Add bnx2i driver.

The CNIC driver controls BNX2 hardware rings and resources used by iSCSI. Most hardware resources for iSCSI are separate from those used for ethernet networking.

iSCSI uses a separate MAC address and IP address. The CNIC driver creates a UIO interface to handle the non-offloaded packets such as ARP, etc in userspace.

drivers/net/cnic.c new file +6 -0

+static int cnic_alloc_context(struct cnic_dev *dev)
+{
+	struct cnic_local *cp = dev->cnic_priv;
+
+	if (CHIP_NUM(cp) == CHIP_NUM_5709) {
+		int i, k, arr_size;

drivers/net/cnic.c new file +121 -0

+		k = 0;
+		for (i = 0; i < 2; i++) {
+			u32 j, reg, off, lo, hi;
+
+			if (i == 0)
+				off = BNX2_PG_CTX_MAP;
+			else
+				off = BNX2_ISCSI_CTX_MAP;
+
+			reg = cnic_reg_rd_ind(dev, off);
+			lo = reg >> 16;
+			hi = reg & 0xffff;
+			for (j = lo; j < hi; j += cp->cids_per_blk, k++)
+				cp->ctx_arr[k].cid = j;
+		}
+
+		cp->ctx_blks = k;
+		if (cp->ctx_blks >= (BNX2_MAX_CID / cp->cids_per_blk)) {
+			cp->ctx_blks = 0;
+			return -ENOMEM;
+		}
+
+		for (i = 0; i < cp->ctx_blks; i++) {
+			cp->ctx_arr[i].ctx =
+				pci_alloc_consistent(dev->pcidev, BCM_PAGE_SIZE,
+						     &cp->ctx_arr[i].mapping);
+			if (cp->ctx_arr[i].ctx == NULL)
+				return -ENOMEM;
+		}
+	}
+	return 0;
+}
+
+static int cnic_alloc_bnx2_resc(struct cnic_dev *dev)
+{
+	struct cnic_local *cp = dev->cnic_priv;
+	struct uio_info *uinfo;
+	int ret;
+
+	ret = cnic_alloc_dma(dev, &cp->kwq_info, KWQ_PAGE_CNT, 1);
+	if (ret)
+		goto error;
+	cp->kwq = (struct kwqe **) cp->kwq_info.pg_arr;
+
+	ret = cnic_alloc_dma(dev, &cp->kcq_info, KCQ_PAGE_CNT, 1);
+	if (ret)
+		goto error;
+	cp->kcq = (struct kcqe **) cp->kcq_info.pg_arr;
+
+	ret = cnic_alloc_context(dev);
+	if (ret)
+		goto error;
+
+	cp->l2_ring_size = 2 * BCM_PAGE_SIZE;
+	cp->l2_ring = pci_alloc_consistent(dev->pcidev, cp->l2_ring_size,
+					   &cp->l2_ring_map);
+	if (!cp->l2_ring)
+		goto error;
+
+	cp->l2_buf_size = (cp->l2_rx_ring_size + 1) * cp->l2_single_buf_size;
+	cp->l2_buf_size = PAGE_ALIGN(cp->l2_buf_size);
+	cp->l2_buf = pci_alloc_consistent(dev->pcidev, cp->l2_buf_size,
+					   &cp->l2_buf_map);
+	if (!cp->l2_buf)
+		goto error;
+
+	uinfo = kzalloc(sizeof(*uinfo), GFP_ATOMIC);
+	if (!uinfo)
+		goto error;
+
+	uinfo->mem[0].addr = dev->netdev->base_addr;
+	uinfo->mem[0].internal_addr = dev->regview;
+	uinfo->mem[0].size = dev->netdev->mem_end - dev->netdev->mem_start;
+	uinfo->mem[0].memtype = UIO_MEM_PHYS;
+
+	uinfo->mem[1].addr = (unsigned long) cp->status_blk & PAGE_MASK;
+	if (cp->ethdev->drv_state & CNIC_DRV_STATE_USING_MSIX)
+		uinfo->mem[1].size = BNX2_SBLK_MSIX_ALIGN_SIZE * 9;
+	else
+		uinfo->mem[1].size = BNX2_SBLK_MSIX_ALIGN_SIZE;
+	uinfo->mem[1].memtype = UIO_MEM_LOGICAL;
+
+	uinfo->mem[2].addr = (unsigned long) cp->l2_ring;
+	uinfo->mem[2].size = cp->l2_ring_size;
+	uinfo->mem[2].memtype = UIO_MEM_LOGICAL;
+
+	uinfo->mem[3].addr = (unsigned long) cp->l2_buf;
+	uinfo->mem[3].size = cp->l2_buf_size;
+	uinfo->mem[3].memtype = UIO_MEM_LOGICAL;
+
+	uinfo->name = "bnx2_cnic";
+	uinfo->version = CNIC_MODULE_VERSION;
+	uinfo->irq = UIO_IRQ_CUSTOM;
+
+	uinfo->open = cnic_uio_open;
+	uinfo->release = cnic_uio_close;
+
+	uinfo->priv = dev;
+
+	ret = uio_register_device(&dev->pcidev->dev, uinfo);
+	if (ret) {
+		kfree(uinfo);
+		goto error;
+	}
+
+	cp->cnic_uinfo = uinfo;
+
+	return 0;
+
+error:
+	cnic_free_resc(dev);
+	return ret;
+}
+
+static inline u32 cnic_kwq_avail(struct cnic_local *cp)
+{
+	return cp->max_kwq_idx -
+		((cp->kwq_prod_idx - cp->kwq_con_idx) & cp->max_kwq_idx);
+}
+
+static int cnic_submit_bnx2_kwqes(struct cnic_dev *dev, struct kwqe

*wqes[], + u32 num_wqes)

drivers/net/cnic.c new file +707 -0

+{
+	struct cnic_local *cp = dev->cnic_priv;
+	struct kwqe *prod_qe;
+	u16 prod, sw_prod, i;
+
+	if (!test_bit(CNIC_F_CNIC_UP, &dev->flags))
+		return -EAGAIN;		/* bnx2 is down */
+
+	spin_lock_bh(&cp->cnic_ulp_lock);
+	if (num_wqes > cnic_kwq_avail(cp) &&
+	    !(cp->cnic_local_flags & CNIC_LCL_FL_KWQ_INIT)) {
+		spin_unlock_bh(&cp->cnic_ulp_lock);
+		return -EAGAIN;
+	}
+
+	cp->cnic_local_flags &= ~CNIC_LCL_FL_KWQ_INIT;
+
+	prod = cp->kwq_prod_idx;
+	sw_prod = prod & MAX_KWQ_IDX;
+	for (i = 0; i < num_wqes; i++) {
+		prod_qe = &cp->kwq[KWQ_PG(sw_prod)][KWQ_IDX(sw_prod)];
+		memcpy(prod_qe, wqes[i], sizeof(struct kwqe));
+		prod++;
+		sw_prod = prod & MAX_KWQ_IDX;
+	}
+	cp->kwq_prod_idx = prod;
+
+	CNIC_WR16(dev, cp->kwq_io_addr, cp->kwq_prod_idx);
+
+	spin_unlock_bh(&cp->cnic_ulp_lock);
+	return 0;
+}
+
+static void service_kcqes(struct cnic_dev *dev, int num_cqes)
+{
+	struct cnic_local *cp = dev->cnic_priv;
+	int i, j;
+
+	i = 0;
+	j = 1;
+	while (num_cqes) {
+		struct cnic_ulp_ops *ulp_ops;
+		int ulp_type;
+		u32 kcqe_op_flag = cp->completed_kcq[i]->kcqe_op_flag;
+		u32 kcqe_layer = kcqe_op_flag & KCQE_FLAGS_LAYER_MASK;
+
+		if (unlikely(kcqe_op_flag & KCQE_RAMROD_COMPLETION))
+			cnic_kwq_completion(dev, 1);
+
+		while (j < num_cqes) {
+			u32 next_op = cp->completed_kcq[i + j]->kcqe_op_flag;
+
+			if ((next_op & KCQE_FLAGS_LAYER_MASK) != kcqe_layer)
+				break;
+
+			if (unlikely(next_op & KCQE_RAMROD_COMPLETION))
+				cnic_kwq_completion(dev, 1);
+			j++;
+		}
+
+		if (kcqe_layer == KCQE_FLAGS_LAYER_MASK_L5_RDMA)
+			ulp_type = CNIC_ULP_RDMA;
+		else if (kcqe_layer == KCQE_FLAGS_LAYER_MASK_L5_ISCSI)
+			ulp_type = CNIC_ULP_ISCSI;
+		else if (kcqe_layer == KCQE_FLAGS_LAYER_MASK_L4)
+			ulp_type = CNIC_ULP_L4;
+		else if (kcqe_layer == KCQE_FLAGS_LAYER_MASK_L2)
+			goto end;
+		else {
+			printk(KERN_ERR PFX "%s: Unknown type of KCQE(0x%x)\n",
+			       dev->netdev->name, kcqe_op_flag);
+			goto end;
+		}
+
+		rcu_read_lock();
+		ulp_ops = rcu_dereference(cp->ulp_ops[ulp_type]);
+		if (likely(ulp_ops)) {
+			ulp_ops->indicate_kcqes(cp->ulp_handle[ulp_type],
+						  cp->completed_kcq + i, j);
+		}
+		rcu_read_unlock();
+end:
+		num_cqes -= j;
+		i += j;
+		j = 1;
+	}
+	return;
+}
+
+static u16 cnic_bnx2_next_idx(u16 idx)
+{
+	return idx + 1;
+}
+
+static u16 cnic_bnx2_hw_idx(u16 idx)
+{
+	return idx;
+}
+
+static int cnic_get_kcqes(struct cnic_dev *dev, u16 hw_prod, u16 *sw_prod)
+{
+	struct cnic_local *cp = dev->cnic_priv;
+	u16 i, ri, last;
+	struct kcqe *kcqe;
+	int kcqe_cnt = 0, last_cnt = 0;
+
+	i = ri = last = *sw_prod;
+	ri &= MAX_KCQ_IDX;
+
+	while ((i != hw_prod) && (kcqe_cnt < MAX_COMPLETED_KCQE)) {
+		kcqe = &cp->kcq[KCQ_PG(ri)][KCQ_IDX(ri)];
+		cp->completed_kcq[kcqe_cnt++] = kcqe;
+		i = cp->next_idx(i);
+		ri = i & MAX_KCQ_IDX;
+		if (likely(!(kcqe->kcqe_op_flag & KCQE_FLAGS_NEXT))) {
+			last_cnt = kcqe_cnt;
+			last = i;
+		}
+	}
+
+	*sw_prod = last;
+	return last_cnt;
+}
+
+static void cnic_chk_bnx2_pkt_rings(struct cnic_local *cp)
+{
+	u16 rx_cons = *cp->rx_cons_ptr;
+	u16 tx_cons = *cp->tx_cons_ptr;
+
+	if (cp->tx_cons != tx_cons || cp->rx_cons != rx_cons) {
+		cp->tx_cons = tx_cons;
+		cp->rx_cons = rx_cons;
+		uio_event_notify(cp->cnic_uinfo);
+	}
+}
+
+static int cnic_service_bnx2(void *data, void *status_blk)
+{
+	struct cnic_dev *dev = data;
+	struct status_block *sblk = status_blk;
+	struct cnic_local *cp = dev->cnic_priv;
+	u32 status_idx = sblk->status_idx;
+	u16 hw_prod, sw_prod;
+	int kcqe_cnt;
+
+	if (unlikely(!test_bit(CNIC_F_CNIC_UP, &dev->flags)))
+		return status_idx;
+
+	cp->kwq_con_idx = *cp->kwq_con_idx_ptr;
+
+	hw_prod = sblk->status_completion_producer_index;
+	sw_prod = cp->kcq_prod_idx;
+	while (sw_prod != hw_prod) {
+		kcqe_cnt = cnic_get_kcqes(dev, hw_prod, &sw_prod);
+		if (kcqe_cnt == 0)
+			goto done;
+
+		service_kcqes(dev, kcqe_cnt);
+
+		/* Tell compiler that status_blk fields can change. */
+		barrier();
+		if (status_idx != sblk->status_idx) {
+			status_idx = sblk->status_idx;
+			cp->kwq_con_idx = *cp->kwq_con_idx_ptr;
+			hw_prod = sblk->status_completion_producer_index;
+		} else
+			break;
+	}
+
+done:
+	CNIC_WR16(dev, cp->kcq_io_addr, sw_prod);
+
+	cp->kcq_prod_idx = sw_prod;
+
+	cnic_chk_bnx2_pkt_rings(cp);
+	return status_idx;
+}
+
+static void cnic_service_bnx2_msix(unsigned long data)
+{
+	struct cnic_dev *dev = (struct cnic_dev *) data;
+	struct cnic_local *cp = dev->cnic_priv;
+	struct status_block_msix *status_blk = cp->bnx2_status_blk;
+	u32 status_idx = status_blk->status_idx;
+	u16 hw_prod, sw_prod;
+	int kcqe_cnt;
+
+	cp->kwq_con_idx = status_blk->status_cmd_consumer_index;
+
+	hw_prod = status_blk->status_completion_producer_index;
+	sw_prod = cp->kcq_prod_idx;
+	while (sw_prod != hw_prod) {
+		kcqe_cnt = cnic_get_kcqes(dev, hw_prod, &sw_prod);
+		if (kcqe_cnt == 0)
+			goto done;
+
+		service_kcqes(dev, kcqe_cnt);
+
+		/* Tell compiler that status_blk fields can change. */
+		barrier();
+		if (status_idx != status_blk->status_idx) {
+			status_idx = status_blk->status_idx;
+			cp->kwq_con_idx = status_blk->status_cmd_consumer_index;
+			hw_prod = status_blk->status_completion_producer_index;
+		} else
+			break;
+	}
+
+done:
+	CNIC_WR16(dev, cp->kcq_io_addr, sw_prod);
+	cp->kcq_prod_idx = sw_prod;
+
+	cnic_chk_bnx2_pkt_rings(cp);
+
+	cp->last_status_idx = status_idx;
+	CNIC_WR(dev, BNX2_PCICFG_INT_ACK_CMD, cp->int_num |
+		BNX2_PCICFG_INT_ACK_CMD_INDEX_VALID | cp->last_status_idx);
+}
+
+static irqreturn_t cnic_irq(int irq, void *dev_instance)
+{
+	struct cnic_dev *dev = dev_instance;
+	struct cnic_local *cp = dev->cnic_priv;
+	u16 prod = cp->kcq_prod_idx & MAX_KCQ_IDX;
+
+	if (cp->ack_int)
+		cp->ack_int(dev);
+
+	prefetch(cp->status_blk);
+	prefetch(&cp->kcq[KCQ_PG(prod)][KCQ_IDX(prod)]);
+
+	if (likely(test_bit(CNIC_F_CNIC_UP, &dev->flags)))
+		tasklet_schedule(&cp->cnic_irq_task);
+
+	return IRQ_HANDLED;
+}
+
+static void cnic_ulp_stop(struct cnic_dev *dev)
+{
+	struct cnic_local *cp = dev->cnic_priv;
+	int if_type;
+
+	rcu_read_lock();
+	for (if_type = 0; if_type < MAX_CNIC_ULP_TYPE; if_type++) {
+		struct cnic_ulp_ops *ulp_ops;
+
+		ulp_ops = rcu_dereference(cp->ulp_ops[if_type]);
+		if (!ulp_ops)
+			continue;
+
+		if (test_and_clear_bit(ULP_F_START, &cp->ulp_flags[if_type]))
+			ulp_ops->cnic_stop(cp->ulp_handle[if_type]);
+	}
+	rcu_read_unlock();
+}
+
+static void cnic_ulp_start(struct cnic_dev *dev)
+{
+	struct cnic_local *cp = dev->cnic_priv;
+	int if_type;
+
+	rcu_read_lock();
+	for (if_type = 0; if_type < MAX_CNIC_ULP_TYPE; if_type++) {
+		struct cnic_ulp_ops *ulp_ops;
+
+		ulp_ops = rcu_dereference(cp->ulp_ops[if_type]);
+		if (!ulp_ops || !ulp_ops->cnic_start)
+			continue;
+
+		if (!test_and_set_bit(ULP_F_START, &cp->ulp_flags[if_type]))
+			ulp_ops->cnic_start(cp->ulp_handle[if_type]);
+	}
+	rcu_read_unlock();
+}
+
+static int cnic_ctl(void *data, struct cnic_ctl_info *info)
+{
+	struct cnic_dev *dev = data;
+
+	switch (info->cmd) {
+	case CNIC_CTL_STOP_CMD:
+		cnic_hold(dev);
+		mutex_lock(&cnic_lock);
+
+		cnic_ulp_stop(dev);
+		cnic_stop_hw(dev);
+
+		mutex_unlock(&cnic_lock);
+		cnic_put(dev);
+		break;
+	case CNIC_CTL_START_CMD:
+		cnic_hold(dev);
+		mutex_lock(&cnic_lock);
+
+		if (!cnic_start_hw(dev))
+			cnic_ulp_start(dev);
+
+		mutex_unlock(&cnic_lock);
+		cnic_put(dev);
+		break;
+	default:
+		return -EINVAL;
+	}
+	return 0;
+}
+
+static void cnic_ulp_init(struct cnic_dev *dev)
+{
+	int i;
+	struct cnic_local *cp = dev->cnic_priv;
+
+	rcu_read_lock();
+	for (i = 0; i < MAX_CNIC_ULP_TYPE_EXT; i++) {
+		struct cnic_ulp_ops *ulp_ops;
+
+		ulp_ops = rcu_dereference(cnic_ulp_tbl[i]);
+		if (!ulp_ops || !ulp_ops->cnic_init)
+			continue;
+
+		if (!test_and_set_bit(ULP_F_INIT, &cp->ulp_flags[i]))
+			ulp_ops->cnic_init(dev);
+
+	}
+	rcu_read_unlock();
+}
+
+static void cnic_ulp_exit(struct cnic_dev *dev)
+{
+	int i;
+	struct cnic_local *cp = dev->cnic_priv;
+
+	rcu_read_lock();
+	for (i = 0; i < MAX_CNIC_ULP_TYPE_EXT; i++) {
+		struct cnic_ulp_ops *ulp_ops;
+
+		ulp_ops = rcu_dereference(cnic_ulp_tbl[i]);
+		if (!ulp_ops || !ulp_ops->cnic_exit)
+			continue;
+
+		if (test_and_clear_bit(ULP_F_INIT, &cp->ulp_flags[i]))
+			ulp_ops->cnic_exit(dev);
+
+	}
+	rcu_read_unlock();
+}
+
+static int cnic_cm_offload_pg(struct cnic_sock *csk)
+{
+	struct cnic_dev *dev = csk->dev;
+	struct l4_kwq_offload_pg *l4kwqe;
+	struct kwqe *wqes[1];
+
+	l4kwqe = (struct l4_kwq_offload_pg *) &csk->kwqe1;
+	memset(l4kwqe, 0, sizeof(*l4kwqe));
+	wqes[0] = (struct kwqe *) l4kwqe;
+
+	l4kwqe->op_code = L4_KWQE_OPCODE_VALUE_OFFLOAD_PG;
+	l4kwqe->flags =
+		L4_LAYER_CODE << L4_KWQ_OFFLOAD_PG_LAYER_CODE_SHIFT;
+	l4kwqe->l2hdr_nbytes = ETH_HLEN;
+
+	l4kwqe->da0 = csk->ha[0];
+	l4kwqe->da1 = csk->ha[1];
+	l4kwqe->da2 = csk->ha[2];
+	l4kwqe->da3 = csk->ha[3];
+	l4kwqe->da4 = csk->ha[4];
+	l4kwqe->da5 = csk->ha[5];
+
+	l4kwqe->sa0 = dev->mac_addr[0];
+	l4kwqe->sa1 = dev->mac_addr[1];
+	l4kwqe->sa2 = dev->mac_addr[2];
+	l4kwqe->sa3 = dev->mac_addr[3];
+	l4kwqe->sa4 = dev->mac_addr[4];
+	l4kwqe->sa5 = dev->mac_addr[5];
+
+	l4kwqe->etype = ETH_P_IP;
+	l4kwqe->ipid_count = DEF_IPID_COUNT;
+	l4kwqe->host_opaque = csk->l5_cid;
+
+	if (csk->vlan_id) {
+		l4kwqe->pg_flags |= L4_KWQ_OFFLOAD_PG_VLAN_TAGGING;
+		l4kwqe->vlan_tag = csk->vlan_id;
+		l4kwqe->l2hdr_nbytes += 4;
+	}
+
+	return dev->submit_kwqes(dev, wqes, 1);
+}
+
+static int cnic_cm_update_pg(struct cnic_sock *csk)
+{
+	struct cnic_dev *dev = csk->dev;
+	struct l4_kwq_update_pg *l4kwqe;
+	struct kwqe *wqes[1];
+
+	l4kwqe = (struct l4_kwq_update_pg *) &csk->kwqe1;
+	memset(l4kwqe, 0, sizeof(*l4kwqe));
+	wqes[0] = (struct kwqe *) l4kwqe;
+
+	l4kwqe->opcode = L4_KWQE_OPCODE_VALUE_UPDATE_PG;
+	l4kwqe->flags =
+		L4_LAYER_CODE << L4_KWQ_UPDATE_PG_LAYER_CODE_SHIFT;
+	l4kwqe->pg_cid = csk->pg_cid;
+
+	l4kwqe->da0 = csk->ha[0];
+	l4kwqe->da1 = csk->ha[1];
+	l4kwqe->da2 = csk->ha[2];
+	l4kwqe->da3 = csk->ha[3];
+	l4kwqe->da4 = csk->ha[4];
+	l4kwqe->da5 = csk->ha[5];
+
+	l4kwqe->pg_host_opaque = csk->l5_cid;
+	l4kwqe->pg_valids = L4_KWQ_UPDATE_PG_VALIDS_DA;
+
+	return dev->submit_kwqes(dev, wqes, 1);
+}
+
+static int cnic_cm_upload_pg(struct cnic_sock *csk)
+{
+	struct cnic_dev *dev = csk->dev;
+	struct l4_kwq_upload *l4kwqe;
+	struct kwqe *wqes[1];
+
+	l4kwqe = (struct l4_kwq_upload *) &csk->kwqe1;
+	memset(l4kwqe, 0, sizeof(*l4kwqe));
+	wqes[0] = (struct kwqe *) l4kwqe;
+
+	l4kwqe->opcode = L4_KWQE_OPCODE_VALUE_UPLOAD_PG;
+	l4kwqe->flags =
+		L4_LAYER_CODE << L4_KWQ_UPLOAD_LAYER_CODE_SHIFT;
+	l4kwqe->cid = csk->pg_cid;
+
+	return dev->submit_kwqes(dev, wqes, 1);
+}
+
+static int cnic_cm_conn_req(struct cnic_sock *csk)
+{
+	struct cnic_dev *dev = csk->dev;
+	struct l4_kwq_connect_req1 *l4kwqe1;
+	struct l4_kwq_connect_req2 *l4kwqe2;
+	struct l4_kwq_connect_req3 *l4kwqe3;
+	struct kwqe *wqes[3];
+	u8 tcp_flags = 0;
+	int num_wqes = 2;
+
+	l4kwqe1 = (struct l4_kwq_connect_req1 *) &csk->kwqe1;
+	l4kwqe2 = (struct l4_kwq_connect_req2 *) &csk->kwqe2;
+	l4kwqe3 = (struct l4_kwq_connect_req3 *) &csk->kwqe3;
+	memset(l4kwqe1, 0, sizeof(*l4kwqe1));
+	memset(l4kwqe2, 0, sizeof(*l4kwqe2));
+	memset(l4kwqe3, 0, sizeof(*l4kwqe3));
+
+	l4kwqe3->op_code = L4_KWQE_OPCODE_VALUE_CONNECT3;
+	l4kwqe3->flags =
+		L4_LAYER_CODE << L4_KWQ_CONNECT_REQ3_LAYER_CODE_SHIFT;
+	l4kwqe3->ka_timeout = csk->ka_timeout;
+	l4kwqe3->ka_interval = csk->ka_interval;
+	l4kwqe3->ka_max_probe_count = csk->ka_max_probe_count;
+	l4kwqe3->tos = csk->tos;
+	l4kwqe3->ttl = csk->ttl;
+	l4kwqe3->snd_seq_scale = csk->snd_seq_scale;
+	l4kwqe3->pmtu = csk->mtu;
+	l4kwqe3->rcv_buf = csk->rcv_buf;
+	l4kwqe3->snd_buf = csk->snd_buf;
+	l4kwqe3->seed = csk->seed;
+
+	wqes[0] = (struct kwqe *) l4kwqe1;
+	if (test_bit(SK_F_IPV6, &csk->flags)) {
+		wqes[1] = (struct kwqe *) l4kwqe2;
+		wqes[2] = (struct kwqe *) l4kwqe3;
+		num_wqes = 3;
+
+		l4kwqe1->conn_flags = L4_KWQ_CONNECT_REQ1_IP_V6;
+		l4kwqe2->op_code = L4_KWQE_OPCODE_VALUE_CONNECT2;
+		l4kwqe2->flags =
+			L4_KWQ_CONNECT_REQ2_LINKED_WITH_NEXT |
+			L4_LAYER_CODE << L4_KWQ_CONNECT_REQ2_LAYER_CODE_SHIFT;
+		l4kwqe2->src_ip_v6_2 = be32_to_cpu(csk->src_ip[1]);
+		l4kwqe2->src_ip_v6_3 = be32_to_cpu(csk->src_ip[2]);
+		l4kwqe2->src_ip_v6_4 = be32_to_cpu(csk->src_ip[3]);
+		l4kwqe2->dst_ip_v6_2 = be32_to_cpu(csk->dst_ip[1]);
+		l4kwqe2->dst_ip_v6_3 = be32_to_cpu(csk->dst_ip[2]);
+		l4kwqe2->dst_ip_v6_4 = be32_to_cpu(csk->dst_ip[3]);
+		l4kwqe3->mss = l4kwqe3->pmtu - sizeof(struct ipv6hdr) -
+			       sizeof(struct tcphdr);
+	} else {
+		wqes[1] = (struct kwqe *) l4kwqe3;
+		l4kwqe3->mss = l4kwqe3->pmtu - sizeof(struct iphdr) -
+			       sizeof(struct tcphdr);
+	}
+
+	l4kwqe1->op_code = L4_KWQE_OPCODE_VALUE_CONNECT1;
+	l4kwqe1->flags =
+		(L4_LAYER_CODE << L4_KWQ_CONNECT_REQ1_LAYER_CODE_SHIFT) |
+		 L4_KWQ_CONNECT_REQ3_LINKED_WITH_NEXT;
+	l4kwqe1->cid = csk->cid;
+	l4kwqe1->pg_cid = csk->pg_cid;
+	l4kwqe1->src_ip = be32_to_cpu(csk->src_ip[0]);
+	l4kwqe1->dst_ip = be32_to_cpu(csk->dst_ip[0]);
+	l4kwqe1->src_port = be16_to_cpu(csk->src_port);
+	l4kwqe1->dst_port = be16_to_cpu(csk->dst_port);
+	if (csk->tcp_flags & SK_TCP_NO_DELAY_ACK)
+		tcp_flags |= L4_KWQ_CONNECT_REQ1_NO_DELAY_ACK;
+	if (csk->tcp_flags & SK_TCP_KEEP_ALIVE)
+		tcp_flags |= L4_KWQ_CONNECT_REQ1_KEEP_ALIVE;
+	if (csk->tcp_flags & SK_TCP_NAGLE)
+		tcp_flags |= L4_KWQ_CONNECT_REQ1_NAGLE_ENABLE;
+	if (csk->tcp_flags & SK_TCP_TIMESTAMP)
+		tcp_flags |= L4_KWQ_CONNECT_REQ1_TIME_STAMP;
+	if (csk->tcp_flags & SK_TCP_SACK)
+		tcp_flags |= L4_KWQ_CONNECT_REQ1_SACK;
+	if (csk->tcp_flags & SK_TCP_SEG_SCALING)
+		tcp_flags |= L4_KWQ_CONNECT_REQ1_SEG_SCALING;
+
+	l4kwqe1->tcp_flags = tcp_flags;
+
+	return dev->submit_kwqes(dev, wqes, num_wqes);
+}
+
+static int cnic_cm_close_req(struct cnic_sock *csk)
+{
+	struct cnic_dev *dev = csk->dev;
+	struct l4_kwq_close_req *l4kwqe;
+	struct kwqe *wqes[1];
+
+	l4kwqe = (struct l4_kwq_close_req *) &csk->kwqe2;
+	memset(l4kwqe, 0, sizeof(*l4kwqe));
+	wqes[0] = (struct kwqe *) l4kwqe;
+
+	l4kwqe->op_code = L4_KWQE_OPCODE_VALUE_CLOSE;
+	l4kwqe->flags = L4_LAYER_CODE << L4_KWQ_CLOSE_REQ_LAYER_CODE_SHIFT;
+	l4kwqe->cid = csk->cid;
+
+	return dev->submit_kwqes(dev, wqes, 1);
+}
+
+static int cnic_cm_abort_req(struct cnic_sock *csk)
+{
+	struct cnic_dev *dev = csk->dev;
+	struct l4_kwq_reset_req *l4kwqe;
+	struct kwqe *wqes[1];
+
+	l4kwqe = (struct l4_kwq_reset_req *) &csk->kwqe2;
+	memset(l4kwqe, 0, sizeof(*l4kwqe));
+	wqes[0] = (struct kwqe *) l4kwqe;
+
+	l4kwqe->op_code = L4_KWQE_OPCODE_VALUE_RESET;
+	l4kwqe->flags = L4_LAYER_CODE << L4_KWQ_RESET_REQ_LAYER_CODE_SHIFT;
+	l4kwqe->cid = csk->cid;
+
+	return dev->submit_kwqes(dev, wqes, 1);
+}
+
+static int cnic_cm_create(struct cnic_dev *dev, int ulp_type, u32 cid,
+			  u32 l5_cid, struct cnic_sock **csk, void *context)
+{
+	struct cnic_local *cp = dev->cnic_priv;
+	struct cnic_sock *csk1;
+
+	if (l5_cid >= MAX_CM_SK_TBL_SZ)
+		return -EINVAL;
+
+	csk1 = &cp->csk_tbl[l5_cid];
+	if (atomic_read(&csk1->ref_count))
+		return -EAGAIN;
+
+	if (test_and_set_bit(SK_F_INUSE, &csk1->flags))
+		return -EBUSY;
+
+	csk1->dev = dev;
+	csk1->cid = cid;
+	csk1->l5_cid = l5_cid;
+	csk1->ulp_type = ulp_type;
+	csk1->context = context;
+
+	csk1->ka_timeout = DEF_KA_TIMEOUT;
+	csk1->ka_interval = DEF_KA_INTERVAL;
+	csk1->ka_max_probe_count = DEF_KA_MAX_PROBE_COUNT;
+	csk1->tos = DEF_TOS;
+	csk1->ttl = DEF_TTL;
+	csk1->snd_seq_scale = DEF_SND_SEQ_SCALE;
+	csk1->rcv_buf = DEF_RCV_BUF;
+	csk1->snd_buf = DEF_SND_BUF;
+	csk1->seed = DEF_SEED;
+
+	*csk = csk1;
+	return 0;
+}
+
+static void cnic_cm_cleanup(struct cnic_sock *csk)
+{
+	if (csk->src_port) {
+		struct cnic_dev *dev = csk->dev;
+		struct cnic_local *cp = dev->cnic_priv;
+
+		cnic_free_id(&cp->csk_port_tbl, csk->src_port);
+		csk->src_port = 0;
+	}
+}
+
+static void cnic_close_conn(struct cnic_sock *csk)
+{
+	if (test_bit(SK_F_PG_OFFLD_COMPLETE, &csk->flags)) {
+		cnic_cm_upload_pg(csk);
+		clear_bit(SK_F_PG_OFFLD_COMPLETE, &csk->flags);
+	}
+	cnic_cm_cleanup(csk);
+}
+
+static int cnic_cm_destroy(struct cnic_sock *csk)
+{
+	if (!cnic_in_use(csk))
+		return -EINVAL;
+
+	csk_hold(csk);
+	clear_bit(SK_F_INUSE, &csk->flags);
+	smp_mb__after_clear_bit();
+	while (atomic_read(&csk->ref_count) != 1)
+		msleep(1);
+	cnic_cm_cleanup(csk);
+
+	csk->flags = 0;
+	csk_put(csk);
+	return 0;
+}
+
+static inline u16 cnic_get_vlan(struct net_device *dev,
+				struct net_device **vlan_dev)
+{
+	if (dev->priv_flags & IFF_802_1Q_VLAN) {
+		*vlan_dev = vlan_dev_real_dev(dev);
+		return vlan_dev_vlan_id(dev);
+	}
+	*vlan_dev = dev;
+	return 0;
+}
+
+static int cnic_get_v4_route(struct sockaddr_in *dst_addr,
+			     struct dst_entry **dst)
+{
+	struct flowi fl;
+	int err;
+	struct rtable *rt;
+
+	memset(&fl, 0, sizeof(fl));
+	fl.nl_u.ip4_u.daddr = dst_addr->sin_addr.s_addr;
+
+	err = ip_route_output_key(&init_net, &rt, &fl);
+	if (!err)
+		*dst = &rt->u.dst;
+	return err;
+}
+
+static int cnic_get_v6_route(struct sockaddr_in6 *dst_addr,
+			     struct dst_entry **dst)
+{
+#if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
+	struct flowi fl;
+
+	memset(&fl, 0, sizeof(fl));
+	ipv6_addr_copy(&fl.fl6_dst, &dst_addr->sin6_addr);
+	if (ipv6_addr_type(&fl.fl6_dst) & IPV6_ADDR_LINKLOCAL)
+		fl.oif = dst_addr->sin6_scope_id;
+
+	*dst = ip6_route_output(&init_net, NULL, &fl);
+	if (*dst)
+		return 0;
+#endif
+
+	return -ENETUNREACH;
+}
+
+static struct cnic_dev *cnic_cm_select_dev(struct sockaddr_in *dst_addr,
+					   int ulp_type)
+{
+	struct cnic_dev *dev = NULL;
+	struct dst_entry *dst;
+	struct net_device *netdev = NULL;
+	int err = -ENETUNREACH;
+
+	if (dst_addr->sin_family == AF_INET)
+		err = cnic_get_v4_route(dst_addr, &dst);
+	else if (dst_addr->sin_family == AF_INET6) {
+		struct sockaddr_in6 *dst_addr6 =
+			(struct sockaddr_in6 *) dst_addr;
+
+		err = cnic_get_v6_route(dst_addr6, &dst);
+	} else
+		return NULL;
+
+	if (err)
+		return NULL;
+
+	if (!dst->dev)
+		goto done;
+
+	cnic_get_vlan(dst->dev, &netdev);
+
+	dev = cnic_from_netdev(netdev);
+
+done:
+	dst_release(dst);
+	if (dev)
+		cnic_put(dev);
+	return dev;
+}
+
+static int cnic_resolve_addr(struct cnic_sock *csk, struct cnic_sockaddr

*saddr) +{

drivers/net/cnic.c new file +7 -0

+	struct cnic_dev *dev = csk->dev;
+	struct cnic_local *cp = dev->cnic_priv;
+
+	return cnic_send_nlmsg(cp, ISCSI_KEVENT_PATH_REQ, csk);
+}
+
+static int cnic_get_route(struct cnic_sock *csk, struct cnic_sockaddr

*saddr) +{

drivers/net/cnic.c new file +85 -0

+	struct cnic_dev *dev = csk->dev;
+	struct cnic_local *cp = dev->cnic_priv;
+	int is_v6, err, rc = -ENETUNREACH;
+	struct dst_entry *dst;
+	struct net_device *realdev;
+	u32 local_port;
+
+	if (saddr->local.v6.sin6_family == AF_INET6 &&
+	    saddr->remote.v6.sin6_family == AF_INET6)
+		is_v6 = 1;
+	else if (saddr->local.v4.sin_family == AF_INET &&
+		 saddr->remote.v4.sin_family == AF_INET)
+		is_v6 = 0;
+	else
+		return -EINVAL;
+
+	clear_bit(SK_F_IPV6, &csk->flags);
+
+	if (is_v6) {
+#if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
+		set_bit(SK_F_IPV6, &csk->flags);
+		err = cnic_get_v6_route(&saddr->remote.v6, &dst);
+		if (err)
+			return err;
+
+		if (!dst || dst->error || !dst->dev)
+			goto err_out;
+
+		memcpy(&csk->dst_ip[0], &saddr->remote.v6.sin6_addr,
+		       sizeof(struct in6_addr));
+		csk->dst_port = saddr->remote.v6.sin6_port;
+		local_port = saddr->local.v6.sin6_port;
+#else
+		return rc;
+#endif
+
+	} else {
+		err = cnic_get_v4_route(&saddr->remote.v4, &dst);
+		if (err)
+			return err;
+
+		if (!dst || dst->error || !dst->dev)
+			goto err_out;
+
+		csk->dst_ip[0] = saddr->remote.v4.sin_addr.s_addr;
+		csk->dst_port = saddr->remote.v4.sin_port;
+		local_port = saddr->local.v4.sin_port;
+	}
+
+	csk->vlan_id = cnic_get_vlan(dst->dev, &realdev);
+	if (realdev != dev->netdev)
+		goto err_out;
+
+	if (local_port >= CNIC_LOCAL_PORT_MIN &&
+	    local_port < CNIC_LOCAL_PORT_MAX) {
+		if (cnic_alloc_id(&cp->csk_port_tbl, local_port))
+			local_port = 0;
+	} else
+		local_port = 0;
+
+	if (!local_port) {
+		local_port = cnic_alloc_new_id(&cp->csk_port_tbl);
+		if (local_port == -1) {
+			rc = -ENOMEM;
+			goto err_out;
+		}
+	}
+	csk->src_port = local_port;
+
+	csk->mtu = dst_mtu(dst);
+	rc = 0;
+
+err_out:
+	dst_release(dst);
+	return rc;
+}
+
+static void cnic_init_csk_state(struct cnic_sock *csk)
+{
+	csk->state = 0;
+	clear_bit(SK_F_OFFLD_SCHED, &csk->flags);
+	clear_bit(SK_F_CLOSING, &csk->flags);
+}
+
+static int cnic_cm_connect(struct cnic_sock *csk, struct cnic_sockaddr

*saddr) +{

drivers/net/cnic.c new file +95 -0

+	int err = 0;
+
+	if (!cnic_in_use(csk))
+		return -EINVAL;
+
+	if (test_and_set_bit(SK_F_CONNECT_START, &csk->flags))
+		return -EINVAL;
+
+	cnic_init_csk_state(csk);
+
+	err = cnic_get_route(csk, saddr);
+	if (err)
+		goto err_out;
+
+	err = cnic_resolve_addr(csk, saddr);
+	if (!err)
+		return 0;
+
+err_out:
+	clear_bit(SK_F_CONNECT_START, &csk->flags);
+	return err;
+}
+
+static int cnic_cm_abort(struct cnic_sock *csk)
+{
+	struct cnic_local *cp = csk->dev->cnic_priv;
+	u32 opcode;
+
+	if (!cnic_in_use(csk))
+		return -EINVAL;
+
+	if (cnic_abort_prep(csk))
+		return cnic_cm_abort_req(csk);
+
+	/* Getting here means that we haven't started connect, or
+	 * connect was not successful.
+	 */
+
+	csk->state = L4_KCQE_OPCODE_VALUE_RESET_COMP;
+	if (test_bit(SK_F_PG_OFFLD_COMPLETE, &csk->flags))
+		opcode = csk->state;
+	else
+		opcode = L5CM_RAMROD_CMD_ID_TERMINATE_OFFLOAD;
+	cp->close_conn(csk, opcode);
+
+	return 0;
+}
+
+static int cnic_cm_close(struct cnic_sock *csk)
+{
+	if (!cnic_in_use(csk))
+		return -EINVAL;
+
+	if (cnic_close_prep(csk)) {
+		csk->state = L4_KCQE_OPCODE_VALUE_CLOSE_COMP;
+		return cnic_cm_close_req(csk);
+	}
+	return 0;
+}
+
+static void cnic_cm_upcall(struct cnic_local *cp, struct cnic_sock *csk,
+			   u8 opcode)
+{
+	struct cnic_ulp_ops *ulp_ops;
+	int ulp_type = csk->ulp_type;
+
+	rcu_read_lock();
+	ulp_ops = rcu_dereference(cp->ulp_ops[ulp_type]);
+	if (ulp_ops) {
+		if (opcode == L4_KCQE_OPCODE_VALUE_CONNECT_COMPLETE)
+			ulp_ops->cm_connect_complete(csk);
+		else if (opcode == L4_KCQE_OPCODE_VALUE_CLOSE_COMP)
+			ulp_ops->cm_close_complete(csk);
+		else if (opcode == L4_KCQE_OPCODE_VALUE_RESET_RECEIVED)
+			ulp_ops->cm_remote_abort(csk);
+		else if (opcode == L4_KCQE_OPCODE_VALUE_RESET_COMP)
+			ulp_ops->cm_abort_complete(csk);
+		else if (opcode == L4_KCQE_OPCODE_VALUE_CLOSE_RECEIVED)
+			ulp_ops->cm_remote_close(csk);
+	}
+	rcu_read_unlock();
+}
+
+static int cnic_cm_set_pg(struct cnic_sock *csk)
+{
+	if (cnic_offld_prep(csk)) {
+		if (test_bit(SK_F_PG_OFFLD_COMPLETE, &csk->flags))
+			cnic_cm_update_pg(csk);
+		else
+			cnic_cm_offload_pg(csk);
+	}
+	return 0;
+}
+
+static void cnic_cm_process_offld_pg(struct cnic_dev *dev, struct l4_kcq

*kcqe) +{

drivers/net/cnic.c new file +76 -0

+	struct cnic_local *cp = dev->cnic_priv;
+	u32 l5_cid = kcqe->pg_host_opaque;
+	u8 opcode = kcqe->op_code;
+	struct cnic_sock *csk = &cp->csk_tbl[l5_cid];
+
+	csk_hold(csk);
+	if (!cnic_in_use(csk))
+		goto done;
+
+	if (opcode == L4_KCQE_OPCODE_VALUE_UPDATE_PG) {
+		clear_bit(SK_F_OFFLD_SCHED, &csk->flags);
+		goto done;
+	}
+	csk->pg_cid = kcqe->pg_cid;
+	set_bit(SK_F_PG_OFFLD_COMPLETE, &csk->flags);
+	cnic_cm_conn_req(csk);
+
+done:
+	csk_put(csk);
+}
+
+static void cnic_cm_process_kcqe(struct cnic_dev *dev, struct kcqe *kcqe)
+{
+	struct cnic_local *cp = dev->cnic_priv;
+	struct l4_kcq *l4kcqe = (struct l4_kcq *) kcqe;
+	u8 opcode = l4kcqe->op_code;
+	u32 l5_cid;
+	struct cnic_sock *csk;
+
+	if (opcode == L4_KCQE_OPCODE_VALUE_OFFLOAD_PG ||
+	    opcode == L4_KCQE_OPCODE_VALUE_UPDATE_PG) {
+		cnic_cm_process_offld_pg(dev, l4kcqe);
+		return;
+	}
+
+	l5_cid = l4kcqe->conn_id;
+	if (opcode & 0x80)
+		l5_cid = l4kcqe->cid;
+	if (l5_cid >= MAX_CM_SK_TBL_SZ)
+		return;
+
+	csk = &cp->csk_tbl[l5_cid];
+	csk_hold(csk);
+
+	if (!cnic_in_use(csk)) {
+		csk_put(csk);
+		return;
+	}
+
+	switch (opcode) {
+	case L4_KCQE_OPCODE_VALUE_CONNECT_COMPLETE:
+		if (l4kcqe->status == 0)
+			set_bit(SK_F_OFFLD_COMPLETE, &csk->flags);
+
+		smp_mb__before_clear_bit();
+		clear_bit(SK_F_OFFLD_SCHED, &csk->flags);
+		cnic_cm_upcall(cp, csk, opcode);
+		break;
+
+	case L4_KCQE_OPCODE_VALUE_RESET_RECEIVED:
+		if (test_and_clear_bit(SK_F_OFFLD_COMPLETE, &csk->flags))
+			csk->state = opcode;
+		/* fall through */
+	case L4_KCQE_OPCODE_VALUE_CLOSE_COMP:
+	case L4_KCQE_OPCODE_VALUE_RESET_COMP:
+		cp->close_conn(csk, opcode);
+		break;
+
+	case L4_KCQE_OPCODE_VALUE_CLOSE_RECEIVED:
+		cnic_cm_upcall(cp, csk, opcode);
+		break;
+	}
+	csk_put(csk);
+}
+
+static void cnic_cm_indicate_kcqe(void *data, struct kcqe *kcqe[], u32

num) +{

drivers/net/cnic.c new file +19 -0

+	struct cnic_dev *dev = data;
+	int i;
+
+	for (i = 0; i < num; i++)
+		cnic_cm_process_kcqe(dev, kcqe[i]);
+}
+
+static struct cnic_ulp_ops cm_ulp_ops = {
+	.indicate_kcqes		= cnic_cm_indicate_kcqe,
+};
+
+static void cnic_cm_free_mem(struct cnic_dev *dev)
+{
+	struct cnic_local *cp = dev->cnic_priv;
+
+	kfree(cp->csk_tbl);
+	cp->csk_tbl = NULL;
+	cnic_free_id_tbl(&cp->csk_port_tbl);
+}

cp->csk_tbl = kzalloc(sizeof(struct cnic_sock) * MAX_CM_SK_TBL_SZ,

                              GFP_KERNEL);

There are some more of these.

From e7cbc3df66fb0b9c7036478be292f2db7be78153 Mon Sep 17 00:00:00 2001

James, the next 4 patches are our latest to add the bnx2i driver. We've fixed up all the issues that Mike brought up. Please consider applying them to scsi-misc-2.6.

New iSCSI driver for Broadcom BNX2 devices.

drivers/scsi/bnx2i/57xx_iscsi_hsi.h new file +4 -0

+/*
+ * iSCSI Async CQE
+ */
+struct bnx2i_async_msg {

Quoted 1 line · Michael Chan

New iSCSI driver for Broadcom BNX2 devices.

...

Quoted 4 lines

drivers/scsi/bnx2i/57xx_iscsi_hsi.h new file +4 -0

+/*
+ * iSCSI Async CQE
+ */
+struct bnx2i_async_msg {

...

drivers/scsi/bnx2i/57xx_iscsi_hsi.h line 50

Quoted 9 lines

drivers/scsi/bnx2i/57xx_iscsi_hsi.h new file +9 -0

+#if defined(__BIG_ENDIAN)
+       u8 async_event;
+       u8 async_vcode;
+       u16 param1;
+#elif defined(__LITTLE_ENDIAN)
+       u16 param1;
+       u8 async_vcode;
+       u8 async_event;
+#endif

...

Michael,
I'm feeling a bit dense and am not seeing why byte data would have to worry about the 32-bit word endianess of the CPU. Can you give an example of why defined(__BIG_ENDIAN) is needed?

Normally the _*ENDIAN defines are used for bit fields, not byte fields.

Byte data addressable by the CPU (e.g. host memory) is at the same offset regardless of endianness of the CPU. I feel like I'm missing something that should be obvious.

Hi Grant, these are what we call "DMA control structures" as opposed to DMA packet data.  Our chips are configured to do an additional 32-bit endian swap on all DMA control structures.  This way, all 32-bit control fields (such as length, status, etc) will come out right when the driver reads these fields.

If everything was defined as u32 in all these control structures, we wouldn't have to add the #ifdef.  u8 and u16 fields have to be defined this way or else big endian CPUs would read the wrong offset.

If you look at some of the control structures in tg3.h, you'll see the same thing.