxProcType as a global variable, and then make mdinit() test directly for
the types of aux process that should have a pendingOpsTable. Having done
that, we might as well also get rid of the random bool flags such as
am_walreceiver that some of the aux processes had grown. (Note that we
could not have fixed the bug by examining those variables in mdinit(),
because it's called from BaseInit() which is run by AuxiliaryProcessMain()
before entering any of the process-type-specific code.)
Back-patch to 9.2, where the problem was introduced by the split-up of
bgwriter and checkpointer processes. The bogus pendingOpsTable exists
in walwriter and walreceiver processes in earlier branches, but absent
any evidence that it causes actual problems there, I'll leave the older
branches alone.
2012-07-18 15:28:10 -04:00
|
|
|
* indexed in the range from 1 to MaxBackends (inclusive), so we use
|
2009-07-31 16:26:23 -04:00
|
|
|
* MaxBackends + AuxProcType + 1 as the index of the slot for an
|
|
|
|
|
* auxiliary process.
|
|
|
|
|
*
|
|
|
|
|
* This will need rethinking if we ever want more than one of a
|
|
|
|
|
* particular auxiliary process type.
|
|
|
|
|
*/
|
Fix management of pendingOpsTable in auxiliary processes.
mdinit() was misusing IsBootstrapProcessingMode() to decide whether to
create an fsync pending-operations table in the current process. This led
to creating a table not only in the startup and checkpointer processes as
intended, but also in the bgwriter process, not to mention other auxiliary
processes such as walwriter and walreceiver. Creation of the table in the
bgwriter is fatal, because it absorbs fsync requests that should have gone
to the checkpointer; instead they just sit in bgwriter local memory and are
never acted on. So writes performed by the bgwriter were not being fsync'd
which could result in data loss after an OS crash. I think there is no
live bug with respect to walwriter and walreceiver because those never
perform any writes of shared buffers; but the potential is there for
future breakage in those processes too.
To fix, make AuxiliaryProcessMain() export the current process's
AuxProcType as a global variable, and then make mdinit() test directly for
the types of aux process that should have a pendingOpsTable. Having done
that, we might as well also get rid of the random bool flags such as
am_walreceiver that some of the aux processes had grown. (Note that we
could not have fixed the bug by examining those variables in mdinit(),
because it's called from BaseInit() which is run by AuxiliaryProcessMain()
before entering any of the process-type-specific code.)
Back-patch to 9.2, where the problem was introduced by the split-up of
bgwriter and checkpointer processes. The bogus pendingOpsTable exists
in walwriter and walreceiver processes in earlier branches, but absent
any evidence that it causes actual problems there, I'll leave the older
branches alone.
2012-07-18 15:28:10 -04:00
|
|
|
ProcSignalInit(MaxBackends + MyAuxProcType + 1);
|
2009-07-31 16:26:23 -04:00
|
|
|
|
2005-08-07 23:12:16 -04:00
|
|
|
/* finish setting up bufmgr.c */
|
|
|
|
|
InitBufferPoolBackend();
|
2006-06-08 19:55:48 -04:00
|
|
|
|
2013-12-18 12:57:20 -05:00
|
|
|
/* register a before-shutdown callback for LWLock cleanup */
|
|
|
|
|
before_shmem_exit(ShutdownAuxiliaryProcess, 0);
|
2003-11-19 10:55:08 -05:00
|
|
|
}
|
2002-09-25 16:31:40 -04:00
|
|
|
|
1999-10-06 17:58:18 -04:00
|
|
|
/*
|
2000-11-09 06:26:00 -05:00
|
|
|
* XLOG operations
|
1997-09-07 01:04:48 -04:00
|
|
|
*/
|
2000-11-21 04:39:57 -05:00
|
|
|
SetProcessingMode(NormalProcessing);
|
2001-09-29 00:02:27 -04:00
|
|
|
|
Fix management of pendingOpsTable in auxiliary processes.
mdinit() was misusing IsBootstrapProcessingMode() to decide whether to
create an fsync pending-operations table in the current process. This led
to creating a table not only in the startup and checkpointer processes as
intended, but also in the bgwriter process, not to mention other auxiliary
processes such as walwriter and walreceiver. Creation of the table in the
bgwriter is fatal, because it absorbs fsync requests that should have gone
to the checkpointer; instead they just sit in bgwriter local memory and are
never acted on. So writes performed by the bgwriter were not being fsync'd
which could result in data loss after an OS crash. I think there is no
live bug with respect to walwriter and walreceiver because those never
perform any writes of shared buffers; but the potential is there for
future breakage in those processes too.
To fix, make AuxiliaryProcessMain() export the current process's
AuxProcType as a global variable, and then make mdinit() test directly for
the types of aux process that should have a pendingOpsTable. Having done
that, we might as well also get rid of the random bool flags such as
am_walreceiver that some of the aux processes had grown. (Note that we
could not have fixed the bug by examining those variables in mdinit(),
because it's called from BaseInit() which is run by AuxiliaryProcessMain()
before entering any of the process-type-specific code.)
Back-patch to 9.2, where the problem was introduced by the split-up of
bgwriter and checkpointer processes. The bogus pendingOpsTable exists
in walwriter and walreceiver processes in earlier branches, but absent
any evidence that it causes actual problems there, I'll leave the older
branches alone.
2012-07-18 15:28:10 -04:00
|
|
|
switch (MyAuxProcType)
|
2000-11-21 04:39:57 -05:00
|
|
|
{
|
2007-03-07 08:35:03 -05:00
|
|
|
case CheckerProcess:
|
2010-04-19 21:38:52 -04:00
|
|
|
/* don't set signals, they're useless here */
|
2007-03-07 08:35:03 -05:00
|
|
|
CheckerModeMain();
|
|
|
|
|
proc_exit(1); /* should never return */
|
2001-09-29 00:02:27 -04:00
|
|
|
|
2007-03-07 08:35:03 -05:00
|
|
|
case BootstrapProcess:
|
2004-05-29 18:48:23 -04:00
|
|
|
bootstrap_signals();
|
2001-09-29 00:02:27 -04:00
|
|
|
BootStrapXLOG();
|
2007-03-07 08:35:03 -05:00
|
|
|
BootstrapModeMain();
|
|
|
|
|
proc_exit(1); /* should never return */
|
2001-09-29 00:02:27 -04:00
|
|
|
|
2007-03-07 08:35:03 -05:00
|
|
|
case StartupProcess:
|
Start background writer during archive recovery. Background writer now performs
its usual buffer cleaning duties during archive recovery, and it's responsible
for performing restartpoints.
This requires some changes in postmaster. When the startup process has done
all the initialization and is ready to start WAL redo, it signals the
postmaster to launch the background writer. The postmaster is signaled again
when the point in recovery is reached where we know that the database is in
consistent state. Postmaster isn't interested in that at the moment, but
that's the point where we could let other backends in to perform read-only
queries. The postmaster is signaled third time when the recovery has ended,
so that postmaster knows that it's safe to start accepting connections.
The startup process now traps SIGTERM, and performs a "clean" shutdown. If
you do a fast shutdown during recovery, a shutdown restartpoint is performed,
like a shutdown checkpoint, and postmaster kills the processes cleanly. You
still have to continue the recovery at next startup, though.
Currently, the background writer is only launched during archive recovery.
We could launch it during crash recovery as well, but it seems better to keep
that codepath as simple as possible, for the sake of robustness. And it
couldn't do any restartpoints during crash recovery anyway, so it wouldn't be
that useful.
log_restartpoints is gone. Use log_checkpoints instead. This is yet to be
documented.
This whole operation is a pre-requisite for Hot Standby, but has some value of
its own whether the hot standby patch makes 8.4 or not.
Simon Riggs, with lots of modifications by me.
2009-02-18 10:58:41 -05:00
|
|
|
/* don't set signals, startup process has its own agenda */
|
|
|
|
|
StartupProcessMain();
|
|
|
|
|
proc_exit(1); /* should never return */
|
2001-09-29 00:02:27 -04:00
|
|
|
|
2007-03-07 08:35:03 -05:00
|
|
|
case BgWriterProcess:
|
2004-05-29 18:48:23 -04:00
|
|
|
/* don't set signals, bgwriter has its own agenda */
|
|
|
|
|
BackgroundWriterMain();
|
|
|
|
|
proc_exit(1); /* should never return */
|
2007-07-24 00:54:09 -04:00
|
|
|
|
2011-11-01 13:14:47 -04:00
|
|
|
case CheckpointerProcess:
|
|
|
|
|
/* don't set signals, checkpointer has its own agenda */
|
|
|
|
|
CheckpointerMain();
|
|
|
|
|
proc_exit(1); /* should never return */
|
|
|
|
|
|
2007-07-24 00:54:09 -04:00
|
|
|
case WalWriterProcess:
|
|
|
|
|
/* don't set signals, walwriter has its own agenda */
|
|
|
|
|
InitXLOGAccess();
|
|
|
|
|
WalWriterMain();
|
|
|
|
|
proc_exit(1); /* should never return */
|
2007-11-15 16:14:46 -05:00
|
|
|
|
2010-01-15 04:19:10 -05:00
|
|
|
case WalReceiverProcess:
|
|
|
|
|
/* don't set signals, walreceiver has its own agenda */
|
2010-01-20 04:16:24 -05:00
|
|
|
WalReceiverMain();
|
2010-01-15 04:19:10 -05:00
|
|
|
proc_exit(1); /* should never return */
|
|
|
|
|
|
2001-09-29 00:02:27 -04:00
|
|
|
default:
|
Fix management of pendingOpsTable in auxiliary processes.
mdinit() was misusing IsBootstrapProcessingMode() to decide whether to
create an fsync pending-operations table in the current process. This led
to creating a table not only in the startup and checkpointer processes as
intended, but also in the bgwriter process, not to mention other auxiliary
processes such as walwriter and walreceiver. Creation of the table in the
bgwriter is fatal, because it absorbs fsync requests that should have gone
to the checkpointer; instead they just sit in bgwriter local memory and are
never acted on. So writes performed by the bgwriter were not being fsync'd
which could result in data loss after an OS crash. I think there is no
live bug with respect to walwriter and walreceiver because those never
perform any writes of shared buffers; but the potential is there for
future breakage in those processes too.
To fix, make AuxiliaryProcessMain() export the current process's
AuxProcType as a global variable, and then make mdinit() test directly for
the types of aux process that should have a pendingOpsTable. Having done
that, we might as well also get rid of the random bool flags such as
am_walreceiver that some of the aux processes had grown. (Note that we
could not have fixed the bug by examining those variables in mdinit(),
because it's called from BaseInit() which is run by AuxiliaryProcessMain()
before entering any of the process-type-specific code.)
Back-patch to 9.2, where the problem was introduced by the split-up of
bgwriter and checkpointer processes. The bogus pendingOpsTable exists
in walwriter and walreceiver processes in earlier branches, but absent
any evidence that it causes actual problems there, I'll leave the older
branches alone.
2012-07-18 15:28:10 -04:00
|
|
|
elog(PANIC, "unrecognized process type: %d", (int) MyAuxProcType);
|
2004-05-29 18:48:23 -04:00
|
|
|
proc_exit(1);
|
1997-09-07 01:04:48 -04:00
|
|
|
}
|
2007-03-07 08:35:03 -05:00
|
|
|
}
|
2001-09-29 00:02:27 -04:00
|
|
|
|
2007-03-07 08:35:03 -05:00
|
|
|
/*
|
|
|
|
|
* In shared memory checker mode, all we really want to do is create shared
|
|
|
|
|
* memory and semaphores (just to prove we can do it with the current GUC
|
2010-04-19 21:38:52 -04:00
|
|
|
* settings). Since, in fact, that was already done by BaseInit(),
|
|
|
|
|
* we have nothing more to do here.
|
2007-03-07 08:35:03 -05:00
|
|
|
*/
|
|
|
|
|
static void
|
|
|
|
|
CheckerModeMain(void)
|
|
|
|
|
{
|
|
|
|
|
proc_exit(0);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* The main entry point for running the backend in bootstrap mode
|
|
|
|
|
*
|
|
|
|
|
* The bootstrap mode is used to initialize the template database.
|
|
|
|
|
* The bootstrap backend doesn't speak SQL, but instead expects
|
|
|
|
|
* commands in a special bootstrap language.
|
|
|
|
|
*/
|
|
|
|
|
static void
|
|
|
|
|
BootstrapModeMain(void)
|
|
|
|
|
{
|
|
|
|
|
int i;
|
|
|
|
|
|
|
|
|
|
Assert(!IsUnderPostmaster);
|
|
|
|
|
|
|
|
|
|
SetProcessingMode(BootstrapProcessing);
|
1997-09-07 01:04:48 -04:00
|
|
|
|
2003-07-14 20:11:14 -04:00
|
|
|
/*
|
2007-03-07 08:35:03 -05:00
|
|
|
* Do backend-like initialization for bootstrap mode
|
2003-07-14 20:11:14 -04:00
|
|
|
*/
|
2007-03-07 08:35:03 -05:00
|
|
|
InitProcess();
|
2010-04-19 21:38:52 -04:00
|
|
|
|
2007-03-07 08:35:03 -05:00
|
|
|
InitPostgres(NULL, InvalidOid, NULL, NULL);
|
2003-07-14 20:11:14 -04:00
|
|
|
|
|
|
|
|
/* Initialize stuff for bootstrap-file processing */
|
1997-09-07 01:04:48 -04:00
|
|
|
for (i = 0; i < MAXATTR; i++)
|
|
|
|
|
{
|
2004-01-07 13:56:30 -05:00
|
|
|
attrtypes[i] = NULL;
|
2008-11-01 21:45:28 -04:00
|
|
|
Nulls[i] = false;
|
1997-09-07 01:04:48 -04:00
|
|
|
}
|
|
|
|
|
|
2001-03-22 01:16:21 -05:00
|
|
|
/*
|
2003-07-14 20:11:14 -04:00
|
|
|
* Process bootstrap input.
|
1997-09-07 01:04:48 -04:00
|
|
|
*/
|
2006-03-06 20:03:12 -05:00
|
|
|
boot_yyparse();
|
1997-09-07 01:04:48 -04:00
|
|
|
|
2010-02-07 15:48:13 -05:00
|
|
|
/*
|
2010-02-25 21:01:40 -05:00
|
|
|
* We should now know about all mapped relations, so it's okay to write
|
|
|
|
|
* out the initial relation mapping files.
|
2010-02-07 15:48:13 -05:00
|
|
|
*/
|
|
|
|
|
RelationMapFinishBootstrap();
|
|
|
|
|
|
2003-07-14 20:11:14 -04:00
|
|
|
/* Clean up and exit */
|
1997-09-07 01:04:48 -04:00
|
|
|
cleanup();
|
2007-03-07 08:35:03 -05:00
|
|
|
proc_exit(0);
|
2004-05-29 18:48:23 -04:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
/* ----------------------------------------------------------------
|
|
|
|
|
* misc functions
|
|
|
|
|
* ----------------------------------------------------------------
|
|
|
|
|
*/
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* Set up signal handling for a bootstrap process
|
|
|
|
|
*/
|
|
|
|
|
static void
|
|
|
|
|
bootstrap_signals(void)
|
|
|
|
|
{
|
|
|
|
|
if (IsUnderPostmaster)
|
|
|
|
|
{
|
2006-11-21 15:59:53 -05:00
|
|
|
/*
|
|
|
|
|
* If possible, make this process a group leader, so that the
|
|
|
|
|
* postmaster can signal any child processes too.
|
|
|
|
|
*/
|
|
|
|
|
#ifdef HAVE_SETSID
|
|
|
|
|
if (setsid() < 0)
|
|
|
|
|
elog(FATAL, "setsid() failed: %m");
|
|
|
|
|
#endif
|
|
|
|
|
|
2004-05-29 18:48:23 -04:00
|
|
|
/*
|
|
|
|
|
* Properly accept or ignore signals the postmaster might send us
|
|
|
|
|
*/
|
|
|
|
|
pqsignal(SIGHUP, SIG_IGN);
|
|
|
|
|
pqsignal(SIGINT, SIG_IGN); /* ignore query-cancel */
|
|
|
|
|
pqsignal(SIGTERM, die);
|
|
|
|
|
pqsignal(SIGQUIT, quickdie);
|
|
|
|
|
pqsignal(SIGALRM, SIG_IGN);
|
|
|
|
|
pqsignal(SIGPIPE, SIG_IGN);
|
|
|
|
|
pqsignal(SIGUSR1, SIG_IGN);
|
|
|
|
|
pqsignal(SIGUSR2, SIG_IGN);
|
1996-07-09 02:22:35 -04:00
|
|
|
|
2004-05-29 18:48:23 -04:00
|
|
|
/*
|
|
|
|
|
* Reset some signals that are accepted by postmaster but not here
|
|
|
|
|
*/
|
|
|
|
|
pqsignal(SIGCHLD, SIG_DFL);
|
|
|
|
|
pqsignal(SIGTTIN, SIG_DFL);
|
|
|
|
|
pqsignal(SIGTTOU, SIG_DFL);
|
|
|
|
|
pqsignal(SIGCONT, SIG_DFL);
|
|
|
|
|
pqsignal(SIGWINCH, SIG_DFL);
|
|
|
|
|
|
|
|
|
|
/*
|
2005-10-14 22:49:52 -04:00
|
|
|
* Unblock signals (they were blocked when the postmaster forked us)
|
2004-05-29 18:48:23 -04:00
|
|
|
*/
|
|
|
|
|
PG_SETMASK(&UnBlockSig);
|
|
|
|
|
}
|
|
|
|
|
else
|
|
|
|
|
{
|
|
|
|
|
/* Set up appropriately for interactive use */
|
|
|
|
|
pqsignal(SIGHUP, die);
|
|
|
|
|
pqsignal(SIGINT, die);
|
|
|
|
|
pqsignal(SIGTERM, die);
|
|
|
|
|
pqsignal(SIGQUIT, die);
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
2006-06-08 19:55:48 -04:00
|
|
|
/*
|
2014-05-06 12:12:18 -04:00
|
|
|
* Begin shutdown of an auxiliary process. This is approximately the equivalent
|
2007-03-07 08:35:03 -05:00
|
|
|
* of ShutdownPostgres() in postinit.c. We can't run transactions in an
|
|
|
|
|
* auxiliary process, so most of the work of AbortTransaction() is not needed,
|
2006-06-08 19:55:48 -04:00
|
|
|
* but we do need to make sure we've released any LWLocks we are holding.
|
|
|
|
|
* (This is only critical during an error exit.)
|
|
|
|
|
*/
|
|
|
|
|
static void
|
2007-03-07 08:35:03 -05:00
|
|
|
ShutdownAuxiliaryProcess(int code, Datum arg)
|
2006-06-08 19:55:48 -04:00
|
|
|
{
|
|
|
|
|
LWLockReleaseAll();
|
|
|
|
|
}
|
|
|
|
|
|
1996-07-09 02:22:35 -04:00
|
|
|
/* ----------------------------------------------------------------
|
1997-09-07 01:04:48 -04:00
|
|
|
* MANUAL BACKEND INTERACTIVE INTERFACE COMMANDS
|
1996-07-09 02:22:35 -04:00
|
|
|
* ----------------------------------------------------------------
|
|
|
|
|
*/
|
|
|
|
|
|
|
|
|
|
/* ----------------
|
1997-09-07 01:04:48 -04:00
|
|
|
* boot_openrel
|
1996-07-09 02:22:35 -04:00
|
|
|
* ----------------
|
|
|
|
|
*/
|
|
|
|
|
void
|
|
|
|
|
boot_openrel(char *relname)
|
|
|
|
|
{
|
1997-09-07 22:41:22 -04:00
|
|
|
int i;
|
1997-09-07 01:04:48 -04:00
|
|
|
struct typmap **app;
|
1998-08-18 22:04:17 -04:00
|
|
|
Relation rel;
|
|
|
|
|
HeapScanDesc scan;
|
1997-09-07 22:41:22 -04:00
|
|
|
HeapTuple tup;
|
1997-09-07 01:04:48 -04:00
|
|
|
|
2005-04-14 16:03:27 -04:00
|
|
|
if (strlen(relname) >= NAMEDATALEN)
|
1998-02-25 23:46:47 -05:00
|
|
|
relname[NAMEDATALEN - 1] = '\0';
|
1997-09-07 01:04:48 -04:00
|
|
|
|
2004-01-07 13:56:30 -05:00
|
|
|
if (Typ == NULL)
|
1997-09-07 01:04:48 -04:00
|
|
|
{
|
2006-08-15 18:36:17 -04:00
|
|
|
/* We can now load the pg_type data */
|
2005-04-14 16:03:27 -04:00
|
|
|
rel = heap_open(TypeRelationId, NoLock);
|
Use an MVCC snapshot, rather than SnapshotNow, for catalog scans.
SnapshotNow scans have the undesirable property that, in the face of
concurrent updates, the scan can fail to see either the old or the new
versions of the row. In many cases, we work around this by requiring
DDL operations to hold AccessExclusiveLock on the object being
modified; in some cases, the existing locking is inadequate and random
failures occur as a result. This commit doesn't change anything
related to locking, but will hopefully pave the way to allowing lock
strength reductions in the future.
The major issue has held us back from making this change in the past
is that taking an MVCC snapshot is significantly more expensive than
using a static special snapshot such as SnapshotNow. However, testing
of various worst-case scenarios reveals that this problem is not
severe except under fairly extreme workloads. To mitigate those
problems, we avoid retaking the MVCC snapshot for each new scan;
instead, we take a new snapshot only when invalidation messages have
been processed. The catcache machinery already requires that
invalidation messages be sent before releasing the related heavyweight
lock; else other backends might rely on locally-cached data rather
than scanning the catalog at all. Thus, making snapshot reuse
dependent on the same guarantees shouldn't break anything that wasn't
already subtly broken.
Patch by me. Review by Michael Paquier and Andres Freund.
2013-07-02 09:47:01 -04:00
|
|
|
scan = heap_beginscan_catalog(rel, 0, NULL);
|
1998-08-18 22:04:17 -04:00
|
|
|
i = 0;
|
2002-05-20 19:51:44 -04:00
|
|
|
while ((tup = heap_getnext(scan, ForwardScanDirection)) != NULL)
|
1998-08-18 22:04:17 -04:00
|
|
|
++i;
|
|
|
|
|
heap_endscan(scan);
|
1997-09-07 01:04:48 -04:00
|
|
|
app = Typ = ALLOC(struct typmap *, i + 1);
|
|
|
|
|
while (i-- > 0)
|
|
|
|
|
*app++ = ALLOC(struct typmap, 1);
|
2004-01-07 13:56:30 -05:00
|
|
|
*app = NULL;
|
Use an MVCC snapshot, rather than SnapshotNow, for catalog scans.
SnapshotNow scans have the undesirable property that, in the face of
concurrent updates, the scan can fail to see either the old or the new
versions of the row. In many cases, we work around this by requiring
DDL operations to hold AccessExclusiveLock on the object being
modified; in some cases, the existing locking is inadequate and random
failures occur as a result. This commit doesn't change anything
related to locking, but will hopefully pave the way to allowing lock
strength reductions in the future.
The major issue has held us back from making this change in the past
is that taking an MVCC snapshot is significantly more expensive than
using a static special snapshot such as SnapshotNow. However, testing
of various worst-case scenarios reveals that this problem is not
severe except under fairly extreme workloads. To mitigate those
problems, we avoid retaking the MVCC snapshot for each new scan;
instead, we take a new snapshot only when invalidation messages have
been processed. The catcache machinery already requires that
invalidation messages be sent before releasing the related heavyweight
lock; else other backends might rely on locally-cached data rather
than scanning the catalog at all. Thus, making snapshot reuse
dependent on the same guarantees shouldn't break anything that wasn't
already subtly broken.
Patch by me. Review by Michael Paquier and Andres Freund.
2013-07-02 09:47:01 -04:00
|
|
|
scan = heap_beginscan_catalog(rel, 0, NULL);
|
1997-09-07 01:04:48 -04:00
|
|
|
app = Typ;
|
2002-05-20 19:51:44 -04:00
|
|
|
while ((tup = heap_getnext(scan, ForwardScanDirection)) != NULL)
|
1997-09-07 01:04:48 -04:00
|
|
|
{
|
2002-07-20 01:16:59 -04:00
|
|
|
(*app)->am_oid = HeapTupleGetOid(tup);
|
2000-06-27 23:33:33 -04:00
|
|
|
memcpy((char *) &(*app)->am_typ,
|
|
|
|
|
(char *) GETSTRUCT(tup),
|
|
|
|
|
sizeof((*app)->am_typ));
|
|
|
|
|
app++;
|
1997-09-07 01:04:48 -04:00
|
|
|
}
|
1998-08-18 22:04:17 -04:00
|
|
|
heap_endscan(scan);
|
1999-09-18 15:08:25 -04:00
|
|
|
heap_close(rel, NoLock);
|
1997-09-07 01:04:48 -04:00
|
|
|
}
|
|
|
|
|
|
2002-04-27 17:24:34 -04:00
|
|
|
if (boot_reldesc != NULL)
|
1997-09-07 01:04:48 -04:00
|
|
|
closerel(NULL);
|
|
|
|
|
|
2003-07-22 19:30:39 -04:00
|
|
|
elog(DEBUG4, "open relation %s, attrsize %d",
|
2009-01-22 15:16:10 -05:00
|
|
|
relname, (int) ATTRIBUTE_FIXED_PART_SIZE);
|
1997-09-07 01:04:48 -04:00
|
|
|
|
2008-09-01 16:42:46 -04:00
|
|
|
boot_reldesc = heap_openrv(makeRangeVar(NULL, relname, -1), NoLock);
|
2002-04-27 17:24:34 -04:00
|
|
|
numattr = boot_reldesc->rd_rel->relnatts;
|
1997-09-07 01:04:48 -04:00
|
|
|
for (i = 0; i < numattr; i++)
|
|
|
|
|
{
|
|
|
|
|
if (attrtypes[i] == NULL)
|
|
|
|
|
attrtypes[i] = AllocateAttribute();
|
|
|
|
|
memmove((char *) attrtypes[i],
|
2002-04-27 17:24:34 -04:00
|
|
|
(char *) boot_reldesc->rd_att->attrs[i],
|
2009-01-22 15:16:10 -05:00
|
|
|
ATTRIBUTE_FIXED_PART_SIZE);
|
1997-09-07 01:04:48 -04:00
|
|
|
|
|
|
|
|
{
|
1998-08-31 23:29:17 -04:00
|
|
|
Form_pg_attribute at = attrtypes[i];
|
1997-09-07 01:04:48 -04:00
|
|
|
|
2003-05-27 13:49:47 -04:00
|
|
|
elog(DEBUG4, "create attribute %d name %s len %d num %d type %u",
|
2001-05-11 21:48:49 -04:00
|
|
|
i, NameStr(at->attname), at->attlen, at->attnum,
|
Commit to match discussed elog() changes. Only update is that LOG is
now just below FATAL in server_min_messages. Added more text to
highlight ordering difference between it and client_min_messages.
---------------------------------------------------------------------------
REALLYFATAL => PANIC
STOP => PANIC
New INFO level the prints to client by default
New LOG level the prints to server log by default
Cause VACUUM information to print only to the client
NOTICE => INFO where purely information messages are sent
DEBUG => LOG for purely server status messages
DEBUG removed, kept as backward compatible
DEBUG5, DEBUG4, DEBUG3, DEBUG2, DEBUG1 added
DebugLvl removed in favor of new DEBUG[1-5] symbols
New server_min_messages GUC parameter with values:
DEBUG[5-1], INFO, NOTICE, ERROR, LOG, FATAL, PANIC
New client_min_messages GUC parameter with values:
DEBUG[5-1], LOG, INFO, NOTICE, ERROR, FATAL, PANIC
Server startup now logged with LOG instead of DEBUG
Remove debug_level GUC parameter
elog() numbers now start at 10
Add test to print error message if older elog() values are passed to elog()
Bootstrap mode now has a -d that requires an argument, like postmaster
2002-03-02 16:39:36 -05:00
|
|
|
at->atttypid);
|
1997-09-07 01:04:48 -04:00
|
|
|
}
|
|
|
|
|
}
|
1996-07-09 02:22:35 -04:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/* ----------------
|
1997-09-07 01:04:48 -04:00
|
|
|
* closerel
|
1996-07-09 02:22:35 -04:00
|
|
|
* ----------------
|
|
|
|
|
*/
|
|
|
|
|
void
|
|
|
|
|
closerel(char *name)
|
|
|
|
|
{
|
1997-09-07 01:04:48 -04:00
|
|
|
if (name)
|
|
|
|
|
{
|
2002-04-27 17:24:34 -04:00
|
|
|
if (boot_reldesc)
|
1997-09-07 01:04:48 -04:00
|
|
|
{
|
2002-04-27 17:24:34 -04:00
|
|
|
if (strcmp(RelationGetRelationName(boot_reldesc), name) != 0)
|
2003-07-22 19:30:39 -04:00
|
|
|
elog(ERROR, "close of %s when %s was expected",
|
2005-04-14 16:03:27 -04:00
|
|
|
name, RelationGetRelationName(boot_reldesc));
|
1997-09-07 01:04:48 -04:00
|
|
|
}
|
|
|
|
|
else
|
2003-07-22 19:30:39 -04:00
|
|
|
elog(ERROR, "close of %s before any relation was opened",
|
1997-09-07 01:04:48 -04:00
|
|
|
name);
|
|
|
|
|
}
|
|
|
|
|
|
2002-04-27 17:24:34 -04:00
|
|
|
if (boot_reldesc == NULL)
|
2001-05-11 21:48:49 -04:00
|
|
|
elog(ERROR, "no open relation to close");
|
1997-09-07 01:04:48 -04:00
|
|
|
else
|
|
|
|
|
{
|
2005-04-14 16:03:27 -04:00
|
|
|
elog(DEBUG4, "close relation %s",
|
|
|
|
|
RelationGetRelationName(boot_reldesc));
|
2002-04-27 17:24:34 -04:00
|
|
|
heap_close(boot_reldesc, NoLock);
|
2004-01-07 13:56:30 -05:00
|
|
|
boot_reldesc = NULL;
|
1997-09-07 01:04:48 -04:00
|
|
|
}
|
1996-07-09 02:22:35 -04:00
|
|
|
}
|
|
|
|
|
|
1997-09-07 01:04:48 -04:00
|
|
|
|
|
|
|
|
|
1996-07-09 02:22:35 -04:00
|
|
|
/* ----------------
|
|
|
|
|
* DEFINEATTR()
|
|
|
|
|
*
|
|
|
|
|
* define a <field,type> pair
|
|
|
|
|
* if there are n fields in a relation to be created, this routine
|
|
|
|
|
* will be called n times
|
|
|
|
|
* ----------------
|
|
|
|
|
*/
|
|
|
|
|
void
|
|
|
|
|
DefineAttr(char *name, char *type, int attnum)
|
|
|
|
|
{
|
1998-08-24 15:04:04 -04:00
|
|
|
Oid typeoid;
|
1997-09-07 01:04:48 -04:00
|
|
|
|
2002-04-27 17:24:34 -04:00
|
|
|
if (boot_reldesc != NULL)
|
1997-09-07 01:04:48 -04:00
|
|
|
{
|
2003-07-22 19:30:39 -04:00
|
|
|
elog(WARNING, "no open relations allowed with CREATE command");
|
2005-04-14 16:03:27 -04:00
|
|
|
closerel(NULL);
|
1997-09-07 01:04:48 -04:00
|
|
|
}
|
|
|
|
|
|
2004-01-07 13:56:30 -05:00
|
|
|
if (attrtypes[attnum] == NULL)
|
1997-09-07 01:04:48 -04:00
|
|
|
attrtypes[attnum] = AllocateAttribute();
|
2009-01-22 15:16:10 -05:00
|
|
|
MemSet(attrtypes[attnum], 0, ATTRIBUTE_FIXED_PART_SIZE);
|
2002-08-02 14:15:10 -04:00
|
|
|
|
|
|
|
|
namestrcpy(&attrtypes[attnum]->attname, name);
|
2003-05-27 13:49:47 -04:00
|
|
|
elog(DEBUG4, "column %s %s", NameStr(attrtypes[attnum]->attname), type);
|
2002-09-22 15:42:52 -04:00
|
|
|
attrtypes[attnum]->attnum = attnum + 1; /* fillatt */
|
2002-08-02 14:15:10 -04:00
|
|
|
|
|
|
|
|
typeoid = gettype(type);
|
|
|
|
|
|
2004-01-07 13:56:30 -05:00
|
|
|
if (Typ != NULL)
|
1997-09-07 01:04:48 -04:00
|
|
|
{
|
|
|
|
|
attrtypes[attnum]->atttypid = Ap->am_oid;
|
2005-03-28 19:17:27 -05:00
|
|
|
attrtypes[attnum]->attlen = Ap->am_typ.typlen;
|
1997-09-07 01:04:48 -04:00
|
|
|
attrtypes[attnum]->attbyval = Ap->am_typ.typbyval;
|
2002-04-24 22:56:56 -04:00
|
|
|
attrtypes[attnum]->attstorage = Ap->am_typ.typstorage;
|
1998-08-24 15:04:04 -04:00
|
|
|
attrtypes[attnum]->attalign = Ap->am_typ.typalign;
|
2011-02-08 16:04:18 -05:00
|
|
|
attrtypes[attnum]->attcollation = Ap->am_typ.typcollation;
|
2004-01-06 18:55:19 -05:00
|
|
|
/* if an array type, assume 1-dimensional attribute */
|
|
|
|
|
if (Ap->am_typ.typelem != InvalidOid && Ap->am_typ.typlen < 0)
|
|
|
|
|
attrtypes[attnum]->attndims = 1;
|
|
|
|
|
else
|
|
|
|
|
attrtypes[attnum]->attndims = 0;
|
1997-09-07 01:04:48 -04:00
|
|
|
}
|
|
|
|
|
else
|
|
|
|
|
{
|
2004-04-01 16:28:47 -05:00
|
|
|
attrtypes[attnum]->atttypid = TypInfo[typeoid].oid;
|
2005-03-28 19:17:27 -05:00
|
|
|
attrtypes[attnum]->attlen = TypInfo[typeoid].len;
|
2004-04-01 16:28:47 -05:00
|
|
|
attrtypes[attnum]->attbyval = TypInfo[typeoid].byval;
|
|
|
|
|
attrtypes[attnum]->attstorage = TypInfo[typeoid].storage;
|
|
|
|
|
attrtypes[attnum]->attalign = TypInfo[typeoid].align;
|
2011-02-08 16:04:18 -05:00
|
|
|
attrtypes[attnum]->attcollation = TypInfo[typeoid].collation;
|
2004-01-06 18:55:19 -05:00
|
|
|
/* if an array type, assume 1-dimensional attribute */
|
2005-03-28 19:17:27 -05:00
|
|
|
if (TypInfo[typeoid].elem != InvalidOid &&
|
|
|
|
|
attrtypes[attnum]->attlen < 0)
|
2004-01-06 18:55:19 -05:00
|
|
|
attrtypes[attnum]->attndims = 1;
|
|
|
|
|
else
|
|
|
|
|
attrtypes[attnum]->attndims = 0;
|
1997-09-07 01:04:48 -04:00
|
|
|
}
|
2004-01-06 18:55:19 -05:00
|
|
|
|
|
|
|
|
attrtypes[attnum]->attstattarget = -1;
|
1998-02-06 14:18:06 -05:00
|
|
|
attrtypes[attnum]->attcacheoff = -1;
|
1998-02-07 01:11:56 -05:00
|
|
|
attrtypes[attnum]->atttypmod = -1;
|
2002-09-22 15:42:52 -04:00
|
|
|
attrtypes[attnum]->attislocal = true;
|
2002-09-04 16:31:48 -04:00
|
|
|
|
2002-08-02 14:15:10 -04:00
|
|
|
/*
|
2005-10-14 22:49:52 -04:00
|
|
|
* Mark as "not null" if type is fixed-width and prior columns are too.
|
|
|
|
|
* This corresponds to case where column can be accessed directly via C
|
|
|
|
|
* struct declaration.
|
2005-03-28 19:17:27 -05:00
|
|
|
*
|
2005-10-14 22:49:52 -04:00
|
|
|
* oidvector and int2vector are also treated as not-nullable, even though
|
|
|
|
|
* they are no longer fixed-width.
|
2002-08-02 14:15:10 -04:00
|
|
|
*/
|
2005-03-28 19:17:27 -05:00
|
|
|
#define MARKNOTNULL(att) \
|
|
|
|
|
((att)->attlen > 0 || \
|
|
|
|
|
(att)->atttypid == OIDVECTOROID || \
|
|
|
|
|
(att)->atttypid == INT2VECTOROID)
|
|
|
|
|
|
|
|
|
|
if (MARKNOTNULL(attrtypes[attnum]))
|
2002-08-02 14:15:10 -04:00
|
|
|
{
|
2002-09-04 16:31:48 -04:00
|
|
|
int i;
|
2002-08-02 14:15:10 -04:00
|
|
|
|
|
|
|
|
for (i = 0; i < attnum; i++)
|
|
|
|
|
{
|
2005-03-28 19:17:27 -05:00
|
|
|
if (!MARKNOTNULL(attrtypes[i]))
|
2002-08-02 14:15:10 -04:00
|
|
|
break;
|
|
|
|
|
}
|
|
|
|
|
if (i == attnum)
|
|
|
|
|
attrtypes[attnum]->attnotnull = true;
|
|
|
|
|
}
|
1996-07-09 02:22:35 -04:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
/* ----------------
|
1997-09-07 01:04:48 -04:00
|
|
|
* InsertOneTuple
|
2001-08-10 14:57:42 -04:00
|
|
|
*
|
|
|
|
|
* If objectid is not zero, it is a specific OID to assign to the tuple.
|
|
|
|
|
* Otherwise, an OID will be assigned (if necessary) by heap_insert.
|
1996-07-09 02:22:35 -04:00
|
|
|
* ----------------
|
|
|
|
|
*/
|
|
|
|
|
void
|
|
|
|
|
InsertOneTuple(Oid objectid)
|
|
|
|
|
{
|
1997-09-07 22:41:22 -04:00
|
|
|
HeapTuple tuple;
|
|
|
|
|
TupleDesc tupDesc;
|
|
|
|
|
int i;
|
1997-09-07 01:04:48 -04:00
|
|
|
|
2003-05-27 13:49:47 -04:00
|
|
|
elog(DEBUG4, "inserting row oid %u, %d columns", objectid, numattr);
|
1997-09-07 01:04:48 -04:00
|
|
|
|
2002-09-01 21:05:06 -04:00
|
|
|
tupDesc = CreateTupleDesc(numattr,
|
|
|
|
|
RelationGetForm(boot_reldesc)->relhasoids,
|
|
|
|
|
attrtypes);
|
2008-11-01 21:45:28 -04:00
|
|
|
tuple = heap_form_tuple(tupDesc, values, Nulls);
|
1997-09-07 01:04:48 -04:00
|
|
|
if (objectid != (Oid) 0)
|
2002-07-20 01:16:59 -04:00
|
|
|
HeapTupleSetOid(tuple, objectid);
|
|
|
|
|
pfree(tupDesc); /* just free's tupDesc, not the attrtypes */
|
2002-09-01 21:05:06 -04:00
|
|
|
|
2002-05-21 18:05:55 -04:00
|
|
|
simple_heap_insert(boot_reldesc, tuple);
|
1999-12-16 17:20:03 -05:00
|
|
|
heap_freetuple(tuple);
|
2003-05-27 13:49:47 -04:00
|
|
|
elog(DEBUG4, "row inserted");
|
1997-09-07 01:04:48 -04:00
|
|
|
|
|
|
|
|
/*
|
2008-11-01 21:45:28 -04:00
|
|
|
* Reset null markers for next tuple
|
1997-09-07 01:04:48 -04:00
|
|
|
*/
|
|
|
|
|
for (i = 0; i < numattr; i++)
|
2008-11-01 21:45:28 -04:00
|
|
|
Nulls[i] = false;
|
1996-07-09 02:22:35 -04:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/* ----------------
|
1997-09-07 01:04:48 -04:00
|
|
|
* InsertOneValue
|
1996-07-09 02:22:35 -04:00
|
|
|
* ----------------
|
|
|
|
|
*/
|
|
|
|
|
void
|
2001-08-10 14:57:42 -04:00
|
|
|
InsertOneValue(char *value, int i)
|
1996-07-09 02:22:35 -04:00
|
|
|
{
|
2004-06-05 20:41:28 -04:00
|
|
|
Oid typoid;
|
2006-08-15 18:36:17 -04:00
|
|
|
int16 typlen;
|
|
|
|
|
bool typbyval;
|
|
|
|
|
char typalign;
|
|
|
|
|
char typdelim;
|
2004-06-05 20:41:28 -04:00
|
|
|
Oid typioparam;
|
|
|
|
|
Oid typinput;
|
|
|
|
|
Oid typoutput;
|
1997-09-07 01:04:48 -04:00
|
|
|
|
2011-09-06 16:05:02 -04:00
|
|
|
AssertArg(i >= 0 && i < MAXATTR);
|
2001-05-11 21:48:49 -04:00
|
|
|
|
2003-07-22 19:30:39 -04:00
|
|
|
elog(DEBUG4, "inserting column %d value \"%s\"", i, value);
|
1997-09-07 01:04:48 -04:00
|
|
|
|
2007-03-07 08:35:03 -05:00
|
|
|
typoid = boot_reldesc->rd_att->attrs[i]->atttypid;
|
2004-06-05 20:41:28 -04:00
|
|
|
|
2006-08-15 18:36:17 -04:00
|
|
|
boot_get_type_io_data(typoid,
|
|
|
|
|
&typlen, &typbyval, &typalign,
|
|
|
|
|
&typdelim, &typioparam,
|
|
|
|
|
&typinput, &typoutput);
|
|
|
|
|
|
2006-04-04 15:35:37 -04:00
|
|
|
values[i] = OidInputFunctionCall(typinput, value, typioparam, -1);
|
Prevent memory leaks from accumulating across printtup() calls.
Historically, printtup() has assumed that it could prevent memory leakage
by pfree'ing the string result of each output function and manually
managing detoasting of toasted values. This amounts to assuming that
datatype output functions never leak any memory internally; an assumption
we've already decided to be bogus elsewhere, for example in COPY OUT.
range_out in particular is known to leak multiple kilobytes per call, as
noted in bug #8573 from Godfried Vanluffelen. While we could go in and fix
that leak, it wouldn't be very notationally convenient, and in any case
there have been and undoubtedly will again be other leaks in other output
functions. So what seems like the best solution is to run the output
functions in a temporary memory context that can be reset after each row,
as we're doing in COPY OUT. Some quick experimentation suggests this is
actually a tad faster than the retail pfree's anyway.
This patch fixes all the variants of printtup, except for debugtup()
which is used in standalone mode. It doesn't seem worth worrying
about query-lifespan leaks in standalone mode, and fixing that case
would be a bit tedious since debugtup() doesn't currently have any
startup or shutdown functions.
While at it, remove manual detoast management from several other
output-function call sites that had copied it from printtup(). This
doesn't make a lot of difference right now, but in view of recent
discussions about supporting "non-flattened" Datums, we're going to
want that code gone eventually anyway.
Back-patch to 9.2 where range_out was introduced. We might eventually
decide to back-patch this further, but in the absence of known major
leaks in older output functions, I'll refrain for now.
2013-11-03 11:33:05 -05:00
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* We use ereport not elog here so that parameters aren't evaluated unless
|
|
|
|
|
* the message is going to be printed, which generally it isn't
|
|
|
|
|
*/
|
|
|
|
|
ereport(DEBUG4,
|
|
|
|
|
(errmsg_internal("inserted -> %s",
|
|
|
|
|
OidOutputFunctionCall(typoutput, values[i]))));
|
1996-07-09 02:22:35 -04:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/* ----------------
|
1997-09-07 01:04:48 -04:00
|
|
|
* InsertOneNull
|
1996-07-09 02:22:35 -04:00
|
|
|
* ----------------
|
|
|
|
|
*/
|
|
|
|
|
void
|
|
|
|
|
InsertOneNull(int i)
|
|
|
|
|
{
|
2003-05-27 13:49:47 -04:00
|
|
|
elog(DEBUG4, "inserting column %d NULL", i);
|
2011-09-06 16:05:02 -04:00
|
|
|
Assert(i >= 0 && i < MAXATTR);
|
2000-05-30 00:25:00 -04:00
|
|
|
values[i] = PointerGetDatum(NULL);
|
2008-11-01 21:45:28 -04:00
|
|
|
Nulls[i] = true;
|
1996-07-09 02:22:35 -04:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/* ----------------
|
1997-09-07 01:04:48 -04:00
|
|
|
* cleanup
|
1996-07-09 02:22:35 -04:00
|
|
|
* ----------------
|
|
|
|
|
*/
|
1997-08-19 17:40:56 -04:00
|
|
|
static void
|
2003-05-28 14:19:09 -04:00
|
|
|
cleanup(void)
|
1996-07-09 02:22:35 -04:00
|
|
|
{
|
2002-08-02 18:36:05 -04:00
|
|
|
if (boot_reldesc != NULL)
|
|
|
|
|
closerel(NULL);
|
1996-07-09 02:22:35 -04:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/* ----------------
|
1997-09-07 01:04:48 -04:00
|
|
|
* gettype
|
2002-04-24 22:56:56 -04:00
|
|
|
*
|
2004-04-01 16:28:47 -05:00
|
|
|
* NB: this is really ugly; it will return an integer index into TypInfo[],
|
2002-04-24 22:56:56 -04:00
|
|
|
* and not an OID at all, until the first reference to a type not known in
|
2004-04-01 16:28:47 -05:00
|
|
|
* TypInfo[]. At that point it will read and cache pg_type in the Typ array,
|
2002-04-24 22:56:56 -04:00
|
|
|
* and subsequently return a real OID (and set the global pointer Ap to
|
2014-05-06 12:12:18 -04:00
|
|
|
* point at the found row in Typ). So caller must check whether Typ is
|
2002-04-24 22:56:56 -04:00
|
|
|
* still NULL to determine what the return value is!
|
1996-07-09 02:22:35 -04:00
|
|
|
* ----------------
|
|
|
|
|
*/
|
1998-08-24 15:04:04 -04:00
|
|
|
static Oid
|
1996-07-09 02:22:35 -04:00
|
|
|
gettype(char *type)
|
|
|
|
|
{
|
1997-09-07 22:41:22 -04:00
|
|
|
int i;
|
1998-08-18 22:04:17 -04:00
|
|
|
Relation rel;
|
|
|
|
|
HeapScanDesc scan;
|
1997-09-07 22:41:22 -04:00
|
|
|
HeapTuple tup;
|
1997-09-07 01:04:48 -04:00
|
|
|
struct typmap **app;
|
|
|
|
|
|
2004-01-07 13:56:30 -05:00
|
|
|
if (Typ != NULL)
|
1997-09-07 01:04:48 -04:00
|
|
|
{
|
2004-01-07 13:56:30 -05:00
|
|
|
for (app = Typ; *app != NULL; app++)
|
1997-09-07 01:04:48 -04:00
|
|
|
{
|
1999-11-07 18:08:36 -05:00
|
|
|
if (strncmp(NameStr((*app)->am_typ.typname), type, NAMEDATALEN) == 0)
|
1997-09-07 01:04:48 -04:00
|
|
|
{
|
|
|
|
|
Ap = *app;
|
1998-08-31 23:29:17 -04:00
|
|
|
return (*app)->am_oid;
|
1997-09-07 01:04:48 -04:00
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
else
|
|
|
|
|
{
|
2002-04-24 22:56:56 -04:00
|
|
|
for (i = 0; i < n_types; i++)
|
1997-09-07 01:04:48 -04:00
|
|
|
{
|
2004-04-01 16:28:47 -05:00
|
|
|
if (strncmp(type, TypInfo[i].name, NAMEDATALEN) == 0)
|
1998-08-31 23:29:17 -04:00
|
|
|
return i;
|
1997-09-07 01:04:48 -04:00
|
|
|
}
|
2003-05-27 13:49:47 -04:00
|
|
|
elog(DEBUG4, "external type: %s", type);
|
2005-04-14 16:03:27 -04:00
|
|
|
rel = heap_open(TypeRelationId, NoLock);
|
Use an MVCC snapshot, rather than SnapshotNow, for catalog scans.
SnapshotNow scans have the undesirable property that, in the face of
concurrent updates, the scan can fail to see either the old or the new
versions of the row. In many cases, we work around this by requiring
DDL operations to hold AccessExclusiveLock on the object being
modified; in some cases, the existing locking is inadequate and random
failures occur as a result. This commit doesn't change anything
related to locking, but will hopefully pave the way to allowing lock
strength reductions in the future.
The major issue has held us back from making this change in the past
is that taking an MVCC snapshot is significantly more expensive than
using a static special snapshot such as SnapshotNow. However, testing
of various worst-case scenarios reveals that this problem is not
severe except under fairly extreme workloads. To mitigate those
problems, we avoid retaking the MVCC snapshot for each new scan;
instead, we take a new snapshot only when invalidation messages have
been processed. The catcache machinery already requires that
invalidation messages be sent before releasing the related heavyweight
lock; else other backends might rely on locally-cached data rather
than scanning the catalog at all. Thus, making snapshot reuse
dependent on the same guarantees shouldn't break anything that wasn't
already subtly broken.
Patch by me. Review by Michael Paquier and Andres Freund.
2013-07-02 09:47:01 -04:00
|
|
|
scan = heap_beginscan_catalog(rel, 0, NULL);
|
1997-09-07 01:04:48 -04:00
|
|
|
i = 0;
|
2002-05-20 19:51:44 -04:00
|
|
|
while ((tup = heap_getnext(scan, ForwardScanDirection)) != NULL)
|
1997-09-07 01:04:48 -04:00
|
|
|
++i;
|
1998-08-18 22:04:17 -04:00
|
|
|
heap_endscan(scan);
|
1997-09-07 01:04:48 -04:00
|
|
|
app = Typ = ALLOC(struct typmap *, i + 1);
|
|
|
|
|
while (i-- > 0)
|
|
|
|
|
*app++ = ALLOC(struct typmap, 1);
|
2004-01-07 13:56:30 -05:00
|
|
|
*app = NULL;
|
Use an MVCC snapshot, rather than SnapshotNow, for catalog scans.
SnapshotNow scans have the undesirable property that, in the face of
concurrent updates, the scan can fail to see either the old or the new
versions of the row. In many cases, we work around this by requiring
DDL operations to hold AccessExclusiveLock on the object being
modified; in some cases, the existing locking is inadequate and random
failures occur as a result. This commit doesn't change anything
related to locking, but will hopefully pave the way to allowing lock
strength reductions in the future.
The major issue has held us back from making this change in the past
is that taking an MVCC snapshot is significantly more expensive than
using a static special snapshot such as SnapshotNow. However, testing
of various worst-case scenarios reveals that this problem is not
severe except under fairly extreme workloads. To mitigate those
problems, we avoid retaking the MVCC snapshot for each new scan;
instead, we take a new snapshot only when invalidation messages have
been processed. The catcache machinery already requires that
invalidation messages be sent before releasing the related heavyweight
lock; else other backends might rely on locally-cached data rather
than scanning the catalog at all. Thus, making snapshot reuse
dependent on the same guarantees shouldn't break anything that wasn't
already subtly broken.
Patch by me. Review by Michael Paquier and Andres Freund.
2013-07-02 09:47:01 -04:00
|
|
|
scan = heap_beginscan_catalog(rel, 0, NULL);
|
1997-09-07 01:04:48 -04:00
|
|
|
app = Typ;
|
2002-05-20 19:51:44 -04:00
|
|
|
while ((tup = heap_getnext(scan, ForwardScanDirection)) != NULL)
|
1997-09-07 01:04:48 -04:00
|
|
|
{
|
2002-07-20 01:16:59 -04:00
|
|
|
(*app)->am_oid = HeapTupleGetOid(tup);
|
1997-09-07 01:04:48 -04:00
|
|
|
memmove((char *) &(*app++)->am_typ,
|
|
|
|
|
(char *) GETSTRUCT(tup),
|
|
|
|
|
sizeof((*app)->am_typ));
|
|
|
|
|
}
|
1998-08-18 22:04:17 -04:00
|
|
|
heap_endscan(scan);
|
1999-09-18 15:08:25 -04:00
|
|
|
heap_close(rel, NoLock);
|
1998-08-31 23:29:17 -04:00
|
|
|
return gettype(type);
|
1997-09-07 01:04:48 -04:00
|
|
|
}
|
2003-07-22 19:30:39 -04:00
|
|
|
elog(ERROR, "unrecognized type \"%s\"", type);
|
1997-09-07 01:04:48 -04:00
|
|
|
/* not reached, here to make compiler happy */
|
|
|
|
|
return 0;
|
1996-07-09 02:22:35 -04:00
|
|
|
}
|
|
|
|
|
|
2006-08-15 18:36:17 -04:00
|
|
|
/* ----------------
|
|
|
|
|
* boot_get_type_io_data
|
|
|
|
|
*
|
|
|
|
|
* Obtain type I/O information at bootstrap time. This intentionally has
|
|
|
|
|
* almost the same API as lsyscache.c's get_type_io_data, except that
|
|
|
|
|
* we only support obtaining the typinput and typoutput routines, not
|
|
|
|
|
* the binary I/O routines. It is exported so that array_in and array_out
|
|
|
|
|
* can be made to work during early bootstrap.
|
|
|
|
|
* ----------------
|
|
|
|
|
*/
|
|
|
|
|
void
|
|
|
|
|
boot_get_type_io_data(Oid typid,
|
|
|
|
|
int16 *typlen,
|
|
|
|
|
bool *typbyval,
|
|
|
|
|
char *typalign,
|
|
|
|
|
char *typdelim,
|
|
|
|
|
Oid *typioparam,
|
|
|
|
|
Oid *typinput,
|
|
|
|
|
Oid *typoutput)
|
|
|
|
|
{
|
|
|
|
|
if (Typ != NULL)
|
|
|
|
|
{
|
|
|
|
|
/* We have the boot-time contents of pg_type, so use it */
|
|
|
|
|
struct typmap **app;
|
|
|
|
|
struct typmap *ap;
|
|
|
|
|
|
|
|
|
|
app = Typ;
|
|
|
|
|
while (*app && (*app)->am_oid != typid)
|
|
|
|
|
++app;
|
|
|
|
|
ap = *app;
|
|
|
|
|
if (ap == NULL)
|
|
|
|
|
elog(ERROR, "type OID %u not found in Typ list", typid);
|
|
|
|
|
|
|
|
|
|
*typlen = ap->am_typ.typlen;
|
|
|
|
|
*typbyval = ap->am_typ.typbyval;
|
|
|
|
|
*typalign = ap->am_typ.typalign;
|
|
|
|
|
*typdelim = ap->am_typ.typdelim;
|
|
|
|
|
|
|
|
|
|
/* XXX this logic must match getTypeIOParam() */
|
|
|
|
|
if (OidIsValid(ap->am_typ.typelem))
|
|
|
|
|
*typioparam = ap->am_typ.typelem;
|
|
|
|
|
else
|
|
|
|
|
*typioparam = typid;
|
|
|
|
|
|
|
|
|
|
*typinput = ap->am_typ.typinput;
|
|
|
|
|
*typoutput = ap->am_typ.typoutput;
|
|
|
|
|
}
|
|
|
|
|
else
|
|
|
|
|
{
|
|
|
|
|
/* We don't have pg_type yet, so use the hard-wired TypInfo array */
|
|
|
|
|
int typeindex;
|
|
|
|
|
|
|
|
|
|
for (typeindex = 0; typeindex < n_types; typeindex++)
|
|
|
|
|
{
|
|
|
|
|
if (TypInfo[typeindex].oid == typid)
|
|
|
|
|
break;
|
|
|
|
|
}
|
|
|
|
|
if (typeindex >= n_types)
|
|
|
|
|
elog(ERROR, "type OID %u not found in TypInfo", typid);
|
|
|
|
|
|
|
|
|
|
*typlen = TypInfo[typeindex].len;
|
|
|
|
|
*typbyval = TypInfo[typeindex].byval;
|
|
|
|
|
*typalign = TypInfo[typeindex].align;
|
|
|
|
|
/* We assume typdelim is ',' for all boot-time types */
|
|
|
|
|
*typdelim = ',';
|
|
|
|
|
|
|
|
|
|
/* XXX this logic must match getTypeIOParam() */
|
|
|
|
|
if (OidIsValid(TypInfo[typeindex].elem))
|
|
|
|
|
*typioparam = TypInfo[typeindex].elem;
|
|
|
|
|
else
|
|
|
|
|
*typioparam = typid;
|
|
|
|
|
|
|
|
|
|
*typinput = TypInfo[typeindex].inproc;
|
|
|
|
|
*typoutput = TypInfo[typeindex].outproc;
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
1996-07-09 02:22:35 -04:00
|
|
|
/* ----------------
|
1997-09-07 01:04:48 -04:00
|
|
|
* AllocateAttribute
|
2009-01-22 15:16:10 -05:00
|
|
|
*
|
|
|
|
|
* Note: bootstrap never sets any per-column ACLs, so we only need
|
|
|
|
|
* ATTRIBUTE_FIXED_PART_SIZE space per attribute.
|
1996-07-09 02:22:35 -04:00
|
|
|
* ----------------
|
|
|
|
|
*/
|
2002-08-02 14:15:10 -04:00
|
|
|
static Form_pg_attribute
|
|
|
|
|
AllocateAttribute(void)
|
1996-07-09 02:22:35 -04:00
|
|
|
{
|
2009-01-22 15:16:10 -05:00
|
|
|
Form_pg_attribute attribute = (Form_pg_attribute) malloc(ATTRIBUTE_FIXED_PART_SIZE);
|
1997-09-07 01:04:48 -04:00
|
|
|
|
|
|
|
|
if (!PointerIsValid(attribute))
|
2003-07-22 19:30:39 -04:00
|
|
|
elog(FATAL, "out of memory");
|
2009-01-22 15:16:10 -05:00
|
|
|
MemSet(attribute, 0, ATTRIBUTE_FIXED_PART_SIZE);
|
1997-09-07 01:04:48 -04:00
|
|
|
|
1998-08-31 23:29:17 -04:00
|
|
|
return attribute;
|
1996-07-09 02:22:35 -04:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/* ----------------
|
1997-09-07 01:04:48 -04:00
|
|
|
* MapArrayTypeName
|
1996-07-09 02:22:35 -04:00
|
|
|
* XXX arrays of "basetype" are always "_basetype".
|
1997-09-07 01:04:48 -04:00
|
|
|
* this is an evil hack inherited from rel. 3.1.
|
1996-07-09 02:22:35 -04:00
|
|
|
* XXX array dimension is thrown away because we
|
1997-09-07 01:04:48 -04:00
|
|
|
* don't support fixed-dimension arrays. again,
|
|
|
|
|
* sickness from 3.1.
|
|
|
|
|
*
|
|
|
|
|
* the string passed in must have a '[' character in it
|
1996-07-09 02:22:35 -04:00
|
|
|
*
|
|
|
|
|
* the string returned is a pointer to static storage and should NOT
|
|
|
|
|
* be freed by the CALLER.
|
|
|
|
|
* ----------------
|
|
|
|
|
*/
|
1998-02-25 23:46:47 -05:00
|
|
|
char *
|
1996-07-09 02:22:35 -04:00
|
|
|
MapArrayTypeName(char *s)
|
|
|
|
|
{
|
1997-09-07 22:41:22 -04:00
|
|
|
int i,
|
|
|
|
|
j;
|
2005-10-14 22:49:52 -04:00
|
|
|
static char newStr[NAMEDATALEN]; /* array type names < NAMEDATALEN long */
|
1996-07-09 02:22:35 -04:00
|
|
|
|
1997-09-07 01:04:48 -04:00
|
|
|
if (s == NULL || s[0] == '\0')
|
|
|
|
|
return s;
|
1996-07-09 02:22:35 -04:00
|
|
|
|
1997-09-07 01:04:48 -04:00
|
|
|
j = 1;
|
|
|
|
|
newStr[0] = '_';
|
|
|
|
|
for (i = 0; i < NAMEDATALEN - 1 && s[i] != '['; i++, j++)
|
|
|
|
|
newStr[j] = s[i];
|
1996-07-09 02:22:35 -04:00
|
|
|
|
1997-09-07 01:04:48 -04:00
|
|
|
newStr[j] = '\0';
|
|
|
|
|
|
|
|
|
|
return newStr;
|
1996-07-09 02:22:35 -04:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
/*
|
1997-09-07 01:04:48 -04:00
|
|
|
* index_register() -- record an index that has been set up for building
|
|
|
|
|
* later.
|
1996-07-09 02:22:35 -04:00
|
|
|
*
|
2006-05-10 19:18:39 -04:00
|
|
|
* At bootstrap time, we define a bunch of indexes on system catalogs.
|
|
|
|
|
* We postpone actually building the indexes until just before we're
|
2014-05-06 12:12:18 -04:00
|
|
|
* finished with initialization, however. This is because the indexes
|
2006-05-10 19:18:39 -04:00
|
|
|
* themselves have catalog entries, and those have to be included in the
|
2014-05-06 12:12:18 -04:00
|
|
|
* indexes on those catalogs. Doing it in two phases is the simplest
|
2006-05-10 19:18:39 -04:00
|
|
|
* way of making sure the indexes have the right contents at the end.
|
1996-07-09 02:22:35 -04:00
|
|
|
*/
|
|
|
|
|
void
|
2002-03-26 14:17:02 -05:00
|
|
|
index_register(Oid heap,
|
|
|
|
|
Oid ind,
|
2000-07-14 18:18:02 -04:00
|
|
|
IndexInfo *indexInfo)
|
1996-07-09 02:22:35 -04:00
|
|
|
{
|
1997-09-07 22:41:22 -04:00
|
|
|
IndexList *newind;
|
|
|
|
|
MemoryContext oldcxt;
|
1997-09-07 01:04:48 -04:00
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* XXX mao 10/31/92 -- don't gc index reldescs, associated info at
|
2014-05-06 12:12:18 -04:00
|
|
|
* bootstrap time. we'll declare the indexes now, but want to create them
|
2005-10-14 22:49:52 -04:00
|
|
|
* later.
|
1997-09-07 01:04:48 -04:00
|
|
|
*/
|
|
|
|
|
|
2000-06-27 23:33:33 -04:00
|
|
|
if (nogc == NULL)
|
2004-01-07 13:56:30 -05:00
|
|
|
nogc = AllocSetContextCreate(NULL,
|
2000-06-27 23:33:33 -04:00
|
|
|
"BootstrapNoGC",
|
|
|
|
|
ALLOCSET_DEFAULT_MINSIZE,
|
|
|
|
|
ALLOCSET_DEFAULT_INITSIZE,
|
|
|
|
|
ALLOCSET_DEFAULT_MAXSIZE);
|
1997-09-07 01:04:48 -04:00
|
|
|
|
2000-06-27 23:33:33 -04:00
|
|
|
oldcxt = MemoryContextSwitchTo(nogc);
|
1997-09-07 01:04:48 -04:00
|
|
|
|
|
|
|
|
newind = (IndexList *) palloc(sizeof(IndexList));
|
2002-03-26 14:17:02 -05:00
|
|
|
newind->il_heap = heap;
|
|
|
|
|
newind->il_ind = ind;
|
2000-07-14 18:18:02 -04:00
|
|
|
newind->il_info = (IndexInfo *) palloc(sizeof(IndexInfo));
|
1997-09-07 01:04:48 -04:00
|
|
|
|
2000-07-14 18:18:02 -04:00
|
|
|
memcpy(newind->il_info, indexInfo, sizeof(IndexInfo));
|
2003-05-28 12:04:02 -04:00
|
|
|
/* expressions will likely be null, but may as well copy it */
|
|
|
|
|
newind->il_info->ii_Expressions = (List *)
|
|
|
|
|
copyObject(indexInfo->ii_Expressions);
|
|
|
|
|
newind->il_info->ii_ExpressionsState = NIL;
|
2001-07-16 01:07:00 -04:00
|
|
|
/* predicate will likely be null, but may as well copy it */
|
|
|
|
|
newind->il_info->ii_Predicate = (List *)
|
|
|
|
|
copyObject(indexInfo->ii_Predicate);
|
2002-12-15 11:17:59 -05:00
|
|
|
newind->il_info->ii_PredicateState = NIL;
|
2009-12-07 00:22:23 -05:00
|
|
|
/* no exclusion constraints at bootstrap time, so no need to copy */
|
|
|
|
|
Assert(indexInfo->ii_ExclusionOps == NULL);
|
|
|
|
|
Assert(indexInfo->ii_ExclusionProcs == NULL);
|
|
|
|
|
Assert(indexInfo->ii_ExclusionStrats == NULL);
|
2000-06-17 19:41:51 -04:00
|
|
|
|
1997-09-07 01:04:48 -04:00
|
|
|
newind->il_next = ILHead;
|
|
|
|
|
ILHead = newind;
|
|
|
|
|
|
|
|
|
|
MemoryContextSwitchTo(oldcxt);
|
1996-07-09 02:22:35 -04:00
|
|
|
}
|
|
|
|
|
|
2006-05-10 19:18:39 -04:00
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* build_indices -- fill in all the indexes registered earlier
|
|
|
|
|
*/
|
1996-07-09 02:22:35 -04:00
|
|
|
void
|
2004-01-07 13:56:30 -05:00
|
|
|
build_indices(void)
|
1996-07-09 02:22:35 -04:00
|
|
|
{
|
2004-01-07 13:56:30 -05:00
|
|
|
for (; ILHead != NULL; ILHead = ILHead->il_next)
|
1997-09-07 01:04:48 -04:00
|
|
|
{
|
2000-07-14 18:18:02 -04:00
|
|
|
Relation heap;
|
|
|
|
|
Relation ind;
|
|
|
|
|
|
2006-07-31 16:09:10 -04:00
|
|
|
/* need not bother with locks during bootstrap */
|
2002-03-26 14:17:02 -05:00
|
|
|
heap = heap_open(ILHead->il_heap, NoLock);
|
2006-07-31 16:09:10 -04:00
|
|
|
ind = index_open(ILHead->il_ind, NoLock);
|
2000-04-12 13:17:23 -04:00
|
|
|
|
2011-04-19 18:50:56 -04:00
|
|
|
index_build(heap, ind, ILHead->il_info, false, false);
|
1997-09-07 01:04:48 -04:00
|
|
|
|
2006-07-31 16:09:10 -04:00
|
|
|
index_close(ind, NoLock);
|
2006-05-10 19:18:39 -04:00
|
|
|
heap_close(heap, NoLock);
|
1997-09-07 01:04:48 -04:00
|
|
|
}
|
1996-07-09 02:22:35 -04:00
|
|
|
}
|