Browse Source

Make sure that local work generation does not proceed indefinitely by limiting it to the average block generation time - 10 minutes.

Doing more work will just waste power and will virtually never produce accepted shares.
nfactor-troky
Con Kolivas 13 years ago
parent
commit
6dcf05ffaf
  1. 49
      main.c

49
main.c

@ -172,6 +172,7 @@ static int accepted, rejected;
int hw_errors; int hw_errors;
static int total_queued, total_staged, lp_staged; static int total_queued, total_staged, lp_staged;
static bool localgen = false; static bool localgen = false;
static bool idlenet = false;
static unsigned int getwork_requested; static unsigned int getwork_requested;
static unsigned int stale_shares; static unsigned int stale_shares;
static unsigned int discarded_work; static unsigned int discarded_work;
@ -925,9 +926,18 @@ static void inc_staged(int inc, bool lp)
if (lp) { if (lp) {
lp_staged += inc; lp_staged += inc;
total_staged += inc; total_staged += inc;
} else if (lp_staged) idlenet = true;
lp_staged--; } else if (lp_staged) {
else if (!--lp_staged) {
unsigned int i;
/* Make sure the watchdog thread doesn't kill the mining
* threads once we unset the idlenet flag */
for (i = 0; i < mining_threads; i++)
gettimeofday(&thr_info[i].last, NULL);
idlenet = false;
}
} else
total_staged += inc; total_staged += inc;
pthread_mutex_unlock(&stgd_lock); pthread_mutex_unlock(&stgd_lock);
} }
@ -1226,6 +1236,7 @@ static void flush_requests(bool longpoll)
static bool get_work(struct work *work, bool queued) static bool get_work(struct work *work, bool queued)
{ {
static struct timeval tv_localgen = {};
struct thr_info *thr = &thr_info[0]; struct thr_info *thr = &thr_info[0];
struct work *work_heap; struct work *work_heap;
bool ret = false; bool ret = false;
@ -1243,9 +1254,24 @@ retry:
/* Only print this message once each time we shift to localgen */ /* Only print this message once each time we shift to localgen */
if (!localgen) { if (!localgen) {
applog(LOG_WARNING, "Server not providing work fast enough, generating work locally");
localgen = true; localgen = true;
applog(LOG_WARNING, "Server not providing work fast enough, generating work locally");
localgen_occasions++; localgen_occasions++;
gettimeofday(&tv_localgen, NULL);
} else {
struct timeval tv_now, diff;
gettimeofday(&tv_now, NULL);
timeval_subtract(&diff, &tv_now, &tv_localgen);
if (diff.tv_sec > 600) {
/* A new block appears on average every 10 mins */
applog(LOG_WARNING, "Server not responding for more than 10 minutes.");
applog(LOG_WARNING, "Further local work generation will only generate rejects.");
applog(LOG_WARNING, "Going idle till network conditions recover.");
/* Force every thread to wait for new work */
inc_staged(mining_threads, true);
goto retry;
}
} }
work_ntime = (uint32_t *)(work->data + 68); work_ntime = (uint32_t *)(work->data + 68);
ntime = be32toh(*work_ntime); ntime = be32toh(*work_ntime);
@ -1254,17 +1280,19 @@ retry:
ret = true; ret = true;
local_work++; local_work++;
goto out; goto out;
} else if (localgen) {
localgen = false;
applog(LOG_WARNING, "Resumed retrieving work from server");
} }
/* wait for 1st response, or get cached response */ /* wait for 1st response, or get cached response */
work_heap = tq_pop(thr->q, NULL); work_heap = tq_pop(thr->q, NULL);
if (unlikely(!work_heap)) { if (unlikely(!work_heap)) {
applog(LOG_WARNING, "Failed to tq_pop in get_work"); applog(LOG_WARNING, "Failed to tq_pop in get_work");
goto out; goto out;
} }
/* If we make it here we have succeeded in getting fresh work */
if (localgen) {
localgen = false;
applog(LOG_WARNING, "Resuming with work from server");
}
dec_queued(); dec_queued();
memcpy(work, work_heap, sizeof(*work)); memcpy(work, work_heap, sizeof(*work));
@ -1972,8 +2000,9 @@ static void *watchdog_thread(void *userdata)
for (i = 0; i < mining_threads; i++) { for (i = 0; i < mining_threads; i++) {
struct thr_info *thr = &thr_info[i]; struct thr_info *thr = &thr_info[i];
/* Do not kill threads waiting on longpoll staged work */ /* Do not kill threads waiting on longpoll staged work
if (now.tv_sec - thr->last.tv_sec > 60 && !lp_staged) { * or idle network */
if (now.tv_sec - thr->last.tv_sec > 60 && !idlenet) {
applog(LOG_ERR, "Attempting to restart thread %d, idle for more than 60 seconds", i); applog(LOG_ERR, "Attempting to restart thread %d, idle for more than 60 seconds", i);
/* Create one mandatory work item */ /* Create one mandatory work item */
inc_staged(1, true); inc_staged(1, true);

Loading…
Cancel
Save