Fix infinite loop on platforms with severely limited aio resources
Problem hit on OSX, where with 4 jobs each with a queue depth larger
than what the system supports, you can get into a situation where
any given process can get EAGAIN on queueing AIO even if it has
nothing queued already. Check for this condition to avoid fio's
IO loop going into a death spiral.
Signed-off-by: Jens Axboe <jaxboe@fusionio.com>
diff --git a/fio.c b/fio.c
index 9e9106d..120431e 100644
--- a/fio.c
+++ b/fio.c
@@ -550,9 +550,10 @@
/*
* if we can queue more, do so. but check if there are
- * completed io_u's first.
+ * completed io_u's first. Note that we can get BUSY even
+ * without IO queued, if the system is resource starved.
*/
- full = queue_full(td) || ret == FIO_Q_BUSY;
+ full = queue_full(td) || (ret == FIO_Q_BUSY && td->cur_depth);
if (full || !td->o.iodepth_batch_complete) {
min_events = min(td->o.iodepth_batch_complete,
td->cur_depth);
@@ -710,9 +711,11 @@
break;
/*
- * See if we need to complete some commands
+ * See if we need to complete some commands. Note that we
+ * can get BUSY even without IO queued, if the system is
+ * resource starved.
*/
- full = queue_full(td) || ret == FIO_Q_BUSY;
+ full = queue_full(td) || (ret == FIO_Q_BUSY && td->cur_depth);
if (full || !td->o.iodepth_batch_complete) {
min_evts = min(td->o.iodepth_batch_complete,
td->cur_depth);