From: mchristi@redhat.com <mchristi@redhat.com>
Date: Wed, 27 May 2009 16:34:05 -0500
Subject: [net] add cnic support to bnx2
Message-id: 1243460050-16253-1-git-send-email-mchristi@redhat.com
O-Subject: [PATCH 1/6] RHEL 5.4: add cnic (bnx2i iscsi) support to bnx2
Bugzilla: 441979

From: Mike Christie <mchristi@redhat.com>

This is for BZ 441979.

Add interface and functions to support a new CNIC driver to drive
the hardware for iSCSI offload.

The upstream version of this patch has been ackd
http://marc.info/?l=linux-netdev&m=124082705613349&w=2
but it has not gone in yet because of dependencies on
other patches that should be going in soon (this should
all go in 2.6.31 when that feature window opens).

I have done some basic testing by using it as the default nic
and running some netperfs and running software iscsi over it.

Broadcom had done and will be doing the majority of the testing.

diff --git a/drivers/net/bnx2.c b/drivers/net/bnx2.c
index 0067e8c..9ca4bcb 100644
--- a/drivers/net/bnx2.c
+++ b/drivers/net/bnx2.c
@@ -48,6 +48,10 @@
 #include <linux/cache.h>
 #include <linux/zlib.h>
 
+#if defined(CONFIG_CNIC) || defined(CONFIG_CNIC_MODULE)
+#define BCM_CNIC 1
+#include "cnic_if.h"
+#endif
 #include "bnx2.h"
 #include "bnx2_fw.h"
 #include "bnx2_fw2.h"
@@ -316,6 +320,158 @@ bnx2_ctx_wr(struct bnx2 *bp, u32 cid_addr, u32 offset, u32 val)
 	spin_unlock_bh(&bp->indirect_lock);
 }
 
+#ifdef BCM_CNIC
+static int
+bnx2_drv_ctl(struct net_device *dev, struct drv_ctl_info *info)
+{
+	struct bnx2 *bp = netdev_priv(dev);
+	struct drv_ctl_io *io = &info->data.io;
+
+	switch (info->cmd) {
+	case DRV_CTL_IO_WR_CMD:
+		bnx2_reg_wr_ind(bp, io->offset, io->data);
+		break;
+	case DRV_CTL_IO_RD_CMD:
+		io->data = bnx2_reg_rd_ind(bp, io->offset);
+		break;
+	case DRV_CTL_CTX_WR_CMD:
+		bnx2_ctx_wr(bp, io->cid_addr, io->offset, io->data);
+		break;
+	default:
+		return -EINVAL;
+	}
+	return 0;
+}
+
+static void bnx2_setup_cnic_irq_info(struct bnx2 *bp)
+{
+	struct cnic_eth_dev *cp = &bp->cnic_eth_dev;
+	struct bnx2_napi *bnapi = &bp->bnx2_napi[0];
+	int sb_id;
+
+	if (bp->flags & BNX2_FLAG_USING_MSIX) {
+		cp->drv_state |= CNIC_DRV_STATE_USING_MSIX;
+		bnapi->cnic_present = 0;
+		sb_id = bp->irq_nvecs;
+		cp->irq_arr[0].irq_flags |= CNIC_IRQ_FL_MSIX;
+	} else {
+		cp->drv_state &= ~CNIC_DRV_STATE_USING_MSIX;
+		bnapi->cnic_tag = bnapi->last_status_idx;
+		bnapi->cnic_present = 1;
+		sb_id = 0;
+		cp->irq_arr[0].irq_flags &= ~CNIC_IRQ_FL_MSIX;
+	}
+
+	cp->irq_arr[0].vector = bp->irq_tbl[sb_id].vector;
+	cp->irq_arr[0].status_blk = (void *)
+		((unsigned long) bnapi->status_blk.msi +
+		(BNX2_SBLK_MSIX_ALIGN_SIZE * sb_id));
+	cp->irq_arr[0].status_blk_num = sb_id;
+	cp->num_irq = 1;
+}
+
+static int bnx2_register_cnic(struct net_device *dev, struct cnic_ops *ops,
+			      void *data)
+{
+	struct bnx2 *bp = netdev_priv(dev);
+	struct cnic_eth_dev *cp = &bp->cnic_eth_dev;
+
+	if (ops == NULL)
+		return -EINVAL;
+
+	if (cp->drv_state & CNIC_DRV_STATE_REGD)
+		return -EBUSY;
+
+	bp->cnic_data = data;
+	rcu_assign_pointer(bp->cnic_ops, ops);
+
+	cp->num_irq = 0;
+	cp->drv_state = CNIC_DRV_STATE_REGD;
+
+	bnx2_setup_cnic_irq_info(bp);
+
+	return 0;
+}
+
+static int bnx2_unregister_cnic(struct net_device *dev)
+{
+	struct bnx2 *bp = netdev_priv(dev);
+	struct bnx2_napi *bnapi = &bp->bnx2_napi[0];
+	struct cnic_eth_dev *cp = &bp->cnic_eth_dev;
+
+	cp->drv_state = 0;
+	bnapi->cnic_present = 0;
+	rcu_assign_pointer(bp->cnic_ops, NULL);
+	synchronize_rcu();
+	return 0;
+}
+
+struct cnic_eth_dev *bnx2_cnic_probe(struct net_device *dev)
+{
+	struct bnx2 *bp = netdev_priv(dev);
+	struct cnic_eth_dev *cp = &bp->cnic_eth_dev;
+
+	cp->drv_owner = THIS_MODULE;
+	cp->chip_id = bp->chip_id;
+	cp->pdev = bp->pdev;
+	cp->io_base = bp->regview;
+	cp->drv_ctl = bnx2_drv_ctl;
+	cp->drv_register_cnic = bnx2_register_cnic;
+	cp->drv_unregister_cnic = bnx2_unregister_cnic;
+
+	return cp;
+}
+EXPORT_SYMBOL(bnx2_cnic_probe);
+
+static void
+bnx2_cnic_stop(struct bnx2 *bp)
+{
+	struct cnic_ops *c_ops;
+	struct cnic_ctl_info info;
+
+	rcu_read_lock();
+	c_ops = rcu_dereference(bp->cnic_ops);
+	if (c_ops) {
+		info.cmd = CNIC_CTL_STOP_CMD;
+		c_ops->cnic_ctl(bp->cnic_data, &info);
+	}
+	rcu_read_unlock();
+}
+
+static void
+bnx2_cnic_start(struct bnx2 *bp)
+{
+	struct cnic_ops *c_ops;
+	struct cnic_ctl_info info;
+
+	rcu_read_lock();
+	c_ops = rcu_dereference(bp->cnic_ops);
+	if (c_ops) {
+		if (!(bp->flags & BNX2_FLAG_USING_MSIX)) {
+			struct bnx2_napi *bnapi = &bp->bnx2_napi[0];
+
+			bnapi->cnic_tag = bnapi->last_status_idx;
+		}
+		info.cmd = CNIC_CTL_START_CMD;
+		c_ops->cnic_ctl(bp->cnic_data, &info);
+	}
+	rcu_read_unlock();
+}
+
+#else
+
+static void
+bnx2_cnic_stop(struct bnx2 *bp)
+{
+}
+
+static void
+bnx2_cnic_start(struct bnx2 *bp)
+{
+}
+
+#endif
+
 static int
 bnx2_read_phy(struct bnx2 *bp, u32 reg, u32 *val)
 {
@@ -489,6 +645,7 @@ bnx2_napi_enable(struct bnx2 *bp)
 static void
 bnx2_netif_stop(struct bnx2 *bp)
 {
+	bnx2_cnic_stop(bp);
 	bnx2_disable_int_sync(bp);
 	if (netif_running(bp->dev)) {
 		bnx2_napi_disable(bp);
@@ -505,6 +662,7 @@ bnx2_netif_start(struct bnx2 *bp)
 			netif_wake_queue(bp->dev);
 			bnx2_napi_enable(bp);
 			bnx2_enable_int(bp);
+			bnx2_cnic_start(bp);
 		}
 	}
 }
@@ -3158,6 +3316,11 @@ bnx2_has_work(struct bnx2_napi *bnapi)
 	if (bnx2_has_fast_work(bnapi))
 		return 1;
 
+#ifdef BCM_CNIC
+	if (bnapi->cnic_present && (bnapi->cnic_tag != sblk->status_idx))
+		return 1;
+#endif
+
 	if ((sblk->status_attn_bits & STATUS_ATTN_EVENTS) !=
 	    (sblk->status_attn_bits_ack & STATUS_ATTN_EVENTS))
 		return 1;
@@ -3187,6 +3350,23 @@ bnx2_chk_missed_msi(struct bnx2 *bp)
 	bp->idle_chk_status_idx = bnapi->last_status_idx;
 }
 
+#ifdef BCM_CNIC
+static void bnx2_poll_cnic(struct bnx2 *bp, struct bnx2_napi *bnapi)
+{
+	struct cnic_ops *c_ops;
+
+	if (!bnapi->cnic_present)
+		return;
+
+	rcu_read_lock();
+	c_ops = rcu_dereference(bp->cnic_ops);
+	if (c_ops)
+		bnapi->cnic_tag = c_ops->cnic_handler(bp->cnic_data,
+						      bnapi->status_blk.msi);
+	rcu_read_unlock();
+}
+#endif
+
 static void bnx2_poll_link(struct bnx2 *bp, struct bnx2_napi *bnapi)
 {
 	struct status_block *sblk = bnapi->status_blk.msi;
@@ -3261,6 +3441,10 @@ static int bnx2_poll(struct net_device *dev, int *budget)
 
 		work_done = bnx2_poll_work(bp, bnapi, work_done, *budget);
 
+#ifdef BCM_CNIC
+		bnx2_poll_cnic(bp, bnapi);
+#endif
+
 		/* bnapi->last_status_idx is used below to tell the hw how
 		 * much work has been processed, so we must read it before
 		 * checking for more work.
@@ -4636,8 +4820,11 @@ bnx2_init_chip(struct bnx2 *bp)
 	val = REG_RD(bp, BNX2_MQ_CONFIG);
 	val &= ~BNX2_MQ_CONFIG_KNL_BYP_BLK_SIZE;
 	val |= BNX2_MQ_CONFIG_KNL_BYP_BLK_SIZE_256;
-	if (CHIP_ID(bp) == CHIP_ID_5709_A0 || CHIP_ID(bp) == CHIP_ID_5709_A1)
-		val |= BNX2_MQ_CONFIG_HALT_DIS;
+	if (CHIP_NUM(bp) == CHIP_NUM_5709) {
+		val |= BNX2_MQ_CONFIG_BIN_MQ_MODE;
+		if (CHIP_REV(bp) == CHIP_REV_Ax)
+			val |= BNX2_MQ_CONFIG_HALT_DIS;
+	}
 
 	REG_WR(bp, BNX2_MQ_CONFIG, val);
 
@@ -5947,7 +6134,7 @@ static void
 bnx2_setup_int_mode(struct bnx2 *bp, int dis_msi)
 {
 	int cpus = num_online_cpus();
-	int msix_vecs = min(cpus + 1, RX_MAX_RSS_RINGS);
+	int msix_vecs = min(cpus + 1, RX_MAX_RINGS);
 
 	bp->irq_tbl[0].handler = bnx2_interrupt;
 	strcpy(bp->irq_tbl[0].name, bp->dev->name);
@@ -7532,7 +7719,7 @@ bnx2_init_board(struct pci_dev *pdev, struct net_device *dev)
 	INIT_WORK(&bp->reset_task, bnx2_reset_task, bp);
 
 	dev->base_addr = dev->mem_start = pci_resource_start(pdev, 0);
-	mem_len = MB_GET_CID_ADDR(TX_TSS_CID + 1);
+	mem_len = MB_GET_CID_ADDR(TX_TSS_CID + TX_MAX_TSS_RINGS + 1);
 	dev->mem_end = dev->mem_start + mem_len;
 	dev->irq = pdev->irq;
 
diff --git a/drivers/net/bnx2.h b/drivers/net/bnx2.h
index cdcbb68..2650627 100644
--- a/drivers/net/bnx2.h
+++ b/drivers/net/bnx2.h
@@ -361,6 +361,9 @@ struct l2_fhdr {
 #define BNX2_L2CTX_CTX_TYPE_CTX_BD_CHN_TYPE_VALUE	 (1<<28)
 
 #define BNX2_L2CTX_HOST_BDIDX				0x00000004
+#define BNX2_L2CTX_STATUSB_NUM_SHIFT			 16
+#define BNX2_L2CTX_STATUSB_NUM(sb_id)			 \
+	(((sb_id) > 0) ? (((sb_id) + 7) << BNX2_L2CTX_STATUSB_NUM_SHIFT) : 0)
 #define BNX2_L2CTX_HOST_BSEQ				0x00000008
 #define BNX2_L2CTX_NX_BSEQ				0x0000000c
 #define BNX2_L2CTX_NX_BDHADDR_HI			0x00000010
@@ -5900,6 +5903,7 @@ struct l2_fhdr {
 #define BNX2_RXP_FTQ_CTL_CUR_DEPTH			 (0x3ffL<<22)
 
 #define BNX2_RXP_SCRATCH				0x000e0000
+#define BNX2_RXP_SCRATCH_RXP_FLOOD			 0x000e0024
 #define BNX2_RXP_SCRATCH_RSS_TBL_SZ			 0x000e0038
 #define BNX2_RXP_SCRATCH_RSS_TBL			 0x000e003c
 #define BNX2_RXP_SCRATCH_RSS_TBL_MAX_ENTRIES		 128
@@ -6528,6 +6532,9 @@ struct l2_fhdr {
 #define RX_CID		0
 #define RX_RSS_CID	4
 #define RX_MAX_RSS_RINGS	7
+#define RX_MAX_RINGS		(RX_MAX_RSS_RINGS + 1)
+#define TX_MAX_TSS_RINGS	7
+#define TX_MAX_RINGS		(TX_MAX_TSS_RINGS + 1)
 
 #define MB_TX_CID_ADDR	MB_GET_CID_ADDR(TX_CID)
 #define MB_RX_CID_ADDR	MB_GET_CID_ADDR(RX_CID)
@@ -6667,6 +6674,11 @@ struct bnx2_napi {
 	u32 			last_status_idx;
 	u32			int_num;
 
+#ifdef BCM_CNIC
+	u32			cnic_tag;
+	int			cnic_present;
+#endif
+
 	struct bnx2_rx_ring_info	rx_ring;
 	struct bnx2_tx_ring_info	tx_ring;
 };
@@ -6717,6 +6729,10 @@ struct bnx2 {
 	int		tx_ring_size;
 	u32		tx_wake_thresh;
 
+#ifdef BCM_CNIC
+	struct cnic_ops		*cnic_ops;
+	void			*cnic_data;
+#endif
 	/* End of fields used in the performance code paths. */
 
 	unsigned int		current_interval;
@@ -6881,6 +6897,10 @@ struct bnx2 {
 
 	u32			idle_chk_status_idx;
 
+#ifdef BCM_CNIC
+	struct cnic_eth_dev	cnic_eth_dev;
+#endif
+
 };
 
 #define REG_RD(bp, offset)					\