Add support for the latest 1G/10G Chelsio adapter, T3.

This driver is required by the Chelsio T3 RDMA driver posted by
Steve Wise.
Signed-off-by: Divy Le Ray <divy@chelsio.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

drivers/net/Kconfig

@@ -2389,6 +2389,24 @@ config CHELSIO_T1_NAPI
 	  NAPI is a driver API designed to reduce CPU and interrupt load
 	  when the driver is receiving lots of packets from the card.
 
+config CHELSIO_T3
+        tristate "Chelsio Communications T3 10Gb Ethernet support"
+        depends on PCI
+        help
+          This driver supports Chelsio T3-based gigabit and 10Gb Ethernet
+          adapters.
+
+          For general information about Chelsio and our products, visit
+          our website at <http://www.chelsio.com>.
+
+          For customer support, please visit our customer support page at
+          <http://www.chelsio.com/support.htm>.
+
+          Please send feedback to <linux-bugs@chelsio.com>.
+
+          To compile this driver as a module, choose M here: the module
+          will be called cxgb3.
+
 config EHEA
 	tristate "eHEA Ethernet support"
 	depends on IBMEBUS

drivers/net/Makefile

@@ -6,6 +6,7 @@ obj-$(CONFIG_E1000) += e1000/
 obj-$(CONFIG_IBM_EMAC) += ibm_emac/
 obj-$(CONFIG_IXGB) += ixgb/
 obj-$(CONFIG_CHELSIO_T1) += chelsio/
+obj-$(CONFIG_CHELSIO_T3) += cxgb3/
 obj-$(CONFIG_EHEA) += ehea/
 obj-$(CONFIG_BONDING) += bonding/
 obj-$(CONFIG_GIANFAR) += gianfar_driver.o

drivers/net/cxgb3/Makefile

+#
+# Chelsio T3 driver
+#
+
+obj-$(CONFIG_CHELSIO_T3) += cxgb3.o
+
+cxgb3-objs := cxgb3_main.o ael1002.o vsc8211.o t3_hw.o mc5.o \
+	      xgmac.o sge.o l2t.o cxgb3_offload.o

drivers/net/cxgb3/adapter.h

+/*
+ * This file is part of the Chelsio T3 Ethernet driver for Linux.
+ *
+ * Copyright (C) 2003-2006 Chelsio Communications.  All rights reserved.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the LICENSE file included in this
+ * release for licensing terms and conditions.
+ */
+
+/* This file should not be included directly.  Include common.h instead. */
+
+#ifndef __T3_ADAPTER_H__
+#define __T3_ADAPTER_H__
+
+#include <linux/pci.h>
+#include <linux/spinlock.h>
+#include <linux/interrupt.h>
+#include <linux/timer.h>
+#include <linux/cache.h>
+#include "t3cdev.h"
+#include <asm/semaphore.h>
+#include <asm/bitops.h>
+#include <asm/io.h>
+
+typedef irqreturn_t(*intr_handler_t) (int, void *);
+
+struct vlan_group;
+
+struct port_info {
+	struct vlan_group *vlan_grp;
+	const struct port_type_info *port_type;
+	u8 port_id;
+	u8 rx_csum_offload;
+	u8 nqsets;
+	u8 first_qset;
+	struct cphy phy;
+	struct cmac mac;
+	struct link_config link_config;
+	struct net_device_stats netstats;
+	int activity;
+};
+
+enum {				/* adapter flags */
+	FULL_INIT_DONE = (1 << 0),
+	USING_MSI = (1 << 1),
+	USING_MSIX = (1 << 2),
+};
+
+struct rx_desc;
+struct rx_sw_desc;
+
+struct sge_fl {			/* SGE per free-buffer list state */
+	unsigned int buf_size;	/* size of each Rx buffer */
+	unsigned int credits;	/* # of available Rx buffers */
+	unsigned int size;	/* capacity of free list */
+	unsigned int cidx;	/* consumer index */
+	unsigned int pidx;	/* producer index */
+	unsigned int gen;	/* free list generation */
+	struct rx_desc *desc;	/* address of HW Rx descriptor ring */
+	struct rx_sw_desc *sdesc;	/* address of SW Rx descriptor ring */
+	dma_addr_t phys_addr;	/* physical address of HW ring start */
+	unsigned int cntxt_id;	/* SGE context id for the free list */
+	unsigned long empty;	/* # of times queue ran out of buffers */
+};
+
+/*
+ * Bundle size for grouping offload RX packets for delivery to the stack.
+ * Don't make this too big as we do prefetch on each packet in a bundle.
+ */
+# define RX_BUNDLE_SIZE 8
+
+struct rsp_desc;
+
+struct sge_rspq {		/* state for an SGE response queue */
+	unsigned int credits;	/* # of pending response credits */
+	unsigned int size;	/* capacity of response queue */
+	unsigned int cidx;	/* consumer index */
+	unsigned int gen;	/* current generation bit */
+	unsigned int polling;	/* is the queue serviced through NAPI? */
+	unsigned int holdoff_tmr;	/* interrupt holdoff timer in 100ns */
+	unsigned int next_holdoff;	/* holdoff time for next interrupt */
+	struct rsp_desc *desc;	/* address of HW response ring */
+	dma_addr_t phys_addr;	/* physical address of the ring */
+	unsigned int cntxt_id;	/* SGE context id for the response q */
+	spinlock_t lock;	/* guards response processing */
+	struct sk_buff *rx_head;	/* offload packet receive queue head */
+	struct sk_buff *rx_tail;	/* offload packet receive queue tail */
+
+	unsigned long offload_pkts;
+	unsigned long offload_bundles;
+	unsigned long eth_pkts;	/* # of ethernet packets */
+	unsigned long pure_rsps;	/* # of pure (non-data) responses */
+	unsigned long imm_data;	/* responses with immediate data */
+	unsigned long rx_drops;	/* # of packets dropped due to no mem */
+	unsigned long async_notif; /* # of asynchronous notification events */
+	unsigned long empty;	/* # of times queue ran out of credits */
+	unsigned long nomem;	/* # of responses deferred due to no mem */
+	unsigned long unhandled_irqs;	/* # of spurious intrs */
+};
+
+struct tx_desc;
+struct tx_sw_desc;
+
+struct sge_txq {		/* state for an SGE Tx queue */
+	unsigned long flags;	/* HW DMA fetch status */
+	unsigned int in_use;	/* # of in-use Tx descriptors */
+	unsigned int size;	/* # of descriptors */
+	unsigned int processed;	/* total # of descs HW has processed */
+	unsigned int cleaned;	/* total # of descs SW has reclaimed */
+	unsigned int stop_thres;	/* SW TX queue suspend threshold */
+	unsigned int cidx;	/* consumer index */
+	unsigned int pidx;	/* producer index */
+	unsigned int gen;	/* current value of generation bit */
+	unsigned int unacked;	/* Tx descriptors used since last COMPL */
+	struct tx_desc *desc;	/* address of HW Tx descriptor ring */
+	struct tx_sw_desc *sdesc;	/* address of SW Tx descriptor ring */
+	spinlock_t lock;	/* guards enqueueing of new packets */
+	unsigned int token;	/* WR token */
+	dma_addr_t phys_addr;	/* physical address of the ring */
+	struct sk_buff_head sendq;	/* List of backpressured offload packets */
+	struct tasklet_struct qresume_tsk;	/* restarts the queue */
+	unsigned int cntxt_id;	/* SGE context id for the Tx q */
+	unsigned long stops;	/* # of times q has been stopped */
+	unsigned long restarts;	/* # of queue restarts */
+};
+
+enum {				/* per port SGE statistics */
+	SGE_PSTAT_TSO,		/* # of TSO requests */
+	SGE_PSTAT_RX_CSUM_GOOD,	/* # of successful RX csum offloads */
+	SGE_PSTAT_TX_CSUM,	/* # of TX checksum offloads */
+	SGE_PSTAT_VLANEX,	/* # of VLAN tag extractions */
+	SGE_PSTAT_VLANINS,	/* # of VLAN tag insertions */
+
+	SGE_PSTAT_MAX		/* must be last */
+};
+
+struct sge_qset {		/* an SGE queue set */
+	struct sge_rspq rspq;
+	struct sge_fl fl[SGE_RXQ_PER_SET];
+	struct sge_txq txq[SGE_TXQ_PER_SET];
+	struct net_device *netdev;	/* associated net device */
+	unsigned long txq_stopped;	/* which Tx queues are stopped */
+	struct timer_list tx_reclaim_timer;	/* reclaims TX buffers */
+	unsigned long port_stats[SGE_PSTAT_MAX];
+} ____cacheline_aligned;
+
+struct sge {
+	struct sge_qset qs[SGE_QSETS];
+	spinlock_t reg_lock;	/* guards non-atomic SGE registers (eg context) */
+};
+
+struct adapter {
+	struct t3cdev tdev;
+	struct list_head adapter_list;
+	void __iomem *regs;
+	struct pci_dev *pdev;
+	unsigned long registered_device_map;
+	unsigned long open_device_map;
+	unsigned long flags;
+
+	const char *name;
+	int msg_enable;
+	unsigned int mmio_len;
+
+	struct adapter_params params;
+	unsigned int slow_intr_mask;
+	unsigned long irq_stats[IRQ_NUM_STATS];
+
+	struct {
+		unsigned short vec;
+		char desc[22];
+	} msix_info[SGE_QSETS + 1];
+
+	/* T3 modules */
+	struct sge sge;
+	struct mc7 pmrx;
+	struct mc7 pmtx;
+	struct mc7 cm;
+	struct mc5 mc5;
+
+	struct net_device *port[MAX_NPORTS];
+	unsigned int check_task_cnt;
+	struct delayed_work adap_check_task;
+	struct work_struct ext_intr_handler_task;
+
+	/*
+	 * Dummy netdevices are needed when using multiple receive queues with
+	 * NAPI as each netdevice can service only one queue.
+	 */
+	struct net_device *dummy_netdev[SGE_QSETS - 1];
+
+	struct dentry *debugfs_root;
+
+	struct mutex mdio_lock;
+	spinlock_t stats_lock;
+	spinlock_t work_lock;
+};
+
+static inline u32 t3_read_reg(struct adapter *adapter, u32 reg_addr)
+{
+	u32 val = readl(adapter->regs + reg_addr);
+
+	CH_DBG(adapter, MMIO, "read register 0x%x value 0x%x\n", reg_addr, val);
+	return val;
+}
+
+static inline void t3_write_reg(struct adapter *adapter, u32 reg_addr, u32 val)
+{
+	CH_DBG(adapter, MMIO, "setting register 0x%x to 0x%x\n", reg_addr, val);
+	writel(val, adapter->regs + reg_addr);
+}
+
+static inline struct port_info *adap2pinfo(struct adapter *adap, int idx)
+{
+	return netdev_priv(adap->port[idx]);
+}
+
+/*
+ * We use the spare atalk_ptr to map a net device to its SGE queue set.
+ * This is a macro so it can be used as l-value.
+ */
+#define dev2qset(netdev) ((netdev)->atalk_ptr)
+
+#define OFFLOAD_DEVMAP_BIT 15
+
+#define tdev2adap(d) container_of(d, struct adapter, tdev)
+
+static inline int offload_running(struct adapter *adapter)
+{
+	return test_bit(OFFLOAD_DEVMAP_BIT, &adapter->open_device_map);
+}
+
+int t3_offload_tx(struct t3cdev *tdev, struct sk_buff *skb);
+
+void t3_os_ext_intr_handler(struct adapter *adapter);
+void t3_os_link_changed(struct adapter *adapter, int port_id, int link_status,
+			int speed, int duplex, int fc);
+
+void t3_sge_start(struct adapter *adap);
+void t3_sge_stop(struct adapter *adap);
+void t3_free_sge_resources(struct adapter *adap);
+void t3_sge_err_intr_handler(struct adapter *adapter);
+intr_handler_t t3_intr_handler(struct adapter *adap, int polling);
+int t3_eth_xmit(struct sk_buff *skb, struct net_device *dev);
+void t3_update_qset_coalesce(struct sge_qset *qs, const struct qset_params *p);
+int t3_sge_alloc_qset(struct adapter *adapter, unsigned int id, int nports,
+		      int irq_vec_idx, const struct qset_params *p,
+		      int ntxq, struct net_device *netdev);
+int t3_get_desc(const struct sge_qset *qs, unsigned int qnum, unsigned int idx,
+		unsigned char *data);
+irqreturn_t t3_sge_intr_msix(int irq, void *cookie);
+
+#endif				/* __T3_ADAPTER_H__ */

drivers/net/cxgb3/ael1002.c

+/*
+ * This file is part of the Chelsio T3 Ethernet driver.
+ *
+ * Copyright (C) 2005-2006 Chelsio Communications.  All rights reserved.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the LICENSE file included in this
+ * release for licensing terms and conditions.
+ */
+
+#include "common.h"
+#include "regs.h"
+
+enum {
+	AEL100X_TX_DISABLE = 9,
+	AEL100X_TX_CONFIG1 = 0xc002,
+	AEL1002_PWR_DOWN_HI = 0xc011,
+	AEL1002_PWR_DOWN_LO = 0xc012,
+	AEL1002_XFI_EQL = 0xc015,
+	AEL1002_LB_EN = 0xc017,
+
+	LASI_CTRL = 0x9002,
+	LASI_STAT = 0x9005
+};
+
+static void ael100x_txon(struct cphy *phy)
+{
+	int tx_on_gpio = phy->addr == 0 ? F_GPIO7_OUT_VAL : F_GPIO2_OUT_VAL;
+
+	msleep(100);
+	t3_set_reg_field(phy->adapter, A_T3DBG_GPIO_EN, 0, tx_on_gpio);
+	msleep(30);
+}
+
+static int ael1002_power_down(struct cphy *phy, int enable)
+{
+	int err;
+
+	err = mdio_write(phy, MDIO_DEV_PMA_PMD, AEL100X_TX_DISABLE, !!enable);
+	if (!err)
+		err = t3_mdio_change_bits(phy, MDIO_DEV_PMA_PMD, MII_BMCR,
+					  BMCR_PDOWN, enable ? BMCR_PDOWN : 0);
+	return err;
+}
+
+static int ael1002_reset(struct cphy *phy, int wait)
+{
+	int err;
+
+	if ((err = ael1002_power_down(phy, 0)) ||
+	    (err = mdio_write(phy, MDIO_DEV_PMA_PMD, AEL100X_TX_CONFIG1, 1)) ||
+	    (err = mdio_write(phy, MDIO_DEV_PMA_PMD, AEL1002_PWR_DOWN_HI, 0)) ||
+	    (err = mdio_write(phy, MDIO_DEV_PMA_PMD, AEL1002_PWR_DOWN_LO, 0)) ||
+	    (err = mdio_write(phy, MDIO_DEV_PMA_PMD, AEL1002_XFI_EQL, 0x18)) ||
+	    (err = t3_mdio_change_bits(phy, MDIO_DEV_PMA_PMD, AEL1002_LB_EN,
+				       0, 1 << 5)))
+		return err;
+	return 0;
+}
+
+static int ael1002_intr_noop(struct cphy *phy)
+{
+	return 0;
+}
+
+static int ael100x_get_link_status(struct cphy *phy, int *link_ok,
+				   int *speed, int *duplex, int *fc)
+{
+	if (link_ok) {
+		unsigned int status;
+		int err = mdio_read(phy, MDIO_DEV_PMA_PMD, MII_BMSR, &status);
+
+		/*
+		 * BMSR_LSTATUS is latch-low, so if it is 0 we need to read it
+		 * once more to get the current link state.
+		 */
+		if (!err && !(status & BMSR_LSTATUS))
+			err = mdio_read(phy, MDIO_DEV_PMA_PMD, MII_BMSR,
+					&status);
+		if (err)
+			return err;
+		*link_ok = !!(status & BMSR_LSTATUS);
+	}
+	if (speed)
+		*speed = SPEED_10000;
+	if (duplex)
+		*duplex = DUPLEX_FULL;
+	return 0;
+}
+
+static struct cphy_ops ael1002_ops = {
+	.reset = ael1002_reset,
+	.intr_enable = ael1002_intr_noop,
+	.intr_disable = ael1002_intr_noop,
+	.intr_clear = ael1002_intr_noop,
+	.intr_handler = ael1002_intr_noop,
+	.get_link_status = ael100x_get_link_status,
+	.power_down = ael1002_power_down,
+};
+
+void t3_ael1002_phy_prep(struct cphy *phy, struct adapter *adapter,
+			 int phy_addr, const struct mdio_ops *mdio_ops)
+{
+	cphy_init(phy, adapter, phy_addr, &ael1002_ops, mdio_ops);
+	ael100x_txon(phy);
+}
+
+static int ael1006_reset(struct cphy *phy, int wait)
+{
+	return t3_phy_reset(phy, MDIO_DEV_PMA_PMD, wait);
+}
+
+static int ael1006_intr_enable(struct cphy *phy)
+{
+	return mdio_write(phy, MDIO_DEV_PMA_PMD, LASI_CTRL, 1);
+}
+
+static int ael1006_intr_disable(struct cphy *phy)
+{
+	return mdio_write(phy, MDIO_DEV_PMA_PMD, LASI_CTRL, 0);
+}
+
+static int ael1006_intr_clear(struct cphy *phy)
+{
+	u32 val;
+
+	return mdio_read(phy, MDIO_DEV_PMA_PMD, LASI_STAT, &val);
+}
+
+static int ael1006_intr_handler(struct cphy *phy)
+{
+	unsigned int status;
+	int err = mdio_read(phy, MDIO_DEV_PMA_PMD, LASI_STAT, &status);
+
+	if (err)
+		return err;
+	return (status & 1) ? cphy_cause_link_change : 0;
+}
+
+static int ael1006_power_down(struct cphy *phy, int enable)
+{
+	return t3_mdio_change_bits(phy, MDIO_DEV_PMA_PMD, MII_BMCR,
+				   BMCR_PDOWN, enable ? BMCR_PDOWN : 0);
+}
+
+static struct cphy_ops ael1006_ops = {
+	.reset = ael1006_reset,
+	.intr_enable = ael1006_intr_enable,
+	.intr_disable = ael1006_intr_disable,
+	.intr_clear = ael1006_intr_clear,
+	.intr_handler = ael1006_intr_handler,
+	.get_link_status = ael100x_get_link_status,
+	.power_down = ael1006_power_down,
+};
+
+void t3_ael1006_phy_prep(struct cphy *phy, struct adapter *adapter,
+			 int phy_addr, const struct mdio_ops *mdio_ops)
+{
+	cphy_init(phy, adapter, phy_addr, &ael1006_ops, mdio_ops);
+	ael100x_txon(phy);
+}
+
+static struct cphy_ops qt2045_ops = {
+	.reset = ael1006_reset,
+	.intr_enable = ael1006_intr_enable,
+	.intr_disable = ael1006_intr_disable,
+	.intr_clear = ael1006_intr_clear,
+	.intr_handler = ael1006_intr_handler,
+	.get_link_status = ael100x_get_link_status,
+	.power_down = ael1006_power_down,
+};
+
+void t3_qt2045_phy_prep(struct cphy *phy, struct adapter *adapter,
+			int phy_addr, const struct mdio_ops *mdio_ops)
+{
+	unsigned int stat;
+
+	cphy_init(phy, adapter, phy_addr, &qt2045_ops, mdio_ops);
+
+	/*
+	 * Some cards where the PHY is supposed to be at address 0 actually
+	 * have it at 1.
+	 */
+	if (!phy_addr && !mdio_read(phy, MDIO_DEV_PMA_PMD, MII_BMSR, &stat) &&
+	    stat == 0xffff)
+		phy->addr = 1;
+}
+
+static int xaui_direct_reset(struct cphy *phy, int wait)
+{
+	return 0;
+}
+
+static int xaui_direct_get_link_status(struct cphy *phy, int *link_ok,
+				       int *speed, int *duplex, int *fc)
+{
+	if (link_ok) {
+		unsigned int status;
+
+		status = t3_read_reg(phy->adapter,
+				     XGM_REG(A_XGM_SERDES_STAT0, phy->addr));
+		*link_ok = !(status & F_LOWSIG0);
+	}
+	if (speed)
+		*speed = SPEED_10000;
+	if (duplex)
+		*duplex = DUPLEX_FULL;
+	return 0;
+}
+
+static int xaui_direct_power_down(struct cphy *phy, int enable)
+{
+	return 0;
+}
+
+static struct cphy_ops xaui_direct_ops = {
+	.reset = xaui_direct_reset,
+	.intr_enable = ael1002_intr_noop,
+	.intr_disable = ael1002_intr_noop,
+	.intr_clear = ael1002_intr_noop,
+	.intr_handler = ael1002_intr_noop,
+	.get_link_status = xaui_direct_get_link_status,
+	.power_down = xaui_direct_power_down,
+};
+
+void t3_xaui_direct_phy_prep(struct cphy *phy, struct adapter *adapter,
+			     int phy_addr, const struct mdio_ops *mdio_ops)
+{
+	cphy_init(phy, adapter, 1, &xaui_direct_ops, mdio_ops);
+}

drivers/net/cxgb3/cxgb3_ctl_defs.h

+/*
+ * Copyright (C) 2003-2006 Chelsio Communications.  All rights reserved.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the LICENSE file included in this
+ * release for licensing terms and conditions.
+ */
+
+#ifndef _CXGB3_OFFLOAD_CTL_DEFS_H
+#define _CXGB3_OFFLOAD_CTL_DEFS_H
+
+enum {
+	GET_MAX_OUTSTANDING_WR,
+	GET_TX_MAX_CHUNK,
+	GET_TID_RANGE,
+	GET_STID_RANGE,
+	GET_RTBL_RANGE,
+	GET_L2T_CAPACITY,
+	GET_MTUS,
+	GET_WR_LEN,
+	GET_IFF_FROM_MAC,
+	GET_DDP_PARAMS,
+	GET_PORTS,
+
+	ULP_ISCSI_GET_PARAMS,
+	ULP_ISCSI_SET_PARAMS,
+
+	RDMA_GET_PARAMS,
+	RDMA_CQ_OP,
+	RDMA_CQ_SETUP,
+	RDMA_CQ_DISABLE,
+	RDMA_CTRL_QP_SETUP,
+	RDMA_GET_MEM,
+};
+
+/*
+ * Structure used to describe a TID range.  Valid TIDs are [base, base+num).
+ */
+struct tid_range {
+	unsigned int base;	/* first TID */
+	unsigned int num;	/* number of TIDs in range */
+};
+
+/*
+ * Structure used to request the size and contents of the MTU table.
+ */
+struct mtutab {
+	unsigned int size;	/* # of entries in the MTU table */
+	const unsigned short *mtus;	/* the MTU table values */
+};
+
+struct net_device;
+
+/*
+ * Structure used to request the adapter net_device owning a given MAC address.
+ */
+struct iff_mac {
+	struct net_device *dev;	/* the net_device */
+	const unsigned char *mac_addr;	/* MAC address to lookup */
+	u16 vlan_tag;
+};
+
+struct pci_dev;
+
+/*
+ * Structure used to request the TCP DDP parameters.
+ */
+struct ddp_params {
+	unsigned int llimit;	/* TDDP region start address */
+	unsigned int ulimit;	/* TDDP region end address */
+	unsigned int tag_mask;	/* TDDP tag mask */
+	struct pci_dev *pdev;
+};
+
+struct adap_ports {
+	unsigned int nports;	/* number of ports on this adapter */
+	struct net_device *lldevs[2];
+};
+