dpaa2-eth: Update FQ taildrop threshold and buffer pool count

Now that we have congestion group taildrop configured at all
times, we can afford to increase the frame queue taildrop
threshold; this will ensure a better response when receiving
bursts of large-sized frames.

Also decouple the buffer pool count from the Rx FQ taildrop
threshold, as above change would increase it too much. Instead,
keep the old count as a hardcoded value.

With the new limits, we try to ensure that:
* we allow enough leeway for large frame bursts (by buffering
enough of them in queues to avoid heavy dropping in case of
bursty traffic, but when overall ingress bandwidth is manageable)
* allow pending frames to be evenly spread between ingress FQs,
regardless of frame size
* avoid dropping frames due to the buffer pool being empty; this
is not a bad behaviour per se, but system overall response is
more linear and predictable when frames are dropped at frame
queue/group level.

Signed-off-by: Ioana Radulescu <[email protected]>
Signed-off-by: Ioana Ciornei <[email protected]>
Signed-off-by: David S. Miller <[email protected]>

Author: Ioana Radulescu

drivers/net/ethernet/freescale/dpaa2/dpaa2-eth.h

@@ -36,24 +36,24 @@
 /* Convert L3 MTU to L2 MFL */
 #define DPAA2_ETH_L2_MAX_FRM(mtu)	((mtu) + VLAN_ETH_HLEN)
 
-/* Set the taildrop threshold (in bytes) to allow the enqueue of several jumbo
- * frames in the Rx queues (length of the current frame is not
- * taken into account when making the taildrop decision)
+/* Set the taildrop threshold (in bytes) to allow the enqueue of a large
+ * enough number of jumbo frames in the Rx queues (length of the current
+ * frame is not taken into account when making the taildrop decision)
  */
-#define DPAA2_ETH_FQ_TAILDROP_THRESH	(64 * 1024)
+#define DPAA2_ETH_FQ_TAILDROP_THRESH	(1024 * 1024)
 
 /* Maximum number of Tx confirmation frames to be processed
  * in a single NAPI call
  */
 #define DPAA2_ETH_TXCONF_PER_NAPI	256
 
-/* Buffer quota per queue. Must be large enough such that for minimum sized
- * frames taildrop kicks in before the bpool gets depleted, so we compute
- * how many 64B frames fit inside the taildrop threshold and add a margin
- * to accommodate the buffer refill delay.
+/* Buffer qouta per channel. We want to keep in check number of ingress frames
+ * in flight: for small sized frames, congestion group taildrop may kick in
+ * first; for large sizes, Rx FQ taildrop threshold will ensure only a
+ * reasonable number of frames will be pending at any given time.
+ * Ingress frame drop due to buffer pool depletion should be a corner case only
  */
-#define DPAA2_ETH_MAX_FRAMES_PER_QUEUE	(DPAA2_ETH_FQ_TAILDROP_THRESH / 64)
-#define DPAA2_ETH_NUM_BUFS		(DPAA2_ETH_MAX_FRAMES_PER_QUEUE + 256)
+#define DPAA2_ETH_NUM_BUFS		1280
 #define DPAA2_ETH_REFILL_THRESH \
 	(DPAA2_ETH_NUM_BUFS - DPAA2_ETH_BUFS_PER_CMD)
 
@@ -63,8 +63,7 @@
  * taildrop kicks in
  */
 #define DPAA2_ETH_CG_TAILDROP_THRESH(priv)				\
-	(DPAA2_ETH_MAX_FRAMES_PER_QUEUE * dpaa2_eth_queue_count(priv) /	\
-	 dpaa2_eth_tc_count(priv))
+	(1024 * dpaa2_eth_queue_count(priv) / dpaa2_eth_tc_count(priv))
 
 /* Maximum number of buffers that can be acquired/released through a single
  * QBMan command