Create discrete persistent submit and get work threads per pool, thus allowing all submitworks belonging to the same pool to reuse the same curl handle, and all getworks to reuse their own handle.

Use separate handles for submission to not make getwork potentially delay share submission which is time critical.
This will allow much more reusing of persistent connections instead of opening new ones which can flood routers.
This mandated a rework of the extra longpoll support (for when pools are switched) and this is managed by restarting longpoll cleanly and waiting for a thread join.

cgminer.c

@@ -378,6 +378,9 @@ sharelog(const char*disposition, const struct work*work) {
 		applog(LOG_ERR, "sharelog fwrite error");
 }
 
+static void *submit_work_thread(void *userdata);
+static void *get_work_thread(void *userdata);
+
 static void add_pool(void)
 {
 	struct pool *pool;
@@ -395,6 +398,11 @@ static void add_pool(void)
 	}
 	/* Make sure the pool doesn't think we've been idle since time 0 */
 	pool->tv_idle.tv_sec = ~0UL;
+
+	if (unlikely(pthread_create(&pool->submit_thread, NULL, submit_work_thread, (void *)pool)))
+		quit(1, "Failed to create pool submit thread");
+	if (unlikely(pthread_create(&pool->getwork_thread, NULL, get_work_thread, (void *)pool)))
+		quit(1, "Failed to create pool getwork thread");
 }
 
 /* Pool variant of test and set */
@@ -1600,18 +1608,13 @@ static bool submit_upstream_work(const struct work *work)
 	bool rc = false;
 	int thr_id = work->thr_id;
 	struct cgpu_info *cgpu = thr_info[thr_id].cgpu;
-	CURL *curl = curl_easy_init();
 	struct pool *pool = work->pool;
+	CURL *curl = pool->submit_curl;
 	bool rolltime;
 	uint32_t *hash32;
 	char hashshow[64+1] = "";
 	bool isblock;
 
-	if (unlikely(!curl)) {
-		applog(LOG_ERR, "CURL initialisation failed");
-		return rc;
-	}
-
 #ifdef __BIG_ENDIAN__
         int swapcounter = 0;
         for (swapcounter = 0; swapcounter < 32; swapcounter++)
@@ -1745,7 +1748,6 @@ static bool submit_upstream_work(const struct work *work)
 out:
 	free(hexstr);
 out_nofree:
-	curl_easy_cleanup(curl);
 	return rc;
 }
 
@@ -1789,41 +1791,26 @@ static void get_benchmark_work(struct work *work)
 	memcpy(work, &bench_block, min_size);
 }
 
-static bool get_upstream_work(struct work *work, bool lagging)
+static bool get_upstream_work(struct work *work)
 {
-	bool rc = false, req_longpoll = false;
+	struct pool *pool = work->pool;
-	struct pool *pool;
+	CURL *curl = pool->getwork_curl;
 	json_t *val = NULL;
+	bool rc = false;
 	int retries = 0;
-	CURL *curl;
 	char *url;
 
-	curl = curl_easy_init();
-	if (unlikely(!curl)) {
-		applog(LOG_ERR, "CURL initialisation failed");
-		return rc;
-	}
-
-	pool = select_pool(lagging);
 	applog(LOG_DEBUG, "DBG: sending %s get RPC call: %s", pool->rpc_url, rpc_req);
 
 	url = pool->rpc_url;
 
-	/* If this is the current pool and supports longpoll but has not sent
-	 * a longpoll, send one now */
-	if (unlikely(want_longpoll && !pool->is_lp && pool == current_pool() &&
-		pool->hdr_path && !pool_tset(pool, &pool->lp_sent))) {
-			req_longpoll = true;
-			url = pool->lp_url;
-	}
-
 retry:
 	/* A single failure response here might be reported as a dead pool and
 	 * there may be temporary denied messages etc. falsely reporting
 	 * failure so retry a few times before giving up */
 	while (!val && retries++ < 3) {
 		val = json_rpc_call(curl, url, pool->rpc_userpass, rpc_req,
-			    false, req_longpoll, &work->rolltime, pool, false);
+			    false, false, &work->rolltime, pool, false);
 	}
 	if (unlikely(!val)) {
 		applog(LOG_DEBUG, "Failed json_rpc_call in get_upstream_work");
@@ -1838,13 +1825,12 @@ retry:
 		goto retry;
 	}
 	work->pool = pool;
-	work->longpoll = req_longpoll;
+	work->longpoll = false;
 	total_getworks++;
 	pool->getwork_requested++;
 
 	json_decref(val);
 out:
-	curl_easy_cleanup(curl);
 
 	return rc;
 }
@@ -2001,13 +1987,40 @@ static void sighandler(int __maybe_unused sig)
 	kill_work();
 }
 
+static void start_longpoll(void);
+static void stop_longpoll(void);
+
+/* One get work thread is created per pool, so as to use one curl handle for
+ * all getwork reqeusts from the same pool, minimising connections opened, but
+ * separate from the submit work curl handle to not delay share submissions due
+ * to getwork traffic */
 static void *get_work_thread(void *userdata)
 {
-	struct workio_cmd *wc = (struct workio_cmd *)userdata;
+	struct pool *pool = (struct pool *)userdata;
+	struct workio_cmd *wc;
+
+	pthread_detach(pthread_self());
+
+	/* getwork_q memory never freed */
+	pool->getwork_q = tq_new();
+	if (!pool->getwork_q)
+		quit(1, "Failed to tq_new in get_work_thread");
+
+	/* getwork_curl never cleared */
+	pool->getwork_curl = curl_easy_init();
+	if (unlikely(!pool->getwork_curl))
+		quit(1, "Failed to initialise pool getwork CURL");
+
+	while ((wc = tq_pop(pool->getwork_q, NULL)) != NULL) {
 		struct work *ret_work;
 		int failures = 0;
 
-	pthread_detach(pthread_self());
+		if (unlikely(want_longpoll && !pool->is_lp && pool == current_pool() &&
+			pool->hdr_path && !pool_tset(pool, &pool->lp_sent))) {
+				stop_longpoll();
+				start_longpoll();
+		}
+
 		ret_work = make_work();
 
 		if (wc->thr)
@@ -2015,13 +2028,15 @@ static void *get_work_thread(void *userdata)
 		else
 			ret_work->thr = NULL;
 
+		ret_work->pool = pool;
+
 		/* obtain new work from bitcoin via JSON-RPC */
-	while (!get_upstream_work(ret_work, wc->lagging)) {
+		while (!get_upstream_work(ret_work)) {
 			if (unlikely((opt_retries >= 0) && (++failures > opt_retries))) {
 				applog(LOG_ERR, "json_rpc_call failed, terminating workio thread");
 				free_work(ret_work);
 				kill_work();
-			goto out;
+				break;
 			}
 
 			/* pause, then restart work-request loop */
@@ -2040,21 +2055,17 @@ static void *get_work_thread(void *userdata)
 			kill_work();
 			free_work(ret_work);
 		}
-
-out:
 		workio_cmd_free(wc);
+	}
+
 	return NULL;
 }
 
 static bool workio_get_work(struct workio_cmd *wc)
 {
-	pthread_t get_thread;
+	struct pool *pool = select_pool(wc->lagging);
 
-	if (unlikely(pthread_create(&get_thread, NULL, get_work_thread, (void *)wc))) {
+	return tq_push(pool->getwork_q, wc);
-		applog(LOG_ERR, "Failed to create get_work_thread");
-		return false;
-	}
-	return true;
 }
 
 static bool stale_work(struct work *work, bool share)
@@ -2080,15 +2091,31 @@ static bool stale_work(struct work *work, bool share)
 	return false;
 }
 
+/* One submit work thread is created per pool, so as to use one curl handle
+ * for all submissions to the same pool, minimising connections opened, but
+ * separate from the getwork curl handle to not delay share submission due to
+ * getwork traffic */
 static void *submit_work_thread(void *userdata)
 {
-	struct workio_cmd *wc = (struct workio_cmd *)userdata;
+	struct pool *pool = (struct pool *)userdata;
-	struct work *work = wc->u.work;
+	struct workio_cmd *wc;
-	struct pool *pool = work->pool;
-	int failures = 0;
 
 	pthread_detach(pthread_self());
 
+	/* submit_q memory never freed */
+	pool->submit_q = tq_new();
+	if (!pool->submit_q )
+		quit(1, "Failed to tq_new in submit_work_thread");
+
+	/* submit_curl never cleared */
+	pool->submit_curl = curl_easy_init();
+	if (unlikely(!pool->submit_curl))
+		quit(1, "Failed to initialise pool submit CURL");
+
+	while ((wc = tq_pop(pool->submit_q, NULL)) != NULL) {
+		struct work *work = wc->u.work;
+		int failures = 0;
+
 		if (stale_work(work, true)) {
 			if (opt_submit_stale)
 				applog(LOG_NOTICE, "Stale share detected, submitting as user requested");
@@ -2099,7 +2126,8 @@ static void *submit_work_thread(void *userdata)
 				sharelog("discard", work);
 				total_stale++;
 				pool->stale_shares++;
-			goto out;
+				workio_cmd_free(wc);
+				continue;
 			}
 		}
 
@@ -2124,20 +2152,15 @@ static void *submit_work_thread(void *userdata)
 			fail_pause += opt_fail_pause;
 		}
 		fail_pause = opt_fail_pause;
-out:
 		workio_cmd_free(wc);
+	}
+
 	return NULL;
 }
 
 static bool workio_submit_work(struct workio_cmd *wc)
 {
-	pthread_t submit_thread;
+	return tq_push(wc->u.work->pool->submit_q, wc);
-
-	if (unlikely(pthread_create(&submit_thread, NULL, submit_work_thread, (void *)wc))) {
-		applog(LOG_ERR, "Failed to create submit_work_thread");
-		return false;
-	}
-	return true;
 }
 
 /* Find the pool that currently has the highest priority */
@@ -2904,9 +2927,6 @@ retry:
 }
 #endif
 
-static void start_longpoll(void);
-static void stop_longpoll(void);
-
 #ifdef HAVE_CURSES
 static void set_options(void)
 {
@@ -3880,7 +3900,6 @@ static void *longpoll_thread(void *userdata)
 	bool rolltime;
 
 	pthread_setcanceltype(PTHREAD_CANCEL_ASYNCHRONOUS, NULL);
-	pthread_detach(pthread_self());
 
 	curl = curl_easy_init();
 	if (unlikely(!curl)) {
@@ -3959,12 +3978,13 @@ out:
 	return NULL;
 }
 
-__maybe_unused
 static void stop_longpoll(void)
 {
 	struct thr_info *thr = &thr_info[longpoll_thr_id];
 
 	thr_info_cancel(thr);
+	if (have_longpoll)
+		pthread_join(thr->pth, NULL);
 	have_longpoll = false;
 	tq_freeze(thr->q);
 }

miner.h

@@ -627,6 +627,15 @@ struct pool {
 	char *rpc_user, *rpc_pass;
 
 	pthread_mutex_t pool_lock;
+
+	struct thread_q *submit_q;
+	struct thread_q *getwork_q;
+
+	pthread_t submit_thread;
+	pthread_t getwork_thread;
+
+	CURL *submit_curl;
+	CURL *getwork_curl;
 };
 
 struct work {