Merge remote-tracking branch 'upstream/master' into ztex-120417

NEWS

@@ -1,3 +1,45 @@
+Version 2.4.0 - May 3, 2012
+
+- Only show longpoll warning once when it has failed.
+- Convert hashes to an unsigned long long as well.
+- Detect pools that have issues represented by endless rejected shares and
+disable them, with a parameter to optionally disable this feature.
+- Bugfix: Use a 64-bit type for hashes_done (miner_thread) since it can overflow
+32-bit on some FPGAs
+- Implement an older header fix for a label existing before the pthread_cleanup
+macro.
+- Limit the number of curls we recruit on communication failures and with
+delaynet enabled to 5 by maintaining a per-pool curl count, and using a pthread
+conditional that wakes up when one is returned to the ring buffer.
+- Generalise add_pool() functions since they're repeated in add_pool_details.
+- Bugfix: Return failure, rather than quit, if BFwrite fails
+- Disable failing devices such that the user can attempt to re-enable them
+- Bugfix: thread_shutdown shouldn't try to free the device, since it's needed
+afterward
+- API bool's and 1TBS fixes
+- Icarus - minimise code delays and name timer variables
+- api.c V1.9 add 'restart' + redesign 'quit' so thread exits cleanly
+- api.c bug - remove extra ']'s in notify command
+- Increase pool watch interval to 30 seconds.
+- Reap curls that are unused for over a minute. This allows connections to be
+closed, thereby allowing the number of curl handles to always be the minimum
+necessary to not delay networking.
+- Use the ringbuffer of curls from the same pool for submit as well as getwork
+threads. Since the curl handles were already connected to the same pool and are
+immediately available, share submission will not be delayed by getworks.
+- Implement a scaleable networking framework designed to cope with any sized
+network requirements, yet minimise the number of connections being reopened. Do
+this by create a ring buffer linked list of curl handles to be used by getwork,
+recruiting extra handles when none is immediately available.
+- There is no need for the submit and getwork curls to be tied to the pool
+struct.
+- Do not recruit extra connection threads if there have been connection errors
+to the pool in question.
+- We should not retry submitting shares indefinitely or we may end up with a
+huge backlog during network outages, so discard stale shares if we failed to
+submit them and they've become stale in the interim.
+
+
 Version 2.3.6 - April 29, 2012
 
 - Shorten stale share messages slightly.

README

@@ -26,7 +26,7 @@ IRC Channel:
 
 irc://irc.freenode.net/cgminer
 
-License: GPLv2.  See COPYING for details.
+License: GPLv3.  See COPYING for details.
 
 READ EXECUTIVE SUMMARY BELOW FOR FIRST TIME USERS!
 
@@ -133,6 +133,7 @@ Options for both config file and command line:
 --log|-l <arg>      Interval in seconds between log output (default: 5)
 --monitor|-m <arg>  Use custom pipe cmd for output messages
 --net-delay         Impose small delays in networking to not overload slow routers
+--no-pool-disable   Do not automatically disable pools that continually reject shares
 --no-submit-stale   Don't submit shares if they are detected as stale
 --pass|-p <arg>     Password for bitcoin JSON-RPC server
 --per-device-stats  Force verbose mode and output per-device statistics
@@ -285,7 +286,6 @@ Current pool management strategy: Failover
 
 S gives you:
 
-[L]ongpoll: On
 [Q]ueue: 1
 [S]cantime: 60
 [E]xpiry: 120
@@ -647,7 +647,7 @@ The STATUS section is:
    This defaults to the cgminer version but is the value of --api-description
    if it was specified at runtime.
 
-For API version 1.8:
+For API version 1.9:
 
 The list of requests - a (*) means it requires privileged access - and replies are:
 
@@ -768,7 +768,7 @@ The list of requests - a (*) means it requires privileged access - and replies a
                               stating success or failure saving the cgminer config
                               to filename
 
- quit (*)      none           There is no status section but just a single "BYE|"
+ quit (*)      none           There is no status section but just a single "BYE"
                               reply before cgminer quits
 
  notify        NOTIFY         The last status and history count of each devices problem
@@ -798,6 +798,9 @@ The list of requests - a (*) means it requires privileged access - and replies a
                               by the 'devs' command
                               e.g. DEVDETAILS=0,Name=GPU,ID=0,Driver=opencl,...|
 
+ restart (*)   none           There is no status section but just a single "RESTART"
+                              reply before cgminer restarts
+
 When you enable, disable or restart a GPU or PGA, you will also get Thread messages
 in the cgminer status window
 
@@ -1005,7 +1008,7 @@ no longer under active development and will not be supported unless someone
 steps up to help maintain it. No binary builds supporting CPU mining will be
 released but CPU mining can be built into cgminer when it is compiled.
 
-Q: I upgraded cgminer version and mu hashrate suddenly dropped!
+Q: I upgraded cgminer version and my hashrate suddenly dropped!
 A: No, you upgraded your SDK version unwittingly between upgrades of cgminer
 and that caused  your hashrate to drop. See the next question.
 

api.c

@@ -158,7 +158,7 @@ static const char SEPARATOR = '|';
 #define SEPSTR "|"
 static const char GPUSEP = ',';
 
-static const char *APIVERSION = "1.8";
+static const char *APIVERSION = "1.9";
 static const char *DEAD = "Dead";
 static const char *SICK = "Sick";
 static const char *NOSTART = "NoStart";
@@ -229,6 +229,7 @@ static const char *OSINFO =
 #define _NOTIFY		"NOTIFY"
 #define _DEVDETAILS	"DEVDETAILS"
 #define _BYE		"BYE"
+#define _RESTART	"RESTART"
 
 static const char ISJSON = '{';
 #define JSON0		"{"
@@ -260,6 +261,7 @@ static const char ISJSON = '{';
 #define JSON_NOTIFY	JSON1 _NOTIFY JSON2
 #define JSON_DEVDETAILS	JSON1 _DEVDETAILS JSON2
 #define JSON_BYE	JSON1 _BYE JSON1
+#define JSON_RESTART	JSON1 _RESTART JSON1
 #define JSON_CLOSE	JSON3
 #define JSON_END	JSON4
 
@@ -476,9 +478,15 @@ struct CODES {
 };
 
 static int my_thr_id = 0;
-static int bye = 0;
+static bool bye;
 static bool ping = true;
 
+// Used to control quit restart access to shutdown variables
+static pthread_mutex_t quit_restart_lock;
+
+static bool do_a_quit;
+static bool do_a_restart;
+
 static time_t when = 0;	// when the request occurred
 
 struct IP4ACCESS {
@@ -612,7 +620,7 @@ static int pgadevice(int pgaid)
 }
 #endif
 
-// All replies (except BYE) start with a message
+// All replies (except BYE and RESTART) start with a message
 //  thus for JSON, message() inserts JSON_START at the front
 //  and send_result() adds JSON_END at the end
 static char *message(int messageid, int paramid, char *param2, bool isjson)
@@ -1811,22 +1819,26 @@ static void gpuvddc(__maybe_unused SOCKETTYPE c, __maybe_unused char *param, boo
 #endif
 }
 
-static void send_result(SOCKETTYPE c, bool isjson);
-
-void doquit(SOCKETTYPE c, __maybe_unused char *param, bool isjson)
+void doquit(__maybe_unused SOCKETTYPE c, __maybe_unused char *param, bool isjson)
 {
 	if (isjson)
 		strcpy(io_buffer, JSON_START JSON_BYE);
 	else
 		strcpy(io_buffer, _BYE);
 
-	send_result(c, isjson);
-	*io_buffer = '\0';
-	bye = 1;
+	bye = true;
+	do_a_quit = true;
+}
 
-        PTH(&thr_info[my_thr_id]) = 0L;
+void dorestart(__maybe_unused SOCKETTYPE c, __maybe_unused char *param, bool isjson)
+{
+	if (isjson)
+		strcpy(io_buffer, JSON_START JSON_RESTART);
+	else
+		strcpy(io_buffer, _RESTART);
 
-	kill_work();
+	bye = true;
+	do_a_restart = true;
 }
 
 void privileged(__maybe_unused SOCKETTYPE c, __maybe_unused char *param, bool isjson)
@@ -1875,7 +1887,7 @@ void notifystatus(int device, struct cgpu_info *cgpu, bool isjson)
 	// ALL counters (and only counters) must start the name with a '*'
 	// Simplifies future external support for adding new counters
 	sprintf(buf, isjson
-		? "%s{\"NOTIFY\":%d,\"Name\":\"%s\",\"ID\":%d,\"Last Well\":%lu,\"Last Not Well\":%lu,\"Reason Not Well\":\"%s\",\"*Thread Fail Init\":%d,\"*Thread Zero Hash\":%d,\"*Thread Fail Queue\":%d,\"*Dev Sick Idle 60s\":%d,\"*Dev Dead Idle 600s\":%d,\"*Dev Nostart\":%d,\"*Dev Over Heat\":%d,\"*Dev Thermal Cutoff\":%d}" JSON_CLOSE
+		? "%s{\"NOTIFY\":%d,\"Name\":\"%s\",\"ID\":%d,\"Last Well\":%lu,\"Last Not Well\":%lu,\"Reason Not Well\":\"%s\",\"*Thread Fail Init\":%d,\"*Thread Zero Hash\":%d,\"*Thread Fail Queue\":%d,\"*Dev Sick Idle 60s\":%d,\"*Dev Dead Idle 600s\":%d,\"*Dev Nostart\":%d,\"*Dev Over Heat\":%d,\"*Dev Thermal Cutoff\":%d}"
 		: "%sNOTIFY=%d,Name=%s,ID=%d,Last Well=%lu,Last Not Well=%lu,Reason Not Well=%s,*Thread Fail Init=%d,*Thread Zero Hash=%d,*Thread Fail Queue=%d,*Dev Sick Idle 60s=%d,*Dev Dead Idle 600s=%d,*Dev Nostart=%d,*Dev Over Heat=%d,*Dev Thermal Cutoff=%d" SEPSTR,
 		(isjson && (device > 0)) ? COMMA : BLANK,
 		device, cgpu->api->name, cgpu->device_id,
@@ -2017,6 +2029,7 @@ struct CMDS {
 	{ "privileged",		privileged,	true },
 	{ "notify",		notify,		false },
 	{ "devdetails",		devdetails,	false },
+	{ "restart",		dorestart,	true },
 	{ NULL,			NULL,		false }
 };
 
@@ -2041,12 +2054,13 @@ static void send_result(SOCKETTYPE c, bool isjson)
 		else
 			applog(LOG_DEBUG, "API: sent %d", n);
 	}
-
 }
 
 static void tidyup(__maybe_unused void *arg)
 {
-	bye = 1;
+	mutex_lock(&quit_restart_lock);
+
+	bye = true;
 
 	if (sock != INVSOCK) {
 		shutdown(sock, SHUT_RDWR);
@@ -2068,6 +2082,8 @@ static void tidyup(__maybe_unused void *arg)
 		free(io_buffer);
 		io_buffer = NULL;
 	}
+
+	mutex_unlock(&quit_restart_lock);
 }
 
 /*
@@ -2172,8 +2188,37 @@ popipo:
 	free(buf);
 }
 
+static void *quit_thread(__maybe_unused void *userdata)
+{
+	// allow thread creator to finish whatever it's doing
+	mutex_lock(&quit_restart_lock);
+	mutex_unlock(&quit_restart_lock);
+
+	if (opt_debug)
+		applog(LOG_DEBUG, "API: killing cgminer");
+
+	kill_work();
+
+	return NULL;
+}
+
+static void *restart_thread(__maybe_unused void *userdata)
+{
+	// allow thread creator to finish whatever it's doing
+	mutex_lock(&quit_restart_lock);
+	mutex_unlock(&quit_restart_lock);
+
+	if (opt_debug)
+		applog(LOG_DEBUG, "API: restarting cgminer");
+
+	app_restart();
+
+	return NULL;
+}
+
 void api(int api_thr_id)
 {
+	struct thr_info bye_thr;
 	char buf[BUFSIZ];
 	char param_buf[BUFSIZ];
 	const char *localaddr = "127.0.0.1";
@@ -2197,6 +2242,8 @@ void api(int api_thr_id)
 	bool did;
 	int i;
 
+	mutex_init(&quit_restart_lock);
+
 	pthread_cleanup_push(tidyup, NULL);
 	my_thr_id = api_thr_id;
 
@@ -2276,7 +2323,7 @@ void api(int api_thr_id)
 	io_buffer = malloc(MYBUFSIZ+1);
 	msg_buffer = malloc(MYBUFSIZ+1);
 
-	while (bye == 0) {
+	while (!bye) {
 		clisiz = sizeof(cli);
 		if (SOCKETFAIL(c = accept(sock, (struct sockaddr *)(&cli), &clisiz))) {
 			applog(LOG_ERR, "API failed (%s)%s", SOCKERRMSG, UNAVAILABLE);
@@ -2407,5 +2454,31 @@ void api(int api_thr_id)
 		CLOSESOCKET(c);
 	}
 die:
+	/* Blank line fix for older compilers since pthread_cleanup_pop is a
+	 * macro that gets confused by a label existing immediately before it
+	 */
+	;
 	pthread_cleanup_pop(true);
+
+	if (opt_debug)
+		applog(LOG_DEBUG, "API: terminating due to: %s",
+				do_a_quit ? "QUIT" : (do_a_restart ? "RESTART" : (bye ? "BYE" : "UNKNOWN!")));
+
+	mutex_lock(&quit_restart_lock);
+
+	if (do_a_restart) {
+		if (thr_info_create(&bye_thr, NULL, restart_thread, &bye_thr)) {
+			mutex_unlock(&quit_restart_lock);
+			quit(1, "API failed to initiate a restart - aborting");
+		}
+		pthread_detach(bye_thr.pth);
+	} else if (do_a_quit) {
+		if (thr_info_create(&bye_thr, NULL, quit_thread, &bye_thr)) {
+			mutex_unlock(&quit_restart_lock);
+			quit(1, "API failed to initiate a clean quit - aborting");
+		}
+		pthread_detach(bye_thr.pth);
+	}
+
+	mutex_unlock(&quit_restart_lock);
 }

cgminer.c

@@ -135,6 +135,7 @@ int opt_api_port = 4028;
 bool opt_api_listen = false;
 bool opt_api_network = false;
 bool opt_delaynet = false;
+bool opt_disable_pool = true;
 
 char *opt_kernel_path;
 char *cgminer_path;
@@ -378,10 +379,8 @@ static void 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)
+/* Return value is ignored if not called from add_pool_details */
+static struct pool *add_pool(void)
 {
 	struct pool *pool;
 
@@ -392,13 +391,14 @@ static void add_pool(void)
 	pools[total_pools++] = pool;
 	if (unlikely(pthread_mutex_init(&pool->pool_lock, NULL)))
 		quit(1, "Failed to pthread_mutex_init in add_pool");
+	if (unlikely(pthread_cond_init(&pool->cr_cond, NULL)))
+		quit(1, "Failed to pthread_cond_init in add_pool");
+	INIT_LIST_HEAD(&pool->curlring);
+
 	/* 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");
+	return pool;
 }
 
 /* Pool variant of test and set */
@@ -836,6 +836,9 @@ static struct opt_table opt_config_table[] = {
 			opt_hidden
 #endif
 	),
+	OPT_WITHOUT_ARG("--no-pool-disable",
+			opt_set_invbool, &opt_disable_pool,
+			"Do not automatically disable pools that continually reject shares"),
 	OPT_WITHOUT_ARG("--no-restart",
 			opt_set_invbool, &opt_restart,
 #ifdef HAVE_OPENCL
@@ -1654,6 +1657,7 @@ static bool submit_upstream_work(const struct work *work, CURL *curl)
 		cgpu->accepted++;
 		total_accepted++;
 		pool->accepted++;
+		pool->seq_rejects = 0;
 		cgpu->last_share_pool = pool->pool_no;
 		cgpu->last_share_pool_time = time(NULL);
 		applog(LOG_DEBUG, "PROOF OF WORK RESULT: true (yay!!!)");
@@ -1675,6 +1679,7 @@ static bool submit_upstream_work(const struct work *work, CURL *curl)
 		cgpu->rejected++;
 		total_rejected++;
 		pool->rejected++;
+		pool->seq_rejects++;
 		applog(LOG_DEBUG, "PROOF OF WORK RESULT: false (booooo)");
 		if (!QUIET) {
 			char where[17];
@@ -1705,6 +1710,22 @@ static bool submit_upstream_work(const struct work *work, CURL *curl)
 			       hashshow, cgpu->api->name, cgpu->device_id, where, reason);
 			sharelog(disposition, work);
 		}
+
+		/* Once we have more than a nominal amount of sequential rejects,
+		 * at least 10 and more than the current utility rate per minute,
+		 * disable the pool because some pool error is likely to have
+		 * ensued. */
+		if (pool->seq_rejects > 10 && opt_disable_pool && total_pools > 1) {
+			double utility = total_accepted / ( total_secs ? total_secs : 1 ) * 60;
+
+			if (pool->seq_rejects > utility) {
+				applog(LOG_WARNING, "Pool %d rejected %d sequential shares, disabling!",
+				       pool->pool_no, pool->seq_rejects);
+				pool->enabled = false;
+				if (pool == current_pool())
+					switch_pools(NULL);
+			}
+		}
 	}
 
 	cgpu->utility = cgpu->accepted / ( total_secs ? total_secs : 1 ) * 60;
@@ -1957,76 +1978,59 @@ static void sighandler(int __maybe_unused sig)
 	kill_work();
 }
 
-/* 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)
+/* Called with pool_lock held. Recruit an extra curl if none are available for
+ * this pool. */
+static void recruit_curl(struct pool *pool)
 {
-	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");
+	struct curl_ent *ce = calloc(sizeof(struct curl_ent), 1);
 
-	/* getwork_curl never cleared */
-	pool->getwork_curl = curl_easy_init();
-	if (unlikely(!pool->getwork_curl))
-		quit(1, "Failed to initialise pool getwork CURL");
+	ce->curl = curl_easy_init();
+	if (unlikely(!ce->curl || !ce))
+		quit(1, "Failed to init in recruit_curl");
 
-	while ((wc = tq_pop(pool->getwork_q, NULL)) != NULL) {
-		struct work *ret_work;
-		int failures = 0;
+	list_add(&ce->node, &pool->curlring);
+	pool->curls++;
+	applog(LOG_DEBUG, "Recruited curl %d for pool %d", pool->curls, pool->pool_no);
+}
 
-		ret_work = make_work();
+/* Grab an available curl if there is one. If not, then recruit extra curls
+ * unless we are in a submit_fail situation, or we have opt_delaynet enabled
+ * and there are already 5 curls in circulation */
+static struct curl_ent *pop_curl_entry(struct pool *pool)
+{
+	struct curl_ent *ce;
 
-		if (wc->thr)
-			ret_work->thr = wc->thr;
+	mutex_lock(&pool->pool_lock);
+	if (!pool->curls)
+		recruit_curl(pool);
+	else if (list_empty(&pool->curlring)) {
+		if ((pool->submit_fail || opt_delaynet) && pool->curls > 4)
+			pthread_cond_wait(&pool->cr_cond, &pool->pool_lock);
 		else
-			ret_work->thr = NULL;
-
-		ret_work->pool = pool;
-
-		/* obtain new work from bitcoin via JSON-RPC */
-		while (!get_upstream_work(ret_work, pool->getwork_curl)) {
-			if (unlikely((opt_retries >= 0) && (++failures > opt_retries))) {
-				applog(LOG_ERR, "json_rpc_call failed, terminating workio thread");
-				free_work(ret_work);
-				kill_work();
-				break;
+			recruit_curl(pool);
 	}
+	ce = list_entry(pool->curlring.next, struct curl_ent, node);
+	list_del(&ce->node);
+	mutex_unlock(&pool->pool_lock);
 
-			/* pause, then restart work-request loop */
-			applog(LOG_DEBUG, "json_rpc_call failed on get work, retry after %d seconds",
-				fail_pause);
-			sleep(fail_pause);
-			fail_pause += opt_fail_pause;
-		}
-		fail_pause = opt_fail_pause;
-
-		applog(LOG_DEBUG, "Pushing work to requesting thread");
-
-		/* send work to requesting thread */
-		if (unlikely(!tq_push(thr_info[stage_thr_id].q, ret_work))) {
-			applog(LOG_ERR, "Failed to tq_push work in workio_get_work");
-			kill_work();
-			free_work(ret_work);
-		}
-		workio_cmd_free(wc);
+	return ce;
 }
 
-	return NULL;
+static void push_curl_entry(struct curl_ent *ce, struct pool *pool)
+{
+	mutex_lock(&pool->pool_lock);
+	list_add_tail(&ce->node, &pool->curlring);
+	gettimeofday(&ce->tv, NULL);
+	pthread_cond_signal(&pool->cr_cond);
+	mutex_unlock(&pool->pool_lock);
 }
 
-static void *get_extra_work(void *userdata)
+static void *get_work_thread(void *userdata)
 {
 	struct workio_cmd *wc = (struct workio_cmd *)userdata;
-	struct work *ret_work = make_work();;
-	CURL *curl = curl_easy_init();
+	struct work *ret_work = make_work();
+	struct curl_ent *ce;
+	struct pool *pool;
 	int failures = 0;
 
 	pthread_detach(pthread_self());
@@ -2038,10 +2042,11 @@ static void *get_extra_work(void *userdata)
 	else
 		ret_work->thr = NULL;
 
-	ret_work->pool = select_pool(wc->lagging);
+	pool = ret_work->pool = select_pool(wc->lagging);
+	ce = pop_curl_entry(pool);
 
 	/* obtain new work from bitcoin via JSON-RPC */
-	while (!get_upstream_work(ret_work, curl)) {
+	while (!get_upstream_work(ret_work, ce->curl)) {
 		if (unlikely((opt_retries >= 0) && (++failures > opt_retries))) {
 			applog(LOG_ERR, "json_rpc_call failed, terminating workio thread");
 			free_work(ret_work);
@@ -2068,7 +2073,7 @@ static void *get_extra_work(void *userdata)
 
 out:
 	workio_cmd_free(wc);
-	curl_easy_cleanup(curl);
+	push_curl_entry(ce, pool);
 	return NULL;
 }
 
@@ -2077,13 +2082,9 @@ out:
  * requests */
 static bool workio_get_work(struct workio_cmd *wc)
 {
-	struct pool *pool = select_pool(wc->lagging);
 	pthread_t get_thread;
 
-	if (list_empty(&pool->getwork_q->q))
-		return tq_push(pool->getwork_q, wc);
-
-	if (unlikely(pthread_create(&get_thread, NULL, get_extra_work, (void *)wc))) {
+	if (unlikely(pthread_create(&get_thread, NULL, get_work_thread, (void *)wc))) {
 		applog(LOG_ERR, "Failed to create get_work_thread");
 		return false;
 	}
@@ -2113,82 +2114,13 @@ 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 pool *pool = (struct pool *)userdata;
-	struct workio_cmd *wc;
-
-	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 (pool->submit_old)
-				applog(LOG_NOTICE, "Stale share, submitting as pool %d requested",
-				       pool->pool_no);
-			else if (opt_submit_stale)
-				applog(LOG_NOTICE, "Stale share from pool %d, submitting as user requested",
-					pool->pool_no);
-			else {
-				applog(LOG_NOTICE, "Stale share from pool %d, discarding",
-					pool->pool_no);
-				sharelog("discard", work);
-				total_stale++;
-				pool->stale_shares++;
-				workio_cmd_free(wc);
-				continue;
-			}
-		}
-
-		/* submit solution to bitcoin via JSON-RPC */
-		while (!submit_upstream_work(work, pool->submit_curl)) {
-			if (!opt_submit_stale && stale_work(work, true) && !pool->submit_old) {
-				applog(LOG_NOTICE, "Stale share detected on submit retry, discarding");
-				total_stale++;
-				pool->stale_shares++;
-				break;
-			}
-			if (unlikely((opt_retries >= 0) && (++failures > opt_retries))) {
-				applog(LOG_ERR, "Failed %d retries ...terminating workio thread", opt_retries);
-				kill_work();
-				break;
-			}
-
-			/* pause, then restart work-request loop */
-			applog(LOG_INFO, "json_rpc_call failed on submit_work, retry after %d seconds",
-				fail_pause);
-			sleep(fail_pause);
-			fail_pause += opt_fail_pause;
-		}
-		fail_pause = opt_fail_pause;
-		workio_cmd_free(wc);
-	}
-
-	return NULL;
-}
-
-static void *submit_extra_work(void *userdata)
+static void *submit_work_thread(void *userdata)
 {
 	struct workio_cmd *wc = (struct workio_cmd *)userdata;
 	struct work *work = wc->u.work;
 	struct pool *pool = work->pool;
-	CURL *curl = curl_easy_init();
+	struct curl_ent *ce;
 	int failures = 0;
 
 	pthread_detach(pthread_self());
@@ -2209,10 +2141,11 @@ static void *submit_extra_work(void *userdata)
 		}
 	}
 
+	ce = pop_curl_entry(pool);
 	/* submit solution to bitcoin via JSON-RPC */
-	while (!submit_upstream_work(work, curl)) {
-		if (!opt_submit_stale && stale_work(work, true)) {
-			applog(LOG_NOTICE, "Stale share detected, discarding");
+	while (!submit_upstream_work(work, ce->curl)) {
+		if (stale_work(work, true)) {
+			applog(LOG_NOTICE, "Share became stale while retrying submit, discarding");
 			total_stale++;
 			pool->stale_shares++;
 			break;
@@ -2230,9 +2163,9 @@ static void *submit_extra_work(void *userdata)
 		fail_pause += opt_fail_pause;
 	}
 	fail_pause = opt_fail_pause;
+	push_curl_entry(ce, pool);
 out:
 	workio_cmd_free(wc);
-	curl_easy_cleanup(curl);
 	return NULL;
 }
 
@@ -2244,10 +2177,7 @@ static bool workio_submit_work(struct workio_cmd *wc)
 {
 	pthread_t submit_thread;
 
-	if (list_empty(&wc->u.work->pool->submit_q->q))
-		return tq_push(wc->u.work->pool->submit_q, wc);
-
-	if (unlikely(pthread_create(&submit_thread, NULL, submit_extra_work, (void *)wc))) {
+	if (unlikely(pthread_create(&submit_thread, NULL, submit_work_thread, (void *)wc))) {
 		applog(LOG_ERR, "Failed to create submit_work_thread");
 		return false;
 	}
@@ -3233,7 +3163,7 @@ static inline void thread_reportout(struct thr_info *thr)
 }
 
 static void hashmeter(int thr_id, struct timeval *diff,
-		      unsigned long hashes_done)
+		      unsigned long long hashes_done)
 {
 	struct timeval temp_tv_end, total_diff;
 	double secs;
@@ -3263,7 +3193,7 @@ static void hashmeter(int thr_id, struct timeval *diff,
 		double thread_rolling = 0.0;
 		int i;
 
-		applog(LOG_DEBUG, "[thread %d: %lu hashes, %.0f khash/sec]",
+		applog(LOG_DEBUG, "[thread %d: %llu hashes, %.0f khash/sec]",
 			thr_id, hashes_done, hashes_done / secs);
 
 		/* Rolling average for each thread and each device */
@@ -3764,8 +3694,8 @@ void *miner_thread(void *userdata)
 	struct timeval tv_start, tv_end, tv_workstart, tv_lastupdate;
 	struct timeval diff, sdiff, wdiff;
 	uint32_t max_nonce = api->can_limit_work ? api->can_limit_work(mythr) : 0xffffffff;
-	uint32_t hashes_done = 0;
-	uint32_t hashes;
+	unsigned long long hashes_done = 0;
+	unsigned long long hashes;
 	struct work *work = make_work();
 	unsigned const int request_interval = opt_scantime * 2 / 3 ? : 1;
 	unsigned const long request_nonce = MAXTHREADS / 3 * 2;
@@ -3828,11 +3758,14 @@ void *miner_thread(void *userdata)
 			}
 
 			if (unlikely(!hashes)) {
+				applog(LOG_ERR, "%s %d failure, disabling!", api->name, cgpu->device_id);
+				cgpu->deven = DEV_DISABLED;
+
 				cgpu->device_last_not_well = time(NULL);
 				cgpu->device_not_well_reason = REASON_THREAD_ZERO_HASH;
 				cgpu->thread_zero_hash_count++;
 
-				goto out;
+				goto disabled;
 			}
 
 			hashes_done += hashes;
@@ -3893,6 +3826,7 @@ void *miner_thread(void *userdata)
 
 			if (unlikely(mythr->pause || cgpu->deven != DEV_ENABLED)) {
 				applog(LOG_WARNING, "Thread %d being disabled", thr_id);
+disabled:
 				mythr->rolling = mythr->cgpu->rolling = 0;
 				applog(LOG_DEBUG, "Popping wakeup ping in miner thread");
 				thread_reportout(mythr);
@@ -4056,8 +3990,9 @@ retry_pool:
 			if (end.tv_sec - start.tv_sec > 30)
 				continue;
 			if (opt_retries == -1 || failures++ < opt_retries) {
+				if (failures == 1)
 					applog(LOG_WARNING,
-					"longpoll failed for %s, sleeping for 30s", pool->lp_url);
+					       "longpoll failed for %s, retrying every 30s", pool->lp_url);
 				sleep(30);
 			} else {
 				applog(LOG_ERR,
@@ -4090,6 +4025,26 @@ void reinit_device(struct cgpu_info *cgpu)
 
 static struct timeval rotate_tv;
 
+/* We reap curls if they are unused for over a minute */
+static void reap_curl(struct pool *pool)
+{
+	struct curl_ent *ent, *iter;
+	struct timeval now;
+
+	gettimeofday(&now, NULL);
+	mutex_lock(&pool->pool_lock);
+	list_for_each_entry_safe(ent, iter, &pool->curlring, node) {
+		if (now.tv_sec - ent->tv.tv_sec > 60) {
+			applog(LOG_DEBUG, "Reaped curl %d from pool %d", pool->curls, pool->pool_no);
+			pool->curls--;
+			list_del(&ent->node);
+			curl_easy_cleanup(ent->curl);
+			free(ent);
+		}
+	}
+	mutex_unlock(&pool->pool_lock);
+}
+
 static void *watchpool_thread(void __maybe_unused *userdata)
 {
 	pthread_setcanceltype(PTHREAD_CANCEL_ASYNCHRONOUS, NULL);
@@ -4103,6 +4058,7 @@ static void *watchpool_thread(void __maybe_unused *userdata)
 		for (i = 0; i < total_pools; i++) {
 			struct pool *pool = pools[i];
 
+			reap_curl(pool);
 			if (!pool->enabled)
 				continue;
 
@@ -4119,7 +4075,7 @@ static void *watchpool_thread(void __maybe_unused *userdata)
 			switch_pools(NULL);
 		}
 
-		sleep(10);
+		sleep(30);
 	}
 	return NULL;
 }
@@ -4435,18 +4391,13 @@ char *curses_input(const char *query)
 
 int add_pool_details(bool live, char *url, char *user, char *pass)
 {
-	struct pool *pool = NULL;
+	struct pool *pool;
 
 	if (total_pools == MAX_POOLS)
 		return ADD_POOL_MAXIMUM;
 
-	pool = calloc(sizeof(struct pool), 1);
-	if (!pool)
-		quit(1, "Failed to realloc pools in add_pool_details");
-	pool->pool_no = total_pools;
-	pool->prio = total_pools;
-	if (unlikely(pthread_mutex_init(&pool->pool_lock, NULL)))
-		quit (1, "Failed to pthread_mutex_init in input_pool");
+	pool = add_pool();
+
 	pool->rpc_url = url;
 	pool->rpc_user = user;
 	pool->rpc_pass = pass;
@@ -4455,19 +4406,11 @@ int add_pool_details(bool live, char *url, char *user, char *pass)
 		quit(1, "Failed to malloc userpass");
 	sprintf(pool->rpc_userpass, "%s:%s", pool->rpc_user, pool->rpc_pass);
 
-	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");
-
 	/* Test the pool is not idle if we're live running, otherwise
 	 * it will be tested separately */
 	pool->enabled = true;
 	if (live && !pool_active(pool, false))
 		pool->idle = true;
-	pools[total_pools++] = pool;
 
 	return ADD_POOL_OK;
 }

configure.ac

@@ -1,8 +1,8 @@
 ##--##--##--##--##--##--##--##--##--##--##--##--##--##--##--##--##
 ##--##--##--##--##--##--##--##--##--##--##--##--##--##--##--##--##
 m4_define([v_maj], [2])
-m4_define([v_min], [3])
-m4_define([v_mic], [6])
+m4_define([v_min], [4])
+m4_define([v_mic], [0])
 ##--##--##--##--##--##--##--##--##--##--##--##--##--##--##--##--##
 m4_define([v_ver], [v_maj.v_min.v_mic])
 m4_define([lt_rev], m4_eval(v_maj + v_min))

driver-bitforce.c

@@ -79,14 +79,18 @@ static void BFgets(char *buf, size_t bufLen, int fd)
 	buf[0] = '\0';
 }
 
-static void BFwrite(int fd, const void *buf, ssize_t bufLen)
+static ssize_t BFwrite2(int fd, const void *buf, ssize_t bufLen)
 {
-	ssize_t ret = write(fd, buf, bufLen);
-
-	if (unlikely(ret != bufLen))
-		quit(1, "BFwrite failed");
+	return write(fd, buf, bufLen);
 }
 
+#define BFwrite(fd, buf, bufLen)  do {  \
+	if ((bufLen) != BFwrite2(fd, buf, bufLen)) {  \
+		applog(LOG_ERR, "Error writing to BitForce (" #buf ")");  \
+		return 0;  \
+	}  \
+} while(0)
+
 #define BFclose(fd) close(fd)
 
 static bool bitforce_detect_one(const char *devpath)

driver-icarus.c

@@ -196,7 +196,7 @@ static int icarus_write(int fd, const void *buf, size_t bufLen)
 
 static bool icarus_detect_one(const char *devpath)
 {
-	struct timeval tv1, tv2;
+	struct timeval tv_start, tv_finish;
 	int fd;
 
 	// Block 171874 nonce = (0xa2870100) = 0x000187a2
@@ -225,11 +225,11 @@ static bool icarus_detect_one(const char *devpath)
 
 	hex2bin(ob_bin, golden_ob, sizeof(ob_bin));
 	icarus_write(fd, ob_bin, sizeof(ob_bin));
-	gettimeofday(&tv1, NULL);
+	gettimeofday(&tv_start, NULL);
 
 	memset(nonce_bin, 0, sizeof(nonce_bin));
 	icarus_gets(nonce_bin, sizeof(nonce_bin), fd, -1, 1);
-	gettimeofday(&tv2, NULL);
+	gettimeofday(&tv_finish, NULL);
 
 	icarus_close(fd);
 
@@ -315,7 +315,9 @@ static uint64_t icarus_scanhash(struct thr_info *thr, struct work *work,
 	char *ob_hex, *nonce_hex;
 	uint32_t nonce;
 	uint32_t hash_count;
-	struct timeval tv1, tv2, elapsed;
+	struct timeval tv_start, tv_finish, elapsed;
+
+	elapsed.tv_sec = elapsed.tv_usec = 0;
 
 	icarus = thr->cgpu;
 	fd = icarus->device_fd;
@@ -329,30 +331,37 @@ static uint64_t icarus_scanhash(struct thr_info *thr, struct work *work,
 	tcflush(fd, TCOFLUSH);
 #endif
 	ret = icarus_write(fd, ob_bin, sizeof(ob_bin));
-	gettimeofday(&tv1, NULL);
+	if (opt_debug)
+		gettimeofday(&tv_start, NULL);
 	if (ret)
 		return 0;	/* This should never happen */
 
+	if (opt_debug) {
 		ob_hex = bin2hex(ob_bin, sizeof(ob_bin));
 		if (ob_hex) {
 			applog(LOG_DEBUG, "Icarus %d sent: %s",
 			       icarus->device_id, ob_hex);
 			free(ob_hex);
 		}
+	}
 
-	/* Icarus will return 8 bytes nonces or nothing */
+	/* Icarus will return 4 bytes nonces or nothing */
 	memset(nonce_bin, 0, sizeof(nonce_bin));
 	ret = icarus_gets(nonce_bin, sizeof(nonce_bin), fd, thr_id,
 						ICARUS_READ_COUNT_DEFAULT);
-	gettimeofday(&tv2, NULL);
+
+	if (opt_debug)
+		gettimeofday(&tv_finish, NULL);
 
 	memcpy((char *)&nonce, nonce_bin, sizeof(nonce_bin));
 
 	// aborted before becoming idle, get new work
         if (nonce == 0 && ret) {
-		timersub(&tv2, &tv1, &elapsed);
+		if (opt_debug) {
+			timersub(&tv_finish, &tv_start, &elapsed);
 			applog(LOG_DEBUG, "Icarus %d no nonce = 0x%08x hashes (%ld.%06lds)",
 				icarus->device_id, ESTIMATE_HASHES, elapsed.tv_sec, elapsed.tv_usec);
+		}
 		return ESTIMATE_HASHES;
 	}
 
@@ -363,7 +372,8 @@ static uint64_t icarus_scanhash(struct thr_info *thr, struct work *work,
 	work->blk.nonce = 0xffffffff;
 	submit_nonce(thr, work, nonce);
 
-	timersub(&tv2, &tv1, &elapsed);
+	if (opt_debug) {
+		timersub(&tv_finish, &tv_start, &elapsed);
 
 		nonce_hex = bin2hex(nonce_bin, sizeof(nonce_bin));
 		if (nonce_hex) {
@@ -371,17 +381,15 @@ static uint64_t icarus_scanhash(struct thr_info *thr, struct work *work,
 			       icarus->device_id, elapsed.tv_sec, elapsed.tv_usec, nonce_hex);
 			free(nonce_hex);
 		}
+	}
 
 	hash_count = (nonce & 0x7fffffff);
-        if (hash_count == 0)
-		hash_count = 2;
-        else {
 	if (hash_count++ == 0x7fffffff)
 		hash_count = 0xffffffff;
 	else
 		hash_count <<= 1;
-        }
 
+	if (opt_debug)
 		applog(LOG_DEBUG, "Icarus %d nonce = 0x%08x = 0x%08x hashes (%ld.%06lds)",
 				icarus->device_id, nonce, hash_count, elapsed.tv_sec, elapsed.tv_usec);
 
@@ -390,21 +398,8 @@ static uint64_t icarus_scanhash(struct thr_info *thr, struct work *work,
 
 static void icarus_shutdown(struct thr_info *thr)
 {
-	struct cgpu_info *icarus;
-
-	if (thr->cgpu) {
-		icarus = thr->cgpu;
-
-		if (icarus->device_path)
-			free(icarus->device_path);
-
-		close(icarus->device_fd);
-
-		devices[icarus->device_id] = NULL;
-		free(icarus);
-
-		thr->cgpu = NULL;
-	}
+	struct cgpu_info *icarus = thr->cgpu;
+	icarus_close(icarus->device_fd);
 }
 
 struct device_api icarus_api = {

miner.h

@@ -598,10 +598,17 @@ typedef struct {
 } dev_blk_ctx;
 #endif
 
+struct curl_ent {
+	CURL *curl;
+	struct list_head node;
+	struct timeval tv;
+};
+
 struct pool {
 	int pool_no;
 	int prio;
 	int accepted, rejected;
+	int seq_rejects;
 
 	bool submit_fail;
 	bool idle;
@@ -635,8 +642,9 @@ struct pool {
 	pthread_t submit_thread;
 	pthread_t getwork_thread;
 
-	CURL *submit_curl;
-	CURL *getwork_curl;
+	int curls;
+	pthread_cond_t cr_cond;
+	struct list_head curlring;
 };
 
 struct work {