mirror of
https://github.com/postgres/postgres.git
synced 2025-04-22 23:02:54 +03:00
"wait" detection and add postmaster start time to help determine if the postmaster is actually using the specified data directory.
1300 lines
35 KiB
C
1300 lines
35 KiB
C
/*-------------------------------------------------------------------------
|
|
*
|
|
* miscinit.c
|
|
* miscellaneous initialization support stuff
|
|
*
|
|
* Portions Copyright (c) 1996-2010, PostgreSQL Global Development Group
|
|
* Portions Copyright (c) 1994, Regents of the University of California
|
|
*
|
|
*
|
|
* IDENTIFICATION
|
|
* src/backend/utils/init/miscinit.c
|
|
*
|
|
*-------------------------------------------------------------------------
|
|
*/
|
|
#include "postgres.h"
|
|
|
|
#include <sys/param.h>
|
|
#include <signal.h>
|
|
#include <sys/file.h>
|
|
#include <sys/stat.h>
|
|
#include <sys/time.h>
|
|
#include <fcntl.h>
|
|
#include <unistd.h>
|
|
#include <grp.h>
|
|
#include <pwd.h>
|
|
#include <netinet/in.h>
|
|
#include <arpa/inet.h>
|
|
#ifdef HAVE_UTIME_H
|
|
#include <utime.h>
|
|
#endif
|
|
|
|
#include "catalog/pg_authid.h"
|
|
#include "mb/pg_wchar.h"
|
|
#include "miscadmin.h"
|
|
#include "postmaster/autovacuum.h"
|
|
#include "postmaster/postmaster.h"
|
|
#include "storage/fd.h"
|
|
#include "storage/ipc.h"
|
|
#include "storage/pg_shmem.h"
|
|
#include "storage/proc.h"
|
|
#include "storage/procarray.h"
|
|
#include "utils/builtins.h"
|
|
#include "utils/guc.h"
|
|
#include "utils/memutils.h"
|
|
#include "utils/syscache.h"
|
|
|
|
|
|
#define DIRECTORY_LOCK_FILE "postmaster.pid"
|
|
/*
|
|
* The lock file contents are:
|
|
*
|
|
* line #
|
|
* 1 pid
|
|
* 2 postmaster start time
|
|
* 3 data directory
|
|
* 4 port #
|
|
* 5 user-specified socket directory
|
|
* (the lines below are added later)
|
|
* 6 first valid listen_address
|
|
* 7 shared memory key
|
|
*/
|
|
|
|
ProcessingMode Mode = InitProcessing;
|
|
|
|
/* Note: we rely on this to initialize as zeroes */
|
|
static char socketLockFile[MAXPGPATH];
|
|
|
|
|
|
/* ----------------------------------------------------------------
|
|
* ignoring system indexes support stuff
|
|
*
|
|
* NOTE: "ignoring system indexes" means we do not use the system indexes
|
|
* for lookups (either in hardwired catalog accesses or in planner-generated
|
|
* plans). We do, however, still update the indexes when a catalog
|
|
* modification is made.
|
|
* ----------------------------------------------------------------
|
|
*/
|
|
|
|
bool IgnoreSystemIndexes = false;
|
|
|
|
|
|
/* ----------------------------------------------------------------
|
|
* database path / name support stuff
|
|
* ----------------------------------------------------------------
|
|
*/
|
|
|
|
void
|
|
SetDatabasePath(const char *path)
|
|
{
|
|
/* This should happen only once per process */
|
|
Assert(!DatabasePath);
|
|
DatabasePath = MemoryContextStrdup(TopMemoryContext, path);
|
|
}
|
|
|
|
/*
|
|
* Set data directory, but make sure it's an absolute path. Use this,
|
|
* never set DataDir directly.
|
|
*/
|
|
void
|
|
SetDataDir(const char *dir)
|
|
{
|
|
char *new;
|
|
|
|
AssertArg(dir);
|
|
|
|
/* If presented path is relative, convert to absolute */
|
|
new = make_absolute_path(dir);
|
|
|
|
if (DataDir)
|
|
free(DataDir);
|
|
DataDir = new;
|
|
}
|
|
|
|
/*
|
|
* Change working directory to DataDir. Most of the postmaster and backend
|
|
* code assumes that we are in DataDir so it can use relative paths to access
|
|
* stuff in and under the data directory. For convenience during path
|
|
* setup, however, we don't force the chdir to occur during SetDataDir.
|
|
*/
|
|
void
|
|
ChangeToDataDir(void)
|
|
{
|
|
AssertState(DataDir);
|
|
|
|
if (chdir(DataDir) < 0)
|
|
ereport(FATAL,
|
|
(errcode_for_file_access(),
|
|
errmsg("could not change directory to \"%s\": %m",
|
|
DataDir)));
|
|
}
|
|
|
|
/*
|
|
* If the given pathname isn't already absolute, make it so, interpreting
|
|
* it relative to the current working directory.
|
|
*
|
|
* Also canonicalizes the path. The result is always a malloc'd copy.
|
|
*
|
|
* Note: interpretation of relative-path arguments during postmaster startup
|
|
* should happen before doing ChangeToDataDir(), else the user will probably
|
|
* not like the results.
|
|
*/
|
|
char *
|
|
make_absolute_path(const char *path)
|
|
{
|
|
char *new;
|
|
|
|
/* Returning null for null input is convenient for some callers */
|
|
if (path == NULL)
|
|
return NULL;
|
|
|
|
if (!is_absolute_path(path))
|
|
{
|
|
char *buf;
|
|
size_t buflen;
|
|
|
|
buflen = MAXPGPATH;
|
|
for (;;)
|
|
{
|
|
buf = malloc(buflen);
|
|
if (!buf)
|
|
ereport(FATAL,
|
|
(errcode(ERRCODE_OUT_OF_MEMORY),
|
|
errmsg("out of memory")));
|
|
|
|
if (getcwd(buf, buflen))
|
|
break;
|
|
else if (errno == ERANGE)
|
|
{
|
|
free(buf);
|
|
buflen *= 2;
|
|
continue;
|
|
}
|
|
else
|
|
{
|
|
free(buf);
|
|
elog(FATAL, "could not get current working directory: %m");
|
|
}
|
|
}
|
|
|
|
new = malloc(strlen(buf) + strlen(path) + 2);
|
|
if (!new)
|
|
ereport(FATAL,
|
|
(errcode(ERRCODE_OUT_OF_MEMORY),
|
|
errmsg("out of memory")));
|
|
sprintf(new, "%s/%s", buf, path);
|
|
free(buf);
|
|
}
|
|
else
|
|
{
|
|
new = strdup(path);
|
|
if (!new)
|
|
ereport(FATAL,
|
|
(errcode(ERRCODE_OUT_OF_MEMORY),
|
|
errmsg("out of memory")));
|
|
}
|
|
|
|
/* Make sure punctuation is canonical, too */
|
|
canonicalize_path(new);
|
|
|
|
return new;
|
|
}
|
|
|
|
|
|
/* ----------------------------------------------------------------
|
|
* User ID state
|
|
*
|
|
* We have to track several different values associated with the concept
|
|
* of "user ID".
|
|
*
|
|
* AuthenticatedUserId is determined at connection start and never changes.
|
|
*
|
|
* SessionUserId is initially the same as AuthenticatedUserId, but can be
|
|
* changed by SET SESSION AUTHORIZATION (if AuthenticatedUserIsSuperuser).
|
|
* This is the ID reported by the SESSION_USER SQL function.
|
|
*
|
|
* OuterUserId is the current user ID in effect at the "outer level" (outside
|
|
* any transaction or function). This is initially the same as SessionUserId,
|
|
* but can be changed by SET ROLE to any role that SessionUserId is a
|
|
* member of. (XXX rename to something like CurrentRoleId?)
|
|
*
|
|
* CurrentUserId is the current effective user ID; this is the one to use
|
|
* for all normal permissions-checking purposes. At outer level this will
|
|
* be the same as OuterUserId, but it changes during calls to SECURITY
|
|
* DEFINER functions, as well as locally in some specialized commands.
|
|
*
|
|
* SecurityRestrictionContext holds flags indicating reason(s) for changing
|
|
* CurrentUserId. In some cases we need to lock down operations that are
|
|
* not directly controlled by privilege settings, and this provides a
|
|
* convenient way to do it.
|
|
* ----------------------------------------------------------------
|
|
*/
|
|
static Oid AuthenticatedUserId = InvalidOid;
|
|
static Oid SessionUserId = InvalidOid;
|
|
static Oid OuterUserId = InvalidOid;
|
|
static Oid CurrentUserId = InvalidOid;
|
|
|
|
/* We also have to remember the superuser state of some of these levels */
|
|
static bool AuthenticatedUserIsSuperuser = false;
|
|
static bool SessionUserIsSuperuser = false;
|
|
|
|
static int SecurityRestrictionContext = 0;
|
|
|
|
/* We also remember if a SET ROLE is currently active */
|
|
static bool SetRoleIsActive = false;
|
|
|
|
|
|
|
|
/*
|
|
* GetUserId - get the current effective user ID.
|
|
*
|
|
* Note: there's no SetUserId() anymore; use SetUserIdAndSecContext().
|
|
*/
|
|
Oid
|
|
GetUserId(void)
|
|
{
|
|
AssertState(OidIsValid(CurrentUserId));
|
|
return CurrentUserId;
|
|
}
|
|
|
|
|
|
/*
|
|
* GetOuterUserId/SetOuterUserId - get/set the outer-level user ID.
|
|
*/
|
|
Oid
|
|
GetOuterUserId(void)
|
|
{
|
|
AssertState(OidIsValid(OuterUserId));
|
|
return OuterUserId;
|
|
}
|
|
|
|
|
|
static void
|
|
SetOuterUserId(Oid userid)
|
|
{
|
|
AssertState(SecurityRestrictionContext == 0);
|
|
AssertArg(OidIsValid(userid));
|
|
OuterUserId = userid;
|
|
|
|
/* We force the effective user ID to match, too */
|
|
CurrentUserId = userid;
|
|
}
|
|
|
|
|
|
/*
|
|
* GetSessionUserId/SetSessionUserId - get/set the session user ID.
|
|
*/
|
|
Oid
|
|
GetSessionUserId(void)
|
|
{
|
|
AssertState(OidIsValid(SessionUserId));
|
|
return SessionUserId;
|
|
}
|
|
|
|
|
|
static void
|
|
SetSessionUserId(Oid userid, bool is_superuser)
|
|
{
|
|
AssertState(SecurityRestrictionContext == 0);
|
|
AssertArg(OidIsValid(userid));
|
|
SessionUserId = userid;
|
|
SessionUserIsSuperuser = is_superuser;
|
|
SetRoleIsActive = false;
|
|
|
|
/* We force the effective user IDs to match, too */
|
|
OuterUserId = userid;
|
|
CurrentUserId = userid;
|
|
}
|
|
|
|
|
|
/*
|
|
* GetUserIdAndSecContext/SetUserIdAndSecContext - get/set the current user ID
|
|
* and the SecurityRestrictionContext flags.
|
|
*
|
|
* Currently there are two valid bits in SecurityRestrictionContext:
|
|
*
|
|
* SECURITY_LOCAL_USERID_CHANGE indicates that we are inside an operation
|
|
* that is temporarily changing CurrentUserId via these functions. This is
|
|
* needed to indicate that the actual value of CurrentUserId is not in sync
|
|
* with guc.c's internal state, so SET ROLE has to be disallowed.
|
|
*
|
|
* SECURITY_RESTRICTED_OPERATION indicates that we are inside an operation
|
|
* that does not wish to trust called user-defined functions at all. This
|
|
* bit prevents not only SET ROLE, but various other changes of session state
|
|
* that normally is unprotected but might possibly be used to subvert the
|
|
* calling session later. An example is replacing an existing prepared
|
|
* statement with new code, which will then be executed with the outer
|
|
* session's permissions when the prepared statement is next used. Since
|
|
* these restrictions are fairly draconian, we apply them only in contexts
|
|
* where the called functions are really supposed to be side-effect-free
|
|
* anyway, such as VACUUM/ANALYZE/REINDEX.
|
|
*
|
|
* Unlike GetUserId, GetUserIdAndSecContext does *not* Assert that the current
|
|
* value of CurrentUserId is valid; nor does SetUserIdAndSecContext require
|
|
* the new value to be valid. In fact, these routines had better not
|
|
* ever throw any kind of error. This is because they are used by
|
|
* StartTransaction and AbortTransaction to save/restore the settings,
|
|
* and during the first transaction within a backend, the value to be saved
|
|
* and perhaps restored is indeed invalid. We have to be able to get
|
|
* through AbortTransaction without asserting in case InitPostgres fails.
|
|
*/
|
|
void
|
|
GetUserIdAndSecContext(Oid *userid, int *sec_context)
|
|
{
|
|
*userid = CurrentUserId;
|
|
*sec_context = SecurityRestrictionContext;
|
|
}
|
|
|
|
void
|
|
SetUserIdAndSecContext(Oid userid, int sec_context)
|
|
{
|
|
CurrentUserId = userid;
|
|
SecurityRestrictionContext = sec_context;
|
|
}
|
|
|
|
|
|
/*
|
|
* InLocalUserIdChange - are we inside a local change of CurrentUserId?
|
|
*/
|
|
bool
|
|
InLocalUserIdChange(void)
|
|
{
|
|
return (SecurityRestrictionContext & SECURITY_LOCAL_USERID_CHANGE) != 0;
|
|
}
|
|
|
|
/*
|
|
* InSecurityRestrictedOperation - are we inside a security-restricted command?
|
|
*/
|
|
bool
|
|
InSecurityRestrictedOperation(void)
|
|
{
|
|
return (SecurityRestrictionContext & SECURITY_RESTRICTED_OPERATION) != 0;
|
|
}
|
|
|
|
|
|
/*
|
|
* These are obsolete versions of Get/SetUserIdAndSecContext that are
|
|
* only provided for bug-compatibility with some rather dubious code in
|
|
* pljava. We allow the userid to be set, but only when not inside a
|
|
* security restriction context.
|
|
*/
|
|
void
|
|
GetUserIdAndContext(Oid *userid, bool *sec_def_context)
|
|
{
|
|
*userid = CurrentUserId;
|
|
*sec_def_context = InLocalUserIdChange();
|
|
}
|
|
|
|
void
|
|
SetUserIdAndContext(Oid userid, bool sec_def_context)
|
|
{
|
|
/* We throw the same error SET ROLE would. */
|
|
if (InSecurityRestrictedOperation())
|
|
ereport(ERROR,
|
|
(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
|
|
errmsg("cannot set parameter \"%s\" within security-restricted operation",
|
|
"role")));
|
|
CurrentUserId = userid;
|
|
if (sec_def_context)
|
|
SecurityRestrictionContext |= SECURITY_LOCAL_USERID_CHANGE;
|
|
else
|
|
SecurityRestrictionContext &= ~SECURITY_LOCAL_USERID_CHANGE;
|
|
}
|
|
|
|
|
|
/*
|
|
* Check if the authenticated user is a replication role
|
|
*/
|
|
bool
|
|
is_authenticated_user_replication_role(void)
|
|
{
|
|
bool result = false;
|
|
HeapTuple utup;
|
|
|
|
utup = SearchSysCache1(AUTHOID, ObjectIdGetDatum(AuthenticatedUserId));
|
|
if (HeapTupleIsValid(utup))
|
|
{
|
|
result = ((Form_pg_authid) GETSTRUCT(utup))->rolreplication;
|
|
ReleaseSysCache(utup);
|
|
}
|
|
return result;
|
|
}
|
|
|
|
/*
|
|
* Initialize user identity during normal backend startup
|
|
*/
|
|
void
|
|
InitializeSessionUserId(const char *rolename)
|
|
{
|
|
HeapTuple roleTup;
|
|
Form_pg_authid rform;
|
|
Oid roleid;
|
|
|
|
/*
|
|
* Don't do scans if we're bootstrapping, none of the system catalogs
|
|
* exist yet, and they should be owned by postgres anyway.
|
|
*/
|
|
AssertState(!IsBootstrapProcessingMode());
|
|
|
|
/* call only once */
|
|
AssertState(!OidIsValid(AuthenticatedUserId));
|
|
|
|
roleTup = SearchSysCache1(AUTHNAME, PointerGetDatum(rolename));
|
|
if (!HeapTupleIsValid(roleTup))
|
|
ereport(FATAL,
|
|
(errcode(ERRCODE_INVALID_AUTHORIZATION_SPECIFICATION),
|
|
errmsg("role \"%s\" does not exist", rolename)));
|
|
|
|
rform = (Form_pg_authid) GETSTRUCT(roleTup);
|
|
roleid = HeapTupleGetOid(roleTup);
|
|
|
|
AuthenticatedUserId = roleid;
|
|
AuthenticatedUserIsSuperuser = rform->rolsuper;
|
|
|
|
/* This sets OuterUserId/CurrentUserId too */
|
|
SetSessionUserId(roleid, AuthenticatedUserIsSuperuser);
|
|
|
|
/* Also mark our PGPROC entry with the authenticated user id */
|
|
/* (We assume this is an atomic store so no lock is needed) */
|
|
MyProc->roleId = roleid;
|
|
|
|
/*
|
|
* These next checks are not enforced when in standalone mode, so that
|
|
* there is a way to recover from sillinesses like "UPDATE pg_authid SET
|
|
* rolcanlogin = false;".
|
|
*/
|
|
if (IsUnderPostmaster)
|
|
{
|
|
/*
|
|
* Is role allowed to login at all?
|
|
*/
|
|
if (!rform->rolcanlogin)
|
|
ereport(FATAL,
|
|
(errcode(ERRCODE_INVALID_AUTHORIZATION_SPECIFICATION),
|
|
errmsg("role \"%s\" is not permitted to log in",
|
|
rolename)));
|
|
|
|
/*
|
|
* Check connection limit for this role.
|
|
*
|
|
* There is a race condition here --- we create our PGPROC before
|
|
* checking for other PGPROCs. If two backends did this at about the
|
|
* same time, they might both think they were over the limit, while
|
|
* ideally one should succeed and one fail. Getting that to work
|
|
* exactly seems more trouble than it is worth, however; instead we
|
|
* just document that the connection limit is approximate.
|
|
*/
|
|
if (rform->rolconnlimit >= 0 &&
|
|
!AuthenticatedUserIsSuperuser &&
|
|
CountUserBackends(roleid) > rform->rolconnlimit)
|
|
ereport(FATAL,
|
|
(errcode(ERRCODE_TOO_MANY_CONNECTIONS),
|
|
errmsg("too many connections for role \"%s\"",
|
|
rolename)));
|
|
}
|
|
|
|
/* Record username and superuser status as GUC settings too */
|
|
SetConfigOption("session_authorization", rolename,
|
|
PGC_BACKEND, PGC_S_OVERRIDE);
|
|
SetConfigOption("is_superuser",
|
|
AuthenticatedUserIsSuperuser ? "on" : "off",
|
|
PGC_INTERNAL, PGC_S_OVERRIDE);
|
|
|
|
ReleaseSysCache(roleTup);
|
|
}
|
|
|
|
|
|
/*
|
|
* Initialize user identity during special backend startup
|
|
*/
|
|
void
|
|
InitializeSessionUserIdStandalone(void)
|
|
{
|
|
/*
|
|
* This function should only be called in single-user mode and in
|
|
* autovacuum workers.
|
|
*/
|
|
AssertState(!IsUnderPostmaster || IsAutoVacuumWorkerProcess());
|
|
|
|
/* call only once */
|
|
AssertState(!OidIsValid(AuthenticatedUserId));
|
|
|
|
AuthenticatedUserId = BOOTSTRAP_SUPERUSERID;
|
|
AuthenticatedUserIsSuperuser = true;
|
|
|
|
SetSessionUserId(BOOTSTRAP_SUPERUSERID, true);
|
|
}
|
|
|
|
|
|
/*
|
|
* Change session auth ID while running
|
|
*
|
|
* Only a superuser may set auth ID to something other than himself. Note
|
|
* that in case of multiple SETs in a single session, the original userid's
|
|
* superuserness is what matters. But we set the GUC variable is_superuser
|
|
* to indicate whether the *current* session userid is a superuser.
|
|
*
|
|
* Note: this is not an especially clean place to do the permission check.
|
|
* It's OK because the check does not require catalog access and can't
|
|
* fail during an end-of-transaction GUC reversion, but we may someday
|
|
* have to push it up into assign_session_authorization.
|
|
*/
|
|
void
|
|
SetSessionAuthorization(Oid userid, bool is_superuser)
|
|
{
|
|
/* Must have authenticated already, else can't make permission check */
|
|
AssertState(OidIsValid(AuthenticatedUserId));
|
|
|
|
if (userid != AuthenticatedUserId &&
|
|
!AuthenticatedUserIsSuperuser)
|
|
ereport(ERROR,
|
|
(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
|
|
errmsg("permission denied to set session authorization")));
|
|
|
|
SetSessionUserId(userid, is_superuser);
|
|
|
|
SetConfigOption("is_superuser",
|
|
is_superuser ? "on" : "off",
|
|
PGC_INTERNAL, PGC_S_OVERRIDE);
|
|
}
|
|
|
|
/*
|
|
* Report current role id
|
|
* This follows the semantics of SET ROLE, ie return the outer-level ID
|
|
* not the current effective ID, and return InvalidOid when the setting
|
|
* is logically SET ROLE NONE.
|
|
*/
|
|
Oid
|
|
GetCurrentRoleId(void)
|
|
{
|
|
if (SetRoleIsActive)
|
|
return OuterUserId;
|
|
else
|
|
return InvalidOid;
|
|
}
|
|
|
|
/*
|
|
* Change Role ID while running (SET ROLE)
|
|
*
|
|
* If roleid is InvalidOid, we are doing SET ROLE NONE: revert to the
|
|
* session user authorization. In this case the is_superuser argument
|
|
* is ignored.
|
|
*
|
|
* When roleid is not InvalidOid, the caller must have checked whether
|
|
* the session user has permission to become that role. (We cannot check
|
|
* here because this routine must be able to execute in a failed transaction
|
|
* to restore a prior value of the ROLE GUC variable.)
|
|
*/
|
|
void
|
|
SetCurrentRoleId(Oid roleid, bool is_superuser)
|
|
{
|
|
/*
|
|
* Get correct info if it's SET ROLE NONE
|
|
*
|
|
* If SessionUserId hasn't been set yet, just do nothing --- the eventual
|
|
* SetSessionUserId call will fix everything. This is needed since we
|
|
* will get called during GUC initialization.
|
|
*/
|
|
if (!OidIsValid(roleid))
|
|
{
|
|
if (!OidIsValid(SessionUserId))
|
|
return;
|
|
|
|
roleid = SessionUserId;
|
|
is_superuser = SessionUserIsSuperuser;
|
|
|
|
SetRoleIsActive = false;
|
|
}
|
|
else
|
|
SetRoleIsActive = true;
|
|
|
|
SetOuterUserId(roleid);
|
|
|
|
SetConfigOption("is_superuser",
|
|
is_superuser ? "on" : "off",
|
|
PGC_INTERNAL, PGC_S_OVERRIDE);
|
|
}
|
|
|
|
|
|
/*
|
|
* Get user name from user oid
|
|
*/
|
|
char *
|
|
GetUserNameFromId(Oid roleid)
|
|
{
|
|
HeapTuple tuple;
|
|
char *result;
|
|
|
|
tuple = SearchSysCache1(AUTHOID, ObjectIdGetDatum(roleid));
|
|
if (!HeapTupleIsValid(tuple))
|
|
ereport(ERROR,
|
|
(errcode(ERRCODE_UNDEFINED_OBJECT),
|
|
errmsg("invalid role OID: %u", roleid)));
|
|
|
|
result = pstrdup(NameStr(((Form_pg_authid) GETSTRUCT(tuple))->rolname));
|
|
|
|
ReleaseSysCache(tuple);
|
|
return result;
|
|
}
|
|
|
|
|
|
/*-------------------------------------------------------------------------
|
|
* Interlock-file support
|
|
*
|
|
* These routines are used to create both a data-directory lockfile
|
|
* ($DATADIR/postmaster.pid) and a Unix-socket-file lockfile ($SOCKFILE.lock).
|
|
* Both kinds of files contain the same info:
|
|
*
|
|
* Owning process' PID
|
|
* Data directory path
|
|
*
|
|
* By convention, the owning process' PID is negated if it is a standalone
|
|
* backend rather than a postmaster. This is just for informational purposes.
|
|
* The path is also just for informational purposes (so that a socket lockfile
|
|
* can be more easily traced to the associated postmaster).
|
|
*
|
|
* A data-directory lockfile can optionally contain a third line, containing
|
|
* the key and ID for the shared memory block used by this postmaster.
|
|
*
|
|
* On successful lockfile creation, a proc_exit callback to remove the
|
|
* lockfile is automatically created.
|
|
*-------------------------------------------------------------------------
|
|
*/
|
|
|
|
/*
|
|
* proc_exit callback to remove a lockfile.
|
|
*/
|
|
static void
|
|
UnlinkLockFile(int status, Datum filename)
|
|
{
|
|
char *fname = (char *) DatumGetPointer(filename);
|
|
|
|
if (fname != NULL)
|
|
{
|
|
if (unlink(fname) != 0)
|
|
{
|
|
/* Should we complain if the unlink fails? */
|
|
}
|
|
free(fname);
|
|
}
|
|
}
|
|
|
|
/*
|
|
* Create a lockfile.
|
|
*
|
|
* filename is the name of the lockfile to create.
|
|
* amPostmaster is used to determine how to encode the output PID.
|
|
* isDDLock and refName are used to determine what error message to produce.
|
|
*/
|
|
static void
|
|
CreateLockFile(const char *filename, bool amPostmaster,
|
|
bool isDDLock, const char *refName)
|
|
{
|
|
int fd;
|
|
char buffer[MAXPGPATH * 3 + 256];
|
|
int ntries;
|
|
int len;
|
|
int encoded_pid;
|
|
pid_t other_pid;
|
|
pid_t my_pid,
|
|
my_p_pid,
|
|
my_gp_pid;
|
|
const char *envvar;
|
|
|
|
/*
|
|
* If the PID in the lockfile is our own PID or our parent's or
|
|
* grandparent's PID, then the file must be stale (probably left over from
|
|
* a previous system boot cycle). We need to check this because of the
|
|
* likelihood that a reboot will assign exactly the same PID as we had in
|
|
* the previous reboot, or one that's only one or two counts larger and
|
|
* hence the lockfile's PID now refers to an ancestor shell process. We
|
|
* allow pg_ctl to pass down its parent shell PID (our grandparent PID)
|
|
* via the environment variable PG_GRANDPARENT_PID; this is so that
|
|
* launching the postmaster via pg_ctl can be just as reliable as
|
|
* launching it directly. There is no provision for detecting
|
|
* further-removed ancestor processes, but if the init script is written
|
|
* carefully then all but the immediate parent shell will be root-owned
|
|
* processes and so the kill test will fail with EPERM. Note that we
|
|
* cannot get a false negative this way, because an existing postmaster
|
|
* would surely never launch a competing postmaster or pg_ctl process
|
|
* directly.
|
|
*/
|
|
my_pid = getpid();
|
|
|
|
#ifndef WIN32
|
|
my_p_pid = getppid();
|
|
#else
|
|
|
|
/*
|
|
* Windows hasn't got getppid(), but doesn't need it since it's not using
|
|
* real kill() either...
|
|
*/
|
|
my_p_pid = 0;
|
|
#endif
|
|
|
|
envvar = getenv("PG_GRANDPARENT_PID");
|
|
if (envvar)
|
|
my_gp_pid = atoi(envvar);
|
|
else
|
|
my_gp_pid = 0;
|
|
|
|
/*
|
|
* We need a loop here because of race conditions. But don't loop forever
|
|
* (for example, a non-writable $PGDATA directory might cause a failure
|
|
* that won't go away). 100 tries seems like plenty.
|
|
*/
|
|
for (ntries = 0;; ntries++)
|
|
{
|
|
/*
|
|
* Try to create the lock file --- O_EXCL makes this atomic.
|
|
*
|
|
* Think not to make the file protection weaker than 0600. See
|
|
* comments below.
|
|
*/
|
|
fd = open(filename, O_RDWR | O_CREAT | O_EXCL, 0600);
|
|
if (fd >= 0)
|
|
break; /* Success; exit the retry loop */
|
|
|
|
/*
|
|
* Couldn't create the pid file. Probably it already exists.
|
|
*/
|
|
if ((errno != EEXIST && errno != EACCES) || ntries > 100)
|
|
ereport(FATAL,
|
|
(errcode_for_file_access(),
|
|
errmsg("could not create lock file \"%s\": %m",
|
|
filename)));
|
|
|
|
/*
|
|
* Read the file to get the old owner's PID. Note race condition
|
|
* here: file might have been deleted since we tried to create it.
|
|
*/
|
|
fd = open(filename, O_RDONLY, 0600);
|
|
if (fd < 0)
|
|
{
|
|
if (errno == ENOENT)
|
|
continue; /* race condition; try again */
|
|
ereport(FATAL,
|
|
(errcode_for_file_access(),
|
|
errmsg("could not open lock file \"%s\": %m",
|
|
filename)));
|
|
}
|
|
if ((len = read(fd, buffer, sizeof(buffer) - 1)) < 0)
|
|
ereport(FATAL,
|
|
(errcode_for_file_access(),
|
|
errmsg("could not read lock file \"%s\": %m",
|
|
filename)));
|
|
close(fd);
|
|
|
|
buffer[len] = '\0';
|
|
encoded_pid = atoi(buffer);
|
|
|
|
/* if pid < 0, the pid is for postgres, not postmaster */
|
|
other_pid = (pid_t) (encoded_pid < 0 ? -encoded_pid : encoded_pid);
|
|
|
|
if (other_pid <= 0)
|
|
elog(FATAL, "bogus data in lock file \"%s\": \"%s\"",
|
|
filename, buffer);
|
|
|
|
/*
|
|
* Check to see if the other process still exists
|
|
*
|
|
* Per discussion above, my_pid, my_p_pid, and my_gp_pid can be
|
|
* ignored as false matches.
|
|
*
|
|
* Normally kill() will fail with ESRCH if the given PID doesn't
|
|
* exist.
|
|
*
|
|
* We can treat the EPERM-error case as okay because that error
|
|
* implies that the existing process has a different userid than we
|
|
* do, which means it cannot be a competing postmaster. A postmaster
|
|
* cannot successfully attach to a data directory owned by a userid
|
|
* other than its own. (This is now checked directly in
|
|
* checkDataDir(), but has been true for a long time because of the
|
|
* restriction that the data directory isn't group- or
|
|
* world-accessible.) Also, since we create the lockfiles mode 600,
|
|
* we'd have failed above if the lockfile belonged to another userid
|
|
* --- which means that whatever process kill() is reporting about
|
|
* isn't the one that made the lockfile. (NOTE: this last
|
|
* consideration is the only one that keeps us from blowing away a
|
|
* Unix socket file belonging to an instance of Postgres being run by
|
|
* someone else, at least on machines where /tmp hasn't got a
|
|
* stickybit.)
|
|
*/
|
|
if (other_pid != my_pid && other_pid != my_p_pid &&
|
|
other_pid != my_gp_pid)
|
|
{
|
|
if (kill(other_pid, 0) == 0 ||
|
|
(errno != ESRCH && errno != EPERM))
|
|
{
|
|
/* lockfile belongs to a live process */
|
|
ereport(FATAL,
|
|
(errcode(ERRCODE_LOCK_FILE_EXISTS),
|
|
errmsg("lock file \"%s\" already exists",
|
|
filename),
|
|
isDDLock ?
|
|
(encoded_pid < 0 ?
|
|
errhint("Is another postgres (PID %d) running in data directory \"%s\"?",
|
|
(int) other_pid, refName) :
|
|
errhint("Is another postmaster (PID %d) running in data directory \"%s\"?",
|
|
(int) other_pid, refName)) :
|
|
(encoded_pid < 0 ?
|
|
errhint("Is another postgres (PID %d) using socket file \"%s\"?",
|
|
(int) other_pid, refName) :
|
|
errhint("Is another postmaster (PID %d) using socket file \"%s\"?",
|
|
(int) other_pid, refName))));
|
|
}
|
|
}
|
|
|
|
/*
|
|
* No, the creating process did not exist. However, it could be that
|
|
* the postmaster crashed (or more likely was kill -9'd by a clueless
|
|
* admin) but has left orphan backends behind. Check for this by
|
|
* looking to see if there is an associated shmem segment that is
|
|
* still in use.
|
|
*
|
|
* Note: because postmaster.pid is written in two steps, we might not
|
|
* find the shmem ID values in it; we can't treat that as an error.
|
|
*/
|
|
if (isDDLock)
|
|
{
|
|
char *ptr = buffer;
|
|
unsigned long id1, id2;
|
|
int lineno;
|
|
|
|
for (lineno = 1; lineno <= LOCK_FILE_LINES - 1; lineno++)
|
|
{
|
|
if ((ptr = strchr(ptr, '\n')) == NULL)
|
|
{
|
|
elog(LOG, "bogus data in \"%s\"", DIRECTORY_LOCK_FILE);
|
|
break;
|
|
}
|
|
ptr++;
|
|
}
|
|
|
|
if (ptr && sscanf(ptr, "%lu %lu", &id1, &id2) == 2)
|
|
{
|
|
if (PGSharedMemoryIsInUse(id1, id2))
|
|
ereport(FATAL,
|
|
(errcode(ERRCODE_LOCK_FILE_EXISTS),
|
|
errmsg("pre-existing shared memory block "
|
|
"(key %lu, ID %lu) is still in use",
|
|
id1, id2),
|
|
errhint("If you're sure there are no old "
|
|
"server processes still running, remove "
|
|
"the shared memory block "
|
|
"or just delete the file \"%s\".",
|
|
filename)));
|
|
}
|
|
}
|
|
|
|
/*
|
|
* Looks like nobody's home. Unlink the file and try again to create
|
|
* it. Need a loop because of possible race condition against other
|
|
* would-be creators.
|
|
*/
|
|
if (unlink(filename) < 0)
|
|
ereport(FATAL,
|
|
(errcode_for_file_access(),
|
|
errmsg("could not remove old lock file \"%s\": %m",
|
|
filename),
|
|
errhint("The file seems accidentally left over, but "
|
|
"it could not be removed. Please remove the file "
|
|
"by hand and try again.")));
|
|
}
|
|
|
|
/*
|
|
* Successfully created the file, now fill it.
|
|
*/
|
|
snprintf(buffer, sizeof(buffer), "%d\n%ld\n%s\n%d\n%s\n",
|
|
amPostmaster ? (int) my_pid : -((int) my_pid),
|
|
(long) MyStartTime, DataDir, PostPortNumber,
|
|
UnixSocketDir);
|
|
errno = 0;
|
|
if (write(fd, buffer, strlen(buffer)) != strlen(buffer))
|
|
{
|
|
int save_errno = errno;
|
|
|
|
close(fd);
|
|
unlink(filename);
|
|
/* if write didn't set errno, assume problem is no disk space */
|
|
errno = save_errno ? save_errno : ENOSPC;
|
|
ereport(FATAL,
|
|
(errcode_for_file_access(),
|
|
errmsg("could not write lock file \"%s\": %m", filename)));
|
|
}
|
|
if (pg_fsync(fd) != 0)
|
|
{
|
|
int save_errno = errno;
|
|
|
|
close(fd);
|
|
unlink(filename);
|
|
errno = save_errno;
|
|
ereport(FATAL,
|
|
(errcode_for_file_access(),
|
|
errmsg("could not write lock file \"%s\": %m", filename)));
|
|
}
|
|
if (close(fd) != 0)
|
|
{
|
|
int save_errno = errno;
|
|
|
|
unlink(filename);
|
|
errno = save_errno;
|
|
ereport(FATAL,
|
|
(errcode_for_file_access(),
|
|
errmsg("could not write lock file \"%s\": %m", filename)));
|
|
}
|
|
|
|
/*
|
|
* Arrange for automatic removal of lockfile at proc_exit.
|
|
*/
|
|
on_proc_exit(UnlinkLockFile, PointerGetDatum(strdup(filename)));
|
|
}
|
|
|
|
/*
|
|
* Create the data directory lockfile.
|
|
*
|
|
* When this is called, we must have already switched the working
|
|
* directory to DataDir, so we can just use a relative path. This
|
|
* helps ensure that we are locking the directory we should be.
|
|
*/
|
|
void
|
|
CreateDataDirLockFile(bool amPostmaster)
|
|
{
|
|
CreateLockFile(DIRECTORY_LOCK_FILE, amPostmaster, true, DataDir);
|
|
}
|
|
|
|
/*
|
|
* Create a lockfile for the specified Unix socket file.
|
|
*/
|
|
void
|
|
CreateSocketLockFile(const char *socketfile, bool amPostmaster)
|
|
{
|
|
char lockfile[MAXPGPATH];
|
|
|
|
snprintf(lockfile, sizeof(lockfile), "%s.lock", socketfile);
|
|
CreateLockFile(lockfile, amPostmaster, false, socketfile);
|
|
/* Save name of lockfile for TouchSocketLockFile */
|
|
strcpy(socketLockFile, lockfile);
|
|
}
|
|
|
|
/*
|
|
* TouchSocketLockFile -- mark socket lock file as recently accessed
|
|
*
|
|
* This routine should be called every so often to ensure that the lock file
|
|
* has a recent mod or access date. That saves it
|
|
* from being removed by overenthusiastic /tmp-directory-cleaner daemons.
|
|
* (Another reason we should never have put the socket file in /tmp...)
|
|
*/
|
|
void
|
|
TouchSocketLockFile(void)
|
|
{
|
|
/* Do nothing if we did not create a socket... */
|
|
if (socketLockFile[0] != '\0')
|
|
{
|
|
/*
|
|
* utime() is POSIX standard, utimes() is a common alternative; if we
|
|
* have neither, fall back to actually reading the file (which only
|
|
* sets the access time not mod time, but that should be enough in
|
|
* most cases). In all paths, we ignore errors.
|
|
*/
|
|
#ifdef HAVE_UTIME
|
|
utime(socketLockFile, NULL);
|
|
#else /* !HAVE_UTIME */
|
|
#ifdef HAVE_UTIMES
|
|
utimes(socketLockFile, NULL);
|
|
#else /* !HAVE_UTIMES */
|
|
int fd;
|
|
char buffer[1];
|
|
|
|
fd = open(socketLockFile, O_RDONLY | PG_BINARY, 0);
|
|
if (fd >= 0)
|
|
{
|
|
read(fd, buffer, sizeof(buffer));
|
|
close(fd);
|
|
}
|
|
#endif /* HAVE_UTIMES */
|
|
#endif /* HAVE_UTIME */
|
|
}
|
|
}
|
|
|
|
|
|
/*
|
|
* Add lines to the data directory lock file. This erases all following
|
|
* lines, but that is OK because lines are added in order.
|
|
*/
|
|
void
|
|
AddToLockFile(int target_line, const char *str)
|
|
{
|
|
int fd;
|
|
int len;
|
|
int lineno;
|
|
char *ptr;
|
|
char buffer[MAXPGPATH * 3 + 256];
|
|
|
|
fd = open(DIRECTORY_LOCK_FILE, O_RDWR | PG_BINARY, 0);
|
|
if (fd < 0)
|
|
{
|
|
ereport(LOG,
|
|
(errcode_for_file_access(),
|
|
errmsg("could not open file \"%s\": %m",
|
|
DIRECTORY_LOCK_FILE)));
|
|
return;
|
|
}
|
|
len = read(fd, buffer, sizeof(buffer) - 100);
|
|
if (len < 0)
|
|
{
|
|
ereport(LOG,
|
|
(errcode_for_file_access(),
|
|
errmsg("could not read from file \"%s\": %m",
|
|
DIRECTORY_LOCK_FILE)));
|
|
close(fd);
|
|
return;
|
|
}
|
|
buffer[len] = '\0';
|
|
|
|
/*
|
|
* Skip over first four lines (PID, pgdata, portnum, socketdir).
|
|
*/
|
|
ptr = buffer;
|
|
for (lineno = 1; lineno < target_line; lineno++)
|
|
{
|
|
if ((ptr = strchr(ptr, '\n')) == NULL)
|
|
{
|
|
elog(LOG, "bogus data in \"%s\"", DIRECTORY_LOCK_FILE);
|
|
close(fd);
|
|
return;
|
|
}
|
|
ptr++;
|
|
}
|
|
|
|
strlcat(buffer, str, sizeof(buffer));
|
|
|
|
/*
|
|
* And rewrite the data. Since we write in a single kernel call, this
|
|
* update should appear atomic to onlookers.
|
|
*/
|
|
len = strlen(buffer);
|
|
errno = 0;
|
|
if (lseek(fd, (off_t) 0, SEEK_SET) != 0 ||
|
|
(int) write(fd, buffer, len) != len)
|
|
{
|
|
/* if write didn't set errno, assume problem is no disk space */
|
|
if (errno == 0)
|
|
errno = ENOSPC;
|
|
ereport(LOG,
|
|
(errcode_for_file_access(),
|
|
errmsg("could not write to file \"%s\": %m",
|
|
DIRECTORY_LOCK_FILE)));
|
|
close(fd);
|
|
return;
|
|
}
|
|
if (pg_fsync(fd) != 0)
|
|
{
|
|
ereport(LOG,
|
|
(errcode_for_file_access(),
|
|
errmsg("could not write to file \"%s\": %m",
|
|
DIRECTORY_LOCK_FILE)));
|
|
}
|
|
if (close(fd) != 0)
|
|
{
|
|
ereport(LOG,
|
|
(errcode_for_file_access(),
|
|
errmsg("could not write to file \"%s\": %m",
|
|
DIRECTORY_LOCK_FILE)));
|
|
}
|
|
}
|
|
|
|
|
|
/*-------------------------------------------------------------------------
|
|
* Version checking support
|
|
*-------------------------------------------------------------------------
|
|
*/
|
|
|
|
/*
|
|
* Determine whether the PG_VERSION file in directory `path' indicates
|
|
* a data version compatible with the version of this program.
|
|
*
|
|
* If compatible, return. Otherwise, ereport(FATAL).
|
|
*/
|
|
void
|
|
ValidatePgVersion(const char *path)
|
|
{
|
|
char full_path[MAXPGPATH];
|
|
FILE *file;
|
|
int ret;
|
|
long file_major,
|
|
file_minor;
|
|
long my_major = 0,
|
|
my_minor = 0;
|
|
char *endptr;
|
|
const char *version_string = PG_VERSION;
|
|
|
|
my_major = strtol(version_string, &endptr, 10);
|
|
if (*endptr == '.')
|
|
my_minor = strtol(endptr + 1, NULL, 10);
|
|
|
|
snprintf(full_path, sizeof(full_path), "%s/PG_VERSION", path);
|
|
|
|
file = AllocateFile(full_path, "r");
|
|
if (!file)
|
|
{
|
|
if (errno == ENOENT)
|
|
ereport(FATAL,
|
|
(errcode(ERRCODE_INVALID_PARAMETER_VALUE),
|
|
errmsg("\"%s\" is not a valid data directory",
|
|
path),
|
|
errdetail("File \"%s\" is missing.", full_path)));
|
|
else
|
|
ereport(FATAL,
|
|
(errcode_for_file_access(),
|
|
errmsg("could not open file \"%s\": %m", full_path)));
|
|
}
|
|
|
|
ret = fscanf(file, "%ld.%ld", &file_major, &file_minor);
|
|
if (ret != 2)
|
|
ereport(FATAL,
|
|
(errcode(ERRCODE_INVALID_PARAMETER_VALUE),
|
|
errmsg("\"%s\" is not a valid data directory",
|
|
path),
|
|
errdetail("File \"%s\" does not contain valid data.",
|
|
full_path),
|
|
errhint("You might need to initdb.")));
|
|
|
|
FreeFile(file);
|
|
|
|
if (my_major != file_major || my_minor != file_minor)
|
|
ereport(FATAL,
|
|
(errcode(ERRCODE_INVALID_PARAMETER_VALUE),
|
|
errmsg("database files are incompatible with server"),
|
|
errdetail("The data directory was initialized by PostgreSQL version %ld.%ld, "
|
|
"which is not compatible with this version %s.",
|
|
file_major, file_minor, version_string)));
|
|
}
|
|
|
|
/*-------------------------------------------------------------------------
|
|
* Library preload support
|
|
*-------------------------------------------------------------------------
|
|
*/
|
|
|
|
/*
|
|
* GUC variables: lists of library names to be preloaded at postmaster
|
|
* start and at backend start
|
|
*/
|
|
char *shared_preload_libraries_string = NULL;
|
|
char *local_preload_libraries_string = NULL;
|
|
|
|
/* Flag telling that we are loading shared_preload_libraries */
|
|
bool process_shared_preload_libraries_in_progress = false;
|
|
|
|
/*
|
|
* load the shared libraries listed in 'libraries'
|
|
*
|
|
* 'gucname': name of GUC variable, for error reports
|
|
* 'restricted': if true, force libraries to be in $libdir/plugins/
|
|
*/
|
|
static void
|
|
load_libraries(const char *libraries, const char *gucname, bool restricted)
|
|
{
|
|
char *rawstring;
|
|
List *elemlist;
|
|
int elevel;
|
|
ListCell *l;
|
|
|
|
if (libraries == NULL || libraries[0] == '\0')
|
|
return; /* nothing to do */
|
|
|
|
/* Need a modifiable copy of string */
|
|
rawstring = pstrdup(libraries);
|
|
|
|
/* Parse string into list of identifiers */
|
|
if (!SplitIdentifierString(rawstring, ',', &elemlist))
|
|
{
|
|
/* syntax error in list */
|
|
pfree(rawstring);
|
|
list_free(elemlist);
|
|
ereport(LOG,
|
|
(errcode(ERRCODE_SYNTAX_ERROR),
|
|
errmsg("invalid list syntax in parameter \"%s\"",
|
|
gucname)));
|
|
return;
|
|
}
|
|
|
|
/*
|
|
* Choose notice level: avoid repeat messages when re-loading a library
|
|
* that was preloaded into the postmaster. (Only possible in EXEC_BACKEND
|
|
* configurations)
|
|
*/
|
|
#ifdef EXEC_BACKEND
|
|
if (IsUnderPostmaster && process_shared_preload_libraries_in_progress)
|
|
elevel = DEBUG2;
|
|
else
|
|
#endif
|
|
elevel = LOG;
|
|
|
|
foreach(l, elemlist)
|
|
{
|
|
char *tok = (char *) lfirst(l);
|
|
char *filename;
|
|
|
|
filename = pstrdup(tok);
|
|
canonicalize_path(filename);
|
|
/* If restricting, insert $libdir/plugins if not mentioned already */
|
|
if (restricted && first_dir_separator(filename) == NULL)
|
|
{
|
|
char *expanded;
|
|
|
|
expanded = palloc(strlen("$libdir/plugins/") + strlen(filename) + 1);
|
|
strcpy(expanded, "$libdir/plugins/");
|
|
strcat(expanded, filename);
|
|
pfree(filename);
|
|
filename = expanded;
|
|
}
|
|
load_file(filename, restricted);
|
|
ereport(elevel,
|
|
(errmsg("loaded library \"%s\"", filename)));
|
|
pfree(filename);
|
|
}
|
|
|
|
pfree(rawstring);
|
|
list_free(elemlist);
|
|
}
|
|
|
|
/*
|
|
* process any libraries that should be preloaded at postmaster start
|
|
*/
|
|
void
|
|
process_shared_preload_libraries(void)
|
|
{
|
|
process_shared_preload_libraries_in_progress = true;
|
|
load_libraries(shared_preload_libraries_string,
|
|
"shared_preload_libraries",
|
|
false);
|
|
process_shared_preload_libraries_in_progress = false;
|
|
}
|
|
|
|
/*
|
|
* process any libraries that should be preloaded at backend start
|
|
*/
|
|
void
|
|
process_local_preload_libraries(void)
|
|
{
|
|
load_libraries(local_preload_libraries_string,
|
|
"local_preload_libraries",
|
|
true);
|
|
}
|
|
|
|
void
|
|
pg_bindtextdomain(const char *domain)
|
|
{
|
|
#ifdef ENABLE_NLS
|
|
if (my_exec_path[0] != '\0')
|
|
{
|
|
char locale_path[MAXPGPATH];
|
|
|
|
get_locale_path(my_exec_path, locale_path);
|
|
bindtextdomain(domain, locale_path);
|
|
pg_bind_textdomain_codeset(domain);
|
|
}
|
|
#endif
|
|
}
|