static void
setup_event_handlers(void)
{
pqsignal(SIGHUP, handle_sighup);
pqsignal(SIGINT, handle_sigint);
pqsignal(SIGTERM, handle_sigint);
}
static void
WorkerCommon(void)
{
/* Establish signal handlers before unblocking signals. */
pqsignal(SIGHUP, sighupHandler);
pqsignal(SIGTERM, sigtermHandler);
/* We're now ready to receive signals */
BackgroundWorkerUnblockSignals();
}
void worker_main(Datum arg)
{
int ret;
StringInfoData buf;
uint32 segment = UInt32GetDatum(arg);
/* Setup signal handlers */
pqsignal(SIGHUP, worker_sighup);
pqsignal(SIGTERM, worker_sigterm);
/* Allow signals */
BackgroundWorkerUnblockSignals();
initialize_worker(segment);
/* Connect to the database */
BackgroundWorkerInitializeConnection(job->datname, job->rolname);
elog(LOG, "%s initialized running job id %d", MyBgworkerEntry->bgw_name, job->job_id);
pgstat_report_appname(MyBgworkerEntry->bgw_name);
/* Initialize the query text */
initStringInfo(&buf);
appendStringInfo(&buf,
"SELECT * FROM %s.%s(%d, NULL)",
job_run_function.schema,
job_run_function.name,
job->job_id);
/* Initialize the SPI subsystem */
SetCurrentStatementStartTimestamp();
StartTransactionCommand();
SPI_connect();
PushActiveSnapshot(GetTransactionSnapshot());
pgstat_report_activity(STATE_RUNNING, buf.data);
SetCurrentStatementStartTimestamp();
/* And run the query */
ret = SPI_execute(buf.data, true, 0);
if (ret < 0)
elog(FATAL, "errors while executing %s", buf.data);
/* Commmit the transaction */
SPI_finish();
PopActiveSnapshot();
CommitTransactionCommand();
pgstat_report_activity(STATE_IDLE, NULL);
proc_exit(0);
}
/*
* proc_exit callback to tear down the JVM
*/
static void _destroyJavaVM(int status, Datum dummy)
{
if(s_javaVM != 0)
{
Invocation ctx;
#ifdef USE_PLJAVA_SIGHANDLERS
#if PG_VERSION_NUM >= 90300
TimeoutId tid;
#else
pqsigfunc saveSigAlrm;
#endif
Invocation_pushInvocation(&ctx, false);
if(sigsetjmp(recoverBuf, 1) != 0)
{
elog(DEBUG2,
"needed to forcibly shut down the Java virtual machine");
s_javaVM = 0;
return;
}
#if PG_VERSION_NUM >= 90300
InitializeTimeouts(); /* establishes SIGALRM handler */
tid = RegisterTimeout(USER_TIMEOUT, terminationTimeoutHandler);
#else
saveSigAlrm = pqsignal(SIGALRM, terminationTimeoutHandler);
enable_sig_alarm(5000, false);
#endif
elog(DEBUG2, "shutting down the Java virtual machine");
JNI_destroyVM(s_javaVM);
#if PG_VERSION_NUM >= 90300
disable_timeout(tid, false);
#else
disable_sig_alarm(false);
pqsignal(SIGALRM, saveSigAlrm);
#endif
#else
Invocation_pushInvocation(&ctx, false);
elog(DEBUG2, "shutting down the Java virtual machine");
JNI_destroyVM(s_javaVM);
#endif
elog(DEBUG2, "done shutting down the Java virtual machine");
s_javaVM = 0;
currentInvocation = 0;
}
}
/*
* hello_main
*
* Main loop processing.
*/
void
hello_main(Datum main_arg)
{
/* Set up the sigterm signal before unblocking them */
pqsignal(SIGTERM, hello_sigterm);
/* We're now ready to receive signals */
BackgroundWorkerUnblockSignals();
while (!got_sigterm)
{
int rc;
/* Wait 10s */
rc = WaitLatch(&MyProc->procLatch,
WL_LATCH_SET | WL_TIMEOUT | WL_POSTMASTER_DEATH,
10000L,
PG_WAIT_EXTENSION);
ResetLatch(&MyProc->procLatch);
/* Emergency bailout if postmaster has died */
if (rc & WL_POSTMASTER_DEATH)
proc_exit(1);
elog(LOG, "Hello World!"); /* Say Hello to the world */
}
proc_exit(0);
}
/*
* Foreign-data wrapper handler function: return a struct with pointers
* to my callback routines.
*/
Datum
jdbc_fdw_handler(PG_FUNCTION_ARGS)
{
FdwRoutine *fdwroutine = makeNode(FdwRoutine);
#if (PG_VERSION_NUM < 90200)
fdwroutine->PlanForeignScan = jdbcPlanForeignScan;
#endif
#if (PG_VERSION_NUM >= 90200)
fdwroutine->GetForeignRelSize = jdbcGetForeignRelSize;
fdwroutine->GetForeignPaths = jdbcGetForeignPaths;
fdwroutine->GetForeignPlan = jdbcGetForeignPlan;
#endif
fdwroutine->ExplainForeignScan = jdbcExplainForeignScan;
fdwroutine->BeginForeignScan = jdbcBeginForeignScan;
fdwroutine->IterateForeignScan = jdbcIterateForeignScan;
fdwroutine->ReScanForeignScan = jdbcReScanForeignScan;
fdwroutine->EndForeignScan = jdbcEndForeignScan;
pqsignal(SIGINT, SIGINTInterruptHandler);
PG_RETURN_POINTER(fdwroutine);
}
/*
* signal handler in case we are interrupted.
*
* The Windows runtime docs at
* http://msdn.microsoft.com/library/en-us/vclib/html/_crt_signal.asp
* specifically forbid a number of things being done from a signal handler,
* including IO, memory allocation and system calls, and only allow jmpbuf
* if you are handling SIGFPE.
*
* I avoided doing the forbidden things by setting a flag instead of calling
* exit_nicely() directly.
*
* Also note the behaviour of Windows with SIGINT, which says this:
* Note SIGINT is not supported for any Win32 application, including
* Windows 98/Me and Windows NT/2000/XP. When a CTRL+C interrupt occurs,
* Win32 operating systems generate a new thread to specifically handle
* that interrupt. This can cause a single-thread application such as UNIX,
* to become multithreaded, resulting in unexpected behavior.
*
* I have no idea how to handle this. (Strange they call UNIX an application!)
* So this will need some testing on Windows.
*/
static void
trapsig(int signum)
{
/* handle systems that reset the handler, like Windows (grr) */
pqsignal(signum, trapsig);
caught_signal = true;
}
void
_PG_init(void)
{
coreIntHandler = pqsignal(SIGINT, handleInterrupt);
#ifdef WIN32
#if POSTGIS_GEOS_VERSION >= 34
GEOS_interruptRegisterCallback(interruptCallback);
#endif
lwgeom_register_interrupt_callback(interruptCallback);
#endif
#if 0
/* Define custom GUC variables. */
DefineCustomIntVariable(
"postgis.debug.level", /* name */
"Sets the debugging level of PostGIS.", /* short_desc */
"This is an experimental configuration.", /* long_desc */
&postgis_debug_level, /* valueAddr */
0, 8, /* min-max */
0, /* bootValue */
PGC_SUSET, /* GucContext context */
GUC_UNIT_MS, /* int flags */
#if POSTGIS_PGSQL_VERSION >= 91
NULL, /* GucStringCheckHook check_hook */
#endif
NULL, /* GucStringAssignHook assign_hook */
NULL /* GucShowHook show_hook */
);
#endif
#if 0
/* Define custom GUC variables. */
DefineCustomStringVariable(
"postgis.greeting.string", /* name */
"Sets the greeting string used on postgis module load.", /* short_desc */
"This is an experimental configuration.", /* long_desc */
&greeting, /* valueAddr */
"Welcome to PostGIS " POSTGIS_VERSION, /* bootValue */
PGC_SUSET, /* GucContext context */
GUC_UNIT_MS, /* int flags */
#if POSTGIS_PGSQL_VERSION >= 91
NULL, /* GucStringCheckHook check_hook */
#endif
NULL, /* GucStringAssignHook assign_hook */
NULL /* GucShowHook show_hook */
);
#endif
/* install PostgreSQL handlers */
pg_install_lwgeom_handlers();
/* initialize geometry backend */
lwgeom_init_backend();
}
static void
config_log_main(Datum main_arg)
{
config_log_objects *objects;
pqsignal(SIGTERM, config_log_sigterm);
pqsignal(SIGHUP, config_log_sighup);
/* We're now ready to receive signals */
BackgroundWorkerUnblockSignals();
/* Connect to database */
BackgroundWorkerInitializeConnection(config_log_database, NULL);
/* Verify expected objects exist */
objects = initialize_objects();
while (!got_sigterm)
{
int rc;
rc = WaitLatch(&MyProc->procLatch,
WL_LATCH_SET | WL_TIMEOUT | WL_POSTMASTER_DEATH,
100000L);
ResetLatch(&MyProc->procLatch);
/* emergency bailout if postmaster has died */
if (rc & WL_POSTMASTER_DEATH)
proc_exit(1);
/*
* In case of a SIGHUP, just reload the configuration.
*/
if (got_sighup)
{
got_sighup = false;
ProcessConfigFile(PGC_SIGHUP);
execute_pg_settings_logger(objects);
}
}
proc_exit(0);
}
void
hello_main(Datum main_arg)
{
/* Register functions for SIGTERM/SIGHUP management */
pqsignal(SIGHUP, hello_sighup);
pqsignal(SIGTERM, hello_sigterm);
/* We're now ready to receive signals */
BackgroundWorkerUnblockSignals();
while (true)
{
/* Wait 1s */
WaitLatch(MyLatch,
WL_LATCH_SET | WL_TIMEOUT | WL_EXIT_ON_PM_DEATH,
1000L,
PG_WAIT_EXTENSION);
ResetLatch(MyLatch);
/* Process signals */
if (got_sighup)
{
/* Process config file */
ProcessConfigFile(PGC_SIGHUP);
got_sighup = false;
ereport(LOG, (errmsg("hello signal: processed SIGHUP")));
}
if (got_sigterm)
{
/* Simply exit */
ereport(LOG, (errmsg("hello signal: processed SIGTERM")));
proc_exit(0);
}
}
/* No problems, so clean exit */
proc_exit(0);
}
int
main(int argc, char *argv[])
{
set_pglocale_pgservice(argv[0], PG_TEXTDOMAIN("pg_test_fsync"));
progname = get_progname(argv[0]);
handle_args(argc, argv);
/* Prevent leaving behind the test file */
pqsignal(SIGINT, signal_cleanup);
pqsignal(SIGTERM, signal_cleanup);
#ifndef WIN32
pqsignal(SIGALRM, process_alarm);
#endif
#ifdef SIGHUP
/* Not defined on win32 */
pqsignal(SIGHUP, signal_cleanup);
#endif
prepare_buf();
test_open();
/* Test using 1 XLOG_BLCKSZ write */
test_sync(1);
/* Test using 2 XLOG_BLCKSZ writes */
test_sync(2);
test_open_syncs();
test_file_descriptor_sync();
test_non_sync();
unlink(filename);
return 0;
}
/* ----------------------------------------------------------------
* proc_exit
*
* this function calls all the callbacks registered
* for it (to free resources) and then calls exit.
*
* This should be the only function to call exit().
* -cim 2/6/90
*
* Unfortunately, we can't really guarantee that add-on code
* obeys the rule of not calling exit() directly. So, while
* this is the preferred way out of the system, we also register
* an atexit callback that will make sure cleanup happens.
* ----------------------------------------------------------------
*/
void
proc_exit(int code)
{
pqsignal(SIGALRM, SIG_IGN);
/* Clean up everything that must be cleaned up */
proc_exit_prepare(code);
#ifdef PROFILE_PID_DIR
{
/*
* If we are profiling ourself then gprof's mcleanup() is about to
* write out a profile to ./gmon.out. Since mcleanup() always uses a
* fixed file name, each backend will overwrite earlier profiles. To
* fix that, we create a separate subdirectory for each backend
* (./gprof/pid) and 'cd' to that subdirectory before we exit() - that
* forces mcleanup() to write each profile into its own directory. We
* end up with something like: $PGDATA/gprof/8829/gmon.out
* $PGDATA/gprof/8845/gmon.out ...
*
* To avoid undesirable disk space bloat, autovacuum workers are
* discriminated against: all their gmon.out files go into the same
* subdirectory. Without this, an installation that is "just sitting
* there" nonetheless eats megabytes of disk space every few seconds.
*
* Note that we do this here instead of in an on_proc_exit() callback
* because we want to ensure that this code executes last - we don't
* want to interfere with any other on_proc_exit() callback. For the
* same reason, we do not include it in proc_exit_prepare ... so if
* you are exiting in the "wrong way" you won't drop your profile in a
* nice place.
*/
char gprofDirName[32];
if (IsAutoVacuumWorkerProcess())
snprintf(gprofDirName, 32, "gprof/avworker");
else
snprintf(gprofDirName, 32, "gprof/%d", (int) getpid());
mkdir("gprof", S_IRWXU | S_IRWXG | S_IRWXO);
mkdir(gprofDirName, S_IRWXU | S_IRWXG | S_IRWXO);
chdir(gprofDirName);
}
#endif
elog(DEBUG3, "exit(%d)", code);
exit(code);
}
/*
* setQFout
* -- handler for -o command line option and \o command
*
* Tries to open file fname (or pipe if fname starts with '|')
* and stores the file handle in pset)
* Upon failure, sets stdout and returns false.
*/
bool
setQFout(const char *fname)
{
bool status = true;
/* Close old file/pipe */
if (pset.queryFout && pset.queryFout != stdout && pset.queryFout != stderr)
{
if (pset.queryFoutPipe)
pclose(pset.queryFout);
else
fclose(pset.queryFout);
}
/* If no filename, set stdout */
if (!fname || fname[0] == '\0')
{
pset.queryFout = stdout;
pset.queryFoutPipe = false;
}
else if (*fname == '|')
{
pset.queryFout = popen(fname + 1, "w");
pset.queryFoutPipe = true;
}
else
{
pset.queryFout = fopen(fname, "w");
pset.queryFoutPipe = false;
}
if (!(pset.queryFout))
{
psql_error("%s: %s\n", fname, strerror(errno));
pset.queryFout = stdout;
pset.queryFoutPipe = false;
status = false;
}
/* Direct signals */
#ifndef WIN32
pqsignal(SIGPIPE, pset.queryFoutPipe ? SIG_IGN : SIG_DFL);
#endif
return status;
}
/*
* Initialize timeout module.
*
* This must be called in every process that wants to use timeouts.
*
* If the process was forked from another one that was also using this
* module, be sure to call this before re-enabling signals; else handlers
* meant to run in the parent process might get invoked in this one.
*/
void
InitializeTimeouts(void)
{
int i;
/* Initialize, or re-initialize, all local state */
disable_alarm();
num_active_timeouts = 0;
for (i = 0; i < MAX_TIMEOUTS; i++)
{
all_timeouts[i].index = i;
all_timeouts[i].indicator = false;
all_timeouts[i].timeout_handler = NULL;
all_timeouts[i].start_time = 0;
all_timeouts[i].fin_time = 0;
}
all_timeouts_initialized = true;
/* Now establish the signal handler */
pqsignal(SIGALRM, handle_sig_alarm);
}
/*
* slashUsage
*
* print out help for the backslash commands
*/
void
slashUsage(unsigned short int pager)
{
FILE *output;
char *currdb;
currdb = PQdb(pset.db);
if (currdb == NULL)
currdb = "";
output = PageOutput(90, pager);
/* if you add/remove a line here, change the row count above */
fprintf(output, _("General\n"));
fprintf(output, _(" \\copyright show PostgreSQL usage and distribution terms\n"));
fprintf(output, _(" \\g [FILE] or ; execute query (and send results to file or |pipe)\n"));
fprintf(output, _(" \\h [NAME] help on syntax of SQL commands, * for all commands\n"));
fprintf(output, _(" \\q quit psql\n"));
fprintf(output, "\n");
fprintf(output, _("Query Buffer\n"));
fprintf(output, _(" \\e [FILE] edit the query buffer (or file) with external editor\n"));
fprintf(output, _(" \\ef [FUNCNAME] edit function definition with external editor\n"));
fprintf(output, _(" \\p show the contents of the query buffer\n"));
fprintf(output, _(" \\r reset (clear) the query buffer\n"));
#ifdef USE_READLINE
fprintf(output, _(" \\s [FILE] display history or save it to file\n"));
#endif
fprintf(output, _(" \\w FILE write query buffer to file\n"));
fprintf(output, "\n");
fprintf(output, _("Input/Output\n"));
fprintf(output, _(" \\copy ... perform SQL COPY with data stream to the client host\n"));
fprintf(output, _(" \\echo [STRING] write string to standard output\n"));
fprintf(output, _(" \\i FILE execute commands from file\n"));
fprintf(output, _(" \\o [FILE] send all query results to file or |pipe\n"));
fprintf(output, _(" \\qecho [STRING] write string to query output stream (see \\o)\n"));
fprintf(output, "\n");
fprintf(output, _("Informational\n"));
fprintf(output, _(" (options: S = show system objects, + = additional detail)\n"));
fprintf(output, _(" \\d[S+] list tables, views, and sequences\n"));
fprintf(output, _(" \\d[S+] NAME describe table, view, sequence, or index\n"));
fprintf(output, _(" \\da[S] [PATTERN] list aggregates\n"));
fprintf(output, _(" \\db[+] [PATTERN] list tablespaces\n"));
fprintf(output, _(" \\dc[S] [PATTERN] list conversions\n"));
fprintf(output, _(" \\dC [PATTERN] list casts\n"));
fprintf(output, _(" \\dd[S] [PATTERN] show comments on objects\n"));
fprintf(output, _(" \\ddp [PATTERN] list default privileges\n"));
fprintf(output, _(" \\dD[S] [PATTERN] list domains\n"));
fprintf(output, _(" \\des[+] [PATTERN] list foreign servers\n"));
fprintf(output, _(" \\deu[+] [PATTERN] list user mappings\n"));
fprintf(output, _(" \\dew[+] [PATTERN] list foreign-data wrappers\n"));
fprintf(output, _(" \\df[antw][S+] [PATRN] list [only agg/normal/trigger/window] functions\n"));
fprintf(output, _(" \\dF[+] [PATTERN] list text search configurations\n"));
fprintf(output, _(" \\dFd[+] [PATTERN] list text search dictionaries\n"));
fprintf(output, _(" \\dFp[+] [PATTERN] list text search parsers\n"));
fprintf(output, _(" \\dFt[+] [PATTERN] list text search templates\n"));
fprintf(output, _(" \\dg[+] [PATTERN] list roles (groups)\n"));
fprintf(output, _(" \\dx[+] [PATTERN] list extensions\n"));
fprintf(output, _(" \\di[S+] [PATTERN] list indexes\n"));
fprintf(output, _(" \\dl list large objects, same as \\lo_list\n"));
fprintf(output, _(" \\dn[+] [PATTERN] list schemas\n"));
fprintf(output, _(" \\do[S] [PATTERN] list operators\n"));
fprintf(output, _(" \\dp [PATTERN] list table, view, and sequence access privileges\n"));
fprintf(output, _(" \\dr[S+] [PATTERN] list foreign tables\n")); /* GPDB Only */
fprintf(output, _(" \\drds [PATRN1 [PATRN2]] list per-database role settings\n"));
fprintf(output, _(" \\ds[S+] [PATTERN] list sequences\n"));
fprintf(output, _(" \\dt[S+] [PATTERN] list tables\n"));
fprintf(output, _(" \\dT[S+] [PATTERN] list data types\n"));
fprintf(output, _(" \\du[+] [PATTERN] list roles (users)\n"));
fprintf(output, _(" \\dv[S+] [PATTERN] list views\n"));
fprintf(output, _(" \\dE [PATTERN] list external tables\n"));
fprintf(output, _(" \\l[+] list all databases\n"));
fprintf(output, _(" \\z [PATTERN] same as \\dp\n"));
fprintf(output, "\n");
fprintf(output, _("Formatting\n"));
fprintf(output, _(" \\a toggle between unaligned and aligned output mode\n"));
fprintf(output, _(" \\C [STRING] set table title, or unset if none\n"));
fprintf(output, _(" \\f [STRING] show or set field separator for unaligned query output\n"));
fprintf(output, _(" \\H toggle HTML output mode (currently %s)\n"),
ON(pset.popt.topt.format == PRINT_HTML));
fprintf(output, _(" \\pset NAME [VALUE] set table output option\n"
" (NAME := {format|border|expanded|fieldsep|footer|null|\n"
" numericlocale|recordsep|tuples_only|title|tableattr|pager})\n"));
fprintf(output, _(" \\t [on|off] show only rows (currently %s)\n"),
ON(pset.popt.topt.tuples_only));
fprintf(output, _(" \\T [STRING] set HTML <table> tag attributes, or unset if none\n"));
fprintf(output, _(" \\x [on|off] toggle expanded output (currently %s)\n"),
ON(pset.popt.topt.expanded));
fprintf(output, "\n");
fprintf(output, _("Connection\n"));
fprintf(output, _(" \\c[onnect] [DBNAME|- USER|- HOST|- PORT|-]\n"
" connect to new database (currently \"%s\")\n"),
PQdb(pset.db));
fprintf(output, _(" \\encoding [ENCODING] show or set client encoding\n"));
fprintf(output, _(" \\password [USERNAME] securely change the password for a user\n"));
fprintf(output, _(" \\conninfo display information about current connection\n"));
fprintf(output, "\n");
fprintf(output, _("Operating System\n"));
fprintf(output, _(" \\cd [DIR] change the current working directory\n"));
fprintf(output, _(" \\timing [on|off] toggle timing of commands (currently %s)\n"),
ON(pset.timing));
fprintf(output, _(" \\! [COMMAND] execute command in shell or start interactive shell\n"));
fprintf(output, "\n");
fprintf(output, _("Variables\n"));
fprintf(output, _(" \\prompt [TEXT] NAME prompt user to set internal variable\n"));
fprintf(output, _(" \\set [NAME [VALUE]] set internal variable, or list all if no parameters\n"));
fprintf(output, _(" \\unset NAME unset (delete) internal variable\n"));
fprintf(output, "\n");
fprintf(output, _("Large Objects\n"));
fprintf(output, _(" \\lo_export LOBOID FILE\n"
" \\lo_import FILE [COMMENT]\n"
" \\lo_list\n"
" \\lo_unlink LOBOID large object operations\n"));
if (output != stdout)
{
pclose(output);
#ifndef WIN32
pqsignal(SIGPIPE, SIG_DFL);
#endif
}
}
/* Set up signal handlers */
void
WalSndSignals(void)
{
/* Set up signal handlers */
pqsignal(SIGHUP, WalSndSigHupHandler); /* set flag to read config
* file */
pqsignal(SIGINT, SIG_IGN); /* not used */
pqsignal(SIGTERM, WalSndShutdownHandler); /* request shutdown */
pqsignal(SIGQUIT, WalSndQuickDieHandler); /* hard crash time */
pqsignal(SIGALRM, SIG_IGN);
pqsignal(SIGPIPE, SIG_IGN);
pqsignal(SIGUSR1, WalSndXLogSendHandler); /* request WAL sending */
pqsignal(SIGUSR2, WalSndLastCycleHandler); /* request a last cycle and
* shutdown */
/* 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);
}
static void
FileRepSubProcess_ConfigureSignals(void)
{
/* Accept Signals */
/* emergency shutdown */
pqsignal(SIGQUIT, FileRepSubProcess_ImmediateShutdownHandler);
/* graceful shutdown */
pqsignal(SIGUSR2, FileRepSubProcess_ShutdownHandler);
/* reload configuration file */
pqsignal(SIGHUP, FileRepSubProcess_SigHupHandler);
/* data or segment state changed */
pqsignal(SIGUSR1, FileRepSubProcess_FileRepStateHandler);
/* Ignore Signals */
pqsignal(SIGTERM, SIG_IGN);
pqsignal(SIGALRM, SIG_IGN);
pqsignal(SIGPIPE, SIG_IGN);
/* Use default action */
pqsignal(SIGCHLD, SIG_DFL);
pqsignal(SIGINT, SIG_DFL);
pqsignal(SIGTTIN, SIG_DFL);
pqsignal(SIGTTOU, SIG_DFL);
pqsignal(SIGCONT, SIG_DFL);
pqsignal(SIGWINCH, SIG_DFL);
#ifdef SIGSEGV
pqsignal(SIGSEGV, FileRepSubProcess_HandleCrash);
#endif
#ifdef SIGILL
pqsignal(SIGILL, FileRepSubProcess_HandleCrash);
#endif
#ifdef SIGBUS
pqsignal(SIGBUS, FileRepSubProcess_HandleCrash);
#endif
}
/*
* PQprint()
*
* Format results of a query for printing.
*
* PQprintOpt is a typedef (structure) that contains
* various flags and options. consult libpq-fe.h for
* details
*
* This function should probably be removed sometime since psql
* doesn't use it anymore. It is unclear to what extent this is used
* by external clients, however.
*/
void
PQprint(FILE *fout, const PGresult *res, const PQprintOpt *po)
{
int nFields;
nFields = PQnfields(res);
if (nFields > 0)
{ /* only print rows with at least 1 field. */
int i,
j;
int nTups;
int *fieldMax = NULL; /* in case we don't use them */
unsigned char *fieldNotNum = NULL;
char *border = NULL;
char **fields = NULL;
const char **fieldNames;
int fieldMaxLen = 0;
int numFieldName;
int fs_len = strlen(po->fieldSep);
int total_line_length = 0;
int usePipe = 0;
char *pagerenv;
#if defined(ENABLE_THREAD_SAFETY) && !defined(WIN32)
sigset_t osigset;
bool sigpipe_masked = false;
bool sigpipe_pending;
#endif
#if !defined(ENABLE_THREAD_SAFETY) && !defined(WIN32)
pqsigfunc oldsigpipehandler = NULL;
#endif
#ifdef TIOCGWINSZ
struct winsize screen_size;
#else
struct winsize
{
int ws_row;
int ws_col;
} screen_size;
#endif
nTups = PQntuples(res);
if (!(fieldNames = (const char **) calloc(nFields, sizeof(char *))))
{
fprintf(stderr, libpq_gettext("out of memory\n"));
abort();
}
if (!(fieldNotNum = (unsigned char *) calloc(nFields, 1)))
{
fprintf(stderr, libpq_gettext("out of memory\n"));
abort();
}
if (!(fieldMax = (int *) calloc(nFields, sizeof(int))))
{
fprintf(stderr, libpq_gettext("out of memory\n"));
abort();
}
for (numFieldName = 0;
po->fieldName && po->fieldName[numFieldName];
numFieldName++)
;
for (j = 0; j < nFields; j++)
{
int len;
const char *s = (j < numFieldName && po->fieldName[j][0]) ?
po->fieldName[j] : PQfname(res, j);
fieldNames[j] = s;
len = s ? strlen(s) : 0;
fieldMax[j] = len;
len += fs_len;
if (len > fieldMaxLen)
fieldMaxLen = len;
total_line_length += len;
}
total_line_length += nFields * strlen(po->fieldSep) + 1;
if (fout == NULL)
fout = stdout;
if (po->pager && fout == stdout && isatty(fileno(stdin)) &&
isatty(fileno(stdout)))
{
/*
* If we think there'll be more than one screen of output, try to
* pipe to the pager program.
*/
#ifdef TIOCGWINSZ
if (ioctl(fileno(stdout), TIOCGWINSZ, &screen_size) == -1 ||
screen_size.ws_col == 0 ||
screen_size.ws_row == 0)
{
screen_size.ws_row = 24;
screen_size.ws_col = 80;
}
#else
screen_size.ws_row = 24;
screen_size.ws_col = 80;
#endif
pagerenv = getenv("PAGER");
if (pagerenv != NULL &&
pagerenv[0] != '\0' &&
!po->html3 &&
((po->expanded &&
nTups * (nFields + 1) >= screen_size.ws_row) ||
(!po->expanded &&
nTups * (total_line_length / screen_size.ws_col + 1) *
(1 + (po->standard != 0)) >= screen_size.ws_row -
(po->header != 0) *
(total_line_length / screen_size.ws_col + 1) * 2
- (po->header != 0) * 2 /* row count and newline */
)))
{
fout = popen(pagerenv, "w");
if (fout)
{
usePipe = 1;
#ifndef WIN32
#ifdef ENABLE_THREAD_SAFETY
if (pq_block_sigpipe(&osigset, &sigpipe_pending) == 0)
sigpipe_masked = true;
#else
oldsigpipehandler = pqsignal(SIGPIPE, SIG_IGN);
#endif /* ENABLE_THREAD_SAFETY */
#endif /* WIN32 */
}
else
fout = stdout;
}
}
if (!po->expanded && (po->align || po->html3))
{
if (!(fields = (char **) calloc(nFields * (nTups + 1), sizeof(char *))))
{
fprintf(stderr, libpq_gettext("out of memory\n"));
abort();
}
}
else if (po->header && !po->html3)
{
if (po->expanded)
{
if (po->align)
fprintf(fout, libpq_gettext("%-*s%s Value\n"),
fieldMaxLen - fs_len, libpq_gettext("Field"), po->fieldSep);
else
fprintf(fout, libpq_gettext("%s%sValue\n"), libpq_gettext("Field"), po->fieldSep);
}
else
{
int len = 0;
for (j = 0; j < nFields; j++)
{
const char *s = fieldNames[j];
fputs(s, fout);
len += strlen(s) + fs_len;
if ((j + 1) < nFields)
fputs(po->fieldSep, fout);
}
fputc('\n', fout);
for (len -= fs_len; len--; fputc('-', fout));
fputc('\n', fout);
}
}
if (po->expanded && po->html3)
{
if (po->caption)
fprintf(fout, "<center><h2>%s</h2></center>\n", po->caption);
else
fprintf(fout,
"<center><h2>"
"Query retrieved %d rows * %d fields"
"</h2></center>\n",
nTups, nFields);
}
for (i = 0; i < nTups; i++)
{
if (po->expanded)
{
if (po->html3)
fprintf(fout,
"<table %s><caption align=\"top\">%d</caption>\n",
po->tableOpt ? po->tableOpt : "", i);
else
fprintf(fout, libpq_gettext("-- RECORD %d --\n"), i);
}
for (j = 0; j < nFields; j++)
do_field(po, res, i, j, fs_len, fields, nFields,
fieldNames, fieldNotNum,
fieldMax, fieldMaxLen, fout);
if (po->html3 && po->expanded)
fputs("</table>\n", fout);
}
if (!po->expanded && (po->align || po->html3))
{
if (po->html3)
{
if (po->header)
{
if (po->caption)
fprintf(fout,
"<table %s><caption align=\"top\">%s</caption>\n",
po->tableOpt ? po->tableOpt : "",
po->caption);
else
fprintf(fout,
"<table %s><caption align=\"top\">"
"Retrieved %d rows * %d fields"
"</caption>\n",
po->tableOpt ? po->tableOpt : "", nTups, nFields);
}
else
fprintf(fout, "<table %s>", po->tableOpt ? po->tableOpt : "");
}
if (po->header)
border = do_header(fout, po, nFields, fieldMax, fieldNames,
fieldNotNum, fs_len, res);
for (i = 0; i < nTups; i++)
output_row(fout, po, nFields, fields,
fieldNotNum, fieldMax, border, i);
free(fields);
if (border)
free(border);
}
if (po->header && !po->html3)
fprintf(fout, "(%d row%s)\n\n", PQntuples(res),
(PQntuples(res) == 1) ? "" : "s");
free(fieldMax);
free(fieldNotNum);
free((void *) fieldNames);
if (usePipe)
{
#ifdef WIN32
_pclose(fout);
#else
pclose(fout);
#ifdef ENABLE_THREAD_SAFETY
/* we can't easily verify if EPIPE occurred, so say it did */
if (sigpipe_masked)
pq_reset_sigpipe(&osigset, sigpipe_pending, true);
#else
pqsignal(SIGPIPE, oldsigpipehandler);
#endif /* ENABLE_THREAD_SAFETY */
#endif /* WIN32 */
}
if (po->html3 && !po->expanded)
fputs("</table>\n", fout);
}
}
void
worker_spi_main(Datum main_arg)
{
int index = DatumGetInt32(main_arg);
worktable *table;
StringInfoData buf;
char name[20];
table = palloc(sizeof(worktable));
sprintf(name, "schema%d", index);
table->schema = pstrdup(name);
table->name = pstrdup("counted");
/* Establish signal handlers before unblocking signals. */
pqsignal(SIGHUP, worker_spi_sighup);
pqsignal(SIGTERM, worker_spi_sigterm);
/* We're now ready to receive signals */
BackgroundWorkerUnblockSignals();
/* Connect to our database */
BackgroundWorkerInitializeConnection("postgres", NULL);
elog(LOG, "%s initialized with %s.%s",
MyBgworkerEntry->bgw_name, table->schema, table->name);
initialize_worker_spi(table);
/*
* Quote identifiers passed to us. Note that this must be done after
* initialize_worker_spi, because that routine assumes the names are not
* quoted.
*
* Note some memory might be leaked here.
*/
table->schema = quote_identifier(table->schema);
table->name = quote_identifier(table->name);
initStringInfo(&buf);
appendStringInfo(&buf,
"WITH deleted AS (DELETE "
"FROM %s.%s "
"WHERE type = 'delta' RETURNING value), "
"total AS (SELECT coalesce(sum(value), 0) as sum "
"FROM deleted) "
"UPDATE %s.%s "
"SET value = %s.value + total.sum "
"FROM total WHERE type = 'total' "
"RETURNING %s.value",
table->schema, table->name,
table->schema, table->name,
table->name,
table->name);
/*
* Main loop: do this until the SIGTERM handler tells us to terminate
*/
while (!got_sigterm)
{
int ret;
int rc;
/*
* Background workers mustn't call usleep() or any direct equivalent:
* instead, they may wait on their process latch, which sleeps as
* necessary, but is awakened if postmaster dies. That way the
* background process goes away immediately in an emergency.
*/
rc = WaitLatch(&MyProc->procLatch,
WL_LATCH_SET | WL_TIMEOUT | WL_POSTMASTER_DEATH,
worker_spi_naptime * 1000L);
ResetLatch(&MyProc->procLatch);
/* emergency bailout if postmaster has died */
if (rc & WL_POSTMASTER_DEATH)
proc_exit(1);
/*
* In case of a SIGHUP, just reload the configuration.
*/
if (got_sighup)
{
got_sighup = false;
ProcessConfigFile(PGC_SIGHUP);
}
/*
* Start a transaction on which we can run queries. Note that each
* StartTransactionCommand() call should be preceded by a
* SetCurrentStatementStartTimestamp() call, which sets both the time
* for the statement we're about the run, and also the transaction
* start time. Also, each other query sent to SPI should probably be
* preceded by SetCurrentStatementStartTimestamp(), so that statement
* start time is always up to date.
*
* The SPI_connect() call lets us run queries through the SPI manager,
* and the PushActiveSnapshot() call creates an "active" snapshot
* which is necessary for queries to have MVCC data to work on.
*
* The pgstat_report_activity() call makes our activity visible
* through the pgstat views.
*/
SetCurrentStatementStartTimestamp();
StartTransactionCommand();
SPI_connect();
PushActiveSnapshot(GetTransactionSnapshot());
pgstat_report_activity(STATE_RUNNING, buf.data);
/* We can now execute queries via SPI */
ret = SPI_execute(buf.data, false, 0);
if (ret != SPI_OK_UPDATE_RETURNING)
elog(FATAL, "cannot select from table %s.%s: error code %d",
table->schema, table->name, ret);
if (SPI_processed > 0)
{
bool isnull;
int32 val;
val = DatumGetInt32(SPI_getbinval(SPI_tuptable->vals[0],
SPI_tuptable->tupdesc,
1, &isnull));
if (!isnull)
elog(LOG, "%s: count in %s.%s is now %d",
MyBgworkerEntry->bgw_name,
table->schema, table->name, val);
}
/*
* And finish our transaction.
*/
SPI_finish();
PopActiveSnapshot();
CommitTransactionCommand();
pgstat_report_activity(STATE_IDLE, NULL);
}
proc_exit(1);
}
/*
* kafka_consume_main
*
* Main function for Kafka consumers running as background workers
*/
void
kafka_consume_main(Datum arg)
{
char err_msg[512];
rd_kafka_topic_conf_t *topic_conf;
rd_kafka_t *kafka;
rd_kafka_topic_t *topic;
rd_kafka_message_t **messages;
const struct rd_kafka_metadata *meta;
struct rd_kafka_metadata_topic topic_meta;
rd_kafka_resp_err_t err;
bool found;
Oid id = (Oid) arg;
ListCell *lc;
KafkaConsumerProc *proc = hash_search(consumer_procs, &id, HASH_FIND, &found);
KafkaConsumer consumer;
CopyStmt *copy;
int valid_brokers = 0;
int i;
int my_partitions = 0;
if (!found)
elog(ERROR, "kafka consumer %d not found", id);
pqsignal(SIGTERM, kafka_consume_main_sigterm);
#define BACKTRACE_SEGFAULTS
#ifdef BACKTRACE_SEGFAULTS
pqsignal(SIGSEGV, debug_segfault);
#endif
/* we're now ready to receive signals */
BackgroundWorkerUnblockSignals();
/* give this proc access to the database */
BackgroundWorkerInitializeConnection(NameStr(proc->dbname), NULL);
/* load saved consumer state */
StartTransactionCommand();
load_consumer_state(proc->consumer_id, &consumer);
copy = get_copy_statement(&consumer);
topic_conf = rd_kafka_topic_conf_new();
kafka = rd_kafka_new(RD_KAFKA_CONSUMER, NULL, err_msg, sizeof(err_msg));
rd_kafka_set_logger(kafka, logger);
/*
* Add all brokers currently in pipeline_kafka_brokers
*/
if (consumer.brokers == NIL)
elog(ERROR, "no valid brokers were found");
foreach(lc, consumer.brokers)
valid_brokers += rd_kafka_brokers_add(kafka, lfirst(lc));
if (!valid_brokers)
elog(ERROR, "no valid brokers were found");
/*
* Set up our topic to read from
*/
topic = rd_kafka_topic_new(kafka, consumer.topic, topic_conf);
err = rd_kafka_metadata(kafka, false, topic, &meta, CONSUMER_TIMEOUT);
if (err != RD_KAFKA_RESP_ERR_NO_ERROR)
elog(ERROR, "failed to acquire metadata: %s", rd_kafka_err2str(err));
Assert(meta->topic_cnt == 1);
topic_meta = meta->topics[0];
load_consumer_offsets(&consumer, &topic_meta, proc->offset);
CommitTransactionCommand();
/*
* Begin consuming all partitions that this process is responsible for
*/
for (i = 0; i < topic_meta.partition_cnt; i++)
{
int partition = topic_meta.partitions[i].id;
Assert(partition <= consumer.num_partitions);
if (partition % consumer.parallelism != proc->partition_group)
continue;
elog(LOG, "[kafka consumer] %s <- %s consuming partition %d from offset %ld",
consumer.rel->relname, consumer.topic, partition, consumer.offsets[partition]);
if (rd_kafka_consume_start(topic, partition, consumer.offsets[partition]) == -1)
elog(ERROR, "failed to start consuming: %s", rd_kafka_err2str(rd_kafka_errno2err(errno)));
my_partitions++;
}
/*
* No point doing anything if we don't have any partitions assigned to us
*/
if (my_partitions == 0)
{
elog(LOG, "[kafka consumer] %s <- %s consumer %d doesn't have any partitions to read from",
consumer.rel->relname, consumer.topic, MyProcPid);
goto done;
}
messages = palloc0(sizeof(rd_kafka_message_t) * consumer.batch_size);
/*
* Consume messages until we are terminated
*/
while (!got_sigterm)
{
ssize_t num_consumed;
int i;
int messages_buffered = 0;
int partition;
StringInfoData buf;
bool xact = false;
for (partition = 0; partition < consumer.num_partitions; partition++)
{
if (partition % consumer.parallelism != proc->partition_group)
continue;
num_consumed = rd_kafka_consume_batch(topic, partition,
CONSUMER_TIMEOUT, messages, consumer.batch_size);
if (num_consumed <= 0)
continue;
if (!xact)
{
StartTransactionCommand();
xact = true;
}
initStringInfo(&buf);
for (i = 0; i < num_consumed; i++)
{
if (messages[i]->payload != NULL)
{
appendBinaryStringInfo(&buf, messages[i]->payload, messages[i]->len);
if (buf.len > 0 && buf.data[buf.len - 1] != '\n')
appendStringInfoChar(&buf, '\n');
messages_buffered++;
}
consumer.offsets[partition] = messages[i]->offset;
rd_kafka_message_destroy(messages[i]);
}
}
if (!xact)
{
pg_usleep(1 * 1000);
continue;
}
/* we don't want to die in the event of any errors */
PG_TRY();
{
if (messages_buffered)
execute_copy(copy, &buf);
}
PG_CATCH();
{
elog(LOG, "[kafka consumer] %s <- %s failed to process batch, dropped %d message%s:",
consumer.rel->relname, consumer.topic, (int) num_consumed, (num_consumed == 1 ? "" : "s"));
EmitErrorReport();
FlushErrorState();
AbortCurrentTransaction();
xact = false;
}
PG_END_TRY();
if (!xact)
StartTransactionCommand();
if (messages_buffered)
save_consumer_state(&consumer, proc->partition_group);
CommitTransactionCommand();
}
done:
hash_search(consumer_procs, &id, HASH_REMOVE, NULL);
rd_kafka_topic_destroy(topic);
rd_kafka_destroy(kafka);
rd_kafka_wait_destroyed(CONSUMER_TIMEOUT);
}
/*
* AutoVacMain
*/
NON_EXEC_STATIC void
AutoVacMain(int argc, char *argv[])
{
ListCell *cell;
List *dblist;
autovac_dbase *db;
TransactionId xidForceLimit;
bool for_xid_wrap;
sigjmp_buf local_sigjmp_buf;
/* we are a postmaster subprocess now */
IsUnderPostmaster = true;
am_autovacuum = true;
/* MPP-4990: Autovacuum always runs as utility-mode */
Gp_role = GP_ROLE_UTILITY;
/* reset MyProcPid */
MyProcPid = getpid();
/* record Start Time for logging */
MyStartTime = time(NULL);
/* Identify myself via ps */
init_ps_display("autovacuum process", "", "", "");
SetProcessingMode(InitProcessing);
/*
* If possible, make this process a group leader, so that the postmaster
* can signal any child processes too. (autovacuum probably never has
* any child processes, but for consistency we make all postmaster
* child processes do this.)
*/
#ifdef HAVE_SETSID
if (setsid() < 0)
elog(FATAL, "setsid() failed: %m");
#endif
/*
* Set up signal handlers. We operate on databases much like a regular
* backend, so we use the same signal handling. See equivalent code in
* tcop/postgres.c.
*
* Currently, we don't pay attention to postgresql.conf changes that
* happen during a single daemon iteration, so we can ignore SIGHUP.
*/
pqsignal(SIGHUP, SIG_IGN);
/*
* SIGINT is used to signal cancelling the current table's vacuum; SIGTERM
* means abort and exit cleanly, and SIGQUIT means abandon ship.
*/
pqsignal(SIGINT, StatementCancelHandler);
pqsignal(SIGTERM, die);
pqsignal(SIGQUIT, quickdie);
pqsignal(SIGALRM, handle_sig_alarm);
pqsignal(SIGPIPE, SIG_IGN);
pqsignal(SIGUSR1, procsignal_sigusr1_handler);
/* We don't listen for async notifies */
pqsignal(SIGUSR2, SIG_IGN);
pqsignal(SIGFPE, FloatExceptionHandler);
pqsignal(SIGCHLD, SIG_DFL);
/* Early initialization */
BaseInit();
/*
* Create a per-backend PGPROC struct in shared memory, except in the
* EXEC_BACKEND case where this was done in SubPostmasterMain. We must do
* this before we can use LWLocks (and in the EXEC_BACKEND case we already
* had to do some stuff with LWLocks).
*/
#ifndef EXEC_BACKEND
InitProcess();
#endif
/*
* If an exception is encountered, processing resumes here.
*
* See notes in postgres.c about the design of this coding.
*/
if (sigsetjmp(local_sigjmp_buf, 1) != 0)
{
/* Prevents interrupts while cleaning up */
HOLD_INTERRUPTS();
/* Report the error to the server log */
EmitErrorReport();
/*
* We can now go away. Note that because we called InitProcess, a
* callback was registered to do ProcKill, which will clean up
* necessary state.
*/
proc_exit(0);
}
/* We can now handle ereport(ERROR) */
PG_exception_stack = &local_sigjmp_buf;
PG_SETMASK(&UnBlockSig);
/*
* Force zero_damaged_pages OFF in the autovac process, even if it is set
* in postgresql.conf. We don't really want such a dangerous option being
* applied non-interactively.
*/
SetConfigOption("zero_damaged_pages", "false", PGC_SUSET, PGC_S_OVERRIDE);
/* Get a list of databases */
dblist = autovac_get_database_list();
/*
* Determine the oldest datfrozenxid/relfrozenxid that we will allow
* to pass without forcing a vacuum. (This limit can be tightened for
* particular tables, but not loosened.)
*/
recentXid = ReadNewTransactionId();
xidForceLimit = recentXid - autovacuum_freeze_max_age;
/* ensure it's a "normal" XID, else TransactionIdPrecedes misbehaves */
if (xidForceLimit < FirstNormalTransactionId)
xidForceLimit -= FirstNormalTransactionId;
/*
* Choose a database to connect to. We pick the database that was least
* recently auto-vacuumed, or one that needs vacuuming to prevent Xid
* wraparound-related data loss. If any db at risk of wraparound is
* found, we pick the one with oldest datfrozenxid,
* independently of autovacuum times.
*
* Note that a database with no stats entry is not considered, except for
* Xid wraparound purposes. The theory is that if no one has ever
* connected to it since the stats were last initialized, it doesn't need
* vacuuming.
*
* XXX This could be improved if we had more info about whether it needs
* vacuuming before connecting to it. Perhaps look through the pgstats
* data for the database's tables? One idea is to keep track of the
* number of new and dead tuples per database in pgstats. However it
* isn't clear how to construct a metric that measures that and not cause
* starvation for less busy databases.
*/
db = NULL;
for_xid_wrap = false;
foreach(cell, dblist)
{
autovac_dbase *tmp = lfirst(cell);
/* Find pgstat entry if any */
tmp->entry = pgstat_fetch_stat_dbentry(tmp->oid);
/* Check to see if this one is at risk of wraparound */
if (TransactionIdPrecedes(tmp->frozenxid, xidForceLimit))
{
if (db == NULL ||
TransactionIdPrecedes(tmp->frozenxid, db->frozenxid))
db = tmp;
for_xid_wrap = true;
continue;
}
else if (for_xid_wrap)
continue; /* ignore not-at-risk DBs */
/*
* Otherwise, skip a database with no pgstat entry; it means it
* hasn't seen any activity.
*/
if (!tmp->entry)
continue;
/*
* Remember the db with oldest autovac time. (If we are here,
* both tmp->entry and db->entry must be non-null.)
*/
if (db == NULL ||
tmp->entry->last_autovac_time < db->entry->last_autovac_time)
db = tmp;
}
static Datum internalCallHandler(bool trusted, PG_FUNCTION_ARGS)
{
Invocation ctx;
Datum retval = 0;
#ifdef USE_PLJAVA_SIGHANDLERS
/*
* Substitute SIGUSR1 handler CatchupInterruptHandler with new local
* handler PLCatchupInterruptHandler, which will strictly let main
* thread of GPDB to process the signal rather than any other threads.
*
* Resolving possible lwlock releasing crash caused by SIGUSR1.
* MPP-23735
*/
if(!s_handlerSubstituted)
{
s_mainThreadIdForHandler = pthread_self();
s_oldHandlerFunc = pqsignal(SIGUSR1, PLCatchupInterruptHandler);
s_handlerSubstituted = true;
}
#endif
/*
* Just in case it could be helpful in offering diagnostics later, hang
* on to an Oid that is known to refer to PL/Java (because it got here).
* It's cheap, and can be followed back to the right language and
* handler function entries later if needed.
*/
*(trusted ? &pljavaTrustedOid : &pljavaUntrustedOid)
= fcinfo->flinfo->fn_oid;
if ( IS_COMPLETE != initstage )
{
initsequencer( initstage, false);
/* Force initial setting
*/
s_currentTrust = !trusted;
}
Invocation_pushInvocation(&ctx, trusted);
PG_TRY();
{
Function function = Function_getFunction(fcinfo);
if(CALLED_AS_TRIGGER(fcinfo))
{
/* Called as a trigger procedure
*/
retval = Function_invokeTrigger(function, fcinfo);
}
else
{
/* Called as a function
*/
retval = Function_invoke(function, fcinfo);
}
Invocation_popInvocation(false);
}
PG_CATCH();
{
Invocation_popInvocation(true);
PG_RE_THROW();
}
PG_END_TRY();
return retval;
}
/*------------ MAIN ----------------------------------------*/
int
main(int argc, char **argv)
{
int c;
progname = get_progname(argv[0]);
if (argc > 1)
{
if (strcmp(argv[1], "--help") == 0 || strcmp(argv[1], "-?") == 0)
{
usage();
exit(0);
}
if (strcmp(argv[1], "--version") == 0 || strcmp(argv[1], "-V") == 0)
{
puts("pg_standby (PostgreSQL) " PG_VERSION);
exit(0);
}
}
#ifndef WIN32
/*
* You can send SIGUSR1 to trigger failover.
*
* Postmaster uses SIGQUIT to request immediate shutdown. The default
* action is to core dump, but we don't want that, so trap it and commit
* suicide without core dump.
*
* We used to use SIGINT and SIGQUIT to trigger failover, but that turned
* out to be a bad idea because postmaster uses SIGQUIT to request
* immediate shutdown. We still trap SIGINT, but that may change in a
* future release.
*
* There's no way to trigger failover via signal on Windows.
*/
(void) pqsignal(SIGUSR1, sighandler);
(void) pqsignal(SIGINT, sighandler); /* deprecated, use SIGUSR1 */
(void) pqsignal(SIGQUIT, sigquit_handler);
#endif
while ((c = getopt(argc, argv, "cdk:lr:s:t:w:")) != -1)
{
switch (c)
{
case 'c': /* Use copy */
restoreCommandType = RESTORE_COMMAND_COPY;
break;
case 'd': /* Debug mode */
debug = true;
break;
case 'k': /* keepfiles */
keepfiles = atoi(optarg);
if (keepfiles < 0)
{
fprintf(stderr, "%s: -k keepfiles must be >= 0\n", progname);
exit(2);
}
break;
case 'l': /* Use link */
/*
* Link feature disabled, possibly permanently. Linking causes
* a problem after recovery ends that is not currently
* resolved by PostgreSQL. 25 Jun 2009
*/
#ifdef NOT_USED
restoreCommandType = RESTORE_COMMAND_LINK;
#endif
break;
case 'r': /* Retries */
maxretries = atoi(optarg);
if (maxretries < 0)
{
fprintf(stderr, "%s: -r maxretries must be >= 0\n", progname);
exit(2);
}
break;
case 's': /* Sleep time */
sleeptime = atoi(optarg);
if (sleeptime <= 0 || sleeptime > 60)
{
fprintf(stderr, "%s: -s sleeptime incorrectly set\n", progname);
exit(2);
}
break;
case 't': /* Trigger file */
triggerPath = strdup(optarg);
break;
case 'w': /* Max wait time */
maxwaittime = atoi(optarg);
if (maxwaittime < 0)
{
fprintf(stderr, "%s: -w maxwaittime incorrectly set\n", progname);
exit(2);
}
break;
default:
fprintf(stderr, "Try \"%s --help\" for more information.\n", progname);
exit(2);
break;
}
}
/*
* Parameter checking - after checking to see if trigger file present
*/
if (argc == 1)
{
fprintf(stderr, "%s: not enough command-line arguments\n", progname);
exit(2);
}
/*
* We will go to the archiveLocation to get nextWALFileName.
* nextWALFileName may not exist yet, which would not be an error, so we
* separate the archiveLocation and nextWALFileName so we can check
* separately whether archiveLocation exists, if not that is an error
*/
if (optind < argc)
{
archiveLocation = argv[optind];
optind++;
}
else
{
fprintf(stderr, "%s: must specify archive location\n", progname);
fprintf(stderr, "Try \"%s --help\" for more information.\n", progname);
exit(2);
}
if (optind < argc)
{
nextWALFileName = argv[optind];
optind++;
}
else
{
fprintf(stderr, "%s: must specify WAL file name as second non-option argument (use \"%%f\")\n", progname);
fprintf(stderr, "Try \"%s --help\" for more information.\n", progname);
exit(2);
}
if (optind < argc)
{
xlogFilePath = argv[optind];
optind++;
}
else
{
fprintf(stderr, "%s: must specify xlog destination as third non-option argument (use \"%%p\")\n", progname);
fprintf(stderr, "Try \"%s --help\" for more information.\n", progname);
exit(2);
}
if (optind < argc)
{
restartWALFileName = argv[optind];
optind++;
}
CustomizableInitialize();
need_cleanup = SetWALFileNameForCleanup();
if (debug)
{
fprintf(stderr, "Trigger file: %s\n", triggerPath ? triggerPath : "<not set>");
fprintf(stderr, "Waiting for WAL file: %s\n", nextWALFileName);
fprintf(stderr, "WAL file path: %s\n", WALFilePath);
fprintf(stderr, "Restoring to: %s\n", xlogFilePath);
fprintf(stderr, "Sleep interval: %d second%s\n",
sleeptime, (sleeptime > 1 ? "s" : " "));
fprintf(stderr, "Max wait interval: %d %s\n",
maxwaittime, (maxwaittime > 0 ? "seconds" : "forever"));
fprintf(stderr, "Command for restore: %s\n", restoreCommand);
fprintf(stderr, "Keep archive history: ");
if (need_cleanup)
fprintf(stderr, "%s and later\n", exclusiveCleanupFileName);
else
fprintf(stderr, "no cleanup required\n");
fflush(stderr);
}
/*
* Check for initial history file: always the first file to be requested
* It's OK if the file isn't there - all other files need to wait
*/
if (strlen(nextWALFileName) > 8 &&
strspn(nextWALFileName, "0123456789ABCDEF") == 8 &&
strcmp(nextWALFileName + strlen(nextWALFileName) - strlen(".history"),
".history") == 0)
{
nextWALFileType = XLOG_HISTORY;
if (RestoreWALFileForRecovery())
exit(0);
else
{
if (debug)
{
fprintf(stderr, "history file not found\n");
fflush(stderr);
}
exit(1);
}
}
/*
* Main wait loop
*/
for (;;)
{
/* Check for trigger file or signal first */
CheckForExternalTrigger();
#ifndef WIN32
if (signaled)
{
Failover = FastFailover;
if (debug)
{
fprintf(stderr, "signaled to exit: fast failover\n");
fflush(stderr);
}
}
#endif
/*
* Check for fast failover immediately, before checking if the
* requested WAL file is available
*/
if (Failover == FastFailover)
exit(1);
if (CustomizableNextWALFileReady())
{
/*
* Once we have restored this file successfully we can remove some
* prior WAL files. If this restore fails we musn't remove any
* file because some of them will be requested again immediately
* after the failed restore, or when we restart recovery.
*/
if (RestoreWALFileForRecovery())
{
if (need_cleanup)
CustomizableCleanupPriorWALFiles();
exit(0);
}
else
{
/* Something went wrong in copying the file */
exit(1);
}
}
/* Check for smart failover if the next WAL file was not available */
if (Failover == SmartFailover)
exit(1);
if (sleeptime <= 60)
pg_usleep(sleeptime * 1000000L);
waittime += sleeptime;
if (waittime >= maxwaittime && maxwaittime > 0)
{
Failover = FastFailover;
if (debug)
{
fprintf(stderr, "Timed out after %d seconds: fast failover\n",
waittime);
fflush(stderr);
}
}
if (debug)
{
fprintf(stderr, "WAL file not present yet.");
if (triggerPath)
fprintf(stderr, " Checking for trigger file...");
fprintf(stderr, "\n");
fflush(stderr);
}
}
}
/* We don't want SIGQUIT to core dump */
static void
sigquit_handler(int sig)
{
pqsignal(SIGINT, SIG_DFL);
kill(getpid(), SIGINT);
}
/*
* Main entry point for checkpointer process
*
* This is invoked from AuxiliaryProcessMain, which has already created the
* basic execution environment, but not enabled signals yet.
*/
void
CheckpointerMain(void)
{
sigjmp_buf local_sigjmp_buf;
MemoryContext checkpointer_context;
CheckpointerShmem->checkpointer_pid = MyProcPid;
/*
* Properly accept or ignore signals the postmaster might send us
*
* Note: we deliberately ignore SIGTERM, because during a standard Unix
* system shutdown cycle, init will SIGTERM all processes at once. We
* want to wait for the backends to exit, whereupon the postmaster will
* tell us it's okay to shut down (via SIGUSR2).
*/
pqsignal(SIGHUP, ChkptSigHupHandler); /* set flag to read config
* file */
pqsignal(SIGINT, ReqCheckpointHandler); /* request checkpoint */
pqsignal(SIGTERM, SIG_IGN); /* ignore SIGTERM */
pqsignal(SIGQUIT, chkpt_quickdie); /* hard crash time */
pqsignal(SIGALRM, SIG_IGN);
pqsignal(SIGPIPE, SIG_IGN);
pqsignal(SIGUSR1, chkpt_sigusr1_handler);
pqsignal(SIGUSR2, ReqShutdownHandler); /* request shutdown */
/*
* 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);
/* We allow SIGQUIT (quickdie) at all times */
sigdelset(&BlockSig, SIGQUIT);
/*
* Initialize so that first time-driven event happens at the correct time.
*/
last_checkpoint_time = last_xlog_switch_time = (pg_time_t) time(NULL);
/*
* Create a resource owner to keep track of our resources (currently only
* buffer pins).
*/
CurrentResourceOwner = ResourceOwnerCreate(NULL, "Checkpointer");
/*
* Create a memory context that we will do all our work in. We do this so
* that we can reset the context during error recovery and thereby avoid
* possible memory leaks. Formerly this code just ran in
* TopMemoryContext, but resetting that would be a really bad idea.
*/
checkpointer_context = AllocSetContextCreate(TopMemoryContext,
"Checkpointer",
ALLOCSET_DEFAULT_SIZES);
MemoryContextSwitchTo(checkpointer_context);
/*
* If an exception is encountered, processing resumes here.
*
* See notes in postgres.c about the design of this coding.
*/
if (sigsetjmp(local_sigjmp_buf, 1) != 0)
{
/* Since not using PG_TRY, must reset error stack by hand */
error_context_stack = NULL;
/* Prevent interrupts while cleaning up */
HOLD_INTERRUPTS();
/* Report the error to the server log */
EmitErrorReport();
/*
* These operations are really just a minimal subset of
* AbortTransaction(). We don't have very many resources to worry
* about in checkpointer, but we do have LWLocks, buffers, and temp
* files.
*/
LWLockReleaseAll();
ConditionVariableCancelSleep();
pgstat_report_wait_end();
AbortBufferIO();
UnlockBuffers();
/* buffer pins are released here: */
ResourceOwnerRelease(CurrentResourceOwner,
RESOURCE_RELEASE_BEFORE_LOCKS,
false, true);
/* we needn't bother with the other ResourceOwnerRelease phases */
AtEOXact_Buffers(false);
AtEOXact_SMgr();
AtEOXact_Files();
AtEOXact_HashTables(false);
/* Warn any waiting backends that the checkpoint failed. */
if (ckpt_active)
{
SpinLockAcquire(&CheckpointerShmem->ckpt_lck);
CheckpointerShmem->ckpt_failed++;
CheckpointerShmem->ckpt_done = CheckpointerShmem->ckpt_started;
SpinLockRelease(&CheckpointerShmem->ckpt_lck);
ckpt_active = false;
}
/*
* Now return to normal top-level context and clear ErrorContext for
* next time.
*/
MemoryContextSwitchTo(checkpointer_context);
FlushErrorState();
/* Flush any leaked data in the top-level context */
MemoryContextResetAndDeleteChildren(checkpointer_context);
/* Now we can allow interrupts again */
RESUME_INTERRUPTS();
/*
* Sleep at least 1 second after any error. A write error is likely
* to be repeated, and we don't want to be filling the error logs as
* fast as we can.
*/
pg_usleep(1000000L);
/*
* Close all open files after any error. This is helpful on Windows,
* where holding deleted files open causes various strange errors.
* It's not clear we need it elsewhere, but shouldn't hurt.
*/
smgrcloseall();
}
/* We can now handle ereport(ERROR) */
PG_exception_stack = &local_sigjmp_buf;
/*
* Unblock signals (they were blocked when the postmaster forked us)
*/
PG_SETMASK(&UnBlockSig);
/*
* Ensure all shared memory values are set correctly for the config. Doing
* this here ensures no race conditions from other concurrent updaters.
*/
UpdateSharedMemoryConfig();
/*
* Advertise our latch that backends can use to wake us up while we're
* sleeping.
*/
ProcGlobal->checkpointerLatch = &MyProc->procLatch;
/*
* Loop forever
*/
for (;;)
{
bool do_checkpoint = false;
int flags = 0;
pg_time_t now;
int elapsed_secs;
int cur_timeout;
int rc;
/* Clear any already-pending wakeups */
ResetLatch(MyLatch);
/*
* Process any requests or signals received recently.
*/
AbsorbFsyncRequests();
if (got_SIGHUP)
{
got_SIGHUP = false;
ProcessConfigFile(PGC_SIGHUP);
/*
* Checkpointer is the last process to shut down, so we ask it to
* hold the keys for a range of other tasks required most of which
* have nothing to do with checkpointing at all.
*
* For various reasons, some config values can change dynamically
* so the primary copy of them is held in shared memory to make
* sure all backends see the same value. We make Checkpointer
* responsible for updating the shared memory copy if the
* parameter setting changes because of SIGHUP.
*/
UpdateSharedMemoryConfig();
}
if (checkpoint_requested)
{
checkpoint_requested = false;
do_checkpoint = true;
BgWriterStats.m_requested_checkpoints++;
}
if (shutdown_requested)
{
/*
* From here on, elog(ERROR) should end with exit(1), not send
* control back to the sigsetjmp block above
*/
ExitOnAnyError = true;
/* Close down the database */
ShutdownXLOG(0, 0);
/* Normal exit from the checkpointer is here */
proc_exit(0); /* done */
}
/*
* Force a checkpoint if too much time has elapsed since the last one.
* Note that we count a timed checkpoint in stats only when this
* occurs without an external request, but we set the CAUSE_TIME flag
* bit even if there is also an external request.
*/
now = (pg_time_t) time(NULL);
elapsed_secs = now - last_checkpoint_time;
if (elapsed_secs >= CheckPointTimeout)
{
if (!do_checkpoint)
BgWriterStats.m_timed_checkpoints++;
do_checkpoint = true;
flags |= CHECKPOINT_CAUSE_TIME;
}
/*
* Do a checkpoint if requested.
*/
if (do_checkpoint)
{
bool ckpt_performed = false;
bool do_restartpoint;
/*
* Check if we should perform a checkpoint or a restartpoint. As a
* side-effect, RecoveryInProgress() initializes TimeLineID if
* it's not set yet.
*/
do_restartpoint = RecoveryInProgress();
/*
* Atomically fetch the request flags to figure out what kind of a
* checkpoint we should perform, and increase the started-counter
* to acknowledge that we've started a new checkpoint.
*/
SpinLockAcquire(&CheckpointerShmem->ckpt_lck);
flags |= CheckpointerShmem->ckpt_flags;
CheckpointerShmem->ckpt_flags = 0;
CheckpointerShmem->ckpt_started++;
SpinLockRelease(&CheckpointerShmem->ckpt_lck);
/*
* The end-of-recovery checkpoint is a real checkpoint that's
* performed while we're still in recovery.
*/
if (flags & CHECKPOINT_END_OF_RECOVERY)
do_restartpoint = false;
/*
* We will warn if (a) too soon since last checkpoint (whatever
* caused it) and (b) somebody set the CHECKPOINT_CAUSE_XLOG flag
* since the last checkpoint start. Note in particular that this
* implementation will not generate warnings caused by
* CheckPointTimeout < CheckPointWarning.
*/
if (!do_restartpoint &&
(flags & CHECKPOINT_CAUSE_XLOG) &&
elapsed_secs < CheckPointWarning)
ereport(LOG,
(errmsg_plural("checkpoints are occurring too frequently (%d second apart)",
"checkpoints are occurring too frequently (%d seconds apart)",
elapsed_secs,
elapsed_secs),
errhint("Consider increasing the configuration parameter \"max_wal_size\".")));
/*
* Initialize checkpointer-private variables used during
* checkpoint.
*/
ckpt_active = true;
if (do_restartpoint)
ckpt_start_recptr = GetXLogReplayRecPtr(NULL);
else
ckpt_start_recptr = GetInsertRecPtr();
ckpt_start_time = now;
ckpt_cached_elapsed = 0;
/*
* Do the checkpoint.
*/
if (!do_restartpoint)
{
CreateCheckPoint(flags);
ckpt_performed = true;
}
else
ckpt_performed = CreateRestartPoint(flags);
/*
* After any checkpoint, close all smgr files. This is so we
* won't hang onto smgr references to deleted files indefinitely.
*/
smgrcloseall();
/*
* Indicate checkpoint completion to any waiting backends.
*/
SpinLockAcquire(&CheckpointerShmem->ckpt_lck);
CheckpointerShmem->ckpt_done = CheckpointerShmem->ckpt_started;
SpinLockRelease(&CheckpointerShmem->ckpt_lck);
if (ckpt_performed)
{
/*
* Note we record the checkpoint start time not end time as
* last_checkpoint_time. This is so that time-driven
* checkpoints happen at a predictable spacing.
*/
last_checkpoint_time = now;
}
else
{
/*
* We were not able to perform the restartpoint (checkpoints
* throw an ERROR in case of error). Most likely because we
* have not received any new checkpoint WAL records since the
* last restartpoint. Try again in 15 s.
*/
last_checkpoint_time = now - CheckPointTimeout + 15;
}
ckpt_active = false;
}
/* Check for archive_timeout and switch xlog files if necessary. */
CheckArchiveTimeout();
/*
* Send off activity statistics to the stats collector. (The reason
* why we re-use bgwriter-related code for this is that the bgwriter
* and checkpointer used to be just one process. It's probably not
* worth the trouble to split the stats support into two independent
* stats message types.)
*/
pgstat_send_bgwriter();
/*
* Sleep until we are signaled or it's time for another checkpoint or
* xlog file switch.
*/
now = (pg_time_t) time(NULL);
elapsed_secs = now - last_checkpoint_time;
if (elapsed_secs >= CheckPointTimeout)
continue; /* no sleep for us ... */
cur_timeout = CheckPointTimeout - elapsed_secs;
if (XLogArchiveTimeout > 0 && !RecoveryInProgress())
{
elapsed_secs = now - last_xlog_switch_time;
if (elapsed_secs >= XLogArchiveTimeout)
continue; /* no sleep for us ... */
cur_timeout = Min(cur_timeout, XLogArchiveTimeout - elapsed_secs);
}
rc = WaitLatch(MyLatch,
WL_LATCH_SET | WL_TIMEOUT | WL_POSTMASTER_DEATH,
cur_timeout * 1000L /* convert to ms */,
WAIT_EVENT_CHECKPOINTER_MAIN);
/*
* Emergency bailout if postmaster has died. This is to avoid the
* necessity for manual cleanup of all postmaster children.
*/
if (rc & WL_POSTMASTER_DEATH)
exit(1);
}
}
void
setup_cancel_handler(void)
{
pqsignal(SIGINT, handle_sigint);
}
/*
* Main entry point for syslogger process
* argc/argv parameters are valid only in EXEC_BACKEND case.
*/
NON_EXEC_STATIC void
SysLoggerMain(int argc, char *argv[])
{
#ifndef WIN32
char logbuffer[READ_BUF_SIZE];
int bytes_in_logbuffer = 0;
#endif
char *currentLogDir;
char *currentLogFilename;
int currentLogRotationAge;
pg_time_t now;
IsUnderPostmaster = true; /* we are a postmaster subprocess now */
MyProcPid = getpid(); /* reset MyProcPid */
MyStartTime = time(NULL); /* set our start time in case we call elog */
now = MyStartTime;
#ifdef EXEC_BACKEND
syslogger_parseArgs(argc, argv);
#endif /* EXEC_BACKEND */
am_syslogger = true;
init_ps_display("logger process", "", "", "");
/*
* If we restarted, our stderr is already redirected into our own input
* pipe. This is of course pretty useless, not to mention that it
* interferes with detecting pipe EOF. Point stderr to /dev/null. This
* assumes that all interesting messages generated in the syslogger will
* come through elog.c and will be sent to write_syslogger_file.
*/
if (redirection_done)
{
int fd = open(DEVNULL, O_WRONLY, 0);
/*
* The closes might look redundant, but they are not: we want to be
* darn sure the pipe gets closed even if the open failed. We can
* survive running with stderr pointing nowhere, but we can't afford
* to have extra pipe input descriptors hanging around.
*/
close(fileno(stdout));
close(fileno(stderr));
if (fd != -1)
{
dup2(fd, fileno(stdout));
dup2(fd, fileno(stderr));
close(fd);
}
}
/*
* Syslogger's own stderr can't be the syslogPipe, so set it back to text
* mode if we didn't just close it. (It was set to binary in
* SubPostmasterMain).
*/
#ifdef WIN32
else
_setmode(_fileno(stderr), _O_TEXT);
#endif
/*
* Also close our copy of the write end of the pipe. This is needed to
* ensure we can detect pipe EOF correctly. (But note that in the restart
* case, the postmaster already did this.)
*/
#ifndef WIN32
if (syslogPipe[1] >= 0)
close(syslogPipe[1]);
syslogPipe[1] = -1;
#else
if (syslogPipe[1])
CloseHandle(syslogPipe[1]);
syslogPipe[1] = 0;
#endif
/*
* If possible, make this process a group leader, so that the postmaster
* can signal any child processes too. (syslogger probably never has any
* child processes, but for consistency we make all postmaster child
* processes do this.)
*/
#ifdef HAVE_SETSID
if (setsid() < 0)
elog(FATAL, "setsid() failed: %m");
#endif
InitializeLatchSupport(); /* needed for latch waits */
/* Initialize private latch for use by signal handlers */
InitLatch(&sysLoggerLatch);
/*
* Properly accept or ignore signals the postmaster might send us
*
* Note: we ignore all termination signals, and instead exit only when all
* upstream processes are gone, to ensure we don't miss any dying gasps of
* broken backends...
*/
pqsignal(SIGHUP, sigHupHandler); /* set flag to read config file */
pqsignal(SIGINT, SIG_IGN);
pqsignal(SIGTERM, SIG_IGN);
pqsignal(SIGQUIT, SIG_IGN);
pqsignal(SIGALRM, SIG_IGN);
pqsignal(SIGPIPE, SIG_IGN);
pqsignal(SIGUSR1, sigUsr1Handler); /* request log rotation */
pqsignal(SIGUSR2, SIG_IGN);
/*
* 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);
PG_SETMASK(&UnBlockSig);
#ifdef WIN32
/* Fire up separate data transfer thread */
InitializeCriticalSection(&sysloggerSection);
EnterCriticalSection(&sysloggerSection);
threadHandle = (HANDLE) _beginthreadex(NULL, 0, pipeThread, NULL, 0, NULL);
if (threadHandle == 0)
elog(FATAL, "could not create syslogger data transfer thread: %m");
#endif /* WIN32 */
/*
* Remember active logfile's name. We recompute this from the reference
* time because passing down just the pg_time_t is a lot cheaper than
* passing a whole file path in the EXEC_BACKEND case.
*/
last_file_name = logfile_getname(first_syslogger_file_time, NULL);
/* remember active logfile parameters */
currentLogDir = pstrdup(Log_directory);
currentLogFilename = pstrdup(Log_filename);
currentLogRotationAge = Log_RotationAge;
/* set next planned rotation time */
set_next_rotation_time();
/* main worker loop */
for (;;)
{
bool time_based_rotation = false;
int size_rotation_for = 0;
long cur_timeout;
int cur_flags;
#ifndef WIN32
int rc;
#endif
/* Clear any already-pending wakeups */
ResetLatch(&sysLoggerLatch);
/*
* Process any requests or signals received recently.
*/
if (got_SIGHUP)
{
got_SIGHUP = false;
ProcessConfigFile(PGC_SIGHUP);
/*
* Check if the log directory or filename pattern changed in
* postgresql.conf. If so, force rotation to make sure we're
* writing the logfiles in the right place.
*/
if (strcmp(Log_directory, currentLogDir) != 0)
{
pfree(currentLogDir);
currentLogDir = pstrdup(Log_directory);
rotation_requested = true;
/*
* Also, create new directory if not present; ignore errors
*/
mkdir(Log_directory, S_IRWXU);
}
if (strcmp(Log_filename, currentLogFilename) != 0)
{
pfree(currentLogFilename);
currentLogFilename = pstrdup(Log_filename);
rotation_requested = true;
}
/*
* If rotation time parameter changed, reset next rotation time,
* but don't immediately force a rotation.
*/
if (currentLogRotationAge != Log_RotationAge)
{
currentLogRotationAge = Log_RotationAge;
set_next_rotation_time();
}
/*
* If we had a rotation-disabling failure, re-enable rotation
* attempts after SIGHUP, and force one immediately.
*/
if (rotation_disabled)
{
rotation_disabled = false;
rotation_requested = true;
}
}
if (Log_RotationAge > 0 && !rotation_disabled)
{
/* Do a logfile rotation if it's time */
now = (pg_time_t) time(NULL);
if (now >= next_rotation_time)
rotation_requested = time_based_rotation = true;
}
if (!rotation_requested && Log_RotationSize > 0 && !rotation_disabled)
{
/* Do a rotation if file is too big */
if (ftell(syslogFile) >= Log_RotationSize * 1024L)
{
rotation_requested = true;
size_rotation_for |= LOG_DESTINATION_STDERR;
}
if (csvlogFile != NULL &&
ftell(csvlogFile) >= Log_RotationSize * 1024L)
{
rotation_requested = true;
size_rotation_for |= LOG_DESTINATION_CSVLOG;
}
}
if (rotation_requested)
{
/*
* Force rotation when both values are zero. It means the request
* was sent by pg_rotate_logfile.
*/
if (!time_based_rotation && size_rotation_for == 0)
size_rotation_for = LOG_DESTINATION_STDERR | LOG_DESTINATION_CSVLOG;
logfile_rotate(time_based_rotation, size_rotation_for);
}
/*
* Calculate time till next time-based rotation, so that we don't
* sleep longer than that. We assume the value of "now" obtained
* above is still close enough. Note we can't make this calculation
* until after calling logfile_rotate(), since it will advance
* next_rotation_time.
*
* Also note that we need to beware of overflow in calculation of the
* timeout: with large settings of Log_RotationAge, next_rotation_time
* could be more than INT_MAX msec in the future. In that case we'll
* wait no more than INT_MAX msec, and try again.
*/
if (Log_RotationAge > 0 && !rotation_disabled)
{
pg_time_t delay;
delay = next_rotation_time - now;
if (delay > 0)
{
if (delay > INT_MAX / 1000)
delay = INT_MAX / 1000;
cur_timeout = delay * 1000L; /* msec */
}
else
cur_timeout = 0;
cur_flags = WL_TIMEOUT;
}
else
{
cur_timeout = -1L;
cur_flags = 0;
}
/*
* Sleep until there's something to do
*/
#ifndef WIN32
rc = WaitLatchOrSocket(&sysLoggerLatch,
WL_LATCH_SET | WL_SOCKET_READABLE | cur_flags,
syslogPipe[0],
cur_timeout);
if (rc & WL_SOCKET_READABLE)
{
int bytesRead;
bytesRead = read(syslogPipe[0],
logbuffer + bytes_in_logbuffer,
sizeof(logbuffer) - bytes_in_logbuffer);
if (bytesRead < 0)
{
if (errno != EINTR)
ereport(LOG,
(errcode_for_socket_access(),
errmsg("could not read from logger pipe: %m")));
}
else if (bytesRead > 0)
{
bytes_in_logbuffer += bytesRead;
process_pipe_input(logbuffer, &bytes_in_logbuffer);
continue;
}
else
{
/*
* Zero bytes read when select() is saying read-ready means
* EOF on the pipe: that is, there are no longer any processes
* with the pipe write end open. Therefore, the postmaster
* and all backends are shut down, and we are done.
*/
pipe_eof_seen = true;
/* if there's any data left then force it out now */
flush_pipe_input(logbuffer, &bytes_in_logbuffer);
}
}
#else /* WIN32 */
/*
* On Windows we leave it to a separate thread to transfer data and
* detect pipe EOF. The main thread just wakes up to handle SIGHUP
* and rotation conditions.
*
* Server code isn't generally thread-safe, so we ensure that only one
* of the threads is active at a time by entering the critical section
* whenever we're not sleeping.
*/
LeaveCriticalSection(&sysloggerSection);
(void) WaitLatch(&sysLoggerLatch,
WL_LATCH_SET | cur_flags,
cur_timeout);
EnterCriticalSection(&sysloggerSection);
#endif /* WIN32 */
if (pipe_eof_seen)
{
/*
* seeing this message on the real stderr is annoying - so we make
* it DEBUG1 to suppress in normal use.
*/
ereport(DEBUG1,
(errmsg("logger shutting down")));
/*
* Normal exit from the syslogger is here. Note that we
* deliberately do not close syslogFile before exiting; this is to
* allow for the possibility of elog messages being generated
* inside proc_exit. Regular exit() will take care of flushing
* and closing stdio channels.
*/
proc_exit(0);
}
}
}
/*
* helpSQL -- help with SQL commands
*
* Note: we assume caller removed any trailing spaces in "topic".
*/
void
helpSQL(const char *topic, unsigned short int pager)
{
#define VALUE_OR_NULL(a) ((a) ? (a) : "")
if (!topic || strlen(topic) == 0)
{
/* Print all the available command names */
int screen_width;
int ncolumns;
int nrows;
FILE *output;
int i;
int j;
#ifdef TIOCGWINSZ
struct winsize screen_size;
if (ioctl(fileno(stdout), TIOCGWINSZ, &screen_size) == -1)
screen_width = 80; /* ioctl failed, assume 80 */
else
screen_width = screen_size.ws_col;
#else
screen_width = 80; /* default assumption */
#endif
ncolumns = (screen_width - 3) / (QL_MAX_CMD_LEN + 1);
ncolumns = Max(ncolumns, 1);
nrows = (QL_HELP_COUNT + (ncolumns - 1)) / ncolumns;
output = PageOutput(nrows + 1, pager);
fputs(_("Available help:\n"), output);
for (i = 0; i < nrows; i++)
{
fprintf(output, " ");
for (j = 0; j < ncolumns - 1; j++)
fprintf(output, "%-*s",
QL_MAX_CMD_LEN + 1,
VALUE_OR_NULL(QL_HELP[i + j * nrows].cmd));
if (i + j * nrows < QL_HELP_COUNT)
fprintf(output, "%s",
VALUE_OR_NULL(QL_HELP[i + j * nrows].cmd));
fputc('\n', output);
}
/* Only close if we used the pager */
if (output != stdout)
{
pclose(output);
#ifndef WIN32
pqsignal(SIGPIPE, SIG_DFL);
#endif
}
}
else
{
int i,
j,
x = 0;
bool help_found = false;
FILE *output;
size_t len,
wordlen;
int nl_count = 0;
/*
* We first try exact match, then first + second words, then first
* word only.
*/
len = strlen(topic);
for (x = 1; x <= 3; x++)
{
if (x > 1) /* Nothing on first pass - try the opening
* word(s) */
{
wordlen = j = 1;
while (topic[j] != ' ' && j++ < len)
wordlen++;
if (x == 2)
{
j++;
while (topic[j] != ' ' && j++ <= len)
wordlen++;
}
if (wordlen >= len) /* Don't try again if the same word */
{
output = PageOutput(nl_count, pager);
break;
}
len = wordlen;
}
/* Count newlines for pager */
for (i = 0; QL_HELP[i].cmd; i++)
{
if (pg_strncasecmp(topic, QL_HELP[i].cmd, len) == 0 ||
strcmp(topic, "*") == 0)
{
nl_count += 5 + QL_HELP[i].nl_count;
/* If we have an exact match, exit. Fixes \h SELECT */
if (pg_strcasecmp(topic, QL_HELP[i].cmd) == 0)
break;
}
}
output = PageOutput(nl_count, pager);
for (i = 0; QL_HELP[i].cmd; i++)
{
if (pg_strncasecmp(topic, QL_HELP[i].cmd, len) == 0 ||
strcmp(topic, "*") == 0)
{
PQExpBufferData buffer;
initPQExpBuffer(&buffer);
QL_HELP[i].syntaxfunc(&buffer);
help_found = true;
fprintf(output, _("Command: %s\n"
"Description: %s\n"
"Syntax:\n%s\n\n"),
QL_HELP[i].cmd,
_(QL_HELP[i].help),
buffer.data);
/* If we have an exact match, exit. Fixes \h SELECT */
if (pg_strcasecmp(topic, QL_HELP[i].cmd) == 0)
break;
}
}
if (help_found) /* Don't keep trying if we got a match */
break;
}
if (!help_found)
fprintf(output, _("No help available for \"%s\".\nTry \\h with no arguments to see available help.\n"), topic);
/* Only close if we used the pager */
if (output != stdout)
{
pclose(output);
#ifndef WIN32
pqsignal(SIGPIPE, SIG_DFL);
#endif
}
}
}
/*
* There are a few ways to arrive in the initsequencer.
* 1. From _PG_init (called exactly once when the library is loaded for ANY
* reason).
* 1a. Because of the command LOAD 'libraryname';
* This case can be distinguished because _PG_init will have found the
* LOAD command and saved the 'libraryname' in pljavaLoadPath.
* 1b. Because of a CREATE FUNCTION naming this library. pljavaLoadPath will
* be NULL.
* 1c. By the first actual use of a PL/Java function, causing this library
* to be loaded. pljavaLoadPath will be NULL. The called function's Oid
* will be available to the call handler once we return from _PG_init,
* but it isn't (easily) available here.
* 2. From the call handler, if initialization isn't complete yet. That can only
* mean something failed in the earlier call to _PG_init, and whatever it was
* is highly likely to fail again. That may lead to the untidyness of
* duplicated diagnostic messages, but for now I like the belt-and-suspenders
* approach of making sure the init sequence gets as many chances as possible
* to succeed.
* 3. From a GUC assign hook, if the user has updated a setting that might allow
* initialization to succeed. It resumes from where it left off.
*
* In all cases, the sequence must progress as far as starting the VM and
* initializing the PL/Java classes. In all cases except 1a, that's enough,
* assuming the language handlers and schema have all been set up already (or,
* in case 1b, the user is intent on setting them up explicitly).
*
* In case 1a, we can go ahead and test for, and create, the schema, functions,
* and language entries as needed, using pljavaLoadPath as the library path
* if creating the language handler functions. One-stop shopping. (The presence
* of pljavaLoadPath in any of the other cases, such as resumption by an assign
* hook, indicates it is really a continuation of case 1a.)
*/
static void initsequencer(enum initstage is, bool tolerant)
{
JVMOptList optList;
Invocation ctx;
jint JNIresult;
char *greeting;
switch (is)
{
case IS_FORMLESS_VOID:
initstage = IS_GUCS_REGISTERED;
case IS_GUCS_REGISTERED:
libjvmlocation = strdup("libjvm.so");
initstage = IS_PLJAVA_ENABLED;
case IS_PLJAVA_ENABLED:
libjvm_handle = pg_dlopen(libjvmlocation);
if ( NULL == libjvm_handle )
{
ereport(ERROR, (
errmsg("Cannot load libjvm.so library, check that it is available in LD_LIBRARY_PATH"),
errdetail("%s", (char *)pg_dlerror())));
goto check_tolerant;
}
initstage = IS_CAND_JVMOPENED;
case IS_CAND_JVMOPENED:
pljava_createvm =
(jint (JNICALL *)(JavaVM **, void **, void *))
pg_dlsym(libjvm_handle, "JNI_CreateJavaVM");
if ( NULL == pljava_createvm )
{
/*
* If it hasn't got the symbol, it can't be the right
* library, so close/unload it so another can be tried.
* Format the dlerror string first: dlclose may clobber it.
*/
char *dle = MemoryContextStrdup(ErrorContext, pg_dlerror());
pg_dlclose(libjvm_handle);
initstage = IS_CAND_JVMLOCATION;
ereport(ERROR, (
errmsg("Cannot start Java VM"),
errdetail("%s", dle),
errhint("Check that libjvm.so is available in LD_LIBRARY_PATH")));
goto check_tolerant;
}
initstage = IS_CREATEVM_SYM_FOUND;
case IS_CREATEVM_SYM_FOUND:
s_javaLogLevel = INFO;
checkIntTimeType();
HashMap_initialize(); /* creates things in TopMemoryContext */
#ifdef PLJAVA_DEBUG
/* Hard setting for debug. Don't forget to recompile...
*/
pljava_debug = 1;
#endif
initstage = IS_MISC_ONCE_DONE;
case IS_MISC_ONCE_DONE:
JVMOptList_init(&optList); /* uses CurrentMemoryContext */
seenVisualVMName = false;
addUserJVMOptions(&optList);
if ( ! seenVisualVMName )
JVMOptList_addVisualVMName(&optList);
JVMOptList_add(&optList, "vfprintf", (void*)my_vfprintf, true);
#ifndef GCJ
JVMOptList_add(&optList, "-Xrs", 0, true);
#endif
effectiveClassPath = getClassPath("-Djava.class.path=");
if(effectiveClassPath != 0)
{
JVMOptList_add(&optList, effectiveClassPath, 0, true);
}
initstage = IS_JAVAVM_OPTLIST;
case IS_JAVAVM_OPTLIST:
JNIresult = initializeJavaVM(&optList); /* frees the optList */
if( JNI_OK != JNIresult )
{
initstage = IS_MISC_ONCE_DONE; /* optList has been freed */
StaticAssertStmt(sizeof(jint) <= sizeof(long int),
"jint wider than long int?!");
ereport(WARNING,
(errmsg("failed to create Java virtual machine"),
errdetail("JNI_CreateJavaVM returned an error code: %ld",
(long int)JNIresult),
jvmStartedAtLeastOnce ?
errhint("Because an earlier attempt during this session "
"did start a VM before failing, this probably means your "
"Java runtime environment does not support more than one "
"VM creation per session. You may need to exit this "
"session and start a new one.") : 0));
goto check_tolerant;
}
jvmStartedAtLeastOnce = true;
elog(DEBUG2, "successfully created Java virtual machine");
initstage = IS_JAVAVM_STARTED;
case IS_JAVAVM_STARTED:
#ifdef USE_PLJAVA_SIGHANDLERS
pqsignal(SIGINT, pljavaStatementCancelHandler);
pqsignal(SIGTERM, pljavaDieHandler);
#endif
/* Register an on_proc_exit handler that destroys the VM
*/
on_proc_exit(_destroyJavaVM, 0);
initstage = IS_SIGHANDLERS;
case IS_SIGHANDLERS:
Invocation_pushBootContext(&ctx);
PG_TRY();
{
initPLJavaClasses();
initJavaSession();
Invocation_popBootContext();
initstage = IS_PLJAVA_FOUND;
}
PG_CATCH();
{
MemoryContextSwitchTo(ctx.upperContext); /* leave ErrorContext */
Invocation_popBootContext();
initstage = IS_MISC_ONCE_DONE;
/* We can't stay here...
*/
if ( tolerant )
reLogWithChangedLevel(WARNING); /* so xact is not aborted */
else
{
EmitErrorReport(); /* no more unwinding, just log it */
/* Seeing an ERROR emitted to the log, without leaving the
* transaction aborted, would violate the principle of least
* astonishment. But at check_tolerant below, another ERROR will
* be thrown immediately, so the transaction effect will be as
* expected and this ERROR will contribute information beyond
* what is in the generic one thrown down there.
*/
FlushErrorState();
}
}
PG_END_TRY();
if ( IS_PLJAVA_FOUND != initstage )
{
/* JVM initialization failed for some reason. Destroy
* the VM if it exists. Perhaps the user will try
* fixing the pljava.classpath and make a new attempt.
*/
ereport(WARNING, (
errmsg("failed to load initial PL/Java classes"),
errhint("The most common reason is that \"pljava_classpath\" "
"needs to be set, naming the proper \"pljava.jar\" file.")
));
_destroyJavaVM(0, 0);
goto check_tolerant;
}
case IS_PLJAVA_FOUND:
greeting = InstallHelper_hello();
ereport(NULL != pljavaLoadPath ? NOTICE : DEBUG1, (
errmsg("PL/Java loaded"),
errdetail("versions:\n%s", greeting)));
pfree(greeting);
if ( NULL != pljavaLoadPath )
InstallHelper_groundwork(); /* sqlj schema, language handlers, ...*/
initstage = IS_COMPLETE;
case IS_COMPLETE:
pljavaLoadingAsExtension = false;
if ( alteredSettingsWereNeeded )
{
/* Use this StringInfoData to conditionally construct part of the
* hint string suggesting ALTER DATABASE ... SET ... FROM CURRENT
* provided the server is >= 9.2 where that will actually work.
* In 9.3, psprintf appeared, which would make this all simpler,
* but if 9.3+ were all that had to be supported, this would all
* be moot anyway. Doing the initStringInfo inside the ereport
* ensures the string is allocated in ErrorContext and won't leak.
* Don't remove the extra parens grouping
* (initStringInfo, appendStringInfo, errhint) ... with the parens,
* that's a comma expression, which is sequenced; without them, they
* are just function parameters with evaluation order unknown.
*/
StringInfoData buf;
#if PG_VERSION_NUM >= 90200
#define MOREHINT \
appendStringInfo(&buf, \
"using ALTER DATABASE %s SET ... FROM CURRENT or ", \
pljavaDbName()),
#else
#define MOREHINT
#endif
ereport(NOTICE, (
errmsg("PL/Java successfully started after adjusting settings"),
(initStringInfo(&buf),
MOREHINT
errhint("The settings that worked should be saved (%s"
"in the \"%s\" file). For a reminder of what has been set, "
"try: SELECT name, setting FROM pg_settings WHERE name LIKE"
" 'pljava.%%' AND source = 'session'",
buf.data,
superuser()
? PG_GETCONFIGOPTION("config_file")
: "postgresql.conf"))));
#undef MOREHINT
if ( loadAsExtensionFailed )
{
ereport(NOTICE, (errmsg(
"PL/Java load successful after failed CREATE EXTENSION"),
errdetail(
"PL/Java is now installed, but not as an extension."),
errhint(
"To correct that, either COMMIT or ROLLBACK, make sure "
"the working settings are saved, exit this session, and "
"in a new session, either: "
"1. if committed, run "
"\"CREATE EXTENSION pljava FROM unpackaged\", or 2. "
"if rolled back, simply \"CREATE EXTENSION pljava\" again."
)));
}
}
return;
default:
ereport(ERROR, (
errmsg("cannot set up PL/Java"),
errdetail(
"An unexpected stage was reached in the startup sequence."),
errhint(
"Please report the circumstances to the PL/Java maintainers.")
));
}
check_tolerant:
if ( pljavaLoadingAsExtension )
{
tolerant = false;
loadAsExtensionFailed = true;
pljavaLoadingAsExtension = false;
}
if ( !tolerant )
{
ereport(ERROR, (
errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
errmsg(
"cannot use PL/Java before successfully completing its setup"),
errhint(
"Check the log for messages closely preceding this one, "
"detailing what step of setup failed and what will be needed, "
"probably setting one of the \"pljava.\" configuration "
"variables, to complete the setup. If there is not enough "
"help in the log, try again with different settings for "
"\"log_min_messages\" or \"log_error_verbosity\".")));
}
}
/*
* Main entry point for bgwriter process
*
* This is invoked from BootstrapMain, which has already created the basic
* execution environment, but not enabled signals yet.
*/
void
BackgroundWriterMain(void)
{
sigjmp_buf local_sigjmp_buf;
MemoryContext bgwriter_context;
am_bg_writer = true;
/*
* If possible, make this process a group leader, so that the postmaster
* can signal any child processes too. (bgwriter probably never has any
* child processes, but for consistency we make all postmaster child
* processes do this.)
*/
#ifdef HAVE_SETSID
if (setsid() < 0)
elog(FATAL, "setsid() failed: %m");
#endif
/*
* Properly accept or ignore signals the postmaster might send us
*
* SIGUSR1 is presently unused; keep it spare in case someday we want this
* process to participate in ProcSignal signalling.
*/
pqsignal(SIGHUP, BgSigHupHandler); /* set flag to read config file */
pqsignal(SIGINT, SIG_IGN); /* as of 9.2 no longer requests checkpoint */
pqsignal(SIGTERM, ReqShutdownHandler); /* shutdown */
pqsignal(SIGQUIT, bg_quickdie); /* hard crash time */
pqsignal(SIGALRM, SIG_IGN);
pqsignal(SIGPIPE, SIG_IGN);
pqsignal(SIGUSR1, SIG_IGN); /* reserve for ProcSignal */
pqsignal(SIGUSR2, SIG_IGN);
/*
* 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);
/* We allow SIGQUIT (quickdie) at all times */
sigdelset(&BlockSig, SIGQUIT);
/*
* Create a resource owner to keep track of our resources (currently only
* buffer pins).
*/
CurrentResourceOwner = ResourceOwnerCreate(NULL, "Background Writer");
/*
* Create a memory context that we will do all our work in. We do this so
* that we can reset the context during error recovery and thereby avoid
* possible memory leaks. Formerly this code just ran in
* TopMemoryContext, but resetting that would be a really bad idea.
*/
bgwriter_context = AllocSetContextCreate(TopMemoryContext,
"Background Writer",
ALLOCSET_DEFAULT_MINSIZE,
ALLOCSET_DEFAULT_INITSIZE,
ALLOCSET_DEFAULT_MAXSIZE);
MemoryContextSwitchTo(bgwriter_context);
/*
* If an exception is encountered, processing resumes here.
*
* See notes in postgres.c about the design of this coding.
*/
if (sigsetjmp(local_sigjmp_buf, 1) != 0)
{
/* Since not using PG_TRY, must reset error stack by hand */
error_context_stack = NULL;
/* Prevent interrupts while cleaning up */
HOLD_INTERRUPTS();
/* Report the error to the server log */
EmitErrorReport();
/*
* These operations are really just a minimal subset of
* AbortTransaction(). We don't have very many resources to worry
* about in bgwriter, but we do have LWLocks, buffers, and temp files.
*/
LWLockReleaseAll();
AbortBufferIO();
UnlockBuffers();
/* buffer pins are released here: */
ResourceOwnerRelease(CurrentResourceOwner,
RESOURCE_RELEASE_BEFORE_LOCKS,
false, true);
/* we needn't bother with the other ResourceOwnerRelease phases */
AtEOXact_Buffers(false);
AtEOXact_Files();
AtEOXact_HashTables(false);
/*
* Now return to normal top-level context and clear ErrorContext for
* next time.
*/
MemoryContextSwitchTo(bgwriter_context);
FlushErrorState();
/* Flush any leaked data in the top-level context */
MemoryContextResetAndDeleteChildren(bgwriter_context);
/* Now we can allow interrupts again */
RESUME_INTERRUPTS();
/*
* Sleep at least 1 second after any error. A write error is likely
* to be repeated, and we don't want to be filling the error logs as
* fast as we can.
*/
pg_usleep(1000000L);
/*
* Close all open files after any error. This is helpful on Windows,
* where holding deleted files open causes various strange errors.
* It's not clear we need it elsewhere, but shouldn't hurt.
*/
smgrcloseall();
}
/* We can now handle ereport(ERROR) */
PG_exception_stack = &local_sigjmp_buf;
/*
* Unblock signals (they were blocked when the postmaster forked us)
*/
PG_SETMASK(&UnBlockSig);
/*
* Use the recovery target timeline ID during recovery
*/
if (RecoveryInProgress())
ThisTimeLineID = GetRecoveryTargetTLI();
/*
* Loop forever
*/
for (;;)
{
/*
* Emergency bailout if postmaster has died. This is to avoid the
* necessity for manual cleanup of all postmaster children.
*/
if (!PostmasterIsAlive())
exit(1);
if (got_SIGHUP)
{
got_SIGHUP = false;
ProcessConfigFile(PGC_SIGHUP);
/* update global shmem state for sync rep */
}
if (shutdown_requested)
{
/*
* From here on, elog(ERROR) should end with exit(1), not send
* control back to the sigsetjmp block above
*/
ExitOnAnyError = true;
/* Normal exit from the bgwriter is here */
proc_exit(0); /* done */
}
/*
* Do one cycle of dirty-buffer writing.
*/
BgBufferSync();
/* Nap for the configured time. */
BgWriterNap();
}
}