bnx2x: Multi-queue
authorEilon Greenstein <eilong@broadcom.com>
Thu, 12 Feb 2009 08:36:11 +0000 (08:36 +0000)
committerDavid S. Miller <davem@davemloft.net>
Mon, 16 Feb 2009 07:31:05 +0000 (23:31 -0800)
Adding Tx multi-queue and enabling multi-queue by default

Signed-off-by: Eilon Greenstein <eilong@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
drivers/net/bnx2x.h
drivers/net/bnx2x_fw_defs.h
drivers/net/bnx2x_main.c

index de094d4b68a085381ecc7c3657b41ab3887eeb85..7edad1f9b3309b555b76c73b0429c4bdbb305326 100644 (file)
 #endif
 
 
+#define BNX2X_MULTI_QUEUE
+
+#define BNX2X_NEW_NAPI
+
 /* error/debug prints */
 
 #define DRV_MODULE_NAME                "bnx2x"
@@ -266,6 +270,7 @@ struct bnx2x_fastpath {
        u64                     tpa_queue_used;
 #endif
 
+       char                    name[IFNAMSIZ];
        struct bnx2x            *bp; /* parent */
 };
 
@@ -680,11 +685,7 @@ struct bnx2x_eth_stats {
                        (offsetof(struct bnx2x_eth_stats, stat_name) / 4)
 
 
-#ifdef BNX2X_MULTI
 #define MAX_CONTEXT                    16
-#else
-#define MAX_CONTEXT                    1
-#endif
 
 union cdu_context {
        struct eth_context eth;
@@ -859,8 +860,9 @@ struct bnx2x {
 #define BNX2X_STATE_DIAG               0xe000
 #define BNX2X_STATE_ERROR              0xf000
 
-       int                     num_queues;
-#define BP_MAX_QUEUES(bp)              (IS_E1HMF(bp) ? 4 : 16)
+       int                     multi_mode;
+       int                     num_rx_queues;
+       int                     num_tx_queues;
 
        u32                     rx_mode;
 #define BNX2X_RX_MODE_NONE             0
@@ -911,11 +913,19 @@ struct bnx2x {
 };
 
 
-#define for_each_queue(bp, var)        for (var = 0; var < bp->num_queues; var++)
+#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 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++)
+#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 < bp->num_queues; var++)
-#define is_multi(bp)           (bp->num_queues > 1)
+                       for (var = 1; var < BNX2X_NUM_QUEUES(bp); var++)
 
 
 void bnx2x_read_dmae(struct bnx2x *bp, u32 src_addr, u32 len32);
@@ -1120,12 +1130,13 @@ static inline u32 reg_poll(struct bnx2x *bp, u32 reg, u32 expected, int ms,
                                 AEU_INPUTS_ATTN_BITS_MISC_PARITY_ERROR)
 
 
-#define MULTI_FLAGS \
+#define MULTI_FLAGS(bp) \
                (TSTORM_ETH_FUNCTION_COMMON_CONFIG_RSS_IPV4_CAPABILITY | \
                 TSTORM_ETH_FUNCTION_COMMON_CONFIG_RSS_IPV4_TCP_CAPABILITY | \
                 TSTORM_ETH_FUNCTION_COMMON_CONFIG_RSS_IPV6_CAPABILITY | \
                 TSTORM_ETH_FUNCTION_COMMON_CONFIG_RSS_IPV6_TCP_CAPABILITY | \
-                TSTORM_ETH_FUNCTION_COMMON_CONFIG_DEFAULT_ENABLE)
+                (bp->multi_mode << \
+                 TSTORM_ETH_FUNCTION_COMMON_CONFIG_RSS_MODE_SHIFT))
 
 #define MULTI_MASK                     0x7f
 
index 2fe14a25ea3ee1d2fe72985fb13cb796236d51c8..9eb1d131179cd04ab58bbbccce9eca921ef5fd08 100644 (file)
 #define ETH_MAX_AGGREGATION_QUEUES_E1 (32)
 #define ETH_MAX_AGGREGATION_QUEUES_E1H (64)
 
+/* ETH RSS modes */
+#define ETH_RSS_MODE_DISABLED 0
+#define ETH_RSS_MODE_REGULAR 1
+
 
 /**
 * This file defines HSI constatnts common to all microcode flows
index 3ca9c969a6889f06e544c29126b88f3cef56be2c..60762f769efc038a3b8a1f04e2cd0ec53019e844 100644 (file)
@@ -73,12 +73,14 @@ MODULE_DESCRIPTION("Broadcom NetXtreme II BCM57710/57711/57711E Driver");
 MODULE_LICENSE("GPL");
 MODULE_VERSION(DRV_MODULE_VERSION);
 
+static int multi_mode = 1;
+module_param(multi_mode, int, 0);
+
 static int disable_tpa;
 static int use_inta;
 static int poll;
 static int debug;
 static int load_count[3]; /* 0-common, 1-port0, 2-port1 */
-static int use_multi;
 
 module_param(disable_tpa, int, 0);
 module_param(use_inta, int, 0);
@@ -89,10 +91,6 @@ MODULE_PARM_DESC(use_inta, "use INT#A instead of MSI-X");
 MODULE_PARM_DESC(poll, "use polling (for debug)");
 MODULE_PARM_DESC(debug, "default debug msglevel");
 
-#ifdef BNX2X_MULTI
-module_param(use_multi, int, 0);
-MODULE_PARM_DESC(use_multi, "use per-CPU queues");
-#endif
 static struct workqueue_struct *bnx2x_wq;
 
 enum bnx2x_board_type {
@@ -845,6 +843,7 @@ static inline u16 bnx2x_tx_avail(struct bnx2x_fastpath *fp)
 static void bnx2x_tx_int(struct bnx2x_fastpath *fp, int work)
 {
        struct bnx2x *bp = fp->bp;
+       struct netdev_queue *txq;
        u16 hw_cons, sw_cons, bd_cons = fp->tx_bd_cons;
        int done = 0;
 
@@ -853,6 +852,7 @@ static void bnx2x_tx_int(struct bnx2x_fastpath *fp, int work)
                return;
 #endif
 
+       txq = netdev_get_tx_queue(bp->dev, fp->index);
        hw_cons = le16_to_cpu(*fp->tx_cons_sb);
        sw_cons = fp->tx_pkt_cons;
 
@@ -882,24 +882,24 @@ static void bnx2x_tx_int(struct bnx2x_fastpath *fp, int work)
        fp->tx_pkt_cons = sw_cons;
        fp->tx_bd_cons = bd_cons;
 
-       /* Need to make the tx_cons update visible to start_xmit()
-        * before checking for netif_queue_stopped().  Without the
+       /* Need to make the tx_bd_cons update visible to start_xmit()
+        * before checking for netif_tx_queue_stopped().  Without the
         * memory barrier, there is a small possibility that start_xmit()
         * will miss it and cause the queue to be stopped forever.
         */
        smp_mb();
 
        /* TBD need a thresh? */
-       if (unlikely(netif_queue_stopped(bp->dev))) {
+       if (unlikely(netif_tx_queue_stopped(txq))) {
 
-               netif_tx_lock(bp->dev);
+               __netif_tx_lock(txq, smp_processor_id());
 
-               if (netif_queue_stopped(bp->dev) &&
+               if ((netif_tx_queue_stopped(txq)) &&
                    (bp->state == BNX2X_STATE_OPEN) &&
                    (bnx2x_tx_avail(fp) >= MAX_SKB_FRAGS + 3))
-                       netif_wake_queue(bp->dev);
+                       netif_tx_wake_queue(txq);
 
-               netif_tx_unlock(bp->dev);
+               __netif_tx_unlock(txq);
        }
 }
 
@@ -1403,8 +1403,8 @@ static inline void bnx2x_update_rx_prod(struct bnx2x *bp,
        mmiowb(); /* keep prod updates ordered */
 
        DP(NETIF_MSG_RX_STATUS,
-          "Wrote: bd_prod %u  cqe_prod %u  sge_prod %u\n",
-          bd_prod, rx_comp_prod, rx_sge_prod);
+          "queue[%d]:  wrote  bd_prod %u  cqe_prod %u  sge_prod %u\n",
+          fp->index, bd_prod, rx_comp_prod, rx_sge_prod);
 }
 
 static int bnx2x_rx_int(struct bnx2x_fastpath *fp, int budget)
@@ -1662,8 +1662,7 @@ static irqreturn_t bnx2x_msix_fp_int(int irq, void *fp_cookie)
 
 static irqreturn_t bnx2x_interrupt(int irq, void *dev_instance)
 {
-       struct net_device *dev = dev_instance;
-       struct bnx2x *bp = netdev_priv(dev);
+       struct bnx2x *bp = netdev_priv(dev_instance);
        u16 status = bnx2x_ack_int(bp);
        u16 mask;
 
@@ -4295,7 +4294,7 @@ static void bnx2x_init_rx_rings(struct bnx2x *bp)
                   "rx_buf_size %d  effective_mtu %d\n",
                   bp->rx_buf_size, bp->dev->mtu + ETH_OVREHEAD);
 
-               for_each_queue(bp, j) {
+               for_each_rx_queue(bp, j) {
                        struct bnx2x_fastpath *fp = &bp->fp[j];
 
                        for (i = 0; i < max_agg_queues; i++) {
@@ -4318,7 +4317,7 @@ static void bnx2x_init_rx_rings(struct bnx2x *bp)
                }
        }
 
-       for_each_queue(bp, j) {
+       for_each_rx_queue(bp, j) {
                struct bnx2x_fastpath *fp = &bp->fp[j];
 
                fp->rx_bd_cons = 0;
@@ -4430,7 +4429,7 @@ static void bnx2x_init_tx_ring(struct bnx2x *bp)
 {
        int i, j;
 
-       for_each_queue(bp, j) {
+       for_each_tx_queue(bp, j) {
                struct bnx2x_fastpath *fp = &bp->fp[j];
 
                for (i = 1; i <= NUM_TX_RINGS; i++) {
@@ -4543,14 +4542,15 @@ static void bnx2x_init_ind_table(struct bnx2x *bp)
        int func = BP_FUNC(bp);
        int i;
 
-       if (!is_multi(bp))
+       if (bp->multi_mode == ETH_RSS_MODE_DISABLED)
                return;
 
-       DP(NETIF_MSG_IFUP, "Initializing indirection table\n");
+       DP(NETIF_MSG_IFUP,
+          "Initializing indirection table  multi_mode %d\n", bp->multi_mode);
        for (i = 0; i < TSTORM_INDIRECTION_TABLE_SIZE; i++)
                REG_WR8(bp, BAR_TSTRORM_INTMEM +
                        TSTORM_INDIRECTION_TABLE_OFFSET(func) + i,
-                       BP_CL_ID(bp) + (i % bp->num_queues));
+                       BP_CL_ID(bp) + (i % bp->num_rx_queues));
 }
 
 static void bnx2x_set_client_config(struct bnx2x *bp)
@@ -4684,7 +4684,7 @@ static void bnx2x_init_internal_func(struct bnx2x *bp)
        u16 max_agg_size;
 
        if (is_multi(bp)) {
-               tstorm_config.config_flags = MULTI_FLAGS;
+               tstorm_config.config_flags = MULTI_FLAGS(bp);
                tstorm_config.rss_result_mask = MULTI_MASK;
        }
        if (IS_E1HMF(bp))
@@ -4764,7 +4764,7 @@ static void bnx2x_init_internal_func(struct bnx2x *bp)
                min((u32)(min((u32)8, (u32)MAX_SKB_FRAGS) *
                          SGE_PAGE_SIZE * PAGES_PER_SGE),
                    (u32)0xffff);
-       for_each_queue(bp, i) {
+       for_each_rx_queue(bp, i) {
                struct bnx2x_fastpath *fp = &bp->fp[i];
 
                REG_WR(bp, BAR_USTRORM_INTMEM +
@@ -5809,20 +5809,19 @@ static void bnx2x_free_mem(struct bnx2x *bp)
        int i;
 
        /* fastpath */
+       /* Common */
        for_each_queue(bp, i) {
 
-               /* Status blocks */
+               /* status blocks */
                BNX2X_PCI_FREE(bnx2x_fp(bp, i, status_blk),
                               bnx2x_fp(bp, i, status_blk_mapping),
                               sizeof(struct host_status_block) +
                               sizeof(struct eth_tx_db_data));
+       }
+       /* Rx */
+       for_each_rx_queue(bp, i) {
 
-               /* fast path rings: tx_buf tx_desc rx_buf rx_desc rx_comp */
-               BNX2X_FREE(bnx2x_fp(bp, i, tx_buf_ring));
-               BNX2X_PCI_FREE(bnx2x_fp(bp, i, tx_desc_ring),
-                              bnx2x_fp(bp, i, tx_desc_mapping),
-                              sizeof(struct eth_tx_bd) * NUM_TX_BD);
-
+               /* fastpath rx rings: rx_buf rx_desc rx_comp */
                BNX2X_FREE(bnx2x_fp(bp, i, rx_buf_ring));
                BNX2X_PCI_FREE(bnx2x_fp(bp, i, rx_desc_ring),
                               bnx2x_fp(bp, i, rx_desc_mapping),
@@ -5839,6 +5838,15 @@ static void bnx2x_free_mem(struct bnx2x *bp)
                               bnx2x_fp(bp, i, rx_sge_mapping),
                               BCM_PAGE_SIZE * NUM_RX_SGE_PAGES);
        }
+       /* Tx */
+       for_each_tx_queue(bp, i) {
+
+               /* fastpath tx rings: tx_buf tx_desc */
+               BNX2X_FREE(bnx2x_fp(bp, i, tx_buf_ring));
+               BNX2X_PCI_FREE(bnx2x_fp(bp, i, tx_desc_ring),
+                              bnx2x_fp(bp, i, tx_desc_mapping),
+                              sizeof(struct eth_tx_bd) * NUM_TX_BD);
+       }
        /* end of fastpath */
 
        BNX2X_PCI_FREE(bp->def_status_blk, bp->def_status_blk_mapping,
@@ -5881,29 +5889,20 @@ static int bnx2x_alloc_mem(struct bnx2x *bp)
        int i;
 
        /* fastpath */
+       /* Common */
        for_each_queue(bp, i) {
                bnx2x_fp(bp, i, bp) = bp;
 
-               /* Status blocks */
+               /* status blocks */
                BNX2X_PCI_ALLOC(bnx2x_fp(bp, i, status_blk),
                                &bnx2x_fp(bp, i, status_blk_mapping),
                                sizeof(struct host_status_block) +
                                sizeof(struct eth_tx_db_data));
+       }
+       /* Rx */
+       for_each_rx_queue(bp, i) {
 
-               bnx2x_fp(bp, i, hw_tx_prods) =
-                               (void *)(bnx2x_fp(bp, i, status_blk) + 1);
-
-               bnx2x_fp(bp, i, tx_prods_mapping) =
-                               bnx2x_fp(bp, i, status_blk_mapping) +
-                               sizeof(struct host_status_block);
-
-               /* fast path rings: tx_buf tx_desc rx_buf rx_desc rx_comp */
-               BNX2X_ALLOC(bnx2x_fp(bp, i, tx_buf_ring),
-                               sizeof(struct sw_tx_bd) * NUM_TX_BD);
-               BNX2X_PCI_ALLOC(bnx2x_fp(bp, i, tx_desc_ring),
-                               &bnx2x_fp(bp, i, tx_desc_mapping),
-                               sizeof(struct eth_tx_bd) * NUM_TX_BD);
-
+               /* fastpath rx rings: rx_buf rx_desc rx_comp */
                BNX2X_ALLOC(bnx2x_fp(bp, i, rx_buf_ring),
                                sizeof(struct sw_rx_bd) * NUM_RX_BD);
                BNX2X_PCI_ALLOC(bnx2x_fp(bp, i, rx_desc_ring),
@@ -5922,6 +5921,23 @@ static int bnx2x_alloc_mem(struct bnx2x *bp)
                                &bnx2x_fp(bp, i, rx_sge_mapping),
                                BCM_PAGE_SIZE * NUM_RX_SGE_PAGES);
        }
+       /* Tx */
+       for_each_tx_queue(bp, i) {
+
+               bnx2x_fp(bp, i, hw_tx_prods) =
+                               (void *)(bnx2x_fp(bp, i, status_blk) + 1);
+
+               bnx2x_fp(bp, i, tx_prods_mapping) =
+                               bnx2x_fp(bp, i, status_blk_mapping) +
+                               sizeof(struct host_status_block);
+
+               /* fastpath tx rings: tx_buf tx_desc */
+               BNX2X_ALLOC(bnx2x_fp(bp, i, tx_buf_ring),
+                               sizeof(struct sw_tx_bd) * NUM_TX_BD);
+               BNX2X_PCI_ALLOC(bnx2x_fp(bp, i, tx_desc_ring),
+                               &bnx2x_fp(bp, i, tx_desc_mapping),
+                               sizeof(struct eth_tx_bd) * NUM_TX_BD);
+       }
        /* end of fastpath */
 
        BNX2X_PCI_ALLOC(bp->def_status_blk, &bp->def_status_blk_mapping,
@@ -5975,7 +5991,7 @@ static void bnx2x_free_tx_skbs(struct bnx2x *bp)
 {
        int i;
 
-       for_each_queue(bp, i) {
+       for_each_tx_queue(bp, i) {
                struct bnx2x_fastpath *fp = &bp->fp[i];
 
                u16 bd_cons = fp->tx_bd_cons;
@@ -5993,7 +6009,7 @@ static void bnx2x_free_rx_skbs(struct bnx2x *bp)
 {
        int i, j;
 
-       for_each_queue(bp, j) {
+       for_each_rx_queue(bp, j) {
                struct bnx2x_fastpath *fp = &bp->fp[j];
 
                for (i = 0; i < NUM_RX_BD; i++) {
@@ -6073,7 +6089,7 @@ static int bnx2x_enable_msix(struct bnx2x *bp)
        }
 
        rc = pci_enable_msix(bp->pdev, &bp->msix_table[0],
-                            bp->num_queues + offset);
+                            BNX2X_NUM_QUEUES(bp) + offset);
        if (rc) {
                DP(NETIF_MSG_IFUP, "MSI-X is not attainable\n");
                return -1;
@@ -6095,19 +6111,32 @@ static int bnx2x_req_msix_irqs(struct bnx2x *bp)
        }
 
        for_each_queue(bp, i) {
+               struct bnx2x_fastpath *fp = &bp->fp[i];
+
+               sprintf(fp->name, "%s.fp%d", bp->dev->name, i);
                rc = request_irq(bp->msix_table[i + offset].vector,
-                                bnx2x_msix_fp_int, 0,
-                                bp->dev->name, &bp->fp[i]);
+                                bnx2x_msix_fp_int, 0, fp->name, fp);
                if (rc) {
-                       BNX2X_ERR("request fp #%d irq failed  rc -%d\n",
-                                 i + offset, -rc);
+                       BNX2X_ERR("request fp #%d irq failed  rc %d\n", i, rc);
                        bnx2x_free_msix_irqs(bp);
                        return -EBUSY;
                }
 
-               bnx2x_fp(bp, i, state) = BNX2X_FP_STATE_IRQ;
+               fp->state = BNX2X_FP_STATE_IRQ;
        }
 
+       i = BNX2X_NUM_QUEUES(bp);
+       if (is_multi(bp))
+               printk(KERN_INFO PFX
+                      "%s: using MSI-X  IRQs: sp %d  fp %d - %d\n",
+                      bp->dev->name, bp->msix_table[0].vector,
+                      bp->msix_table[offset].vector,
+                      bp->msix_table[offset + i - 1].vector);
+       else
+               printk(KERN_INFO PFX "%s: using MSI-X  IRQs: sp %d  fp %d\n",
+                      bp->dev->name, bp->msix_table[0].vector,
+                      bp->msix_table[offset + i - 1].vector);
+
        return 0;
 }
 
@@ -6127,7 +6156,7 @@ static void bnx2x_napi_enable(struct bnx2x *bp)
 {
        int i;
 
-       for_each_queue(bp, i)
+       for_each_rx_queue(bp, i)
                napi_enable(&bnx2x_fp(bp, i, napi));
 }
 
@@ -6135,7 +6164,7 @@ static void bnx2x_napi_disable(struct bnx2x *bp)
 {
        int i;
 
-       for_each_queue(bp, i)
+       for_each_rx_queue(bp, i)
                napi_disable(&bnx2x_fp(bp, i, napi));
 }
 
@@ -6143,10 +6172,10 @@ static void bnx2x_netif_start(struct bnx2x *bp)
 {
        if (atomic_dec_and_test(&bp->intr_sem)) {
                if (netif_running(bp->dev)) {
-                       if (bp->state == BNX2X_STATE_OPEN)
-                               netif_wake_queue(bp->dev);
                        bnx2x_napi_enable(bp);
                        bnx2x_int_enable(bp);
+                       if (bp->state == BNX2X_STATE_OPEN)
+                               netif_tx_wake_all_queues(bp->dev);
                }
        }
 }
@@ -6320,16 +6349,19 @@ static int bnx2x_setup_leading(struct bnx2x *bp)
 
 static int bnx2x_setup_multi(struct bnx2x *bp, int index)
 {
+       struct bnx2x_fastpath *fp = &bp->fp[index];
+
        /* reset IGU state */
-       bnx2x_ack_sb(bp, bp->fp[index].sb_id, CSTORM_ID, 0, IGU_INT_ENABLE, 0);
+       bnx2x_ack_sb(bp, fp->sb_id, CSTORM_ID, 0, IGU_INT_ENABLE, 0);
 
        /* SETUP ramrod */
-       bp->fp[index].state = BNX2X_FP_STATE_OPENING;
-       bnx2x_sp_post(bp, RAMROD_CMD_ID_ETH_CLIENT_SETUP, index, 0, index, 0);
+       fp->state = BNX2X_FP_STATE_OPENING;
+       bnx2x_sp_post(bp, RAMROD_CMD_ID_ETH_CLIENT_SETUP, index, 0,
+                     fp->cl_id, 0);
 
        /* Wait for completion */
        return bnx2x_wait_ramrod(bp, BNX2X_FP_STATE_OPEN, index,
-                                &(bp->fp[index].state), 0);
+                                &(fp->state), 0);
 }
 
 static int bnx2x_poll(struct napi_struct *napi, int budget);
@@ -6340,6 +6372,7 @@ static int bnx2x_nic_load(struct bnx2x *bp, int load_mode)
 {
        u32 load_code;
        int i, rc = 0;
+       int num_queues;
 #ifdef BNX2X_STOP_ON_ERROR
        if (unlikely(bp->panic))
                return -EPERM;
@@ -6348,22 +6381,22 @@ static int bnx2x_nic_load(struct bnx2x *bp, int load_mode)
        bp->state = BNX2X_STATE_OPENING_WAIT4_LOAD;
 
        if (use_inta) {
-               bp->num_queues = 1;
-
+               num_queues = 1;
+               bp->num_rx_queues = num_queues;
+               bp->num_tx_queues = num_queues;
+               DP(NETIF_MSG_IFUP,
+                  "set number of queues to %d\n", num_queues);
        } else {
-               if ((use_multi > 1) && (use_multi <= BP_MAX_QUEUES(bp)))
-                       /* user requested number */
-                       bp->num_queues = use_multi;
-
-               else if (use_multi)
-                       bp->num_queues = min_t(u32, num_online_cpus(),
-                                              BP_MAX_QUEUES(bp));
+               if (bp->multi_mode == ETH_RSS_MODE_REGULAR)
+                       num_queues = min_t(u32, num_online_cpus(),
+                                          BNX2X_MAX_QUEUES(bp));
                else
-                       bp->num_queues = 1;
-
-               DP(NETIF_MSG_IFUP,
-                  "set number of queues to %d\n", bp->num_queues);
-
+                       num_queues = 1;
+               bp->num_rx_queues = num_queues;
+               bp->num_tx_queues = num_queues;
+               DP(NETIF_MSG_IFUP, "set number of rx queues to %d"
+                  "  number of tx queues to %d\n",
+                  bp->num_rx_queues, bp->num_tx_queues);
                /* if we can't use MSI-X we only need one fp,
                 * so try to enable MSI-X with the requested number of fp's
                 * and fallback to MSI or legacy INTx with one fp
@@ -6371,26 +6404,30 @@ static int bnx2x_nic_load(struct bnx2x *bp, int load_mode)
                rc = bnx2x_enable_msix(bp);
                if (rc) {
                        /* failed to enable MSI-X */
-                       bp->num_queues = 1;
-                       if (use_multi)
-                               BNX2X_ERR("Multi requested but failed"
-                                         " to enable MSI-X\n");
+                       num_queues = 1;
+                       bp->num_rx_queues = num_queues;
+                       bp->num_tx_queues = num_queues;
+                       if (bp->multi_mode)
+                               BNX2X_ERR("Multi requested but failed to "
+                                         "enable MSI-X  set number of "
+                                         "queues to %d\n", num_queues);
                }
        }
+       bp->dev->real_num_tx_queues = bp->num_tx_queues;
 
        if (bnx2x_alloc_mem(bp))
                return -ENOMEM;
 
-       for_each_queue(bp, i)
+       for_each_rx_queue(bp, i)
                bnx2x_fp(bp, i, disable_tpa) =
                                        ((bp->flags & TPA_ENABLE_FLAG) == 0);
 
-       for_each_queue(bp, i)
+       for_each_rx_queue(bp, i)
                netif_napi_add(bp->dev, &bnx2x_fp(bp, i, napi),
                               bnx2x_poll, 128);
 
 #ifdef BNX2X_STOP_ON_ERROR
-       for_each_queue(bp, i) {
+       for_each_rx_queue(bp, i) {
                struct bnx2x_fastpath *fp = &bp->fp[i];
 
                fp->poll_no_work = 0;
@@ -6512,13 +6549,13 @@ static int bnx2x_nic_load(struct bnx2x *bp, int load_mode)
        switch (load_mode) {
        case LOAD_NORMAL:
                /* Tx queue should be only reenabled */
-               netif_wake_queue(bp->dev);
+               netif_tx_wake_all_queues(bp->dev);
                /* Initialize the receive filter. */
                bnx2x_set_rx_mode(bp->dev);
                break;
 
        case LOAD_OPEN:
-               netif_start_queue(bp->dev);
+               netif_tx_start_all_queues(bp->dev);
                /* Initialize the receive filter. */
                bnx2x_set_rx_mode(bp->dev);
                break;
@@ -6551,14 +6588,14 @@ load_error3:
        bp->port.pmf = 0;
        /* Free SKBs, SGEs, TPA pool and driver internals */
        bnx2x_free_skbs(bp);
-       for_each_queue(bp, i)
+       for_each_rx_queue(bp, i)
                bnx2x_free_rx_sge_range(bp, bp->fp + i, NUM_RX_SGE);
 load_error2:
        /* Release IRQs */
        bnx2x_free_irq(bp);
 load_error1:
        bnx2x_napi_disable(bp);
-       for_each_queue(bp, i)
+       for_each_rx_queue(bp, i)
                netif_napi_del(&bnx2x_fp(bp, i, napi));
        bnx2x_free_mem(bp);
 
@@ -6569,15 +6606,16 @@ load_error1:
 
 static int bnx2x_stop_multi(struct bnx2x *bp, int index)
 {
+       struct bnx2x_fastpath *fp = &bp->fp[index];
        int rc;
 
        /* halt the connection */
-       bp->fp[index].state = BNX2X_FP_STATE_HALTING;
-       bnx2x_sp_post(bp, RAMROD_CMD_ID_ETH_HALT, index, 0, index, 0);
+       fp->state = BNX2X_FP_STATE_HALTING;
+       bnx2x_sp_post(bp, RAMROD_CMD_ID_ETH_HALT, index, 0, fp->cl_id, 0);
 
        /* Wait for completion */
        rc = bnx2x_wait_ramrod(bp, BNX2X_FP_STATE_HALTED, index,
-                              &(bp->fp[index].state), 1);
+                              &(fp->state), 1);
        if (rc) /* timeout */
                return rc;
 
@@ -6586,7 +6624,7 @@ static int bnx2x_stop_multi(struct bnx2x *bp, int index)
 
        /* Wait for completion */
        rc = bnx2x_wait_ramrod(bp, BNX2X_FP_STATE_CLOSED, index,
-                              &(bp->fp[index].state), 1);
+                              &(fp->state), 1);
        return rc;
 }
 
@@ -6734,8 +6772,8 @@ static int bnx2x_nic_unload(struct bnx2x *bp, int unload_mode)
        /* Release IRQs */
        bnx2x_free_irq(bp);
 
-       /* Wait until tx fast path tasks complete */
-       for_each_queue(bp, i) {
+       /* Wait until tx fastpath tasks complete */
+       for_each_tx_queue(bp, i) {
                struct bnx2x_fastpath *fp = &bp->fp[i];
 
                cnt = 1000;
@@ -6867,9 +6905,9 @@ unload_error:
 
        /* Free SKBs, SGEs, TPA pool and driver internals */
        bnx2x_free_skbs(bp);
-       for_each_queue(bp, i)
+       for_each_rx_queue(bp, i)
                bnx2x_free_rx_sge_range(bp, bp->fp + i, NUM_RX_SGE);
-       for_each_queue(bp, i)
+       for_each_rx_queue(bp, i)
                netif_napi_del(&bnx2x_fp(bp, i, napi));
        bnx2x_free_mem(bp);
 
@@ -7597,6 +7635,15 @@ static int __devinit bnx2x_init_bp(struct bnx2x *bp)
                printk(KERN_ERR PFX
                       "MCP disabled, must load devices in order!\n");
 
+       /* Set multi queue mode */
+       if ((multi_mode != ETH_RSS_MODE_DISABLED) && (!use_inta)) {
+               printk(KERN_ERR PFX
+                     "Multi disabled since INTA is requested\n");
+               multi_mode = ETH_RSS_MODE_DISABLED;
+       }
+       bp->multi_mode = multi_mode;
+
+
        /* Set TPA flags */
        if (disable_tpa) {
                bp->flags &= ~TPA_ENABLE_FLAG;
@@ -9569,6 +9616,7 @@ static int bnx2x_start_xmit(struct sk_buff *skb, struct net_device *dev)
 {
        struct bnx2x *bp = netdev_priv(dev);
        struct bnx2x_fastpath *fp;
+       struct netdev_queue *txq;
        struct sw_tx_bd *tx_buf;
        struct eth_tx_bd *tx_bd;
        struct eth_tx_parse_bd *pbd = NULL;
@@ -9585,12 +9633,14 @@ static int bnx2x_start_xmit(struct sk_buff *skb, struct net_device *dev)
                return NETDEV_TX_BUSY;
 #endif
 
-       fp_index = (smp_processor_id() % bp->num_queues);
+       fp_index = skb_get_queue_mapping(skb);
+       txq = netdev_get_tx_queue(dev, fp_index);
+
        fp = &bp->fp[fp_index];
 
        if (unlikely(bnx2x_tx_avail(fp) < (skb_shinfo(skb)->nr_frags + 3))) {
                bp->eth_stats.driver_xoff++,
-               netif_stop_queue(dev);
+               netif_tx_stop_queue(txq);
                BNX2X_ERR("BUG! Tx ring full when queue awake!\n");
                return NETDEV_TX_BUSY;
        }
@@ -9829,10 +9879,10 @@ static int bnx2x_start_xmit(struct sk_buff *skb, struct net_device *dev)
                /* We want bnx2x_tx_int to "see" the updated tx_bd_prod
                   if we put Tx into XOFF state. */
                smp_mb();
-               netif_stop_queue(dev);
+               netif_tx_stop_queue(txq);
                bp->eth_stats.driver_xoff++;
                if (bnx2x_tx_avail(fp) >= MAX_SKB_FRAGS + 3)
-                       netif_wake_queue(dev);
+                       netif_tx_wake_queue(txq);
        }
        fp->tx_pkt++;
 
@@ -10324,7 +10374,7 @@ static int __devinit bnx2x_init_one(struct pci_dev *pdev,
                printk(KERN_INFO "%s", version);
 
        /* dev zeroed in init_etherdev */
-       dev = alloc_etherdev(sizeof(*bp));
+       dev = alloc_etherdev_mq(sizeof(*bp), MAX_CONTEXT);
        if (!dev) {
                printk(KERN_ERR PFX "Cannot allocate net device\n");
                return -ENOMEM;
@@ -10496,9 +10546,9 @@ static int bnx2x_eeh_nic_unload(struct bnx2x *bp)
 
        /* Free SKBs, SGEs, TPA pool and driver internals */
        bnx2x_free_skbs(bp);
-       for_each_queue(bp, i)
+       for_each_rx_queue(bp, i)
                bnx2x_free_rx_sge_range(bp, bp->fp + i, NUM_RX_SGE);
-       for_each_queue(bp, i)
+       for_each_rx_queue(bp, i)
                netif_napi_del(&bnx2x_fp(bp, i, napi));
        bnx2x_free_mem(bp);