Subject: Re: logical replication launcher crash on buildfarm
To: Robert Haas <robertmhaas@gmail.com>, Andres Freund <andres@anarazel.de>
References: <20170316032833.2zxfnqe3sl3ktrbt@alap3.anarazel.de>
 <20170316034246.3uurliezczmka6cn@alap3.anarazel.de>
 <e56bda7f-9d29-6900-0f90-96be8219b930@2ndquadrant.com>
 <20170316085322.crffknkgee5s6air@alap3.anarazel.de>
 <368e64f6-ee9a-f09f-82b4-a33b61b28d36@2ndquadrant.com>
 <20170327165023.t224nralrspk6udc@alap3.anarazel.de>
 <CA+Tgmoa3C7=gVvmwvvKWT6+mkm=XGSDqF4bXu6=fohArafWHMQ@mail.gmail.com>
 <52d78669-0523-4e60-a263-841c96210256@2ndquadrant.com>
 <8941c9b1-f6e9-1056-39f2-7bc7336c4d00@2ndquadrant.com>
 <20170328020422.i4srzzgq4zbzu2ci@alap3.anarazel.de>
 <CA+TgmobBTZ_eDbTmx-pfUndZDKPBAcj+r_3CQjEHFkp99kE2Yg@mail.gmail.com>
Cc: "pgsql-hackers@postgresql.org" <pgsql-hackers@postgresql.org>,
 Petr Jelinek <petr@2ndquadrant.com>, Peter Eisentraut <peter_e@gmx.net>
From: Petr Jelinek <petr.jelinek@2ndquadrant.com>
Message-ID: <e3c35f3a-426d-b1ca-7fe5-9241c2717013@2ndquadrant.com>
Date: Tue, 28 Mar 2017 05:20:44 +0200
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:45.0) Gecko/20100101
 Thunderbird/45.7.0
MIME-Version: 1.0
In-Reply-To: 
 <CA+TgmobBTZ_eDbTmx-pfUndZDKPBAcj+r_3CQjEHFkp99kE2Yg@mail.gmail.com>
Content-Type: multipart/mixed;
 boundary="------------B9262B1C47B9ACB2C5A1A658"
Precedence: bulk
Sender: pgsql-hackers-owner@postgresql.org

This is a multi-part message in MIME format.
--------------B9262B1C47B9ACB2C5A1A658
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: 7bit

On 28/03/17 04:46, Robert Haas wrote:
> On Mon, Mar 27, 2017 at 10:04 PM, Andres Freund <andres@anarazel.de> wrote:
>>> Btw now that I look at the code, I guess we'll want to get rid of
>>> bgw_main completely in HEAD given that we can't guarantee it will be
>>> valid even for shared_preload_library libraries. For older branches I
>>> would leave things as they are in this regard as there don't seem to be
>>> any immediate issue for standard binaries.
>>
>> As long as you fix it so culicidae is happy (in 9.6) ;).  I think it's
>> fine to just introduce bgw_builtin_id or such, and leave the bgw_main
>> code in place in < HEAD.
> 
> I wasn't thinking of introducing bgw_builtin_id.  My idea was just
> along the lines of
> 
> if (bgw_library_name == NULL && bgw_function_name != NULL)
> {
>     if (strcmp(bgw_function_name, "ParallelQueryMain") == 0)
>        ParallelQueryMain(blah);
>     else if (strcmp(bgw_function_name, "LogicalReplicationMain") == 0)
>        LogicalReplicationMain(blah);
> }
> 
> I think something like that is certainly better for the back-branches,
> because it doesn't cause an ABI break.  But I think it would also be
> fine for master.
> 

I had something slightly more complex like the attached in mind.

-- 
  Petr Jelinek                  http://www.2ndQuadrant.com/
  PostgreSQL Development, 24x7 Support, Training & Services

--------------B9262B1C47B9ACB2C5A1A658
Content-Type: binary/octet-stream;
 name="0001-Remove-bgw_main-from-bgworker-and-always-use-library.patch"
Content-Transfer-Encoding: 7bit
Content-Disposition: attachment;
 filename*0="0001-Remove-bgw_main-from-bgworker-and-always-use-library.pa";
 filename*1="tch"

From d548ff5467a6880122b5807add3d3c1febd8022e Mon Sep 17 00:00:00 2001
From: Petr Jelinek <pjmodos@pjmodos.net>
Date: Tue, 28 Mar 2017 04:21:46 +0200
Subject: [PATCH] Remove bgw_main from bgworker and always use library/function
 for loading main entry point

---
 src/backend/access/transam/parallel.c      |  6 +--
 src/backend/postmaster/bgworker.c          | 81 +++++++++++++++++++-----------
 src/backend/replication/logical/launcher.c |  6 ++-
 src/include/access/parallel.h              |  2 +
 src/include/postmaster/bgworker.h          |  1 -
 src/test/modules/test_shm_mq/setup.c       |  1 -
 src/test/modules/worker_spi/worker_spi.c   |  4 +-
 7 files changed, 64 insertions(+), 37 deletions(-)

diff --git a/src/backend/access/transam/parallel.c b/src/backend/access/transam/parallel.c
index 3e0ee87..b3d3853 100644
--- a/src/backend/access/transam/parallel.c
+++ b/src/backend/access/transam/parallel.c
@@ -110,7 +110,6 @@ static dlist_head pcxt_list = DLIST_STATIC_INIT(pcxt_list);
 /* Private functions. */
 static void HandleParallelMessage(ParallelContext *pcxt, int i, StringInfo msg);
 static void ParallelExtensionTrampoline(dsm_segment *seg, shm_toc *toc);
-static void ParallelWorkerMain(Datum main_arg);
 static void WaitForParallelWorkersToExit(ParallelContext *pcxt);
 
 
@@ -458,7 +457,8 @@ LaunchParallelWorkers(ParallelContext *pcxt)
 		| BGWORKER_CLASS_PARALLEL;
 	worker.bgw_start_time = BgWorkerStart_ConsistentState;
 	worker.bgw_restart_time = BGW_NEVER_RESTART;
-	worker.bgw_main = ParallelWorkerMain;
+	sprintf(worker.bgw_library_name, "postgres");
+	sprintf(worker.bgw_function_name, "ParallelWorkerMain");
 	worker.bgw_main_arg = UInt32GetDatum(dsm_segment_handle(pcxt->seg));
 	worker.bgw_notify_pid = MyProcPid;
 	memset(&worker.bgw_extra, 0, BGW_EXTRALEN);
@@ -931,7 +931,7 @@ AtEOXact_Parallel(bool isCommit)
 /*
  * Main entrypoint for parallel workers.
  */
-static void
+void
 ParallelWorkerMain(Datum main_arg)
 {
 	dsm_segment *seg;
diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c
index 10e0f88..39a098c 100644
--- a/src/backend/postmaster/bgworker.c
+++ b/src/backend/postmaster/bgworker.c
@@ -15,12 +15,14 @@
 #include <unistd.h>
 
 #include "libpq/pqsignal.h"
+#include "access/parallel.h"
 #include "miscadmin.h"
 #include "pgstat.h"
 #include "port/atomics.h"
 #include "postmaster/bgworker_internals.h"
 #include "postmaster/postmaster.h"
 #include "replication/logicallauncher.h"
+#include "replication/logicalworker.h"
 #include "storage/dsm.h"
 #include "storage/ipc.h"
 #include "storage/latch.h"
@@ -109,14 +111,26 @@ struct BackgroundWorkerHandle
 static BackgroundWorkerArray *BackgroundWorkerData;
 
 /*
- * List of workers that are allowed to be started outside of
- * shared_preload_libraries.
+ * List of internal background workers. These are used for mapping the
+ * function name to actual function when building with EXEC_BACKEND and also
+ * to allow these to be loaded outside of shared_preload_libraries.
  */
-static const bgworker_main_type InternalBGWorkers[] = {
-	ApplyLauncherMain,
-	NULL
+typedef struct InternalBGWorkerMain
+{
+	char			   *bgw_function_name;
+	bgworker_main_type	bgw_main;
+} InternalBGWorkerMain;
+
+static const InternalBGWorkerMain InternalBGWorkers[] = {
+	{"ParallelWorkerMain", ParallelWorkerMain},
+	{"ApplyLauncherMain", ApplyLauncherMain},
+	{"ApplyWorkerMain", ApplyWorkerMain},
+	/* Dummy entry marking end of the array. */
+	{NULL, NULL}
 };
 
+static bgworker_main_type GetInternalBgWorkerMain(BackgroundWorker *worker);
+
 /*
  * Calculate shared memory needed.
  */
@@ -341,7 +355,6 @@ BackgroundWorkerStateChange(void)
 		rw->rw_worker.bgw_flags = slot->worker.bgw_flags;
 		rw->rw_worker.bgw_start_time = slot->worker.bgw_start_time;
 		rw->rw_worker.bgw_restart_time = slot->worker.bgw_restart_time;
-		rw->rw_worker.bgw_main = slot->worker.bgw_main;
 		rw->rw_worker.bgw_main_arg = slot->worker.bgw_main_arg;
 		memcpy(rw->rw_worker.bgw_extra, slot->worker.bgw_extra, BGW_EXTRALEN);
 
@@ -763,17 +776,14 @@ StartBackgroundWorker(void)
 	}
 
 	/*
-	 * If bgw_main is set, we use that value as the initial entrypoint.
-	 * However, if the library containing the entrypoint wasn't loaded at
-	 * postmaster startup time, passing it as a direct function pointer is not
-	 * possible.  To work around that, we allow callers for whom a function
-	 * pointer is not available to pass a library name (which will be loaded,
-	 * if necessary) and a function name (which will be looked up in the named
-	 * library).
+	 * For internal workers set the entry point to known function address.
+	 * Otherwise use the entry point specified by library name (which will
+	 * be loaded, if necessary) and a function name (which will be looked up
+	 * in the named library).
 	 */
-	if (worker->bgw_main != NULL)
-		entrypt = worker->bgw_main;
-	else
+	entrypt = GetInternalBgWorkerMain(worker);
+
+	if (entrypt == NULL)
 		entrypt = (bgworker_main_type)
 			load_external_function(worker->bgw_library_name,
 								   worker->bgw_function_name,
@@ -806,23 +816,13 @@ RegisterBackgroundWorker(BackgroundWorker *worker)
 {
 	RegisteredBgWorker *rw;
 	static int	numworkers = 0;
-	bool		internal = false;
-	int			i;
 
 	if (!IsUnderPostmaster)
 		ereport(DEBUG1,
 		 (errmsg("registering background worker \"%s\"", worker->bgw_name)));
 
-	for (i = 0; InternalBGWorkers[i]; i++)
-	{
-		if (worker->bgw_main == InternalBGWorkers[i])
-		{
-			internal = true;
-			break;
-		}
-	}
-
-	if (!process_shared_preload_libraries_in_progress && !internal)
+	if (!process_shared_preload_libraries_in_progress &&
+		GetInternalBgWorkerMain(worker) == NULL)
 	{
 		if (!IsUnderPostmaster)
 			ereport(LOG,
@@ -1152,3 +1152,28 @@ TerminateBackgroundWorker(BackgroundWorkerHandle *handle)
 	if (signal_postmaster)
 		SendPostmasterSignal(PMSIGNAL_BACKGROUND_WORKER_CHANGE);
 }
+
+/*
+ * Search the known internal worker array and return its main function
+ * pointer if found.
+ *
+ * Returns NULL if not known internal worker.
+ */
+static bgworker_main_type
+GetInternalBgWorkerMain(BackgroundWorker *worker)
+{
+	int i;
+
+	/* Internal workers always have to use postgres as library name. */
+	if (strncmp(worker->bgw_library_name, "postgres", 8) != 0)
+		return NULL;
+
+	for (i = 0; InternalBGWorkers[i].bgw_function_name; i++)
+	{
+		if (strncmp(InternalBGWorkers[i].bgw_function_name,
+					worker->bgw_function_name, BGW_MAXLEN) == 0)
+			return InternalBGWorkers[i].bgw_main;
+	}
+
+	return NULL;
+}
diff --git a/src/backend/replication/logical/launcher.c b/src/backend/replication/logical/launcher.c
index 0d3ec27..226a168 100644
--- a/src/backend/replication/logical/launcher.c
+++ b/src/backend/replication/logical/launcher.c
@@ -295,7 +295,8 @@ logicalrep_worker_launch(Oid dbid, Oid subid, const char *subname, Oid userid,
 	bgw.bgw_flags =	BGWORKER_SHMEM_ACCESS |
 		BGWORKER_BACKEND_DATABASE_CONNECTION;
 	bgw.bgw_start_time = BgWorkerStart_RecoveryFinished;
-	bgw.bgw_main = ApplyWorkerMain;
+	snprintf(bgw.bgw_library_name, BGW_MAXLEN, "postgres");
+	snprintf(bgw.bgw_function_name, BGW_MAXLEN, "ApplyWorkerMain");
 	if (OidIsValid(relid))
 		snprintf(bgw.bgw_name, BGW_MAXLEN,
 				 "logical replication worker for subscription %u sync %u", subid, relid);
@@ -553,7 +554,8 @@ ApplyLauncherRegister(void)
 	bgw.bgw_flags =	BGWORKER_SHMEM_ACCESS |
 		BGWORKER_BACKEND_DATABASE_CONNECTION;
 	bgw.bgw_start_time = BgWorkerStart_RecoveryFinished;
-	bgw.bgw_main = ApplyLauncherMain;
+	snprintf(bgw.bgw_library_name, BGW_MAXLEN, "postgres");
+	snprintf(bgw.bgw_function_name, BGW_MAXLEN, "ApplyLauncherMain");
 	snprintf(bgw.bgw_name, BGW_MAXLEN,
 			 "logical replication launcher");
 	bgw.bgw_restart_time = 5;
diff --git a/src/include/access/parallel.h b/src/include/access/parallel.h
index 96ce480..5065a38 100644
--- a/src/include/access/parallel.h
+++ b/src/include/access/parallel.h
@@ -67,4 +67,6 @@ extern void AtEOXact_Parallel(bool isCommit);
 extern void AtEOSubXact_Parallel(bool isCommit, SubTransactionId mySubId);
 extern void ParallelWorkerReportLastRecEnd(XLogRecPtr last_xlog_end);
 
+extern void ParallelWorkerMain(Datum main_arg);
+
 #endif   /* PARALLEL_H */
diff --git a/src/include/postmaster/bgworker.h b/src/include/postmaster/bgworker.h
index 128e92c..7cc4c89 100644
--- a/src/include/postmaster/bgworker.h
+++ b/src/include/postmaster/bgworker.h
@@ -91,7 +91,6 @@ typedef struct BackgroundWorker
 	int			bgw_flags;
 	BgWorkerStartTime bgw_start_time;
 	int			bgw_restart_time;		/* in seconds, or BGW_NEVER_RESTART */
-	bgworker_main_type bgw_main;
 	char		bgw_library_name[BGW_MAXLEN];	/* only if bgw_main is NULL */
 	char		bgw_function_name[BGW_MAXLEN];	/* only if bgw_main is NULL */
 	Datum		bgw_main_arg;
diff --git a/src/test/modules/test_shm_mq/setup.c b/src/test/modules/test_shm_mq/setup.c
index e3c024e..319a67f 100644
--- a/src/test/modules/test_shm_mq/setup.c
+++ b/src/test/modules/test_shm_mq/setup.c
@@ -216,7 +216,6 @@ setup_background_workers(int nworkers, dsm_segment *seg)
 	worker.bgw_flags = BGWORKER_SHMEM_ACCESS;
 	worker.bgw_start_time = BgWorkerStart_ConsistentState;
 	worker.bgw_restart_time = BGW_NEVER_RESTART;
-	worker.bgw_main = NULL;		/* new worker might not have library loaded */
 	sprintf(worker.bgw_library_name, "test_shm_mq");
 	sprintf(worker.bgw_function_name, "test_shm_mq_main");
 	snprintf(worker.bgw_name, BGW_MAXLEN, "test_shm_mq");
diff --git a/src/test/modules/worker_spi/worker_spi.c b/src/test/modules/worker_spi/worker_spi.c
index 72ab846..421ec76 100644
--- a/src/test/modules/worker_spi/worker_spi.c
+++ b/src/test/modules/worker_spi/worker_spi.c
@@ -347,7 +347,8 @@ _PG_init(void)
 		BGWORKER_BACKEND_DATABASE_CONNECTION;
 	worker.bgw_start_time = BgWorkerStart_RecoveryFinished;
 	worker.bgw_restart_time = BGW_NEVER_RESTART;
-	worker.bgw_main = worker_spi_main;
+	sprintf(worker.bgw_library_name, "worker_spi");
+	sprintf(worker.bgw_function_name, "worker_spi_main");
 	worker.bgw_notify_pid = 0;
 
 	/*
@@ -378,7 +379,6 @@ worker_spi_launch(PG_FUNCTION_ARGS)
 		BGWORKER_BACKEND_DATABASE_CONNECTION;
 	worker.bgw_start_time = BgWorkerStart_RecoveryFinished;
 	worker.bgw_restart_time = BGW_NEVER_RESTART;
-	worker.bgw_main = NULL;		/* new worker might not have library loaded */
 	sprintf(worker.bgw_library_name, "worker_spi");
 	sprintf(worker.bgw_function_name, "worker_spi_main");
 	snprintf(worker.bgw_name, BGW_MAXLEN, "worker %d", i);
-- 
2.7.4


--------------B9262B1C47B9ACB2C5A1A658
Content-Type: text/plain
Content-Disposition: inline
Content-Transfer-Encoding: 8bit
MIME-Version: 1.0


-- 
Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers

--------------B9262B1C47B9ACB2C5A1A658--