bnx2x: Using the new FW
The new FW improves the packets per second rate. It required a lot of change in
the FW which implies many changes in the driver to support it. It is now also
possible for the driver to use a separate MSI-X vector for Rx and Tx - this also
add some to the complicity of this change.
All things said - after this patch, practically all performance matrixes show
improvement.
Though Vladislav Zolotarov is not signed on this patch, he did most of the job
and deserves credit for that.
Signed-off-by: Eilon Greenstein <eilong@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
diff --git a/drivers/net/bnx2x.h b/drivers/net/bnx2x.h
index 8bd80fc..16ccba8 100644
--- a/drivers/net/bnx2x.h
+++ b/drivers/net/bnx2x.h
@@ -142,6 +142,9 @@
struct sw_tx_bd {
struct sk_buff *skb;
u16 first_bd;
+ u8 flags;
+/* Set on the first BD descriptor when there is a split BD */
+#define BNX2X_TSO_SPLIT_BD (1<<0)
};
struct sw_rx_page {
@@ -149,6 +152,11 @@
DECLARE_PCI_UNMAP_ADDR(mapping)
};
+union db_prod {
+ struct doorbell_set_prod data;
+ u32 raw;
+};
+
/* MC hsi */
#define BCM_PAGE_SHIFT 12
@@ -234,15 +242,14 @@
struct napi_struct napi;
+ u8 is_rx_queue;
+
struct host_status_block *status_blk;
dma_addr_t status_blk_mapping;
- struct eth_tx_db_data *hw_tx_prods;
- dma_addr_t tx_prods_mapping;
-
struct sw_tx_bd *tx_buf_ring;
- struct eth_tx_bd *tx_desc_ring;
+ union eth_tx_bd_types *tx_desc_ring;
dma_addr_t tx_desc_mapping;
struct sw_rx_bd *rx_buf_ring; /* BDs mappings ring */
@@ -272,6 +279,8 @@
u8 cl_id; /* eth client id */
u8 sb_id; /* status block number in HW */
+ union db_prod tx_db;
+
u16 tx_pkt_prod;
u16 tx_pkt_cons;
u16 tx_bd_prod;
@@ -309,21 +318,24 @@
struct xstorm_per_client_stats old_xclient;
struct bnx2x_eth_q_stats eth_q_stats;
- char name[IFNAMSIZ];
+ /* The size is calculated using the following:
+ sizeof name field from netdev structure +
+ 4 ('-Xx-' string) +
+ 4 (for the digits and to make it DWORD aligned) */
+#define FP_NAME_SIZE (sizeof(((struct net_device *)0)->name) + 8)
+ char name[FP_NAME_SIZE];
struct bnx2x *bp; /* parent */
};
#define bnx2x_fp(bp, nr, var) (bp->fp[nr].var)
-#define BNX2X_HAS_WORK(fp) (bnx2x_has_rx_work(fp) || bnx2x_has_tx_work(fp))
-
/* MC hsi */
#define MAX_FETCH_BD 13 /* HW max BDs per packet */
#define RX_COPY_THRESH 92
#define NUM_TX_RINGS 16
-#define TX_DESC_CNT (BCM_PAGE_SIZE / sizeof(struct eth_tx_bd))
+#define TX_DESC_CNT (BCM_PAGE_SIZE / sizeof(union eth_tx_bd_types))
#define MAX_TX_DESC_CNT (TX_DESC_CNT - 1)
#define NUM_TX_BD (TX_DESC_CNT * NUM_TX_RINGS)
#define MAX_TX_BD (NUM_TX_BD - 1)
@@ -395,7 +407,7 @@
#define DPM_TRIGER_TYPE 0x40
#define DOORBELL(bp, cid, val) \
do { \
- writel((u32)val, (bp)->doorbells + (BCM_PAGE_SIZE * cid) + \
+ writel((u32)(val), bp->doorbells + (BCM_PAGE_SIZE * (cid)) + \
DPM_TRIGER_TYPE); \
} while (0)
@@ -902,8 +914,6 @@
u16 rx_quick_cons_trip;
u16 rx_ticks_int;
u16 rx_ticks;
-/* Maximal coalescing timeout in us */
-#define BNX2X_MAX_COALESCE_TOUT (0xf0*12)
u32 lin_cnt;
@@ -985,19 +995,20 @@
};
-#define BNX2X_MAX_QUEUES(bp) (IS_E1HMF(bp) ? (MAX_CONTEXT / E1HVN_MAX) : \
- MAX_CONTEXT)
-#define BNX2X_NUM_QUEUES(bp) max(bp->num_rx_queues, bp->num_tx_queues)
-#define is_multi(bp) (BNX2X_NUM_QUEUES(bp) > 1)
+#define BNX2X_MAX_QUEUES(bp) (IS_E1HMF(bp) ? (MAX_CONTEXT/(2 * E1HVN_MAX)) \
+ : (MAX_CONTEXT/2))
+#define BNX2X_NUM_QUEUES(bp) (bp->num_rx_queues + bp->num_tx_queues)
+#define is_multi(bp) (BNX2X_NUM_QUEUES(bp) > 2)
#define for_each_rx_queue(bp, var) \
for (var = 0; var < bp->num_rx_queues; var++)
#define for_each_tx_queue(bp, var) \
- for (var = 0; var < bp->num_tx_queues; var++)
+ for (var = bp->num_rx_queues; \
+ var < BNX2X_NUM_QUEUES(bp); var++)
#define for_each_queue(bp, var) \
for (var = 0; var < BNX2X_NUM_QUEUES(bp); var++)
#define for_each_nondefault_queue(bp, var) \
- for (var = 1; var < BNX2X_NUM_QUEUES(bp); var++)
+ for (var = 1; var < bp->num_rx_queues; var++)
void bnx2x_read_dmae(struct bnx2x *bp, u32 src_addr, u32 len32);