iwlwifi: pcie: implement GRO without NAPI

Use the new NAPI infrastructure added to mac80211 to get
GRO. We don't really implement NAPI since we don't have
a real poll function and we never schedule a NAPI poll.
Instead of this, we collect all the packets we got from a
single interrupt and then call napi_gro_flush().

This allows us to benefit from GRO. In half duplex medium
like WiFi, its main advantage is that it reduces the number
of TCP Acks, hence improving the TCP Rx performance.

Since we call the Rx path with a spinlock held, remove
the might_sleep mention from the op_mode's API.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Reviewed-by: Ido Yariv <ido@wizery.com>
[Squash different patches and rewrite the commit message]
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>

3 files changed, 37 insertions, 4 deletions

diff --git a/drivers/net/wireless/iwlwifi/pcie/internal.h b/drivers/net/wireless/iwlwifi/pcie/internal.h
index 9091513ea738..e2694686ebfc 100644
--- a/drivers/net/wireless/iwlwifi/pcie/internal.h
+++ b/drivers/net/wireless/iwlwifi/pcie/internal.h
@@ -270,6 +270,9 @@ struct iwl_trans_pcie {
 	struct iwl_trans *trans;
 	struct iwl_drv *drv;
 
+	struct net_device napi_dev;
+	struct napi_struct napi;
+
 	/* INT ICT Table */
 	__le32 *ict_tbl;
 	dma_addr_t ict_tbl_dma;
diff --git a/drivers/net/wireless/iwlwifi/pcie/rx.c b/drivers/net/wireless/iwlwifi/pcie/rx.c
index fdfa3969cac9..e8e5afcaf42b 100644
--- a/drivers/net/wireless/iwlwifi/pcie/rx.c
+++ b/drivers/net/wireless/iwlwifi/pcie/rx.c
@@ -673,7 +673,6 @@ static void iwl_pcie_rx_handle_rb(struct iwl_trans *trans,
 	/* Reuse the page if possible. For notification packets and
 	 * SKBs that fail to Rx correctly, add them back into the
 	 * rx_free list for reuse later. */
-	spin_lock(&rxq->lock);
 	if (rxb->page != NULL) {
 		rxb->page_dma =
 			dma_map_page(trans->dev, rxb->page, 0,
@@ -694,7 +693,6 @@ static void iwl_pcie_rx_handle_rb(struct iwl_trans *trans,
 		}
 	} else
 		list_add_tail(&rxb->list, &rxq->rx_used);
-	spin_unlock(&rxq->lock);
 }
 
 /*
@@ -709,6 +707,8 @@ static void iwl_pcie_rx_handle(struct iwl_trans *trans)
 	u32 count = 8;
 	int total_empty;
 
+restart:
+	spin_lock(&rxq->lock);
 	/* uCode's read index (stored in shared DRAM) indicates the last Rx
 	 * buffer that the driver may process (last buffer filled by ucode). */
 	r = le16_to_cpu(ACCESS_ONCE(rxq->rb_stts->closed_rb_num)) & 0x0FFF;
@@ -743,18 +743,25 @@ static void iwl_pcie_rx_handle(struct iwl_trans *trans)
 			count++;
 			if (count >= 8) {
 				rxq->read = i;
+				spin_unlock(&rxq->lock);
 				iwl_pcie_rx_replenish_now(trans);
 				count = 0;
+				goto restart;
 			}
 		}
 	}
 
 	/* Backtrack one entry */
 	rxq->read = i;
+	spin_unlock(&rxq->lock);
+
 	if (fill_rx)
 		iwl_pcie_rx_replenish_now(trans);
 	else
 		iwl_pcie_rxq_restock(trans);
+
+	if (trans_pcie->napi.poll)
+		napi_gro_flush(&trans_pcie->napi, false);
 }
 
 /*
@@ -1068,8 +1075,6 @@ irqreturn_t iwl_pcie_irq_handler(int irq, void *dev_id)
 		iwl_write8(trans, CSR_INT_PERIODIC_REG,
 			    CSR_INT_PERIODIC_DIS);
 
-		iwl_pcie_rx_handle(trans);
-
 		/*
 		 * Enable periodic interrupt in 8 msec only if we received
 		 * real RX interrupt (instead of just periodic int), to catch
@@ -1082,6 +1087,10 @@ irqreturn_t iwl_pcie_irq_handler(int irq, void *dev_id)
 				   CSR_INT_PERIODIC_ENA);
 
 		isr_stats->rx++;
+
+		local_bh_disable();
+		iwl_pcie_rx_handle(trans);
+		local_bh_enable();
 	}
 
 	/* This "Tx" DMA channel is used only for loading uCode */
diff --git a/drivers/net/wireless/iwlwifi/pcie/trans.c b/drivers/net/wireless/iwlwifi/pcie/trans.c
index dcfd6d866d09..97e6bd826880 100644
--- a/drivers/net/wireless/iwlwifi/pcie/trans.c
+++ b/drivers/net/wireless/iwlwifi/pcie/trans.c
@@ -1053,6 +1053,12 @@ static void iwl_trans_pcie_write_prph(struct iwl_trans *trans, u32 addr,
 	iwl_trans_pcie_write32(trans, HBUS_TARG_PRPH_WDAT, val);
 }
 
+static int iwl_pcie_dummy_napi_poll(struct napi_struct *napi, int budget)
+{
+	WARN_ON(1);
+	return 0;
+}
+
 static void iwl_trans_pcie_configure(struct iwl_trans *trans,
 				     const struct iwl_trans_config *trans_cfg)
 {
@@ -1079,6 +1085,18 @@ static void iwl_trans_pcie_configure(struct iwl_trans *trans,
 
 	trans_pcie->command_names = trans_cfg->command_names;
 	trans_pcie->bc_table_dword = trans_cfg->bc_table_dword;
+
+	/* Initialize NAPI here - it should be before registering to mac80211
+	 * in the opmode but after the HW struct is allocated.
+	 * As this function may be called again in some corner cases don't
+	 * do anything if NAPI was already initialized.
+	 */
+	if (!trans_pcie->napi.poll && trans->op_mode->ops->napi_add) {
+		init_dummy_netdev(&trans_pcie->napi_dev);
+		iwl_op_mode_napi_add(trans->op_mode, &trans_pcie->napi,
+				     &trans_pcie->napi_dev,
+				     iwl_pcie_dummy_napi_poll, 64);
+	}
 }
 
 void iwl_trans_pcie_free(struct iwl_trans *trans)
@@ -1099,6 +1117,9 @@ void iwl_trans_pcie_free(struct iwl_trans *trans)
 	pci_disable_device(trans_pcie->pci_dev);
 	kmem_cache_destroy(trans->dev_cmd_pool);
 
+	if (trans_pcie->napi.poll)
+		netif_napi_del(&trans_pcie->napi);
+
 	kfree(trans);
 }