GOSTSec/sgminer
1
0
Fork 0
mirror of https://github.com/GOSTSec/sgminer synced 2025-01-30 08:24:26 +00:00
Code Issues Releases Wiki Activity

So we can pthread cancel a stuck thread, but can't pthread join it.

Browse Source 
And we can't tq freeze the old queue or we could hang here.
And we need to re-initialise all the opencl contexts so send a test command and if it succeeds reinit the card.
This commit is contained in:
Con Kolivas 2011-07-30 13:28:25 +10:00
parent 74a07eb461
commit 59b640870c
1 changed files with 21 additions and 8 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

29
main.c
View File

@ -3133,13 +3133,13 @@ static void *gpuminer_thread(void *userdata)
} }
} }
if (unlikely(!gpu_devices[gpu])) { if (unlikely(!gpu_devices[gpu])) {
applog(LOG_WARNING, "Thread %d being disabled\n", thr_id); applog(LOG_WARNING, "Thread %d being disabled", thr_id);
mythr->rolling = mythr->cgpu->rolling = 0; mythr->rolling = mythr->cgpu->rolling = 0;
if (opt_debug) if (opt_debug)
applog(LOG_DEBUG, "Popping wakeup ping in gpuminer thread"); applog(LOG_DEBUG, "Popping wakeup ping in gpuminer thread");
tq_pop(mythr->q, NULL); /* Ignore ping that's popped */ tq_pop(mythr->q, NULL); /* Ignore ping that's popped */
applog(LOG_WARNING, "Thread %d being re-enabled\n", thr_id); applog(LOG_WARNING, "Thread %d being re-enabled", thr_id);
} }
} }
out: out:
@ -3361,7 +3361,9 @@ static void *reinit_gpu(void *userdata)
int thr_id; int thr_id;
_clState *clState; _clState *clState;
gpus[gpu].status = LIFE_DEAD; /* Send threads message to stop */
gpu_devices[gpu] = false;
sleep(5);
for (thr_id = 0; thr_id < gpu_threads; thr_id ++) { for (thr_id = 0; thr_id < gpu_threads; thr_id ++) {
if (dev_from_id(thr_id) != gpu) if (dev_from_id(thr_id) != gpu)
@ -3370,11 +3372,14 @@ static void *reinit_gpu(void *userdata)
clState = clStates[thr_id]; clState = clStates[thr_id];
thr = &thr_info[thr_id]; thr = &thr_info[thr_id];
thr->rolling = thr->cgpu->rolling = 0; thr->rolling = thr->cgpu->rolling = 0;
tq_freeze(thr->q);
if (!pthread_cancel(*thr->pth)) { if (!pthread_cancel(*thr->pth)) {
pthread_join(*thr->pth, NULL); applog(LOG_WARNING, "Thread still exists, killing it off");
free(thr->q); free(thr->q);
} } else
applog(LOG_WARNING, "Thread no longer exists");
sleep(5);
/* Lose this ram cause we may get stuck here! */
//tq_freeze(thr->q);
thr->q = tq_new(); thr->q = tq_new();
if (!thr->q) if (!thr->q)
@ -3408,6 +3413,14 @@ static void *reinit_gpu(void *userdata)
applog(LOG_WARNING, "Thread %d restarted", thr_id); applog(LOG_WARNING, "Thread %d restarted", thr_id);
} }
/* Try to re-enable it */
gpu_devices[gpu] = true;
for (thr_id = 0; thr_id < gpu_threads; thr_id ++) {
thr = &thr_info[thr_id];
if (dev_from_id(thr_id) == gpu)
tq_push(thr->q, &ping);
}
return NULL; return NULL;
} }
#endif #endif
@ -3523,9 +3536,9 @@ static void *watchdog_thread(void *userdata)
applog(LOG_ERR, "Thread %d idle for more than 60 seconds, GPU %d declared SICK!", i, gpu); applog(LOG_ERR, "Thread %d idle for more than 60 seconds, GPU %d declared SICK!", i, gpu);
/* Sent it a ping, it might respond */ /* Sent it a ping, it might respond */
tq_push(thr->q, &ping); tq_push(thr->q, &ping);
} else if (now.tv_sec - thr->last.tv_sec > 600 && gpus[i].status == LIFE_SICK) { } else if (now.tv_sec - thr->last.tv_sec > 300 && gpus[i].status == LIFE_SICK) {
gpus[gpu].status = LIFE_DEAD; gpus[gpu].status = LIFE_DEAD;
applog(LOG_ERR, "Thread %d idle for more than 10 minutes, GPU %d declared DEAD!", i, gpu); applog(LOG_ERR, "Thread %d idle for more than 5 minutes, GPU %d declared DEAD!", i, gpu);
applog(LOG_ERR, "Attempting to restart GPU"); applog(LOG_ERR, "Attempting to restart GPU");
reinit_device(thr->cgpu); reinit_device(thr->cgpu);
break; break;