cxgb4vf: Add main T4 PCI-E SR-IOV Virtual Function driver for cxgb4vf

Add main T4 PCI-E SR-IOV Virtual Function driver for "cxgb4vf".

Signed-off-by: Casey Leedom
Signed-off-by: David S. Miller <davem@davemloft.net>

2 files changed, 3446 insertions, 0 deletions

diff --git a/drivers/net/cxgb4vf/adapter.h b/drivers/net/cxgb4vf/adapter.h
new file mode 100644
index 00000000000..8ea01962e04
--- /dev/null
+++ b/drivers/net/cxgb4vf/adapter.h
@@ -0,0 +1,540 @@
+/*
+ * This file is part of the Chelsio T4 PCI-E SR-IOV Virtual Function Ethernet
+ * driver for Linux.
+ *
+ * Copyright (c) 2009-2010 Chelsio Communications, Inc. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer in the documentation and/or other materials
+ *        provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+/*
+ * This file should not be included directly.  Include t4vf_common.h instead.
+ */
+
+#ifndef __CXGB4VF_ADAPTER_H__
+#define __CXGB4VF_ADAPTER_H__
+
+#include <linux/pci.h>
+#include <linux/spinlock.h>
+#include <linux/skbuff.h>
+#include <linux/if_ether.h>
+#include <linux/netdevice.h>
+
+#include "../cxgb4/t4_hw.h"
+
+/*
+ * Constants of the implementation.
+ */
+enum {
+	MAX_NPORTS	= 1,		/* max # of "ports" */
+	MAX_PORT_QSETS	= 8,		/* max # of Queue Sets / "port" */
+	MAX_ETH_QSETS	= MAX_NPORTS*MAX_PORT_QSETS,
+
+	/*
+	 * MSI-X interrupt index usage.
+	 */
+	MSIX_FW		= 0,		/* MSI-X index for firmware Q */
+	MSIX_NIQFLINT	= 1,		/* MSI-X index base for Ingress Qs */
+	MSIX_EXTRAS	= 1,
+	MSIX_ENTRIES	= MAX_ETH_QSETS + MSIX_EXTRAS,
+
+	/*
+	 * The maximum number of Ingress and Egress Queues is determined by
+	 * the maximum number of "Queue Sets" which we support plus any
+	 * ancillary queues.  Each "Queue Set" requires one Ingress Queue
+	 * for RX Packet Ingress Event notifications and two Egress Queues for
+	 * a Free List and an Ethernet TX list.
+	 */
+	INGQ_EXTRAS	= 2,		/* firmware event queue and */
+					/*   forwarded interrupts */
+	MAX_INGQ	= MAX_ETH_QSETS+INGQ_EXTRAS,
+	MAX_EGRQ	= MAX_ETH_QSETS*2,
+};
+
+/*
+ * Forward structure definition references.
+ */
+struct adapter;
+struct sge_eth_rxq;
+struct sge_rspq;
+
+/*
+ * Per-"port" information.  This is really per-Virtual Interface information
+ * but the use of the "port" nomanclature makes it easier to go back and forth
+ * between the PF and VF drivers ...
+ */
+struct port_info {
+	struct adapter *adapter;	/* our adapter */
+	struct vlan_group *vlan_grp;	/* out VLAN group */
+	u16 viid;			/* virtual interface ID */
+	s16 xact_addr_filt;		/* index of our MAC address filter */
+	u16 rss_size;			/* size of VI's RSS table slice */
+	u8 pidx;			/* index into adapter port[] */
+	u8 port_id;			/* physical port ID */
+	u8 rx_offload;			/* CSO, etc. */
+	u8 nqsets;			/* # of "Queue Sets" */
+	u8 first_qset;			/* index of first "Queue Set" */
+	struct link_config link_cfg;	/* physical port configuration */
+};
+
+/* port_info.rx_offload flags */
+enum {
+	RX_CSO = 1 << 0,
+};
+
+/*
+ * Scatter Gather Engine resources for the "adapter".  Our ingress and egress
+ * queues are organized into "Queue Sets" with one ingress and one egress
+ * queue per Queue Set.  These Queue Sets are aportionable between the "ports"
+ * (Virtual Interfaces).  One extra ingress queue is used to receive
+ * asynchronous messages from the firmware.  Note that the "Queue IDs" that we
+ * use here are really "Relative Queue IDs" which are returned as part of the
+ * firmware command to allocate queues.  These queue IDs are relative to the
+ * absolute Queue ID base of the section of the Queue ID space allocated to
+ * the PF/VF.
+ */
+
+/*
+ * SGE free-list queue state.
+ */
+struct rx_sw_desc;
+struct sge_fl {
+	unsigned int avail;		/* # of available RX buffers */
+	unsigned int pend_cred;		/* new buffers since last FL DB ring */
+	unsigned int cidx;		/* consumer index */
+	unsigned int pidx;		/* producer index */
+	unsigned long alloc_failed;	/* # of buffer allocation failures */
+	unsigned long large_alloc_failed;
+	unsigned long starving;		/* # of times FL was found starving */
+
+	/*
+	 * Write-once/infrequently fields.
+	 * -------------------------------
+	 */
+
+	unsigned int cntxt_id;		/* SGE relative QID for the free list */
+	unsigned int abs_id;		/* SGE absolute QID for the free list */
+	unsigned int size;		/* capacity of free list */
+	struct rx_sw_desc *sdesc;	/* address of SW RX descriptor ring */
+	__be64 *desc;			/* address of HW RX descriptor ring */
+	dma_addr_t addr;		/* PCI bus address of hardware ring */
+};
+
+/*
+ * An ingress packet gather list.
+ */
+struct pkt_gl {
+	skb_frag_t frags[MAX_SKB_FRAGS];
+	void *va;			/* virtual address of first byte */
+	unsigned int nfrags;		/* # of fragments */
+	unsigned int tot_len;		/* total length of fragments */
+};
+
+typedef int (*rspq_handler_t)(struct sge_rspq *, const __be64 *,
+			      const struct pkt_gl *);
+
+/*
+ * State for an SGE Response Queue.
+ */
+struct sge_rspq {
+	struct napi_struct napi;	/* NAPI scheduling control */
+	const __be64 *cur_desc;		/* current descriptor in queue */
+	unsigned int cidx;		/* consumer index */
+	u8 gen;				/* current generation bit */
+	u8 next_intr_params;		/* holdoff params for next interrupt */
+	int offset;			/* offset into current FL buffer */
+
+	unsigned int unhandled_irqs;	/* bogus interrupts */
+
+	/*
+	 * Write-once/infrequently fields.
+	 * -------------------------------
+	 */
+
+	u8 intr_params;			/* interrupt holdoff parameters */
+	u8 pktcnt_idx;			/* interrupt packet threshold */
+	u8 idx;				/* queue index within its group */
+	u16 cntxt_id;			/* SGE rel QID for the response Q */
+	u16 abs_id;			/* SGE abs QID for the response Q */
+	__be64 *desc;			/* address of hardware response ring */
+	dma_addr_t phys_addr;		/* PCI bus address of ring */
+	unsigned int iqe_len;		/* entry size */
+	unsigned int size;		/* capcity of response Q */
+	struct adapter *adapter;	/* our adapter */
+	struct net_device *netdev;	/* associated net device */
+	rspq_handler_t handler;		/* the handler for this response Q */
+};
+
+/*
+ * Ethernet queue statistics
+ */
+struct sge_eth_stats {
+	unsigned long pkts;		/* # of ethernet packets */
+	unsigned long lro_pkts;		/* # of LRO super packets */
+	unsigned long lro_merged;	/* # of wire packets merged by LRO */
+	unsigned long rx_cso;		/* # of Rx checksum offloads */
+	unsigned long vlan_ex;		/* # of Rx VLAN extractions */
+	unsigned long rx_drops;		/* # of packets dropped due to no mem */
+};
+
+/*
+ * State for an Ethernet Receive Queue.
+ */
+struct sge_eth_rxq {
+	struct sge_rspq rspq;		/* Response Queue */
+	struct sge_fl fl;		/* Free List */
+	struct sge_eth_stats stats;	/* receive statistics */
+};
+
+/*
+ * SGE Transmit Queue state.  This contains all of the resources associated
+ * with the hardware status of a TX Queue which is a circular ring of hardware
+ * TX Descriptors.  For convenience, it also contains a pointer to a parallel
+ * "Software Descriptor" array but we don't know anything about it here other
+ * than its type name.
+ */
+struct tx_desc {
+	/*
+	 * Egress Queues are measured in units of SGE_EQ_IDXSIZE by the
+	 * hardware: Sizes, Producer and Consumer indices, etc.
+	 */
+	__be64 flit[SGE_EQ_IDXSIZE/sizeof(__be64)];
+};
+struct tx_sw_desc;
+struct sge_txq {
+	unsigned int in_use;		/* # of in-use TX descriptors */
+	unsigned int size;		/* # of descriptors */
+	unsigned int cidx;		/* SW consumer index */
+	unsigned int pidx;		/* producer index */
+	unsigned long stops;		/* # of times queue has been stopped */
+	unsigned long restarts;		/* # of queue restarts */
+
+	/*
+	 * Write-once/infrequently fields.
+	 * -------------------------------
+	 */
+
+	unsigned int cntxt_id;		/* SGE relative QID for the TX Q */
+	unsigned int abs_id;		/* SGE absolute QID for the TX Q */
+	struct tx_desc *desc;		/* address of HW TX descriptor ring */
+	struct tx_sw_desc *sdesc;	/* address of SW TX descriptor ring */
+	struct sge_qstat *stat;		/* queue status entry */
+	dma_addr_t phys_addr;		/* PCI bus address of hardware ring */
+};
+
+/*
+ * State for an Ethernet Transmit Queue.
+ */
+struct sge_eth_txq {
+	struct sge_txq q;		/* SGE TX Queue */
+	struct netdev_queue *txq;	/* associated netdev TX queue */
+	unsigned long tso;		/* # of TSO requests */
+	unsigned long tx_cso;		/* # of TX checksum offloads */
+	unsigned long vlan_ins;		/* # of TX VLAN insertions */
+	unsigned long mapping_err;	/* # of I/O MMU packet mapping errors */
+};
+
+/*
+ * The complete set of Scatter/Gather Engine resources.
+ */
+struct sge {
+	/*
+	 * Our "Queue Sets" ...
+	 */
+	struct sge_eth_txq ethtxq[MAX_ETH_QSETS];
+	struct sge_eth_rxq ethrxq[MAX_ETH_QSETS];
+
+	/*
+	 * Extra ingress queues for asynchronous firmware events and
+	 * forwarded interrupts (when in MSI mode).
+	 */
+	struct sge_rspq fw_evtq ____cacheline_aligned_in_smp;
+
+	struct sge_rspq intrq ____cacheline_aligned_in_smp;
+	spinlock_t intrq_lock;
+
+	/*
+	 * State for managing "starving Free Lists" -- Free Lists which have
+	 * fallen below a certain threshold of buffers available to the
+	 * hardware and attempts to refill them up to that threshold have
+	 * failed.  We have a regular "slow tick" timer process which will
+	 * make periodic attempts to refill these starving Free Lists ...
+	 */
+	DECLARE_BITMAP(starving_fl, MAX_EGRQ);
+	struct timer_list rx_timer;
+
+	/*
+	 * State for cleaning up completed TX descriptors.
+	 */
+	struct timer_list tx_timer;
+
+	/*
+	 * Write-once/infrequently fields.
+	 * -------------------------------
+	 */
+
+	u16 max_ethqsets;		/* # of available Ethernet queue sets */
+	u16 ethqsets;			/* # of active Ethernet queue sets */
+	u16 ethtxq_rover;		/* Tx queue to clean up next */
+	u16 timer_val[SGE_NTIMERS];	/* interrupt holdoff timer array */
+	u8 counter_val[SGE_NCOUNTERS];	/* interrupt RX threshold array */
+
+	/*
+	 * Reverse maps from Absolute Queue IDs to associated queue pointers.
+	 * The absolute Queue IDs are in a compact range which start at a
+	 * [potentially large] Base Queue ID.  We perform the reverse map by
+	 * first converting the Absolute Queue ID into a Relative Queue ID by
+	 * subtracting off the Base Queue ID and then use a Relative Queue ID
+	 * indexed table to get the pointer to the corresponding software
+	 * queue structure.
+	 */
+	unsigned int egr_base;
+	unsigned int ingr_base;
+	void *egr_map[MAX_EGRQ];
+	struct sge_rspq *ingr_map[MAX_INGQ];
+};
+
+/*
+ * Utility macros to convert Absolute- to Relative-Queue indices and Egress-
+ * and Ingress-Queues.  The EQ_MAP() and IQ_MAP() macros which provide
+ * pointers to Ingress- and Egress-Queues can be used as both L- and R-values
+ */
+#define EQ_IDX(s, abs_id) ((unsigned int)((abs_id) - (s)->egr_base))
+#define IQ_IDX(s, abs_id) ((unsigned int)((abs_id) - (s)->ingr_base))
+
+#define EQ_MAP(s, abs_id) ((s)->egr_map[EQ_IDX(s, abs_id)])
+#define IQ_MAP(s, abs_id) ((s)->ingr_map[IQ_IDX(s, abs_id)])
+
+/*
+ * Macro to iterate across Queue Sets ("rxq" is a historic misnomer).
+ */
+#define for_each_ethrxq(sge, iter) \
+	for (iter = 0; iter < (sge)->ethqsets; iter++)
+
+/*
+ * Per-"adapter" (Virtual Function) information.
+ */
+struct adapter {
+	/* PCI resources */
+	void __iomem *regs;
+	struct pci_dev *pdev;
+	struct device *pdev_dev;
+
+	/* "adapter" resources */
+	unsigned long registered_device_map;
+	unsigned long open_device_map;
+	unsigned long flags;
+	struct adapter_params params;
+
+	/* queue and interrupt resources */
+	struct {
+		unsigned short vec;
+		char desc[22];
+	} msix_info[MSIX_ENTRIES];
+	struct sge sge;
+
+	/* Linux network device resources */
+	struct net_device *port[MAX_NPORTS];
+	const char *name;
+	unsigned int msg_enable;
+
+	/* debugfs resources */
+	struct dentry *debugfs_root;
+
+	/* various locks */
+	spinlock_t stats_lock;
+};
+
+enum { /* adapter flags */
+	FULL_INIT_DONE     = (1UL << 0),
+	USING_MSI          = (1UL << 1),
+	USING_MSIX         = (1UL << 2),
+	QUEUES_BOUND       = (1UL << 3),
+};
+
+/*
+ * The following register read/write routine definitions are required by
+ * the common code.
+ */
+
+/**
+ * t4_read_reg - read a HW register
+ * @adapter: the adapter
+ * @reg_addr: the register address
+ *
+ * Returns the 32-bit value of the given HW register.
+ */
+static inline u32 t4_read_reg(struct adapter *adapter, u32 reg_addr)
+{
+	return readl(adapter->regs + reg_addr);
+}
+
+/**
+ * t4_write_reg - write a HW register
+ * @adapter: the adapter
+ * @reg_addr: the register address
+ * @val: the value to write
+ *
+ * Write a 32-bit value into the given HW register.
+ */
+static inline void t4_write_reg(struct adapter *adapter, u32 reg_addr, u32 val)
+{
+	writel(val, adapter->regs + reg_addr);
+}
+
+#ifndef readq
+static inline u64 readq(const volatile void __iomem *addr)
+{
+	return readl(addr) + ((u64)readl(addr + 4) << 32);
+}
+
+static inline void writeq(u64 val, volatile void __iomem *addr)
+{
+	writel(val, addr);
+	writel(val >> 32, addr + 4);
+}
+#endif
+
+/**
+ * t4_read_reg64 - read a 64-bit HW register
+ * @adapter: the adapter
+ * @reg_addr: the register address
+ *
+ * Returns the 64-bit value of the given HW register.
+ */
+static inline u64 t4_read_reg64(struct adapter *adapter, u32 reg_addr)
+{
+	return readq(adapter->regs + reg_addr);
+}
+
+/**
+ * t4_write_reg64 - write a 64-bit HW register
+ * @adapter: the adapter
+ * @reg_addr: the register address
+ * @val: the value to write
+ *
+ * Write a 64-bit value into the given HW register.
+ */
+static inline void t4_write_reg64(struct adapter *adapter, u32 reg_addr,
+				  u64 val)
+{
+	writeq(val, adapter->regs + reg_addr);
+}
+
+/**
+ * port_name - return the string name of a port
+ * @adapter: the adapter
+ * @pidx: the port index
+ *
+ * Return the string name of the selected port.
+ */
+static inline const char *port_name(struct adapter *adapter, int pidx)
+{
+	return adapter->port[pidx]->name;
+}
+
+/**
+ * t4_os_set_hw_addr - store a port's MAC address in SW
+ * @adapter: the adapter
+ * @pidx: the port index
+ * @hw_addr: the Ethernet address
+ *
+ * Store the Ethernet address of the given port in SW.  Called by the common
+ * code when it retrieves a port's Ethernet address from EEPROM.
+ */
+static inline void t4_os_set_hw_addr(struct adapter *adapter, int pidx,
+				     u8 hw_addr[])
+{
+	memcpy(adapter->port[pidx]->dev_addr, hw_addr, ETH_ALEN);
+	memcpy(adapter->port[pidx]->perm_addr, hw_addr, ETH_ALEN);
+}
+
+/**
+ * netdev2pinfo - return the port_info structure associated with a net_device
+ * @dev: the netdev
+ *
+ * Return the struct port_info associated with a net_device
+ */
+static inline struct port_info *netdev2pinfo(const struct net_device *dev)
+{
+	return netdev_priv(dev);
+}
+
+/**
+ * adap2pinfo - return the port_info of a port
+ * @adap: the adapter
+ * @pidx: the port index
+ *
+ * Return the port_info structure for the adapter.
+ */
+static inline struct port_info *adap2pinfo(struct adapter *adapter, int pidx)
+{
+	return netdev_priv(adapter->port[pidx]);
+}
+
+/**
+ * netdev2adap - return the adapter structure associated with a net_device
+ * @dev: the netdev
+ *
+ * Return the struct adapter associated with a net_device
+ */
+static inline struct adapter *netdev2adap(const struct net_device *dev)
+{
+	return netdev2pinfo(dev)->adapter;
+}
+
+/*
+ * OS "Callback" function declarations.  These are functions that the OS code
+ * is "contracted" to provide for the common code.
+ */
+void t4vf_os_link_changed(struct adapter *, int, int);
+
+/*
+ * SGE function prototype declarations.
+ */
+int t4vf_sge_alloc_rxq(struct adapter *, struct sge_rspq *, bool,
+		       struct net_device *, int,
+		       struct sge_fl *, rspq_handler_t);
+int t4vf_sge_alloc_eth_txq(struct adapter *, struct sge_eth_txq *,
+			   struct net_device *, struct netdev_queue *,
+			   unsigned int);
+void t4vf_free_sge_resources(struct adapter *);
+
+int t4vf_eth_xmit(struct sk_buff *, struct net_device *);
+int t4vf_ethrx_handler(struct sge_rspq *, const __be64 *,
+		       const struct pkt_gl *);
+
+irq_handler_t t4vf_intr_handler(struct adapter *);
+irqreturn_t t4vf_sge_intr_msix(int, void *);
+
+int t4vf_sge_init(struct adapter *);
+void t4vf_sge_start(struct adapter *);
+void t4vf_sge_stop(struct adapter *);
+
+#endif /* __CXGB4VF_ADAPTER_H__ */
diff --git a/drivers/net/cxgb4vf/cxgb4vf_main.c b/drivers/net/cxgb4vf/cxgb4vf_main.c
new file mode 100644
index 00000000000..bd73ff5b51b
--- /dev/null
+++ b/drivers/net/cxgb4vf/cxgb4vf_main.c
@@ -0,0 +1,2906 @@
+/*
+ * This file is part of the Chelsio T4 PCI-E SR-IOV Virtual Function Ethernet
+ * driver for Linux.
+ *
+ * Copyright (c) 2009-2010 Chelsio Communications, Inc. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer in the documentation and/or other materials
+ *        provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#include <linux/version.h>
+#include <linux/module.h>
+#include <linux/moduleparam.h>
+#include <linux/init.h>
+#include <linux/pci.h>
+#include <linux/dma-mapping.h>
+#include <linux/netdevice.h>
+#include <linux/etherdevice.h>
+#include <linux/debugfs.h>
+#include <linux/ethtool.h>
+
+#include "t4vf_common.h"
+#include "t4vf_defs.h"
+
+#include "../cxgb4/t4_regs.h"
+#include "../cxgb4/t4_msg.h"
+
+/*
+ * Generic information about the driver.
+ */
+#define DRV_VERSION "1.0.0"
+#define DRV_DESC "Chelsio T4 Virtual Function (VF) Network Driver"
+
+/*
+ * Module Parameters.
+ * ==================
+ */
+
+/*
+ * Default ethtool "message level" for adapters.
+ */
+#define DFLT_MSG_ENABLE (NETIF_MSG_DRV | NETIF_MSG_PROBE | NETIF_MSG_LINK | \
+			 NETIF_MSG_TIMER | NETIF_MSG_IFDOWN | NETIF_MSG_IFUP |\
+			 NETIF_MSG_RX_ERR | NETIF_MSG_TX_ERR)
+
+static int dflt_msg_enable = DFLT_MSG_ENABLE;
+
+module_param(dflt_msg_enable, int, 0644);
+MODULE_PARM_DESC(dflt_msg_enable,
+		 "default adapter ethtool message level bitmap");
+
+/*
+ * The driver uses the best interrupt scheme available on a platform in the
+ * order MSI-X then MSI.  This parameter determines which of these schemes the
+ * driver may consider as follows:
+ *
+ *     msi = 2: choose from among MSI-X and MSI
+ *     msi = 1: only consider MSI interrupts
+ *
+ * Note that unlike the Physical Function driver, this Virtual Function driver
+ * does _not_ support legacy INTx interrupts (this limitation is mandated by
+ * the PCI-E SR-IOV standard).
+ */
+#define MSI_MSIX	2
+#define MSI_MSI		1
+#define MSI_DEFAULT	MSI_MSIX
+
+static int msi = MSI_DEFAULT;
+
+module_param(msi, int, 0644);
+MODULE_PARM_DESC(msi, "whether to use MSI-X or MSI");
+
+/*
+ * Fundamental constants.
+ * ======================
+ */
+
+enum {
+	MAX_TXQ_ENTRIES		= 16384,
+	MAX_RSPQ_ENTRIES	= 16384,
+	MAX_RX_BUFFERS		= 16384,
+
+	MIN_TXQ_ENTRIES		= 32,
+	MIN_RSPQ_ENTRIES	= 128,
+	MIN_FL_ENTRIES		= 16,
+
+	/*
+	 * For purposes of manipulating the Free List size we need to
+	 * recognize that Free Lists are actually Egress Queues (the host
+	 * produces free buffers which the hardware consumes), Egress Queues
+	 * indices are all in units of Egress Context Units bytes, and free
+	 * list entries are 64-bit PCI DMA addresses.  And since the state of
+	 * the Producer Index == the Consumer Index implies an EMPTY list, we
+	 * always have at least one Egress Unit's worth of Free List entries
+	 * unused.  See sge.c for more details ...
+	 */
+	EQ_UNIT = SGE_EQ_IDXSIZE,
+	FL_PER_EQ_UNIT = EQ_UNIT / sizeof(__be64),
+	MIN_FL_RESID = FL_PER_EQ_UNIT,
+};
+
+/*
+ * Global driver state.
+ * ====================
+ */
+
+static struct dentry *cxgb4vf_debugfs_root;
+
+/*
+ * OS "Callback" functions.
+ * ========================
+ */
+
+/*
+ * The link status has changed on the indicated "port" (Virtual Interface).
+ */
+void t4vf_os_link_changed(struct adapter *adapter, int pidx, int link_ok)
+{
+	struct net_device *dev = adapter->port[pidx];
+
+	/*
+	 * If the port is disabled or the current recorded "link up"
+	 * status matches the new status, just return.
+	 */
+	if (!netif_running(dev) || link_ok == netif_carrier_ok(dev))
+		return;
+
+	/*
+	 * Tell the OS that the link status has changed and print a short
+	 * informative message on the console about the event.
+	 */
+	if (link_ok) {
+		const char *s;
+		const char *fc;
+		const struct port_info *pi = netdev_priv(dev);
+
+		netif_carrier_on(dev);
+
+		switch (pi->link_cfg.speed) {
+		case SPEED_10000:
+			s = "10Gbps";
+			break;
+
+		case SPEED_1000:
+			s = "1000Mbps";
+			break;
+
+		case SPEED_100:
+			s = "100Mbps";
+			break;
+
+		default:
+			s = "unknown";
+			break;
+		}
+
+		switch (pi->link_cfg.fc) {
+		case PAUSE_RX:
+			fc = "RX";
+			break;
+
+		case PAUSE_TX:
+			fc = "TX";
+			break;
+
+		case PAUSE_RX|PAUSE_TX:
+			fc = "RX/TX";
+			break;
+
+		default:
+			fc = "no";
+			break;
+		}
+
+		printk(KERN_INFO "%s: link up, %s, full-duplex, %s PAUSE\n",
+		       dev->name, s, fc);
+	} else {
+		netif_carrier_off(dev);
+		printk(KERN_INFO "%s: link down\n", dev->name);
+	}
+}
+
+/*
+ * Net device operations.
+ * ======================
+ */
+
+/*
+ * Record our new VLAN Group and enable/disable hardware VLAN Tag extraction
+ * based on whether the specified VLAN Group pointer is NULL or not.
+ */
+static void cxgb4vf_vlan_rx_register(struct net_device *dev,
+				     struct vlan_group *grp)
+{
+	struct port_info *pi = netdev_priv(dev);
+
+	pi->vlan_grp = grp;
+	t4vf_set_rxmode(pi->adapter, pi->viid, -1, -1, -1, -1, grp != NULL, 0);
+}
+
+/*
+ * Perform the MAC and PHY actions needed to enable a "port" (Virtual
+ * Interface).
+ */
+static int link_start(struct net_device *dev)
+{
+	int ret;
+	struct port_info *pi = netdev_priv(dev);
+
+	/*
+	 * We do not set address filters and promiscuity here, the stack does
+	 * that step explicitly.
+	 */
+	ret = t4vf_set_rxmode(pi->adapter, pi->viid, dev->mtu, -1, -1, -1, -1,
+			      true);
+	if (ret == 0) {
+		ret = t4vf_change_mac(pi->adapter, pi->viid,
+				      pi->xact_addr_filt, dev->dev_addr, true);
+		if (ret >= 0) {
+			pi->xact_addr_filt = ret;
+			ret = 0;
+		}
+	}
+
+	/*
+	 * We don't need to actually "start the link" itself since the
+	 * firmware will do that for us when the first Virtual Interface
+	 * is enabled on a port.
+	 */
+	if (ret == 0)
+		ret = t4vf_enable_vi(pi->adapter, pi->viid, true, true);
+	return ret;
+}
+
+/*
+ * Name the MSI-X interrupts.
+ */
+static void name_msix_vecs(struct adapter *adapter)
+{
+	int namelen = sizeof(adapter->msix_info[0].desc) - 1;
+	int pidx;
+
+	/*
+	 * Firmware events.
+	 */
+	snprintf(adapter->msix_info[MSIX_FW].desc, namelen,
+		 "%s-FWeventq", adapter->name);
+	adapter->msix_info[MSIX_FW].desc[namelen] = 0;
+
+	/*
+	 * Ethernet queues.
+	 */
+	for_each_port(adapter, pidx) {
+		struct net_device *dev = adapter->port[pidx];
+		const struct port_info *pi = netdev_priv(dev);
+		int qs, msi;
+
+		for (qs = 0, msi = MSIX_NIQFLINT;
+		     qs < pi->nqsets;
+		     qs++, msi++) {
+			snprintf(adapter->msix_info[msi].desc, namelen,
+				 "%s-%d", dev->name, qs);
+			adapter->msix_info[msi].desc[namelen] = 0;
+		}
+	}
+}
+
+/*
+ * Request all of our MSI-X resources.
+ */
+static int request_msix_queue_irqs(struct adapter *adapter)
+{
+	struct sge *s = &adapter->sge;
+	int rxq, msi, err;
+
+	/*
+	 * Firmware events.
+	 */
+	err = request_irq(adapter->msix_info[MSIX_FW].vec, t4vf_sge_intr_msix,
+			  0, adapter->msix_info[MSIX_FW].desc, &s->fw_evtq);
+	if (err)
+		return err;
+
+	/*
+	 * Ethernet queues.
+	 */
+	msi = MSIX_NIQFLINT;
+	for_each_ethrxq(s, rxq) {
+		err = request_irq(adapter->msix_info[msi].vec,
+				  t4vf_sge_intr_msix, 0,
+				  adapter->msix_info[msi].desc,
+				  &s->ethrxq[rxq].rspq);
+		if (err)
+			goto err_free_irqs;
+		msi++;
+	}
+	return 0;
+
+err_free_irqs:
+	while (--rxq >= 0)
+		free_irq(adapter->msix_info[--msi].vec, &s->ethrxq[rxq].rspq);
+	free_irq(adapter->msix_info[MSIX_FW].vec, &s->fw_evtq);
+	return err;
+}
+
+/*
+ * Free our MSI-X resources.
+ */
+static void free_msix_queue_irqs(struct adapter *adapter)
+{
+	struct sge *s = &adapter->sge;
+	int rxq, msi;
+
+	free_irq(adapter->msix_info[MSIX_FW].vec, &s->fw_evtq);
+	msi = MSIX_NIQFLINT;
+	for_each_ethrxq(s, rxq)
+		free_irq(adapter->msix_info[msi++].vec,
+			 &s->ethrxq[rxq].rspq);
+}
+
+/*
+ * Turn on NAPI and start up interrupts on a response queue.
+ */
+static void qenable(struct sge_rspq *rspq)
+{
+	napi_enable(&rspq->napi);
+
+	/*
+	 * 0-increment the Going To Sleep register to start the timer and
+	 * enable interrupts.
+	 */
+	t4_write_reg(rspq->adapter, T4VF_SGE_BASE_ADDR + SGE_VF_GTS,
+		     CIDXINC(0) |
+		     SEINTARM(rspq->intr_params) |
+		     INGRESSQID(rspq->cntxt_id));
+}
+
+/*
+ * Enable NAPI scheduling and interrupt generation for all Receive Queues.
+ */
+static void enable_rx(struct adapter *adapter)
+{
+	int rxq;
+	struct sge *s = &adapter->sge;
+
+	for_each_ethrxq(s, rxq)
+		qenable(&s->ethrxq[rxq].rspq);
+	qenable(&s->fw_evtq);
+
+	/*
+	 * The interrupt queue doesn't use NAPI so we do the 0-increment of
+	 * its Going To Sleep register here to get it started.
+	 */
+	if (adapter->flags & USING_MSI)
+		t4_write_reg(adapter, T4VF_SGE_BASE_ADDR + SGE_VF_GTS,
+			     CIDXINC(0) |
+			     SEINTARM(s->intrq.intr_params) |
+			     INGRESSQID(s->intrq.cntxt_id));
+
+}
+
+/*
+ * Wait until all NAPI handlers are descheduled.
+ */
+static void quiesce_rx(struct adapter *adapter)
+{
+	struct sge *s = &adapter->sge;
+	int rxq;
+
+	for_each_ethrxq(s, rxq)
+		napi_disable(&s->ethrxq[rxq].rspq.napi);
+	napi_disable(&s->fw_evtq.napi);
+}
+
+/*
+ * Response queue handler for the firmware event queue.
+ */
+static int fwevtq_handler(struct sge_rspq *rspq, const __be64 *rsp,
+			  const struct pkt_gl *gl)
+{
+	/*
+	 * Extract response opcode and get pointer to CPL message body.
+	 */
+	struct adapter *adapter = rspq->adapter;
+	u8 opcode = ((const struct rss_header *)rsp)->opcode;
+	void *cpl = (void *)(rsp + 1);
+
+	switch (opcode) {
+	case CPL_FW6_MSG: {
+		/*
+		 * We've received an asynchronous message from the firmware.
+		 */
+		const struct cpl_fw6_msg *fw_msg = cpl;
+		if (fw_msg->type == FW6_TYPE_CMD_RPL)
+			t4vf_handle_fw_rpl(adapter, fw_msg->data);
+		break;
+	}
+
+	case CPL_SGE_EGR_UPDATE: {
+		/*
+		 * We've received an Egress Queue status update message.
+		 * We get these, as the SGE is currently configured, when
+		 * the firmware passes certain points in processing our
+		 * TX Ethernet Queue.  We use these updates to determine
+		 * when we may need to restart a TX Ethernet Queue which
+		 * was stopped for lack of free slots ...
+		 */
+		const struct cpl_sge_egr_update *p = (void *)cpl;
+		unsigned int qid = EGR_QID(be32_to_cpu(p->opcode_qid));
+		struct sge *s = &adapter->sge;
+		struct sge_txq *tq;
+		struct sge_eth_txq *txq;
+		unsigned int eq_idx;
+		int hw_cidx, reclaimable, in_use;
+
+		/*
+		 * Perform sanity checking on the Queue ID to make sure it
+		 * really refers to one of our TX Ethernet Egress Queues which
+		 * is active and matches the queue's ID.  None of these error
+		 * conditions should ever happen so we may want to either make
+		 * them fatal and/or conditionalized under DEBUG.
+		 */
+		eq_idx = EQ_IDX(s, qid);
+		if (unlikely(eq_idx >= MAX_EGRQ)) {
+			dev_err(adapter->pdev_dev,
+				"Egress Update QID %d out of range\n", qid);
+			break;
+		}
+		tq = s->egr_map[eq_idx];
+		if (unlikely(tq == NULL)) {
+			dev_err(adapter->pdev_dev,
+				"Egress Update QID %d TXQ=NULL\n", qid);
+			break;
+		}
+		txq = container_of(tq, struct sge_eth_txq, q);
+		if (unlikely(tq->abs_id != qid)) {
+			dev_err(adapter->pdev_dev,
+				"Egress Update QID %d refers to TXQ %d\n",
+				qid, tq->abs_id);
+			break;
+		}
+
+		/*
+		 * Skip TX Queues which aren't stopped.
+		 */
+		if (likely(!netif_tx_queue_stopped(txq->txq)))
+			break;
+
+		/*
+		 * Skip stopped TX Queues which have more than half of their
+		 * DMA rings occupied with unacknowledged writes.
+		 */
+		hw_cidx = be16_to_cpu(txq->q.stat->cidx);
+		reclaimable = hw_cidx - txq->q.cidx;
+		if (reclaimable < 0)
+			reclaimable += txq->q.size;
+		in_use = txq->q.in_use - reclaimable;
+		if (in_use >= txq->q.size/2)
+			break;
+
+		/*
+		 * Restart a stopped TX Queue which has less than half of its
+		 * TX ring in use ...
+		 */
+		txq->q.restarts++;
+		netif_tx_wake_queue(txq->txq);
+		break;
+	}
+
+	default:
+		dev_err(adapter->pdev_dev,
+			"unexpected CPL %#x on FW event queue\n", opcode);
+	}
+
+	return 0;
+}
+
+/*
+ * Allocate SGE TX/RX response queues.  Determine how many sets of SGE queues
+ * to use and initializes them.  We support multiple "Queue Sets" per port if
+ * we have MSI-X, otherwise just one queue set per port.
+ */
+static int setup_sge_queues(struct adapter *adapter)
+{
+	struct sge *s = &adapter->sge;
+	int err, pidx, msix;
+
+	/*
+	 * Clear "Queue Set" Free List Starving and TX Queue Mapping Error
+	 * state.
+	 */
+	bitmap_zero(s->starving_fl, MAX_EGRQ);
+
+	/*
+	 * If we're using MSI interrupt mode we need to set up a "forwarded
+	 * interrupt" queue which we'll set up with our MSI vector.  The rest
+	 * of the ingress queues will be set up to forward their interrupts to
+	 * this queue ...  This must be first since t4vf_sge_alloc_rxq() uses
+	 * the intrq's queue ID as the interrupt forwarding queue for the
+	 * subsequent calls ...
+	 */
+	if (adapter->flags & USING_MSI) {
+		err = t4vf_sge_alloc_rxq(adapter, &s->intrq, false,
+					 adapter->port[0], 0, NULL, NULL);
+		if (err)
+			goto err_free_queues;
+	}
+
+	/*
+	 * Allocate our ingress queue for asynchronous firmware messages.
+	 */
+	err = t4vf_sge_alloc_rxq(adapter, &s->fw_evtq, true, adapter->port[0],
+				 MSIX_FW, NULL, fwevtq_handler);
+	if (err)
+		goto err_free_queues;
+
+	/*
+	 * Allocate each "port"'s initial Queue Sets.  These can be changed
+	 * later on ... up to the point where any interface on the adapter is
+	 * brought up at which point lots of things get nailed down
+	 * permanently ...
+	 */
+	msix = MSIX_NIQFLINT;
+	for_each_port(adapter, pidx) {
+		struct net_device *dev = adapter->port[pidx];
+		struct port_info *pi = netdev_priv(dev);
+		struct sge_eth_rxq *rxq = &s->ethrxq[pi->first_qset];
+		struct sge_eth_txq *txq = &s->ethtxq[pi->first_qset];
+		int nqsets = (adapter->flags & USING_MSIX) ? pi->nqsets : 1;
+		int qs;
+
+		for (qs = 0; qs < nqsets; qs++, rxq++, txq++) {
+			err = t4vf_sge_alloc_rxq(adapter, &rxq->rspq, false,
+						 dev, msix++,
+						 &rxq->fl, t4vf_ethrx_handler);
+			if (err)
+				goto err_free_queues;
+
+			err = t4vf_sge_alloc_eth_txq(adapter, txq, dev,
+					     netdev_get_tx_queue(dev, qs),
+					     s->fw_evtq.cntxt_id);
+			if (err)
+				goto err_free_queues;
+
+			rxq->rspq.idx = qs;
+			memset(&rxq->stats, 0, sizeof(rxq->stats));
+		}