ibmveth: Add optional flush of rx buffer
Santiago Leon [Fri, 3 Sep 2010 18:28:41 +0000 (18:28 +0000)]
On some machines we can improve the bandwidth by ensuring rx buffers are
not in the cache. Add a module option that is disabled by default that flushes
rx buffers on insertion.

Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Santiago Leon <santil@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

drivers/net/ibmveth.c

index d8a8984..1685e23 100644 (file)
@@ -127,6 +127,10 @@ module_param(rx_copybreak, uint, 0644);
 MODULE_PARM_DESC(rx_copybreak,
        "Maximum size of packet that is copied to a new buffer on receive");
 
+static unsigned int rx_flush __read_mostly = 0;
+module_param(rx_flush, uint, 0644);
+MODULE_PARM_DESC(rx_flush, "Flush receive buffers before use");
+
 struct ibmveth_stat {
        char name[ETH_GSTRING_LEN];
        int offset;
@@ -234,6 +238,14 @@ static int ibmveth_alloc_buffer_pool(struct ibmveth_buff_pool *pool)
        return 0;
 }
 
+static inline void ibmveth_flush_buffer(void *addr, unsigned long length)
+{
+       unsigned long offset;
+
+       for (offset = 0; offset < length; offset += SMP_CACHE_BYTES)
+               asm("dcbfl %0,%1" :: "b" (addr), "r" (offset));
+}
+
 /* replenish the buffers for a pool.  note that we don't need to
  * skb_reserve these since they are used for incoming...
  */
@@ -286,6 +298,12 @@ static void ibmveth_replenish_buffer_pool(struct ibmveth_adapter *adapter, struc
                desc.fields.flags_len = IBMVETH_BUF_VALID | pool->buff_size;
                desc.fields.address = dma_addr;
 
+               if (rx_flush) {
+                       unsigned int len = min(pool->buff_size,
+                                               adapter->netdev->mtu +
+                                               IBMVETH_BUFF_OH);
+                       ibmveth_flush_buffer(skb->data, len);
+               }
                lpar_rc = h_add_logical_lan_buffer(adapter->vdev->unit_address, desc.desc);
 
                if (lpar_rc != H_SUCCESS)
@@ -1095,6 +1113,9 @@ static int ibmveth_poll(struct napi_struct *napi, int budget)
                                skb_copy_to_linear_data(new_skb,
                                                        skb->data + offset,
                                                        length);
+                               if (rx_flush)
+                                       ibmveth_flush_buffer(skb->data,
+                                               length + offset);
                                skb = new_skb;
                                ibmveth_rxq_recycle_buffer(adapter);
                        } else {