zebra: Limit queue depth in dplane_fpm_nl

donaldsharp · donaldsharp · commit 8926ac198487 · 2024-09-05T15:52:05.000-04:00
The dplane providers have a concept of input queues
and output queues.  These queues are chained together
during normal operation.  The code in zebra also has
a feedback mechanism where the MetaQ will not run when
the first input queue is backed up.  Having the dplane_fpm_nl
code grab all contexts when it is backed up prevents
this system from behaving appropriately.

Modify the code to not add to the dplane_fpm_nl's internal
queue when it is already full.  This will allow the backpressure
to work appropriately in zebra proper.

Signed-off-by: Donald Sharp &lt;sharpd@nvidia.com&gt;
diff --git a/zebra/dplane_fpm_nl.c b/zebra/dplane_fpm_nl.c
@@ -1678,6 +1678,25 @@ static int fpm_nl_process(struct zebra_dplane_provider *prov)
 
 	fnc = dplane_provider_get_data(prov);
 	limit = dplane_provider_get_work_limit(prov);
+
+	frr_with_mutex (&fnc->ctxqueue_mutex) {
+		cur_queue = dplane_ctx_queue_count(&fnc->ctxqueue);
+	}
+
+	if (cur_queue >= (uint64_t)limit) {
+		if (IS_ZEBRA_DEBUG_FPM)
+			zlog_debug("%s: Already at a limit(%" PRIu64
+				   ") of internal work, hold off",
+				   __func__, cur_queue);
+		limit = 0;
+	} else if (cur_queue != 0) {
+		if (IS_ZEBRA_DEBUG_FPM)
+			zlog_debug("%s: current queue is %" PRIu64
+				   ", limiting to lesser amount of %" PRIu64,
+				   __func__, cur_queue, limit - cur_queue);
+		limit -= cur_queue;
+	}
+
 	for (counter = 0; counter < limit; counter++) {
 		ctx = dplane_provider_dequeue_in_ctx(prov);
 		if (ctx == NULL)
